1724f4c0dSSunil V L acpi= [HW,ACPI,X86,ARM64,RISCV64] 2e52347bdSJani Nikula Advanced Configuration and Power Interface 3e52347bdSJani Nikula Format: { force | on | off | strict | noirq | rsdt | 4e52347bdSJani Nikula copy_dsdt } 5e52347bdSJani Nikula force -- enable ACPI if default was off 6724f4c0dSSunil V L on -- enable ACPI but allow fallback to DT [arm64,riscv64] 7e52347bdSJani Nikula off -- disable ACPI if default was on 8e52347bdSJani Nikula noirq -- do not use ACPI for IRQ routing 9e52347bdSJani Nikula strict -- Be less tolerant of platforms that are not 10e52347bdSJani Nikula strictly ACPI specification compliant. 11e52347bdSJani Nikula rsdt -- prefer RSDT over (default) XSDT 12e52347bdSJani Nikula copy_dsdt -- copy DSDT to memory 13724f4c0dSSunil V L For ARM64 and RISCV64, ONLY "acpi=off", "acpi=on" or 14724f4c0dSSunil V L "acpi=force" are available 15e52347bdSJani Nikula 16151f4e2bSMauro Carvalho Chehab See also Documentation/power/runtime_pm.rst, pci=noacpi 17e52347bdSJani Nikula 18e52347bdSJani Nikula acpi_apic_instance= [ACPI, IOAPIC] 19e52347bdSJani Nikula Format: <int> 20e52347bdSJani Nikula 2: use 2nd APIC table, if available 21e52347bdSJani Nikula 1,0: use 1st APIC table 22e52347bdSJani Nikula default: 0 23e52347bdSJani Nikula 24e52347bdSJani Nikula acpi_backlight= [HW,ACPI] 255fd769c2SRandy Dunlap { vendor | video | native | none } 265fd769c2SRandy Dunlap If set to vendor, prefer vendor-specific driver 27e52347bdSJani Nikula (e.g. thinkpad_acpi, sony_acpi, etc.) instead 28e52347bdSJani Nikula of the ACPI video.ko driver. 295fd769c2SRandy Dunlap If set to video, use the ACPI video.ko driver. 305fd769c2SRandy Dunlap If set to native, use the device's native backlight mode. 315fd769c2SRandy Dunlap If set to none, disable the ACPI backlight interface. 32e52347bdSJani Nikula 33e52347bdSJani Nikula acpi_force_32bit_fadt_addr 34e52347bdSJani Nikula force FADT to use 32 bit addresses rather than the 35e52347bdSJani Nikula 64 bit X_* addresses. Some firmware have broken 64 36e52347bdSJani Nikula bit addresses for force ACPI ignore these and use 37e52347bdSJani Nikula the older legacy 32 bit addresses. 38e52347bdSJani Nikula 39e52347bdSJani Nikula acpica_no_return_repair [HW, ACPI] 40e52347bdSJani Nikula Disable AML predefined validation mechanism 41e52347bdSJani Nikula This mechanism can repair the evaluation result to make 42e52347bdSJani Nikula the return objects more ACPI specification compliant. 43e52347bdSJani Nikula This option is useful for developers to identify the 44e52347bdSJani Nikula root cause of an AML interpreter issue when the issue 45e52347bdSJani Nikula has something to do with the repair mechanism. 46e52347bdSJani Nikula 47e52347bdSJani Nikula acpi.debug_layer= [HW,ACPI,ACPI_DEBUG] 48e52347bdSJani Nikula acpi.debug_level= [HW,ACPI,ACPI_DEBUG] 49e52347bdSJani Nikula Format: <int> 50e52347bdSJani Nikula CONFIG_ACPI_DEBUG must be enabled to produce any ACPI 51e52347bdSJani Nikula debug output. Bits in debug_layer correspond to a 52e52347bdSJani Nikula _COMPONENT in an ACPI source file, e.g., 53866d6cdfSRafael J. Wysocki #define _COMPONENT ACPI_EVENTS 54e52347bdSJani Nikula Bits in debug_level correspond to a level in 55e52347bdSJani Nikula ACPI_DEBUG_PRINT statements, e.g., 56e52347bdSJani Nikula ACPI_DEBUG_PRINT((ACPI_DB_INFO, ... 57e52347bdSJani Nikula The debug_level mask defaults to "info". See 58cb1aaebeSMauro Carvalho Chehab Documentation/firmware-guide/acpi/debug.rst for more information about 59e52347bdSJani Nikula debug layers and levels. 60e52347bdSJani Nikula 61e52347bdSJani Nikula Enable processor driver info messages: 62e52347bdSJani Nikula acpi.debug_layer=0x20000000 63e52347bdSJani Nikula Enable AML "Debug" output, i.e., stores to the Debug 64e52347bdSJani Nikula object while interpreting AML: 65e52347bdSJani Nikula acpi.debug_layer=0xffffffff acpi.debug_level=0x2 66e52347bdSJani Nikula Enable all messages related to ACPI hardware: 67e52347bdSJani Nikula acpi.debug_layer=0x2 acpi.debug_level=0xffffffff 68e52347bdSJani Nikula 69e52347bdSJani Nikula Some values produce so much output that the system is 70e52347bdSJani Nikula unusable. The "log_buf_len" parameter may be useful 71e52347bdSJani Nikula if you need to capture more output. 72e52347bdSJani Nikula 73e52347bdSJani Nikula acpi_enforce_resources= [ACPI] 74e52347bdSJani Nikula { strict | lax | no } 75e52347bdSJani Nikula Check for resource conflicts between native drivers 76e52347bdSJani Nikula and ACPI OperationRegions (SystemIO and SystemMemory 77e52347bdSJani Nikula only). IO ports and memory declared in ACPI might be 78e52347bdSJani Nikula used by the ACPI subsystem in arbitrary AML code and 79e52347bdSJani Nikula can interfere with legacy drivers. 80e52347bdSJani Nikula strict (default): access to resources claimed by ACPI 81e52347bdSJani Nikula is denied; legacy drivers trying to access reserved 82e52347bdSJani Nikula resources will fail to bind to device using them. 83e52347bdSJani Nikula lax: access to resources claimed by ACPI is allowed; 84e52347bdSJani Nikula legacy drivers trying to access reserved resources 85e52347bdSJani Nikula will bind successfully but a warning message is logged. 86e52347bdSJani Nikula no: ACPI OperationRegions are not marked as reserved, 87e52347bdSJani Nikula no further checks are performed. 88e52347bdSJani Nikula 89e52347bdSJani Nikula acpi_force_table_verification [HW,ACPI] 90e52347bdSJani Nikula Enable table checksum verification during early stage. 91e52347bdSJani Nikula By default, this is disabled due to x86 early mapping 92e52347bdSJani Nikula size limitation. 93e52347bdSJani Nikula 94e52347bdSJani Nikula acpi_irq_balance [HW,ACPI] 95e52347bdSJani Nikula ACPI will balance active IRQs 96e52347bdSJani Nikula default in APIC mode 97e52347bdSJani Nikula 98e52347bdSJani Nikula acpi_irq_nobalance [HW,ACPI] 99e52347bdSJani Nikula ACPI will not move active IRQs (default) 100e52347bdSJani Nikula default in PIC mode 101e52347bdSJani Nikula 102e52347bdSJani Nikula acpi_irq_isa= [HW,ACPI] If irq_balance, mark listed IRQs used by ISA 103e52347bdSJani Nikula Format: <irq>,<irq>... 104e52347bdSJani Nikula 105e52347bdSJani Nikula acpi_irq_pci= [HW,ACPI] If irq_balance, clear listed IRQs for 106e52347bdSJani Nikula use by PCI 107e52347bdSJani Nikula Format: <irq>,<irq>... 108e52347bdSJani Nikula 1099c4aa1eeSLv Zheng acpi_mask_gpe= [HW,ACPI] 1109c4aa1eeSLv Zheng Due to the existence of _Lxx/_Exx, some GPEs triggered 1119c4aa1eeSLv Zheng by unsupported hardware/firmware features can result in 1129c4aa1eeSLv Zheng GPE floodings that cannot be automatically disabled by 1139c4aa1eeSLv Zheng the GPE dispatcher. 1149c4aa1eeSLv Zheng This facility can be used to prevent such uncontrolled 1159c4aa1eeSLv Zheng GPE floodings. 116d3121e64SAndy Shevchenko Format: <byte> or <bitmap-list> 1179c4aa1eeSLv Zheng 118e52347bdSJani Nikula acpi_no_auto_serialize [HW,ACPI] 119e52347bdSJani Nikula Disable auto-serialization of AML methods 120e52347bdSJani Nikula AML control methods that contain the opcodes to create 121e52347bdSJani Nikula named objects will be marked as "Serialized" by the 122e52347bdSJani Nikula auto-serialization feature. 123e52347bdSJani Nikula This feature is enabled by default. 124e52347bdSJani Nikula This option allows to turn off the feature. 125e52347bdSJani Nikula 126e52347bdSJani Nikula acpi_no_memhotplug [ACPI] Disable memory hotplug. Useful for kdump 127e52347bdSJani Nikula kernels. 128e52347bdSJani Nikula 129e52347bdSJani Nikula acpi_no_static_ssdt [HW,ACPI] 130e52347bdSJani Nikula Disable installation of static SSDTs at early boot time 131e52347bdSJani Nikula By default, SSDTs contained in the RSDT/XSDT will be 132e52347bdSJani Nikula installed automatically and they will appear under 133e52347bdSJani Nikula /sys/firmware/acpi/tables. 134e52347bdSJani Nikula This option turns off this feature. 135e52347bdSJani Nikula Note that specifying this option does not affect 136e52347bdSJani Nikula dynamic table installation which will install SSDT 137e52347bdSJani Nikula tables to /sys/firmware/acpi/tables/dynamic. 138e52347bdSJani Nikula 1393f9e12e0SJean Delvare acpi_no_watchdog [HW,ACPI,WDT] 1403f9e12e0SJean Delvare Ignore the ACPI-based watchdog interface (WDAT) and let 1413f9e12e0SJean Delvare a native driver control the watchdog device instead. 1423f9e12e0SJean Delvare 143e52347bdSJani Nikula acpi_rsdp= [ACPI,EFI,KEXEC] 144e52347bdSJani Nikula Pass the RSDP address to the kernel, mostly used 145e52347bdSJani Nikula on machines running EFI runtime service to boot the 146e52347bdSJani Nikula second kernel for kdump. 147e52347bdSJani Nikula 148e52347bdSJani Nikula acpi_os_name= [HW,ACPI] Tell ACPI BIOS the name of the OS 149e52347bdSJani Nikula Format: To spoof as Windows 98: ="Microsoft Windows" 150e52347bdSJani Nikula 151e52347bdSJani Nikula acpi_rev_override [ACPI] Override the _REV object to return 5 (instead 152e52347bdSJani Nikula of 2 which is mandated by ACPI 6) as the supported ACPI 153e52347bdSJani Nikula specification revision (when using this switch, it may 154e52347bdSJani Nikula be necessary to carry out a cold reboot _twice_ in a 155e52347bdSJani Nikula row to make it take effect on the platform firmware). 156e52347bdSJani Nikula 157e52347bdSJani Nikula acpi_osi= [HW,ACPI] Modify list of supported OS interface strings 158e52347bdSJani Nikula acpi_osi="string1" # add string1 159e52347bdSJani Nikula acpi_osi="!string2" # remove string2 160e52347bdSJani Nikula acpi_osi=!* # remove all strings 161e52347bdSJani Nikula acpi_osi=! # disable all built-in OS vendor 162e52347bdSJani Nikula strings 163e52347bdSJani Nikula acpi_osi=!! # enable all built-in OS vendor 164e52347bdSJani Nikula strings 165e52347bdSJani Nikula acpi_osi= # disable all strings 166e52347bdSJani Nikula 167e52347bdSJani Nikula 'acpi_osi=!' can be used in combination with single or 168e52347bdSJani Nikula multiple 'acpi_osi="string1"' to support specific OS 169e52347bdSJani Nikula vendor string(s). Note that such command can only 170e52347bdSJani Nikula affect the default state of the OS vendor strings, thus 171e52347bdSJani Nikula it cannot affect the default state of the feature group 172e52347bdSJani Nikula strings and the current state of the OS vendor strings, 173e52347bdSJani Nikula specifying it multiple times through kernel command line 174e52347bdSJani Nikula is meaningless. This command is useful when one do not 175e52347bdSJani Nikula care about the state of the feature group strings which 176e52347bdSJani Nikula should be controlled by the OSPM. 177e52347bdSJani Nikula Examples: 178e52347bdSJani Nikula 1. 'acpi_osi=! acpi_osi="Windows 2000"' is equivalent 179e52347bdSJani Nikula to 'acpi_osi="Windows 2000" acpi_osi=!', they all 180e52347bdSJani Nikula can make '_OSI("Windows 2000")' TRUE. 181e52347bdSJani Nikula 182e52347bdSJani Nikula 'acpi_osi=' cannot be used in combination with other 183e52347bdSJani Nikula 'acpi_osi=' command lines, the _OSI method will not 184e52347bdSJani Nikula exist in the ACPI namespace. NOTE that such command can 185e52347bdSJani Nikula only affect the _OSI support state, thus specifying it 186e52347bdSJani Nikula multiple times through kernel command line is also 187e52347bdSJani Nikula meaningless. 188e52347bdSJani Nikula Examples: 189e52347bdSJani Nikula 1. 'acpi_osi=' can make 'CondRefOf(_OSI, Local1)' 190e52347bdSJani Nikula FALSE. 191e52347bdSJani Nikula 192e52347bdSJani Nikula 'acpi_osi=!*' can be used in combination with single or 193e52347bdSJani Nikula multiple 'acpi_osi="string1"' to support specific 194e52347bdSJani Nikula string(s). Note that such command can affect the 195e52347bdSJani Nikula current state of both the OS vendor strings and the 196e52347bdSJani Nikula feature group strings, thus specifying it multiple times 197e52347bdSJani Nikula through kernel command line is meaningful. But it may 198e52347bdSJani Nikula still not able to affect the final state of a string if 199e52347bdSJani Nikula there are quirks related to this string. This command 200e52347bdSJani Nikula is useful when one want to control the state of the 201e52347bdSJani Nikula feature group strings to debug BIOS issues related to 202e52347bdSJani Nikula the OSPM features. 203e52347bdSJani Nikula Examples: 204e52347bdSJani Nikula 1. 'acpi_osi="Module Device" acpi_osi=!*' can make 205e52347bdSJani Nikula '_OSI("Module Device")' FALSE. 206e52347bdSJani Nikula 2. 'acpi_osi=!* acpi_osi="Module Device"' can make 207e52347bdSJani Nikula '_OSI("Module Device")' TRUE. 208e52347bdSJani Nikula 3. 'acpi_osi=! acpi_osi=!* acpi_osi="Windows 2000"' is 209e52347bdSJani Nikula equivalent to 210e52347bdSJani Nikula 'acpi_osi=!* acpi_osi=! acpi_osi="Windows 2000"' 211e52347bdSJani Nikula and 212e52347bdSJani Nikula 'acpi_osi=!* acpi_osi="Windows 2000" acpi_osi=!', 213e52347bdSJani Nikula they all will make '_OSI("Windows 2000")' TRUE. 214e52347bdSJani Nikula 215e52347bdSJani Nikula acpi_pm_good [X86] 216e52347bdSJani Nikula Override the pmtimer bug detection: force the kernel 217e52347bdSJani Nikula to assume that this machine's pmtimer latches its value 218e52347bdSJani Nikula and always returns good values. 219e52347bdSJani Nikula 220e52347bdSJani Nikula acpi_sci= [HW,ACPI] ACPI System Control Interrupt trigger mode 221e52347bdSJani Nikula Format: { level | edge | high | low } 222e52347bdSJani Nikula 223e52347bdSJani Nikula acpi_skip_timer_override [HW,ACPI] 224e52347bdSJani Nikula Recognize and ignore IRQ0/pin2 Interrupt Override. 225e52347bdSJani Nikula For broken nForce2 BIOS resulting in XT-PIC timer. 226e52347bdSJani Nikula 227e52347bdSJani Nikula acpi_sleep= [HW,ACPI] Sleep options 22874d95555SDavid Woodhouse Format: { s3_bios, s3_mode, s3_beep, s4_hwsig, 22974d95555SDavid Woodhouse s4_nohwsig, old_ordering, nonvs, 23074d95555SDavid Woodhouse sci_force_enable, nobl } 231151f4e2bSMauro Carvalho Chehab See Documentation/power/video.rst for information on 232e52347bdSJani Nikula s3_bios and s3_mode. 233e52347bdSJani Nikula s3_beep is for debugging; it makes the PC's speaker beep 234e52347bdSJani Nikula as soon as the kernel's real-mode entry point is called. 23574d95555SDavid Woodhouse s4_hwsig causes the kernel to check the ACPI hardware 23674d95555SDavid Woodhouse signature during resume from hibernation, and gracefully 23774d95555SDavid Woodhouse refuse to resume if it has changed. This complies with 23874d95555SDavid Woodhouse the ACPI specification but not with reality, since 23974d95555SDavid Woodhouse Windows does not do this and many laptops do change it 24074d95555SDavid Woodhouse on docking. So the default behaviour is to allow resume 24174d95555SDavid Woodhouse and simply warn when the signature changes, unless the 24274d95555SDavid Woodhouse s4_hwsig option is enabled. 243e52347bdSJani Nikula s4_nohwsig prevents ACPI hardware signature from being 24474d95555SDavid Woodhouse used (or even warned about) during resume. 245e52347bdSJani Nikula old_ordering causes the ACPI 1.0 ordering of the _PTS 246e52347bdSJani Nikula control method, with respect to putting devices into 247e52347bdSJani Nikula low power states, to be enforced (the ACPI 2.0 ordering 248e52347bdSJani Nikula of _PTS is used by default). 249e52347bdSJani Nikula nonvs prevents the kernel from saving/restoring the 250e52347bdSJani Nikula ACPI NVS memory during suspend/hibernation and resume. 251e52347bdSJani Nikula sci_force_enable causes the kernel to set SCI_EN directly 252e52347bdSJani Nikula on resume from S1/S3 (which is against the ACPI spec, 253e52347bdSJani Nikula but some broken systems don't work without it). 25457044031SRafael J. Wysocki nobl causes the internal blacklist of systems known to 25557044031SRafael J. Wysocki behave incorrectly in some ways with respect to system 25657044031SRafael J. Wysocki suspend and resume to be ignored (use wisely). 257e52347bdSJani Nikula 258e52347bdSJani Nikula acpi_use_timer_override [HW,ACPI] 259e52347bdSJani Nikula Use timer override. For some broken Nvidia NF5 boards 260e52347bdSJani Nikula that require a timer override, but don't have HPET 261e52347bdSJani Nikula 262e52347bdSJani Nikula add_efi_memmap [EFI; X86] Include EFI memory map in 263e52347bdSJani Nikula kernel's map of available physical RAM. 264e52347bdSJani Nikula 265e52347bdSJani Nikula agp= [AGP] 266e52347bdSJani Nikula { off | try_unsupported } 267e52347bdSJani Nikula off: disable AGP support 268e52347bdSJani Nikula try_unsupported: try to drive unsupported chipsets 269e52347bdSJani Nikula (may crash computer or cause data corruption) 270e52347bdSJani Nikula 271e52347bdSJani Nikula ALSA [HW,ALSA] 2721ca2c806SMauro Carvalho Chehab See Documentation/sound/alsa-configuration.rst 273e52347bdSJani Nikula 274e52347bdSJani Nikula alignment= [KNL,ARM] 275e52347bdSJani Nikula Allow the default userspace alignment fault handler 276e52347bdSJani Nikula behaviour to be specified. Bit 0 enables warnings, 277e52347bdSJani Nikula bit 1 enables fixups, and bit 2 sends a segfault. 278e52347bdSJani Nikula 279e52347bdSJani Nikula align_va_addr= [X86-64] 280e52347bdSJani Nikula Align virtual addresses by clearing slice [14:12] when 281e52347bdSJani Nikula allocating a VMA at process creation time. This option 282e52347bdSJani Nikula gives you up to 3% performance improvement on AMD F15h 283e52347bdSJani Nikula machines (where it is enabled by default) for a 284e52347bdSJani Nikula CPU-intensive style benchmark, and it can vary highly in 285e52347bdSJani Nikula a microbenchmark depending on workload and compiler. 286e52347bdSJani Nikula 287e52347bdSJani Nikula 32: only for 32-bit processes 288e52347bdSJani Nikula 64: only for 64-bit processes 289e52347bdSJani Nikula on: enable for both 32- and 64-bit processes 290e52347bdSJani Nikula off: disable for both 32- and 64-bit processes 291e52347bdSJani Nikula 292e52347bdSJani Nikula alloc_snapshot [FTRACE] 293e52347bdSJani Nikula Allocate the ftrace snapshot buffer on boot up when the 294e52347bdSJani Nikula main buffer is allocated. This is handy if debugging 295e52347bdSJani Nikula and you need to use tracing_snapshot() on boot up, and 296e52347bdSJani Nikula do not want to use tracing_snapshot_alloc() as it needs 297e52347bdSJani Nikula to be done where GFP_KERNEL allocations are allowed. 298e52347bdSJani Nikula 299ead7de46SWill Deacon allow_mismatched_32bit_el0 [ARM64] 300ead7de46SWill Deacon Allow execve() of 32-bit applications and setting of the 301ead7de46SWill Deacon PER_LINUX32 personality on systems where only a strict 302ead7de46SWill Deacon subset of the CPUs support 32-bit EL0. When this 303ead7de46SWill Deacon parameter is present, the set of CPUs supporting 32-bit 304ead7de46SWill Deacon EL0 is indicated by /sys/devices/system/cpu/aarch32_el0 305ead7de46SWill Deacon and hot-unplug operations may be restricted. 306ead7de46SWill Deacon 307e4624435SJonathan Corbet See Documentation/arch/arm64/asymmetric-32bit.rst for more 308702f4387SWill Deacon information. 309702f4387SWill Deacon 310e52347bdSJani Nikula amd_iommu= [HW,X86-64] 311e52347bdSJani Nikula Pass parameters to the AMD IOMMU driver in the system. 312e52347bdSJani Nikula Possible values are: 3131d479f16SJohn Garry fullflush - Deprecated, equivalent to iommu.strict=1 314e52347bdSJani Nikula off - do not initialize any AMD IOMMU found in 315e52347bdSJani Nikula the system 316e52347bdSJani Nikula force_isolation - Force device isolation for all 317e52347bdSJani Nikula devices. The IOMMU driver is not 318e52347bdSJani Nikula allowed anymore to lift isolation 319e52347bdSJani Nikula requirements as needed. This option 320e52347bdSJani Nikula does not override iommu=pt 321b1e650dbSJoerg Roedel force_enable - Force enable the IOMMU on platforms known 322b1e650dbSJoerg Roedel to be buggy with IOMMU enabled. Use this 323b1e650dbSJoerg Roedel option with care. 324d799a183SVasant Hegde pgtbl_v1 - Use v1 page table for DMA-API (Default). 325d799a183SVasant Hegde pgtbl_v2 - Use v2 page table for DMA-API. 32666419036SSuravee Suthikulpanit irtcachedis - Disable Interrupt Remapping Table (IRT) caching. 327e52347bdSJani Nikula 328e52347bdSJani Nikula amd_iommu_dump= [HW,X86-64] 329e52347bdSJani Nikula Enable AMD IOMMU driver option to dump the ACPI table 330e52347bdSJani Nikula for AMD IOMMU. With this option enabled, AMD IOMMU 331e52347bdSJani Nikula driver will print ACPI tables for AMD IOMMU during 332e52347bdSJani Nikula IOMMU initialization. 333e52347bdSJani Nikula 334e52347bdSJani Nikula amd_iommu_intr= [HW,X86-64] 335e52347bdSJani Nikula Specifies one of the following AMD IOMMU interrupt 336e52347bdSJani Nikula remapping modes: 337e52347bdSJani Nikula legacy - Use legacy interrupt remapping mode. 338e52347bdSJani Nikula vapic - Use virtual APIC mode, which allows IOMMU 339e52347bdSJani Nikula to inject interrupts directly into guest. 340e52347bdSJani Nikula This mode requires kvm-amd.avic=1. 341e52347bdSJani Nikula (Default when IOMMU HW support is present.) 342e52347bdSJani Nikula 3433e6e0780SWyes Karny amd_pstate= [X86] 3443e6e0780SWyes Karny disable 3453e6e0780SWyes Karny Do not enable amd_pstate as the default 3463e6e0780SWyes Karny scaling driver for the supported processors 3473e6e0780SWyes Karny passive 3482dd6d0ebSWyes Karny Use amd_pstate with passive mode as a scaling driver. 3492dd6d0ebSWyes Karny In this mode autonomous selection is disabled. 3502dd6d0ebSWyes Karny Driver requests a desired performance level and platform 3512dd6d0ebSWyes Karny tries to match the same performance level if it is 3522dd6d0ebSWyes Karny satisfied by guaranteed performance level. 3533e6e0780SWyes Karny active 3543e6e0780SWyes Karny Use amd_pstate_epp driver instance as the scaling driver, 3553e6e0780SWyes Karny driver provides a hint to the hardware if software wants 3563e6e0780SWyes Karny to bias toward performance (0x0) or energy efficiency (0xff) 3573e6e0780SWyes Karny to the CPPC firmware. then CPPC power algorithm will 3583e6e0780SWyes Karny calculate the runtime workload and adjust the realtime cores 3593e6e0780SWyes Karny frequency. 3602dd6d0ebSWyes Karny guided 3612dd6d0ebSWyes Karny Activate guided autonomous mode. Driver requests minimum and 3622dd6d0ebSWyes Karny maximum performance level and the platform autonomously 3632dd6d0ebSWyes Karny selects a performance level in this range and appropriate 3642dd6d0ebSWyes Karny to the current workload. 3653e6e0780SWyes Karny 366e52347bdSJani Nikula amijoy.map= [HW,JOY] Amiga joystick support 367e52347bdSJani Nikula Map of devices attached to JOY0DAT and JOY1DAT 368e52347bdSJani Nikula Format: <a>,<b> 3691752118dSTom Saeger See also Documentation/input/joydev/joystick.rst 370e52347bdSJani Nikula 371e52347bdSJani Nikula analog.map= [HW,JOY] Analog joystick and gamepad support 372e52347bdSJani Nikula Specifies type or capabilities of an analog joystick 373e52347bdSJani Nikula connected to one of 16 gameports 374e52347bdSJani Nikula Format: <type1>,<type2>,..<type16> 375e52347bdSJani Nikula 376e52347bdSJani Nikula apc= [HW,SPARC] 377e52347bdSJani Nikula Power management functions (SPARCstation-4/5 + deriv.) 378e52347bdSJani Nikula Format: noidle 379e52347bdSJani Nikula Disable APC CPU standby support. SPARCstation-Fox does 380e52347bdSJani Nikula not play well with APC CPU idle - disable it if you have 381e52347bdSJani Nikula APC and your system crashes randomly. 382e52347bdSJani Nikula 38364e05d11SDou Liyang apic= [APIC,X86] Advanced Programmable Interrupt Controller 384806654a9SWill Deacon Change the output verbosity while booting 385e52347bdSJani Nikula Format: { quiet (default) | verbose | debug } 386e52347bdSJani Nikula Change the amount of debugging information output 387e52347bdSJani Nikula when initialising the APIC and IO-APIC components. 38864e05d11SDou Liyang For X86-32, this can also be used to specify an APIC 38964e05d11SDou Liyang driver name. 39064e05d11SDou Liyang Format: apic=driver_name 39164e05d11SDou Liyang Examples: apic=bigsmp 392e52347bdSJani Nikula 393e52347bdSJani Nikula apic_extnmi= [APIC,X86] External NMI delivery setting 394e52347bdSJani Nikula Format: { bsp (default) | all | none } 395e52347bdSJani Nikula bsp: External NMI is delivered only to CPU 0 396e52347bdSJani Nikula all: External NMIs are broadcast to all CPUs as a 397e52347bdSJani Nikula backup of CPU 0 398e52347bdSJani Nikula none: External NMI is masked for all CPUs. This is 399e52347bdSJani Nikula useful so that a dump capture kernel won't be 400e52347bdSJani Nikula shot down by NMI 401e52347bdSJani Nikula 402e52347bdSJani Nikula autoconf= [IPV6] 40319093313SMauro Carvalho Chehab See Documentation/networking/ipv6.rst. 404e52347bdSJani Nikula 405e52347bdSJani Nikula apm= [APM] Advanced Power Management 406e52347bdSJani Nikula See header of arch/x86/kernel/apm_32.c. 407e52347bdSJani Nikula 40842551b8dSRandy Dunlap apparmor= [APPARMOR] Disable or enable AppArmor at boot time 40942551b8dSRandy Dunlap Format: { "0" | "1" } 41042551b8dSRandy Dunlap See security/apparmor/Kconfig help text 41142551b8dSRandy Dunlap 0 -- disable. 41242551b8dSRandy Dunlap 1 -- enable. 41342551b8dSRandy Dunlap Default value is set via kernel config option. 41442551b8dSRandy Dunlap 415e52347bdSJani Nikula arcrimi= [HW,NET] ARCnet - "RIM I" (entirely mem-mapped) cards 416e52347bdSJani Nikula Format: <io>,<irq>,<nodeID> 417e52347bdSJani Nikula 41893ad55b7SMarc Zyngier arm64.nobti [ARM64] Unconditionally disable Branch Target 41993ad55b7SMarc Zyngier Identification support 42093ad55b7SMarc Zyngier 421eb38cc80SRandy Dunlap arm64.nomops [ARM64] Unconditionally disable Memory Copy and Memory 422eb38cc80SRandy Dunlap Set instructions support 423f8da5752SMarc Zyngier 4247a062ce3SYee Lee arm64.nomte [ARM64] Unconditionally disable Memory Tagging Extension 4257a062ce3SYee Lee support 4267a062ce3SYee Lee 427eb38cc80SRandy Dunlap arm64.nopauth [ARM64] Unconditionally disable Pointer Authentication 428eb38cc80SRandy Dunlap support 429504ee236SMarc Zyngier 430b3000e21SMarc Zyngier arm64.nosme [ARM64] Unconditionally disable Scalable Matrix 431b3000e21SMarc Zyngier Extension support 432b3000e21SMarc Zyngier 433eb38cc80SRandy Dunlap arm64.nosve [ARM64] Unconditionally disable Scalable Vector 434eb38cc80SRandy Dunlap Extension support 4353e1dedb2SKristina Martsenko 436e52347bdSJani Nikula ataflop= [HW,M68k] 437e52347bdSJani Nikula 438e52347bdSJani Nikula atarimouse= [HW,MOUSE] Atari Mouse 439e52347bdSJani Nikula 440e52347bdSJani Nikula atkbd.extra= [HW] Enable extra LEDs and keys on IBM RapidAccess, 441e52347bdSJani Nikula EzKey and similar keyboards 442e52347bdSJani Nikula 443e52347bdSJani Nikula atkbd.reset= [HW] Reset keyboard during initialization 444e52347bdSJani Nikula 445e52347bdSJani Nikula atkbd.set= [HW] Select keyboard code set 446e52347bdSJani Nikula Format: <int> (2 = AT (default), 3 = PS/2) 447e52347bdSJani Nikula 448e52347bdSJani Nikula atkbd.scroll= [HW] Enable scroll wheel on MS Office and similar 449e52347bdSJani Nikula keyboards 450e52347bdSJani Nikula 451e52347bdSJani Nikula atkbd.softraw= [HW] Choose between synthetic and real raw mode 452e52347bdSJani Nikula Format: <bool> (0 = real, 1 = synthetic (default)) 453e52347bdSJani Nikula 454e52347bdSJani Nikula atkbd.softrepeat= [HW] 455e52347bdSJani Nikula Use software keyboard repeat 456e52347bdSJani Nikula 457e52347bdSJani Nikula audit= [KNL] Enable the audit sub-system 45811dd2666SGreg Edwards Format: { "0" | "1" | "off" | "on" } 45911dd2666SGreg Edwards 0 | off - kernel audit is disabled and can not be 46011dd2666SGreg Edwards enabled until the next reboot 461e52347bdSJani Nikula unset - kernel audit is initialized but disabled and 462e52347bdSJani Nikula will be fully enabled by the userspace auditd. 46311dd2666SGreg Edwards 1 | on - kernel audit is initialized and partially 46411dd2666SGreg Edwards enabled, storing at most audit_backlog_limit 46511dd2666SGreg Edwards messages in RAM until it is fully enabled by the 46611dd2666SGreg Edwards userspace auditd. 467e52347bdSJani Nikula Default: unset 468e52347bdSJani Nikula 469e52347bdSJani Nikula audit_backlog_limit= [KNL] Set the audit queue size limit. 470e52347bdSJani Nikula Format: <int> (must be >=0) 471e52347bdSJani Nikula Default: 64 472e52347bdSJani Nikula 473e52347bdSJani Nikula bau= [X86_UV] Enable the BAU on SGI UV. The default 474e52347bdSJani Nikula behavior is to disable the BAU (i.e. bau=0). 475e52347bdSJani Nikula Format: { "0" | "1" } 476e52347bdSJani Nikula 0 - Disable the BAU. 477e52347bdSJani Nikula 1 - Enable the BAU. 478e52347bdSJani Nikula unset - Disable the BAU. 479e52347bdSJani Nikula 480e52347bdSJani Nikula baycom_epp= [HW,AX25] 481e52347bdSJani Nikula Format: <io>,<mode> 482e52347bdSJani Nikula 483e52347bdSJani Nikula baycom_par= [HW,AX25] BayCom Parallel Port AX.25 Modem 484e52347bdSJani Nikula Format: <io>,<mode> 485e52347bdSJani Nikula See header of drivers/net/hamradio/baycom_par.c. 486e52347bdSJani Nikula 487e52347bdSJani Nikula baycom_ser_fdx= [HW,AX25] 488e52347bdSJani Nikula BayCom Serial Port AX.25 Modem (Full Duplex Mode) 489e52347bdSJani Nikula Format: <io>,<irq>,<mode>[,<baud>] 490e52347bdSJani Nikula See header of drivers/net/hamradio/baycom_ser_fdx.c. 491e52347bdSJani Nikula 492e52347bdSJani Nikula baycom_ser_hdx= [HW,AX25] 493e52347bdSJani Nikula BayCom Serial Port AX.25 Modem (Half Duplex Mode) 494e52347bdSJani Nikula Format: <io>,<irq>,<mode> 495e52347bdSJani Nikula See header of drivers/net/hamradio/baycom_ser_hdx.c. 496e52347bdSJani Nikula 497389cfd96SRandy Dunlap bert_disable [ACPI] 498389cfd96SRandy Dunlap Disable BERT OS support on buggy BIOSes. 499389cfd96SRandy Dunlap 500389cfd96SRandy Dunlap bgrt_disable [ACPI][X86] 501389cfd96SRandy Dunlap Disable BGRT to avoid flickering OEM logo. 502389cfd96SRandy Dunlap 503e52347bdSJani Nikula blkdevparts= Manual partition parsing of block device(s) for 504e52347bdSJani Nikula embedded devices based on command line input. 505898bd37aSMauro Carvalho Chehab See Documentation/block/cmdline-partition.rst 506e52347bdSJani Nikula 507e52347bdSJani Nikula boot_delay= Milliseconds to delay each printk during boot. 508a568375bSBjorn Helgaas Only works if CONFIG_BOOT_PRINTK_DELAY is enabled, 509a568375bSBjorn Helgaas and you may also have to specify "lpj=". Boot_delay 510a568375bSBjorn Helgaas values larger than 10 seconds (10000) are assumed 511a568375bSBjorn Helgaas erroneous and ignored. 512e52347bdSJani Nikula Format: integer 513e52347bdSJani Nikula 5147495e092SSteven Rostedt (VMware) bootconfig [KNL] 5157495e092SSteven Rostedt (VMware) Extended command line options can be added to an initrd 5167495e092SSteven Rostedt (VMware) and this will cause the kernel to look for it. 5177495e092SSteven Rostedt (VMware) 5187495e092SSteven Rostedt (VMware) See Documentation/admin-guide/bootconfig.rst 5197495e092SSteven Rostedt (VMware) 520e52347bdSJani Nikula bttv.card= [HW,V4L] bttv (bt848 + bt878 based grabber cards) 521e52347bdSJani Nikula bttv.radio= Most important insmod options are available as 522e52347bdSJani Nikula kernel args too. 52332e2eae2SMauro Carvalho Chehab bttv.pll= See Documentation/admin-guide/media/bttv.rst 524e52347bdSJani Nikula bttv.tuner= 525e52347bdSJani Nikula 526e52347bdSJani Nikula bulk_remove=off [PPC] This parameter disables the use of the pSeries 527e52347bdSJani Nikula firmware feature for flushing multiple hpte entries 528e52347bdSJani Nikula at a time. 529e52347bdSJani Nikula 530e52347bdSJani Nikula c101= [NET] Moxa C101 synchronous serial card 531e52347bdSJani Nikula 532e52347bdSJani Nikula cachesize= [BUGS=X86-32] Override level 2 CPU cache size detection. 533e52347bdSJani Nikula Sometimes CPU hardware bugs make them report the cache 534e52347bdSJani Nikula size incorrectly. The kernel will attempt work arounds 535e52347bdSJani Nikula to fix known problems, but for some CPUs it is not 536e52347bdSJani Nikula possible to determine what the correct size should be. 537e52347bdSJani Nikula This option provides an override for these situations. 538e52347bdSJani Nikula 5393fc46fc9SMartin Kepplinger carrier_timeout= 5403fc46fc9SMartin Kepplinger [NET] Specifies amount of time (in seconds) that 5413fc46fc9SMartin Kepplinger the kernel should wait for a network carrier. By default 5423fc46fc9SMartin Kepplinger it waits 120 seconds. 5433fc46fc9SMartin Kepplinger 544e52347bdSJani Nikula ca_keys= [KEYS] This parameter identifies a specific key(s) on 545e52347bdSJani Nikula the system trusted keyring to be used for certificate 546e52347bdSJani Nikula trust validation. 547e52347bdSJani Nikula format: { id:<keyid> | builtin } 548e52347bdSJani Nikula 549e52347bdSJani Nikula cca= [MIPS] Override the kernel pages' cache coherency 550e52347bdSJani Nikula algorithm. Accepted values range from 0 to 7 551e52347bdSJani Nikula inclusive. See arch/mips/include/asm/pgtable-bits.h 552e52347bdSJani Nikula for platform specific values (SB1, Loongson3 and 553e52347bdSJani Nikula others). 554e52347bdSJani Nikula 555e52347bdSJani Nikula ccw_timeout_log [S390] 55637002bc6SCosta Shulyupin See Documentation/arch/s390/common_io.rst for details. 557e52347bdSJani Nikula 5583958e2d0SSuren Baghdasaryan cgroup_disable= [KNL] Disable a particular controller or optional feature 5593958e2d0SSuren Baghdasaryan Format: {name of the controller(s) or feature(s) to disable} 560e52347bdSJani Nikula The effects of cgroup_disable=foo are: 561e52347bdSJani Nikula - foo isn't auto-mounted if you mount all cgroups in 562e52347bdSJani Nikula a single hierarchy 563e52347bdSJani Nikula - foo isn't visible as an individually mountable 564e52347bdSJani Nikula subsystem 5653958e2d0SSuren Baghdasaryan - if foo is an optional feature then the feature is 5663958e2d0SSuren Baghdasaryan disabled and corresponding cgroup files are not 5673958e2d0SSuren Baghdasaryan created 568e52347bdSJani Nikula {Currently only "memory" controller deal with this and 569e52347bdSJani Nikula cut the overhead, others just disable the usage. So 570e52347bdSJani Nikula only cgroup_disable=memory is actually worthy} 5713958e2d0SSuren Baghdasaryan Specifying "pressure" disables per-cgroup pressure 5723958e2d0SSuren Baghdasaryan stall information accounting feature 573e52347bdSJani Nikula 5743fc9c12dSTejun Heo cgroup_no_v1= [KNL] Disable cgroup controllers and named hierarchies in v1 5753fc9c12dSTejun Heo Format: { { controller | "all" | "named" } 5763fc9c12dSTejun Heo [,{ controller | "all" | "named" }...] } 577e52347bdSJani Nikula Like cgroup_disable, but only applies to cgroup v1; 578e52347bdSJani Nikula the blacklisted controllers remain available in cgroup2. 5793fc9c12dSTejun Heo "all" blacklists all controllers and "named" disables 5803fc9c12dSTejun Heo named mounts. Specifying both "all" and "named" disables 5813fc9c12dSTejun Heo all v1 hierarchies. 582e52347bdSJani Nikula 583e52347bdSJani Nikula cgroup.memory= [KNL] Pass options to the cgroup memory controller. 584e52347bdSJani Nikula Format: <string> 585e52347bdSJani Nikula nosocket -- Disable socket memory accounting. 586e52347bdSJani Nikula nokmem -- Disable kernel memory accounting. 587b6c1a8afSYafang Shao nobpf -- Disable BPF memory accounting. 588e52347bdSJani Nikula 5898d6d51edSRandy Dunlap checkreqprot= [SELINUX] Set initial checkreqprot flag value. 590e52347bdSJani Nikula Format: { "0" | "1" } 591e52347bdSJani Nikula See security/selinux/Kconfig help text. 592e52347bdSJani Nikula 0 -- check protection applied by kernel (includes 593e52347bdSJani Nikula any implied execute protection). 594e52347bdSJani Nikula 1 -- check protection requested by application. 595e52347bdSJani Nikula Default value is set via a kernel config option. 596e52347bdSJani Nikula Value can be changed at runtime via 597d41415ebSStephen Smalley /sys/fs/selinux/checkreqprot. 598e9c38f9fSStephen Smalley Setting checkreqprot to 1 is deprecated. 599e52347bdSJani Nikula 600e52347bdSJani Nikula cio_ignore= [S390] 60137002bc6SCosta Shulyupin See Documentation/arch/s390/common_io.rst for details. 602d2fc83c1SRandy Dunlap 60388a61892SLinus Torvalds clearcpuid=X[,X...] [X86] 604d2fc83c1SRandy Dunlap Disable CPUID feature X for the kernel. See 605d2fc83c1SRandy Dunlap arch/x86/include/asm/cpufeatures.h for the valid bit 60688a61892SLinus Torvalds numbers X. Note the Linux-specific bits are not necessarily 60788a61892SLinus Torvalds stable over kernel options, but the vendor-specific 608d2fc83c1SRandy Dunlap ones should be. 60988a61892SLinus Torvalds X can also be a string as appearing in the flags: line 61088a61892SLinus Torvalds in /proc/cpuinfo which does not have the above 61188a61892SLinus Torvalds instability issue. However, not all features have names 61288a61892SLinus Torvalds in /proc/cpuinfo. 61388a61892SLinus Torvalds Note that using this option will taint your kernel. 614d2fc83c1SRandy Dunlap Also note that user programs calling CPUID directly 615d2fc83c1SRandy Dunlap or using the feature without checking anything 616d2fc83c1SRandy Dunlap will still see it. This just prevents it from 617d2fc83c1SRandy Dunlap being used by the kernel or shown in /proc/cpuinfo. 618d2fc83c1SRandy Dunlap Also note the kernel might malfunction if you disable 619d2fc83c1SRandy Dunlap some critical bits. 620d2fc83c1SRandy Dunlap 621e52347bdSJani Nikula clk_ignore_unused 622e52347bdSJani Nikula [CLK] 623e52347bdSJani Nikula Prevents the clock framework from automatically gating 624e52347bdSJani Nikula clocks that have not been explicitly enabled by a Linux 625e52347bdSJani Nikula device driver but are enabled in hardware at reset or 626e52347bdSJani Nikula by the bootloader/firmware. Note that this does not 627e52347bdSJani Nikula force such clocks to be always-on nor does it reserve 628e52347bdSJani Nikula those clocks in any way. This parameter is useful for 629e52347bdSJani Nikula debug and development, but should not be needed on a 630e52347bdSJani Nikula platform with proper driver support. For more 63118bcaa4eSMauro Carvalho Chehab information, see Documentation/driver-api/clk.rst. 632e52347bdSJani Nikula 633e52347bdSJani Nikula clock= [BUGS=X86-32, HW] gettimeofday clocksource override. 634e52347bdSJani Nikula [Deprecated] 635e52347bdSJani Nikula Forces specified clocksource (if available) to be used 636e52347bdSJani Nikula when calculating gettimeofday(). If specified 637e52347bdSJani Nikula clocksource is not available, it defaults to PIT. 638e52347bdSJani Nikula Format: { pit | tsc | cyclone | pmtmr } 639e52347bdSJani Nikula 640e52347bdSJani Nikula clocksource= Override the default clocksource 641e52347bdSJani Nikula Format: <string> 642e52347bdSJani Nikula Override the default clocksource and use the clocksource 643e52347bdSJani Nikula with the name specified. 644e52347bdSJani Nikula Some clocksource names to choose from, depending on 645e52347bdSJani Nikula the platform: 646e52347bdSJani Nikula [all] jiffies (this is the base, fallback clocksource) 647e52347bdSJani Nikula [ACPI] acpi_pm 648e52347bdSJani Nikula [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2, 649e52347bdSJani Nikula pxa_timer,timer3,32k_counter,timer0_1 650e52347bdSJani Nikula [X86-32] pit,hpet,tsc; 651e52347bdSJani Nikula scx200_hrt on Geode; cyclone on IBM x440 652e52347bdSJani Nikula [MIPS] MIPS 653e52347bdSJani Nikula [PARISC] cr16 654e52347bdSJani Nikula [S390] tod 655e52347bdSJani Nikula [SH] SuperH 656e52347bdSJani Nikula [SPARC64] tick 657e52347bdSJani Nikula [X86-64] hpet,tsc 658e52347bdSJani Nikula 659e52347bdSJani Nikula clocksource.arm_arch_timer.evtstrm= 660e52347bdSJani Nikula [ARM,ARM64] 661e52347bdSJani Nikula Format: <bool> 662e52347bdSJani Nikula Enable/disable the eventstream feature of the ARM 663e52347bdSJani Nikula architected timer so that code using WFE-based polling 664e52347bdSJani Nikula loops can be debugged more effectively on production 665e52347bdSJani Nikula systems. 666e52347bdSJani Nikula 667db3a34e1SPaul E. McKenney clocksource.max_cswd_read_retries= [KNL] 668db3a34e1SPaul E. McKenney Number of clocksource_watchdog() retries due to 669db3a34e1SPaul E. McKenney external delays before the clock will be marked 6701a562067SWaiman Long unstable. Defaults to two retries, that is, 6711a562067SWaiman Long three attempts to read the clock under test. 672db3a34e1SPaul E. McKenney 673fa218f1cSPaul E. McKenney clocksource.verify_n_cpus= [KNL] 674fa218f1cSPaul E. McKenney Limit the number of CPUs checked for clocksources 675fa218f1cSPaul E. McKenney marked with CLOCK_SOURCE_VERIFY_PERCPU that 676fa218f1cSPaul E. McKenney are marked unstable due to excessive skew. 677fa218f1cSPaul E. McKenney A negative value says to check all CPUs, while 678fa218f1cSPaul E. McKenney zero says not to check any. Values larger than 679fa218f1cSPaul E. McKenney nr_cpu_ids are silently truncated to nr_cpu_ids. 680fa218f1cSPaul E. McKenney The actual CPUs are chosen randomly, with 681fa218f1cSPaul E. McKenney no replacement if the same CPU is chosen twice. 682fa218f1cSPaul E. McKenney 6831253b9b8SPaul E. McKenney clocksource-wdtest.holdoff= [KNL] 6841253b9b8SPaul E. McKenney Set the time in seconds that the clocksource 6851253b9b8SPaul E. McKenney watchdog test waits before commencing its tests. 6861253b9b8SPaul E. McKenney Defaults to zero when built as a module and to 6871253b9b8SPaul E. McKenney 10 seconds when built into the kernel. 6881253b9b8SPaul E. McKenney 689e52347bdSJani Nikula cma=nn[MG]@[start[MG][-end[MG]]] 690c372e741STian Tao [KNL,CMA] 691e52347bdSJani Nikula Sets the size of kernel global memory area for 692e52347bdSJani Nikula contiguous memory allocations and optionally the 693e52347bdSJani Nikula placement constraint by the physical address range of 694e52347bdSJani Nikula memory allocations. A value of 0 disables CMA 695e52347bdSJani Nikula altogether. For more information, see 6960b1abd1fSChristoph Hellwig kernel/dma/contiguous.c 697e52347bdSJani Nikula 698b7176c26SBarry Song cma_pernuma=nn[MG] 69922e4a348SYajun Deng [KNL,CMA] 700b7176c26SBarry Song Sets the size of kernel per-numa memory area for 701b7176c26SBarry Song contiguous memory allocations. A value of 0 disables 702b7176c26SBarry Song per-numa CMA altogether. And If this option is not 703dbeb56feSRandy Dunlap specified, the default value is 0. 704b7176c26SBarry Song With per-numa CMA enabled, DMA users on node nid will 705b7176c26SBarry Song first try to allocate buffer from the pernuma area 706b7176c26SBarry Song which is located in node nid, if the allocation fails, 707b7176c26SBarry Song they will fallback to the global default memory area. 708e52347bdSJani Nikula 709bf29bfaaSYajun Deng numa_cma=<node>:nn[MG][,<node>:nn[MG]] 710bf29bfaaSYajun Deng [KNL,CMA] 711bf29bfaaSYajun Deng Sets the size of kernel numa memory area for 712bf29bfaaSYajun Deng contiguous memory allocations. It will reserve CMA 713bf29bfaaSYajun Deng area for the specified node. 714bf29bfaaSYajun Deng 715bf29bfaaSYajun Deng With numa CMA enabled, DMA users on node nid will 716bf29bfaaSYajun Deng first try to allocate buffer from the numa area 717bf29bfaaSYajun Deng which is located in node nid, if the allocation fails, 718bf29bfaaSYajun Deng they will fallback to the global default memory area. 719bf29bfaaSYajun Deng 720e52347bdSJani Nikula cmo_free_hint= [PPC] Format: { yes | no } 721e52347bdSJani Nikula Specify whether pages are marked as being inactive 722e52347bdSJani Nikula when they are freed. This is used in CMO environments 723e52347bdSJani Nikula to determine OS memory pressure for page stealing by 724e52347bdSJani Nikula a hypervisor. 725e52347bdSJani Nikula Default: yes 726e52347bdSJani Nikula 727e52347bdSJani Nikula coherent_pool=nn[KMG] [ARM,KNL] 728e52347bdSJani Nikula Sets the size of memory pool for coherent, atomic dma 729e52347bdSJani Nikula allocations, by default set to 256K. 730e52347bdSJani Nikula 731e52347bdSJani Nikula com20020= [HW,NET] ARCnet - COM20020 chipset 732e52347bdSJani Nikula Format: 733e52347bdSJani Nikula <io>[,<irq>[,<nodeID>[,<backplane>[,<ckp>[,<timeout>]]]]] 734e52347bdSJani Nikula 735e52347bdSJani Nikula com90io= [HW,NET] ARCnet - COM90xx chipset (IO-mapped buffers) 736e52347bdSJani Nikula Format: <io>[,<irq>] 737e52347bdSJani Nikula 738e52347bdSJani Nikula com90xx= [HW,NET] 739e52347bdSJani Nikula ARCnet - COM90xx chipset (memory-mapped buffers) 740e52347bdSJani Nikula Format: <io>[,<irq>[,<memstart>]] 741e52347bdSJani Nikula 742e52347bdSJani Nikula condev= [HW,S390] console device 743e52347bdSJani Nikula conmode= 744e52347bdSJani Nikula 7451f3307cfSThomas Richter con3215_drop= [S390] 3215 console drop mode. 7461f3307cfSThomas Richter Format: y|n|Y|N|1|0 7471f3307cfSThomas Richter When set to true, drop data on the 3215 console when 7481f3307cfSThomas Richter the console buffer is full. In this case the 7491f3307cfSThomas Richter operator using a 3270 terminal emulator (for example 7501f3307cfSThomas Richter x3270) does not have to enter the clear key for the 7511f3307cfSThomas Richter console output to advance and the kernel to continue. 7521f3307cfSThomas Richter This leads to a much faster boot time when a 3270 7531f3307cfSThomas Richter terminal emulator is active. If no 3270 terminal 7541f3307cfSThomas Richter emulator is used, this parameter has no effect. 7551f3307cfSThomas Richter 756e52347bdSJani Nikula console= [KNL] Output console device and options. 757e52347bdSJani Nikula 758e52347bdSJani Nikula tty<n> Use the virtual console device <n>. 759e52347bdSJani Nikula 760e52347bdSJani Nikula ttyS<n>[,options] 761e52347bdSJani Nikula ttyUSB0[,options] 762e52347bdSJani Nikula Use the specified serial port. The options are of 763e52347bdSJani Nikula the form "bbbbpnf", where "bbbb" is the baud rate, 764e52347bdSJani Nikula "p" is parity ("n", "o", or "e"), "n" is number of 765e52347bdSJani Nikula bits, and "f" is flow control ("r" for RTS or 766e52347bdSJani Nikula omit it). Default is "9600n8". 767e52347bdSJani Nikula 768e52347bdSJani Nikula See Documentation/admin-guide/serial-console.rst for more 769e52347bdSJani Nikula information. See 770d9d6ef25SMauro Carvalho Chehab Documentation/networking/netconsole.rst for an 771e52347bdSJani Nikula alternative. 772e52347bdSJani Nikula 773e52347bdSJani Nikula uart[8250],io,<addr>[,options] 774e52347bdSJani Nikula uart[8250],mmio,<addr>[,options] 775e52347bdSJani Nikula uart[8250],mmio16,<addr>[,options] 776e52347bdSJani Nikula uart[8250],mmio32,<addr>[,options] 777e52347bdSJani Nikula uart[8250],0x<addr>[,options] 778e52347bdSJani Nikula Start an early, polled-mode console on the 8250/16550 779e52347bdSJani Nikula UART at the specified I/O port or MMIO address, 780e52347bdSJani Nikula switching to the matching ttyS device later. 781e52347bdSJani Nikula MMIO inter-register address stride is either 8-bit 782e52347bdSJani Nikula (mmio), 16-bit (mmio16), or 32-bit (mmio32). 783e52347bdSJani Nikula If none of [io|mmio|mmio16|mmio32], <addr> is assumed 784e52347bdSJani Nikula to be equivalent to 'mmio'. 'options' are specified in 785e52347bdSJani Nikula the same format described for ttyS above; if unspecified, 786e52347bdSJani Nikula the h/w is not re-initialized. 787e52347bdSJani Nikula 788e52347bdSJani Nikula hvc<n> Use the hypervisor console device <n>. This is for 789e52347bdSJani Nikula both Xen and PowerPC hypervisors. 790e52347bdSJani Nikula 79196b02f2fSRandy Dunlap { null | "" } 79296b02f2fSRandy Dunlap Use to disable console output, i.e., to have kernel 79396b02f2fSRandy Dunlap console messages discarded. 79496b02f2fSRandy Dunlap This must be the only console= parameter used on the 79596b02f2fSRandy Dunlap kernel command line. 79696b02f2fSRandy Dunlap 797e52347bdSJani Nikula If the device connected to the port is not a TTY but a braille 798e52347bdSJani Nikula device, prepend "brl," before the device type, for instance 799e52347bdSJani Nikula console=brl,ttyS0 800e52347bdSJani Nikula For now, only VisioBraille is supported. 801e52347bdSJani Nikula 802cca10d58SSergey Senozhatsky console_msg_format= 803cca10d58SSergey Senozhatsky [KNL] Change console messages format 804cca10d58SSergey Senozhatsky default 805cca10d58SSergey Senozhatsky By default we print messages on consoles in 806cca10d58SSergey Senozhatsky "[time stamp] text\n" format (time stamp may not be 807cca10d58SSergey Senozhatsky printed, depending on CONFIG_PRINTK_TIME or 808cca10d58SSergey Senozhatsky `printk_time' param). 809cca10d58SSergey Senozhatsky syslog 810cca10d58SSergey Senozhatsky Switch to syslog format: "<%u>[time stamp] text\n" 811cca10d58SSergey Senozhatsky IOW, each message will have a facility and loglevel 812cca10d58SSergey Senozhatsky prefix. The format is similar to one used by syslog() 813cca10d58SSergey Senozhatsky syscall, or to executing "dmesg -S --raw" or to reading 814cca10d58SSergey Senozhatsky from /proc/kmsg. 815cca10d58SSergey Senozhatsky 816e52347bdSJani Nikula consoleblank= [KNL] The console blank (screen saver) timeout in 817ac0a314cSDaniel Xu seconds. A value of 0 disables the blank timer. 818ac0a314cSDaniel Xu Defaults to 0. 819e52347bdSJani Nikula 820e52347bdSJani Nikula coredump_filter= 821e52347bdSJani Nikula [KNL] Change the default value for 822e52347bdSJani Nikula /proc/<pid>/coredump_filter. 823cd4ca341SJimmy Assarsson See also Documentation/filesystems/proc.rst. 824e52347bdSJani Nikula 82562a31ce1SLeo Yan coresight_cpu_debug.enable 82662a31ce1SLeo Yan [ARM,ARM64] 82762a31ce1SLeo Yan Format: <bool> 82862a31ce1SLeo Yan Enable/disable the CPU sampling based debugging. 82962a31ce1SLeo Yan 0: default value, disable debugging 83062a31ce1SLeo Yan 1: enable debugging at boot time 83162a31ce1SLeo Yan 832389cfd96SRandy Dunlap cpcihp_generic= [HW,PCI] Generic port I/O CompactPCI driver 833389cfd96SRandy Dunlap Format: 834389cfd96SRandy Dunlap <first_slot>,<last_slot>,<port>,<enum_bit>[,<debug>] 835389cfd96SRandy Dunlap 836e52347bdSJani Nikula cpuidle.off=1 [CPU_IDLE] 837e52347bdSJani Nikula disable the cpuidle sub-system 838e52347bdSJani Nikula 83961cb5758SRafael J. Wysocki cpuidle.governor= 84061cb5758SRafael J. Wysocki [CPU_IDLE] Name of the cpuidle governor to use. 84161cb5758SRafael J. Wysocki 842d82f2692SLen Brown cpufreq.off=1 [CPU_FREQ] 843d82f2692SLen Brown disable the cpufreq sub-system 844d82f2692SLen Brown 8458412b456SQuentin Perret cpufreq.default_governor= 8468412b456SQuentin Perret [CPU_FREQ] Name of the default cpufreq governor or 8478412b456SQuentin Perret policy to use. This governor must be registered in the 8488412b456SQuentin Perret kernel before the cpufreq driver probes. 8498412b456SQuentin Perret 850e52347bdSJani Nikula cpu_init_udelay=N 851e52347bdSJani Nikula [X86] Delay for N microsec between assert and de-assert 852e52347bdSJani Nikula of APIC INIT to start processors. This delay occurs 853e52347bdSJani Nikula on every CPU online, such as boot, and resume from suspend. 854e52347bdSJani Nikula Default: 10000 855e52347bdSJani Nikula 85618415f33SThomas Gleixner cpuhp.parallel= 85718415f33SThomas Gleixner [SMP] Enable/disable parallel bringup of secondary CPUs 85818415f33SThomas Gleixner Format: <bool> 85918415f33SThomas Gleixner Default is enabled if CONFIG_HOTPLUG_PARALLEL=y. Otherwise 86018415f33SThomas Gleixner the parameter has no effect. 86118415f33SThomas Gleixner 862389cfd96SRandy Dunlap crash_kexec_post_notifiers 863389cfd96SRandy Dunlap Run kdump after running panic-notifiers and dumping 864389cfd96SRandy Dunlap kmsg. This only for the users who doubt kdump always 865389cfd96SRandy Dunlap succeeds in any situation. 866389cfd96SRandy Dunlap Note that this also increases risks of kdump failure, 867389cfd96SRandy Dunlap because some panic notifiers can make the crashed 868389cfd96SRandy Dunlap kernel more unstable. 869e52347bdSJani Nikula 870e52347bdSJani Nikula crashkernel=size[KMG][@offset[KMG]] 871e52347bdSJani Nikula [KNL] Using kexec, Linux can switch to a 'crash kernel' 872e52347bdSJani Nikula upon panic. This parameter reserves the physical 873e52347bdSJani Nikula memory region [offset, offset + size] for that kernel 874e52347bdSJani Nikula image. If '@offset' is omitted, then a suitable offset 875b9ac3849SDave Young is selected automatically. 87633f0dd97SChen Jiahao [KNL, X86-64, ARM64, RISCV] Select a region under 4G first, and 877b9ac3849SDave Young fall back to reserve region above 4G when '@offset' 878b9ac3849SDave Young hasn't been specified. 879330d4810SMauro Carvalho Chehab See Documentation/admin-guide/kdump/kdump.rst for further details. 880e52347bdSJani Nikula 881e52347bdSJani Nikula crashkernel=range1:size1[,range2:size2,...][@offset] 882e52347bdSJani Nikula [KNL] Same as above, but depends on the memory 883e52347bdSJani Nikula in the running system. The syntax of range is 884e52347bdSJani Nikula start-[end] where start and end are both 885e52347bdSJani Nikula a memory unit (amount[KMG]). See also 886330d4810SMauro Carvalho Chehab Documentation/admin-guide/kdump/kdump.rst for an example. 887e52347bdSJani Nikula 888e52347bdSJani Nikula crashkernel=size[KMG],high 88933f0dd97SChen Jiahao [KNL, X86-64, ARM64, RISCV] range could be above 4G. 89033f0dd97SChen Jiahao Allow kernel to allocate physical memory region from top, 89133f0dd97SChen Jiahao so could be above 4G if system have more than 4G ram 89233f0dd97SChen Jiahao installed. Otherwise memory region will be allocated 89333f0dd97SChen Jiahao below 4G, if available. 894e52347bdSJani Nikula It will be ignored if crashkernel=X is specified. 895e52347bdSJani Nikula crashkernel=size[KMG],low 89633f0dd97SChen Jiahao [KNL, X86-64, ARM64, RISCV] range under 4G. When crashkernel=X,high 897e52347bdSJani Nikula is passed, kernel could allocate physical memory region 898e52347bdSJani Nikula above 4G, that cause second kernel crash on system 899e52347bdSJani Nikula that require some amount of low memory, e.g. swiotlb 900e52347bdSJani Nikula requires at least 64M+32K low memory, also enough extra 901e52347bdSJani Nikula low memory is needed to make sure DMA buffers for 32-bit 9025832f1aeSZhen Lei devices won't run out. Kernel would try to allocate 903a149cf00SZhen Lei default size of memory below 4G automatically. The default 904a149cf00SZhen Lei size is platform dependent. 905a149cf00SZhen Lei --> x86: max(swiotlb_size_or_default() + 8MiB, 256MiB) 906a149cf00SZhen Lei --> arm64: 128MiB 90733f0dd97SChen Jiahao --> riscv: 128MiB 9088f0f104eSZhen Lei This one lets the user specify own low range under 4G 909e52347bdSJani Nikula for second kernel instead. 910e52347bdSJani Nikula 0: to disable low allocation. 911e52347bdSJani Nikula It will be ignored when crashkernel=X,high is not used 912e52347bdSJani Nikula or memory reserved is below 4G. 913e52347bdSJani Nikula 914e52347bdSJani Nikula cryptomgr.notests 915e52347bdSJani Nikula [KNL] Disable crypto self-tests 916e52347bdSJani Nikula 917e52347bdSJani Nikula cs89x0_dma= [HW,NET] 918e52347bdSJani Nikula Format: <dma> 919e52347bdSJani Nikula 920e52347bdSJani Nikula cs89x0_media= [HW,NET] 921e52347bdSJani Nikula Format: { rj45 | aui | bnc } 922e52347bdSJani Nikula 923203e4358SPaul E. McKenney csdlock_debug= [KNL] Enable or disable debug add-ons of cross-CPU 924203e4358SPaul E. McKenney function call handling. When switched on, 925203e4358SPaul E. McKenney additional debug data is printed to the console 926203e4358SPaul E. McKenney in case a hanging CPU is detected, and that 927203e4358SPaul E. McKenney CPU is pinged again in order to try to resolve 928203e4358SPaul E. McKenney the hang situation. The default value of this 929203e4358SPaul E. McKenney option depends on the CSD_LOCK_WAIT_DEBUG_DEFAULT 930c5219860SPaul E. McKenney Kconfig option. 9318d0968ccSJuergen Gross 932e52347bdSJani Nikula dasd= [HW,NET] 933e52347bdSJani Nikula See header of drivers/s390/block/dasd_devmap.c. 934e52347bdSJani Nikula 935e52347bdSJani Nikula db9.dev[2|3]= [HW,JOY] Multisystem joystick support via parallel port 936e52347bdSJani Nikula (one device per port) 937e52347bdSJani Nikula Format: <port#>,<type> 9381752118dSTom Saeger See also Documentation/input/devices/joystick-parport.rst 939e52347bdSJani Nikula 940e52347bdSJani Nikula debug [KNL] Enable kernel debugging (events log level). 941e52347bdSJani Nikula 9423672476eSTobin C. Harding debug_boot_weak_hash 9433672476eSTobin C. Harding [KNL] Enable printing [hashed] pointers early in the 9443672476eSTobin C. Harding boot sequence. If enabled, we use a weak hash instead 9453672476eSTobin C. Harding of siphash to hash pointers. Use this option if you are 9463672476eSTobin C. Harding seeing instances of '(___ptrval___)') and need to see a 9473672476eSTobin C. Harding value (hashed pointer) instead. Cryptographically 9483672476eSTobin C. Harding insecure, please do not use on production kernels. 9493672476eSTobin C. Harding 950e52347bdSJani Nikula debug_locks_verbose= 9515831c0f7SPeter Zijlstra [KNL] verbose locking self-tests 9525831c0f7SPeter Zijlstra Format: <int> 953e52347bdSJani Nikula Print debugging info while doing the locking API 954e52347bdSJani Nikula self-tests. 9555831c0f7SPeter Zijlstra Bitmask for the various LOCKTYPE_ tests. Defaults to 0 9565831c0f7SPeter Zijlstra (no extra messages), setting it to -1 (all bits set) 9575831c0f7SPeter Zijlstra will print _a_lot_ more information - normally only 9585831c0f7SPeter Zijlstra useful to lockdep developers. 959e52347bdSJani Nikula 960e52347bdSJani Nikula debug_objects [KNL] Enable object debugging 961e52347bdSJani Nikula 962e52347bdSJani Nikula debug_guardpage_minorder= 963e52347bdSJani Nikula [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this 964e52347bdSJani Nikula parameter allows control of the order of pages that will 965e52347bdSJani Nikula be intentionally kept free (and hence protected) by the 966e52347bdSJani Nikula buddy allocator. Bigger value increase the probability 967e52347bdSJani Nikula of catching random memory corruption, but reduce the 968e52347bdSJani Nikula amount of memory for normal system use. The maximum 969e52347bdSJani Nikula possible value is MAX_ORDER/2. Setting this parameter 970e52347bdSJani Nikula to 1 or 2 should be enough to identify most random 971e52347bdSJani Nikula memory corruption problems caused by bugs in kernel or 972e52347bdSJani Nikula driver code when a CPU writes to (or reads from) a 973e52347bdSJani Nikula random memory location. Note that there exists a class 974e52347bdSJani Nikula of memory corruptions problems caused by buggy H/W or 975dbeb56feSRandy Dunlap F/W or by drivers badly programming DMA (basically when 976e52347bdSJani Nikula memory is written at bus level and the CPU MMU is 977e52347bdSJani Nikula bypassed) which are not detectable by 978e52347bdSJani Nikula CONFIG_DEBUG_PAGEALLOC, hence this option will not help 979e52347bdSJani Nikula tracking down these problems. 980e52347bdSJani Nikula 981e52347bdSJani Nikula debug_pagealloc= 9823972f6bbSVlastimil Babka [KNL] When CONFIG_DEBUG_PAGEALLOC is set, this parameter 9833972f6bbSVlastimil Babka enables the feature at boot time. By default, it is 9843972f6bbSVlastimil Babka disabled and the system will work mostly the same as a 9853972f6bbSVlastimil Babka kernel built without CONFIG_DEBUG_PAGEALLOC. 9868974558fSVlastimil Babka Note: to get most of debug_pagealloc error reports, it's 9878974558fSVlastimil Babka useful to also enable the page_owner functionality. 988e52347bdSJani Nikula on: enable the feature 989e52347bdSJani Nikula 990a24c6f7bSPeter Enderborg debugfs= [KNL] This parameter enables what is exposed to userspace 991a24c6f7bSPeter Enderborg and debugfs internal clients. 992a24c6f7bSPeter Enderborg Format: { on, no-mount, off } 993a24c6f7bSPeter Enderborg on: All functions are enabled. 994a24c6f7bSPeter Enderborg no-mount: 995a24c6f7bSPeter Enderborg Filesystem is not registered but kernel clients can 996a24c6f7bSPeter Enderborg access APIs and a crashkernel can be used to read 997a24c6f7bSPeter Enderborg its content. There is nothing to mount. 998a24c6f7bSPeter Enderborg off: Filesystem is not registered and clients 999a24c6f7bSPeter Enderborg get a -EPERM as result when trying to register files 1000a24c6f7bSPeter Enderborg or directories within debugfs. 1001a24c6f7bSPeter Enderborg This is equivalent of the runtime functionality if 1002a24c6f7bSPeter Enderborg debugfs was not enabled in the kernel at all. 1003a24c6f7bSPeter Enderborg Default value is set in build-time with a kernel configuration. 1004a24c6f7bSPeter Enderborg 1005e52347bdSJani Nikula debugpat [X86] Enable PAT debugging 1006e52347bdSJani Nikula 1007e52347bdSJani Nikula default_hugepagesz= 1008282f4214SMike Kravetz [HW] The size of the default HugeTLB page. This is 1009282f4214SMike Kravetz the size represented by the legacy /proc/ hugepages 1010282f4214SMike Kravetz APIs. In addition, this is the default hugetlb size 1011282f4214SMike Kravetz used for shmget(), mmap() and mounting hugetlbfs 1012282f4214SMike Kravetz filesystems. If not specified, defaults to the 1013282f4214SMike Kravetz architecture's default huge page size. Huge page 1014282f4214SMike Kravetz sizes are architecture dependent. See also 1015282f4214SMike Kravetz Documentation/admin-guide/mm/hugetlbpage.rst. 1016282f4214SMike Kravetz Format: size[KMG] 1017e52347bdSJani Nikula 101825b4e70dSRob Herring deferred_probe_timeout= 101925b4e70dSRob Herring [KNL] Debugging option to set a timeout in seconds for 102025b4e70dSRob Herring deferred probe to give up waiting on dependencies to 102125b4e70dSRob Herring probe. Only specific dependencies (subsystems or 10222b28a1a8SSaravana Kannan drivers) that have opted in will be ignored. A timeout 10232b28a1a8SSaravana Kannan of 0 will timeout at the end of initcalls. If the time 10242b28a1a8SSaravana Kannan out hasn't expired, it'll be restarted by each 10252b28a1a8SSaravana Kannan successful driver registration. This option will also 102625b4e70dSRob Herring dump out devices still on the deferred probe list after 102725b4e70dSRob Herring retrying. 102825b4e70dSRob Herring 1029d2fc83c1SRandy Dunlap delayacct [KNL] Enable per-task delay accounting 1030d2fc83c1SRandy Dunlap 10311b089084SArmin Wolf dell_smm_hwmon.ignore_dmi= 10321b089084SArmin Wolf [HW] Continue probing hardware even if DMI data 10331b089084SArmin Wolf indicates that the driver is running on unsupported 10341b089084SArmin Wolf hardware. 10351b089084SArmin Wolf 10361b089084SArmin Wolf dell_smm_hwmon.force= 10371b089084SArmin Wolf [HW] Activate driver even if SMM BIOS signature does 10381b089084SArmin Wolf not match list of supported models and enable otherwise 10391b089084SArmin Wolf blacklisted features. 10401b089084SArmin Wolf 10411b089084SArmin Wolf dell_smm_hwmon.power_status= 10421b089084SArmin Wolf [HW] Report power status in /proc/i8k 10431b089084SArmin Wolf (disabled by default). 10441b089084SArmin Wolf 10451b089084SArmin Wolf dell_smm_hwmon.restricted= 10461b089084SArmin Wolf [HW] Allow controlling fans only if SYS_ADMIN 10471b089084SArmin Wolf capability is set. 10481b089084SArmin Wolf 104999fdc587SArmin Wolf dell_smm_hwmon.fan_mult= 105099fdc587SArmin Wolf [HW] Factor to multiply fan speed with. 105199fdc587SArmin Wolf 105299fdc587SArmin Wolf dell_smm_hwmon.fan_max= 105399fdc587SArmin Wolf [HW] Maximum configurable fan speed. 105499fdc587SArmin Wolf 1055c65e6815SMikhail Zaslonko dfltcc= [HW,S390] 1056c65e6815SMikhail Zaslonko Format: { on | off | def_only | inf_only | always } 1057c65e6815SMikhail Zaslonko on: s390 zlib hardware support for compression on 1058c65e6815SMikhail Zaslonko level 1 and decompression (default) 1059c65e6815SMikhail Zaslonko off: No s390 zlib hardware support 1060c65e6815SMikhail Zaslonko def_only: s390 zlib hardware support for deflate 1061c65e6815SMikhail Zaslonko only (compression on level 1) 1062c65e6815SMikhail Zaslonko inf_only: s390 zlib hardware support for inflate 1063c65e6815SMikhail Zaslonko only (decompression) 1064c65e6815SMikhail Zaslonko always: Same as 'on' but ignores the selected compression 1065c65e6815SMikhail Zaslonko level always using hardware support (used for debugging) 1066c65e6815SMikhail Zaslonko 1067e52347bdSJani Nikula dhash_entries= [KNL] 1068e52347bdSJani Nikula Set number of hash buckets for dentry cache. 1069e52347bdSJani Nikula 1070e52347bdSJani Nikula disable_1tb_segments [PPC] 1071e52347bdSJani Nikula Disables the use of 1TB hash page table segments. This 1072e52347bdSJani Nikula causes the kernel to fall back to 256MB segments which 1073e52347bdSJani Nikula can be useful when debugging issues that require an SLB 1074e52347bdSJani Nikula miss to occur. 1075e52347bdSJani Nikula 1076e52347bdSJani Nikula disable= [IPV6] 107719093313SMauro Carvalho Chehab See Documentation/networking/ipv6.rst. 1078e52347bdSJani Nikula 1079e52347bdSJani Nikula disable_radix [PPC] 1080e52347bdSJani Nikula Disable RADIX MMU mode on POWER9 1081e52347bdSJani Nikula 10822275d7b5SNicholas Piggin disable_tlbie [PPC] 10832275d7b5SNicholas Piggin Disable TLBIE instruction. Currently does not work 10842275d7b5SNicholas Piggin with KVM, with HASH MMU, or with coherent accelerators. 10852275d7b5SNicholas Piggin 1086e52347bdSJani Nikula disable_cpu_apicid= [X86,APIC,SMP] 1087e52347bdSJani Nikula Format: <int> 1088e52347bdSJani Nikula The number of initial APIC ID for the 1089e52347bdSJani Nikula corresponding CPU to be disabled at boot, 1090e52347bdSJani Nikula mostly used for the kdump 2nd kernel to 1091e52347bdSJani Nikula disable BSP to wake up multiple CPUs without 1092e52347bdSJani Nikula causing system reset or hang due to sending 1093e52347bdSJani Nikula INIT from AP to BSP. 1094e52347bdSJani Nikula 1095e52347bdSJani Nikula disable_ddw [PPC/PSERIES] 1096255bf90fSRandy Dunlap Disable Dynamic DMA Window support. Use this 1097e52347bdSJani Nikula to workaround buggy firmware. 1098e52347bdSJani Nikula 1099e52347bdSJani Nikula disable_ipv6= [IPV6] 110019093313SMauro Carvalho Chehab See Documentation/networking/ipv6.rst. 1101e52347bdSJani Nikula 1102e52347bdSJani Nikula disable_mtrr_cleanup [X86] 1103e52347bdSJani Nikula The kernel tries to adjust MTRR layout from continuous 1104e52347bdSJani Nikula to discrete, to make X server driver able to add WB 1105e52347bdSJani Nikula entry later. This parameter disables that. 1106e52347bdSJani Nikula 1107e52347bdSJani Nikula disable_mtrr_trim [X86, Intel and AMD only] 1108e52347bdSJani Nikula By default the kernel will trim any uncacheable 1109e52347bdSJani Nikula memory out of your available memory pool based on 1110e52347bdSJani Nikula MTRR settings. This parameter disables that behavior, 1111e52347bdSJani Nikula possibly causing your machine to run very slowly. 1112e52347bdSJani Nikula 1113e52347bdSJani Nikula disable_timer_pin_1 [X86] 1114e52347bdSJani Nikula Disable PIN 1 of APIC timer 1115e52347bdSJani Nikula Can be useful to work around chipset bugs. 1116e52347bdSJani Nikula 1117e52347bdSJani Nikula dis_ucode_ldr [X86] Disable the microcode loader. 1118e52347bdSJani Nikula 1119e52347bdSJani Nikula dma_debug=off If the kernel is compiled with DMA_API_DEBUG support, 1120e52347bdSJani Nikula this option disables the debugging code at boot. 1121e52347bdSJani Nikula 1122e52347bdSJani Nikula dma_debug_entries=<number> 1123e52347bdSJani Nikula This option allows to tune the number of preallocated 1124e52347bdSJani Nikula entries for DMA-API debugging code. One entry is 1125e52347bdSJani Nikula required per DMA-API allocation. Use this if the 1126e52347bdSJani Nikula DMA-API debugging code disables itself because the 1127e52347bdSJani Nikula architectural default is too low. 1128e52347bdSJani Nikula 1129e52347bdSJani Nikula dma_debug_driver=<driver_name> 1130e52347bdSJani Nikula With this option the DMA-API debugging driver 1131e52347bdSJani Nikula filter feature can be enabled at boot time. Just 1132e52347bdSJani Nikula pass the driver to filter for as the parameter. 1133e52347bdSJani Nikula The filter can be disabled or changed to another 1134e52347bdSJani Nikula driver later using sysfs. 1135e52347bdSJani Nikula 113677018fb9SPawan Gupta reg_file_data_sampling= 113777018fb9SPawan Gupta [X86] Controls mitigation for Register File Data 113877018fb9SPawan Gupta Sampling (RFDS) vulnerability. RFDS is a CPU 113977018fb9SPawan Gupta vulnerability which may allow userspace to infer 114077018fb9SPawan Gupta kernel data values previously stored in floating point 114177018fb9SPawan Gupta registers, vector registers, or integer registers. 114277018fb9SPawan Gupta RFDS only affects Intel Atom processors. 114377018fb9SPawan Gupta 114477018fb9SPawan Gupta on: Turns ON the mitigation. 114577018fb9SPawan Gupta off: Turns OFF the mitigation. 114677018fb9SPawan Gupta 114777018fb9SPawan Gupta This parameter overrides the compile time default set 114877018fb9SPawan Gupta by CONFIG_MITIGATION_RFDS. Mitigation cannot be 114977018fb9SPawan Gupta disabled when other VERW based mitigations (like MDS) 115077018fb9SPawan Gupta are enabled. In order to disable RFDS mitigation all 115177018fb9SPawan Gupta VERW based mitigations need to be disabled. 115277018fb9SPawan Gupta 115377018fb9SPawan Gupta For details see: 115477018fb9SPawan Gupta Documentation/admin-guide/hw-vuln/reg-file-data-sampling.rst 115577018fb9SPawan Gupta 11561ea61b68SFeng Tang driver_async_probe= [KNL] 1157f79f662eSSaravana Kannan List of driver names to be probed asynchronously. * 1158f79f662eSSaravana Kannan matches with all driver names. If * is specified, the 1159f79f662eSSaravana Kannan rest of the listed driver names are those that will NOT 1160f79f662eSSaravana Kannan match the *. 11611ea61b68SFeng Tang Format: <driver_name1>,<driver_name2>... 11621ea61b68SFeng Tang 116353fd40a9SJani Nikula drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>] 1164e52347bdSJani Nikula Broken monitors, graphic adapters, KVMs and EDIDless 1165e52347bdSJani Nikula panels may send no or incorrect EDID data sets. 1166e52347bdSJani Nikula This parameter allows to specify an EDID data sets 1167e52347bdSJani Nikula in the /lib/firmware directory that are used instead. 1168e52347bdSJani Nikula Generic built-in EDID data sets are used, if one of 1169e52347bdSJani Nikula edid/1024x768.bin, edid/1280x1024.bin, 1170e52347bdSJani Nikula edid/1680x1050.bin, or edid/1920x1080.bin is given 1171e52347bdSJani Nikula and no file with the same name exists. Details and 1172e52347bdSJani Nikula instructions how to build your own EDID data are 1173cd4ca341SJimmy Assarsson available in Documentation/admin-guide/edid.rst. An EDID 1174e52347bdSJani Nikula data set will only be used for a particular connector, 1175e52347bdSJani Nikula if its name and a colon are prepended to the EDID 1176e52347bdSJani Nikula name. Each connector may use a unique EDID data 1177e52347bdSJani Nikula set by separating the files with a comma. An EDID 1178e52347bdSJani Nikula data set with no connector name will be used for 1179e52347bdSJani Nikula any connectors not explicitly specified. 1180e52347bdSJani Nikula 1181e52347bdSJani Nikula dscc4.setup= [NET] 1182e52347bdSJani Nikula 1183a2b05b7aSNicholas Piggin dt_cpu_ftrs= [PPC] 1184a2b05b7aSNicholas Piggin Format: {"off" | "known"} 1185a2b05b7aSNicholas Piggin Control how the dt_cpu_ftrs device-tree binding is 1186a2b05b7aSNicholas Piggin used for CPU feature discovery and setup (if it 1187a2b05b7aSNicholas Piggin exists). 1188a2b05b7aSNicholas Piggin off: Do not use it, fall back to legacy cpu table. 1189a2b05b7aSNicholas Piggin known: Do not pass through unknown features to guests 1190a2b05b7aSNicholas Piggin or userspace, only those that the kernel is aware of. 1191a2b05b7aSNicholas Piggin 1192e7aa8c2eSLinus Torvalds dump_apple_properties [X86] 1193e7aa8c2eSLinus Torvalds Dump name and content of EFI device properties on 1194e7aa8c2eSLinus Torvalds x86 Macs. Useful for driver authors to determine 1195e7aa8c2eSLinus Torvalds what data is available or for reverse-engineering. 1196e7aa8c2eSLinus Torvalds 1197e52347bdSJani Nikula dyndbg[="val"] [KNL,DYNAMIC_DEBUG] 1198307e3ee9SRandy Dunlap <module>.dyndbg[="val"] 1199e52347bdSJani Nikula Enable debug messages at boot time. See 1200787e3075SSteffen Maier Documentation/admin-guide/dynamic-debug-howto.rst 1201787e3075SSteffen Maier for details. 1202e52347bdSJani Nikula 1203e52347bdSJani Nikula early_ioremap_debug [KNL] 1204e52347bdSJani Nikula Enable debug messages in early_ioremap support. This 1205e52347bdSJani Nikula is useful for tracking down temporary early mappings 1206e52347bdSJani Nikula which are not unmapped. 1207e52347bdSJani Nikula 1208e52347bdSJani Nikula earlycon= [KNL] Output early console device and options. 1209e52347bdSJani Nikula 1210e18409c0SChristoph Hellwig When used with no options, the early console is 1211e18409c0SChristoph Hellwig determined by stdout-path property in device tree's 1212e18409c0SChristoph Hellwig chosen node or the ACPI SPCR table if supported by 1213e18409c0SChristoph Hellwig the platform. 1214e52347bdSJani Nikula 1215e52347bdSJani Nikula cdns,<addr>[,options] 1216e52347bdSJani Nikula Start an early, polled-mode console on a Cadence 1217e52347bdSJani Nikula (xuartps) serial port at the specified address. Only 1218e52347bdSJani Nikula supported option is baud rate. If baud rate is not 1219e52347bdSJani Nikula specified, the serial port must already be setup and 1220e52347bdSJani Nikula configured. 1221e52347bdSJani Nikula 1222374b30f2SRicardo Ribalda uart[8250],io,<addr>[,options[,uartclk]] 1223374b30f2SRicardo Ribalda uart[8250],mmio,<addr>[,options[,uartclk]] 1224374b30f2SRicardo Ribalda uart[8250],mmio32,<addr>[,options[,uartclk]] 1225374b30f2SRicardo Ribalda uart[8250],mmio32be,<addr>[,options[,uartclk]] 1226e52347bdSJani Nikula uart[8250],0x<addr>[,options] 1227e52347bdSJani Nikula Start an early, polled-mode console on the 8250/16550 1228e52347bdSJani Nikula UART at the specified I/O port or MMIO address. 1229e52347bdSJani Nikula MMIO inter-register address stride is either 8-bit 1230e52347bdSJani Nikula (mmio) or 32-bit (mmio32 or mmio32be). 1231e52347bdSJani Nikula If none of [io|mmio|mmio32|mmio32be], <addr> is assumed 1232e52347bdSJani Nikula to be equivalent to 'mmio'. 'options' are specified 1233e52347bdSJani Nikula in the same format described for "console=ttyS<n>"; if 1234374b30f2SRicardo Ribalda unspecified, the h/w is not initialized. 'uartclk' is 1235374b30f2SRicardo Ribalda the uart clock frequency; if unspecified, it is set 1236374b30f2SRicardo Ribalda to 'BASE_BAUD' * 16. 1237e52347bdSJani Nikula 1238e52347bdSJani Nikula pl011,<addr> 1239e52347bdSJani Nikula pl011,mmio32,<addr> 1240e52347bdSJani Nikula Start an early, polled-mode console on a pl011 serial 1241e52347bdSJani Nikula port at the specified address. The pl011 serial port 1242e52347bdSJani Nikula must already be setup and configured. Options are not 1243e52347bdSJani Nikula yet supported. If 'mmio32' is specified, then only 1244e52347bdSJani Nikula the driver will use only 32-bit accessors to read/write 1245e52347bdSJani Nikula the device registers. 1246e52347bdSJani Nikula 12474bc2bd5aSStafford Horne liteuart,<addr> 12484bc2bd5aSStafford Horne Start an early console on a litex serial port at the 12494bc2bd5aSStafford Horne specified address. The serial port must already be 12504bc2bd5aSStafford Horne setup and configured. Options are not yet supported. 12514bc2bd5aSStafford Horne 1252e52347bdSJani Nikula meson,<addr> 1253e52347bdSJani Nikula Start an early, polled-mode console on a meson serial 1254e52347bdSJani Nikula port at the specified address. The serial port must 1255e52347bdSJani Nikula already be setup and configured. Options are not yet 1256e52347bdSJani Nikula supported. 1257e52347bdSJani Nikula 1258e52347bdSJani Nikula msm_serial,<addr> 1259e52347bdSJani Nikula Start an early, polled-mode console on an msm serial 1260e52347bdSJani Nikula port at the specified address. The serial port 1261e52347bdSJani Nikula must already be setup and configured. Options are not 1262e52347bdSJani Nikula yet supported. 1263e52347bdSJani Nikula 1264e52347bdSJani Nikula msm_serial_dm,<addr> 1265e52347bdSJani Nikula Start an early, polled-mode console on an msm serial 1266e52347bdSJani Nikula dm port at the specified address. The serial port 1267e52347bdSJani Nikula must already be setup and configured. Options are not 1268e52347bdSJani Nikula yet supported. 1269e52347bdSJani Nikula 1270e36361d7SAndreas Färber owl,<addr> 1271e36361d7SAndreas Färber Start an early, polled-mode console on a serial port 1272e36361d7SAndreas Färber of an Actions Semi SoC, such as S500 or S900, at the 1273e36361d7SAndreas Färber specified address. The serial port must already be 1274e36361d7SAndreas Färber setup and configured. Options are not yet supported. 1275e36361d7SAndreas Färber 1276c10b1332SManivannan Sadhasivam rda,<addr> 1277c10b1332SManivannan Sadhasivam Start an early, polled-mode console on a serial port 1278c10b1332SManivannan Sadhasivam of an RDA Micro SoC, such as RDA8810PL, at the 1279c10b1332SManivannan Sadhasivam specified address. The serial port must already be 1280c10b1332SManivannan Sadhasivam setup and configured. Options are not yet supported. 1281c10b1332SManivannan Sadhasivam 128282f12ab3SPalmer Dabbelt sbi 128382f12ab3SPalmer Dabbelt Use RISC-V SBI (Supervisor Binary Interface) for early 128482f12ab3SPalmer Dabbelt console. 128582f12ab3SPalmer Dabbelt 1286e52347bdSJani Nikula smh Use ARM semihosting calls for early console. 1287e52347bdSJani Nikula 1288e52347bdSJani Nikula s3c2410,<addr> 1289e52347bdSJani Nikula s3c2412,<addr> 1290e52347bdSJani Nikula s3c2440,<addr> 1291e52347bdSJani Nikula s3c6400,<addr> 1292e52347bdSJani Nikula s5pv210,<addr> 1293e52347bdSJani Nikula exynos4210,<addr> 1294e52347bdSJani Nikula Use early console provided by serial driver available 1295e52347bdSJani Nikula on Samsung SoCs, requires selecting proper type and 1296e52347bdSJani Nikula a correct base address of the selected UART port. The 1297e52347bdSJani Nikula serial port must already be setup and configured. 1298e52347bdSJani Nikula Options are not yet supported. 1299e52347bdSJani Nikula 1300ec84aa0aSMartin Blumenstingl lantiq,<addr> 1301ec84aa0aSMartin Blumenstingl Start an early, polled-mode console on a lantiq serial 1302ec84aa0aSMartin Blumenstingl (lqasc) port at the specified address. The serial port 1303ec84aa0aSMartin Blumenstingl must already be setup and configured. Options are not 1304ec84aa0aSMartin Blumenstingl yet supported. 1305ec84aa0aSMartin Blumenstingl 1306e52347bdSJani Nikula lpuart,<addr> 1307e52347bdSJani Nikula lpuart32,<addr> 1308e52347bdSJani Nikula Use early console provided by Freescale LP UART driver 1309e52347bdSJani Nikula found on Freescale Vybrid and QorIQ LS1021A processors. 1310e52347bdSJani Nikula A valid base address must be provided, and the serial 1311e52347bdSJani Nikula port must already be setup and configured. 1312e52347bdSJani Nikula 13137fe068dbSJonathan Neuschäfer ec_imx21,<addr> 13147fe068dbSJonathan Neuschäfer ec_imx6q,<addr> 13157fe068dbSJonathan Neuschäfer Start an early, polled-mode, output-only console on the 13167fe068dbSJonathan Neuschäfer Freescale i.MX UART at the specified address. The UART 13177fe068dbSJonathan Neuschäfer must already be setup and configured. 13187fe068dbSJonathan Neuschäfer 1319f7c864e7SAndre Przywara ar3700_uart,<addr> 1320e52347bdSJani Nikula Start an early, polled-mode console on the 1321e52347bdSJani Nikula Armada 3700 serial port at the specified 1322e52347bdSJani Nikula address. The serial port must already be setup 1323e52347bdSJani Nikula and configured. Options are not yet supported. 1324e52347bdSJani Nikula 132543f1831bSKarthikeyan Ramasubramanian qcom_geni,<addr> 132643f1831bSKarthikeyan Ramasubramanian Start an early, polled-mode console on a Qualcomm 132743f1831bSKarthikeyan Ramasubramanian Generic Interface (GENI) based serial port at the 132843f1831bSKarthikeyan Ramasubramanian specified address. The serial port must already be 132943f1831bSKarthikeyan Ramasubramanian setup and configured. Options are not yet supported. 133043f1831bSKarthikeyan Ramasubramanian 133169c1f396SArd Biesheuvel efifb,[options] 133269c1f396SArd Biesheuvel Start an early, unaccelerated console on the EFI 133369c1f396SArd Biesheuvel memory mapped framebuffer (if available). On cache 133469c1f396SArd Biesheuvel coherent non-x86 systems that use system memory for 133569c1f396SArd Biesheuvel the framebuffer, pass the 'ram' option so that it is 133669c1f396SArd Biesheuvel mapped with the correct attributes. 133769c1f396SArd Biesheuvel 133809864c1cSStefan-gabriel Mirea linflex,<addr> 13399905f32aSStefan-Gabriel Mirea Use early console provided by Freescale LINFlexD UART 134009864c1cSStefan-gabriel Mirea serial driver for NXP S32V234 SoCs. A valid base 134109864c1cSStefan-gabriel Mirea address must be provided, and the serial port must 134209864c1cSStefan-gabriel Mirea already be setup and configured. 134309864c1cSStefan-gabriel Mirea 13444ba66a97SArnd Bergmann earlyprintk= [X86,SH,ARM,M68k,S390] 1345e52347bdSJani Nikula earlyprintk=vga 134689175cf7SHeiko Carstens earlyprintk=sclp 1347e52347bdSJani Nikula earlyprintk=xen 1348e52347bdSJani Nikula earlyprintk=serial[,ttySn[,baudrate]] 1349e52347bdSJani Nikula earlyprintk=serial[,0x...[,baudrate]] 1350e52347bdSJani Nikula earlyprintk=ttySn[,baudrate] 1351e52347bdSJani Nikula earlyprintk=dbgp[debugController#] 1352d2266bbfSFeng Tang earlyprintk=pciserial[,force],bus:device.function[,baudrate] 13531b5aeebfSLu Baolu earlyprintk=xdbc[xhciController#] 1354e52347bdSJani Nikula 1355e52347bdSJani Nikula earlyprintk is useful when the kernel crashes before 1356e52347bdSJani Nikula the normal console is initialized. It is not enabled by 1357e52347bdSJani Nikula default because it has some cosmetic problems. 1358e52347bdSJani Nikula 1359e52347bdSJani Nikula Append ",keep" to not disable it when the real console 1360e52347bdSJani Nikula takes over. 1361e52347bdSJani Nikula 136282850028SAkihiko Odaki Only one of vga, serial, or usb debug port can 1363e52347bdSJani Nikula be used at a time. 1364e52347bdSJani Nikula 1365e52347bdSJani Nikula Currently only ttyS0 and ttyS1 may be specified by 1366e52347bdSJani Nikula name. Other I/O ports may be explicitly specified 1367e52347bdSJani Nikula on some architectures (x86 and arm at least) by 1368e52347bdSJani Nikula replacing ttySn with an I/O port address, like this: 1369e52347bdSJani Nikula earlyprintk=serial,0x1008,115200 1370e52347bdSJani Nikula You can find the port for a given device in 1371e52347bdSJani Nikula /proc/tty/driver/serial: 1372e52347bdSJani Nikula 2: uart:ST16650V2 port:00001008 irq:18 ... 1373e52347bdSJani Nikula 1374e52347bdSJani Nikula Interaction with the standard serial driver is not 1375e52347bdSJani Nikula very good. 1376e52347bdSJani Nikula 137782850028SAkihiko Odaki The VGA output is eventually overwritten by 1378e52347bdSJani Nikula the real console. 1379e52347bdSJani Nikula 138042bc9716SJan Beulich The xen option can only be used in Xen domains. 1381e52347bdSJani Nikula 138289175cf7SHeiko Carstens The sclp output can only be used on s390. 138389175cf7SHeiko Carstens 1384d2266bbfSFeng Tang The optional "force" to "pciserial" enables use of a 1385d2266bbfSFeng Tang PCI device even when its classcode is not of the 1386d2266bbfSFeng Tang UART class. 1387d2266bbfSFeng Tang 1388e52347bdSJani Nikula edac_report= [HW,EDAC] Control how to report EDAC event 1389e52347bdSJani Nikula Format: {"on" | "off" | "force"} 1390e52347bdSJani Nikula on: enable EDAC to report H/W event. May be overridden 1391e52347bdSJani Nikula by other higher priority error reporting module. 1392e52347bdSJani Nikula off: disable H/W event reporting through EDAC. 1393e52347bdSJani Nikula force: enforce the use of EDAC to report H/W event. 1394e52347bdSJani Nikula default: on. 1395e52347bdSJani Nikula 1396e52347bdSJani Nikula edd= [EDD] 1397e52347bdSJani Nikula Format: {"off" | "on" | "skip[mbr]"} 1398e52347bdSJani Nikula 1399e52347bdSJani Nikula efi= [EFI] 1400c03f739fSHeinrich Schuchardt Format: { "debug", "disable_early_pci_dma", 1401c03f739fSHeinrich Schuchardt "nochunk", "noruntime", "nosoftreserve", 1402fb1201aeSArd Biesheuvel "novamap", "no_disable_early_pci_dma" } 1403c03f739fSHeinrich Schuchardt debug: enable misc debug output. 1404c03f739fSHeinrich Schuchardt disable_early_pci_dma: disable the busmaster bit on all 1405c03f739fSHeinrich Schuchardt PCI bridges while in the EFI boot stub. 1406e52347bdSJani Nikula nochunk: disable reading files in "chunks" in the EFI 1407e52347bdSJani Nikula boot stub, as chunking can cause problems with some 1408e52347bdSJani Nikula firmware implementations. 1409e52347bdSJani Nikula noruntime : disable EFI runtime services support 1410b617c526SDan Williams nosoftreserve: The EFI_MEMORY_SP (Specific Purpose) 1411b617c526SDan Williams attribute may cause the kernel to reserve the 1412b617c526SDan Williams memory range for a memory mapping driver to 1413b617c526SDan Williams claim. Specify efi=nosoftreserve to disable this 1414b617c526SDan Williams reservation and treat the memory by its base type 1415b617c526SDan Williams (i.e. EFI_CONVENTIONAL_MEMORY / "System RAM"). 1416c03f739fSHeinrich Schuchardt novamap: do not call SetVirtualAddressMap(). 14174444f854SMatthew Garrett no_disable_early_pci_dma: Leave the busmaster bit set 14184444f854SMatthew Garrett on all PCI bridges while in the EFI boot stub 1419e52347bdSJani Nikula 1420e52347bdSJani Nikula efi_no_storage_paranoia [EFI; X86] 1421e52347bdSJani Nikula Using this parameter you can use more than 50% of 1422e52347bdSJani Nikula your efi variable storage. Use this parameter only if 1423e52347bdSJani Nikula you are really sure that your UEFI does sane gc and 1424e52347bdSJani Nikula fulfills the spec otherwise your board may brick. 1425e52347bdSJani Nikula 1426e52347bdSJani Nikula efi_fake_mem= nn[KMG]@ss[KMG]:aa[,nn[KMG]@ss[KMG]:aa,..] [EFI; X86] 1427e52347bdSJani Nikula Add arbitrary attribute to specific memory range by 1428e52347bdSJani Nikula updating original EFI memory map. 1429e52347bdSJani Nikula Region of memory which aa attribute is added to is 1430e52347bdSJani Nikula from ss to ss+nn. 1431199c8471SDan Williams 1432e52347bdSJani Nikula If efi_fake_mem=2G@4G:0x10000,2G@0x10a0000000:0x10000 1433e52347bdSJani Nikula is specified, EFI_MEMORY_MORE_RELIABLE(0x10000) 1434e52347bdSJani Nikula attribute is added to range 0x100000000-0x180000000 and 1435e52347bdSJani Nikula 0x10a0000000-0x1120000000. 1436e52347bdSJani Nikula 1437199c8471SDan Williams If efi_fake_mem=8G@9G:0x40000 is specified, the 1438199c8471SDan Williams EFI_MEMORY_SP(0x40000) attribute is added to 1439199c8471SDan Williams range 0x240000000-0x43fffffff. 1440199c8471SDan Williams 1441e52347bdSJani Nikula Using this parameter you can do debugging of EFI memmap 1442199c8471SDan Williams related features. For example, you can do debugging of 1443e52347bdSJani Nikula Address Range Mirroring feature even if your box 1444199c8471SDan Williams doesn't support it, or mark specific memory as 1445199c8471SDan Williams "soft reserved". 1446e52347bdSJani Nikula 1447e52347bdSJani Nikula efivar_ssdt= [EFI; X86] Name of an EFI variable that contains an SSDT 1448e52347bdSJani Nikula that is to be dynamically loaded by Linux. If there are 1449e52347bdSJani Nikula multiple variables with the same name but with different 1450e52347bdSJani Nikula vendor GUIDs, all of them will be loaded. See 1451cb1aaebeSMauro Carvalho Chehab Documentation/admin-guide/acpi/ssdt-overlays.rst for details. 1452e52347bdSJani Nikula 1453e52347bdSJani Nikula 1454e52347bdSJani Nikula eisa_irq_edge= [PARISC,HW] 1455e52347bdSJani Nikula See header of drivers/parisc/eisa.c. 1456e52347bdSJani Nikula 1457d2fc83c1SRandy Dunlap ekgdboc= [X86,KGDB] Allow early kernel console debugging 1458d2fc83c1SRandy Dunlap Format: ekgdboc=kbd 1459d2fc83c1SRandy Dunlap 1460d2fc83c1SRandy Dunlap This is designed to be used in conjunction with 1461d2fc83c1SRandy Dunlap the boot argument: earlyprintk=vga 1462d2fc83c1SRandy Dunlap 1463d2fc83c1SRandy Dunlap This parameter works in place of the kgdboc parameter 1464d2fc83c1SRandy Dunlap but can only be used if the backing tty is available 1465d2fc83c1SRandy Dunlap very early in the boot process. For early debugging 1466d2fc83c1SRandy Dunlap via a serial port see kgdboc_earlycon instead. 1467d2fc83c1SRandy Dunlap 1468e52347bdSJani Nikula elanfreq= [X86-32] 1469e52347bdSJani Nikula See comment before function elanfreq_setup() in 1470e52347bdSJani Nikula arch/x86/kernel/cpu/cpufreq/elanfreq.c. 1471e52347bdSJani Nikula 1472e52347bdSJani Nikula elfcorehdr=[size[KMG]@]offset[KMG] [IA64,PPC,SH,X86,S390] 1473e52347bdSJani Nikula Specifies physical address of start of kernel core 1474e52347bdSJani Nikula image elf header and optionally the size. Generally 1475e52347bdSJani Nikula kexec loader will pass this option to capture kernel. 1476330d4810SMauro Carvalho Chehab See Documentation/admin-guide/kdump/kdump.rst for details. 1477e52347bdSJani Nikula 1478e52347bdSJani Nikula enable_mtrr_cleanup [X86] 1479e52347bdSJani Nikula The kernel tries to adjust MTRR layout from continuous 1480e52347bdSJani Nikula to discrete, to make X server driver able to add WB 1481e52347bdSJani Nikula entry later. This parameter enables that. 1482e52347bdSJani Nikula 1483e52347bdSJani Nikula enable_timer_pin_1 [X86] 1484e52347bdSJani Nikula Enable PIN 1 of APIC timer 1485e52347bdSJani Nikula Can be useful to work around chipset bugs 1486e52347bdSJani Nikula (in particular on some ATI chipsets). 1487e52347bdSJani Nikula The kernel tries to set a reasonable default. 1488e52347bdSJani Nikula 14898d6d51edSRandy Dunlap enforcing= [SELINUX] Set initial enforcing status. 1490e52347bdSJani Nikula Format: {"0" | "1"} 1491e52347bdSJani Nikula See security/selinux/Kconfig help text. 1492e52347bdSJani Nikula 0 -- permissive (log only, no denials). 1493e52347bdSJani Nikula 1 -- enforcing (deny and log). 1494e52347bdSJani Nikula Default value is 0. 1495d41415ebSStephen Smalley Value can be changed at runtime via 1496d41415ebSStephen Smalley /sys/fs/selinux/enforce. 1497e52347bdSJani Nikula 1498e52347bdSJani Nikula erst_disable [ACPI] 1499e52347bdSJani Nikula Disable Error Record Serialization Table (ERST) 1500e52347bdSJani Nikula support. 1501e52347bdSJani Nikula 1502e52347bdSJani Nikula ether= [HW,NET] Ethernet cards parameters 1503e52347bdSJani Nikula This option is obsoleted by the "netdev=" option, which 1504e52347bdSJani Nikula has equivalent usage. See its documentation for details. 1505e52347bdSJani Nikula 1506e52347bdSJani Nikula evm= [EVM] 1507e52347bdSJani Nikula Format: { "fix" } 1508e52347bdSJani Nikula Permit 'security.evm' to be updated regardless of 1509e52347bdSJani Nikula current integrity status. 1510e52347bdSJani Nikula 1511c4f20f14SLi Zhe early_page_ext [KNL] Enforces page_ext initialization to earlier 1512c4f20f14SLi Zhe stages so cover more early boot allocations. 1513c4f20f14SLi Zhe Please note that as side effect some optimizations 1514c4f20f14SLi Zhe might be disabled to achieve that (e.g. parallelized 1515c4f20f14SLi Zhe memory initialization is disabled) so the boot process 1516c4f20f14SLi Zhe might take longer, especially on systems with a lot of 1517c4f20f14SLi Zhe memory. Available with CONFIG_PAGE_EXTENSION=y. 1518c4f20f14SLi Zhe 1519e52347bdSJani Nikula failslab= 15202c739cedSAlbert van der Linde fail_usercopy= 1521e52347bdSJani Nikula fail_page_alloc= 1522e52347bdSJani Nikula fail_make_request=[KNL] 1523e52347bdSJani Nikula General fault injection mechanism. 1524e52347bdSJani Nikula Format: <interval>,<probability>,<space>,<times> 1525e52347bdSJani Nikula See also Documentation/fault-injection/. 1526e52347bdSJani Nikula 1527316cdaa1SMahesh Bandewar fb_tunnels= [NET] 1528316cdaa1SMahesh Bandewar Format: { initns | none } 1529316cdaa1SMahesh Bandewar See Documentation/admin-guide/sysctl/net.rst for 1530316cdaa1SMahesh Bandewar fb_tunnels_only_for_init_ns 1531316cdaa1SMahesh Bandewar 1532e52347bdSJani Nikula floppy= [HW] 1533e7751617SMauro Carvalho Chehab See Documentation/admin-guide/blockdev/floppy.rst. 1534e52347bdSJani Nikula 1535e52347bdSJani Nikula force_pal_cache_flush 1536e52347bdSJani Nikula [IA-64] Avoid check_sal_cache_flush which may hang on 1537e52347bdSJani Nikula buggy SAL_CACHE_FLUSH implementations. Using this 1538e52347bdSJani Nikula parameter will force ia64_sal_cache_flush to call 1539e52347bdSJani Nikula ia64_pal_cache_flush instead of SAL_CACHE_FLUSH. 1540e52347bdSJani Nikula 1541e52347bdSJani Nikula forcepae [X86-32] 1542e52347bdSJani Nikula Forcefully enable Physical Address Extension (PAE). 1543e52347bdSJani Nikula Many Pentium M systems disable PAE but may have a 1544e52347bdSJani Nikula functionally usable PAE implementation. 1545e52347bdSJani Nikula Warning: use of this parameter will taint the kernel 1546e52347bdSJani Nikula and may cause unknown problems. 1547e52347bdSJani Nikula 1548e52347bdSJani Nikula ftrace=[tracer] 1549e52347bdSJani Nikula [FTRACE] will set and start the specified tracer 1550e52347bdSJani Nikula as early as possible in order to facilitate early 1551e52347bdSJani Nikula boot debugging. 1552e52347bdSJani Nikula 1553380af29bSSteven Rostedt (Google) ftrace_boot_snapshot 1554380af29bSSteven Rostedt (Google) [FTRACE] On boot up, a snapshot will be taken of the 1555380af29bSSteven Rostedt (Google) ftrace ring buffer that can be read at: 1556380af29bSSteven Rostedt (Google) /sys/kernel/tracing/snapshot. 1557380af29bSSteven Rostedt (Google) This is useful if you need tracing information from kernel 1558380af29bSSteven Rostedt (Google) boot up that is likely to be overridden by user space 1559380af29bSSteven Rostedt (Google) start up functionality. 1560380af29bSSteven Rostedt (Google) 15619c1c251dSSteven Rostedt (Google) Optionally, the snapshot can also be defined for a tracing 15629c1c251dSSteven Rostedt (Google) instance that was created by the trace_instance= command 15639c1c251dSSteven Rostedt (Google) line parameter. 15649c1c251dSSteven Rostedt (Google) 15659c1c251dSSteven Rostedt (Google) trace_instance=foo,sched_switch ftrace_boot_snapshot=foo 15669c1c251dSSteven Rostedt (Google) 15679c1c251dSSteven Rostedt (Google) The above will cause the "foo" tracing instance to trigger 15689c1c251dSSteven Rostedt (Google) a snapshot at the end of boot up. 15699c1c251dSSteven Rostedt (Google) 1570e52347bdSJani Nikula ftrace_dump_on_oops[=orig_cpu] 1571e52347bdSJani Nikula [FTRACE] will dump the trace buffers on oops. 1572e52347bdSJani Nikula If no parameter is passed, ftrace will dump 1573e52347bdSJani Nikula buffers of all CPUs, but if you pass orig_cpu, it will 1574e52347bdSJani Nikula dump only the buffer of the CPU that triggered the 1575e52347bdSJani Nikula oops. 1576e52347bdSJani Nikula 1577e52347bdSJani Nikula ftrace_filter=[function-list] 1578e52347bdSJani Nikula [FTRACE] Limit the functions traced by the function 157925942e5eSRandy Dunlap tracer at boot up. function-list is a comma-separated 1580e52347bdSJani Nikula list of functions. This list can be changed at run 1581e52347bdSJani Nikula time by the set_ftrace_filter file in the debugfs 1582e52347bdSJani Nikula tracing directory. 1583e52347bdSJani Nikula 1584e52347bdSJani Nikula ftrace_notrace=[function-list] 1585e52347bdSJani Nikula [FTRACE] Do not trace the functions specified in 1586e52347bdSJani Nikula function-list. This list can be changed at run time 1587e52347bdSJani Nikula by the set_ftrace_notrace file in the debugfs 1588e52347bdSJani Nikula tracing directory. 1589e52347bdSJani Nikula 1590e52347bdSJani Nikula ftrace_graph_filter=[function-list] 1591e52347bdSJani Nikula [FTRACE] Limit the top level callers functions traced 1592e52347bdSJani Nikula by the function graph tracer at boot up. 159325942e5eSRandy Dunlap function-list is a comma-separated list of functions 1594e52347bdSJani Nikula that can be changed at run time by the 1595e52347bdSJani Nikula set_graph_function file in the debugfs tracing directory. 1596e52347bdSJani Nikula 1597e52347bdSJani Nikula ftrace_graph_notrace=[function-list] 1598e52347bdSJani Nikula [FTRACE] Do not trace from the functions specified in 159925942e5eSRandy Dunlap function-list. This list is a comma-separated list of 1600e52347bdSJani Nikula functions that can be changed at run time by the 1601e52347bdSJani Nikula set_graph_notrace file in the debugfs tracing directory. 1602e52347bdSJani Nikula 160365a50c65STodd Brandt ftrace_graph_max_depth=<uint> 160465a50c65STodd Brandt [FTRACE] Used with the function graph tracer. This is 160565a50c65STodd Brandt the max depth it will trace into a function. This value 160665a50c65STodd Brandt can be changed at run time by the max_graph_depth file 160765a50c65STodd Brandt in the tracefs tracing directory. default: 0 (no limit) 160865a50c65STodd Brandt 16098375e74fSSaravana Kannan fw_devlink= [KNL] Create device links between consumer and supplier 16108375e74fSSaravana Kannan devices by scanning the firmware to infer the 16118375e74fSSaravana Kannan consumer/supplier relationships. This feature is 16128375e74fSSaravana Kannan especially useful when drivers are loaded as modules as 16138375e74fSSaravana Kannan it ensures proper ordering of tasks like device probing 16148375e74fSSaravana Kannan (suppliers first, then consumers), supplier boot state 16158375e74fSSaravana Kannan clean up (only after all consumers have probed), 16168375e74fSSaravana Kannan suspend/resume & runtime PM (consumers first, then 16178375e74fSSaravana Kannan suppliers). 16188375e74fSSaravana Kannan Format: { off | permissive | on | rpm } 16198375e74fSSaravana Kannan off -- Don't create device links from firmware info. 16208375e74fSSaravana Kannan permissive -- Create device links from firmware info 16218375e74fSSaravana Kannan but use it only for ordering boot state clean 16228375e74fSSaravana Kannan up (sync_state() calls). 16238375e74fSSaravana Kannan on -- Create device links from firmware info and use it 16248375e74fSSaravana Kannan to enforce probe and suspend/resume ordering. 16258375e74fSSaravana Kannan rpm -- Like "on", but also use to order runtime PM. 16268375e74fSSaravana Kannan 162719d0f5f6SSaravana Kannan fw_devlink.strict=<bool> 162819d0f5f6SSaravana Kannan [KNL] Treat all inferred dependencies as mandatory 162919d0f5f6SSaravana Kannan dependencies. This only applies for fw_devlink=on|rpm. 163019d0f5f6SSaravana Kannan Format: <bool> 163119d0f5f6SSaravana Kannan 1632ffbe08a8SSaravana Kannan fw_devlink.sync_state = 1633ffbe08a8SSaravana Kannan [KNL] When all devices that could probe have finished 1634ffbe08a8SSaravana Kannan probing, this parameter controls what to do with 1635ffbe08a8SSaravana Kannan devices that haven't yet received their sync_state() 1636ffbe08a8SSaravana Kannan calls. 1637ffbe08a8SSaravana Kannan Format: { strict | timeout } 1638ffbe08a8SSaravana Kannan strict -- Default. Continue waiting on consumers to 1639ffbe08a8SSaravana Kannan probe successfully. 1640ffbe08a8SSaravana Kannan timeout -- Give up waiting on consumers and call 1641ffbe08a8SSaravana Kannan sync_state() on any devices that haven't yet 1642ffbe08a8SSaravana Kannan received their sync_state() calls after 1643ffbe08a8SSaravana Kannan deferred_probe_timeout has expired or by 1644ffbe08a8SSaravana Kannan late_initcall() if !CONFIG_MODULES. 1645ffbe08a8SSaravana Kannan 1646e52347bdSJani Nikula gamecon.map[2|3]= 1647e52347bdSJani Nikula [HW,JOY] Multisystem joystick and NES/SNES/PSX pad 1648e52347bdSJani Nikula support via parallel port (up to 5 devices per port) 1649e52347bdSJani Nikula Format: <port#>,<pad1>,<pad2>,<pad3>,<pad4>,<pad5> 16501752118dSTom Saeger See also Documentation/input/devices/joystick-parport.rst 1651e52347bdSJani Nikula 1652e52347bdSJani Nikula gamma= [HW,DRM] 1653e52347bdSJani Nikula 1654be3a5b0eSRandy Dunlap gart_fix_e820= [X86-64] disable the fix e820 for K8 GART 1655e52347bdSJani Nikula Format: off | on 1656e52347bdSJani Nikula default: on 1657e52347bdSJani Nikula 16588974eb58SDaniel Sneddon gather_data_sampling= 16598974eb58SDaniel Sneddon [X86,INTEL] Control the Gather Data Sampling (GDS) 16608974eb58SDaniel Sneddon mitigation. 16618974eb58SDaniel Sneddon 16628974eb58SDaniel Sneddon Gather Data Sampling is a hardware vulnerability which 16638974eb58SDaniel Sneddon allows unprivileged speculative access to data which was 16648974eb58SDaniel Sneddon previously stored in vector registers. 16658974eb58SDaniel Sneddon 16668974eb58SDaniel Sneddon This issue is mitigated by default in updated microcode. 16678974eb58SDaniel Sneddon The mitigation may have a performance impact but can be 1668553a5c03SDaniel Sneddon disabled. On systems without the microcode mitigation 1669553a5c03SDaniel Sneddon disabling AVX serves as a mitigation. 1670553a5c03SDaniel Sneddon 1671553a5c03SDaniel Sneddon force: Disable AVX to mitigate systems without 1672553a5c03SDaniel Sneddon microcode mitigation. No effect if the microcode 1673553a5c03SDaniel Sneddon mitigation is present. Known to cause crashes in 1674553a5c03SDaniel Sneddon userspace with buggy AVX enumeration. 16758974eb58SDaniel Sneddon 16768974eb58SDaniel Sneddon off: Disable GDS mitigation. 16778974eb58SDaniel Sneddon 1678e52347bdSJani Nikula gcov_persist= [GCOV] When non-zero (default), profiling data for 1679e52347bdSJani Nikula kernel modules is saved and remains accessible via 1680e52347bdSJani Nikula debugfs, even when the module is unloaded/reloaded. 1681e52347bdSJani Nikula When zero, profiling data is discarded and associated 1682e52347bdSJani Nikula debugfs files are removed at module unload time. 1683e52347bdSJani Nikula 168447512cfdSThomas Gleixner goldfish [X86] Enable the goldfish android emulator platform. 168547512cfdSThomas Gleixner Don't use this when you are not running on the 168647512cfdSThomas Gleixner android emulator 168747512cfdSThomas Gleixner 16883eb52226SAlexander Dahl gpio-mockup.gpio_mockup_ranges 16893eb52226SAlexander Dahl [HW] Sets the ranges of gpiochip of for this device. 16903eb52226SAlexander Dahl Format: <start1>,<end1>,<start2>,<end2>... 16916984a320SAlexander Dahl gpio-mockup.gpio_mockup_named_lines 16926984a320SAlexander Dahl [HW] Let the driver know GPIO lines should be named. 16933eb52226SAlexander Dahl 1694e52347bdSJani Nikula gpt [EFI] Forces disk with valid GPT signature but 1695e52347bdSJani Nikula invalid Protective MBR to be treated as GPT. If the 1696e52347bdSJani Nikula primary GPT is corrupted, it enables the backup/alternate 1697e52347bdSJani Nikula GPT to be used instead. 1698e52347bdSJani Nikula 1699e52347bdSJani Nikula grcan.enable0= [HW] Configuration of physical interface 0. Determines 1700e52347bdSJani Nikula the "Enable 0" bit of the configuration register. 1701e52347bdSJani Nikula Format: 0 | 1 1702e52347bdSJani Nikula Default: 0 1703e52347bdSJani Nikula grcan.enable1= [HW] Configuration of physical interface 1. Determines 1704e52347bdSJani Nikula the "Enable 0" bit of the configuration register. 1705e52347bdSJani Nikula Format: 0 | 1 1706e52347bdSJani Nikula Default: 0 1707e52347bdSJani Nikula grcan.select= [HW] Select which physical interface to use. 1708e52347bdSJani Nikula Format: 0 | 1 1709e52347bdSJani Nikula Default: 0 1710e52347bdSJani Nikula grcan.txsize= [HW] Sets the size of the tx buffer. 1711e52347bdSJani Nikula Format: <unsigned int> such that (txsize & ~0x1fffc0) == 0. 1712e52347bdSJani Nikula Default: 1024 1713e52347bdSJani Nikula grcan.rxsize= [HW] Sets the size of the rx buffer. 1714e52347bdSJani Nikula Format: <unsigned int> such that (rxsize & ~0x1fffc0) == 0. 1715e52347bdSJani Nikula Default: 1024 1716e52347bdSJani Nikula 1717d2fc83c1SRandy Dunlap hardened_usercopy= 1718d2fc83c1SRandy Dunlap [KNL] Under CONFIG_HARDENED_USERCOPY, whether 1719d2fc83c1SRandy Dunlap hardening is enabled for this boot. Hardened 1720d2fc83c1SRandy Dunlap usercopy checking is used to protect the kernel 1721d2fc83c1SRandy Dunlap from reading or writing beyond known memory 1722d2fc83c1SRandy Dunlap allocation boundaries as a proactive defense 1723d2fc83c1SRandy Dunlap against bounds-checking flaws in the kernel's 1724d2fc83c1SRandy Dunlap copy_to_user()/copy_from_user() interface. 1725d2fc83c1SRandy Dunlap on Perform hardened usercopy checks (default). 1726d2fc83c1SRandy Dunlap off Disable hardened usercopy checks. 1727d2fc83c1SRandy Dunlap 1728e52347bdSJani Nikula hardlockup_all_cpu_backtrace= 1729e52347bdSJani Nikula [KNL] Should the hard-lockup detector generate 1730e52347bdSJani Nikula backtraces on all cpus. 1731f117955aSGuilherme G. Piccoli Format: 0 | 1 1732e52347bdSJani Nikula 1733e52347bdSJani Nikula hashdist= [KNL,NUMA] Large hashes allocated during boot 1734e52347bdSJani Nikula are distributed across NUMA nodes. Defaults on 1735e52347bdSJani Nikula for 64-bit NUMA, off otherwise. 1736e52347bdSJani Nikula Format: 0 | 1 (for off | on) 1737e52347bdSJani Nikula 1738e52347bdSJani Nikula hcl= [IA-64] SGI's Hardware Graph compatibility layer 1739e52347bdSJani Nikula 1740e52347bdSJani Nikula hd= [EIDE] (E)IDE hard drive subsystem geometry 1741e52347bdSJani Nikula Format: <cyl>,<head>,<sect> 1742e52347bdSJani Nikula 1743e52347bdSJani Nikula hest_disable [ACPI] 1744e52347bdSJani Nikula Disable Hardware Error Source Table (HEST) support; 1745e52347bdSJani Nikula corresponding firmware-first mode error processing 1746e52347bdSJani Nikula logic will be disabled. 1747e52347bdSJani Nikula 1748d2fc83c1SRandy Dunlap hibernate= [HIBERNATION] 1749d2fc83c1SRandy Dunlap noresume Don't check if there's a hibernation image 1750d2fc83c1SRandy Dunlap present during boot. 1751d2fc83c1SRandy Dunlap nocompress Don't compress/decompress hibernation images. 1752d2fc83c1SRandy Dunlap no Disable hibernation and resume. 1753d2fc83c1SRandy Dunlap protect_image Turn on image protection during restoration 1754d2fc83c1SRandy Dunlap (that will set all pages holding image data 1755d2fc83c1SRandy Dunlap during restoration read-only). 1756d2fc83c1SRandy Dunlap 1757e52347bdSJani Nikula highmem=nn[KMG] [KNL,BOOT] forces the highmem zone to have an exact 1758e52347bdSJani Nikula size of <nn>. This works even on boxes that have no 1759e52347bdSJani Nikula highmem otherwise. This also works to reduce highmem 1760e52347bdSJani Nikula size on bigger boxes. 1761e52347bdSJani Nikula 1762e52347bdSJani Nikula highres= [KNL] Enable/disable high resolution timer mode. 1763e52347bdSJani Nikula Valid parameters: "on", "off" 1764e52347bdSJani Nikula Default: "on" 1765e52347bdSJani Nikula 1766e52347bdSJani Nikula hlt [BUGS=ARM,SH] 1767e52347bdSJani Nikula 17685a704629SDan Moulding hostname= [KNL] Set the hostname (aka UTS nodename). 17695a704629SDan Moulding Format: <string> 17705a704629SDan Moulding This allows setting the system's hostname during early 17715a704629SDan Moulding startup. This sets the name returned by gethostname. 17725a704629SDan Moulding Using this parameter to set the hostname makes it 17735a704629SDan Moulding possible to ensure the hostname is correctly set before 17745a704629SDan Moulding any userspace processes run, avoiding the possibility 17755a704629SDan Moulding that a process may call gethostname before the hostname 17765a704629SDan Moulding has been explicitly set, resulting in the calling 17775a704629SDan Moulding process getting an incorrect result. The string must 17785a704629SDan Moulding not exceed the maximum allowed hostname length (usually 17795a704629SDan Moulding 64 characters) and will be truncated otherwise. 17805a704629SDan Moulding 1781e52347bdSJani Nikula hpet= [X86-32,HPET] option to control HPET usage 1782e52347bdSJani Nikula Format: { enable (default) | disable | force | 1783e52347bdSJani Nikula verbose } 1784e52347bdSJani Nikula disable: disable HPET and use PIT instead 1785e52347bdSJani Nikula force: allow force enabled of undocumented chips (ICH4, 1786e52347bdSJani Nikula VIA, nVidia) 1787e52347bdSJani Nikula verbose: show contents of HPET registers during setup 1788e52347bdSJani Nikula 1789e52347bdSJani Nikula hpet_mmap= [X86, HPET_MMAP] Allow userspace to mmap HPET 1790e52347bdSJani Nikula registers. Default set by CONFIG_HPET_MMAP_DEFAULT. 1791e52347bdSJani Nikula 1792282f4214SMike Kravetz hugepages= [HW] Number of HugeTLB pages to allocate at boot. 1793282f4214SMike Kravetz If this follows hugepagesz (below), it specifies 1794282f4214SMike Kravetz the number of pages of hugepagesz to be allocated. 1795282f4214SMike Kravetz If this is the first HugeTLB parameter on the command 1796282f4214SMike Kravetz line, it specifies the number of pages to allocate for 1797b5389086SZhenguo Yao the default huge page size. If using node format, the 1798b5389086SZhenguo Yao number of pages to allocate per-node can be specified. 1799b5389086SZhenguo Yao See also Documentation/admin-guide/mm/hugetlbpage.rst. 1800b5389086SZhenguo Yao Format: <integer> or (node format) 1801b5389086SZhenguo Yao <node>:<integer>[,<node>:<integer>] 1802282f4214SMike Kravetz 1803282f4214SMike Kravetz hugepagesz= 1804282f4214SMike Kravetz [HW] The size of the HugeTLB pages. This is used in 1805282f4214SMike Kravetz conjunction with hugepages (above) to allocate huge 1806282f4214SMike Kravetz pages of a specific size at boot. The pair 1807282f4214SMike Kravetz hugepagesz=X hugepages=Y can be specified once for 1808282f4214SMike Kravetz each supported huge page size. Huge page sizes are 1809282f4214SMike Kravetz architecture dependent. See also 1810282f4214SMike Kravetz Documentation/admin-guide/mm/hugetlbpage.rst. 1811282f4214SMike Kravetz Format: size[KMG] 1812e52347bdSJani Nikula 1813389cfd96SRandy Dunlap hugetlb_cma= [HW,CMA] The size of a CMA area used for allocation 1814389cfd96SRandy Dunlap of gigantic hugepages. Or using node format, the size 1815389cfd96SRandy Dunlap of a CMA area per node can be specified. 1816389cfd96SRandy Dunlap Format: nn[KMGTPE] or (node format) 1817389cfd96SRandy Dunlap <node>:nn[KMGTPE][,<node>:nn[KMGTPE]] 1818389cfd96SRandy Dunlap 1819389cfd96SRandy Dunlap Reserve a CMA area of given size and allocate gigantic 1820389cfd96SRandy Dunlap hugepages using the CMA allocator. If enabled, the 1821389cfd96SRandy Dunlap boot-time allocation of gigantic hugepages is skipped. 1822389cfd96SRandy Dunlap 1823e9fdff87SMuchun Song hugetlb_free_vmemmap= 1824dbeb56feSRandy Dunlap [KNL] Requires CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP 1825e9fdff87SMuchun Song enabled. 1826dff03381SMuchun Song Control if HugeTLB Vmemmap Optimization (HVO) is enabled. 1827e9fdff87SMuchun Song Allows heavy hugetlb users to free up some more 1828e7d32485SMuchun Song memory (7 * PAGE_SIZE for each 2MB hugetlb page). 1829dff03381SMuchun Song Format: { on | off (default) } 1830e9fdff87SMuchun Song 1831dff03381SMuchun Song on: enable HVO 1832dff03381SMuchun Song off: disable HVO 1833e9fdff87SMuchun Song 183447010c04SMuchun Song Built with CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON=y, 1835e6d41f12SMuchun Song the default is on. 1836e6d41f12SMuchun Song 183766361095SMuchun Song Note that the vmemmap pages may be allocated from the added 183866361095SMuchun Song memory block itself when memory_hotplug.memmap_on_memory is 183966361095SMuchun Song enabled, those vmemmap pages cannot be optimized even if this 184066361095SMuchun Song feature is enabled. Other vmemmap pages not allocated from 184166361095SMuchun Song the added memory block itself do not be affected. 18424bab4964SMuchun Song 1843a49d9c0aSOmar Sandoval hung_task_panic= 1844a49d9c0aSOmar Sandoval [KNL] Should the hung task detector generate panics. 1845f117955aSGuilherme G. Piccoli Format: 0 | 1 1846a49d9c0aSOmar Sandoval 1847b467f3efSVlastimil Babka A value of 1 instructs the kernel to panic when a 1848a49d9c0aSOmar Sandoval hung task is detected. The default value is controlled 1849a49d9c0aSOmar Sandoval by the CONFIG_BOOTPARAM_HUNG_TASK_PANIC build-time 1850a49d9c0aSOmar Sandoval option. The value selected by this boot parameter can 1851a49d9c0aSOmar Sandoval be changed later by the kernel.hung_task_panic sysctl. 1852a49d9c0aSOmar Sandoval 1853e52347bdSJani Nikula hvc_iucv= [S390] Number of z/VM IUCV hypervisor console (HVC) 1854e52347bdSJani Nikula terminal devices. Valid values: 0..8 1855e52347bdSJani Nikula hvc_iucv_allow= [S390] Comma-separated list of z/VM user IDs. 1856e52347bdSJani Nikula If specified, z/VM IUCV HVC accepts connections 1857e52347bdSJani Nikula from listed z/VM user IDs only. 18583a025de6SYi Sun 18593a025de6SYi Sun hv_nopvspin [X86,HYPER_V] Disables the paravirt spinlock optimizations 18603a025de6SYi Sun which allow the hypervisor to 'idle' the 18613a025de6SYi Sun guest on lock contention. 18623a025de6SYi Sun 1863e52347bdSJani Nikula i2c_bus= [HW] Override the default board specific I2C bus speed 1864e52347bdSJani Nikula or register an additional I2C bus that is not 1865e52347bdSJani Nikula registered from board initialization code. 1866e52347bdSJani Nikula Format: 1867e52347bdSJani Nikula <bus_id>,<clkrate> 1868e52347bdSJani Nikula 1869e52347bdSJani Nikula i8042.debug [HW] Toggle i8042 debug mode 1870e52347bdSJani Nikula i8042.unmask_kbd_data 1871e52347bdSJani Nikula [HW] Enable printing of interrupt data from the KBD port 1872e52347bdSJani Nikula (disabled by default, and as a pre-condition 1873e52347bdSJani Nikula requires that i8042.debug=1 be enabled) 1874e52347bdSJani Nikula i8042.direct [HW] Put keyboard port into non-translated mode 1875e52347bdSJani Nikula i8042.dumbkbd [HW] Pretend that controller can only read data from 1876e52347bdSJani Nikula keyboard and cannot control its state 1877e52347bdSJani Nikula (Don't attempt to blink the leds) 1878e52347bdSJani Nikula i8042.noaux [HW] Don't check for auxiliary (== mouse) port 1879e52347bdSJani Nikula i8042.nokbd [HW] Don't check/create keyboard port 1880e52347bdSJani Nikula i8042.noloop [HW] Disable the AUX Loopback command while probing 1881e52347bdSJani Nikula for the AUX port 1882e52347bdSJani Nikula i8042.nomux [HW] Don't check presence of an active multiplexing 1883e52347bdSJani Nikula controller 1884e52347bdSJani Nikula i8042.nopnp [HW] Don't use ACPIPnP / PnPBIOS to discover KBD/AUX 1885e52347bdSJani Nikula controllers 1886e52347bdSJani Nikula i8042.notimeout [HW] Ignore timeout condition signalled by controller 1887e52347bdSJani Nikula i8042.reset [HW] Reset the controller during init, cleanup and 1888e52347bdSJani Nikula suspend-to-ram transitions, only during s2r 1889e52347bdSJani Nikula transitions, or never reset 1890e52347bdSJani Nikula Format: { 1 | Y | y | 0 | N | n } 1891e52347bdSJani Nikula 1, Y, y: always reset controller 1892e52347bdSJani Nikula 0, N, n: don't ever reset controller 1893e52347bdSJani Nikula Default: only on s2r transitions on x86; most other 1894e52347bdSJani Nikula architectures force reset to be always executed 1895e52347bdSJani Nikula i8042.unlock [HW] Unlock (ignore) the keylock 1896e52347bdSJani Nikula i8042.kbdreset [HW] Reset device connected to KBD port 18979222ba68STakashi Iwai i8042.probe_defer 18989222ba68STakashi Iwai [HW] Allow deferred probing upon i8042 probe errors 1899e52347bdSJani Nikula 1900e52347bdSJani Nikula i810= [HW,DRM] 1901e52347bdSJani Nikula 1902e52347bdSJani Nikula i915.invert_brightness= 1903e52347bdSJani Nikula [DRM] Invert the sense of the variable that is used to 1904e52347bdSJani Nikula set the brightness of the panel backlight. Normally a 1905e52347bdSJani Nikula brightness value of 0 indicates backlight switched off, 1906e52347bdSJani Nikula and the maximum of the brightness value sets the backlight 1907e52347bdSJani Nikula to maximum brightness. If this parameter is set to 0 1908e52347bdSJani Nikula (default) and the machine requires it, or this parameter 1909e52347bdSJani Nikula is set to 1, a brightness value of 0 sets the backlight 1910e52347bdSJani Nikula to maximum brightness, and the maximum of the brightness 1911e52347bdSJani Nikula value switches the backlight off. 1912e52347bdSJani Nikula -1 -- never invert brightness 1913e52347bdSJani Nikula 0 -- machine default 1914e52347bdSJani Nikula 1 -- force brightness inversion 1915e52347bdSJani Nikula 1916e52347bdSJani Nikula icn= [HW,ISDN] 1917e52347bdSJani Nikula Format: <io>[,<membase>[,<icn_id>[,<icn_id2>]]] 1918e52347bdSJani Nikula 1919e52347bdSJani Nikula 1920e52347bdSJani Nikula idle= [X86] 1921e52347bdSJani Nikula Format: idle=poll, idle=halt, idle=nomwait 1922e52347bdSJani Nikula Poll forces a polling idle loop that can slightly 1923e52347bdSJani Nikula improve the performance of waking up a idle CPU, but 1924e52347bdSJani Nikula will use a lot of power and make the system run hot. 1925e52347bdSJani Nikula Not recommended. 1926e52347bdSJani Nikula idle=halt: Halt is forced to be used for CPU idle. 1927e52347bdSJani Nikula In such case C2/C3 won't be used again. 1928e52347bdSJani Nikula idle=nomwait: Disable mwait for CPU C-states 1929e52347bdSJani Nikula 193003d939c7SDave Jiang idxd.sva= [HW] 193103d939c7SDave Jiang Format: <bool> 193203d939c7SDave Jiang Allow force disabling of Shared Virtual Memory (SVA) 193303d939c7SDave Jiang support for the idxd driver. By default it is set to 193403d939c7SDave Jiang true (1). 193503d939c7SDave Jiang 1936ade8a86bSDave Jiang idxd.tc_override= [HW] 1937ade8a86bSDave Jiang Format: <bool> 1938ade8a86bSDave Jiang Allow override of default traffic class configuration 1939ade8a86bSDave Jiang for the device. By default it is set to false (0). 1940ade8a86bSDave Jiang 1941e52347bdSJani Nikula ieee754= [MIPS] Select IEEE Std 754 conformance mode 1942e52347bdSJani Nikula Format: { strict | legacy | 2008 | relaxed } 1943e52347bdSJani Nikula Default: strict 1944e52347bdSJani Nikula 1945e52347bdSJani Nikula Choose which programs will be accepted for execution 1946e52347bdSJani Nikula based on the IEEE 754 NaN encoding(s) supported by 1947e52347bdSJani Nikula the FPU and the NaN encoding requested with the value 1948e52347bdSJani Nikula of an ELF file header flag individually set by each 1949e52347bdSJani Nikula binary. Hardware implementations are permitted to 1950e52347bdSJani Nikula support either or both of the legacy and the 2008 NaN 1951e52347bdSJani Nikula encoding mode. 1952e52347bdSJani Nikula 1953e52347bdSJani Nikula Available settings are as follows: 1954e52347bdSJani Nikula strict accept binaries that request a NaN encoding 1955e52347bdSJani Nikula supported by the FPU 1956e52347bdSJani Nikula legacy only accept legacy-NaN binaries, if supported 1957e52347bdSJani Nikula by the FPU 1958e52347bdSJani Nikula 2008 only accept 2008-NaN binaries, if supported 1959e52347bdSJani Nikula by the FPU 1960e52347bdSJani Nikula relaxed accept any binaries regardless of whether 1961e52347bdSJani Nikula supported by the FPU 1962e52347bdSJani Nikula 1963e52347bdSJani Nikula The FPU emulator is always able to support both NaN 1964e52347bdSJani Nikula encodings, so if no FPU hardware is present or it has 1965e52347bdSJani Nikula been disabled with 'nofpu', then the settings of 1966e52347bdSJani Nikula 'legacy' and '2008' strap the emulator accordingly, 1967e52347bdSJani Nikula 'relaxed' straps the emulator for both legacy-NaN and 1968e52347bdSJani Nikula 2008-NaN, whereas 'strict' enables legacy-NaN only on 1969e52347bdSJani Nikula legacy processors and both NaN encodings on MIPS32 or 1970e52347bdSJani Nikula MIPS64 CPUs. 1971e52347bdSJani Nikula 1972e52347bdSJani Nikula The setting for ABS.fmt/NEG.fmt instruction execution 1973e52347bdSJani Nikula mode generally follows that for the NaN encoding, 1974e52347bdSJani Nikula except where unsupported by hardware. 1975e52347bdSJani Nikula 1976e52347bdSJani Nikula ignore_loglevel [KNL] 1977e52347bdSJani Nikula Ignore loglevel setting - this will print /all/ 1978e52347bdSJani Nikula kernel messages to the console. Useful for debugging. 1979e52347bdSJani Nikula We also add it as printk module parameter, so users 1980e52347bdSJani Nikula could change it dynamically, usually by 1981e52347bdSJani Nikula /sys/module/printk/parameters/ignore_loglevel. 1982e52347bdSJani Nikula 1983e52347bdSJani Nikula ignore_rlimit_data 1984e52347bdSJani Nikula Ignore RLIMIT_DATA setting for data mappings, 1985e52347bdSJani Nikula print warning at first misuse. Can be changed via 1986e52347bdSJani Nikula /sys/module/kernel/parameters/ignore_rlimit_data. 1987e52347bdSJani Nikula 1988e52347bdSJani Nikula ihash_entries= [KNL] 1989e52347bdSJani Nikula Set number of hash buckets for inode cache. 1990e52347bdSJani Nikula 1991e52347bdSJani Nikula ima_appraise= [IMA] appraise integrity measurements 1992e52347bdSJani Nikula Format: { "off" | "enforce" | "fix" | "log" } 1993e52347bdSJani Nikula default: "enforce" 1994e52347bdSJani Nikula 199541475a3eSPetr Vorel ima_appraise_tcb [IMA] Deprecated. Use ima_policy= instead. 1996e52347bdSJani Nikula The builtin appraise policy appraises all files 1997e52347bdSJani Nikula owned by uid=0. 1998e52347bdSJani Nikula 1999d68a6fe9SMimi Zohar ima_canonical_fmt [IMA] 2000d68a6fe9SMimi Zohar Use the canonical format for the binary runtime 2001d68a6fe9SMimi Zohar measurements, instead of host native format. 2002d68a6fe9SMimi Zohar 2003e52347bdSJani Nikula ima_hash= [IMA] 2004e52347bdSJani Nikula Format: { md5 | sha1 | rmd160 | sha256 | sha384 2005e52347bdSJani Nikula | sha512 | ... } 2006e52347bdSJani Nikula default: "sha1" 2007e52347bdSJani Nikula 2008e52347bdSJani Nikula The list of supported hash algorithms is defined 2009e52347bdSJani Nikula in crypto/hash_info.h. 2010e52347bdSJani Nikula 2011e52347bdSJani Nikula ima_policy= [IMA] 201233ce9549SMimi Zohar The builtin policies to load during IMA setup. 20139e67028eSMimi Zohar Format: "tcb | appraise_tcb | secure_boot | 201403cee168SLakshmi Ramasubramanian fail_securely | critical_data" 201533ce9549SMimi Zohar 201633ce9549SMimi Zohar The "tcb" policy measures all programs exec'd, files 201733ce9549SMimi Zohar mmap'd for exec, and all files opened with the read 201833ce9549SMimi Zohar mode bit set by either the effective uid (euid=0) or 201933ce9549SMimi Zohar uid=0. 202033ce9549SMimi Zohar 202133ce9549SMimi Zohar The "appraise_tcb" policy appraises the integrity of 202241475a3eSPetr Vorel all files owned by root. 2023e52347bdSJani Nikula 2024503ceaefSMimi Zohar The "secure_boot" policy appraises the integrity 2025503ceaefSMimi Zohar of files (eg. kexec kernel image, kernel modules, 2026503ceaefSMimi Zohar firmware, policy, etc) based on file signatures. 2027e52347bdSJani Nikula 20289e67028eSMimi Zohar The "fail_securely" policy forces file signature 20299e67028eSMimi Zohar verification failure also on privileged mounted 20309e67028eSMimi Zohar filesystems with the SB_I_UNVERIFIABLE_SIGNATURE 20319e67028eSMimi Zohar flag. 20329e67028eSMimi Zohar 203303cee168SLakshmi Ramasubramanian The "critical_data" policy measures kernel integrity 203403cee168SLakshmi Ramasubramanian critical data. 203503cee168SLakshmi Ramasubramanian 2036e52347bdSJani Nikula ima_tcb [IMA] Deprecated. Use ima_policy= instead. 2037e52347bdSJani Nikula Load a policy which meets the needs of the Trusted 2038e52347bdSJani Nikula Computing Base. This means IMA will measure all 2039e52347bdSJani Nikula programs exec'd, files mmap'd for exec, and all files 2040e52347bdSJani Nikula opened for read by uid=0. 2041e52347bdSJani Nikula 2042e52347bdSJani Nikula ima_template= [IMA] 2043e52347bdSJani Nikula Select one of defined IMA measurements template formats. 2044989dc725SMimi Zohar Formats: { "ima" | "ima-ng" | "ima-ngv2" | "ima-sig" | 2045989dc725SMimi Zohar "ima-sigv2" } 2046e52347bdSJani Nikula Default: "ima-ng" 2047e52347bdSJani Nikula 2048e52347bdSJani Nikula ima_template_fmt= 2049e52347bdSJani Nikula [IMA] Define a custom template format. 2050e52347bdSJani Nikula Format: { "field1|...|fieldN" } 2051e52347bdSJani Nikula 2052e52347bdSJani Nikula ima.ahash_minsize= [IMA] Minimum file size for asynchronous hash usage 2053e52347bdSJani Nikula Format: <min_file_size> 2054e52347bdSJani Nikula Set the minimal file size for using asynchronous hash. 2055e52347bdSJani Nikula If left unspecified, ahash usage is disabled. 2056e52347bdSJani Nikula 2057e52347bdSJani Nikula ahash performance varies for different data sizes on 2058e52347bdSJani Nikula different crypto accelerators. This option can be used 2059e52347bdSJani Nikula to achieve the best performance for a particular HW. 2060e52347bdSJani Nikula 2061e52347bdSJani Nikula ima.ahash_bufsize= [IMA] Asynchronous hash buffer size 2062e52347bdSJani Nikula Format: <bufsize> 2063e52347bdSJani Nikula Set hashing buffer size. Default: 4k. 2064e52347bdSJani Nikula 2065e52347bdSJani Nikula ahash performance varies for different chunk sizes on 2066e52347bdSJani Nikula different crypto accelerators. This option can be used 2067e52347bdSJani Nikula to achieve best performance for particular HW. 2068e52347bdSJani Nikula 2069e52347bdSJani Nikula init= [KNL] 2070e52347bdSJani Nikula Format: <full_path> 2071e52347bdSJani Nikula Run specified binary instead of /sbin/init as init 2072e52347bdSJani Nikula process. 2073e52347bdSJani Nikula 2074e52347bdSJani Nikula initcall_debug [KNL] Trace initcalls as they are executed. Useful 2075e52347bdSJani Nikula for working out where the kernel is dying during 2076e52347bdSJani Nikula startup. 2077e52347bdSJani Nikula 2078e52347bdSJani Nikula initcall_blacklist= [KNL] Do not execute a comma-separated list of 2079e52347bdSJani Nikula initcall functions. Useful for debugging built-in 2080e52347bdSJani Nikula modules and initcalls. 2081e52347bdSJani Nikula 2082e7cb072eSRasmus Villemoes initramfs_async= [KNL] 2083e7cb072eSRasmus Villemoes Format: <bool> 2084e7cb072eSRasmus Villemoes Default: 1 2085e7cb072eSRasmus Villemoes This parameter controls whether the initramfs 2086e7cb072eSRasmus Villemoes image is unpacked asynchronously, concurrently 2087e7cb072eSRasmus Villemoes with devices being probed and 2088e7cb072eSRasmus Villemoes initialized. This should normally just work, 2089e7cb072eSRasmus Villemoes but as a debugging aid, one can get the 2090e7cb072eSRasmus Villemoes historical behaviour of the initramfs 2091e7cb072eSRasmus Villemoes unpacking being completed before device_ and 2092e7cb072eSRasmus Villemoes late_ initcalls. 2093e7cb072eSRasmus Villemoes 2094e52347bdSJani Nikula initrd= [BOOT] Specify the location of the initial ramdisk 2095e52347bdSJani Nikula 2096694cfd87SRonald G. Minnich initrdmem= [KNL] Specify a physical address and size from which to 2097694cfd87SRonald G. Minnich load the initrd. If an initrd is compiled in or 2098694cfd87SRonald G. Minnich specified in the bootparams, it takes priority over this 2099694cfd87SRonald G. Minnich setting. 2100694cfd87SRonald G. Minnich Format: ss[KMG],nn[KMG] 2101694cfd87SRonald G. Minnich Default is 0, 0 2102694cfd87SRonald G. Minnich 21036471384aSAlexander Potapenko init_on_alloc= [MM] Fill newly allocated pages and heap objects with 21046471384aSAlexander Potapenko zeroes. 21056471384aSAlexander Potapenko Format: 0 | 1 21066471384aSAlexander Potapenko Default set by CONFIG_INIT_ON_ALLOC_DEFAULT_ON. 21076471384aSAlexander Potapenko 21086471384aSAlexander Potapenko init_on_free= [MM] Fill freed pages and heap objects with zeroes. 21096471384aSAlexander Potapenko Format: 0 | 1 21106471384aSAlexander Potapenko Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. 21116471384aSAlexander Potapenko 2112be3a5b0eSRandy Dunlap init_pkru= [X86] Specify the default memory protection keys rights 2113e52347bdSJani Nikula register contents for all processes. 0x55555554 by 2114e52347bdSJani Nikula default (disallow access to all but pkey 0). Can 2115e52347bdSJani Nikula override in debugfs after boot. 2116e52347bdSJani Nikula 2117e52347bdSJani Nikula inport.irq= [HW] Inport (ATI XL and Microsoft) busmouse driver 2118e52347bdSJani Nikula Format: <irq> 2119e52347bdSJani Nikula 2120be3a5b0eSRandy Dunlap int_pln_enable [X86] Enable power limit notification interrupt 2121e52347bdSJani Nikula 2122e52347bdSJani Nikula integrity_audit=[IMA] 2123e52347bdSJani Nikula Format: { "0" | "1" } 2124e52347bdSJani Nikula 0 -- basic integrity auditing messages. (Default) 2125e52347bdSJani Nikula 1 -- additional integrity auditing messages. 2126e52347bdSJani Nikula 2127e52347bdSJani Nikula intel_iommu= [DMAR] Intel IOMMU driver (DMAR) option 2128e52347bdSJani Nikula on 2129e52347bdSJani Nikula Enable intel iommu driver. 2130e52347bdSJani Nikula off 2131e52347bdSJani Nikula Disable intel iommu driver. 2132e52347bdSJani Nikula igfx_off [Default Off] 2133e52347bdSJani Nikula By default, gfx is mapped as normal device. If a gfx 2134e52347bdSJani Nikula device has a dedicated DMAR unit, the DMAR unit is 2135e52347bdSJani Nikula bypassed by not enabling DMAR with this option. In 2136e52347bdSJani Nikula this case, gfx device will use physical address for 2137e52347bdSJani Nikula DMA. 2138e52347bdSJani Nikula strict [Default Off] 21391d479f16SJohn Garry Deprecated, equivalent to iommu.strict=1. 2140e52347bdSJani Nikula sp_off [Default Off] 2141e52347bdSJani Nikula By default, super page will be supported if Intel IOMMU 2142e52347bdSJani Nikula has the capability. With this option, super page will 2143e52347bdSJani Nikula not be supported. 2144792fb43cSLu Baolu sm_on 2145792fb43cSLu Baolu Enable the Intel IOMMU scalable mode if the hardware 2146792fb43cSLu Baolu advertises that it has support for the scalable mode 2147792fb43cSLu Baolu translation. 2148792fb43cSLu Baolu sm_off 2149792fb43cSLu Baolu Disallow use of the Intel IOMMU scalable mode. 2150bfd20f1cSShaohua Li tboot_noforce [Default Off] 2151bfd20f1cSShaohua Li Do not force the Intel IOMMU enabled under tboot. 2152bfd20f1cSShaohua Li By default, tboot will force Intel IOMMU on, which 2153bfd20f1cSShaohua Li could harm performance of some high-throughput 2154bfd20f1cSShaohua Li devices like 40GBit network cards, even if identity 2155bfd20f1cSShaohua Li mapping is enabled. 2156bfd20f1cSShaohua Li Note that using this option lowers the security 2157bfd20f1cSShaohua Li provided by tboot because it makes the system 2158bfd20f1cSShaohua Li vulnerable to DMA attacks. 2159e52347bdSJani Nikula 2160e52347bdSJani Nikula intel_idle.max_cstate= [KNL,HW,ACPI,X86] 2161e52347bdSJani Nikula 0 disables intel_idle and fall back on acpi_idle. 2162e52347bdSJani Nikula 1 to 9 specify maximum depth of C-state. 2163e52347bdSJani Nikula 2164e52347bdSJani Nikula intel_pstate= [X86] 2165e52347bdSJani Nikula disable 2166e52347bdSJani Nikula Do not enable intel_pstate as the default 2167e52347bdSJani Nikula scaling driver for the supported processors 2168f02c20d9SNatesh Sharma active 2169f02c20d9SNatesh Sharma Use intel_pstate driver to bypass the scaling 2170f02c20d9SNatesh Sharma governors layer of cpufreq and provides it own 2171f02c20d9SNatesh Sharma algorithms for p-state selection. There are two 2172f02c20d9SNatesh Sharma P-state selection algorithms provided by 2173f02c20d9SNatesh Sharma intel_pstate in the active mode: powersave and 2174f02c20d9SNatesh Sharma performance. The way they both operate depends 2175f02c20d9SNatesh Sharma on whether or not the hardware managed P-states 2176f02c20d9SNatesh Sharma (HWP) feature has been enabled in the processor 2177f02c20d9SNatesh Sharma and possibly on the processor model. 21787b9dc3f7SLinus Torvalds passive 21797b9dc3f7SLinus Torvalds Use intel_pstate as a scaling driver, but configure it 21807b9dc3f7SLinus Torvalds to work with generic cpufreq governors (instead of 21817b9dc3f7SLinus Torvalds enabling its internal governor). This mode cannot be 21827b9dc3f7SLinus Torvalds used along with the hardware-managed P-states (HWP) 21837b9dc3f7SLinus Torvalds feature. 2184e52347bdSJani Nikula force 2185e52347bdSJani Nikula Enable intel_pstate on systems that prohibit it by default 2186e52347bdSJani Nikula in favor of acpi-cpufreq. Forcing the intel_pstate driver 2187e52347bdSJani Nikula instead of acpi-cpufreq may disable platform features, such 2188e52347bdSJani Nikula as thermal controls and power capping, that rely on ACPI 2189e52347bdSJani Nikula P-States information being indicated to OSPM and therefore 2190e52347bdSJani Nikula should be used with caution. This option does not work with 2191e52347bdSJani Nikula processors that aren't supported by the intel_pstate driver 2192e52347bdSJani Nikula or on platforms that use pcc-cpufreq instead of acpi-cpufreq. 2193e52347bdSJani Nikula no_hwp 2194e52347bdSJani Nikula Do not enable hardware P state control (HWP) 2195e52347bdSJani Nikula if available. 2196e52347bdSJani Nikula hwp_only 2197e52347bdSJani Nikula Only load intel_pstate on systems which support 2198e52347bdSJani Nikula hardware P state control (HWP) if available. 2199e52347bdSJani Nikula support_acpi_ppc 2200e52347bdSJani Nikula Enforce ACPI _PPC performance limits. If the Fixed ACPI 2201e52347bdSJani Nikula Description Table, specifies preferred power management 2202e52347bdSJani Nikula profile as "Enterprise Server" or "Performance Server", 2203e52347bdSJani Nikula then this feature is turned on by default. 22047b9dc3f7SLinus Torvalds per_cpu_perf_limits 22057b9dc3f7SLinus Torvalds Allow per-logical-CPU P-State performance control limits using 22067b9dc3f7SLinus Torvalds cpufreq sysfs interface 2207e52347bdSJani Nikula 2208e52347bdSJani Nikula intremap= [X86-64, Intel-IOMMU] 2209e52347bdSJani Nikula on enable Interrupt Remapping (default) 2210e52347bdSJani Nikula off disable Interrupt Remapping 2211e52347bdSJani Nikula nosid disable Source ID checking 2212e52347bdSJani Nikula no_x2apic_optout 2213e52347bdSJani Nikula BIOS x2APIC opt-out request will be ignored 2214e52347bdSJani Nikula nopost disable Interrupt Posting 2215e52347bdSJani Nikula 2216e52347bdSJani Nikula iomem= Disable strict checking of access to MMIO memory 2217e52347bdSJani Nikula strict regions from userspace. 2218e52347bdSJani Nikula relaxed 2219e52347bdSJani Nikula 2220be3a5b0eSRandy Dunlap iommu= [X86] 2221e52347bdSJani Nikula off 2222e52347bdSJani Nikula force 2223e52347bdSJani Nikula noforce 2224e52347bdSJani Nikula biomerge 2225e52347bdSJani Nikula panic 2226e52347bdSJani Nikula nopanic 2227e52347bdSJani Nikula merge 2228e52347bdSJani Nikula nomerge 2229e52347bdSJani Nikula soft 2230be3a5b0eSRandy Dunlap pt [X86] 2231be3a5b0eSRandy Dunlap nopt [X86] 2232e52347bdSJani Nikula nobypass [PPC/POWERNV] 2233e52347bdSJani Nikula Disable IOMMU bypass, using IOMMU for PCI devices. 2234e52347bdSJani Nikula 22353542dcb1SRobin Murphy iommu.forcedac= [ARM64, X86] Control IOVA allocation for PCI devices. 22363542dcb1SRobin Murphy Format: { "0" | "1" } 22373542dcb1SRobin Murphy 0 - Try to allocate a 32-bit DMA address first, before 22383542dcb1SRobin Murphy falling back to the full range if needed. 22393542dcb1SRobin Murphy 1 - Allocate directly from the full usable range, 22403542dcb1SRobin Murphy forcing Dual Address Cycle for PCI cards supporting 22413542dcb1SRobin Murphy greater than 32-bit addressing. 22423542dcb1SRobin Murphy 2243531353e6SRobin Murphy iommu.strict= [ARM64, X86] Configure TLB invalidation behaviour 224468a6efe8SZhen Lei Format: { "0" | "1" } 224568a6efe8SZhen Lei 0 - Lazy mode. 224668a6efe8SZhen Lei Request that DMA unmap operations use deferred 224768a6efe8SZhen Lei invalidation of hardware TLBs, for increased 224868a6efe8SZhen Lei throughput at the cost of reduced device isolation. 224968a6efe8SZhen Lei Will fall back to strict mode if not supported by 225068a6efe8SZhen Lei the relevant IOMMU driver. 2251712d8f20SZhen Lei 1 - Strict mode. 225268a6efe8SZhen Lei DMA unmap operations invalidate IOMMU hardware TLBs 225368a6efe8SZhen Lei synchronously. 2254e96763ecSRobin Murphy unset - Use value of CONFIG_IOMMU_DEFAULT_DMA_{LAZY,STRICT}. 2255e96763ecSRobin Murphy Note: on x86, strict mode specified via one of the 2256e96763ecSRobin Murphy legacy driver-specific options takes precedence. 225768a6efe8SZhen Lei 2258fccb4e3bSWill Deacon iommu.passthrough= 2259c8fb436bSJoerg Roedel [ARM64, X86] Configure DMA to bypass the IOMMU by default. 2260fccb4e3bSWill Deacon Format: { "0" | "1" } 2261fccb4e3bSWill Deacon 0 - Use IOMMU translation for DMA. 2262fccb4e3bSWill Deacon 1 - Bypass the IOMMU for DMA. 22639d723b4cSRobin Murphy unset - Use value of CONFIG_IOMMU_DEFAULT_PASSTHROUGH. 2264e52347bdSJani Nikula 22657c42376eSRandy Dunlap io7= [HW] IO7 for Marvel-based Alpha systems 2266e52347bdSJani Nikula See comment before marvel_specify_io7 in 2267e52347bdSJani Nikula arch/alpha/kernel/core_marvel.c. 2268e52347bdSJani Nikula 2269e52347bdSJani Nikula io_delay= [X86] I/O delay method 2270e52347bdSJani Nikula 0x80 2271e52347bdSJani Nikula Standard port 0x80 based delay 2272e52347bdSJani Nikula 0xed 2273e52347bdSJani Nikula Alternate port 0xed based delay (needed on some systems) 2274e52347bdSJani Nikula udelay 2275e52347bdSJani Nikula Simple two microseconds delay 2276e52347bdSJani Nikula none 2277e52347bdSJani Nikula No delay 2278e52347bdSJani Nikula 2279e52347bdSJani Nikula ip= [IP_PNP] 22803eb30c51SNiklas Söderlund See Documentation/admin-guide/nfs/nfsroot.rst. 2281e52347bdSJani Nikula 22825ac893b8SWaiman Long ipcmni_extend [KNL] Extend the maximum number of unique System V 22835ac893b8SWaiman Long IPC identifiers from 32,768 to 16,777,216. 22845ac893b8SWaiman Long 2285e52347bdSJani Nikula irqaffinity= [SMP] Set the default irq affinity mask 2286e52347bdSJani Nikula The argument is a cpu list, as described above. 2287e52347bdSJani Nikula 22880962289bSMarc Zyngier irqchip.gicv2_force_probe= 22890962289bSMarc Zyngier [ARM, ARM64] 22900962289bSMarc Zyngier Format: <bool> 22910962289bSMarc Zyngier Force the kernel to look for the second 4kB page 22920962289bSMarc Zyngier of a GICv2 controller even if the memory range 22930962289bSMarc Zyngier exposed by the device tree is too small. 22940962289bSMarc Zyngier 2295f736d65dSMarc Zyngier irqchip.gicv3_nolpi= 2296f736d65dSMarc Zyngier [ARM, ARM64] 2297f736d65dSMarc Zyngier Force the kernel to ignore the availability of 2298f736d65dSMarc Zyngier LPIs (and by consequence ITSs). Intended for system 2299f736d65dSMarc Zyngier that use the kernel as a bootloader, and thus want 2300f736d65dSMarc Zyngier to let secondary kernels in charge of setting up 2301f736d65dSMarc Zyngier LPIs. 2302f736d65dSMarc Zyngier 2303bc3c03ccSJulien Thierry irqchip.gicv3_pseudo_nmi= [ARM64] 2304bc3c03ccSJulien Thierry Enables support for pseudo-NMIs in the kernel. This 2305bc3c03ccSJulien Thierry requires the kernel to be built with 2306bc3c03ccSJulien Thierry CONFIG_ARM64_PSEUDO_NMI. 2307bc3c03ccSJulien Thierry 2308e52347bdSJani Nikula irqfixup [HW] 2309e52347bdSJani Nikula When an interrupt is not handled search all handlers 2310e52347bdSJani Nikula for it. Intended to get systems with badly broken 2311e52347bdSJani Nikula firmware running. 2312e52347bdSJani Nikula 2313e52347bdSJani Nikula irqpoll [HW] 2314e52347bdSJani Nikula When an interrupt is not handled search all handlers 2315e52347bdSJani Nikula for it. Also check all handlers each timer 2316e52347bdSJani Nikula interrupt. Intended to get systems with badly broken 2317e52347bdSJani Nikula firmware running. 2318e52347bdSJani Nikula 2319e52347bdSJani Nikula isapnp= [ISAPNP] 2320e52347bdSJani Nikula Format: <RDP>,<reset>,<pci_scan>,<verbosity> 2321e52347bdSJani Nikula 2322d94d1053SFrederic Weisbecker isolcpus= [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance. 2323b0d40d2bSFrederic Weisbecker [Deprecated - use cpusets instead] 2324b0d40d2bSFrederic Weisbecker Format: [flag-list,]<cpu-list> 2325e52347bdSJani Nikula 2326b0d40d2bSFrederic Weisbecker Specify one or more CPUs to isolate from disturbances 2327b0d40d2bSFrederic Weisbecker specified in the flag list (default: domain): 2328b0d40d2bSFrederic Weisbecker 2329b0d40d2bSFrederic Weisbecker nohz 2330b0d40d2bSFrederic Weisbecker Disable the tick when a single task runs. 2331083c6eeaSFrederic Weisbecker 2332083c6eeaSFrederic Weisbecker A residual 1Hz tick is offloaded to workqueues, which you 2333083c6eeaSFrederic Weisbecker need to affine to housekeeping through the global 2334083c6eeaSFrederic Weisbecker workqueue's affinity configured via the 2335083c6eeaSFrederic Weisbecker /sys/devices/virtual/workqueue/cpumask sysfs file, or 2336083c6eeaSFrederic Weisbecker by using the 'domain' flag described below. 2337083c6eeaSFrederic Weisbecker 2338083c6eeaSFrederic Weisbecker NOTE: by default the global workqueue runs on all CPUs, 2339083c6eeaSFrederic Weisbecker so to protect individual CPUs the 'cpumask' file has to 2340083c6eeaSFrederic Weisbecker be configured manually after bootup. 2341083c6eeaSFrederic Weisbecker 2342b0d40d2bSFrederic Weisbecker domain 2343b0d40d2bSFrederic Weisbecker Isolate from the general SMP balancing and scheduling 2344b0d40d2bSFrederic Weisbecker algorithms. Note that performing domain isolation this way 2345b0d40d2bSFrederic Weisbecker is irreversible: it's not possible to bring back a CPU to 2346b0d40d2bSFrederic Weisbecker the domains once isolated through isolcpus. It's strongly 2347b0d40d2bSFrederic Weisbecker advised to use cpusets instead to disable scheduler load 2348b0d40d2bSFrederic Weisbecker balancing through the "cpuset.sched_load_balance" file. 2349b0d40d2bSFrederic Weisbecker It offers a much more flexible interface where CPUs can 2350b0d40d2bSFrederic Weisbecker move in and out of an isolated set anytime. 2351b0d40d2bSFrederic Weisbecker 2352b0d40d2bSFrederic Weisbecker You can move a process onto or off an "isolated" CPU via 2353b0d40d2bSFrederic Weisbecker the CPU affinity syscalls or cpuset. 2354e52347bdSJani Nikula <cpu number> begins at 0 and the maximum value is 2355e52347bdSJani Nikula "number of CPUs in system - 1". 2356e52347bdSJani Nikula 235711ea68f5SMing Lei managed_irq 235811ea68f5SMing Lei 235911ea68f5SMing Lei Isolate from being targeted by managed interrupts 236011ea68f5SMing Lei which have an interrupt mask containing isolated 236111ea68f5SMing Lei CPUs. The affinity of managed interrupts is 236211ea68f5SMing Lei handled by the kernel and cannot be changed via 236311ea68f5SMing Lei the /proc/irq/* interfaces. 236411ea68f5SMing Lei 236511ea68f5SMing Lei This isolation is best effort and only effective 236611ea68f5SMing Lei if the automatically assigned interrupt mask of a 236711ea68f5SMing Lei device queue contains isolated and housekeeping 236811ea68f5SMing Lei CPUs. If housekeeping CPUs are online then such 236911ea68f5SMing Lei interrupts are directed to the housekeeping CPU 237011ea68f5SMing Lei so that IO submitted on the housekeeping CPU 237111ea68f5SMing Lei cannot disturb the isolated CPU. 237211ea68f5SMing Lei 237311ea68f5SMing Lei If a queue's affinity mask contains only isolated 237411ea68f5SMing Lei CPUs then this parameter has no effect on the 237511ea68f5SMing Lei interrupt routing decision, though interrupts are 237611ea68f5SMing Lei only delivered when tasks running on those 237711ea68f5SMing Lei isolated CPUs submit IO. IO submitted on 237811ea68f5SMing Lei housekeeping CPUs has no influence on those 237911ea68f5SMing Lei queues. 238011ea68f5SMing Lei 2381b0d40d2bSFrederic Weisbecker The format of <cpu-list> is described above. 2382b0d40d2bSFrederic Weisbecker 2383e52347bdSJani Nikula iucv= [HW,NET] 2384e52347bdSJani Nikula 2385be3a5b0eSRandy Dunlap ivrs_ioapic [HW,X86-64] 2386e52347bdSJani Nikula Provide an override to the IOAPIC-ID<->DEVICE-ID 2387bbe3a106SSuravee Suthikulpanit mapping provided in the IVRS ACPI table. 2388bbe3a106SSuravee Suthikulpanit By default, PCI segment is 0, and can be omitted. 23891198d231SKim Phillips 23901198d231SKim Phillips For example, to map IOAPIC-ID decimal 10 to 23911198d231SKim Phillips PCI segment 0x1 and PCI device 00:14.0, 23921198d231SKim Phillips write the parameter as: 23931198d231SKim Phillips ivrs_ioapic=10@0001:00:14.0 23941198d231SKim Phillips 23951198d231SKim Phillips Deprecated formats: 2396bbe3a106SSuravee Suthikulpanit * To map IOAPIC-ID decimal 10 to PCI device 00:14.0 2397bbe3a106SSuravee Suthikulpanit write the parameter as: 2398e52347bdSJani Nikula ivrs_ioapic[10]=00:14.0 2399bbe3a106SSuravee Suthikulpanit * To map IOAPIC-ID decimal 10 to PCI segment 0x1 and 2400bbe3a106SSuravee Suthikulpanit PCI device 00:14.0 write the parameter as: 2401bbe3a106SSuravee Suthikulpanit ivrs_ioapic[10]=0001:00:14.0 2402e52347bdSJani Nikula 2403be3a5b0eSRandy Dunlap ivrs_hpet [HW,X86-64] 2404e52347bdSJani Nikula Provide an override to the HPET-ID<->DEVICE-ID 2405bbe3a106SSuravee Suthikulpanit mapping provided in the IVRS ACPI table. 2406bbe3a106SSuravee Suthikulpanit By default, PCI segment is 0, and can be omitted. 24071198d231SKim Phillips 24081198d231SKim Phillips For example, to map HPET-ID decimal 10 to 24091198d231SKim Phillips PCI segment 0x1 and PCI device 00:14.0, 24101198d231SKim Phillips write the parameter as: 24111198d231SKim Phillips ivrs_hpet=10@0001:00:14.0 24121198d231SKim Phillips 24131198d231SKim Phillips Deprecated formats: 2414bbe3a106SSuravee Suthikulpanit * To map HPET-ID decimal 0 to PCI device 00:14.0 2415bbe3a106SSuravee Suthikulpanit write the parameter as: 2416e52347bdSJani Nikula ivrs_hpet[0]=00:14.0 2417bbe3a106SSuravee Suthikulpanit * To map HPET-ID decimal 10 to PCI segment 0x1 and 2418bbe3a106SSuravee Suthikulpanit PCI device 00:14.0 write the parameter as: 2419bbe3a106SSuravee Suthikulpanit ivrs_ioapic[10]=0001:00:14.0 2420e52347bdSJani Nikula 2421be3a5b0eSRandy Dunlap ivrs_acpihid [HW,X86-64] 2422e52347bdSJani Nikula Provide an override to the ACPI-HID:UID<->DEVICE-ID 2423bbe3a106SSuravee Suthikulpanit mapping provided in the IVRS ACPI table. 24241198d231SKim Phillips By default, PCI segment is 0, and can be omitted. 2425bbe3a106SSuravee Suthikulpanit 2426bbe3a106SSuravee Suthikulpanit For example, to map UART-HID:UID AMD0020:0 to 2427bbe3a106SSuravee Suthikulpanit PCI segment 0x1 and PCI device ID 00:14.5, 2428bbe3a106SSuravee Suthikulpanit write the parameter as: 24291198d231SKim Phillips ivrs_acpihid=AMD0020:0@0001:00:14.5 2430bbe3a106SSuravee Suthikulpanit 24311198d231SKim Phillips Deprecated formats: 24321198d231SKim Phillips * To map UART-HID:UID AMD0020:0 to PCI segment is 0, 24331198d231SKim Phillips PCI device ID 00:14.5, write the parameter as: 2434e52347bdSJani Nikula ivrs_acpihid[00:14.5]=AMD0020:0 24351198d231SKim Phillips * To map UART-HID:UID AMD0020:0 to PCI segment 0x1 and 24361198d231SKim Phillips PCI device ID 00:14.5, write the parameter as: 24371198d231SKim Phillips ivrs_acpihid[0001:00:14.5]=AMD0020:0 2438e52347bdSJani Nikula 2439e52347bdSJani Nikula js= [HW,JOY] Analog joystick 24401752118dSTom Saeger See Documentation/input/joydev/joystick.rst. 2441e52347bdSJani Nikula 2442b0845ce5SMark Rutland kasan_multi_shot 2443b0845ce5SMark Rutland [KNL] Enforce KASAN (Kernel Address Sanitizer) to print 2444b0845ce5SMark Rutland report on every invalid memory access. Without this 2445b0845ce5SMark Rutland parameter KASAN will print report only for the first 2446b0845ce5SMark Rutland invalid access. 2447b0845ce5SMark Rutland 244842551b8dSRandy Dunlap keep_bootcon [KNL] 244942551b8dSRandy Dunlap Do not unregister boot console at start. This is only 245042551b8dSRandy Dunlap useful for debugging when something happens in the window 245142551b8dSRandy Dunlap between unregistering the boot console and initializing 245242551b8dSRandy Dunlap the real console. 245342551b8dSRandy Dunlap 2454e52347bdSJani Nikula keepinitrd [HW,ARM] 2455e52347bdSJani Nikula 2456e52347bdSJani Nikula kernelcore= [KNL,X86,IA-64,PPC] 2457a5c6d650SDavid Rientjes Format: nn[KMGTPE] | nn% | "mirror" 2458a5c6d650SDavid Rientjes This parameter specifies the amount of memory usable by 2459a5c6d650SDavid Rientjes the kernel for non-movable allocations. The requested 2460a5c6d650SDavid Rientjes amount is spread evenly throughout all nodes in the 2461a5c6d650SDavid Rientjes system as ZONE_NORMAL. The remaining memory is used for 2462a5c6d650SDavid Rientjes movable memory in its own zone, ZONE_MOVABLE. In the 2463a5c6d650SDavid Rientjes event, a node is too small to have both ZONE_NORMAL and 2464a5c6d650SDavid Rientjes ZONE_MOVABLE, kernelcore memory will take priority and 2465a5c6d650SDavid Rientjes other nodes will have a larger ZONE_MOVABLE. 2466a5c6d650SDavid Rientjes 2467a5c6d650SDavid Rientjes ZONE_MOVABLE is used for the allocation of pages that 2468a5c6d650SDavid Rientjes may be reclaimed or moved by the page migration 2469a5c6d650SDavid Rientjes subsystem. Note that allocations like PTEs-from-HighMem 2470a5c6d650SDavid Rientjes still use the HighMem zone if it exists, and the Normal 2471e52347bdSJani Nikula zone if it does not. 2472e52347bdSJani Nikula 2473a5c6d650SDavid Rientjes It is possible to specify the exact amount of memory in 2474a5c6d650SDavid Rientjes the form of "nn[KMGTPE]", a percentage of total system 2475a5c6d650SDavid Rientjes memory in the form of "nn%", or "mirror". If "mirror" 2476e52347bdSJani Nikula option is specified, mirrored (reliable) memory is used 2477e52347bdSJani Nikula for non-movable allocations and remaining memory is used 2478a5c6d650SDavid Rientjes for Movable pages. "nn[KMGTPE]", "nn%", and "mirror" 2479a5c6d650SDavid Rientjes are exclusive, so you cannot specify multiple forms. 2480e52347bdSJani Nikula 2481e52347bdSJani Nikula kgdbdbgp= [KGDB,HW] kgdb over EHCI usb debug port. 2482e52347bdSJani Nikula Format: <Controller#>[,poll interval] 2483e52347bdSJani Nikula The controller # is the number of the ehci usb debug 2484e52347bdSJani Nikula port as it is probed via PCI. The poll interval is 2485e52347bdSJani Nikula optional and is the number seconds in between 2486e52347bdSJani Nikula each poll cycle to the debug port in case you need 2487e52347bdSJani Nikula the functionality for interrupting the kernel with 2488e52347bdSJani Nikula gdb or control-c on the dbgp connection. When 2489e52347bdSJani Nikula not using this parameter you use sysrq-g to break into 2490e52347bdSJani Nikula the kernel debugger. 2491e52347bdSJani Nikula 2492e52347bdSJani Nikula kgdboc= [KGDB,HW] kgdb over consoles. 2493e52347bdSJani Nikula Requires a tty driver that supports console polling, 2494e52347bdSJani Nikula or a supported polling keyboard driver (non-usb). 2495e52347bdSJani Nikula Serial only format: <serial_device>[,baud] 2496e52347bdSJani Nikula keyboard only format: kbd 2497e52347bdSJani Nikula keyboard and serial format: kbd,<serial_device>[,baud] 2498e52347bdSJani Nikula Optional Kernel mode setting: 2499e52347bdSJani Nikula kms, kbd format: kms,kbd 2500e52347bdSJani Nikula kms, kbd and serial format: kms,kbd,<ser_dev>[,baud] 2501e52347bdSJani Nikula 2502f71fc3bcSDouglas Anderson kgdboc_earlycon= [KGDB,HW] 2503f71fc3bcSDouglas Anderson If the boot console provides the ability to read 2504f71fc3bcSDouglas Anderson characters and can work in polling mode, you can use 2505f71fc3bcSDouglas Anderson this parameter to tell kgdb to use it as a backend 2506f71fc3bcSDouglas Anderson until the normal console is registered. Intended to 2507f71fc3bcSDouglas Anderson be used together with the kgdboc parameter which 2508f71fc3bcSDouglas Anderson specifies the normal console to transition to. 2509f71fc3bcSDouglas Anderson 2510f71fc3bcSDouglas Anderson The name of the early console should be specified 2511f71fc3bcSDouglas Anderson as the value of this parameter. Note that the name of 2512f71fc3bcSDouglas Anderson the early console might be different than the tty 2513f71fc3bcSDouglas Anderson name passed to kgdboc. It's OK to leave the value 2514f71fc3bcSDouglas Anderson blank and the first boot console that implements 2515f71fc3bcSDouglas Anderson read() will be picked. 2516f71fc3bcSDouglas Anderson 2517e52347bdSJani Nikula kgdbwait [KGDB] Stop kernel execution and enter the 2518e52347bdSJani Nikula kernel debugger at the earliest opportunity. 2519e52347bdSJani Nikula 2520497de97eSRandy Dunlap kmac= [MIPS] Korina ethernet MAC address. 2521e52347bdSJani Nikula Configure the RouterBoard 532 series on-chip 2522e52347bdSJani Nikula Ethernet adapter MAC address. 2523e52347bdSJani Nikula 2524e52347bdSJani Nikula kmemleak= [KNL] Boot-time kmemleak enable/disable 2525e52347bdSJani Nikula Valid arguments: on, off 2526e52347bdSJani Nikula Default: on 2527e52347bdSJani Nikula Built with CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF=y, 2528e52347bdSJani Nikula the default is off. 2529e52347bdSJani Nikula 2530970988e1SMasami Hiramatsu kprobe_event=[probe-list] 2531970988e1SMasami Hiramatsu [FTRACE] Add kprobe events and enable at boot time. 2532970988e1SMasami Hiramatsu The probe-list is a semicolon delimited list of probe 2533970988e1SMasami Hiramatsu definitions. Each definition is same as kprobe_events 2534970988e1SMasami Hiramatsu interface, but the parameters are comma delimited. 2535970988e1SMasami Hiramatsu For example, to add a kprobe event on vfs_read with 2536970988e1SMasami Hiramatsu arg1 and arg2, add to the command line; 2537970988e1SMasami Hiramatsu 2538970988e1SMasami Hiramatsu kprobe_event=p,vfs_read,$arg1,$arg2 2539970988e1SMasami Hiramatsu 2540970988e1SMasami Hiramatsu See also Documentation/trace/kprobetrace.rst "Kernel 2541970988e1SMasami Hiramatsu Boot Parameter" section. 2542970988e1SMasami Hiramatsu 2543de190555SJeremy Linton kpti= [ARM64] Control page table isolation of user 2544de190555SJeremy Linton and kernel address spaces. 2545de190555SJeremy Linton Default: enabled on cores which need mitigation. 2546de190555SJeremy Linton 0: force disabled 2547de190555SJeremy Linton 1: force enabled 2548de190555SJeremy Linton 2549d20a6ba5SJoe Fradley kunit.enable= [KUNIT] Enable executing KUnit tests. Requires 2550d20a6ba5SJoe Fradley CONFIG_KUNIT to be set to be fully enabled. The 2551d20a6ba5SJoe Fradley default value can be overridden via 2552d20a6ba5SJoe Fradley KUNIT_DEFAULT_ENABLED. 2553d20a6ba5SJoe Fradley Default is 1 (enabled) 2554d20a6ba5SJoe Fradley 2555e52347bdSJani Nikula kvm.ignore_msrs=[KVM] Ignore guest accesses to unhandled MSRs. 2556e52347bdSJani Nikula Default is 0 (don't ignore, but inject #GP) 2557e52347bdSJani Nikula 2558a3fe5dbdSDavid Matlack kvm.eager_page_split= 2559a3fe5dbdSDavid Matlack [KVM,X86] Controls whether or not KVM will try to 2560a3fe5dbdSDavid Matlack proactively split all huge pages during dirty logging. 2561a3fe5dbdSDavid Matlack Eager page splitting reduces interruptions to vCPU 2562a3fe5dbdSDavid Matlack execution by eliminating the write-protection faults 2563a3fe5dbdSDavid Matlack and MMU lock contention that would otherwise be 2564a3fe5dbdSDavid Matlack required to split huge pages lazily. 2565a3fe5dbdSDavid Matlack 2566a3fe5dbdSDavid Matlack VM workloads that rarely perform writes or that write 2567a3fe5dbdSDavid Matlack only to a small region of VM memory may benefit from 2568a3fe5dbdSDavid Matlack disabling eager page splitting to allow huge pages to 2569a3fe5dbdSDavid Matlack still be used for reads. 2570a3fe5dbdSDavid Matlack 2571a3fe5dbdSDavid Matlack The behavior of eager page splitting depends on whether 2572a3fe5dbdSDavid Matlack KVM_DIRTY_LOG_INITIALLY_SET is enabled or disabled. If 2573a3fe5dbdSDavid Matlack disabled, all huge pages in a memslot will be eagerly 2574a3fe5dbdSDavid Matlack split when dirty logging is enabled on that memslot. If 2575cb00a70bSDavid Matlack enabled, eager page splitting will be performed during 2576cb00a70bSDavid Matlack the KVM_CLEAR_DIRTY ioctl, and only for the pages being 2577cb00a70bSDavid Matlack cleared. 2578a3fe5dbdSDavid Matlack 2579ada51a9dSDavid Matlack Eager page splitting is only supported when kvm.tdp_mmu=Y. 2580a3fe5dbdSDavid Matlack 2581a3fe5dbdSDavid Matlack Default is Y (on). 2582a3fe5dbdSDavid Matlack 2583c4ae60e4SLiran Alon kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface. 2584c4ae60e4SLiran Alon Default is false (don't support). 2585c4ae60e4SLiran Alon 2586b8e8c830SPaolo Bonzini kvm.nx_huge_pages= 2587b8e8c830SPaolo Bonzini [KVM] Controls the software workaround for the 2588b8e8c830SPaolo Bonzini X86_BUG_ITLB_MULTIHIT bug. 2589b8e8c830SPaolo Bonzini force : Always deploy workaround. 2590b8e8c830SPaolo Bonzini off : Never deploy workaround. 2591b8e8c830SPaolo Bonzini auto : Deploy workaround based on the presence of 2592b8e8c830SPaolo Bonzini X86_BUG_ITLB_MULTIHIT. 2593b8e8c830SPaolo Bonzini 2594b8e8c830SPaolo Bonzini Default is 'auto'. 2595b8e8c830SPaolo Bonzini 2596b8e8c830SPaolo Bonzini If the software workaround is enabled for the host, 2597b8e8c830SPaolo Bonzini guests do need not to enable it for nested guests. 2598b8e8c830SPaolo Bonzini 25991aa9b957SJunaid Shahid kvm.nx_huge_pages_recovery_ratio= 26001aa9b957SJunaid Shahid [KVM] Controls how many 4KiB pages are periodically zapped 26011aa9b957SJunaid Shahid back to huge pages. 0 disables the recovery, otherwise if 26021aa9b957SJunaid Shahid the value is N KVM will zap 1/Nth of the 4KiB pages every 26034dfe4f40SJunaid Shahid period (see below). The default is 60. 26044dfe4f40SJunaid Shahid 26054dfe4f40SJunaid Shahid kvm.nx_huge_pages_recovery_period_ms= 26064dfe4f40SJunaid Shahid [KVM] Controls the time period at which KVM zaps 4KiB pages 26074dfe4f40SJunaid Shahid back to huge pages. If the value is a non-zero N, KVM will 26084dfe4f40SJunaid Shahid zap a portion (see ratio above) of the pages every N msecs. 26094dfe4f40SJunaid Shahid If the value is 0 (the default), KVM will pick a period based 26104dfe4f40SJunaid Shahid on the ratio, such that a page is zapped after 1 hour on average. 26111aa9b957SJunaid Shahid 2612f41dd67dSYan-Jie Wang kvm-amd.nested= [KVM,AMD] Control nested virtualization feature in 2613f41dd67dSYan-Jie Wang KVM/SVM. Default is 1 (enabled). 2614e52347bdSJani Nikula 2615f41dd67dSYan-Jie Wang kvm-amd.npt= [KVM,AMD] Control KVM's use of Nested Page Tables, 2616f41dd67dSYan-Jie Wang a.k.a. Two-Dimensional Page Tables. Default is 1 2617f41dd67dSYan-Jie Wang (enabled). Disable by KVM if hardware lacks support 2618f41dd67dSYan-Jie Wang for NPT. 2619e52347bdSJani Nikula 2620d8b369c4SDavid Brazdil kvm-arm.mode= 2621d8b369c4SDavid Brazdil [KVM,ARM] Select one of KVM/arm64's modes of operation. 2622d8b369c4SDavid Brazdil 2623b6a68b97SMarc Zyngier none: Forcefully disable KVM. 2624b6a68b97SMarc Zyngier 26251945a067SMarc Zyngier nvhe: Standard nVHE-based mode, without support for 26261945a067SMarc Zyngier protected guests. 26271945a067SMarc Zyngier 2628d8b369c4SDavid Brazdil protected: nVHE-based mode with support for guests whose 2629d8b369c4SDavid Brazdil state is kept private from the host. 2630d8b369c4SDavid Brazdil 2631675cabc8SJintack Lim nested: VHE-based mode with support for nested 2632675cabc8SJintack Lim virtualization. Requires at least ARMv8.3 2633675cabc8SJintack Lim hardware. 2634675cabc8SJintack Lim 263553e8ce13SAlexandru Elisei Defaults to VHE/nVHE based on hardware support. Setting 263653e8ce13SAlexandru Elisei mode to "protected" will disable kexec and hibernation 2637675cabc8SJintack Lim for the host. "nested" is experimental and should be 2638675cabc8SJintack Lim used with extreme caution. 2639d8b369c4SDavid Brazdil 2640e23f62f7SMarc Zyngier kvm-arm.vgic_v3_group0_trap= 2641e23f62f7SMarc Zyngier [KVM,ARM] Trap guest accesses to GICv3 group-0 2642e23f62f7SMarc Zyngier system registers 2643e23f62f7SMarc Zyngier 2644182936eeSMarc Zyngier kvm-arm.vgic_v3_group1_trap= 2645182936eeSMarc Zyngier [KVM,ARM] Trap guest accesses to GICv3 group-1 2646182936eeSMarc Zyngier system registers 2647182936eeSMarc Zyngier 2648ff89511eSMarc Zyngier kvm-arm.vgic_v3_common_trap= 2649ff89511eSMarc Zyngier [KVM,ARM] Trap guest accesses to GICv3 common 2650ff89511eSMarc Zyngier system registers 2651ff89511eSMarc Zyngier 2652a7546054SMarc Zyngier kvm-arm.vgic_v4_enable= 2653a7546054SMarc Zyngier [KVM,ARM] Allow use of GICv4 for direct injection of 2654a7546054SMarc Zyngier LPIs. 2655a7546054SMarc Zyngier 2656aed26eebSSatheesh Rajendran kvm_cma_resv_ratio=n [PPC] 2657aed26eebSSatheesh Rajendran Reserves given percentage from system memory area for 2658aed26eebSSatheesh Rajendran contiguous memory allocation for KVM hash pagetable 2659aed26eebSSatheesh Rajendran allocation. 2660aed26eebSSatheesh Rajendran By default it reserves 5% of total system memory. 2661aed26eebSSatheesh Rajendran Format: <integer> 2662aed26eebSSatheesh Rajendran Default: 5 2663aed26eebSSatheesh Rajendran 2664f41dd67dSYan-Jie Wang kvm-intel.ept= [KVM,Intel] Control KVM's use of Extended Page Tables, 2665f41dd67dSYan-Jie Wang a.k.a. Two-Dimensional Page Tables. Default is 1 2666f41dd67dSYan-Jie Wang (enabled). Disable by KVM if hardware lacks support 2667f41dd67dSYan-Jie Wang for EPT. 2668e52347bdSJani Nikula 2669e52347bdSJani Nikula kvm-intel.emulate_invalid_guest_state= 2670f41dd67dSYan-Jie Wang [KVM,Intel] Control whether to emulate invalid guest 2671f41dd67dSYan-Jie Wang state. Ignored if kvm-intel.enable_unrestricted_guest=1, 2672f41dd67dSYan-Jie Wang as guest state is never invalid for unrestricted 2673f41dd67dSYan-Jie Wang guests. This param doesn't apply to nested guests (L2), 2674f41dd67dSYan-Jie Wang as KVM never emulates invalid L2 guest state. 2675f41dd67dSYan-Jie Wang Default is 1 (enabled). 2676e52347bdSJani Nikula 2677e52347bdSJani Nikula kvm-intel.flexpriority= 2678f41dd67dSYan-Jie Wang [KVM,Intel] Control KVM's use of FlexPriority feature 2679d56b699dSBjorn Helgaas (TPR shadow). Default is 1 (enabled). Disable by KVM if 2680f41dd67dSYan-Jie Wang hardware lacks support for it. 2681e52347bdSJani Nikula 2682e52347bdSJani Nikula kvm-intel.nested= 2683f41dd67dSYan-Jie Wang [KVM,Intel] Control nested virtualization feature in 2684f41dd67dSYan-Jie Wang KVM/VMX. Default is 1 (enabled). 2685e52347bdSJani Nikula 2686e52347bdSJani Nikula kvm-intel.unrestricted_guest= 2687f41dd67dSYan-Jie Wang [KVM,Intel] Control KVM's use of unrestricted guest 2688f41dd67dSYan-Jie Wang feature (virtualized real and unpaged mode). Default 2689f41dd67dSYan-Jie Wang is 1 (enabled). Disable by KVM if EPT is disabled or 2690f41dd67dSYan-Jie Wang hardware lacks support for it. 2691e52347bdSJani Nikula 2692a399477eSKonrad Rzeszutek Wilk kvm-intel.vmentry_l1d_flush=[KVM,Intel] Mitigation for L1 Terminal Fault 2693a399477eSKonrad Rzeszutek Wilk CVE-2018-3620. 2694a399477eSKonrad Rzeszutek Wilk 2695a399477eSKonrad Rzeszutek Wilk Valid arguments: never, cond, always 2696a399477eSKonrad Rzeszutek Wilk 2697a399477eSKonrad Rzeszutek Wilk always: L1D cache flush on every VMENTER. 2698a399477eSKonrad Rzeszutek Wilk cond: Flush L1D on VMENTER only when the code between 2699a399477eSKonrad Rzeszutek Wilk VMEXIT and VMENTER can leak host memory. 2700a399477eSKonrad Rzeszutek Wilk never: Disables the mitigation 2701a399477eSKonrad Rzeszutek Wilk 2702a399477eSKonrad Rzeszutek Wilk Default is cond (do L1 cache flush in specific instances) 2703a399477eSKonrad Rzeszutek Wilk 2704f41dd67dSYan-Jie Wang kvm-intel.vpid= [KVM,Intel] Control KVM's use of Virtual Processor 2705f41dd67dSYan-Jie Wang Identification feature (tagged TLBs). Default is 1 2706f41dd67dSYan-Jie Wang (enabled). Disable by KVM if hardware lacks support 2707f41dd67dSYan-Jie Wang for it. 2708e52347bdSJani Nikula 2709b7fe54f6SBalbir Singh l1d_flush= [X86,INTEL] 2710b7fe54f6SBalbir Singh Control mitigation for L1D based snooping vulnerability. 2711b7fe54f6SBalbir Singh 2712b7fe54f6SBalbir Singh Certain CPUs are vulnerable to an exploit against CPU 2713b7fe54f6SBalbir Singh internal buffers which can forward information to a 2714b7fe54f6SBalbir Singh disclosure gadget under certain conditions. 2715b7fe54f6SBalbir Singh 2716b7fe54f6SBalbir Singh In vulnerable processors, the speculatively 2717b7fe54f6SBalbir Singh forwarded data can be used in a cache side channel 2718b7fe54f6SBalbir Singh attack, to access data to which the attacker does 2719b7fe54f6SBalbir Singh not have direct access. 2720b7fe54f6SBalbir Singh 2721b7fe54f6SBalbir Singh This parameter controls the mitigation. The 2722b7fe54f6SBalbir Singh options are: 2723b7fe54f6SBalbir Singh 2724b7fe54f6SBalbir Singh on - enable the interface for the mitigation 2725b7fe54f6SBalbir Singh 2726d90a7a0eSJiri Kosina l1tf= [X86] Control mitigation of the L1TF vulnerability on 2727d90a7a0eSJiri Kosina affected CPUs 2728d90a7a0eSJiri Kosina 2729d90a7a0eSJiri Kosina The kernel PTE inversion protection is unconditionally 2730d90a7a0eSJiri Kosina enabled and cannot be disabled. 2731d90a7a0eSJiri Kosina 2732d90a7a0eSJiri Kosina full 2733d90a7a0eSJiri Kosina Provides all available mitigations for the 2734d90a7a0eSJiri Kosina L1TF vulnerability. Disables SMT and 2735d90a7a0eSJiri Kosina enables all mitigations in the 2736d90a7a0eSJiri Kosina hypervisors, i.e. unconditional L1D flush. 2737d90a7a0eSJiri Kosina 2738d90a7a0eSJiri Kosina SMT control and L1D flush control via the 2739d90a7a0eSJiri Kosina sysfs interface is still possible after 2740d90a7a0eSJiri Kosina boot. Hypervisors will issue a warning 2741d90a7a0eSJiri Kosina when the first VM is started in a 2742d90a7a0eSJiri Kosina potentially insecure configuration, 2743d90a7a0eSJiri Kosina i.e. SMT enabled or L1D flush disabled. 2744d90a7a0eSJiri Kosina 2745d90a7a0eSJiri Kosina full,force 2746d90a7a0eSJiri Kosina Same as 'full', but disables SMT and L1D 2747d90a7a0eSJiri Kosina flush runtime control. Implies the 2748d90a7a0eSJiri Kosina 'nosmt=force' command line option. 2749d90a7a0eSJiri Kosina (i.e. sysfs control of SMT is disabled.) 2750d90a7a0eSJiri Kosina 2751d90a7a0eSJiri Kosina flush 2752d90a7a0eSJiri Kosina Leaves SMT enabled and enables the default 2753d90a7a0eSJiri Kosina hypervisor mitigation, i.e. conditional 2754d90a7a0eSJiri Kosina L1D flush. 2755d90a7a0eSJiri Kosina 2756d90a7a0eSJiri Kosina SMT control and L1D flush control via the 2757d90a7a0eSJiri Kosina sysfs interface is still possible after 2758d90a7a0eSJiri Kosina boot. Hypervisors will issue a warning 2759d90a7a0eSJiri Kosina when the first VM is started in a 2760d90a7a0eSJiri Kosina potentially insecure configuration, 2761d90a7a0eSJiri Kosina i.e. SMT enabled or L1D flush disabled. 2762d90a7a0eSJiri Kosina 2763d90a7a0eSJiri Kosina flush,nosmt 2764d90a7a0eSJiri Kosina 2765d90a7a0eSJiri Kosina Disables SMT and enables the default 2766d90a7a0eSJiri Kosina hypervisor mitigation. 2767d90a7a0eSJiri Kosina 2768d90a7a0eSJiri Kosina SMT control and L1D flush control via the 2769d90a7a0eSJiri Kosina sysfs interface is still possible after 2770d90a7a0eSJiri Kosina boot. Hypervisors will issue a warning 2771d90a7a0eSJiri Kosina when the first VM is started in a 2772d90a7a0eSJiri Kosina potentially insecure configuration, 2773d90a7a0eSJiri Kosina i.e. SMT enabled or L1D flush disabled. 2774d90a7a0eSJiri Kosina 2775d90a7a0eSJiri Kosina flush,nowarn 2776d90a7a0eSJiri Kosina Same as 'flush', but hypervisors will not 2777d90a7a0eSJiri Kosina warn when a VM is started in a potentially 2778d90a7a0eSJiri Kosina insecure configuration. 2779d90a7a0eSJiri Kosina 2780d90a7a0eSJiri Kosina off 2781d90a7a0eSJiri Kosina Disables hypervisor mitigations and doesn't 2782d90a7a0eSJiri Kosina emit any warnings. 27835b5e4d62SMichal Hocko It also drops the swap size and available 27845b5e4d62SMichal Hocko RAM limit restriction on both hypervisor and 27855b5e4d62SMichal Hocko bare metal. 2786d90a7a0eSJiri Kosina 2787d90a7a0eSJiri Kosina Default is 'flush'. 2788d90a7a0eSJiri Kosina 278965fd4cb6SThomas Gleixner For details see: Documentation/admin-guide/hw-vuln/l1tf.rst 2790d90a7a0eSJiri Kosina 2791e52347bdSJani Nikula l2cr= [PPC] 2792e52347bdSJani Nikula 2793e52347bdSJani Nikula l3cr= [PPC] 2794e52347bdSJani Nikula 2795e52347bdSJani Nikula lapic [X86-32,APIC] Enable the local APIC even if BIOS 2796e52347bdSJani Nikula disabled it. 2797e52347bdSJani Nikula 2798622381e6SRandy Dunlap lapic= [X86,APIC] Do not use TSC deadline 2799e52347bdSJani Nikula value for LAPIC timer one-shot implementation. Default 2800e52347bdSJani Nikula back to the programmable timer unit in the LAPIC. 2801622381e6SRandy Dunlap Format: notscdeadline 2802e52347bdSJani Nikula 2803e52347bdSJani Nikula lapic_timer_c2_ok [X86,APIC] trust the local apic timer 2804e52347bdSJani Nikula in C2 power state. 2805e52347bdSJani Nikula 2806e52347bdSJani Nikula libata.dma= [LIBATA] DMA control 2807e52347bdSJani Nikula libata.dma=0 Disable all PATA and SATA DMA 2808e52347bdSJani Nikula libata.dma=1 PATA and SATA Disk DMA only 2809e52347bdSJani Nikula libata.dma=2 ATAPI (CDROM) DMA only 2810e52347bdSJani Nikula libata.dma=4 Compact Flash DMA only 2811e52347bdSJani Nikula Combinations also work, so libata.dma=3 enables DMA 2812e52347bdSJani Nikula for disks and CDROMs, but not CFs. 2813e52347bdSJani Nikula 2814e52347bdSJani Nikula libata.ignore_hpa= [LIBATA] Ignore HPA limit 2815e52347bdSJani Nikula libata.ignore_hpa=0 keep BIOS limits (default) 2816e52347bdSJani Nikula libata.ignore_hpa=1 ignore limits, using full disk 2817e52347bdSJani Nikula 2818e52347bdSJani Nikula libata.noacpi [LIBATA] Disables use of ACPI in libata suspend/resume 2819e52347bdSJani Nikula when set. 2820e52347bdSJani Nikula Format: <int> 2821e52347bdSJani Nikula 2822fa82cabbSDamien Le Moal libata.force= [LIBATA] Force configurations. The format is a comma- 2823fa82cabbSDamien Le Moal separated list of "[ID:]VAL" where ID is PORT[.DEVICE]. 2824fa82cabbSDamien Le Moal PORT and DEVICE are decimal numbers matching port, link 2825fa82cabbSDamien Le Moal or device. Basically, it matches the ATA ID string 2826fa82cabbSDamien Le Moal printed on console by libata. If the whole ID part is 2827fa82cabbSDamien Le Moal omitted, the last PORT and DEVICE values are used. If 2828fa82cabbSDamien Le Moal ID hasn't been specified yet, the configuration applies 2829fa82cabbSDamien Le Moal to all ports, links and devices. 2830e52347bdSJani Nikula 2831e52347bdSJani Nikula If only DEVICE is omitted, the parameter applies to 2832e52347bdSJani Nikula the port and all links and devices behind it. DEVICE 2833e52347bdSJani Nikula number of 0 either selects the first device or the 2834e52347bdSJani Nikula first fan-out link behind PMP device. It does not 2835e52347bdSJani Nikula select the host link. DEVICE number of 15 selects the 2836e52347bdSJani Nikula host link and device attached to it. 2837e52347bdSJani Nikula 2838e52347bdSJani Nikula The VAL specifies the configuration to force. As long 2839fa82cabbSDamien Le Moal as there is no ambiguity, shortcut notation is allowed. 2840e52347bdSJani Nikula For example, both 1.5 and 1.5G would work for 1.5Gbps. 2841e52347bdSJani Nikula The following configurations can be forced. 2842e52347bdSJani Nikula 2843e52347bdSJani Nikula * Cable type: 40c, 80c, short40c, unk, ign or sata. 2844e52347bdSJani Nikula Any ID with matching PORT is used. 2845e52347bdSJani Nikula 2846e52347bdSJani Nikula * SATA link speed limit: 1.5Gbps or 3.0Gbps. 2847e52347bdSJani Nikula 2848e52347bdSJani Nikula * Transfer mode: pio[0-7], mwdma[0-4] and udma[0-7]. 2849e52347bdSJani Nikula udma[/][16,25,33,44,66,100,133] notation is also 2850e52347bdSJani Nikula allowed. 2851e52347bdSJani Nikula 2852fa82cabbSDamien Le Moal * nohrst, nosrst, norst: suppress hard, soft and both 2853fa82cabbSDamien Le Moal resets. 2854fa82cabbSDamien Le Moal 2855fa82cabbSDamien Le Moal * rstonce: only attempt one reset during hot-unplug 2856fa82cabbSDamien Le Moal link recovery. 2857fa82cabbSDamien Le Moal 2858fa82cabbSDamien Le Moal * [no]dbdelay: Enable or disable the extra 200ms delay 2859fa82cabbSDamien Le Moal before debouncing a link PHY and device presence 2860fa82cabbSDamien Le Moal detection. 2861fa82cabbSDamien Le Moal 2862e52347bdSJani Nikula * [no]ncq: Turn on or off NCQ. 2863e52347bdSJani Nikula 2864fa82cabbSDamien Le Moal * [no]ncqtrim: Enable or disable queued DSM TRIM. 2865e52347bdSJani Nikula 2866fa82cabbSDamien Le Moal * [no]ncqati: Enable or disable NCQ trim on ATI chipset. 2867e52347bdSJani Nikula 2868fa82cabbSDamien Le Moal * [no]trim: Enable or disable (unqueued) TRIM. 2869e52347bdSJani Nikula 2870fa82cabbSDamien Le Moal * trim_zero: Indicate that TRIM command zeroes data. 2871e52347bdSJani Nikula 2872fa82cabbSDamien Le Moal * max_trim_128m: Set 128M maximum trim size limit. 2873fa82cabbSDamien Le Moal 2874fa82cabbSDamien Le Moal * [no]dma: Turn on or off DMA transfers. 2875fa82cabbSDamien Le Moal 2876fa82cabbSDamien Le Moal * atapi_dmadir: Enable ATAPI DMADIR bridge support. 2877fa82cabbSDamien Le Moal 2878fa82cabbSDamien Le Moal * atapi_mod16_dma: Enable the use of ATAPI DMA for 2879fa82cabbSDamien Le Moal commands that are not a multiple of 16 bytes. 2880fa82cabbSDamien Le Moal 2881fa82cabbSDamien Le Moal * [no]dmalog: Enable or disable the use of the 2882fa82cabbSDamien Le Moal READ LOG DMA EXT command to access logs. 2883fa82cabbSDamien Le Moal 2884fa82cabbSDamien Le Moal * [no]iddevlog: Enable or disable access to the 2885fa82cabbSDamien Le Moal identify device data log. 2886fa82cabbSDamien Le Moal 2887fa82cabbSDamien Le Moal * [no]logdir: Enable or disable access to the general 2888fa82cabbSDamien Le Moal purpose log directory. 2889fa82cabbSDamien Le Moal 2890fa82cabbSDamien Le Moal * max_sec_128: Set transfer size limit to 128 sectors. 2891fa82cabbSDamien Le Moal 2892fa82cabbSDamien Le Moal * max_sec_1024: Set or clear transfer size limit to 2893fa82cabbSDamien Le Moal 1024 sectors. 2894fa82cabbSDamien Le Moal 2895fa82cabbSDamien Le Moal * max_sec_lba48: Set or clear transfer size limit to 2896fa82cabbSDamien Le Moal 65535 sectors. 2897fa82cabbSDamien Le Moal 2898fa82cabbSDamien Le Moal * [no]lpm: Enable or disable link power management. 2899fa82cabbSDamien Le Moal 2900fa82cabbSDamien Le Moal * [no]setxfer: Indicate if transfer speed mode setting 2901fa82cabbSDamien Le Moal should be skipped. 2902fa82cabbSDamien Le Moal 29034d2e4980SDamien Le Moal * [no]fua: Disable or enable FUA (Force Unit Access) 29044d2e4980SDamien Le Moal support for devices supporting this feature. 29054d2e4980SDamien Le Moal 2906fa82cabbSDamien Le Moal * dump_id: Dump IDENTIFY data. 2907e52347bdSJani Nikula 2908e52347bdSJani Nikula * disable: Disable this device. 2909e52347bdSJani Nikula 2910e52347bdSJani Nikula If there are multiple matching configurations changing 2911e52347bdSJani Nikula the same attribute, the last one is used. 2912e52347bdSJani Nikula 29136b99e6e6SRandy Dunlap load_ramdisk= [RAM] [Deprecated] 2914e52347bdSJani Nikula 2915e52347bdSJani Nikula lockd.nlm_grace_period=P [NFS] Assign grace period. 2916e52347bdSJani Nikula Format: <integer> 2917e52347bdSJani Nikula 2918e52347bdSJani Nikula lockd.nlm_tcpport=N [NFS] Assign TCP port. 2919e52347bdSJani Nikula Format: <integer> 2920e52347bdSJani Nikula 2921e52347bdSJani Nikula lockd.nlm_timeout=T [NFS] Assign timeout value. 2922e52347bdSJani Nikula Format: <integer> 2923e52347bdSJani Nikula 2924e52347bdSJani Nikula lockd.nlm_udpport=M [NFS] Assign UDP port. 2925e52347bdSJani Nikula Format: <integer> 2926e52347bdSJani Nikula 2927000d388eSMatthew Garrett lockdown= [SECURITY] 2928000d388eSMatthew Garrett { integrity | confidentiality } 2929000d388eSMatthew Garrett Enable the kernel lockdown feature. If set to 2930000d388eSMatthew Garrett integrity, kernel features that allow userland to 2931000d388eSMatthew Garrett modify the running kernel are disabled. If set to 2932000d388eSMatthew Garrett confidentiality, kernel features that allow userland 2933000d388eSMatthew Garrett to extract confidential information from the kernel 2934000d388eSMatthew Garrett are also disabled. 2935000d388eSMatthew Garrett 2936e52347bdSJani Nikula locktorture.nreaders_stress= [KNL] 2937e52347bdSJani Nikula Set the number of locking read-acquisition kthreads. 2938e52347bdSJani Nikula Defaults to being automatically set based on the 2939e52347bdSJani Nikula number of online CPUs. 2940e52347bdSJani Nikula 2941e52347bdSJani Nikula locktorture.nwriters_stress= [KNL] 2942e52347bdSJani Nikula Set the number of locking write-acquisition kthreads. 2943e52347bdSJani Nikula 2944e52347bdSJani Nikula locktorture.onoff_holdoff= [KNL] 2945e52347bdSJani Nikula Set time (s) after boot for CPU-hotplug testing. 2946e52347bdSJani Nikula 2947e52347bdSJani Nikula locktorture.onoff_interval= [KNL] 2948e52347bdSJani Nikula Set time (s) between CPU-hotplug operations, or 2949e52347bdSJani Nikula zero to disable CPU-hotplug testing. 2950e52347bdSJani Nikula 2951e52347bdSJani Nikula locktorture.shuffle_interval= [KNL] 2952e52347bdSJani Nikula Set task-shuffle interval (jiffies). Shuffling 2953e52347bdSJani Nikula tasks allows some CPUs to go into dyntick-idle 2954e52347bdSJani Nikula mode during the locktorture test. 2955e52347bdSJani Nikula 2956e52347bdSJani Nikula locktorture.shutdown_secs= [KNL] 2957e52347bdSJani Nikula Set time (s) after boot system shutdown. This 2958e52347bdSJani Nikula is useful for hands-off automated testing. 2959e52347bdSJani Nikula 2960e52347bdSJani Nikula locktorture.stat_interval= [KNL] 2961e52347bdSJani Nikula Time (s) between statistics printk()s. 2962e52347bdSJani Nikula 2963e52347bdSJani Nikula locktorture.stutter= [KNL] 2964e52347bdSJani Nikula Time (s) to stutter testing, for example, 2965e52347bdSJani Nikula specifying five seconds causes the test to run for 2966e52347bdSJani Nikula five seconds, wait for five seconds, and so on. 2967e52347bdSJani Nikula This tests the locking primitive's ability to 2968e52347bdSJani Nikula transition abruptly to and from idle. 2969e52347bdSJani Nikula 2970e52347bdSJani Nikula locktorture.torture_type= [KNL] 2971e52347bdSJani Nikula Specify the locking implementation to test. 2972e52347bdSJani Nikula 29735d248bb3SDietmar Eggemann locktorture.writer_fifo= [KNL] 29745d248bb3SDietmar Eggemann Run the write-side locktorture kthreads at 29755d248bb3SDietmar Eggemann sched_set_fifo() real-time priority. 29765d248bb3SDietmar Eggemann 2977e52347bdSJani Nikula locktorture.verbose= [KNL] 2978e52347bdSJani Nikula Enable additional printk() statements. 2979e52347bdSJani Nikula 2980e52347bdSJani Nikula logibm.irq= [HW,MOUSE] Logitech Bus Mouse Driver 2981e52347bdSJani Nikula Format: <irq> 2982e52347bdSJani Nikula 2983e52347bdSJani Nikula loglevel= All Kernel Messages with a loglevel smaller than the 2984e52347bdSJani Nikula console loglevel will be printed to the console. It can 2985e52347bdSJani Nikula also be changed with klogd or other programs. The 2986e52347bdSJani Nikula loglevels are defined as follows: 2987e52347bdSJani Nikula 2988e52347bdSJani Nikula 0 (KERN_EMERG) system is unusable 2989e52347bdSJani Nikula 1 (KERN_ALERT) action must be taken immediately 2990e52347bdSJani Nikula 2 (KERN_CRIT) critical conditions 2991e52347bdSJani Nikula 3 (KERN_ERR) error conditions 2992e52347bdSJani Nikula 4 (KERN_WARNING) warning conditions 2993e52347bdSJani Nikula 5 (KERN_NOTICE) normal but significant condition 2994e52347bdSJani Nikula 6 (KERN_INFO) informational 2995e52347bdSJani Nikula 7 (KERN_DEBUG) debug-level messages 2996e52347bdSJani Nikula 2997e52347bdSJani Nikula log_buf_len=n[KMG] Sets the size of the printk ring buffer, 2998e52347bdSJani Nikula in bytes. n must be a power of two and greater 2999e52347bdSJani Nikula than the minimal size. The minimal size is defined 3000e52347bdSJani Nikula by LOG_BUF_SHIFT kernel config parameter. There is 3001e52347bdSJani Nikula also CONFIG_LOG_CPU_MAX_BUF_SHIFT config parameter 3002e52347bdSJani Nikula that allows to increase the default size depending on 3003e52347bdSJani Nikula the number of CPUs. See init/Kconfig for more details. 3004e52347bdSJani Nikula 3005e52347bdSJani Nikula logo.nologo [FB] Disables display of the built-in Linux logo. 3006e52347bdSJani Nikula This may be used to provide more screen space for 3007e52347bdSJani Nikula kernel log messages and is useful when debugging 3008e52347bdSJani Nikula kernel boot problems. 3009e52347bdSJani Nikula 3010e52347bdSJani Nikula lp=0 [LP] Specify parallel ports to use, e.g, 3011e52347bdSJani Nikula lp=port[,port...] lp=none,parport0 (lp0 not configured, lp1 uses 3012e52347bdSJani Nikula lp=reset first parallel port). 'lp=0' disables the 3013e52347bdSJani Nikula lp=auto printer driver. 'lp=reset' (which can be 3014e52347bdSJani Nikula specified in addition to the ports) causes 3015e52347bdSJani Nikula attached printers to be reset. Using 3016e52347bdSJani Nikula lp=port1,port2,... specifies the parallel ports 3017e52347bdSJani Nikula to associate lp devices with, starting with 3018e52347bdSJani Nikula lp0. A port specification may be 'none' to skip 3019e52347bdSJani Nikula that lp device, or a parport name such as 3020e52347bdSJani Nikula 'parport0'. Specifying 'lp=auto' instead of a 3021e52347bdSJani Nikula port specification list means that device IDs 3022e52347bdSJani Nikula from each port should be examined, to see if 3023e52347bdSJani Nikula an IEEE 1284-compliant printer is attached; if 3024e52347bdSJani Nikula so, the driver will manage that printer. 3025e52347bdSJani Nikula See also header of drivers/char/lp.c. 3026e52347bdSJani Nikula 3027e52347bdSJani Nikula lpj=n [KNL] 3028e52347bdSJani Nikula Sets loops_per_jiffy to given constant, thus avoiding 3029e52347bdSJani Nikula time-consuming boot-time autodetection (up to 250 ms per 3030e52347bdSJani Nikula CPU). 0 enables autodetection (default). To determine 3031e52347bdSJani Nikula the correct value for your kernel, boot with normal 3032e52347bdSJani Nikula autodetection and see what value is printed. Note that 3033e52347bdSJani Nikula on SMP systems the preset will be applied to all CPUs, 3034e52347bdSJani Nikula which is likely to cause problems if your CPUs need 3035e52347bdSJani Nikula significantly divergent settings. An incorrect value 3036e52347bdSJani Nikula will cause delays in the kernel to be wrong, leading to 3037e52347bdSJani Nikula unpredictable I/O errors and other breakage. Although 3038e52347bdSJani Nikula unlikely, in the extreme case this might damage your 3039e52347bdSJani Nikula hardware. 3040e52347bdSJani Nikula 3041e52347bdSJani Nikula ltpc= [NET] 3042e52347bdSJani Nikula Format: <io>,<irq>,<dma> 3043e52347bdSJani Nikula 30449b8c7c14SKees Cook lsm.debug [SECURITY] Enable LSM initialization debugging output. 30459b8c7c14SKees Cook 304679f7865dSKees Cook lsm=lsm1,...,lsmN 304779f7865dSKees Cook [SECURITY] Choose order of LSM initialization. This 304889a9684eSKees Cook overrides CONFIG_LSM, and the "security=" parameter. 304979f7865dSKees Cook 3050e52347bdSJani Nikula machvec= [IA-64] Force the use of a particular machine-vector 3051e52347bdSJani Nikula (machvec) in a generic kernel. 3052df43acacSChristoph Hellwig Example: machvec=hpzx1 3053e52347bdSJani Nikula 305442769488SRandy Dunlap machtype= [Loongson] Share the same kernel image file between 305542769488SRandy Dunlap different yeeloong laptops. 3056e52347bdSJani Nikula Example: machtype=lemote-yeeloong-2f-7inch 3057e52347bdSJani Nikula 305859bdbbd5SRandy Dunlap max_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory greater 3059e52347bdSJani Nikula than or equal to this physical address is ignored. 3060e52347bdSJani Nikula 3061e52347bdSJani Nikula maxcpus= [SMP] Maximum number of processors that an SMP kernel 3062e52347bdSJani Nikula will bring up during bootup. maxcpus=n : n >= 0 limits 3063e52347bdSJani Nikula the kernel to bring up 'n' processors. Surely after 3064e52347bdSJani Nikula bootup you can bring up the other plugged cpu by executing 3065e52347bdSJani Nikula "echo 1 > /sys/devices/system/cpu/cpuX/online". So maxcpus 3066e52347bdSJani Nikula only takes effect during system bootup. 3067e52347bdSJani Nikula While n=0 is a special case, it is equivalent to "nosmp", 3068e52347bdSJani Nikula which also disables the IO APIC. 3069e52347bdSJani Nikula 3070e52347bdSJani Nikula max_loop= [LOOP] The number of loop block devices that get 3071e52347bdSJani Nikula (loop.max_loop) unconditionally pre-created at init time. The default 3072e52347bdSJani Nikula number is configured by BLK_DEV_LOOP_MIN_COUNT. Instead 3073e52347bdSJani Nikula of statically allocating a predefined number, loop 3074e52347bdSJani Nikula devices can be requested on-demand with the 3075e52347bdSJani Nikula /dev/loop-control interface. 3076e52347bdSJani Nikula 3077e52347bdSJani Nikula mce [X86-32] Machine Check Exception 3078e52347bdSJani Nikula 3079ff61f079SJonathan Corbet mce=option [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst 3080e52347bdSJani Nikula 3081e52347bdSJani Nikula md= [HW] RAID subsystems devices and level 3082e52347bdSJani Nikula See Documentation/admin-guide/md.rst. 3083e52347bdSJani Nikula 3084e52347bdSJani Nikula mdacon= [MDA] 3085e52347bdSJani Nikula Format: <first>,<last> 3086e52347bdSJani Nikula Specifies range of consoles to be captured by the MDA. 3087e52347bdSJani Nikula 3088bc124170SThomas Gleixner mds= [X86,INTEL] 3089bc124170SThomas Gleixner Control mitigation for the Micro-architectural Data 3090bc124170SThomas Gleixner Sampling (MDS) vulnerability. 3091bc124170SThomas Gleixner 3092bc124170SThomas Gleixner Certain CPUs are vulnerable to an exploit against CPU 3093bc124170SThomas Gleixner internal buffers which can forward information to a 3094bc124170SThomas Gleixner disclosure gadget under certain conditions. 3095bc124170SThomas Gleixner 3096bc124170SThomas Gleixner In vulnerable processors, the speculatively 3097bc124170SThomas Gleixner forwarded data can be used in a cache side channel 3098bc124170SThomas Gleixner attack, to access data to which the attacker does 3099bc124170SThomas Gleixner not have direct access. 3100bc124170SThomas Gleixner 3101bc124170SThomas Gleixner This parameter controls the MDS mitigation. The 3102bc124170SThomas Gleixner options are: 3103bc124170SThomas Gleixner 3104bc124170SThomas Gleixner full - Enable MDS mitigation on vulnerable CPUs 3105d71eb0ceSJosh Poimboeuf full,nosmt - Enable MDS mitigation and disable 3106d71eb0ceSJosh Poimboeuf SMT on vulnerable CPUs 3107bc124170SThomas Gleixner off - Unconditionally disable MDS mitigation 3108bc124170SThomas Gleixner 310964870ed1SWaiman Long On TAA-affected machines, mds=off can be prevented by 311064870ed1SWaiman Long an active TAA mitigation as both vulnerabilities are 311164870ed1SWaiman Long mitigated with the same mechanism so in order to disable 311264870ed1SWaiman Long this mitigation, you need to specify tsx_async_abort=off 311364870ed1SWaiman Long too. 311464870ed1SWaiman Long 3115bc124170SThomas Gleixner Not specifying this option is equivalent to 3116bc124170SThomas Gleixner mds=full. 3117bc124170SThomas Gleixner 31185999bbe7SThomas Gleixner For details see: Documentation/admin-guide/hw-vuln/mds.rst 31195999bbe7SThomas Gleixner 312075c05fabSMike Rapoport mem=nn[KMG] [HEXAGON] Set the memory size. 312175c05fabSMike Rapoport Must be specified, otherwise memory size will be 0. 312275c05fabSMike Rapoport 3123e52347bdSJani Nikula mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory 3124f3cd4c86SBaoquan He Amount of memory to be used in cases as follows: 3125f3cd4c86SBaoquan He 3126f3cd4c86SBaoquan He 1 for test; 3127f3cd4c86SBaoquan He 2 when the kernel is not able to see the whole system memory; 3128f3cd4c86SBaoquan He 3 memory that lies after 'mem=' boundary is excluded from 3129f3cd4c86SBaoquan He the hypervisor, then assigned to KVM guests. 313075c05fabSMike Rapoport 4 to limit the memory available for kdump kernel. 313175c05fabSMike Rapoport 313275c05fabSMike Rapoport [ARC,MICROBLAZE] - the limit applies only to low memory, 313375c05fabSMike Rapoport high memory is not affected. 313475c05fabSMike Rapoport 313575c05fabSMike Rapoport [ARM64] - only limits memory covered by the linear 313675c05fabSMike Rapoport mapping. The NOMAP regions are not affected. 3137f3cd4c86SBaoquan He 3138e52347bdSJani Nikula [X86] Work as limiting max address. Use together 3139e52347bdSJani Nikula with memmap= to avoid physical address space collisions. 3140e52347bdSJani Nikula Without memmap= PCI devices could be placed at addresses 3141e52347bdSJani Nikula belonging to unused RAM. 3142e52347bdSJani Nikula 3143f3cd4c86SBaoquan He Note that this only takes effects during boot time since 3144f3cd4c86SBaoquan He in above case 3, memory may need be hot added after boot 3145f3cd4c86SBaoquan He if system memory of hypervisor is not sufficient. 3146f3cd4c86SBaoquan He 314775c05fabSMike Rapoport mem=nn[KMG]@ss[KMG] 314875c05fabSMike Rapoport [ARM,MIPS] - override the memory layout reported by 314975c05fabSMike Rapoport firmware. 315075c05fabSMike Rapoport Define a memory region of size nn[KMG] starting at 315175c05fabSMike Rapoport ss[KMG]. 315275c05fabSMike Rapoport Multiple different regions can be specified with 315375c05fabSMike Rapoport multiple mem= parameters on the command line. 315475c05fabSMike Rapoport 3155e52347bdSJani Nikula mem=nopentium [BUGS=X86-32] Disable usage of 4MB pages for kernel 3156e52347bdSJani Nikula memory. 3157e52347bdSJani Nikula 3158389cfd96SRandy Dunlap memblock=debug [KNL] Enable memblock debug messages. 3159389cfd96SRandy Dunlap 3160e52347bdSJani Nikula memchunk=nn[KMG] 3161e52347bdSJani Nikula [KNL,SH] Allow user to override the default size for 3162e52347bdSJani Nikula per-device physically contiguous DMA buffers. 3163e52347bdSJani Nikula 31645797f5f9SMa Wupeng memhp_default_state=online/offline/online_kernel/online_movable 3165e52347bdSJani Nikula [KNL] Set the initial state for the memory hotplug 3166e52347bdSJani Nikula onlining policy. If not specified, the default value is 3167e52347bdSJani Nikula set according to the 3168e52347bdSJani Nikula CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE kernel config 3169e52347bdSJani Nikula option. 3170cb1aaebeSMauro Carvalho Chehab See Documentation/admin-guide/mm/memory-hotplug.rst. 3171e52347bdSJani Nikula 3172e52347bdSJani Nikula memmap=exactmap [KNL,X86] Enable setting of an exact 3173e52347bdSJani Nikula E820 memory map, as specified by the user. 3174e52347bdSJani Nikula Such memmap=exactmap lines can be constructed based on 3175e52347bdSJani Nikula BIOS output or other requirements. See the memmap=nn@ss 3176e52347bdSJani Nikula option description. 3177e52347bdSJani Nikula 3178e52347bdSJani Nikula memmap=nn[KMG]@ss[KMG] 31794c8e3de4SBarry Song [KNL, X86, MIPS, XTENSA] Force usage of a specific region of memory. 3180e52347bdSJani Nikula Region of memory to be used is from ss to ss+nn. 31818fcc9bc3SBaoquan He If @ss[KMG] is omitted, it is equivalent to mem=nn[KMG], 31828fcc9bc3SBaoquan He which limits max address to nn[KMG]. 31838fcc9bc3SBaoquan He Multiple different regions can be specified, 31848fcc9bc3SBaoquan He comma delimited. 31858fcc9bc3SBaoquan He Example: 31868fcc9bc3SBaoquan He memmap=100M@2G,100M#3G,1G!1024G 3187e52347bdSJani Nikula 3188e52347bdSJani Nikula memmap=nn[KMG]#ss[KMG] 3189e52347bdSJani Nikula [KNL,ACPI] Mark specific memory as ACPI data. 3190e52347bdSJani Nikula Region of memory to be marked is from ss to ss+nn. 3191e52347bdSJani Nikula 3192e52347bdSJani Nikula memmap=nn[KMG]$ss[KMG] 3193e52347bdSJani Nikula [KNL,ACPI] Mark specific memory as reserved. 3194e52347bdSJani Nikula Region of memory to be reserved is from ss to ss+nn. 3195e52347bdSJani Nikula Example: Exclude memory from 0x18690000-0x1869ffff 3196e52347bdSJani Nikula memmap=64K$0x18690000 3197e52347bdSJani Nikula or 3198e52347bdSJani Nikula memmap=0x10000$0x18690000 31998fcc9bc3SBaoquan He Some bootloaders may need an escape character before '$', 32008fcc9bc3SBaoquan He like Grub2, otherwise '$' and the following number 32018fcc9bc3SBaoquan He will be eaten. 3202e52347bdSJani Nikula 3203e52347bdSJani Nikula memmap=nn[KMG]!ss[KMG] 3204e52347bdSJani Nikula [KNL,X86] Mark specific memory as protected. 3205e52347bdSJani Nikula Region of memory to be used, from ss to ss+nn. 3206e52347bdSJani Nikula The memory region may be marked as e820 type 12 (0xc) 3207e52347bdSJani Nikula and is NVDIMM or ADR memory. 3208e52347bdSJani Nikula 3209ef61f8a3SJan H. Schönherr memmap=<size>%<offset>-<oldtype>+<newtype> 3210ef61f8a3SJan H. Schönherr [KNL,ACPI] Convert memory within the specified region 3211ef61f8a3SJan H. Schönherr from <oldtype> to <newtype>. If "-<oldtype>" is left 3212ef61f8a3SJan H. Schönherr out, the whole region will be marked as <newtype>, 3213ef61f8a3SJan H. Schönherr even if previously unavailable. If "+<newtype>" is left 3214ef61f8a3SJan H. Schönherr out, matching memory will be removed. Types are 3215ef61f8a3SJan H. Schönherr specified as e820 types, e.g., 1 = RAM, 2 = reserved, 3216ef61f8a3SJan H. Schönherr 3 = ACPI, 12 = PRAM. 3217ef61f8a3SJan H. Schönherr 3218e52347bdSJani Nikula memory_corruption_check=0/1 [X86] 3219e52347bdSJani Nikula Some BIOSes seem to corrupt the first 64k of 3220e52347bdSJani Nikula memory when doing things like suspend/resume. 3221e52347bdSJani Nikula Setting this option will scan the memory 3222e52347bdSJani Nikula looking for corruption. Enabling this will 3223e52347bdSJani Nikula both detect corruption and prevent the kernel 3224e52347bdSJani Nikula from using the memory being corrupted. 3225e52347bdSJani Nikula However, its intended as a diagnostic tool; if 3226e52347bdSJani Nikula repeatable BIOS-originated corruption always 3227e52347bdSJani Nikula affects the same memory, you can use memmap= 3228e52347bdSJani Nikula to prevent the kernel from using that memory. 3229e52347bdSJani Nikula 3230e52347bdSJani Nikula memory_corruption_check_size=size [X86] 3231e52347bdSJani Nikula By default it checks for corruption in the low 3232e52347bdSJani Nikula 64k, making this memory unavailable for normal 3233e52347bdSJani Nikula use. Use this parameter to scan for 3234e52347bdSJani Nikula corruption in more or less memory. 3235e52347bdSJani Nikula 3236e52347bdSJani Nikula memory_corruption_check_period=seconds [X86] 3237e52347bdSJani Nikula By default it checks for corruption every 60 3238e52347bdSJani Nikula seconds. Use this parameter to check at some 3239e52347bdSJani Nikula other rate. 0 disables periodic checking. 3240e52347bdSJani Nikula 3241e3a9d9fcSOscar Salvador memory_hotplug.memmap_on_memory 3242e3a9d9fcSOscar Salvador [KNL,X86,ARM] Boolean flag to enable this feature. 3243e3a9d9fcSOscar Salvador Format: {on | off (default)} 3244e3a9d9fcSOscar Salvador When enabled, runtime hotplugged memory will 324566361095SMuchun Song allocate its internal metadata (struct pages, 324666361095SMuchun Song those vmemmap pages cannot be optimized even 324766361095SMuchun Song if hugetlb_free_vmemmap is enabled) from the 324866361095SMuchun Song hotadded memory which will allow to hotadd a 324966361095SMuchun Song lot of memory without requiring additional 325066361095SMuchun Song memory to do so. 3251e3a9d9fcSOscar Salvador This feature is disabled by default because it 3252e3a9d9fcSOscar Salvador has some implication on large (e.g. GB) 3253e3a9d9fcSOscar Salvador allocations in some configurations (e.g. small 3254e3a9d9fcSOscar Salvador memory blocks). 3255e3a9d9fcSOscar Salvador The state of the flag can be read in 3256e3a9d9fcSOscar Salvador /sys/module/memory_hotplug/parameters/memmap_on_memory. 3257e3a9d9fcSOscar Salvador Note that even when enabled, there are a few cases where 3258e3a9d9fcSOscar Salvador the feature is not effective. 3259e3a9d9fcSOscar Salvador 3260376e3fdeSFinn Thain memtest= [KNL,X86,ARM,M68K,PPC,RISCV] Enable memtest 3261e52347bdSJani Nikula Format: <integer> 3262e52347bdSJani Nikula default : 0 <disable> 3263e52347bdSJani Nikula Specifies the number of memtest passes to be 3264e52347bdSJani Nikula performed. Each pass selects another test 3265e52347bdSJani Nikula pattern from a given set of patterns. Memtest 3266e52347bdSJani Nikula fills the memory with this pattern, validates 3267e52347bdSJani Nikula memory contents and reserves bad memory 3268e52347bdSJani Nikula regions that are detected. 3269e52347bdSJani Nikula 3270c262f3b9STom Lendacky mem_encrypt= [X86-64] AMD Secure Memory Encryption (SME) control 3271c262f3b9STom Lendacky Valid arguments: on, off 3272ecd16da3SBorislav Petkov (AMD) Default: off 3273c262f3b9STom Lendacky mem_encrypt=on: Activate SME 3274c262f3b9STom Lendacky mem_encrypt=off: Do not activate SME 3275c262f3b9STom Lendacky 32767ac3945dSMauro Carvalho Chehab Refer to Documentation/virt/kvm/x86/amd-memory-encryption.rst 3277c262f3b9STom Lendacky for details on when memory encryption can be activated. 3278c262f3b9STom Lendacky 32797b9dc3f7SLinus Torvalds mem_sleep_default= [SUSPEND] Default system suspend mode: 32807b9dc3f7SLinus Torvalds s2idle - Suspend-To-Idle 32817b9dc3f7SLinus Torvalds shallow - Power-On Suspend or equivalent (if supported) 32827b9dc3f7SLinus Torvalds deep - Suspend-To-RAM or equivalent (if supported) 328358e7cb9eSRafael J. Wysocki See Documentation/admin-guide/pm/sleep-states.rst. 32847b9dc3f7SLinus Torvalds 3285e52347bdSJani Nikula mem.devmem= Activate the /dev/mem device 3286e52347bdSJani Nikula Format: <bool> (1/Y/y=enable, 0/N/n=disable) 3287e52347bdSJani Nikula 3288e52347bdSJani Nikula mfgpt_irq= [IA-32] Specify the IRQ to use for the 3289e52347bdSJani Nikula Multi-Function General Purpose Timers on AMD Geode 3290e52347bdSJani Nikula platforms. 3291e52347bdSJani Nikula 3292e52347bdSJani Nikula mfgptfix [X86-32] Fix MFGPT timers on AMD Geode platforms when 3293e52347bdSJani Nikula the BIOS has incorrectly applied a workaround. TinyBIOS 3294e52347bdSJani Nikula version 0.98 is known to be affected, 0.99 fixes the 3295e52347bdSJani Nikula problem by letting the user disable the workaround. 329659bdbbd5SRandy Dunlap 3297e52347bdSJani Nikula mga= [HW,DRM] 3298e52347bdSJani Nikula 3299e52347bdSJani Nikula min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this 3300e52347bdSJani Nikula physical address is ignored. 3301e52347bdSJani Nikula 3302e52347bdSJani Nikula mini2440= [ARM,HW,KNL] 3303e52347bdSJani Nikula Format:[0..2][b][c][t] 3304e52347bdSJani Nikula Default: "0tb" 3305e52347bdSJani Nikula MINI2440 configuration specification: 3306e52347bdSJani Nikula 0 - The attached screen is the 3.5" TFT 3307e52347bdSJani Nikula 1 - The attached screen is the 7" TFT 3308e52347bdSJani Nikula 2 - The VGA Shield is attached (1024x768) 3309e52347bdSJani Nikula Leaving out the screen size parameter will not load 3310e52347bdSJani Nikula the TFT driver, and the framebuffer will be left 3311e52347bdSJani Nikula unconfigured. 3312e52347bdSJani Nikula b - Enable backlight. The TFT backlight pin will be 3313e52347bdSJani Nikula linked to the kernel VESA blanking code and a GPIO 3314e52347bdSJani Nikula LED. This parameter is not necessary when using the 3315e52347bdSJani Nikula VGA shield. 3316e52347bdSJani Nikula c - Enable the s3c camera interface. 3317e52347bdSJani Nikula t - Reserved for enabling touchscreen support. The 33186b2484e1SAlexander A. Klimov touchscreen support is not enabled in the mainstream 3319e52347bdSJani Nikula kernel as of 2.6.30, a preliminary port can be found 332098af8452SJosh Poimboeuf in the "bleeding edge" mini2440 support kernel at 3321a111b7c0SJosh Poimboeuf https://repo.or.cz/w/linux-2.6/mini2440.git 3322a111b7c0SJosh Poimboeuf 3323d68be4c4SJosh Poimboeuf mitigations= 3324d68be4c4SJosh Poimboeuf [X86,PPC,S390,ARM64] Control optional mitigations for 332598af8452SJosh Poimboeuf CPU vulnerabilities. This is a set of curated, 3326976b74faSSean Christopherson arch-independent options, each of which is an 3327976b74faSSean Christopherson aggregation of existing arch-specific options. 3328976b74faSSean Christopherson 332998af8452SJosh Poimboeuf Note, "mitigations" is supported if and only if the 333098af8452SJosh Poimboeuf kernel was built with CPU_MITIGATIONS=y. 333198af8452SJosh Poimboeuf 333298af8452SJosh Poimboeuf off 33338974eb58SDaniel Sneddon Disable all optional CPU mitigations. This 33348974eb58SDaniel Sneddon improves system performance, but it may also 33358974eb58SDaniel Sneddon expose users to several CPU vulnerabilities. 3336d68be4c4SJosh Poimboeuf Equivalent to: if nokaslr then kpti=0 [ARM64] 33375c14068fSJosh Poimboeuf gather_data_sampling=off [X86] 33388974eb58SDaniel Sneddon kvm.nx_huge_pages=off [X86] 3339f7964378SNicholas Piggin l1tf=off [X86] 33409a32a7e7SNicholas Piggin mds=off [X86] 33418974eb58SDaniel Sneddon mmio_stale_data=off [X86] 33428974eb58SDaniel Sneddon no_entry_flush [PPC] 33438974eb58SDaniel Sneddon no_uaccess_flush [PPC] 33448974eb58SDaniel Sneddon nobp=0 [S390] 33458974eb58SDaniel Sneddon nopti [X86,PPC] 334677018fb9SPawan Gupta nospectre_bhb [ARM64] 3347ea304a8bSEiichi Tsukata nospectre_v1 [X86,PPC] 33488974eb58SDaniel Sneddon nospectre_v2 [X86,PPC,S390,ARM64] 334939053a34SJosh Poimboeuf reg_file_data_sampling=off [X86] 33508974eb58SDaniel Sneddon retbleed=off [X86] 33518974eb58SDaniel Sneddon spec_store_bypass_disable=off [X86,PPC] 33528974eb58SDaniel Sneddon spectre_bhi=off [X86] 33538974eb58SDaniel Sneddon spectre_v2_user=off [X86] 3354b8e8c830SPaolo Bonzini srbds=off [X86,INTEL] 3355b8e8c830SPaolo Bonzini ssbd=force-off [ARM64] 3356b8e8c830SPaolo Bonzini tsx_async_abort=off [X86] 3357b8e8c830SPaolo Bonzini 3358b8e8c830SPaolo Bonzini Exceptions: 335998af8452SJosh Poimboeuf This does not have any effect on 336098af8452SJosh Poimboeuf kvm.nx_huge_pages when 336198af8452SJosh Poimboeuf kvm.nx_huge_pages=force. 336298af8452SJosh Poimboeuf 336398af8452SJosh Poimboeuf auto (default) 336498af8452SJosh Poimboeuf Mitigate all CPU vulnerabilities, but leave SMT 336598af8452SJosh Poimboeuf enabled, even if it's vulnerable. This is for 3366d68be4c4SJosh Poimboeuf users who don't want to be surprised by SMT 336798af8452SJosh Poimboeuf getting disabled across kernel upgrades, or who 336898af8452SJosh Poimboeuf have other ways of avoiding SMT-based attacks. 336998af8452SJosh Poimboeuf Equivalent to: (default behavior) 337098af8452SJosh Poimboeuf 337198af8452SJosh Poimboeuf auto,nosmt 3372d68be4c4SJosh Poimboeuf Mitigate all CPU vulnerabilities, disabling SMT 33735c14068fSJosh Poimboeuf if needed. This is for users who always want to 3374a7a248c5SPawan Gupta be fully mitigated, even if it means losing SMT. 33758cb861e9SPawan Gupta Equivalent to: l1tf=flush,nosmt [X86] 3376ea304a8bSEiichi Tsukata mds=full,nosmt [X86] 337798af8452SJosh Poimboeuf tsx_async_abort=full,nosmt [X86] 3378e52347bdSJani Nikula mmio_stale_data=full,nosmt [X86] 3379e52347bdSJani Nikula retbleed=auto,nosmt [X86] 3380e52347bdSJani Nikula 3381e52347bdSJani Nikula mminit_loglevel= 3382e52347bdSJani Nikula [KNL] When CONFIG_DEBUG_MEMORY_INIT is set, this 3383e52347bdSJani Nikula parameter allows control of the logging verbosity for 3384e52347bdSJani Nikula the additional memory initialisation checks. A value 3385e52347bdSJani Nikula of 0 disables mminit logging and a level of 4 will 33868cb861e9SPawan Gupta log everything. Information is printed at KERN_DEBUG 33878cb861e9SPawan Gupta so loglevel=8 may also need to be specified. 33888cb861e9SPawan Gupta 33898cb861e9SPawan Gupta mmio_stale_data= 33908cb861e9SPawan Gupta [X86,INTEL] Control mitigation for the Processor 33918cb861e9SPawan Gupta MMIO Stale Data vulnerabilities. 33928cb861e9SPawan Gupta 33938cb861e9SPawan Gupta Processor MMIO Stale Data is a class of 33948cb861e9SPawan Gupta vulnerabilities that may expose data after an MMIO 33958cb861e9SPawan Gupta operation. Exposed data could originate or end in 33968cb861e9SPawan Gupta the same CPU buffers as affected by MDS and TAA. 33978cb861e9SPawan Gupta Therefore, similar to MDS and TAA, the mitigation 33988cb861e9SPawan Gupta is to clear the affected CPU buffers. 33998cb861e9SPawan Gupta 34008cb861e9SPawan Gupta This parameter controls the mitigation. The 34018cb861e9SPawan Gupta options are: 34028cb861e9SPawan Gupta 34038cb861e9SPawan Gupta full - Enable mitigation on vulnerable CPUs 34048cb861e9SPawan Gupta 34058cb861e9SPawan Gupta full,nosmt - Enable mitigation and disable SMT on 34068cb861e9SPawan Gupta vulnerable CPUs. 34078cb861e9SPawan Gupta 34088cb861e9SPawan Gupta off - Unconditionally disable mitigation 34098cb861e9SPawan Gupta 34108cb861e9SPawan Gupta On MDS or TAA affected machines, 34118cb861e9SPawan Gupta mmio_stale_data=off can be prevented by an active 34128cb861e9SPawan Gupta MDS or TAA mitigation as these vulnerabilities are 34138cb861e9SPawan Gupta mitigated with the same mechanism so in order to 34148cb861e9SPawan Gupta disable this mitigation, you need to specify 34158cb861e9SPawan Gupta mds=off and tsx_async_abort=off too. 34168cb861e9SPawan Gupta 34178cb861e9SPawan Gupta Not specifying this option is equivalent to 34188cb861e9SPawan Gupta mmio_stale_data=full. 34198cb861e9SPawan Gupta 342042551b8dSRandy Dunlap For details see: 342142551b8dSRandy Dunlap Documentation/admin-guide/hw-vuln/processor_mmio_stale_data.rst 342242551b8dSRandy Dunlap 342342551b8dSRandy Dunlap <module>.async_probe[=<bool>] [KNL] 342442551b8dSRandy Dunlap If no <bool> value is specified or if the value 342542551b8dSRandy Dunlap specified is not a valid <bool>, enable asynchronous 342642551b8dSRandy Dunlap probe on this module. Otherwise, enable/disable 3427ae39e9edSSaravana Kannan asynchronous probe on this module as indicated by the 3428ae39e9edSSaravana Kannan <bool> value. See also: module.async_probe 3429ae39e9edSSaravana Kannan 3430ae39e9edSSaravana Kannan module.async_probe=<bool> 3431ae39e9edSSaravana Kannan [KNL] When set to true, modules will use async probing 3432ae39e9edSSaravana Kannan by default. To enable/disable async probing for a 3433ae39e9edSSaravana Kannan specific module, use the module specific control that 3434ae39e9edSSaravana Kannan is documented under <module>.async_probe. When both 3435ae39e9edSSaravana Kannan module.async_probe and <module>.async_probe are 34368660484eSLuis Chamberlain specified, <module>.async_probe takes precedence for 34378660484eSLuis Chamberlain the specific module. 34388660484eSLuis Chamberlain 34398660484eSLuis Chamberlain module.enable_dups_trace 34408660484eSLuis Chamberlain [KNL] When CONFIG_MODULE_DEBUG_AUTOLOAD_DUPS is set, 34418660484eSLuis Chamberlain this means that duplicate request_module() calls will 3442e52347bdSJani Nikula trigger a WARN_ON() instead of a pr_warn(). Note that 3443e52347bdSJani Nikula if MODULE_DEBUG_AUTOLOAD_DUPS_TRACE is set, WARN_ON()s 3444e52347bdSJani Nikula will always be issued and this option does nothing. 3445e52347bdSJani Nikula module.sig_enforce 3446e52347bdSJani Nikula [KNL] When CONFIG_MODULE_SIG is set, this means that 3447e52347bdSJani Nikula modules without (valid) signatures will fail to load. 3448e52347bdSJani Nikula Note that if CONFIG_MODULE_SIG_FORCE is set, that 3449e52347bdSJani Nikula is always true, so this option does nothing. 3450e52347bdSJani Nikula 3451e52347bdSJani Nikula module_blacklist= [KNL] Do not load a comma-separated list of 3452e52347bdSJani Nikula modules. Useful for debugging problem modules. 3453e52347bdSJani Nikula 3454e52347bdSJani Nikula mousedev.tap_time= 3455e52347bdSJani Nikula [MOUSE] Maximum time between finger touching and 3456e52347bdSJani Nikula leaving touchpad surface for touch to be considered 3457e52347bdSJani Nikula a tap and be reported as a left button click (for 3458e52347bdSJani Nikula touchpads working in absolute mode only). 3459e52347bdSJani Nikula Format: <msecs> 3460e52347bdSJani Nikula mousedev.xres= [MOUSE] Horizontal screen resolution, used for devices 3461e52347bdSJani Nikula reporting absolute coordinates, such as tablets 3462a5c6d650SDavid Rientjes mousedev.yres= [MOUSE] Vertical screen resolution, used for devices 3463a5c6d650SDavid Rientjes reporting absolute coordinates, such as tablets 3464a5c6d650SDavid Rientjes 3465a5c6d650SDavid Rientjes movablecore= [KNL,X86,IA-64,PPC] 3466a5c6d650SDavid Rientjes Format: nn[KMGTPE] | nn% 3467a5c6d650SDavid Rientjes This parameter is the complement to kernelcore=, it 3468a5c6d650SDavid Rientjes specifies the amount of memory used for migratable 3469a5c6d650SDavid Rientjes allocations. If both kernelcore and movablecore is 3470e52347bdSJani Nikula specified, then kernelcore will be at *least* the 3471e52347bdSJani Nikula specified value but may be more. If movablecore on its 3472e52347bdSJani Nikula own is specified, the administrator must be careful 3473f70029bbSMichal Hocko that the amount of memory usable for all allocations 3474f70029bbSMichal Hocko is not too small. 3475f70029bbSMichal Hocko 3476f70029bbSMichal Hocko movable_node [KNL] Boot-time switch to make hotplugable memory 3477f70029bbSMichal Hocko NUMA nodes to be movable. This means that the memory 3478e52347bdSJani Nikula of such nodes will be usable only for movable 3479e52347bdSJani Nikula allocations which rules out almost all kernel 3480e52347bdSJani Nikula allocations. Use with caution! 3481e52347bdSJani Nikula 3482e52347bdSJani Nikula MTD_Partition= [MTD] 3483e52347bdSJani Nikula Format: <name>,<region-number>,<size>,<offset> 3484e52347bdSJani Nikula 3485e52347bdSJani Nikula MTD_Region= [MTD] Format: 3486fb251124SJonathan Neuschäfer <name>,<region-number>[,<base>,<size>,<buswidth>,<altbuswidth>] 3487e52347bdSJani Nikula 3488e52347bdSJani Nikula mtdparts= [MTD] 3489e52347bdSJani Nikula See drivers/mtd/parsers/cmdlinepart.c 3490e52347bdSJani Nikula 34910f12999eSKrzysztof Kozlowski mtdset= [ARM] 3492e52347bdSJani Nikula ARM/S3C2412 JIVE boot control 3493e52347bdSJani Nikula 3494e52347bdSJani Nikula See arch/arm/mach-s3c/mach-jive.c 3495e52347bdSJani Nikula 3496e52347bdSJani Nikula mtouchusb.raw_coordinates= 3497a4316603SJuergen Gross [HW] Make the MicroTouch USB driver use raw coordinates 3498a4316603SJuergen Gross ('y', default) or cooked coordinates ('n') 3499a4316603SJuergen Gross 3500a4316603SJuergen Gross mtrr=debug [X86] 3501e52347bdSJani Nikula Enable printing debug information related to MTRR 3502e52347bdSJani Nikula registers at boot time. 3503e52347bdSJani Nikula 3504e52347bdSJani Nikula mtrr_chunk_size=nn[KMG] [X86] 3505e52347bdSJani Nikula used for mtrr cleanup. It is largest continuous chunk 3506e52347bdSJani Nikula that could hold holes aka. UC entries. 3507e52347bdSJani Nikula 3508e52347bdSJani Nikula mtrr_gran_size=nn[KMG] [X86] 3509e52347bdSJani Nikula Used for mtrr cleanup. It is granularity of mtrr block. 3510e52347bdSJani Nikula Default is 1. 3511e52347bdSJani Nikula Large value could prevent small alignment from 3512e52347bdSJani Nikula using up MTRRs. 3513e52347bdSJani Nikula 3514e52347bdSJani Nikula mtrr_spare_reg_nr=n [X86] 3515e52347bdSJani Nikula Format: <integer> 3516e52347bdSJani Nikula Range: 0,7 : spare reg number 3517e52347bdSJani Nikula Default : 1 3518389cfd96SRandy Dunlap Used for mtrr cleanup. It is spare mtrr entries number. 3519389cfd96SRandy Dunlap Set to 2 or more if your graphical card needs more. 3520389cfd96SRandy Dunlap 3521389cfd96SRandy Dunlap multitce=off [PPC] This parameter disables the use of the pSeries 3522e52347bdSJani Nikula firmware feature for updating multiple TCE entries 3523e52347bdSJani Nikula at a time. 3524e52347bdSJani Nikula 3525e52347bdSJani Nikula n2= [NET] SDL Inc. RISCom/N2 synchronous serial card 3526e52347bdSJani Nikula 3527e52347bdSJani Nikula netdev= [NET] Network devices parameters 3528e52347bdSJani Nikula Format: <irq>,<io>,<mem_start>,<mem_end>,<name> 3529e52347bdSJani Nikula Note that mem_start is often overloaded to mean 3530e52347bdSJani Nikula something different and driver-specific. 3531389cfd96SRandy Dunlap This usage is only documented in each driver source 3532389cfd96SRandy Dunlap file if at all. 3533389cfd96SRandy Dunlap 3534389cfd96SRandy Dunlap netpoll.carrier_timeout= 3535389cfd96SRandy Dunlap [NET] Specifies amount of time (in seconds) that 3536e52347bdSJani Nikula netpoll should wait for a carrier. By default netpoll 3537e52347bdSJani Nikula waits 4 seconds. 3538e52347bdSJani Nikula 3539e52347bdSJani Nikula nf_conntrack.acct= 3540e52347bdSJani Nikula [NETFILTER] Enable connection tracking flow accounting 3541e52347bdSJani Nikula 0 to disable accounting 3542c500488fSRandy Dunlap 1 to enable accounting 3543c500488fSRandy Dunlap Default value is 0. 3544c500488fSRandy Dunlap 3545e52347bdSJani Nikula nfs.cache_getent= 3546c500488fSRandy Dunlap [NFS] sets the pathname to the program which is used 3547c500488fSRandy Dunlap to update the NFS client cache entries. 3548c500488fSRandy Dunlap 3549e52347bdSJani Nikula nfs.cache_getent_timeout= 3550e52347bdSJani Nikula [NFS] sets the timeout after which an attempt to 3551e52347bdSJani Nikula update a cache entry is deemed to have failed. 3552e52347bdSJani Nikula 3553e52347bdSJani Nikula nfs.callback_nr_threads= 3554e52347bdSJani Nikula [NFSv4] set the total number of threads that the 3555e52347bdSJani Nikula NFS client will assign to service NFSv4 callback 3556e52347bdSJani Nikula requests. 3557e52347bdSJani Nikula 3558e52347bdSJani Nikula nfs.callback_tcpport= 3559e52347bdSJani Nikula [NFS] set the TCP port on which the NFSv4 callback 3560e52347bdSJani Nikula channel should listen. 3561e52347bdSJani Nikula 3562e52347bdSJani Nikula nfs.enable_ino64= 3563e52347bdSJani Nikula [NFS] enable 64-bit inode numbers. 3564e52347bdSJani Nikula If zero, the NFS client will fake up a 32-bit inode 3565e52347bdSJani Nikula number for the readdir() and stat() syscalls instead 3566c500488fSRandy Dunlap of returning the full 64-bit number. 3567c500488fSRandy Dunlap The default is to return 64-bit inode numbers. 3568c500488fSRandy Dunlap 3569c500488fSRandy Dunlap nfs.idmap_cache_timeout= 3570e52347bdSJani Nikula [NFS] set the maximum lifetime for idmapper cache 3571e52347bdSJani Nikula entries. 3572e52347bdSJani Nikula 3573e52347bdSJani Nikula nfs.max_session_cb_slots= 3574e52347bdSJani Nikula [NFSv4.1] Sets the maximum number of session 3575e52347bdSJani Nikula slots the client will assign to the callback 3576e52347bdSJani Nikula channel. This determines the maximum number of 3577e52347bdSJani Nikula callbacks the client will process in parallel for 3578e52347bdSJani Nikula a particular server. 3579e52347bdSJani Nikula 3580e52347bdSJani Nikula nfs.max_session_slots= 3581e52347bdSJani Nikula [NFSv4.1] Sets the maximum number of session slots 3582e52347bdSJani Nikula the client will attempt to negotiate with the server. 3583e52347bdSJani Nikula This limits the number of simultaneous RPC requests 3584e52347bdSJani Nikula that the client can send to the NFSv4.1 server. 3585e52347bdSJani Nikula Note that there is little point in setting this 3586e52347bdSJani Nikula value higher than the max_tcp_slot_table_limit. 3587e52347bdSJani Nikula 3588e52347bdSJani Nikula nfs.nfs4_disable_idmapping= 3589e52347bdSJani Nikula [NFSv4] When set to the default of '1', this option 3590e52347bdSJani Nikula ensures that both the RPC level authentication 3591e52347bdSJani Nikula scheme and the NFS level operations agree to use 3592e52347bdSJani Nikula numeric uids/gids if the mount is using the 3593e52347bdSJani Nikula 'sec=sys' security flavour. In effect it is 3594e52347bdSJani Nikula disabling idmapping, which can make migration from 3595e52347bdSJani Nikula legacy NFSv2/v3 systems to NFSv4 easier. 3596e52347bdSJani Nikula Servers that do not support this mode of operation 3597c500488fSRandy Dunlap will be autodetected by the client, and it will fall 3598e52347bdSJani Nikula back to using the idmapper. 3599e52347bdSJani Nikula To turn off this behaviour, set the value to '0'. 3600e52347bdSJani Nikula 3601e52347bdSJani Nikula nfs.nfs4_unique_id= 3602e52347bdSJani Nikula [NFS4] Specify an additional fixed unique ident- 3603e52347bdSJani Nikula ification string that NFSv4 clients can insert into 3604e52347bdSJani Nikula their nfs_client_id4 string. This is typically a 3605e52347bdSJani Nikula UUID that is generated at system install time. 3606e52347bdSJani Nikula 3607e52347bdSJani Nikula nfs.recover_lost_locks= 3608e52347bdSJani Nikula [NFSv4] Attempt to recover locks that were lost due 3609e52347bdSJani Nikula to a lease timeout on the server. Please note that 3610e52347bdSJani Nikula doing this risks data corruption, since there are 3611e52347bdSJani Nikula no guarantees that the file will remain unchanged 3612e52347bdSJani Nikula after the locks are lost. 3613e52347bdSJani Nikula If you want to enable the kernel legacy behaviour of 3614e52347bdSJani Nikula attempting to recover these locks, then set this 3615e52347bdSJani Nikula parameter to '1'. 3616c500488fSRandy Dunlap The default parameter value of '0' causes the kernel 3617c500488fSRandy Dunlap not to attempt recovery of lost locks. 3618c500488fSRandy Dunlap 3619c500488fSRandy Dunlap nfs.send_implementation_id= 3620c500488fSRandy Dunlap [NFSv4.1] Send client implementation identification 3621c500488fSRandy Dunlap information in exchange_id requests. 3622c500488fSRandy Dunlap If zero, no implementation identification information 3623c500488fSRandy Dunlap will be sent. 3624e52347bdSJani Nikula The default is to send the implementation identification 3625e52347bdSJani Nikula information. 3626e52347bdSJani Nikula 3627e52347bdSJani Nikula nfs4.layoutstats_timer= 3628e52347bdSJani Nikula [NFSv4.2] Change the rate at which the kernel sends 3629e52347bdSJani Nikula layoutstats to the pNFS metadata server. 3630e52347bdSJani Nikula 3631e52347bdSJani Nikula Setting this to value to 0 causes the kernel to use 3632e52347bdSJani Nikula whatever value is the default set by the layout 36336d91929aSJ. Bruce Fields driver. A non-zero value sets the minimum interval 36346d91929aSJ. Bruce Fields in seconds between layoutstats transmissions. 36356d91929aSJ. Bruce Fields 36366d91929aSJ. Bruce Fields nfsd.inter_copy_offload_enable= 36376d91929aSJ. Bruce Fields [NFSv4.2] When set to 1, the server will support 3638c500488fSRandy Dunlap server-to-server copies for which this server is 3639c500488fSRandy Dunlap the destination of the copy. 3640c500488fSRandy Dunlap 3641c500488fSRandy Dunlap nfsd.nfs4_disable_idmapping= 3642c500488fSRandy Dunlap [NFSv4] When set to the default of '1', the NFSv4 3643c500488fSRandy Dunlap server will return only numeric uids and gids to 3644c500488fSRandy Dunlap clients using auth_sys, and will accept numeric uids 36456d91929aSJ. Bruce Fields and gids from such clients. This is intended to ease 36466d91929aSJ. Bruce Fields migration from NFSv2/v3. 36476d91929aSJ. Bruce Fields 36486d91929aSJ. Bruce Fields nfsd.nfsd4_ssc_umount_timeout= 36496d91929aSJ. Bruce Fields [NFSv4.2] When used as the destination of a 36506d91929aSJ. Bruce Fields server-to-server copy, knfsd temporarily mounts 36516d91929aSJ. Bruce Fields the source server. It caches the mount in case 36526d91929aSJ. Bruce Fields it will be needed again, and discards it if not 3653c500488fSRandy Dunlap used for the number of milliseconds specified by 3654c500488fSRandy Dunlap this parameter. 3655e52347bdSJani Nikula 3656c500488fSRandy Dunlap nfsaddrs= [NFS] Deprecated. Use ip= instead. 3657c500488fSRandy Dunlap See Documentation/admin-guide/nfs/nfsroot.rst. 3658c500488fSRandy Dunlap 3659c500488fSRandy Dunlap nfsroot= [NFS] nfs root filesystem for disk-less boxes. 3660c500488fSRandy Dunlap See Documentation/admin-guide/nfs/nfsroot.rst. 36616d91929aSJ. Bruce Fields 3662160c7ba3SPaul E. McKenney nfsrootdebug [NFS] enable nfsroot debugging messages. 3663160c7ba3SPaul E. McKenney See Documentation/admin-guide/nfs/nfsroot.rst. 3664160c7ba3SPaul E. McKenney 3665160c7ba3SPaul E. McKenney nmi_backtrace.backtrace_idle [KNL] 3666c0c74acbSHans-Christian Noren Egtvedt Dump stacks even of idle CPUs in response to an 3667e52347bdSJani Nikula NMI stack-backtrace request. 3668e52347bdSJani Nikula 3669e52347bdSJani Nikula nmi_debug= [KNL,SH] Specify one or more actions to take 3670e52347bdSJani Nikula when a NMI is triggered. 3671e52347bdSJani Nikula Format: [state][,regs][,debounce][,die] 3672e52347bdSJani Nikula 3673e52347bdSJani Nikula nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels 3674e52347bdSJani Nikula Format: [panic,][nopanic,][num] 3675e52347bdSJani Nikula Valid num: 0 or 1 367693285c01SZhenzhong Duan 0 - turn hardlockup detector in nmi_watchdog off 367793285c01SZhenzhong Duan 1 - turn hardlockup detector in nmi_watchdog on 367893285c01SZhenzhong Duan When panic is specified, panic when an NMI watchdog 3679e52347bdSJani Nikula timeout occurs (or 'nopanic' to not panic on an NMI 3680e52347bdSJani Nikula watchdog, if CONFIG_BOOTPARAM_HARDLOCKUP_PANIC is set) 3681e52347bdSJani Nikula To disable both hard and soft lockup detectors, 3682e52347bdSJani Nikula please see 'nowatchdog'. 3683d22881dcSScott Wood This is useful when you use a panic=... timeout and 3684d22881dcSScott Wood need the box quickly up again. 3685d22881dcSScott Wood 3686e52347bdSJani Nikula These settings can be accessed at runtime via 3687e52347bdSJani Nikula the nmi_watchdog and hardlockup_panic sysctls. 3688e52347bdSJani Nikula 3689e52347bdSJani Nikula no387 [BUGS=X86-32] Tells the kernel to use the 387 maths 369026e7aacbSAlexandre Ghiti emulation library even if a 387 maths coprocessor 369126e7aacbSAlexandre Ghiti is present. 369226e7aacbSAlexandre Ghiti 369326e7aacbSAlexandre Ghiti no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces 3694372fddf7SKirill A. Shutemov kernel to use 3-level paging instead. 3695372fddf7SKirill A. Shutemov 3696e52347bdSJani Nikula no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces 3697e52347bdSJani Nikula kernel to use 4-level paging instead. 3698e52347bdSJani Nikula 3699e52347bdSJani Nikula noaliencache [MM, NUMA, SLAB] Disables the allocation of alien 3700e52347bdSJani Nikula caches in the slab allocator. Saves per-node memory, 3701e52347bdSJani Nikula but will impact performance. 3702686140a1SVasily Gorbik 3703686140a1SVasily Gorbik noalign [KNL,ARM] 3704686140a1SVasily Gorbik 3705e52347bdSJani Nikula noaltinstr [S390] Disables alternative instructions patching 3706e52347bdSJani Nikula (CPU alternatives feature). 3707e52347bdSJani Nikula 3708e52347bdSJani Nikula noapic [SMP,APIC] Tells the kernel to not make use of any 3709e52347bdSJani Nikula IOAPICs that may be present in the system. 3710e52347bdSJani Nikula 3711e52347bdSJani Nikula noautogroup Disable scheduler automatic task group creation. 3712a894a8a5SRandy Dunlap 3713a894a8a5SRandy Dunlap nocache [ARM] 3714a894a8a5SRandy Dunlap 3715a894a8a5SRandy Dunlap no_console_suspend 3716a894a8a5SRandy Dunlap [HW] Never suspend the console 3717a894a8a5SRandy Dunlap Disable suspending of consoles during suspend and 3718a894a8a5SRandy Dunlap hibernate operations. Once disabled, debugging 3719a894a8a5SRandy Dunlap messages can reach various consoles while the rest 3720a894a8a5SRandy Dunlap of the system is being put to sleep (ie, while 3721a894a8a5SRandy Dunlap debugging driver suspend/resume hooks). This may 3722a894a8a5SRandy Dunlap not work reliably with all consoles, but is known 3723a894a8a5SRandy Dunlap to work with serial and VGA consoles. 3724a894a8a5SRandy Dunlap To facilitate more flexible debugging, we also add 3725a894a8a5SRandy Dunlap console_suspend, a printk module parameter to control 3726a894a8a5SRandy Dunlap it. Users could use console_suspend (usually 3727a894a8a5SRandy Dunlap /sys/module/printk/parameters/console_suspend) to 3728a894a8a5SRandy Dunlap turn on/off it dynamically. 3729a894a8a5SRandy Dunlap 3730e52347bdSJani Nikula no_debug_objects 3731e52347bdSJani Nikula [KNL] Disable object debugging 3732e52347bdSJani Nikula 3733e52347bdSJani Nikula nodsp [SH] Disable hardware DSP at boot time. 3734f7964378SNicholas Piggin 3735f7964378SNicholas Piggin noefi Disable EFI runtime services support. 3736e52347bdSJani Nikula 3737e52347bdSJani Nikula no_entry_flush [PPC] Don't flush the L1-D cache when entering the kernel. 3738e52347bdSJani Nikula 3739e52347bdSJani Nikula noexec [IA-64] 3740e52347bdSJani Nikula 3741e52347bdSJani Nikula noexec32 [X86-64] 3742e52347bdSJani Nikula This affects only 32-bit executables. 3743e52347bdSJani Nikula noexec32=on: enable non-executable mappings (default) 3744e52347bdSJani Nikula read doesn't imply executable mappings 3745a894a8a5SRandy Dunlap noexec32=off: disable non-executable mappings 3746a894a8a5SRandy Dunlap read implies executable mappings 3747a894a8a5SRandy Dunlap 3748a894a8a5SRandy Dunlap no_file_caps Tells the kernel not to honor file capabilities. The 3749e52347bdSJani Nikula only way then for a file to be executed with privilege 3750e52347bdSJani Nikula is to be setuid root or executed by root. 3751a894a8a5SRandy Dunlap 3752a894a8a5SRandy Dunlap nofpu [MIPS,SH] Disable hardware FPU at boot time. 3753e52347bdSJani Nikula 3754e52347bdSJani Nikula nofsgsbase [X86] Disables FSGSBASE instructions. 3755e52347bdSJani Nikula 3756e52347bdSJani Nikula nofxsr [BUGS=X86-32] Disables x86 floating point extended 3757e52347bdSJani Nikula register save and restore. The kernel will only save 3758e52347bdSJani Nikula legacy floating-point registers on task switch. 3759e52347bdSJani Nikula 3760e52347bdSJani Nikula nohalt [IA-64] Tells the kernel not to use the power saving 3761e52347bdSJani Nikula function PAL_HALT_LIGHT when idle. This increases 3762e52347bdSJani Nikula power-consumption. On the positive side, it reduces 3763e52347bdSJani Nikula interrupt wake-up latency, which may improve performance 37645ead723aSTimur Tabi in certain environments such as networked servers or 37655ead723aSTimur Tabi real-time systems. 37665ead723aSTimur Tabi 37675ead723aSTimur Tabi no_hash_pointers 37685ead723aSTimur Tabi Force pointers printed to the console or buffers to be 37695ead723aSTimur Tabi unhashed. By default, when a pointer is printed via %p 37705ead723aSTimur Tabi format string, that pointer is "hashed", i.e. obscured 37715ead723aSTimur Tabi by hashing the pointer value. This is a security feature 37725ead723aSTimur Tabi that hides actual kernel addresses from unprivileged 37735ead723aSTimur Tabi users, but it also makes debugging the kernel more 377484842911SChristophe Leroy difficult since unequal pointers can no longer be 37755ead723aSTimur Tabi compared. However, if this command-line option is 37765ead723aSTimur Tabi specified, then all normal pointers will have their true 37775ead723aSTimur Tabi value printed. This option should only be specified when 3778e52347bdSJani Nikula debugging the kernel. Please do not use on production 3779e52347bdSJani Nikula kernels. 37800ceef6e9SVaibhav Jain 3781975fd3c2SJiaxun Yang nohibernate [HIBERNATION] Disable hibernation and resume. 3782a894a8a5SRandy Dunlap 3783a894a8a5SRandy Dunlap nohlt [ARM,ARM64,MICROBLAZE,MIPS,PPC,SH] Forces the kernel to 3784a894a8a5SRandy Dunlap busy wait in do_idle() and not use the arch_cpu_idle() 3785a894a8a5SRandy Dunlap implementation; requires CONFIG_GENERIC_IDLE_POLL_SETUP 3786a894a8a5SRandy Dunlap to be effective. This is useful on platforms where the 3787a894a8a5SRandy Dunlap sleep(SH) or wfi(ARM,ARM64) instructions do not work 3788a894a8a5SRandy Dunlap correctly or when doing power measurements to evaluate 3789a894a8a5SRandy Dunlap the impact of the sleep instructions. This is also 3790a894a8a5SRandy Dunlap useful when using JTAG debugger. 3791a894a8a5SRandy Dunlap 3792a894a8a5SRandy Dunlap nohugeiomap [KNL,X86,PPC,ARM64] Disable kernel huge I/O mappings. 3793e52347bdSJani Nikula 3794e52347bdSJani Nikula nohugevmalloc [KNL,X86,PPC,ARM64] Disable kernel huge vmalloc mappings. 3795e52347bdSJani Nikula 3796e52347bdSJani Nikula nohz= [KNL] Boottime enable/disable dynamic ticks 3797d94d1053SFrederic Weisbecker Valid arguments: on, off 3798e52347bdSJani Nikula Default: on 3799e52347bdSJani Nikula 3800e52347bdSJani Nikula nohz_full= [KNL,BOOT,SMP,ISOL] 3801e52347bdSJani Nikula The argument is a cpu list, as described above. 3802f99bcb2cSPaul E. McKenney In kernels built with CONFIG_NO_HZ_FULL=y, set 3803f99bcb2cSPaul E. McKenney the specified list of CPUs whose tick will be stopped 3804f99bcb2cSPaul E. McKenney whenever possible. The boot CPU will be forced outside 3805f99bcb2cSPaul E. McKenney the range to maintain the timekeeping. Any CPUs 3806e52347bdSJani Nikula in this list will have their RCU callbacks offloaded, 3807b37a667cSJoel Fernandes just as if they had also been called out in the 3808b37a667cSJoel Fernandes rcu_nocbs= boot parameter. 3809b37a667cSJoel Fernandes 3810e52347bdSJani Nikula Note that this argument takes precedence over 3811e52347bdSJani Nikula the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 3812e52347bdSJani Nikula 3813e52347bdSJani Nikula noinitrd [RAM] Tells the kernel not to load any configured 3814e52347bdSJani Nikula initial RAM disk. 3815e52347bdSJani Nikula 3816e52347bdSJani Nikula nointremap [X86-64, Intel-IOMMU] Do not enable interrupt 3817e52347bdSJani Nikula remapping. 3818e52347bdSJani Nikula [Deprecated - use intremap=off] 3819e52347bdSJani Nikula 3820e52347bdSJani Nikula nointroute [IA-64] 3821a894a8a5SRandy Dunlap 3822a894a8a5SRandy Dunlap noinvpcid [X86] Disable the INVPCID cpu feature. 3823a894a8a5SRandy Dunlap 3824a894a8a5SRandy Dunlap noiotrap [SH] Disables trapped I/O port accesses. 3825a894a8a5SRandy Dunlap 3826a894a8a5SRandy Dunlap noirqdebug [X86-32] Disables the code which attempts to detect and 3827a894a8a5SRandy Dunlap disable unhandled interrupt sources. 3828e52347bdSJani Nikula 3829e52347bdSJani Nikula noisapnp [ISAPNP] Disables ISA PnP code. 383042551b8dSRandy Dunlap 383142551b8dSRandy Dunlap nojitter [IA-64] Disables jitter checking for ITC timers. 383242551b8dSRandy Dunlap 383342551b8dSRandy Dunlap nokaslr [KNL] 383442551b8dSRandy Dunlap When CONFIG_RANDOMIZE_BASE is set, this disables 3835e52347bdSJani Nikula kernel and module base offset ASLR (Address Space 3836e52347bdSJani Nikula Layout Randomization). 3837e52347bdSJani Nikula 3838a894a8a5SRandy Dunlap no-kvmapf [X86,KVM] Disable paravirtualized asynchronous page 3839e52347bdSJani Nikula fault handling. 3840e52347bdSJani Nikula 3841e52347bdSJani Nikula no-kvmclock [X86,KVM] Disable paravirtualized KVM clock driver 3842e52347bdSJani Nikula 3843e52347bdSJani Nikula nolapic [X86-32,APIC] Do not enable or use the local APIC. 3844e52347bdSJani Nikula 3845e52347bdSJani Nikula nolapic_timer [X86-32,APIC] Do not use the local APIC timer. 3846e52347bdSJani Nikula 3847e52347bdSJani Nikula nomca [IA-64] Disable machine check abort handling 3848e52347bdSJani Nikula 3849e52347bdSJani Nikula nomce [X86-32] Disable Machine Check Exception 3850e52347bdSJani Nikula 38519a758d87SThomas Zimmermann nomfgpt [X86-32] Disable Multi-Function General Purpose 38529a758d87SThomas Zimmermann Timer usage (for AMD Geode machines). 38539a758d87SThomas Zimmermann 38549a758d87SThomas Zimmermann nomodeset Disable kernel modesetting. Most systems' firmware 38559a758d87SThomas Zimmermann sets up a display mode and provides framebuffer memory 38569a758d87SThomas Zimmermann for output. With nomodeset, DRM and fbdev drivers will 38579a758d87SThomas Zimmermann not load if they could possibly displace the pre- 3858b22a15a5SJavier Martinez Canillas initialized output. Only the system framebuffer will 38599a758d87SThomas Zimmermann be available for use. The respective drivers will not 3860b22a15a5SJavier Martinez Canillas perform display-mode changes or accelerated rendering. 3861e52347bdSJani Nikula 3862e52347bdSJani Nikula Useful as error fallback, or for testing and debugging. 3863a894a8a5SRandy Dunlap 3864a894a8a5SRandy Dunlap nomodule Disable module load 3865a894a8a5SRandy Dunlap 3866a894a8a5SRandy Dunlap nonmi_ipi [X86] Disable using NMI IPIs during panic/reboot to 3867e52347bdSJani Nikula shutdown the other cpus. Instead use the REBOOT_VECTOR 3868e52347bdSJani Nikula irq. 3869e52347bdSJani Nikula 38700790c9aaSAndy Lutomirski nopat [X86] Disable PAT (page attribute table extension of 38710790c9aaSAndy Lutomirski pagetables) support. 387242551b8dSRandy Dunlap 387342551b8dSRandy Dunlap nopcid [X86-64] Disable the PCID cpu feature. 387442551b8dSRandy Dunlap 3875a894a8a5SRandy Dunlap nopku [X86] Disable Memory Protection Keys CPU feature found 3876a894a8a5SRandy Dunlap in some Intel CPUs. 3877a894a8a5SRandy Dunlap 387842551b8dSRandy Dunlap nopti [X86-64] 387942551b8dSRandy Dunlap Equivalent to pti=off 388042551b8dSRandy Dunlap 388142551b8dSRandy Dunlap nopv= [X86,XEN,KVM,HYPER_V,VMWARE] 388242551b8dSRandy Dunlap Disables the PV optimizations forcing the guest to run 388342551b8dSRandy Dunlap as generic guest with no PV drivers. Currently support 388442551b8dSRandy Dunlap XEN HVM, KVM, HYPER_V and VMWARE guest. 388542551b8dSRandy Dunlap 388642551b8dSRandy Dunlap nopvspin [X86,XEN,KVM] 388742551b8dSRandy Dunlap Disables the qspinlock slow path using PV optimizations 3888e52347bdSJani Nikula which allow the hypervisor to 'idle' the guest on lock 3889e52347bdSJani Nikula contention. 3890e52347bdSJani Nikula 3891e52347bdSJani Nikula norandmaps Don't use address space randomization. Equivalent to 3892e52347bdSJani Nikula echo 0 > /proc/sys/kernel/randomize_va_space 3893e52347bdSJani Nikula 3894e52347bdSJani Nikula noreplace-smp [X86-32,SMP] Don't replace SMP instructions 3895e52347bdSJani Nikula with UP alternatives 3896e52347bdSJani Nikula 3897a894a8a5SRandy Dunlap noresume [SWSUSP] Disables resume and restores original swap 3898a894a8a5SRandy Dunlap space. 3899e52347bdSJani Nikula 3900e52347bdSJani Nikula nosbagart [IA-64] 3901e52347bdSJani Nikula 3902e52347bdSJani Nikula no-scroll [VGA] Disables scrollback. 390338853a30SJarkko Sakkinen This is required for the Braillex ib80-piezo Braille 390438853a30SJarkko Sakkinen reader made by F.H. Papenmeier (Germany). 3905a894a8a5SRandy Dunlap 3906a894a8a5SRandy Dunlap nosgx [X86-64,SGX] Disables Intel SGX kernel support. 3907a894a8a5SRandy Dunlap 3908a894a8a5SRandy Dunlap nosmap [PPC] 3909a894a8a5SRandy Dunlap Disable SMAP (Supervisor Mode Access Prevention) 3910a894a8a5SRandy Dunlap even if it is supported by processor. 3911a894a8a5SRandy Dunlap 3912a894a8a5SRandy Dunlap nosmep [PPC64s] 3913e52347bdSJani Nikula Disable SMEP (Supervisor Mode Execution Prevention) 3914e52347bdSJani Nikula even if it is supported by processor. 3915e52347bdSJani Nikula 391673c58e7eSMichael Ellerman nosmp [SMP] Tells an SMP kernel to act as a UP kernel, 3917a894a8a5SRandy Dunlap and disable the IO APIC. legacy for "maxcpus=0". 3918a894a8a5SRandy Dunlap 391973c58e7eSMichael Ellerman nosmt [KNL,MIPS,PPC,S390] Disable symmetric multithreading (SMT). 3920a894a8a5SRandy Dunlap Equivalent to smt=1. 3921a894a8a5SRandy Dunlap 3922a894a8a5SRandy Dunlap [KNL,X86,PPC] Disable symmetric multithreading (SMT). 3923e52347bdSJani Nikula nosmt=force: Force disable SMT, cannot be undone 3924e52347bdSJani Nikula via the sysfs control file. 3925a894a8a5SRandy Dunlap 3926a894a8a5SRandy Dunlap nosoftlockup [KNL] Disable the soft-lockup detector. 3927a894a8a5SRandy Dunlap 3928a894a8a5SRandy Dunlap nospec_store_bypass_disable 3929a894a8a5SRandy Dunlap [HW] Disable all mitigations for the Speculative Store Bypass vulnerability 3930a894a8a5SRandy Dunlap 3931a894a8a5SRandy Dunlap nospectre_bhb [ARM64] Disable all mitigations for Spectre-BHB (branch 3932a894a8a5SRandy Dunlap history injection) vulnerability. System may allow data leaks 3933a894a8a5SRandy Dunlap with this option. 3934a894a8a5SRandy Dunlap 3935a894a8a5SRandy Dunlap nospectre_v1 [X86,PPC] Disable mitigations for Spectre Variant 1 3936a894a8a5SRandy Dunlap (bounds check bypass). With this option data leaks are 3937a894a8a5SRandy Dunlap possible in the system. 3938a894a8a5SRandy Dunlap 3939a894a8a5SRandy Dunlap nospectre_v2 [X86,PPC_E500,ARM64] Disable all mitigations for 3940a894a8a5SRandy Dunlap the Spectre variant 2 (indirect branch prediction) 3941a894a8a5SRandy Dunlap vulnerability. System may allow data leaks with this 3942a894a8a5SRandy Dunlap option. 3943a894a8a5SRandy Dunlap 3944a894a8a5SRandy Dunlap no-steal-acc [X86,PV_OPS,ARM64,PPC/PSERIES] Disable paravirtualized 3945e52347bdSJani Nikula steal time accounting. steal time is computed, but 3946e52347bdSJani Nikula won't influence scheduler behaviour 3947a894a8a5SRandy Dunlap 3948a894a8a5SRandy Dunlap nosync [HW,M68K] Disables sync negotiation for all devices. 3949a894a8a5SRandy Dunlap 3950a894a8a5SRandy Dunlap no_timer_check [X86,APIC] Disables the code which tests for 3951a894a8a5SRandy Dunlap broken timer IRQ sources. 3952a894a8a5SRandy Dunlap 3953a894a8a5SRandy Dunlap no_uaccess_flush 3954a894a8a5SRandy Dunlap [PPC] Don't flush the L1-D cache after accessing user data. 3955a894a8a5SRandy Dunlap 3956a894a8a5SRandy Dunlap novmcoredd [KNL,KDUMP] 3957a894a8a5SRandy Dunlap Disable device dump. Device dump allows drivers to 3958a894a8a5SRandy Dunlap append dump data to vmcore so you can collect driver 3959a894a8a5SRandy Dunlap specified debug info. Drivers can append the data 3960a894a8a5SRandy Dunlap without any limit and this data is stored in memory, 3961a894a8a5SRandy Dunlap so this may cause significant memory stress. Disabling 3962a894a8a5SRandy Dunlap device dump can help save memory but the driver debug 3963a894a8a5SRandy Dunlap data will be no longer available. This parameter 3964a894a8a5SRandy Dunlap is only available when CONFIG_PROC_VMCORE_DEVICE_DUMP 3965a894a8a5SRandy Dunlap is set. 3966a894a8a5SRandy Dunlap 3967a894a8a5SRandy Dunlap no-vmw-sched-clock 3968e52347bdSJani Nikula [X86,PV_OPS] Disable paravirtualized VMware scheduler 3969e52347bdSJani Nikula clock and use the default one. 3970e52347bdSJani Nikula 3971e52347bdSJani Nikula nowatchdog [KNL] Disable both lockup detectors, i.e. 3972e52347bdSJani Nikula soft-lockup and NMI watchdog (hard-lockup). 3973e52347bdSJani Nikula 3974e52347bdSJani Nikula nowb [ARM] 3975b8d1d163SDaniel Sneddon 3976b8d1d163SDaniel Sneddon nox2apic [X86-64,APIC] Do not enable x2APIC mode. 3977b8d1d163SDaniel Sneddon 3978b8d1d163SDaniel Sneddon NOTE: this parameter will be ignored on systems with the 3979a894a8a5SRandy Dunlap LEGACY_XAPIC_DISABLED bit set in the 3980a894a8a5SRandy Dunlap IA32_XAPIC_DISABLE_STATUS MSR. 3981a894a8a5SRandy Dunlap 3982a894a8a5SRandy Dunlap noxsave [BUGS=X86] Disables x86 extended register state save 3983a894a8a5SRandy Dunlap and restore using xsave. The kernel will fallback to 3984a894a8a5SRandy Dunlap enabling legacy floating-point and sse state. 3985a894a8a5SRandy Dunlap 3986a894a8a5SRandy Dunlap noxsaveopt [X86] Disables xsaveopt used in saving x86 extended 3987a894a8a5SRandy Dunlap register states. The kernel will fall back to use 3988a894a8a5SRandy Dunlap xsave to save the states. By using this parameter, 3989a894a8a5SRandy Dunlap performance of saving the states is degraded because 3990a894a8a5SRandy Dunlap xsave doesn't support modified optimization while 3991a894a8a5SRandy Dunlap xsaveopt supports it on xsaveopt enabled systems. 3992a894a8a5SRandy Dunlap 3993a894a8a5SRandy Dunlap noxsaves [X86] Disables xsaves and xrstors used in saving and 3994a894a8a5SRandy Dunlap restoring x86 extended register state in compacted 3995a894a8a5SRandy Dunlap form of xsave area. The kernel will fall back to use 3996a894a8a5SRandy Dunlap xsaveopt and xrstor to save and restore the states 3997a894a8a5SRandy Dunlap in standard form of xsave area. By using this 399835b55ef2SNoam Camus parameter, xsave area per process might occupy more 399935b55ef2SNoam Camus memory on xsaves enabled systems. 400035b55ef2SNoam Camus 400135b55ef2SNoam Camus nps_mtm_hs_ctr= [KNL,ARC] 400235b55ef2SNoam Camus This parameter sets the maximum duration, in 400335b55ef2SNoam Camus cycles, each HW thread of the CTOP can run 400435b55ef2SNoam Camus without interruptions, before HW switches it. 400535b55ef2SNoam Camus The actual maximum duration is 16 times this 400635b55ef2SNoam Camus parameter's value. 4007e52347bdSJani Nikula Format: integer between 1 and 255 4008e52347bdSJani Nikula Default: 255 4009e52347bdSJani Nikula 4010e52347bdSJani Nikula nptcg= [IA-64] Override max number of concurrent global TLB 4011e52347bdSJani Nikula purges which is reported from either PAL_VM_SUMMARY or 4012e52347bdSJani Nikula SAL PALO. 4013e52347bdSJani Nikula 4014e52347bdSJani Nikula nr_cpus= [SMP] Maximum number of processors that an SMP kernel 4015e52347bdSJani Nikula could support. nr_cpus=n : n >= 1 limits the kernel to 4016e52347bdSJani Nikula support 'n' processors. It could be larger than the 4017e52347bdSJani Nikula number of already plugged CPU during bootup, later in 4018e52347bdSJani Nikula runtime you can physically add extra cpu until it reaches 4019e52347bdSJani Nikula n. So during boot up some boot time memory for per-cpu 4020e52347bdSJani Nikula variables need be pre-allocated for later physical cpu 4021e52347bdSJani Nikula hot plugging. 4022544ef682SBarry Song 4023544ef682SBarry Song nr_uarts= [SERIAL] maximum number of UARTs to be registered. 4024544ef682SBarry Song 402500b072c0SBarry Song numa=off [KNL, ARM64, PPC, RISCV, SPARC, X86] Disable NUMA, Only 402600b072c0SBarry Song set up a single NUMA node spanning all memory. 4027e52347bdSJani Nikula 4028e52347bdSJani Nikula numa_balancing= [KNL,ARM64,PPC,RISCV,S390,X86] Enable or disable automatic 4029e52347bdSJani Nikula NUMA balancing. 4030c9bff3eeSMichal Hocko Allowed values are enable and disable 4031e52347bdSJani Nikula 403257043247SMauro Carvalho Chehab numa_zonelist_order= [KNL, BOOT] Select zonelist order for NUMA. 4033e52347bdSJani Nikula 'node', 'default' can be specified 4034e52347bdSJani Nikula This can be set from sysctl after boot. 4035a74e2a22SMauro Carvalho Chehab See Documentation/admin-guide/sysctl/vm.rst for details. 4036e52347bdSJani Nikula 4037e52347bdSJani Nikula ohci1394_dma=early [HW] enable debugging via the ohci1394 driver. 4038e52347bdSJani Nikula See Documentation/core-api/debugging-via-ohci1394.rst for more 4039e52347bdSJani Nikula info. 4040e52347bdSJani Nikula 4041e52347bdSJani Nikula olpc_ec_timeout= [OLPC] ms delay when issuing EC commands 4042e52347bdSJani Nikula Rather than timing out after 20 ms if an EC 4043e52347bdSJani Nikula command is not properly ACKed, override the length 4044e52347bdSJani Nikula of the timeout. We have interrupts disabled while 4045e52347bdSJani Nikula waiting for the ACK, so if this is set too high 4046e52347bdSJani Nikula interrupts *may* be lost! 4047e52347bdSJani Nikula 4048e52347bdSJani Nikula omap_mux= [OMAP] Override bootloader pin multiplexing. 4049e52347bdSJani Nikula Format: <mux_mode0.mode_name=value>... 4050d2fc83c1SRandy Dunlap For example, to override I2C bus2: 4051d2fc83c1SRandy Dunlap omap_mux=i2c2_scl.i2c2_scl=0x100,i2c2_sda.i2c2_sda=0x100 4052d2fc83c1SRandy Dunlap 4053d2fc83c1SRandy Dunlap onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration 4054d2fc83c1SRandy Dunlap 4055d2fc83c1SRandy Dunlap Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] 4056d2fc83c1SRandy Dunlap 4057d2fc83c1SRandy Dunlap boundary - index of last SLC block on Flex-OneNAND. 4058d2fc83c1SRandy Dunlap The remaining blocks are configured as MLC blocks. 4059d2fc83c1SRandy Dunlap lock - Configure if Flex-OneNAND boundary should be locked. 4060e52347bdSJani Nikula Once locked, the boundary cannot be changed. 4061e52347bdSJani Nikula 1 indicates lock status, 0 indicates unlock status. 4062e52347bdSJani Nikula 4063e52347bdSJani Nikula oops=panic Always panic on oopses. Default is to just kill the 4064e52347bdSJani Nikula process, but there is a small probability of 4065e52347bdSJani Nikula deadlocking the machine. 4066e900a918SDan Williams This will also cause panics on machine check exceptions. 4067e900a918SDan Williams Useful together with panic=30 to trigger a reboot. 4068e900a918SDan Williams 4069e900a918SDan Williams page_alloc.shuffle= 4070e900a918SDan Williams [KNL] Boolean flag to control whether the page allocator 4071e900a918SDan Williams should randomize its free lists. The randomization may 4072e900a918SDan Williams be automatically enabled if the kernel detects it is 4073e900a918SDan Williams running on a platform with a direct-mapped memory-side 4074e900a918SDan Williams cache, and this parameter can be used to 4075e900a918SDan Williams override/disable that behavior. The state of the flag 4076e52347bdSJani Nikula can be read from sysfs at: 4077e52347bdSJani Nikula /sys/module/page_alloc/parameters/shuffle. 4078e52347bdSJani Nikula 4079e52347bdSJani Nikula page_owner= [KNL] Boot-time page_owner enabling option. 4080e52347bdSJani Nikula Storage of the information about who allocated 4081e52347bdSJani Nikula each page is disabled in default. With this switch, 4082e52347bdSJani Nikula we can turn it on. 40838c9a134cSKees Cook on: enable the feature 40848c9a134cSKees Cook 40858c9a134cSKees Cook page_poison= [KNL] Boot-time parameter changing the state of 4086e52347bdSJani Nikula poisoning on the buddy allocator, available with 4087e52347bdSJani Nikula CONFIG_PAGE_POISONING=y. 4088f58780a8SGavin Shan off: turn off poisoning (default) 4089f58780a8SGavin Shan on: turn on poisoning 4090f58780a8SGavin Shan 4091f58780a8SGavin Shan page_reporting.page_reporting_order= 409223baf831SKirill A. Shutemov [KNL] Minimal page reporting order 4093f58780a8SGavin Shan Format: <integer> 4094e52347bdSJani Nikula Adjust the minimal page reporting order. The page 4095e52347bdSJani Nikula reporting is disabled when it exceeds MAX_ORDER. 4096e52347bdSJani Nikula 4097e52347bdSJani Nikula panic= [KNL] Kernel behaviour on panic: delay <timeout> 4098e52347bdSJani Nikula timeout > 0: seconds before rebooting 4099e52347bdSJani Nikula timeout = 0: wait forever 4100db38d5c1SRafael Aquini timeout < 0: reboot immediately 4101db38d5c1SRafael Aquini Format: <timeout> 4102db38d5c1SRafael Aquini 4103db38d5c1SRafael Aquini panic_on_taint= Bitmask for conditionally calling panic() in add_taint() 4104db38d5c1SRafael Aquini Format: <hex>[,nousertaint] 4105db38d5c1SRafael Aquini Hexadecimal bitmask representing the set of TAINT flags 4106db38d5c1SRafael Aquini that will cause the kernel to panic when add_taint() is 4107db38d5c1SRafael Aquini called with any of the flags in this set. 4108db38d5c1SRafael Aquini The optional switch "nousertaint" can be utilized to 4109db38d5c1SRafael Aquini prevent userspace forced crashes by writing to sysctl 4110db38d5c1SRafael Aquini /proc/sys/kernel/tainted any flagset matching with the 4111db38d5c1SRafael Aquini bitmask set on panic_on_taint. 4112db38d5c1SRafael Aquini See Documentation/admin-guide/tainted-kernels.rst for 411357ada235SOlaf Hering extra details on the taint flags that users can pick 4114e52347bdSJani Nikula to compose the bitmask to assign to panic_on_taint. 4115e52347bdSJani Nikula 4116b4047e53SRandy Dunlap panic_on_warn=1 panic() instead of WARN(). Useful to cause kdump 4117b4047e53SRandy Dunlap on a WARN(). 4118b4047e53SRandy Dunlap 4119b4047e53SRandy Dunlap panic_print= Bitmask for printing system info when panic happens. 4120b4047e53SRandy Dunlap User can chose combination of the following bits: 4121b4047e53SRandy Dunlap bit 0: print all tasks info 4122b4047e53SRandy Dunlap bit 1: print system memory info 4123b4047e53SRandy Dunlap bit 2: print timer info 4124b4047e53SRandy Dunlap bit 3: print locks info if CONFIG_LOCKDEP is on 4125b4047e53SRandy Dunlap bit 4: print ftrace buffer 4126b4047e53SRandy Dunlap bit 5: print all printk messages in buffer 4127b4047e53SRandy Dunlap bit 6: print all CPUs backtrace (if available in the arch) 4128b4047e53SRandy Dunlap *Be aware* that this option may print a _lot_ of lines, 4129b4047e53SRandy Dunlap so there are risks of losing older messages in the log. 4130e52347bdSJani Nikula Use this option carefully, maybe worth to setup a 4131e52347bdSJani Nikula bigger log buffer with "log_buf_len" along with this. 4132e52347bdSJani Nikula 4133e52347bdSJani Nikula parkbd.port= [HW] Parallel port number the keyboard adapter is 4134e52347bdSJani Nikula connected to, default is 0. 4135e52347bdSJani Nikula Format: <parport#> 4136e52347bdSJani Nikula parkbd.mode= [HW] Parallel port keyboard adapter mode of operation, 4137e52347bdSJani Nikula 0 for XT, 1 for AT (default is AT). 4138e52347bdSJani Nikula Format: <mode> 4139e52347bdSJani Nikula 4140e52347bdSJani Nikula parport= [HW,PPT] Specify parallel ports. 0 disables. 4141e52347bdSJani Nikula Format: { 0 | auto | 0xBBB[,IRQ[,DMA]] } 4142e52347bdSJani Nikula Use 'auto' to force the driver to use any 4143e52347bdSJani Nikula IRQ/DMA settings detected (the default is to 4144e52347bdSJani Nikula ignore detected IRQ/DMA settings because of 4145e52347bdSJani Nikula possible conflicts). You can specify the base 4146e52347bdSJani Nikula address, IRQ, and DMA settings; IRQ and DMA 4147e52347bdSJani Nikula should be numbers, or 'auto' (for using detected 4148e52347bdSJani Nikula settings on that particular port), or 'nofifo' 4149e52347bdSJani Nikula (to avoid using a FIFO even if it is detected). 4150e52347bdSJani Nikula Parallel ports are assigned in the order they 4151e52347bdSJani Nikula are specified on the command line, starting 4152e52347bdSJani Nikula with parport0. 4153e52347bdSJani Nikula 4154e52347bdSJani Nikula parport_init_mode= [HW,PPT] 4155e52347bdSJani Nikula Configure VIA parallel port to operate in 4156e52347bdSJani Nikula a specific mode. This is necessary on Pegasos 4157e52347bdSJani Nikula computer where firmware has no options for setting 4158e52347bdSJani Nikula up parallel port mode and sets it to spp. 4159426e2c6aSMaciej W. Rozycki Currently this function knows 686a and 8231 chips. 4160426e2c6aSMaciej W. Rozycki Format: [spp|ps2|epp|ecp|ecpepp] 4161426e2c6aSMaciej W. Rozycki 4162426e2c6aSMaciej W. Rozycki pata_legacy.all= [HW,LIBATA] 4163426e2c6aSMaciej W. Rozycki Format: <int> 4164426e2c6aSMaciej W. Rozycki Set to non-zero to probe primary and secondary ISA 4165426e2c6aSMaciej W. Rozycki port ranges on PCI systems where no PCI PATA device 4166426e2c6aSMaciej W. Rozycki has been found at either range. Disabled by default. 4167426e2c6aSMaciej W. Rozycki 4168426e2c6aSMaciej W. Rozycki pata_legacy.autospeed= [HW,LIBATA] 4169426e2c6aSMaciej W. Rozycki Format: <int> 4170426e2c6aSMaciej W. Rozycki Set to non-zero if a chip is present that snoops speed 4171426e2c6aSMaciej W. Rozycki changes. Disabled by default. 4172426e2c6aSMaciej W. Rozycki 4173426e2c6aSMaciej W. Rozycki pata_legacy.ht6560a= [HW,LIBATA] 4174426e2c6aSMaciej W. Rozycki Format: <int> 4175426e2c6aSMaciej W. Rozycki Set to 1, 2, or 3 for HT 6560A on the primary channel, 4176426e2c6aSMaciej W. Rozycki the secondary channel, or both channels respectively. 4177426e2c6aSMaciej W. Rozycki Disabled by default. 4178426e2c6aSMaciej W. Rozycki 4179426e2c6aSMaciej W. Rozycki pata_legacy.ht6560b= [HW,LIBATA] 4180426e2c6aSMaciej W. Rozycki Format: <int> 4181426e2c6aSMaciej W. Rozycki Set to 1, 2, or 3 for HT 6560B on the primary channel, 4182426e2c6aSMaciej W. Rozycki the secondary channel, or both channels respectively. 4183426e2c6aSMaciej W. Rozycki Disabled by default. 4184426e2c6aSMaciej W. Rozycki 4185426e2c6aSMaciej W. Rozycki pata_legacy.iordy_mask= [HW,LIBATA] 4186426e2c6aSMaciej W. Rozycki Format: <int> 4187426e2c6aSMaciej W. Rozycki IORDY enable mask. Set individual bits to allow IORDY 4188426e2c6aSMaciej W. Rozycki for the respective channel. Bit 0 is for the first 4189426e2c6aSMaciej W. Rozycki legacy channel handled by this driver, bit 1 is for 4190426e2c6aSMaciej W. Rozycki the second channel, and so on. The sequence will often 4191426e2c6aSMaciej W. Rozycki correspond to the primary legacy channel, the secondary 4192426e2c6aSMaciej W. Rozycki legacy channel, and so on, but the handling of a PCI 4193426e2c6aSMaciej W. Rozycki bus and the use of other driver options may interfere 4194426e2c6aSMaciej W. Rozycki with the sequence. By default IORDY is allowed across 4195426e2c6aSMaciej W. Rozycki all channels. 4196426e2c6aSMaciej W. Rozycki 4197426e2c6aSMaciej W. Rozycki pata_legacy.opti82c46x= [HW,LIBATA] 4198426e2c6aSMaciej W. Rozycki Format: <int> 4199426e2c6aSMaciej W. Rozycki Set to 1, 2, or 3 for Opti 82c611A on the primary 4200426e2c6aSMaciej W. Rozycki channel, the secondary channel, or both channels 4201426e2c6aSMaciej W. Rozycki respectively. Disabled by default. 4202426e2c6aSMaciej W. Rozycki 4203426e2c6aSMaciej W. Rozycki pata_legacy.opti82c611a= [HW,LIBATA] 4204426e2c6aSMaciej W. Rozycki Format: <int> 4205426e2c6aSMaciej W. Rozycki Set to 1, 2, or 3 for Opti 82c465MV on the primary 4206426e2c6aSMaciej W. Rozycki channel, the secondary channel, or both channels 4207426e2c6aSMaciej W. Rozycki respectively. Disabled by default. 4208426e2c6aSMaciej W. Rozycki 4209426e2c6aSMaciej W. Rozycki pata_legacy.pio_mask= [HW,LIBATA] 4210426e2c6aSMaciej W. Rozycki Format: <int> 4211426e2c6aSMaciej W. Rozycki PIO mode mask for autospeed devices. Set individual 4212426e2c6aSMaciej W. Rozycki bits to allow the use of the respective PIO modes. 4213426e2c6aSMaciej W. Rozycki Bit 0 is for mode 0, bit 1 is for mode 1, and so on. 4214426e2c6aSMaciej W. Rozycki All modes allowed by default. 4215426e2c6aSMaciej W. Rozycki 4216426e2c6aSMaciej W. Rozycki pata_legacy.probe_all= [HW,LIBATA] 4217426e2c6aSMaciej W. Rozycki Format: <int> 42187d33004dSMaciej W. Rozycki Set to non-zero to probe tertiary and further ISA 42197d33004dSMaciej W. Rozycki port ranges on PCI systems. Disabled by default. 42207d33004dSMaciej W. Rozycki 42217d33004dSMaciej W. Rozycki pata_legacy.probe_mask= [HW,LIBATA] 42227d33004dSMaciej W. Rozycki Format: <int> 42237d33004dSMaciej W. Rozycki Probe mask for legacy ISA PATA ports. Depending on 42247d33004dSMaciej W. Rozycki platform configuration and the use of other driver 42257d33004dSMaciej W. Rozycki options up to 6 legacy ports are supported: 0x1f0, 42267d33004dSMaciej W. Rozycki 0x170, 0x1e8, 0x168, 0x1e0, 0x160, however probing 42277d33004dSMaciej W. Rozycki of individual ports can be disabled by setting the 42287d33004dSMaciej W. Rozycki corresponding bits in the mask to 1. Bit 0 is for 4229426e2c6aSMaciej W. Rozycki the first port in the list above (0x1f0), and so on. 4230426e2c6aSMaciej W. Rozycki By default all supported ports are probed. 4231426e2c6aSMaciej W. Rozycki 4232426e2c6aSMaciej W. Rozycki pata_legacy.qdi= [HW,LIBATA] 4233426e2c6aSMaciej W. Rozycki Format: <int> 4234426e2c6aSMaciej W. Rozycki Set to non-zero to probe QDI controllers. By default 4235426e2c6aSMaciej W. Rozycki set to 1 if CONFIG_PATA_QDI_MODULE, 0 otherwise. 4236426e2c6aSMaciej W. Rozycki 4237426e2c6aSMaciej W. Rozycki pata_legacy.winbond= [HW,LIBATA] 4238426e2c6aSMaciej W. Rozycki Format: <int> 4239426e2c6aSMaciej W. Rozycki Set to non-zero to probe Winbond controllers. Use 4240426e2c6aSMaciej W. Rozycki the standard I/O port (0x130) if 1, otherwise the 4241426e2c6aSMaciej W. Rozycki value given is the I/O port to use (typically 0x1b0). 42426ddcec95SMaciej W. Rozycki By default set to 1 if CONFIG_PATA_WINBOND_VLB_MODULE, 42436ddcec95SMaciej W. Rozycki 0 otherwise. 42446ddcec95SMaciej W. Rozycki 42456ddcec95SMaciej W. Rozycki pata_platform.pio_mask= [HW,LIBATA] 42466ddcec95SMaciej W. Rozycki Format: <int> 42476ddcec95SMaciej W. Rozycki Supported PIO mode mask. Set individual bits to allow 42486ddcec95SMaciej W. Rozycki the use of the respective PIO modes. Bit 0 is for 4249b4047e53SRandy Dunlap mode 0, bit 1 is for mode 1, and so on. Mode 0 only 4250e52347bdSJani Nikula allowed by default. 4251e52347bdSJani Nikula 4252e52347bdSJani Nikula pause_on_oops=<int> 4253e52347bdSJani Nikula Halt all CPUs after the first oops has been printed for 4254e52347bdSJani Nikula the specified number of seconds. This is to be used if 4255e52347bdSJani Nikula your oopses keep scrolling off the screen. 425607d8d7e5SLogan Gunthorpe 425707d8d7e5SLogan Gunthorpe pcbit= [HW,ISDN] 425807d8d7e5SLogan Gunthorpe 425907d8d7e5SLogan Gunthorpe pci=option[,option...] [PCI] various PCI subsystem options. 426007d8d7e5SLogan Gunthorpe 426107d8d7e5SLogan Gunthorpe Some options herein operate on a specific device 426245db3370SLogan Gunthorpe or a set of devices (<pci_dev>). These are 426307d8d7e5SLogan Gunthorpe specified in one of the following formats: 426407d8d7e5SLogan Gunthorpe 426507d8d7e5SLogan Gunthorpe [<domain>:]<bus>:<dev>.<func>[/<dev>.<func>]* 426607d8d7e5SLogan Gunthorpe pci:<vendor>:<device>[:<subvendor>:<subdevice>] 426707d8d7e5SLogan Gunthorpe 426807d8d7e5SLogan Gunthorpe Note: the first format specifies a PCI 426907d8d7e5SLogan Gunthorpe bus/device/function address which may change 427007d8d7e5SLogan Gunthorpe if new hardware is inserted, if motherboard 427145db3370SLogan Gunthorpe firmware changes, or due to changes caused 427245db3370SLogan Gunthorpe by other kernel parameters. If the 427345db3370SLogan Gunthorpe domain is left unspecified, it is 427445db3370SLogan Gunthorpe taken to be zero. Optionally, a path 427545db3370SLogan Gunthorpe to a device through multiple device/function 427607d8d7e5SLogan Gunthorpe addresses can be specified after the base 427707d8d7e5SLogan Gunthorpe address (this is more robust against 427807d8d7e5SLogan Gunthorpe renumbering issues). The second format 427907d8d7e5SLogan Gunthorpe selects devices using IDs from the 428011eb0e0eSSinan Kaya configuration space which may match multiple 4281e52347bdSJani Nikula devices in the system. 4282e52347bdSJani Nikula 4283e52347bdSJani Nikula earlydump dump PCI config space before the kernel 4284e52347bdSJani Nikula changes anything 4285e52347bdSJani Nikula off [X86] don't probe for the PCI bus 4286e52347bdSJani Nikula bios [X86-32] force use of PCI BIOS, don't access 4287e52347bdSJani Nikula the hardware directly. Use this if your machine 4288e52347bdSJani Nikula has a non-standard PCI host bridge. 4289e52347bdSJani Nikula nobios [X86-32] disallow use of PCI BIOS, only direct 4290e52347bdSJani Nikula hardware access methods are allowed. Use this 4291e52347bdSJani Nikula if you experience crashes upon bootup and you 4292e52347bdSJani Nikula suspect they are caused by the BIOS. 4293e52347bdSJani Nikula conf1 [X86] Force use of PCI Configuration Access 4294e52347bdSJani Nikula Mechanism 1 (config address in IO port 0xCF8, 4295e52347bdSJani Nikula data in IO port 0xCFC, both 32-bit). 4296e52347bdSJani Nikula conf2 [X86] Force use of PCI Configuration Access 4297e52347bdSJani Nikula Mechanism 2 (IO port 0xCF8 is an 8-bit port for 4298e52347bdSJani Nikula the function, IO port 0xCFA, also 8-bit, sets 4299e52347bdSJani Nikula bus number. The config space is then accessed 4300e52347bdSJani Nikula through ports 0xC000-0xCFFF). 4301e52347bdSJani Nikula See http://wiki.osdev.org/PCI for more info 4302e52347bdSJani Nikula on the configuration access mechanisms. 4303e52347bdSJani Nikula noaer [PCIE] If the PCIEAER kernel config parameter is 4304e52347bdSJani Nikula enabled, this kernel boot option can be used to 4305e52347bdSJani Nikula disable the use of PCIE advanced error reporting. 4306e52347bdSJani Nikula nodomains [PCI] Disable support for multiple PCI 4307e52347bdSJani Nikula root domains (aka PCI segments, in ACPI-speak). 4308e52347bdSJani Nikula nommconf [X86] Disable use of MMCONFIG for PCI 4309e52347bdSJani Nikula Configuration 4310e52347bdSJani Nikula check_enable_amd_mmconf [X86] check for and enable 4311e52347bdSJani Nikula properly configured MMIO access to PCI 4312e52347bdSJani Nikula config space on AMD family 10h CPU 4313e52347bdSJani Nikula nomsi [MSI] If the PCI_MSI kernel config parameter is 4314e52347bdSJani Nikula enabled, this kernel boot option can be used to 4315e52347bdSJani Nikula disable the use of MSI interrupts system-wide. 4316e52347bdSJani Nikula noioapicquirk [APIC] Disable all boot interrupt quirks. 4317e52347bdSJani Nikula Safety option to keep boot IRQs enabled. This 4318e52347bdSJani Nikula should never be necessary. 4319e52347bdSJani Nikula ioapicreroute [APIC] Enable rerouting of boot IRQs to the 4320e52347bdSJani Nikula primary IO-APIC for bridges that cannot disable 4321e52347bdSJani Nikula boot IRQs. This fixes a source of spurious IRQs 4322e52347bdSJani Nikula when the system masks IRQs. 4323e52347bdSJani Nikula noioapicreroute [APIC] Disable workaround that uses the 4324e52347bdSJani Nikula boot IRQ equivalent of an IRQ that connects to 4325e52347bdSJani Nikula a chipset where boot IRQs cannot be disabled. 4326e52347bdSJani Nikula The opposite of ioapicreroute. 4327e52347bdSJani Nikula biosirq [X86-32] Use PCI BIOS calls to get the interrupt 4328e52347bdSJani Nikula routing table. These calls are known to be buggy 4329e52347bdSJani Nikula on several machines and they hang the machine 4330e52347bdSJani Nikula when used, but on other computers it's the only 4331e52347bdSJani Nikula way to get the interrupt routing table. Try 4332e52347bdSJani Nikula this option if the kernel is unable to allocate 4333e52347bdSJani Nikula IRQs or discover secondary PCI buses on your 4334e52347bdSJani Nikula motherboard. 4335e52347bdSJani Nikula rom [X86] Assign address space to expansion ROMs. 4336e52347bdSJani Nikula Use with caution as certain devices share 4337e52347bdSJani Nikula address decoders between ROMs and other 4338e52347bdSJani Nikula resources. 4339e52347bdSJani Nikula norom [X86] Do not assign address space to 4340e52347bdSJani Nikula expansion ROMs that do not already have 4341e52347bdSJani Nikula BIOS assigned address ranges. 4342e52347bdSJani Nikula nobar [X86] Do not assign address space to the 4343e52347bdSJani Nikula BARs that weren't assigned by the BIOS. 4344e52347bdSJani Nikula irqmask=0xMMMM [X86] Set a bit mask of IRQs allowed to be 4345e52347bdSJani Nikula assigned automatically to PCI devices. You can 4346e52347bdSJani Nikula make the kernel exclude IRQs of your ISA cards 4347e52347bdSJani Nikula this way. 4348e52347bdSJani Nikula pirqaddr=0xAAAAA [X86] Specify the physical address 4349e52347bdSJani Nikula of the PIRQ table (normally generated 4350e52347bdSJani Nikula by the BIOS) if it is outside the 4351e52347bdSJani Nikula F0000h-100000h range. 4352e52347bdSJani Nikula lastbus=N [X86] Scan all buses thru bus #N. Can be 4353e52347bdSJani Nikula useful if the kernel is unable to find your 4354e52347bdSJani Nikula secondary buses and you want to tell it 4355e52347bdSJani Nikula explicitly which ones they are. 4356e52347bdSJani Nikula assign-busses [X86] Always assign all PCI bus 4357e52347bdSJani Nikula numbers ourselves, overriding 4358e52347bdSJani Nikula whatever the firmware may have done. 4359e52347bdSJani Nikula usepirqmask [X86] Honor the possible IRQ mask stored 4360e52347bdSJani Nikula in the BIOS $PIR table. This is needed on 4361e52347bdSJani Nikula some systems with broken BIOSes, notably 4362e52347bdSJani Nikula some HP Pavilion N5400 and Omnibook XE3 4363e52347bdSJani Nikula notebooks. This will have no effect if ACPI 4364e52347bdSJani Nikula IRQ routing is enabled. 4365e52347bdSJani Nikula noacpi [X86] Do not use ACPI for IRQ routing 4366e52347bdSJani Nikula or for PCI scanning. 4367e52347bdSJani Nikula use_crs [X86] Use PCI host bridge window information 4368e52347bdSJani Nikula from ACPI. On BIOSes from 2008 or later, this 4369e52347bdSJani Nikula is enabled by default. If you need to use this, 4370fa6dae5dSHans de Goede please report a bug. 4371fa6dae5dSHans de Goede nocrs [X86] Ignore PCI host bridge windows from ACPI. 4372fa6dae5dSHans de Goede If you need to use this, please report a bug. 4373fa6dae5dSHans de Goede use_e820 [X86] Use E820 reservations to exclude parts of 4374fa6dae5dSHans de Goede PCI host bridge windows. This is a workaround 4375fa6dae5dSHans de Goede for BIOS defects in host bridge _CRS methods. 4376fa6dae5dSHans de Goede If you need to use this, please report a bug to 4377fa6dae5dSHans de Goede <linux-pci@vger.kernel.org>. 4378fa6dae5dSHans de Goede no_e820 [X86] Ignore E820 reservations for PCI host 4379e52347bdSJani Nikula bridge windows. This is the default on modern 4380e52347bdSJani Nikula hardware. If you need to use this, please report 4381e52347bdSJani Nikula a bug to <linux-pci@vger.kernel.org>. 4382e52347bdSJani Nikula routeirq Do IRQ routing for all PCI devices. 4383e52347bdSJani Nikula This is normally done in pci_enable_device(), 4384e52347bdSJani Nikula so this option is a temporary workaround 4385e52347bdSJani Nikula for broken drivers that don't call it. 4386e52347bdSJani Nikula skip_isa_align [X86] do not align io start addr, so can 4387e52347bdSJani Nikula handle more pci cards 4388e52347bdSJani Nikula noearly [X86] Don't do any early type 1 scanning. 4389e52347bdSJani Nikula This might help on some broken boards which 4390e52347bdSJani Nikula machine check when some devices' config space 4391e52347bdSJani Nikula is read. But various workarounds are disabled 4392e52347bdSJani Nikula and some IOMMU drivers will not work. 4393e52347bdSJani Nikula bfsort Sort PCI devices into breadth-first order. 4394e52347bdSJani Nikula This sorting is done to get a device 4395e52347bdSJani Nikula order compatible with older (<= 2.4) kernels. 4396e52347bdSJani Nikula nobfsort Don't sort PCI devices into breadth-first order. 4397e52347bdSJani Nikula pcie_bus_tune_off Disable PCIe MPS (Max Payload Size) 4398e52347bdSJani Nikula tuning and use the BIOS-configured MPS defaults. 4399e52347bdSJani Nikula pcie_bus_safe Set every device's MPS to the largest value 4400e52347bdSJani Nikula supported by all devices below the root complex. 4401e52347bdSJani Nikula pcie_bus_perf Set device MPS to the largest allowable MPS 4402e52347bdSJani Nikula based on its parent bus. Also set MRRS (Max 4403e52347bdSJani Nikula Read Request Size) to the largest supported 4404e52347bdSJani Nikula value (no larger than the MPS that the device 4405e52347bdSJani Nikula or bus can support) for best performance. 4406e52347bdSJani Nikula pcie_bus_peer2peer Set every device's MPS to 128B, which 4407e52347bdSJani Nikula every device is guaranteed to support. This 4408e52347bdSJani Nikula configuration allows peer-to-peer DMA between 4409e52347bdSJani Nikula any pair of devices, possibly at the cost of 4410e52347bdSJani Nikula reduced performance. This also guarantees 4411e52347bdSJani Nikula that hot-added devices will work. 4412e52347bdSJani Nikula cbiosize=nn[KMG] The fixed amount of bus space which is 4413e52347bdSJani Nikula reserved for the CardBus bridge's IO window. 4414e52347bdSJani Nikula The default value is 256 bytes. 4415e52347bdSJani Nikula cbmemsize=nn[KMG] The fixed amount of bus space which is 4416e52347bdSJani Nikula reserved for the CardBus bridge's memory 441707d8d7e5SLogan Gunthorpe window. The default value is 64 megabytes. 4418e52347bdSJani Nikula resource_alignment= 441907d8d7e5SLogan Gunthorpe Format: 442007d8d7e5SLogan Gunthorpe [<order of align>@]<pci_dev>[; ...] 4421e52347bdSJani Nikula Specifies alignment and device to reassign 4422e52347bdSJani Nikula aligned memory resources. How to 44233b1b1ce3SAlexey Kardashevskiy specify the device is described above. 4424e52347bdSJani Nikula If <order of align> is not specified, 4425e52347bdSJani Nikula PAGE_SIZE is used as alignment. 4426e52347bdSJani Nikula A PCI-PCI bridge can be specified if resource 4427e52347bdSJani Nikula windows need to be expanded. 44283b1b1ce3SAlexey Kardashevskiy To specify the alignment for several 44293b1b1ce3SAlexey Kardashevskiy instances of a device, the PCI vendor, 4430e52347bdSJani Nikula device, subvendor, and subdevice may be 4431bba50659SVidya Sagar specified, e.g., 12@pci:8086:9c22:103c:198f 4432bba50659SVidya Sagar for 4096-byte alignment. 4433bba50659SVidya Sagar ecrc= Enable/disable PCIe ECRC (transaction layer 4434e52347bdSJani Nikula end-to-end CRC checking). Only effective if 4435e52347bdSJani Nikula OS has native AER control (either granted by 4436e52347bdSJani Nikula ACPI _OSC or forced via "pcie_ports=native") 4437e52347bdSJani Nikula bios: Use BIOS/firmware settings. This is the 4438e52347bdSJani Nikula the default. 4439e52347bdSJani Nikula off: Turn ECRC off 4440e52347bdSJani Nikula on: Turn ECRC on. 4441d7b8a217SNicholas Johnson hpiosize=nn[KMG] The fixed amount of bus space which is 4442d7b8a217SNicholas Johnson reserved for hotplug bridge's IO window. 4443d7b8a217SNicholas Johnson Default size is 256 bytes. 4444d7b8a217SNicholas Johnson hpmmiosize=nn[KMG] The fixed amount of bus space which is 4445d7b8a217SNicholas Johnson reserved for hotplug bridge's MMIO window. 4446d7b8a217SNicholas Johnson Default size is 2 megabytes. 4447e52347bdSJani Nikula hpmmioprefsize=nn[KMG] The fixed amount of bus space which is 4448d7b8a217SNicholas Johnson reserved for hotplug bridge's MMIO_PREF window. 4449d7b8a217SNicholas Johnson Default size is 2 megabytes. 4450e52347bdSJani Nikula hpmemsize=nn[KMG] The fixed amount of bus space which is 4451e52347bdSJani Nikula reserved for hotplug bridge's MMIO and 4452e52347bdSJani Nikula MMIO_PREF window. 4453e52347bdSJani Nikula Default size is 2 megabytes. 4454e52347bdSJani Nikula hpbussize=nn The minimum amount of additional bus numbers 4455e52347bdSJani Nikula reserved for buses below a hotplug bridge. 4456e52347bdSJani Nikula Default is 1. 4457e52347bdSJani Nikula realloc= Enable/disable reallocating PCI bridge resources 4458e52347bdSJani Nikula if allocations done by BIOS are too small to 4459e52347bdSJani Nikula accommodate resources required by all child 4460e52347bdSJani Nikula devices. 4461e52347bdSJani Nikula off: Turn realloc off 4462cef74409SGil Kupfer on: Turn realloc on 4463cef74409SGil Kupfer realloc same as realloc=on 4464e52347bdSJani Nikula noari do not use PCIe ARI. 4465e52347bdSJani Nikula noats [PCIE, Intel-IOMMU, AMD-IOMMU] 4466e52347bdSJani Nikula do not use PCIe ATS (and IOMMU device IOTLB). 4467f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= pcie_scan_all Scan all possible PCIe devices. Otherwise we 4468f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= only look for one device below a PCIe downstream 4469f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= port. 4470f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= big_root_window Try to add a big 64bit memory window to the PCIe 4471f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= root complex on AMD CPUs. Some GFX hardware 4472f32ab754S=?UTF-8?q?Christian=20K=C3=B6nig?= can resize a BAR to allow access to all VRAM. 4473aaca43fdSLogan Gunthorpe Adding the window is slightly risky (it may 4474aaca43fdSLogan Gunthorpe conflict with unreported devices), so this 4475aaca43fdSLogan Gunthorpe taints the kernel. 4476aaca43fdSLogan Gunthorpe disable_acs_redir=<pci_dev>[; ...] 4477aaca43fdSLogan Gunthorpe Specify one or more PCI devices (in the format 4478aaca43fdSLogan Gunthorpe specified above) separated by semicolons. 4479aaca43fdSLogan Gunthorpe Each device specified will have the PCI ACS 4480aaca43fdSLogan Gunthorpe redirect capabilities forced off which will 4481aaca43fdSLogan Gunthorpe allow P2P traffic between devices through 4482fbfe07d4SSebastian Ott bridges without forcing it upstream. Note: 448356271303SSebastian Ott this removes isolation between devices and 4484de267a7cSPierre Morel may put more devices in an IOMMU group. 4485de267a7cSPierre Morel force_floating [S390] Force usage of floating interrupts. 4486e52347bdSJani Nikula nomio [S390] Do not use MIO instructions. 4487e52347bdSJani Nikula norid [S390] ignore the RID field and force use of 4488e52347bdSJani Nikula one PCI domain per PCI function 4489e52347bdSJani Nikula 4490e52347bdSJani Nikula pcie_aspm= [PCIE] Forcibly enable or disable PCIe Active State Power 4491e52347bdSJani Nikula Management. 4492e52347bdSJani Nikula off Disable ASPM. 44934c0fd764SBjorn Helgaas force Enable ASPM even on devices that claim not to support it. 44944c0fd764SBjorn Helgaas WARNING: Forcing ASPM on may cause system lockups. 44954c0fd764SBjorn Helgaas 44964c0fd764SBjorn Helgaas pcie_ports= [PCIE] PCIe port services handling: 44974c0fd764SBjorn Helgaas native Use native PCIe services (PME, AER, DPC, PCIe hotplug) 449835a0b237SOlof Johansson even if the platform doesn't give the OS permission to 449935a0b237SOlof Johansson use them. This may cause conflicts if the platform 45004c0fd764SBjorn Helgaas also tries to use these services. 45014c0fd764SBjorn Helgaas dpc-native Use native PCIe service for DPC only. May 4502e52347bdSJani Nikula cause conflicts if firmware uses AER or DPC. 4503e52347bdSJani Nikula compat Disable native PCIe services (PME, AER, DPC, PCIe 4504e52347bdSJani Nikula hotplug). 4505e52347bdSJani Nikula 4506e52347bdSJani Nikula pcie_port_pm= [PCIE] PCIe port power management handling: 4507e52347bdSJani Nikula off Disable power management of all PCIe ports 4508e52347bdSJani Nikula force Forcibly enable power management of all PCIe ports 4509e52347bdSJani Nikula 4510e52347bdSJani Nikula pcie_pme= [PCIE,PM] Native PCIe PME signaling options: 4511e52347bdSJani Nikula nomsi Do not use MSI for native PCIe PME signaling (this makes 4512e52347bdSJani Nikula all PCIe root ports use INTx for all services). 4513e52347bdSJani Nikula 4514e52347bdSJani Nikula pcmv= [HW,PCMCIA] BadgePAD 4 4515e52347bdSJani Nikula 4516e52347bdSJani Nikula pd_ignore_unused 4517e52347bdSJani Nikula [PM] 4518e52347bdSJani Nikula Keep all power-domains already enabled by bootloader on, 4519e52347bdSJani Nikula even if no driver has claimed them. This is useful 4520e52347bdSJani Nikula for debug and development, but should not be 4521e52347bdSJani Nikula needed on a platform with proper driver support. 4522e52347bdSJani Nikula 4523e52347bdSJani Nikula pdcchassis= [PARISC,HW] Disable/Enable PDC Chassis Status codes at 4524e52347bdSJani Nikula boot time. 4525e52347bdSJani Nikula Format: { 0 | 1 } 4526e52347bdSJani Nikula See arch/parisc/kernel/pdc_chassis.c 4527e52347bdSJani Nikula 4528e52347bdSJani Nikula percpu_alloc= Select which percpu first chunk allocator to use. 4529e52347bdSJani Nikula Currently supported values are "embed" and "page". 4530e52347bdSJani Nikula Archs may support subset or none of the selections. 4531e52347bdSJani Nikula See comments in mm/percpu.c for details on each 4532e52347bdSJani Nikula allocator. This parameter is primarily for debugging 4533ff61f079SJonathan Corbet and performance comparison. 4534e52347bdSJani Nikula 4535e52347bdSJani Nikula pirq= [SMP,APIC] Manual mp-table setup 4536e52347bdSJani Nikula See Documentation/arch/x86/i386/IO-APIC.rst. 45373ba9b1b8STom Saeger 4538e52347bdSJani Nikula plip= [PPT,NET] Parallel port network link 4539e52347bdSJani Nikula Format: { parport<nr> | timid | 0 } 4540e52347bdSJani Nikula See also Documentation/admin-guide/parport.rst. 4541e52347bdSJani Nikula 4542e52347bdSJani Nikula pmtmr= [X86] Manual setup of pmtmr I/O Port. 45430a4b4327SNicholas Piggin Override pmtimer IOPort with a hex value. 45440a4b4327SNicholas Piggin e.g. pmtmr=0x508 45450a4b4327SNicholas Piggin 45460a4b4327SNicholas Piggin pmu_override= [PPC] Override the PMU. 45470a4b4327SNicholas Piggin This option takes over the PMU facility, so it is no 45480a4b4327SNicholas Piggin longer usable by perf. Setting this option starts the 45490a4b4327SNicholas Piggin PMU counters by setting MMCR0 to 0 (the FC bit is 45500a4b4327SNicholas Piggin cleared). If a number is given, then MMCR1 is set to 4551db96a759SChen Yu that number, otherwise (e.g., 'pmu_override=on'), MMCR1 4552db96a759SChen Yu remains 0. 4553db96a759SChen Yu 4554e52347bdSJani Nikula pm_debug_messages [SUSPEND,KNL] 4555e52347bdSJani Nikula Enable suspend/resume debug messages during boot up. 4556e52347bdSJani Nikula 4557e52347bdSJani Nikula pnp.debug=1 [PNP] 4558e52347bdSJani Nikula Enable PNP debug messages (depends on the 4559e52347bdSJani Nikula CONFIG_PNP_DEBUG_MESSAGES option). Change at run-time 4560e52347bdSJani Nikula via /sys/module/pnp/parameters/debug. We always show 4561e52347bdSJani Nikula current resource usage; turning this on also shows 4562e52347bdSJani Nikula possible settings and some assignment information. 4563e52347bdSJani Nikula 4564e52347bdSJani Nikula pnpacpi= [ACPI] 4565e52347bdSJani Nikula { off } 4566e52347bdSJani Nikula 4567e52347bdSJani Nikula pnpbios= [ISAPNP] 4568e52347bdSJani Nikula { on | off | curr | res | no-curr | no-res } 4569e52347bdSJani Nikula 4570e52347bdSJani Nikula pnp_reserve_irq= 4571e52347bdSJani Nikula [ISAPNP] Exclude IRQs for the autoconfiguration 4572e52347bdSJani Nikula 4573e52347bdSJani Nikula pnp_reserve_dma= 4574e52347bdSJani Nikula [ISAPNP] Exclude DMAs for the autoconfiguration 4575e52347bdSJani Nikula 4576e52347bdSJani Nikula pnp_reserve_io= [ISAPNP] Exclude I/O ports for the autoconfiguration 4577e52347bdSJani Nikula Ranges are in pairs (I/O port base and size). 4578e52347bdSJani Nikula 4579e52347bdSJani Nikula pnp_reserve_mem= 4580e52347bdSJani Nikula [ISAPNP] Exclude memory regions for the 4581e52347bdSJani Nikula autoconfiguration. 4582e52347bdSJani Nikula Ranges are in pairs (memory base and size). 4583e52347bdSJani Nikula 4584e52347bdSJani Nikula ports= [IP_VS_FTP] IPVS ftp helper module 4585e52347bdSJani Nikula Default is 21. 4586e52347bdSJani Nikula Up to 8 (IP_VS_APP_MAX_PORTS) ports 4587c3cbd075SBalbir Singh may be specified. 4588c3cbd075SBalbir Singh Format: <port>,<port>.... 4589c3cbd075SBalbir Singh 4590c3cbd075SBalbir Singh powersave=off [PPC] This option disables power saving features. 4591c3cbd075SBalbir Singh It specifically disables cpuidle and sets the 4592c3cbd075SBalbir Singh platform machine description specific power_save 4593e52347bdSJani Nikula function to NULL. On Idle the CPU just reduces 4594e52347bdSJani Nikula execution priority. 4595e52347bdSJani Nikula 4596e52347bdSJani Nikula ppc_strict_facility_enable 4597e52347bdSJani Nikula [PPC] This option catches any kernel floating point, 4598e52347bdSJani Nikula Altivec, VSX and SPE outside of regions specifically 459907fd1761SCyril Bur allowed (eg kernel_enable_fpu()/kernel_disable_fpu()). 460007fd1761SCyril Bur There is some performance impact when enabling this. 460107fd1761SCyril Bur 460207fd1761SCyril Bur ppc_tm= [PPC] 46036ef869e0SMichal Hocko Format: {"off"} 46046ef869e0SMichal Hocko Disable Hardware Transactional Memory 46056ef869e0SMichal Hocko 46066ef869e0SMichal Hocko preempt= [KNL] 46076ef869e0SMichal Hocko Select preemption mode if you have CONFIG_PREEMPT_DYNAMIC 46086ef869e0SMichal Hocko none - Limited to cond_resched() calls 46096ef869e0SMichal Hocko voluntary - Limited to cond_resched() and might_sleep() calls 4610e52347bdSJani Nikula full - Any section that isn't explicitly preempt disabled 4611e52347bdSJani Nikula can be preempted anytime. 4612e52347bdSJani Nikula 4613e52347bdSJani Nikula print-fatal-signals= 4614e52347bdSJani Nikula [KNL] debug: print fatal signals 4615e52347bdSJani Nikula 4616e52347bdSJani Nikula If enabled, warn about various signal handling 4617e52347bdSJani Nikula related application anomalies: too many signals, 4618e52347bdSJani Nikula too many POSIX.1 timers, fatal signals causing a 4619e52347bdSJani Nikula coredump - etc. 4620e52347bdSJani Nikula 4621e52347bdSJani Nikula If you hit the warning due to signal overflow, 4622e52347bdSJani Nikula you might want to try "ulimit -i unlimited". 4623e52347bdSJani Nikula 4624e52347bdSJani Nikula default: off. 4625e52347bdSJani Nikula 4626e52347bdSJani Nikula printk.always_kmsg_dump= 4627e52347bdSJani Nikula Trigger kmsg_dump for cases other than kernel oops or 4628e52347bdSJani Nikula panics 462910102a89SDmitry Safonov Format: <bool> (1/Y/y=enable, 0/N/n=disable) 463010102a89SDmitry Safonov default: disabled 463110102a89SDmitry Safonov 463210102a89SDmitry Safonov printk.console_no_auto_verbose= 463310102a89SDmitry Safonov Disable console loglevel raise on oops, panic 463410102a89SDmitry Safonov or lockdep-detected issues (only if lock debug is on). 463510102a89SDmitry Safonov With an exception to setups with low baudrate on 463610102a89SDmitry Safonov serial console, keeping this 0 is a good choice 463710102a89SDmitry Safonov in order to provide more debug information. 4638e52347bdSJani Nikula Format: <bool> 4639e52347bdSJani Nikula default: 0 (auto_verbose is enabled) 4640e52347bdSJani Nikula 4641e52347bdSJani Nikula printk.devkmsg={on,off,ratelimit} 4642e52347bdSJani Nikula Control writing to /dev/kmsg. 4643e52347bdSJani Nikula on - unlimited logging to /dev/kmsg from userspace 4644e52347bdSJani Nikula off - logging to /dev/kmsg disabled 4645e52347bdSJani Nikula ratelimit - ratelimit the logging 4646e52347bdSJani Nikula Default: ratelimit 4647e52347bdSJani Nikula 4648e52347bdSJani Nikula printk.time= Show timing data prefixed to each printk message line 4649e52347bdSJani Nikula Format: <bool> (1/Y/y=enable, 0/N/n=disable) 4650e52347bdSJani Nikula 4651e52347bdSJani Nikula processor.max_cstate= [HW,ACPI] 4652e52347bdSJani Nikula Limit processor to maximum C-state 4653e52347bdSJani Nikula max_cstate=9 overrides any DMI blacklist limit. 4654e52347bdSJani Nikula 4655e52347bdSJani Nikula processor.nocst [HW,ACPI] 4656e52347bdSJani Nikula Ignore the _CST method to determine C-states, 4657e7e61fc0SRandy Dunlap instead using the legacy FADT method 4658e7e61fc0SRandy Dunlap 4659e7e61fc0SRandy Dunlap profile= [KNL] Enable kernel profiling via /proc/profile 4660e52347bdSJani Nikula Format: [<profiletype>,]<number> 4661e52347bdSJani Nikula Param: <profiletype>: "schedule", "sleep", or "kvm" 4662e52347bdSJani Nikula [defaults to kernel profiling] 4663e52347bdSJani Nikula Param: "schedule" - profile schedule points. 4664e7e61fc0SRandy Dunlap Param: "sleep" - profile D-state sleeping (millisecs). 4665e7e61fc0SRandy Dunlap Requires CONFIG_SCHEDSTATS 4666e52347bdSJani Nikula Param: "kvm" - profile VM exits. 46676b99e6e6SRandy Dunlap Param: <number> - step/bucket size as a power of 2 for 4668e52347bdSJani Nikula statistical time based profiling. 4669ecdc5d84SVasily Gorbik 4670ecdc5d84SVasily Gorbik prompt_ramdisk= [RAM] [Deprecated] 4671ecdc5d84SVasily Gorbik 4672ecdc5d84SVasily Gorbik prot_virt= [S390] enable hosting protected virtual machines 4673ecdc5d84SVasily Gorbik isolated from the hypervisor (if hardware supports 4674e0c27447SJohannes Weiner that). 4675e0c27447SJohannes Weiner Format: <bool> 4676e0c27447SJohannes Weiner 4677e0c27447SJohannes Weiner psi= [KNL] Enable or disable pressure stall information 4678e52347bdSJani Nikula tracking. 4679e52347bdSJani Nikula Format: <bool> 4680e52347bdSJani Nikula 4681e52347bdSJani Nikula psmouse.proto= [HW,MOUSE] Highest PS2 mouse protocol extension to 4682e52347bdSJani Nikula probe for; one of (bare|imps|exps|lifebook|any). 4683e52347bdSJani Nikula psmouse.rate= [HW,MOUSE] Set desired mouse report rate, in reports 4684e52347bdSJani Nikula per second. 4685e52347bdSJani Nikula psmouse.resetafter= [HW,MOUSE] 4686e52347bdSJani Nikula Try to reset the device after so many bad packets 4687e52347bdSJani Nikula (0 = never). 4688e52347bdSJani Nikula psmouse.resolution= 4689e52347bdSJani Nikula [HW,MOUSE] Set desired mouse resolution, in dpi. 4690e52347bdSJani Nikula psmouse.smartscroll= 4691e52347bdSJani Nikula [HW,MOUSE] Controls Logitech smartscroll autorepeat. 4692e52347bdSJani Nikula 0 = disabled, 1 = enabled (default). 4693be3a5b0eSRandy Dunlap 469401c9b17bSDave Hansen pstore.backend= Specify the name of the pstore backend to use 469501c9b17bSDave Hansen 469601c9b17bSDave Hansen pti= [X86-64] Control Page Table Isolation of user and 469701c9b17bSDave Hansen kernel address spaces. Disabling this feature 469801c9b17bSDave Hansen removes hardening, but improves performance of 469901c9b17bSDave Hansen system calls and interrupts. 470001c9b17bSDave Hansen 470101c9b17bSDave Hansen on - unconditionally enable 470201c9b17bSDave Hansen off - unconditionally disable 470301c9b17bSDave Hansen auto - kernel detects whether your CPU model is 470401c9b17bSDave Hansen vulnerable to issues that PTI mitigates 4705e52347bdSJani Nikula 4706e52347bdSJani Nikula Not specifying this option is equivalent to pti=auto. 4707e52347bdSJani Nikula 4708e52347bdSJani Nikula pty.legacy_count= 4709e52347bdSJani Nikula [KNL] Number of legacy pty's. Overwrites compiled-in 4710e52347bdSJani Nikula default number. 4711e52347bdSJani Nikula 4712e52347bdSJani Nikula quiet [KNL] Disable most log messages 471342551b8dSRandy Dunlap 471442551b8dSRandy Dunlap r128= [HW,DRM] 471542551b8dSRandy Dunlap 471642551b8dSRandy Dunlap radix_hcall_invalidate=on [PPC/PSERIES] 4717e52347bdSJani Nikula Disable RADIX GTSE feature and use hcall for TLB 4718e52347bdSJani Nikula invalidate. 4719e52347bdSJani Nikula 4720e52347bdSJani Nikula raid= [HW,RAID] 4721e7751617SMauro Carvalho Chehab See Documentation/admin-guide/md.rst. 4722e52347bdSJani Nikula 47236b99e6e6SRandy Dunlap ramdisk_size= [RAM] Sizes of RAM disks in kilobytes 47246b99e6e6SRandy Dunlap See Documentation/admin-guide/blockdev/ramdisk.rst. 4725b9b01a56SJason A. Donenfeld 4726b9b01a56SJason A. Donenfeld ramdisk_start= [RAM] RAM disk image start address 4727b9b01a56SJason A. Donenfeld 4728b9b01a56SJason A. Donenfeld random.trust_cpu=off 47299b254366SKees Cook [KNL] Disable trusting the use of the CPU's 4730b9b01a56SJason A. Donenfeld random number generator (if available) to 4731b9b01a56SJason A. Donenfeld initialize the kernel's RNG. 4732b9b01a56SJason A. Donenfeld 4733b9b01a56SJason A. Donenfeld random.trust_bootloader=off 4734d97c68d1SJason A. Donenfeld [KNL] Disable trusting the use of the a seed 473539218ff4SKees Cook passed by the bootloader (if available) to 473639218ff4SKees Cook initialize the kernel's RNG. 473739218ff4SKees Cook 473839218ff4SKees Cook randomize_kstack_offset= 473939218ff4SKees Cook [KNL] Enable or disable kernel stack offset 474039218ff4SKees Cook randomization, which provides roughly 5 bits of 474139218ff4SKees Cook entropy, frustrating memory corruption attacks 474239218ff4SKees Cook that depend on stack address determinism or 474339218ff4SKees Cook cross-syscall address exposures. This is only 474439218ff4SKees Cook available on architectures that have defined 474539218ff4SKees Cook CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET. 4746011d8261SBorislav Petkov Format: <bool> (1/Y/y=enable, 0/N/n=disable) 4747011d8261SBorislav Petkov Default is CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT. 4748011d8261SBorislav Petkov 4749011d8261SBorislav Petkov ras=option[,option,...] [KNL] RAS-specific options 4750011d8261SBorislav Petkov 4751011d8261SBorislav Petkov cec_disable [X86] 4752d2cf0854SFrederic Weisbecker Disable the Correctable Errors Collector, 4753d2cf0854SFrederic Weisbecker see CONFIG_RAS_CEC help text. 4754d2cf0854SFrederic Weisbecker 4755e52347bdSJani Nikula rcu_nocbs[=cpu-list] 4756d2cf0854SFrederic Weisbecker [KNL] The optional argument is a cpu list, 4757d2cf0854SFrederic Weisbecker as described above. 4758d2cf0854SFrederic Weisbecker 4759d2cf0854SFrederic Weisbecker In kernels built with CONFIG_RCU_NOCB_CPU=y, 4760d2cf0854SFrederic Weisbecker enable the no-callback CPU mode, which prevents 4761d2cf0854SFrederic Weisbecker such CPUs' callbacks from being invoked in 4762d2cf0854SFrederic Weisbecker softirq context. Invocation of such CPUs' RCU 4763d2cf0854SFrederic Weisbecker callbacks will instead be offloaded to "rcuox/N" 4764d2cf0854SFrederic Weisbecker kthreads created for that purpose, where "x" is 4765d2cf0854SFrederic Weisbecker "p" for RCU-preempt, "s" for RCU-sched, and "g" 4766d2cf0854SFrederic Weisbecker for the kthreads that mediate grace periods; and 4767d2cf0854SFrederic Weisbecker "N" is the CPU number. This reduces OS jitter on 4768d2cf0854SFrederic Weisbecker the offloaded CPUs, which can be useful for HPC 4769d2cf0854SFrederic Weisbecker and real-time workloads. It can also improve 4770d2cf0854SFrederic Weisbecker energy efficiency for asymmetric multiprocessors. 4771d2cf0854SFrederic Weisbecker 4772d2cf0854SFrederic Weisbecker If a cpulist is passed as an argument, the specified 4773d2cf0854SFrederic Weisbecker list of CPUs is set to no-callback mode from boot. 4774d2cf0854SFrederic Weisbecker 4775d2cf0854SFrederic Weisbecker Otherwise, if the '=' sign and the cpulist 4776e52347bdSJani Nikula arguments are omitted, no CPU will be set to 4777b37a667cSJoel Fernandes no-callback mode from boot but the mode may be 4778b37a667cSJoel Fernandes toggled at runtime via cpusets. 4779b37a667cSJoel Fernandes 4780e52347bdSJani Nikula Note that this argument takes precedence over 4781e52347bdSJani Nikula the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. 4782e52347bdSJani Nikula 4783e52347bdSJani Nikula rcu_nocb_poll [KNL] 4784e52347bdSJani Nikula Rather than requiring that offloaded CPUs 4785e52347bdSJani Nikula (specified by rcu_nocbs= above) explicitly 4786e52347bdSJani Nikula awaken the corresponding "rcuoN" kthreads, 4787e52347bdSJani Nikula make these kthreads poll for callbacks. 4788e52347bdSJani Nikula This improves the real-time response for the 4789e52347bdSJani Nikula offloaded CPUs by relieving them of the need to 4790e52347bdSJani Nikula wake up the corresponding kthread, but degrades 4791e52347bdSJani Nikula energy efficiency by requiring that the kthreads 4792e52347bdSJani Nikula periodically wake up to do the polling. 4793e52347bdSJani Nikula 4794e52347bdSJani Nikula rcutree.blimit= [KNL] 4795e52347bdSJani Nikula Set maximum number of finished RCU callbacks to 4796e52347bdSJani Nikula process in one batch. 4797e52347bdSJani Nikula 4798e52347bdSJani Nikula rcutree.dump_tree= [KNL] 4799e52347bdSJani Nikula Dump the structure of the rcu_node combining tree 4800e52347bdSJani Nikula out at early boot. This is used for diagnostic 4801e52347bdSJani Nikula purposes, to verify correct tree setup. 480290040c9eSPaul E. McKenney 4803e52347bdSJani Nikula rcutree.gp_cleanup_delay= [KNL] 4804e52347bdSJani Nikula Set the number of jiffies to delay each step of 4805e52347bdSJani Nikula RCU grace-period cleanup. 480690040c9eSPaul E. McKenney 4807e52347bdSJani Nikula rcutree.gp_init_delay= [KNL] 4808e52347bdSJani Nikula Set the number of jiffies to delay each step of 4809e52347bdSJani Nikula RCU grace-period initialization. 4810e52347bdSJani Nikula 4811e52347bdSJani Nikula rcutree.gp_preinit_delay= [KNL] 481290040c9eSPaul E. McKenney Set the number of jiffies to delay each step of 4813e52347bdSJani Nikula RCU grace-period pre-initialization, that is, 4814e52347bdSJani Nikula the propagation of recent CPU-hotplug changes up 4815e52347bdSJani Nikula the rcu_node combining tree. 4816e52347bdSJani Nikula 4817e52347bdSJani Nikula rcutree.jiffies_till_first_fqs= [KNL] 4818e52347bdSJani Nikula Set delay from grace-period initialization to 4819e52347bdSJani Nikula first attempt to force quiescent states. 4820e52347bdSJani Nikula Units are jiffies, minimum value is zero, 4821e52347bdSJani Nikula and maximum value is HZ. 4822e52347bdSJani Nikula 4823e52347bdSJani Nikula rcutree.jiffies_till_next_fqs= [KNL] 4824e52347bdSJani Nikula Set delay between subsequent attempts to force 48251a4762b9SPaul E. McKenney quiescent states. Units are jiffies, minimum 48261a4762b9SPaul E. McKenney value is one, and maximum value is HZ. 48271a4762b9SPaul E. McKenney 48281a4762b9SPaul E. McKenney rcutree.jiffies_till_sched_qs= [KNL] 48291a4762b9SPaul E. McKenney Set required age in jiffies for a 48301a4762b9SPaul E. McKenney given grace period before RCU starts 48311a4762b9SPaul E. McKenney soliciting quiescent-state help from 48321a4762b9SPaul E. McKenney rcu_note_context_switch() and cond_resched(). 48331a4762b9SPaul E. McKenney If not specified, the kernel will calculate 48341a4762b9SPaul E. McKenney a value based on the most recent settings 48351a4762b9SPaul E. McKenney of rcutree.jiffies_till_first_fqs 48361a4762b9SPaul E. McKenney and rcutree.jiffies_till_next_fqs. 48371a4762b9SPaul E. McKenney This calculated value may be viewed in 48381a4762b9SPaul E. McKenney rcutree.jiffies_to_sched_qs. Any attempt to set 4839e52347bdSJani Nikula rcutree.jiffies_to_sched_qs will be cheerfully 4840e52347bdSJani Nikula overwritten. 4841e52347bdSJani Nikula 4842e52347bdSJani Nikula rcutree.kthread_prio= [KNL,BOOT] 4843e52347bdSJani Nikula Set the SCHED_FIFO priority of the RCU per-CPU 4844e52347bdSJani Nikula kthreads (rcuc/N). This value is also used for 4845e52347bdSJani Nikula the priority of the RCU boost threads (rcub/N) 4846e52347bdSJani Nikula and for the RCU grace-period kthreads (rcu_bh, 4847e52347bdSJani Nikula rcu_preempt, and rcu_sched). If RCU_BOOST is 4848e52347bdSJani Nikula set, valid values are 1-99 and the default is 1 4849a469948bSAlison Chaiken (the least-favored priority). Otherwise, when 4850a469948bSAlison Chaiken RCU_BOOST is not set, valid values are 0-99 and 4851e52347bdSJani Nikula the default is zero (non-realtime operation). 485289f7f291SPaul E. McKenney When RCU_NOCB_CPU is set, also adjust the 485389f7f291SPaul E. McKenney priority of NOCB callback kthreads. 485489f7f291SPaul E. McKenney 485589f7f291SPaul E. McKenney rcutree.nocb_nobypass_lim_per_jiffy= [KNL] 485689f7f291SPaul E. McKenney On callback-offloaded (rcu_nocbs) CPUs, 485789f7f291SPaul E. McKenney RCU reduces the lock contention that would 485889f7f291SPaul E. McKenney otherwise be caused by callback floods through 485989f7f291SPaul E. McKenney use of the ->nocb_bypass list. However, in the 486089f7f291SPaul E. McKenney common non-flooded case, RCU queues directly to 486189f7f291SPaul E. McKenney the main ->cblist in order to avoid the extra 486289f7f291SPaul E. McKenney overhead of the ->nocb_bypass list and its lock. 486389f7f291SPaul E. McKenney But if there are too many callbacks queued during 486489f7f291SPaul E. McKenney a single jiffy, RCU pre-queues the callbacks into 4865e52347bdSJani Nikula the ->nocb_bypass queue. The definition of "too 4866e52347bdSJani Nikula many" is supplied by this kernel boot parameter. 4867e52347bdSJani Nikula 4868e52347bdSJani Nikula rcutree.qhimark= [KNL] 4869e52347bdSJani Nikula Set threshold of queued RCU callbacks beyond which 4870e52347bdSJani Nikula batch limiting is disabled. 4871e52347bdSJani Nikula 4872e52347bdSJani Nikula rcutree.qlowmark= [KNL] 4873b2b00ddfSPaul E. McKenney Set threshold of queued RCU callbacks below which 4874b2b00ddfSPaul E. McKenney batch limiting is re-enabled. 4875b2b00ddfSPaul E. McKenney 4876b2b00ddfSPaul E. McKenney rcutree.qovld= [KNL] 4877b2b00ddfSPaul E. McKenney Set threshold of queued RCU callbacks beyond which 4878b2b00ddfSPaul E. McKenney RCU's force-quiescent-state scan will aggressively 4879b2b00ddfSPaul E. McKenney enlist help from cond_resched() and sched IPIs to 4880b2b00ddfSPaul E. McKenney help CPUs more quickly reach quiescent states. 4881b2b00ddfSPaul E. McKenney Set to less than zero to make this be set based 488251823ca6SPaul E. McKenney on rcutree.qhimark at boot time and to zero to 488351823ca6SPaul E. McKenney disable more aggressive help enlistment. 488451823ca6SPaul E. McKenney 488551823ca6SPaul E. McKenney rcutree.rcu_delay_page_cache_fill_msec= [KNL] 488651823ca6SPaul E. McKenney Set the page-cache refill delay (in milliseconds) 488751823ca6SPaul E. McKenney in response to low-memory conditions. The range 488851823ca6SPaul E. McKenney of permitted values is in the range 0:100000. 488951823ca6SPaul E. McKenney 489051823ca6SPaul E. McKenney rcutree.rcu_divisor= [KNL] 489151823ca6SPaul E. McKenney Set the shift-right count to use to compute 489251823ca6SPaul E. McKenney the callback-invocation batch limit bl from 489351823ca6SPaul E. McKenney the number of callbacks queued on this CPU. 489451823ca6SPaul E. McKenney The result will be bounded below by the value of 489551823ca6SPaul E. McKenney the rcutree.blimit kernel parameter. Every bl 489651823ca6SPaul E. McKenney callbacks, the softirq handler will exit in 489751823ca6SPaul E. McKenney order to allow the CPU to do other work. 489851823ca6SPaul E. McKenney 489951823ca6SPaul E. McKenney Please note that this callback-invocation batch 490051823ca6SPaul E. McKenney limit applies only to non-offloaded callback 490151823ca6SPaul E. McKenney invocation. Offloaded callbacks are instead 490251823ca6SPaul E. McKenney invoked in the context of an rcuoc kthread, which 490351823ca6SPaul E. McKenney scheduler will preempt as it does any other task. 490451823ca6SPaul E. McKenney 490551823ca6SPaul E. McKenney rcutree.rcu_fanout_exact= [KNL] 490651823ca6SPaul E. McKenney Disable autobalancing of the rcu_node combining 490751823ca6SPaul E. McKenney tree. This is used by rcutorture, and might 490851823ca6SPaul E. McKenney possibly be useful for architectures having high 490951823ca6SPaul E. McKenney cache-to-cache transfer latencies. 491051823ca6SPaul E. McKenney 491151823ca6SPaul E. McKenney rcutree.rcu_fanout_leaf= [KNL] 491251823ca6SPaul E. McKenney Change the number of CPUs assigned to each 491351823ca6SPaul E. McKenney leaf rcu_node structure. Useful for very 491451823ca6SPaul E. McKenney large systems, which will choose the value 64, 491551823ca6SPaul E. McKenney and for NUMA systems with large remote-access 491651823ca6SPaul E. McKenney latencies, which will choose a value aligned 491751823ca6SPaul E. McKenney with the appropriate hardware boundaries. 491851823ca6SPaul E. McKenney 491951823ca6SPaul E. McKenney rcutree.rcu_min_cached_objs= [KNL] 492051823ca6SPaul E. McKenney Minimum number of objects which are cached and 492151823ca6SPaul E. McKenney maintained per one CPU. Object size is equal 492251823ca6SPaul E. McKenney to PAGE_SIZE. The cache allows to reduce the 492351823ca6SPaul E. McKenney pressure to page allocator, also it makes the 492451823ca6SPaul E. McKenney whole algorithm to behave better in low memory 492551823ca6SPaul E. McKenney condition. 492651823ca6SPaul E. McKenney 492751823ca6SPaul E. McKenney rcutree.rcu_nocb_gp_stride= [KNL] 492851823ca6SPaul E. McKenney Set the number of NOCB callback kthreads in 492951823ca6SPaul E. McKenney each group, which defaults to the square root 493051823ca6SPaul E. McKenney of the number of CPUs. Larger numbers reduce 493151823ca6SPaul E. McKenney the wakeup overhead on the global grace-period 4932e3c50dfbSPaul E. McKenney kthread, but increases that same overhead on 4933e3c50dfbSPaul E. McKenney each group's NOCB grace-period kthread. 4934e3c50dfbSPaul E. McKenney 4935e3c50dfbSPaul E. McKenney rcutree.rcu_kick_kthreads= [KNL] 4936e3c50dfbSPaul E. McKenney Cause the grace-period kthread to get an extra 4937e3c50dfbSPaul E. McKenney wake_up() if it sleeps three times longer than 4938e3c50dfbSPaul E. McKenney it should at force-quiescent-state time. 4939fb611249SPaul E. McKenney This wake_up() will be accompanied by a 4940fb611249SPaul E. McKenney WARN_ONCE() splat and an ftrace_dump(). 4941fb611249SPaul E. McKenney 4942fb611249SPaul E. McKenney rcutree.rcu_resched_ns= [KNL] 4943fb611249SPaul E. McKenney Limit the time spend invoking a batch of RCU 4944fb611249SPaul E. McKenney callbacks to the specified number of nanoseconds. 4945fb611249SPaul E. McKenney By default, this limit is checked only once 49463d29aaf1SPaul E. McKenney every 32 callbacks in order to limit the pain 49473d29aaf1SPaul E. McKenney inflicted by local_clock() overhead. 49483d29aaf1SPaul E. McKenney 49493d29aaf1SPaul E. McKenney rcutree.rcu_unlock_delay= [KNL] 49503d29aaf1SPaul E. McKenney In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, 49513d29aaf1SPaul E. McKenney this specifies an rcu_read_unlock()-time delay 49523d29aaf1SPaul E. McKenney in microseconds. This defaults to zero. 49533d29aaf1SPaul E. McKenney Larger delays increase the probability of 49543d29aaf1SPaul E. McKenney catching RCU pointer leaks, that is, buggy use 49552ccaff10SPaul E. McKenney of RCU-protected pointers after the relevant 49562ccaff10SPaul E. McKenney rcu_read_unlock() has completed. 49572ccaff10SPaul E. McKenney 49582ccaff10SPaul E. McKenney rcutree.sysrq_rcu= [KNL] 49592ccaff10SPaul E. McKenney Commandeer a sysrq key to dump out Tree RCU's 496051823ca6SPaul E. McKenney rcu_node tree with an eye towards determining 496151823ca6SPaul E. McKenney why a new grace period has not yet started. 496251823ca6SPaul E. McKenney 496351823ca6SPaul E. McKenney rcutree.use_softirq= [KNL] 496451823ca6SPaul E. McKenney If set to zero, move all RCU_SOFTIRQ processing to 496551823ca6SPaul E. McKenney per-CPU rcuc kthreads. Defaults to a non-zero 496651823ca6SPaul E. McKenney value, meaning that RCU_SOFTIRQ is used by default. 496751823ca6SPaul E. McKenney Specify rcutree.use_softirq=0 to use rcuc kthreads. 496851823ca6SPaul E. McKenney 496951823ca6SPaul E. McKenney But note that CONFIG_PREEMPT_RT=y kernels disable 49704e88ec4aSPaul E. McKenney this kernel boot parameter, forcibly setting it 4971881ed593SPaul E. McKenney to zero. 4972881ed593SPaul E. McKenney 4973881ed593SPaul E. McKenney rcuscale.gp_async= [KNL] 49744e88ec4aSPaul E. McKenney Measure performance of asynchronous 4975881ed593SPaul E. McKenney grace-period primitives such as call_rcu(). 4976881ed593SPaul E. McKenney 4977881ed593SPaul E. McKenney rcuscale.gp_async_max= [KNL] 4978881ed593SPaul E. McKenney Specify the maximum number of outstanding 4979881ed593SPaul E. McKenney callbacks per writer thread. When a writer 4980881ed593SPaul E. McKenney thread exceeds this limit, it invokes the 49814e88ec4aSPaul E. McKenney corresponding flavor of rcu_barrier() to allow 4982e52347bdSJani Nikula previously posted callbacks to drain. 4983e52347bdSJani Nikula 4984e52347bdSJani Nikula rcuscale.gp_exp= [KNL] 49854e88ec4aSPaul E. McKenney Measure performance of expedited synchronous 4986e52347bdSJani Nikula grace-period primitives. 4987e52347bdSJani Nikula 4988e52347bdSJani Nikula rcuscale.holdoff= [KNL] 4989e52347bdSJani Nikula Set test-start holdoff period. The purpose of 4990e52347bdSJani Nikula this parameter is to delay the start of the 49912d7b2b34SPaul E. McKenney test until boot completes in order to avoid 49922d7b2b34SPaul E. McKenney interference. 49932d7b2b34SPaul E. McKenney 49942d7b2b34SPaul E. McKenney rcuscale.kfree_by_call_rcu= [KNL] 49952d7b2b34SPaul E. McKenney In kernels built with CONFIG_RCU_LAZY=y, test 49962d7b2b34SPaul E. McKenney call_rcu() instead of kfree_rcu(). 49972d7b2b34SPaul E. McKenney 49982d7b2b34SPaul E. McKenney rcuscale.kfree_mult= [KNL] 49992d7b2b34SPaul E. McKenney Instead of allocating an object of size kfree_obj, 50004e88ec4aSPaul E. McKenney allocate one of kfree_mult * sizeof(kfree_obj). 5001e6e78b00SJoel Fernandes (Google) Defaults to 1. 5002e6e78b00SJoel Fernandes (Google) 5003686fe1bfSUladzislau Rezki (Sony) rcuscale.kfree_rcu_test= [KNL] 5004686fe1bfSUladzislau Rezki (Sony) Set to measure performance of kfree_rcu() flooding. 5005686fe1bfSUladzislau Rezki (Sony) 5006686fe1bfSUladzislau Rezki (Sony) rcuscale.kfree_rcu_test_double= [KNL] 5007686fe1bfSUladzislau Rezki (Sony) Test the double-argument variant of kfree_rcu(). 5008686fe1bfSUladzislau Rezki (Sony) If this parameter has the same value as 5009686fe1bfSUladzislau Rezki (Sony) rcuscale.kfree_rcu_test_single, both the single- 5010686fe1bfSUladzislau Rezki (Sony) and double-argument variants are tested. 5011686fe1bfSUladzislau Rezki (Sony) 5012686fe1bfSUladzislau Rezki (Sony) rcuscale.kfree_rcu_test_single= [KNL] 5013686fe1bfSUladzislau Rezki (Sony) Test the single-argument variant of kfree_rcu(). 5014686fe1bfSUladzislau Rezki (Sony) If this parameter has the same value as 50154e88ec4aSPaul E. McKenney rcuscale.kfree_rcu_test_double, both the single- 5016e6e78b00SJoel Fernandes (Google) and double-argument variants are tested. 5017e6e78b00SJoel Fernandes (Google) 50184e88ec4aSPaul E. McKenney rcuscale.kfree_nthreads= [KNL] 5019e6e78b00SJoel Fernandes (Google) The number of threads running loops of kfree_rcu(). 5020e6e78b00SJoel Fernandes (Google) 50214e88ec4aSPaul E. McKenney rcuscale.kfree_alloc_num= [KNL] 50224e88ec4aSPaul E. McKenney Number of allocations and frees done in an iteration. 5023e6e78b00SJoel Fernandes (Google) 5024e6e78b00SJoel Fernandes (Google) rcuscale.kfree_loops= [KNL] 50257221f493SPaul E. McKenney Number of loops doing rcuscale.kfree_alloc_num number 50267221f493SPaul E. McKenney of allocations and frees. 50277221f493SPaul E. McKenney 50287221f493SPaul E. McKenney rcuscale.minruntime= [KNL] 50297221f493SPaul E. McKenney Set the minimum test run time in seconds. This 50307221f493SPaul E. McKenney does not affect the data-collection interval, 50314e88ec4aSPaul E. McKenney but instead allows better measurement of things 5032e52347bdSJani Nikula like CPU consumption. 5033e52347bdSJani Nikula 5034e52347bdSJani Nikula rcuscale.nreaders= [KNL] 5035e52347bdSJani Nikula Set number of RCU readers. The value -1 selects 5036e52347bdSJani Nikula N, where N is the number of CPUs. A value 5037e52347bdSJani Nikula "n" less than -1 selects N-n+1, where N is again 5038e52347bdSJani Nikula the number of CPUs. For example, -2 selects N 5039e52347bdSJani Nikula (the number of CPUs), -3 selects N+1, and so on. 50404e88ec4aSPaul E. McKenney A value of "n" less than or equal to -N selects 5041e52347bdSJani Nikula a single reader. 50424e88ec4aSPaul E. McKenney 5043e52347bdSJani Nikula rcuscale.nwriters= [KNL] 5044e52347bdSJani Nikula Set number of RCU writers. The values operate 50452d7b2b34SPaul E. McKenney the same as for rcuscale.nreaders. 5046820687a7SPaul E. McKenney N, where N is the number of CPUs 5047820687a7SPaul E. McKenney 50484e88ec4aSPaul E. McKenney rcuscale.scale_type= [KNL] 5049e52347bdSJani Nikula Specify the RCU implementation to test. 5050e52347bdSJani Nikula 5051e52347bdSJani Nikula rcuscale.shutdown= [KNL] 5052e52347bdSJani Nikula Shut the system down after performance tests 50534e88ec4aSPaul E. McKenney complete. This is useful for hands-off automated 5054e52347bdSJani Nikula testing. 5055e52347bdSJani Nikula 50564e88ec4aSPaul E. McKenney rcuscale.verbose= [KNL] 5057820687a7SPaul E. McKenney Enable additional printk() statements. 5058820687a7SPaul E. McKenney 5059820687a7SPaul E. McKenney rcuscale.writer_holdoff= [KNL] 5060820687a7SPaul E. McKenney Write-side holdoff between grace periods, 50612226f3dcSPaul E. McKenney in microseconds. The default of zero says 50622226f3dcSPaul E. McKenney no holdoff. 50632226f3dcSPaul E. McKenney 50642226f3dcSPaul E. McKenney rcuscale.writer_holdoff_jiffies= [KNL] 50652226f3dcSPaul E. McKenney Additional write-side holdoff between grace 5066e52347bdSJani Nikula periods, but in jiffies. The default of zero 5067e52347bdSJani Nikula says no holdoff. 5068e52347bdSJani Nikula 5069e52347bdSJani Nikula rcutorture.fqs_duration= [KNL] 5070e52347bdSJani Nikula Set duration of force_quiescent_state bursts 5071e52347bdSJani Nikula in microseconds. 5072e52347bdSJani Nikula 5073e52347bdSJani Nikula rcutorture.fqs_holdoff= [KNL] 5074e52347bdSJani Nikula Set holdoff time within force_quiescent_state bursts 5075e52347bdSJani Nikula in microseconds. 5076e52347bdSJani Nikula 5077e52347bdSJani Nikula rcutorture.fqs_stutter= [KNL] 5078ed8f6fb2SPaul E. McKenney Set wait time between force_quiescent_state bursts 507982e31003SPaul E. McKenney in seconds. 508082e31003SPaul E. McKenney 5081ed8f6fb2SPaul E. McKenney rcutorture.fwd_progress= [KNL] 508282e31003SPaul E. McKenney Specifies the number of kthreads to be used 508382e31003SPaul E. McKenney for RCU grace-period forward-progress testing 508482e31003SPaul E. McKenney for the types of RCU supporting this notion. 5085ed8f6fb2SPaul E. McKenney Defaults to 1 kthread, values less than zero or 5086ed8f6fb2SPaul E. McKenney greater than the number of CPUs cause the number 5087ed8f6fb2SPaul E. McKenney of CPUs to be used. 5088ed8f6fb2SPaul E. McKenney 5089ed8f6fb2SPaul E. McKenney rcutorture.fwd_progress_div= [KNL] 5090ed8f6fb2SPaul E. McKenney Specify the fraction of a CPU-stall-warning 5091ed8f6fb2SPaul E. McKenney period to do tight-loop forward-progress testing. 5092ed8f6fb2SPaul E. McKenney 5093ed8f6fb2SPaul E. McKenney rcutorture.fwd_progress_holdoff= [KNL] 5094ed8f6fb2SPaul E. McKenney Number of seconds to wait between successive 5095ed8f6fb2SPaul E. McKenney forward-progress tests. 5096ed8f6fb2SPaul E. McKenney 5097ed8f6fb2SPaul E. McKenney rcutorture.fwd_progress_need_resched= [KNL] 5098ed8f6fb2SPaul E. McKenney Enclose cond_resched() calls within checks for 5099e52347bdSJani Nikula need_resched() during tight-loop forward-progress 5100e52347bdSJani Nikula testing. 5101e52347bdSJani Nikula 5102e52347bdSJani Nikula rcutorture.gp_cond= [KNL] 5103e52347bdSJani Nikula Use conditional/asynchronous update-side 5104e52347bdSJani Nikula primitives, if available. 5105e52347bdSJani Nikula 5106e52347bdSJani Nikula rcutorture.gp_exp= [KNL] 5107e52347bdSJani Nikula Use expedited update-side primitives, if available. 5108e52347bdSJani Nikula 5109e52347bdSJani Nikula rcutorture.gp_normal= [KNL] 5110e52347bdSJani Nikula Use normal (non-expedited) asynchronous 5111e52347bdSJani Nikula update-side primitives, if available. 5112e52347bdSJani Nikula 5113e52347bdSJani Nikula rcutorture.gp_sync= [KNL] 5114e52347bdSJani Nikula Use normal (non-expedited) synchronous 5115e52347bdSJani Nikula update-side primitives, if available. If all 5116e52347bdSJani Nikula of rcutorture.gp_cond=, rcutorture.gp_exp=, 5117e52347bdSJani Nikula rcutorture.gp_normal=, and rcutorture.gp_sync= 5118d6855142SPaul E. McKenney are zero, rcutorture acts as if is interpreted 5119d6855142SPaul E. McKenney they are all non-zero. 5120d6855142SPaul E. McKenney 5121d6855142SPaul E. McKenney rcutorture.irqreader= [KNL] 5122d6855142SPaul E. McKenney Run RCU readers from irq handlers, or, more 5123d6855142SPaul E. McKenney accurately, from a timer handler. Not all RCU 5124d6855142SPaul E. McKenney flavors take kindly to this sort of thing. 5125d6855142SPaul E. McKenney 5126d6855142SPaul E. McKenney rcutorture.leakpointer= [KNL] 5127d6855142SPaul E. McKenney Leak an RCU-protected pointer out of the reader. 5128d6855142SPaul E. McKenney This can of course result in splats, and is 5129d6855142SPaul E. McKenney intended to test the ability of things like 5130e52347bdSJani Nikula CONFIG_RCU_STRICT_GRACE_PERIOD=y to detect 5131e52347bdSJani Nikula such leaks. 5132e52347bdSJani Nikula 5133e52347bdSJani Nikula rcutorture.n_barrier_cbs= [KNL] 5134e52347bdSJani Nikula Set callbacks/threads for rcu_barrier() testing. 5135e52347bdSJani Nikula 5136e52347bdSJani Nikula rcutorture.nfakewriters= [KNL] 5137e52347bdSJani Nikula Set number of concurrent RCU writers. These just 51382c4319bdSPaul E. McKenney stress RCU, they don't participate in the actual 51392c4319bdSPaul E. McKenney test, hence the "fake". 51402c4319bdSPaul E. McKenney 51412c4319bdSPaul E. McKenney rcutorture.nocbs_nthreads= [KNL] 51422c4319bdSPaul E. McKenney Set number of RCU callback-offload togglers. 51432c4319bdSPaul E. McKenney Zero (the default) disables toggling. 51442c4319bdSPaul E. McKenney 51452c4319bdSPaul E. McKenney rcutorture.nocbs_toggle= [KNL] 5146e52347bdSJani Nikula Set the delay in milliseconds between successive 5147e52347bdSJani Nikula callback-offload toggling attempts. 5148e52347bdSJani Nikula 5149e52347bdSJani Nikula rcutorture.nreaders= [KNL] 5150e52347bdSJani Nikula Set number of RCU readers. The value -1 selects 5151e52347bdSJani Nikula N-1, where N is the number of CPUs. A value 5152e52347bdSJani Nikula "n" less than -1 selects N-n-2, where N is again 5153e52347bdSJani Nikula the number of CPUs. For example, -2 selects N 5154e52347bdSJani Nikula (the number of CPUs), -3 selects N+1, and so on. 5155e52347bdSJani Nikula 5156e52347bdSJani Nikula rcutorture.object_debug= [KNL] 5157e52347bdSJani Nikula Enable debug-object double-call_rcu() testing. 5158e52347bdSJani Nikula 5159e52347bdSJani Nikula rcutorture.onoff_holdoff= [KNL] 5160028be12bSPaul E. McKenney Set time (s) after boot for CPU-hotplug testing. 5161028be12bSPaul E. McKenney 5162e52347bdSJani Nikula rcutorture.onoff_interval= [KNL] 51634a5f133cSPaul E. McKenney Set time (jiffies) between CPU-hotplug operations, 51644a5f133cSPaul E. McKenney or zero to disable CPU-hotplug testing. 51654a5f133cSPaul E. McKenney 51664a5f133cSPaul E. McKenney rcutorture.read_exit= [KNL] 51674a5f133cSPaul E. McKenney Set the number of read-then-exit kthreads used 51684a5f133cSPaul E. McKenney to test the interaction of RCU updaters and 51694a5f133cSPaul E. McKenney task-exit processing. 51704a5f133cSPaul E. McKenney 51714a5f133cSPaul E. McKenney rcutorture.read_exit_burst= [KNL] 51724a5f133cSPaul E. McKenney The number of times in a given read-then-exit 51734a5f133cSPaul E. McKenney episode that a set of read-then-exit kthreads 51744a5f133cSPaul E. McKenney is spawned. 51754a5f133cSPaul E. McKenney 51764a5f133cSPaul E. McKenney rcutorture.read_exit_delay= [KNL] 5177e52347bdSJani Nikula The delay, in seconds, between successive 5178e52347bdSJani Nikula read-then-exit testing episodes. 5179e52347bdSJani Nikula 5180e52347bdSJani Nikula rcutorture.shuffle_interval= [KNL] 5181e52347bdSJani Nikula Set task-shuffle interval (s). Shuffling tasks 5182e52347bdSJani Nikula allows some CPUs to go into dyntick-idle mode 5183e52347bdSJani Nikula during the rcutorture test. 5184e52347bdSJani Nikula 5185e52347bdSJani Nikula rcutorture.shutdown_secs= [KNL] 5186e52347bdSJani Nikula Set time (s) after boot system shutdown. This 5187e52347bdSJani Nikula is useful for hands-off automated testing. 5188e52347bdSJani Nikula 5189e52347bdSJani Nikula rcutorture.stall_cpu= [KNL] 519019a8ff95SPaul E. McKenney Duration of CPU stall (s) to test RCU CPU stall 519119a8ff95SPaul E. McKenney warnings, zero to disable. 51929e5d61c0SZqiang 51939e5d61c0SZqiang rcutorture.stall_cpu_block= [KNL] 51949e5d61c0SZqiang Sleep while stalling if set. This will result 51959e5d61c0SZqiang in warnings from preemptible RCU in addition to 51969e5d61c0SZqiang any other stall-related activity. Note that 51979e5d61c0SZqiang in kernels built with CONFIG_PREEMPTION=n and 51989e5d61c0SZqiang CONFIG_PREEMPT_COUNT=y, this parameter will 51999e5d61c0SZqiang cause the CPU to pass through a quiescent state. 52009e5d61c0SZqiang Given CONFIG_PREEMPTION=n, this will suppress 52019e5d61c0SZqiang RCU CPU stall warnings, but will instead result 52029e5d61c0SZqiang in scheduling-while-atomic splats. 520319a8ff95SPaul E. McKenney 5204e52347bdSJani Nikula Use of this module parameter results in splats. 5205e52347bdSJani Nikula 5206e52347bdSJani Nikula 52072b1516e5SPaul E. McKenney rcutorture.stall_cpu_holdoff= [KNL] 52082b1516e5SPaul E. McKenney Time to wait (s) after boot before inducing stall. 52092b1516e5SPaul E. McKenney 521055b2dcf5SPaul E. McKenney rcutorture.stall_cpu_irqsoff= [KNL] 521155b2dcf5SPaul E. McKenney Disable interrupts while stalling if set. 521255b2dcf5SPaul E. McKenney 521355b2dcf5SPaul E. McKenney rcutorture.stall_gp_kthread= [KNL] 521455b2dcf5SPaul E. McKenney Duration (s) of forced sleep within RCU 521555b2dcf5SPaul E. McKenney grace-period kthread to test RCU CPU stall 521655b2dcf5SPaul E. McKenney warnings, zero to disable. If both stall_cpu 5217e52347bdSJani Nikula and stall_gp_kthread are specified, the 5218e52347bdSJani Nikula kthread is starved first, then the CPU. 5219e52347bdSJani Nikula 5220e52347bdSJani Nikula rcutorture.stat_interval= [KNL] 5221e52347bdSJani Nikula Time (s) between statistics printk()s. 5222e52347bdSJani Nikula 5223e52347bdSJani Nikula rcutorture.stutter= [KNL] 5224e52347bdSJani Nikula Time (s) to stutter testing, for example, specifying 5225e52347bdSJani Nikula five seconds causes the test to run for five seconds, 5226e52347bdSJani Nikula wait for five seconds, and so on. This tests RCU's 5227e52347bdSJani Nikula ability to transition abruptly to and from idle. 5228e52347bdSJani Nikula 5229e52347bdSJani Nikula rcutorture.test_boost= [KNL] 5230e52347bdSJani Nikula Test RCU priority boosting? 0=no, 1=maybe, 2=yes. 5231e52347bdSJani Nikula "Maybe" means test if the RCU implementation 5232e52347bdSJani Nikula under test support RCU priority boosting. 5233e52347bdSJani Nikula 5234e52347bdSJani Nikula rcutorture.test_boost_duration= [KNL] 5235e52347bdSJani Nikula Duration (s) of each individual boost test. 5236e52347bdSJani Nikula 5237e52347bdSJani Nikula rcutorture.test_boost_interval= [KNL] 5238e52347bdSJani Nikula Interval (s) between each boost test. 5239e52347bdSJani Nikula 5240e52347bdSJani Nikula rcutorture.test_no_idle_hz= [KNL] 5241e52347bdSJani Nikula Test RCU's dyntick-idle handling. See also the 5242e52347bdSJani Nikula rcutorture.shuffle_interval parameter. 5243e52347bdSJani Nikula 5244e52347bdSJani Nikula rcutorture.torture_type= [KNL] 5245e52347bdSJani Nikula Specify the RCU implementation to test. 5246e52347bdSJani Nikula 5247cdc694b2SPaul E. McKenney rcutorture.verbose= [KNL] 5248cdc694b2SPaul E. McKenney Enable additional printk() statements. 5249cdc694b2SPaul E. McKenney 5250cdc694b2SPaul E. McKenney rcupdate.rcu_cpu_stall_ftrace_dump= [KNL] 5251e52347bdSJani Nikula Dump ftrace buffer after reporting RCU CPU 5252e52347bdSJani Nikula stall warning. 5253e52347bdSJani Nikula 525458c53360SPaul E. McKenney rcupdate.rcu_cpu_stall_suppress= [KNL] 525558c53360SPaul E. McKenney Suppress RCU CPU stall warning messages. 525658c53360SPaul E. McKenney 525758c53360SPaul E. McKenney rcupdate.rcu_cpu_stall_suppress_at_boot= [KNL] 525858c53360SPaul E. McKenney Suppress RCU CPU stall warning messages and 525958c53360SPaul E. McKenney rcutorture writer stall warnings that occur 5260e52347bdSJani Nikula during early boot, that is, during the time 5261e52347bdSJani Nikula before the init task is spawned. 526228b3ae42SUladzislau Rezki 526328b3ae42SUladzislau Rezki rcupdate.rcu_cpu_stall_timeout= [KNL] 526428b3ae42SUladzislau Rezki Set timeout for RCU CPU stall warning messages. 526528b3ae42SUladzislau Rezki The value is in seconds and the maximum allowed 526628b3ae42SUladzislau Rezki value is 300 seconds. 526728b3ae42SUladzislau Rezki 526828b3ae42SUladzislau Rezki rcupdate.rcu_exp_cpu_stall_timeout= [KNL] 526928b3ae42SUladzislau Rezki Set timeout for expedited RCU CPU stall warning 527028b3ae42SUladzislau Rezki messages. The value is in milliseconds 527128b3ae42SUladzislau Rezki and the maximum allowed value is 21000 527228b3ae42SUladzislau Rezki milliseconds. Please note that this value is 527328b3ae42SUladzislau Rezki adjusted to an arch timer tick resolution. 5274e52347bdSJani Nikula Setting this to zero causes the value from 5275be42f00bSZhen Lei rcupdate.rcu_cpu_stall_timeout to be used (after 5276be42f00bSZhen Lei conversion from seconds to milliseconds). 5277be42f00bSZhen Lei 5278be42f00bSZhen Lei rcupdate.rcu_cpu_stall_cputime= [KNL] 5279be42f00bSZhen Lei Provide statistics on the cputime and count of 5280be42f00bSZhen Lei interrupts and tasks during the sampling period. For 528192987fe8SPaul E. McKenney multiple continuous RCU stalls, all sampling periods 528292987fe8SPaul E. McKenney begin at half of the first RCU stall timeout. 528392987fe8SPaul E. McKenney 528492987fe8SPaul E. McKenney rcupdate.rcu_exp_stall_task_details= [KNL] 528592987fe8SPaul E. McKenney Print stack dumps of any tasks blocking the 5286e52347bdSJani Nikula current expedited RCU grace period during an 5287e52347bdSJani Nikula expedited RCU CPU stall warning. 5288e52347bdSJani Nikula 5289e52347bdSJani Nikula rcupdate.rcu_expedited= [KNL] 5290e52347bdSJani Nikula Use expedited grace-period primitives, for 5291e52347bdSJani Nikula example, synchronize_rcu_expedited() instead 5292e52347bdSJani Nikula of synchronize_rcu(). This reduces latency, 5293e52347bdSJani Nikula but can increase CPU utilization, degrade 5294e52347bdSJani Nikula real-time latency, and degrade energy efficiency. 5295e52347bdSJani Nikula No effect on CONFIG_TINY_RCU kernels. 5296e52347bdSJani Nikula 5297e52347bdSJani Nikula rcupdate.rcu_normal= [KNL] 5298e52347bdSJani Nikula Use only normal grace-period primitives, 5299e52347bdSJani Nikula for example, synchronize_rcu() instead of 5300e52347bdSJani Nikula synchronize_rcu_expedited(). This improves 5301e52347bdSJani Nikula real-time latency, CPU utilization, and 5302e52347bdSJani Nikula energy efficiency, but can expose users to 5303e52347bdSJani Nikula increased grace-period latency. This parameter 5304e52347bdSJani Nikula overrides rcupdate.rcu_expedited. No effect on 5305e52347bdSJani Nikula CONFIG_TINY_RCU kernels. 5306e52347bdSJani Nikula 5307e52347bdSJani Nikula rcupdate.rcu_normal_after_boot= [KNL] 5308e52347bdSJani Nikula Once boot has completed (that is, after 5309e52347bdSJani Nikula rcu_end_inkernel_boot() has been invoked), use 531036221e10SJulia Cartwright only normal grace-period primitives. No effect 531136221e10SJulia Cartwright on CONFIG_TINY_RCU kernels. 531236221e10SJulia Cartwright 531336221e10SJulia Cartwright But note that CONFIG_PREEMPT_RT=y kernels enables 531436221e10SJulia Cartwright this kernel boot parameter, forcibly setting 531536221e10SJulia Cartwright it to the value one, that is, converting any 531636221e10SJulia Cartwright post-boot attempt at an expedited RCU grace 5317fd796e41SPaul E. McKenney period to instead use normal non-expedited 5318fd796e41SPaul E. McKenney grace-period processing. 5319fd796e41SPaul E. McKenney 5320fd796e41SPaul E. McKenney rcupdate.rcu_task_collapse_lim= [KNL] 5321fd796e41SPaul E. McKenney Set the maximum number of callbacks present 5322fd796e41SPaul E. McKenney at the beginning of a grace period that allows 5323fd796e41SPaul E. McKenney the RCU Tasks flavors to collapse back to using 5324fd796e41SPaul E. McKenney a single callback queue. This switching only 5325ab97152fSPaul E. McKenney occurs when rcupdate.rcu_task_enqueue_lim is 5326ab97152fSPaul E. McKenney set to the default value of -1. 5327ab97152fSPaul E. McKenney 5328ab97152fSPaul E. McKenney rcupdate.rcu_task_contend_lim= [KNL] 5329ab97152fSPaul E. McKenney Set the minimum number of callback-queuing-time 5330ab97152fSPaul E. McKenney lock-contention events per jiffy required to 5331ab97152fSPaul E. McKenney cause the RCU Tasks flavors to switch to per-CPU 5332ab97152fSPaul E. McKenney callback queuing. This switching only occurs 53338610b656SPaul E. McKenney when rcupdate.rcu_task_enqueue_lim is set to 53348610b656SPaul E. McKenney the default value of -1. 53358610b656SPaul E. McKenney 53368610b656SPaul E. McKenney rcupdate.rcu_task_enqueue_lim= [KNL] 53378610b656SPaul E. McKenney Set the number of callback queues to use for the 53388610b656SPaul E. McKenney RCU Tasks family of RCU flavors. The default 53398610b656SPaul E. McKenney of -1 allows this to be automatically (and 5340b0afa0f0SPaul E. McKenney dynamically) adjusted. This parameter is intended 5341b0afa0f0SPaul E. McKenney for use in testing. 5342b0afa0f0SPaul E. McKenney 5343b0afa0f0SPaul E. McKenney rcupdate.rcu_task_ipi_delay= [KNL] 5344b0afa0f0SPaul E. McKenney Set time in jiffies during which RCU tasks will 5345b0afa0f0SPaul E. McKenney avoid sending IPIs, starting with the beginning 5346b0afa0f0SPaul E. McKenney of a given grace period. Setting a large 5347db13710aSPaul E. McKenney number avoids disturbing real-time workloads, 5348db13710aSPaul E. McKenney but lengthens grace periods. 5349db13710aSPaul E. McKenney 5350db13710aSPaul E. McKenney rcupdate.rcu_task_lazy_lim= [KNL] 5351db13710aSPaul E. McKenney Number of callbacks on a given CPU that will 5352db13710aSPaul E. McKenney cancel laziness on that CPU. Use -1 to disable 5353db13710aSPaul E. McKenney cancellation of laziness, but be advised that 5354f2539003SPaul E. McKenney doing so increases the danger of OOM due to 5355f2539003SPaul E. McKenney callback flooding. 5356f2539003SPaul E. McKenney 5357f2539003SPaul E. McKenney rcupdate.rcu_task_stall_info= [KNL] 5358f2539003SPaul E. McKenney Set initial timeout in jiffies for RCU task stall 5359f2539003SPaul E. McKenney informational messages, which give some indication 5360f2539003SPaul E. McKenney of the problem for those not patient enough to 5361f2539003SPaul E. McKenney wait for ten minutes. Informational messages are 5362f2539003SPaul E. McKenney only printed prior to the stall-warning message 5363f2539003SPaul E. McKenney for a given grace period. Disable with a value 5364f2539003SPaul E. McKenney less than or equal to zero. Defaults to ten 5365f2539003SPaul E. McKenney seconds. A change in value does not take effect 5366f2539003SPaul E. McKenney until the beginning of the next grace period. 5367f2539003SPaul E. McKenney 5368f2539003SPaul E. McKenney rcupdate.rcu_task_stall_info_mult= [KNL] 5369f2539003SPaul E. McKenney Multiplier for time interval between successive 5370f2539003SPaul E. McKenney RCU task stall informational messages for a given 5371f2539003SPaul E. McKenney RCU tasks grace period. This value is clamped 5372f2539003SPaul E. McKenney to one through ten, inclusive. It defaults to 5373f2539003SPaul E. McKenney the value three, so that the first informational 5374f2539003SPaul E. McKenney message is printed 10 seconds into the grace 5375f2539003SPaul E. McKenney period, the second at 40 seconds, the third at 5376e52347bdSJani Nikula 160 seconds, and then the stall warning at 600 5377f2539003SPaul E. McKenney seconds would prevent a fourth at 640 seconds. 5378f2539003SPaul E. McKenney 5379f2539003SPaul E. McKenney rcupdate.rcu_task_stall_timeout= [KNL] 5380f2539003SPaul E. McKenney Set timeout in jiffies for RCU task stall 5381f2539003SPaul E. McKenney warning messages. Disable with a value less 5382e52347bdSJani Nikula than or equal to zero. Defaults to ten minutes. 5383450d461aSPaul E. McKenney A change in value does not take effect until 5384450d461aSPaul E. McKenney the beginning of the next grace period. 5385450d461aSPaul E. McKenney 5386450d461aSPaul E. McKenney rcupdate.rcu_tasks_lazy_ms= [KNL] 5387450d461aSPaul E. McKenney Set timeout in milliseconds RCU Tasks asynchronous 5388450d461aSPaul E. McKenney callback batching for call_rcu_tasks(). 5389450d461aSPaul E. McKenney A negative value will take the default. A value 5390450d461aSPaul E. McKenney of zero will disable batching. Batching is 5391450d461aSPaul E. McKenney always disabled for synchronize_rcu_tasks(). 5392450d461aSPaul E. McKenney 5393450d461aSPaul E. McKenney rcupdate.rcu_tasks_rude_lazy_ms= [KNL] 5394450d461aSPaul E. McKenney Set timeout in milliseconds RCU Tasks 5395450d461aSPaul E. McKenney Rude asynchronous callback batching for 5396450d461aSPaul E. McKenney call_rcu_tasks_rude(). A negative value 5397450d461aSPaul E. McKenney will take the default. A value of zero will 5398450d461aSPaul E. McKenney disable batching. Batching is always disabled 5399450d461aSPaul E. McKenney for synchronize_rcu_tasks_rude(). 5400450d461aSPaul E. McKenney 5401450d461aSPaul E. McKenney rcupdate.rcu_tasks_trace_lazy_ms= [KNL] 5402450d461aSPaul E. McKenney Set timeout in milliseconds RCU Tasks 5403450d461aSPaul E. McKenney Trace asynchronous callback batching for 5404450d461aSPaul E. McKenney call_rcu_tasks_trace(). A negative value 5405450d461aSPaul E. McKenney will take the default. A value of zero will 5406e52347bdSJani Nikula disable batching. Batching is always disabled 5407e52347bdSJani Nikula for synchronize_rcu_tasks_trace(). 5408e52347bdSJani Nikula 5409e52347bdSJani Nikula rcupdate.rcu_self_test= [KNL] 5410e52347bdSJani Nikula Run the RCU early boot self tests 5411e52347bdSJani Nikula 5412e52347bdSJani Nikula rdinit= [KNL] 5413e52347bdSJani Nikula Format: <full_path> 5414c49a0a80STom Lendacky Run specified binary instead of /init from the ramdisk, 5415c49a0a80STom Lendacky used for early userspace startup. See initrd. 5416c49a0a80STom Lendacky 5417c49a0a80STom Lendacky rdrand= [X86] 5418c49a0a80STom Lendacky force - Override the decision by the kernel to hide the 5419c49a0a80STom Lendacky advertisement of RDRAND support (this affects 5420c49a0a80STom Lendacky certain AMD processors because of buggy BIOS 54211d9807fcSTony Luck support, specifically around the suspend/resume 54221d9807fcSTony Luck path). 542331516de3SFenghua Yu 5424a76f65c8SBabu Moger rdt= [HW,X86,RDT] 54251d9807fcSTony Luck Turn on/off individual RDT features. List is: 54261d9807fcSTony Luck cmt, mbmtotal, mbmlocal, l3cat, l3cdp, l2cat, l2cdp, 54271d9807fcSTony Luck mba, smba, bmec. 5428e52347bdSJani Nikula E.g. to turn on cmt and turn off mba use: 5429e52347bdSJani Nikula rdt=cmt,!mba 543012febc18SPaul Gortmaker 5431e52347bdSJani Nikula reboot= [KNL] 5432e52347bdSJani Nikula Format (x86 or x86_64): 5433e52347bdSJani Nikula [w[arm] | c[old] | h[ard] | s[oft] | g[pio]] | d[efault] \ 5434b287a25aSAaro Koskinen [[,]s[mp]#### \ 5435b287a25aSAaro Koskinen [[,]b[ios] | a[cpi] | k[bd] | t[riple] | e[fi] | p[ci]] \ 5436b287a25aSAaro Koskinen [[,]f[orce] 5437e52347bdSJani Nikula Where reboot_mode is one of warm (soft) or cold (hard) or gpio 5438e52347bdSJani Nikula (prefix with 'panic_' to set mode for panic 5439e52347bdSJani Nikula reboot only), 5440e52347bdSJani Nikula reboot_type is one of bios, acpi, kbd, triple, efi, or pci, 5441e52347bdSJani Nikula reboot_force is either force or not specified, 54421fbeb3a8SPaul E. McKenney reboot_cpu is s[mp]#### with #### being the processor 5443847dd70aSPaul E. McKenney to be used for rebooting. 5444847dd70aSPaul E. McKenney 5445847dd70aSPaul E. McKenney refscale.holdoff= [KNL] 5446847dd70aSPaul E. McKenney Set test-start holdoff period. The purpose of 5447847dd70aSPaul E. McKenney this parameter is to delay the start of the 54481fbeb3a8SPaul E. McKenney test until boot completes in order to avoid 5449847dd70aSPaul E. McKenney interference. 5450847dd70aSPaul E. McKenney 5451847dd70aSPaul E. McKenney refscale.loops= [KNL] 5452847dd70aSPaul E. McKenney Set the number of loops over the synchronization 5453847dd70aSPaul E. McKenney primitive under test. Increasing this number 5454847dd70aSPaul E. McKenney reduces noise due to loop start/end overhead, 5455847dd70aSPaul E. McKenney but the default has already reduced the per-pass 54561fbeb3a8SPaul E. McKenney noise to a handful of picoseconds on ca. 2020 5457847dd70aSPaul E. McKenney x86 laptops. 5458847dd70aSPaul E. McKenney 5459847dd70aSPaul E. McKenney refscale.nreaders= [KNL] 5460847dd70aSPaul E. McKenney Set number of readers. The default value of -1 54611fbeb3a8SPaul E. McKenney selects N, where N is roughly 75% of the number 5462847dd70aSPaul E. McKenney of CPUs. A value of zero is an interesting choice. 5463847dd70aSPaul E. McKenney 5464847dd70aSPaul E. McKenney refscale.nruns= [KNL] 54651fbeb3a8SPaul E. McKenney Set number of runs, each of which is dumped onto 5466847dd70aSPaul E. McKenney the console log. 5467847dd70aSPaul E. McKenney 5468847dd70aSPaul E. McKenney refscale.readdelay= [KNL] 54691fbeb3a8SPaul E. McKenney Set the read-side critical-section duration, 54701fbeb3a8SPaul E. McKenney measured in microseconds. 54711fbeb3a8SPaul E. McKenney 54721fbeb3a8SPaul E. McKenney refscale.scale_type= [KNL] 5473847dd70aSPaul E. McKenney Specify the read-protection implementation to test. 5474847dd70aSPaul E. McKenney 54754e88ec4aSPaul E. McKenney refscale.shutdown= [KNL] 54764e88ec4aSPaul E. McKenney Shut down the system at the end of the performance 5477847dd70aSPaul E. McKenney test. This defaults to 1 (shut it down) when 54781fbeb3a8SPaul E. McKenney refscale is built into the kernel and to 0 (leave 5479847dd70aSPaul E. McKenney it running) when refscale is built as a module. 5480847dd70aSPaul E. McKenney 5481e76506f0SPaul E. McKenney refscale.verbose= [KNL] 5482e76506f0SPaul E. McKenney Enable additional printk() statements. 5483e76506f0SPaul E. McKenney 5484e76506f0SPaul E. McKenney refscale.verbose_batched= [KNL] 5485e76506f0SPaul E. McKenney Batch the additional printk() statements. If zero 5486e76506f0SPaul E. McKenney (the default) or negative, print everything. Otherwise, 5487e52347bdSJani Nikula print every Nth verbose statement, where N is the value 5488e52347bdSJani Nikula specified. 5489da82c92fSMauro Carvalho Chehab 5490e52347bdSJani Nikula relax_domain_level= 5491ffd2e8dfSBjorn Helgaas [KNL, SMP] Set scheduler's default relax_domain_level. 5492ffd2e8dfSBjorn Helgaas See Documentation/admin-guide/cgroup-v1/cpusets.rst. 5493ffd2e8dfSBjorn Helgaas 5494ffd2e8dfSBjorn Helgaas reserve= [KNL,BUGS] Force kernel to ignore I/O ports or memory 5495ffd2e8dfSBjorn Helgaas Format: <base1>,<size1>[,<base2>,<size2>,...] 5496e52347bdSJani Nikula Reserve I/O ports or memory so the kernel won't use 5497e52347bdSJani Nikula them. If <base> is less than 0x10000, the region 5498e52347bdSJani Nikula is assumed to be I/O ports; otherwise it is memory. 5499e52347bdSJani Nikula 5500e52347bdSJani Nikula reservetop= [X86-32] 5501e52347bdSJani Nikula Format: nn[KMG] 5502e52347bdSJani Nikula Reserves a hole at the top of the kernel virtual 5503e52347bdSJani Nikula address space. 5504e52347bdSJani Nikula 5505e52347bdSJani Nikula reset_devices [KNL] Force drivers to reset the underlying device 5506e52347bdSJani Nikula during initialization. 5507e52347bdSJani Nikula 5508e52347bdSJani Nikula resume= [SWSUSP] 5509e52347bdSJani Nikula Specify the partition device for software suspend 5510e52347bdSJani Nikula Format: 5511e52347bdSJani Nikula {/dev/<dev> | PARTUUID=<uuid> | <int>:<int> | <hex>} 5512e52347bdSJani Nikula 5513e52347bdSJani Nikula resume_offset= [SWSUSP] 5514151f4e2bSMauro Carvalho Chehab Specify the offset from the beginning of the partition 5515e52347bdSJani Nikula given by "resume=" at which the swap header is located, 5516e52347bdSJani Nikula in <PAGE_SIZE> units (needed only for swap files). 5517e52347bdSJani Nikula See Documentation/power/swsusp-and-swap-files.rst 5518e52347bdSJani Nikula 5519e52347bdSJani Nikula resumedelay= [HIBERNATION] Delay (in seconds) to pause before attempting to 5520e52347bdSJani Nikula read the resume files 5521e52347bdSJani Nikula 5522e52347bdSJani Nikula resumewait [HIBERNATION] Wait (indefinitely) for resume device to show up. 5523e52347bdSJani Nikula Useful for devices that are detected asynchronously 5524e52347bdSJani Nikula (e.g. USB and MMC devices). 55257fbf47c7SAlexandre Chartre 55267fbf47c7SAlexandre Chartre retain_initrd [RAM] Keep initrd memory after extraction 55277fbf47c7SAlexandre Chartre 55287fbf47c7SAlexandre Chartre retbleed= [X86] Control mitigation of RETBleed (Arbitrary 5529e6cfcddaSKim Phillips Speculative Code Execution with Return Instructions) 5530e6cfcddaSKim Phillips vulnerability. 5531e6cfcddaSKim Phillips 5532e6cfcddaSKim Phillips AMD-based UNRET and IBPB mitigations alone do not stop 5533e6cfcddaSKim Phillips sibling threads from influencing the predictions of other 5534e6cfcddaSKim Phillips sibling threads. For that reason, STIBP is used on pro- 5535e8ec1b6eSKim Phillips cessors that support it, and mitigate SMT on processors 55367fbf47c7SAlexandre Chartre that don't. 5537e8ec1b6eSKim Phillips 5538e8ec1b6eSKim Phillips off - no mitigation 5539e8ec1b6eSKim Phillips auto - automatically select a migitation 5540e8ec1b6eSKim Phillips auto,nosmt - automatically select a mitigation, 5541e6cfcddaSKim Phillips disabling SMT if necessary for 5542e6cfcddaSKim Phillips the full mitigation (only on Zen1 5543e6cfcddaSKim Phillips and older without STIBP). 5544e6cfcddaSKim Phillips ibpb - On AMD, mitigate short speculation 5545e6cfcddaSKim Phillips windows on basic block boundaries too. 5546e6cfcddaSKim Phillips Safe, highest perf impact. It also 5547e6cfcddaSKim Phillips enables STIBP if present. Not suitable 5548e6cfcddaSKim Phillips on Intel. 5549e6cfcddaSKim Phillips ibpb,nosmt - Like "ibpb" above but will disable SMT 5550e6cfcddaSKim Phillips when STIBP is not available. This is 5551e6cfcddaSKim Phillips the alternative for systems which do not 5552e6cfcddaSKim Phillips have STIBP. 5553e6cfcddaSKim Phillips unret - Force enable untrained return thunks, 5554e6cfcddaSKim Phillips only effective on AMD f15h-f17h based 5555e6cfcddaSKim Phillips systems. 55567fbf47c7SAlexandre Chartre unret,nosmt - Like unret, but will disable SMT when STIBP 55577fbf47c7SAlexandre Chartre is not available. This is the alternative for 55587fbf47c7SAlexandre Chartre systems which do not have STIBP. 55597fbf47c7SAlexandre Chartre 55607fbf47c7SAlexandre Chartre Selecting 'auto' will choose a mitigation method at run 55617fbf47c7SAlexandre Chartre time according to the CPU. 5562e52347bdSJani Nikula 5563e52347bdSJani Nikula Not specifying this option is equivalent to retbleed=auto. 5564e52347bdSJani Nikula 5565e52347bdSJani Nikula rfkill.default_state= 5566e52347bdSJani Nikula 0 "airplane mode". All wifi, bluetooth, wimax, gps, fm, 5567e52347bdSJani Nikula etc. communication is blocked by default. 5568e52347bdSJani Nikula 1 Unblocked. 5569e52347bdSJani Nikula 5570e52347bdSJani Nikula rfkill.master_switch_mode= 5571e52347bdSJani Nikula 0 The "airplane mode" button does nothing. 5572e52347bdSJani Nikula 1 The "airplane mode" button toggles between everything 5573e52347bdSJani Nikula blocked and the previous configuration. 5574e52347bdSJani Nikula 2 The "airplane mode" button toggles between everything 5575e52347bdSJani Nikula blocked and everything unblocked. 5576e52347bdSJani Nikula 5577e16fd002SGrzegorz Andrejczuk rhash_entries= [KNL,NET] 5578e16fd002SGrzegorz Andrejczuk Set number of hash buckets for route cache 5579e16fd002SGrzegorz Andrejczuk 5580e16fd002SGrzegorz Andrejczuk ring3mwait=disable 5581496ea826SConor Dooley [KNL] Disable ring 3 MONITOR/MWAIT feature on supported 5582496ea826SConor Dooley CPUs. 5583496ea826SConor Dooley 5584496ea826SConor Dooley riscv_isa_fallback [RISCV] 5585496ea826SConor Dooley When CONFIG_RISCV_ISA_FALLBACK is not enabled, permit 5586496ea826SConor Dooley falling back to detecting extension support by parsing 5587496ea826SConor Dooley "riscv,isa" property on devicetree systems when the 5588e52347bdSJani Nikula replacement properties are not found. See the Kconfig 5589e52347bdSJani Nikula entry for RISCV_ISA_FALLBACK. 5590e52347bdSJani Nikula 5591e52347bdSJani Nikula ro [KNL] Mount root device read-only on boot 5592e52347bdSJani Nikula 55932e8cff0aSMark Rutland rodata= [KNL] 55942e8cff0aSMark Rutland on Mark read-only kernel memory as read-only (default). 5595e52347bdSJani Nikula off Leave read-only kernel memory writable for debugging. 5596e52347bdSJani Nikula full Mark read-only kernel memory and aliases as read-only 5597e52347bdSJani Nikula [arm64] 5598e52347bdSJani Nikula 5599e52347bdSJani Nikula rockchip.usb_uart 5600e52347bdSJani Nikula Enable the uart passthrough on the designated usb port 5601e52347bdSJani Nikula on Rockchip SoCs. When active, the signals of the 5602e52347bdSJani Nikula debug-uart get routed to the D+ and D- pins of the usb 5603c0c1a7dcSChristoph Hellwig port and the regular usb controller gets disabled. 5604702f3189SChristoph Hellwig 5605702f3189SChristoph Hellwig root= [KNL] Root filesystem 5606c0c1a7dcSChristoph Hellwig Usually this a a block device specifier of some kind, 5607c0c1a7dcSChristoph Hellwig see the early_lookup_bdev comment in 5608c0c1a7dcSChristoph Hellwig block/early-lookup.c for details. 5609e52347bdSJani Nikula Alternatively this can be "ram" for the legacy initial 5610e52347bdSJani Nikula ramdisk, "nfs" and "cifs" for root on a network file 5611e52347bdSJani Nikula system, or "mtd" and "ubi" for mounting from raw flash. 5612e52347bdSJani Nikula 5613e52347bdSJani Nikula rootdelay= [KNL] Delay (in seconds) to pause before attempting to 5614e52347bdSJani Nikula mount the root filesystem 5615e52347bdSJani Nikula 5616e52347bdSJani Nikula rootflags= [KNL] Set root filesystem mount option string 5617e52347bdSJani Nikula 5618e52347bdSJani Nikula rootfstype= [KNL] Set root filesystem type 5619e52347bdSJani Nikula 5620e52347bdSJani Nikula rootwait [KNL] Wait (indefinitely) for root device to show up. 562145071e1cSLoic Poulain Useful for devices that are detected asynchronously 562245071e1cSLoic Poulain (e.g. USB and MMC devices). 562345071e1cSLoic Poulain 562445071e1cSLoic Poulain rootwait= [KNL] Maximum time (in seconds) to wait for root device 5625e52347bdSJani Nikula to show up before attempting to mount the root 5626e52347bdSJani Nikula filesystem. 5627e52347bdSJani Nikula 5628e52347bdSJani Nikula rproc_mem=nn[KMG][@address] 5629e52347bdSJani Nikula [KNL,ARM,CMA] Remoteproc physical memory block. 5630e52347bdSJani Nikula Memory area to be used by remote processor image, 5631e52347bdSJani Nikula managed by CMA. 5632e52347bdSJani Nikula 5633e52347bdSJani Nikula rw [KNL] Mount root device read-write on boot 5634e52347bdSJani Nikula 5635e52347bdSJani Nikula S [KNL] Run init in single mode 5636e52347bdSJani Nikula 5637e52347bdSJani Nikula s390_iommu= [HW,S390] 5638e52347bdSJani Nikula Set s390 IOTLB flushing mode 5639e52347bdSJani Nikula strict 5640e52347bdSJani Nikula With strict flushing every unmap operation will result in 56416aefbf1cSNiklas Schnelle an IOTLB flush. Default is lazy flushing before reuse, 56426aefbf1cSNiklas Schnelle which is faster. 56436aefbf1cSNiklas Schnelle 56446aefbf1cSNiklas Schnelle s390_iommu_aperture= [KNL,S390] 56456aefbf1cSNiklas Schnelle Specifies the size of the per device DMA address space 56466aefbf1cSNiklas Schnelle accessible through the DMA and IOMMU APIs as a decimal 56476aefbf1cSNiklas Schnelle factor of the size of main memory. 56486aefbf1cSNiklas Schnelle The default is 1 meaning that one can concurrently use 56496aefbf1cSNiklas Schnelle as many DMA addresses as physical memory is installed, 56506aefbf1cSNiklas Schnelle if supported by hardware, and thus map all of memory 56516aefbf1cSNiklas Schnelle once. With a value of 2 one can map all of memory twice 56526aefbf1cSNiklas Schnelle and so on. As a special case a factor of 0 imposes no 5653e52347bdSJani Nikula restrictions other than those given by hardware at the 5654e52347bdSJani Nikula cost of significant additional memory use for tables. 5655e52347bdSJani Nikula 56569406415fSPeter Zijlstra sa1100ir [NET] 5657e52347bdSJani Nikula See drivers/net/irda/sa1100_ir.c. 5658e52347bdSJani Nikula 5659e52347bdSJani Nikula sched_verbose [KNL] Enables verbose scheduler debug messages. 5660e52347bdSJani Nikula 5661e52347bdSJani Nikula schedstats= [KNL,X86] Enable or disable scheduled statistics. 5662e52347bdSJani Nikula Allowed values are enable and disable. This feature 566305289b90SThara Gopinath incurs a small amount of overhead in the scheduler 566405289b90SThara Gopinath but is useful for debugging and performance tuning. 566505289b90SThara Gopinath 566605289b90SThara Gopinath sched_thermal_decay_shift= 566705289b90SThara Gopinath [KNL, SMP] Set a decay shift for scheduler thermal 566805289b90SThara Gopinath pressure signal. Thermal pressure signal follows the 566905289b90SThara Gopinath default decay period of other scheduler pelt 567005289b90SThara Gopinath signals(usually 32 ms but configurable). Setting 567105289b90SThara Gopinath sched_thermal_decay_shift will left shift the decay 567205289b90SThara Gopinath period for the thermal pressure signal by the shift 567305289b90SThara Gopinath value. 567405289b90SThara Gopinath i.e. with the default pelt decay period of 32 ms 567505289b90SThara Gopinath sched_thermal_decay_shift thermal pressure decay pr 567605289b90SThara Gopinath 1 64 ms 567705289b90SThara Gopinath 2 128 ms 567805289b90SThara Gopinath and so on. 5679e9d338a0SPaul E. McKenney Format: integer between 0 and 10 5680e9d338a0SPaul E. McKenney Default is 0. 5681e9d338a0SPaul E. McKenney 5682e9d338a0SPaul E. McKenney scftorture.holdoff= [KNL] 5683e9d338a0SPaul E. McKenney Number of seconds to hold off before starting 5684e9d338a0SPaul E. McKenney test. Defaults to zero for module insertion and 5685e9d338a0SPaul E. McKenney to 10 seconds for built-in smp_call_function() 5686e9d338a0SPaul E. McKenney tests. 5687e9d338a0SPaul E. McKenney 5688e9d338a0SPaul E. McKenney scftorture.longwait= [KNL] 5689e9d338a0SPaul E. McKenney Request ridiculously long waits randomly selected 5690e9d338a0SPaul E. McKenney up to the chosen limit in seconds. Zero (the 5691e9d338a0SPaul E. McKenney default) disables this feature. Please note 5692e9d338a0SPaul E. McKenney that requesting even small non-zero numbers of 5693e9d338a0SPaul E. McKenney seconds can result in RCU CPU stall warnings, 5694e9d338a0SPaul E. McKenney softlockup complaints, and so on. 5695e9d338a0SPaul E. McKenney 5696e9d338a0SPaul E. McKenney scftorture.nthreads= [KNL] 5697e9d338a0SPaul E. McKenney Number of kthreads to spawn to invoke the 5698e9d338a0SPaul E. McKenney smp_call_function() family of functions. 5699e9d338a0SPaul E. McKenney The default of -1 specifies a number of kthreads 5700e9d338a0SPaul E. McKenney equal to the number of CPUs. 5701e9d338a0SPaul E. McKenney 5702e9d338a0SPaul E. McKenney scftorture.onoff_holdoff= [KNL] 5703e9d338a0SPaul E. McKenney Number seconds to wait after the start of the 5704e9d338a0SPaul E. McKenney test before initiating CPU-hotplug operations. 5705e9d338a0SPaul E. McKenney 5706e9d338a0SPaul E. McKenney scftorture.onoff_interval= [KNL] 5707e9d338a0SPaul E. McKenney Number seconds to wait between successive 5708e9d338a0SPaul E. McKenney CPU-hotplug operations. Specifying zero (which 5709e9d338a0SPaul E. McKenney is the default) disables CPU-hotplug operations. 5710e9d338a0SPaul E. McKenney 5711e9d338a0SPaul E. McKenney scftorture.shutdown_secs= [KNL] 5712e9d338a0SPaul E. McKenney The number of seconds following the start of the 5713e9d338a0SPaul E. McKenney test after which to shut down the system. The 5714e9d338a0SPaul E. McKenney default of zero avoids shutting down the system. 5715e9d338a0SPaul E. McKenney Non-zero values are useful for automated tests. 5716e9d338a0SPaul E. McKenney 5717e9d338a0SPaul E. McKenney scftorture.stat_interval= [KNL] 5718e9d338a0SPaul E. McKenney The number of seconds between outputting the 5719e9d338a0SPaul E. McKenney current test statistics to the console. A value 5720e9d338a0SPaul E. McKenney of zero disables statistics output. 5721e9d338a0SPaul E. McKenney 5722e9d338a0SPaul E. McKenney scftorture.stutter_cpus= [KNL] 5723e9d338a0SPaul E. McKenney The number of jiffies to wait between each change 5724e9d338a0SPaul E. McKenney to the set of CPUs under test. 5725e9d338a0SPaul E. McKenney 5726e9d338a0SPaul E. McKenney scftorture.use_cpus_read_lock= [KNL] 5727e9d338a0SPaul E. McKenney Use use_cpus_read_lock() instead of the default 5728e9d338a0SPaul E. McKenney preempt_disable() to disable CPU hotplug 5729e9d338a0SPaul E. McKenney while invoking one of the smp_call_function*() 5730e9d338a0SPaul E. McKenney functions. 5731e9d338a0SPaul E. McKenney 5732e9d338a0SPaul E. McKenney scftorture.verbose= [KNL] 5733e9d338a0SPaul E. McKenney Enable additional printk() statements. 5734e9d338a0SPaul E. McKenney 5735e9d338a0SPaul E. McKenney scftorture.weight_single= [KNL] 5736e9d338a0SPaul E. McKenney The probability weighting to use for the 5737e9d338a0SPaul E. McKenney smp_call_function_single() function with a zero 5738e9d338a0SPaul E. McKenney "wait" parameter. A value of -1 selects the 5739e9d338a0SPaul E. McKenney default if all other weights are -1. However, 5740e9d338a0SPaul E. McKenney if at least one weight has some other value, a 5741e9d338a0SPaul E. McKenney value of -1 will instead select a weight of zero. 5742e9d338a0SPaul E. McKenney 5743e9d338a0SPaul E. McKenney scftorture.weight_single_wait= [KNL] 5744e9d338a0SPaul E. McKenney The probability weighting to use for the 5745e9d338a0SPaul E. McKenney smp_call_function_single() function with a 5746e9d338a0SPaul E. McKenney non-zero "wait" parameter. See weight_single. 5747e9d338a0SPaul E. McKenney 5748e9d338a0SPaul E. McKenney scftorture.weight_many= [KNL] 5749e9d338a0SPaul E. McKenney The probability weighting to use for the 5750e9d338a0SPaul E. McKenney smp_call_function_many() function with a zero 5751e9d338a0SPaul E. McKenney "wait" parameter. See weight_single. 5752e9d338a0SPaul E. McKenney Note well that setting a high probability for 5753e9d338a0SPaul E. McKenney this weighting can place serious IPI load 5754e9d338a0SPaul E. McKenney on the system. 5755e9d338a0SPaul E. McKenney 5756e9d338a0SPaul E. McKenney scftorture.weight_many_wait= [KNL] 5757e9d338a0SPaul E. McKenney The probability weighting to use for the 5758e9d338a0SPaul E. McKenney smp_call_function_many() function with a 5759e9d338a0SPaul E. McKenney non-zero "wait" parameter. See weight_single 5760e9d338a0SPaul E. McKenney and weight_many. 5761e9d338a0SPaul E. McKenney 5762e9d338a0SPaul E. McKenney scftorture.weight_all= [KNL] 5763e9d338a0SPaul E. McKenney The probability weighting to use for the 5764e9d338a0SPaul E. McKenney smp_call_function_all() function with a zero 5765e9d338a0SPaul E. McKenney "wait" parameter. See weight_single and 5766e9d338a0SPaul E. McKenney weight_many. 5767e9d338a0SPaul E. McKenney 5768e9d338a0SPaul E. McKenney scftorture.weight_all_wait= [KNL] 5769e9d338a0SPaul E. McKenney The probability weighting to use for the 5770e9d338a0SPaul E. McKenney smp_call_function_all() function with a 5771e52347bdSJani Nikula non-zero "wait" parameter. See weight_single 5772e52347bdSJani Nikula and weight_many. 5773e52347bdSJani Nikula 5774e52347bdSJani Nikula skew_tick= [KNL] Offset the periodic timer tick per cpu to mitigate 5775e52347bdSJani Nikula xtime_lock contention on larger systems, and/or RCU lock 5776e52347bdSJani Nikula contention on all systems with CONFIG_MAXSMP set. 5777e52347bdSJani Nikula Format: { "0" | "1" } 5778e52347bdSJani Nikula 0 -- disable. (may be 1 via CONFIG_CMDLINE="skew_tick=1" 5779e52347bdSJani Nikula 1 -- enable. 578089a9684eSKees Cook Note: increases power consumption, thus should only be 578189a9684eSKees Cook enabled if running jitter sensitive (HPC/RT) workloads. 578289a9684eSKees Cook 5783e52347bdSJani Nikula security= [SECURITY] Choose a legacy "major" security module to 5784e52347bdSJani Nikula enable at boot. This has been deprecated by the 5785e52347bdSJani Nikula "lsm=" parameter. 5786e52347bdSJani Nikula 5787e52347bdSJani Nikula selinux= [SELINUX] Disable or enable SELinux at boot time. 5788e52347bdSJani Nikula Format: { "0" | "1" } 5789d41415ebSStephen Smalley See security/selinux/Kconfig help text. 5790e52347bdSJani Nikula 0 -- disable. 5791e52347bdSJani Nikula 1 -- enable. 5792e52347bdSJani Nikula Default value is 1. 5793ff61f079SJonathan Corbet 5794ba37a143SMichael Roth serialnumber [BUGS=X86-32] 5795e52347bdSJani Nikula 5796e52347bdSJani Nikula sev=option[,option...] [X86-64] See Documentation/arch/x86/x86_64/boot-options.rst 5797e52347bdSJani Nikula 579842551b8dSRandy Dunlap shapers= [NET] 579942551b8dSRandy Dunlap Maximal number of shapers. 580042551b8dSRandy Dunlap 580142551b8dSRandy Dunlap show_lapic= [APIC,X86] Advanced Programmable Interrupt Controller 580242551b8dSRandy Dunlap Limit apic dumping. The parameter defines the maximal 580342551b8dSRandy Dunlap number of local apics being dumped. Also it is possible 580442551b8dSRandy Dunlap to set it to "all" by meaning -- no limit here. 580542551b8dSRandy Dunlap Format: { 1 (default) | 2 | ... | all }. 580642551b8dSRandy Dunlap The parameter valid if only apic=debug or 5807e52347bdSJani Nikula apic=verbose is specified. 5808e52347bdSJani Nikula Example: apic=debug show_lapic=all 5809e52347bdSJani Nikula 5810e52347bdSJani Nikula simeth= [IA-64] 5811e52347bdSJani Nikula simscsi= 581282edd9d5SRafael Aquini 581382edd9d5SRafael Aquini slram= [HW,MTD] 581482edd9d5SRafael Aquini 581582edd9d5SRafael Aquini slab_merge [MM] 5816e52347bdSJani Nikula Enable merging of slabs with similar size when the 5817e52347bdSJani Nikula kernel is built without CONFIG_SLAB_MERGE_DEFAULT. 5818e52347bdSJani Nikula 58197660a6fdSKees Cook slab_nomerge [MM] 58207660a6fdSKees Cook Disable merging of slabs with similar size. May be 58217660a6fdSKees Cook necessary if there is some reason to distinguish 58227660a6fdSKees Cook allocs to different slabs, especially in hardened 58237660a6fdSKees Cook environments where the risk of heap overflows and 58247660a6fdSKees Cook layout control by attackers can usually be 58257660a6fdSKees Cook frustrated by disabling merging. This will reduce 58267660a6fdSKees Cook most of the exposure of a heap attack to a single 5827ee65728eSMike Rapoport cache (risks via metadata attacks are mostly 5828e52347bdSJani Nikula unchanged). Debug options disable merging on their 5829e52347bdSJani Nikula own. 5830e52347bdSJani Nikula For more information see Documentation/mm/slub.rst. 5831e52347bdSJani Nikula 5832e52347bdSJani Nikula slab_max_order= [MM, SLAB] 5833e52347bdSJani Nikula Determines the maximum allowed order for slabs. 5834e52347bdSJani Nikula A high setting may cause OOMs due to memory 5835e17f1dfbSVlastimil Babka fragmentation. Defaults to 1 for systems with 5836e52347bdSJani Nikula more than 32MB of RAM, 0 otherwise. 5837e52347bdSJani Nikula 5838e52347bdSJani Nikula slub_debug[=options[,slabs][;[options[,slabs]]...] [MM, SLUB] 5839e52347bdSJani Nikula Enabling slub_debug allows one to determine the 5840e52347bdSJani Nikula culprit if slab objects become corrupted. Enabling 5841ee65728eSMike Rapoport slub_debug can create guard zones around objects and 5842e52347bdSJani Nikula may poison objects when not in use. Also tracks the 5843e52347bdSJani Nikula last alloc / free. For more information see 5844e52347bdSJani Nikula Documentation/mm/slub.rst. 5845e52347bdSJani Nikula 5846e52347bdSJani Nikula slub_max_order= [MM, SLUB] 5847ee65728eSMike Rapoport Determines the maximum allowed order for slabs. 5848e52347bdSJani Nikula A high setting may cause OOMs due to memory 5849e52347bdSJani Nikula fragmentation. For more information see 5850e52347bdSJani Nikula Documentation/mm/slub.rst. 5851e52347bdSJani Nikula 5852e52347bdSJani Nikula slub_min_objects= [MM, SLUB] 5853e52347bdSJani Nikula The minimum number of objects per slab. SLUB will 5854e52347bdSJani Nikula increase the slab order up to slub_max_order to 5855e52347bdSJani Nikula generate a sufficiently large slab able to contain 5856ee65728eSMike Rapoport the number of objects indicated. The higher the number 5857e52347bdSJani Nikula of objects the smaller the overhead of tracking slabs 5858e52347bdSJani Nikula and the less frequently locks need to be acquired. 5859e52347bdSJani Nikula For more information see Documentation/mm/slub.rst. 5860e52347bdSJani Nikula 5861ee65728eSMike Rapoport slub_min_order= [MM, SLUB] 5862e52347bdSJani Nikula Determines the minimum page order for slabs. Must be 586382edd9d5SRafael Aquini lower than slub_max_order. 586482edd9d5SRafael Aquini For more information see Documentation/mm/slub.rst. 586582edd9d5SRafael Aquini 5866e52347bdSJani Nikula slub_merge [MM, SLUB] 5867e52347bdSJani Nikula Same with slab_merge. 5868e52347bdSJani Nikula 5869e52347bdSJani Nikula slub_nomerge [MM, SLUB] 5870e52347bdSJani Nikula Same with slab_nomerge. This is supported for legacy. 5871e52347bdSJani Nikula See slab_nomerge for more information. 5872e52347bdSJani Nikula 58733791a223SPaul E. McKenney smart2= [HW] 58743791a223SPaul E. McKenney Format: <io1>[,<io2>[,...,<io8>]] 58753791a223SPaul E. McKenney 58763791a223SPaul E. McKenney smp.csd_lock_timeout= [KNL] 58773791a223SPaul E. McKenney Specify the period of time in milliseconds 58783791a223SPaul E. McKenney that smp_call_function() and friends will wait 58793791a223SPaul E. McKenney for a CPU to release the CSD lock. This is 58803791a223SPaul E. McKenney useful when diagnosing bugs involving CPUs 58813791a223SPaul E. McKenney disabling interrupts for extended periods 58823791a223SPaul E. McKenney of time. Defaults to 5,000 milliseconds, and 58833791a223SPaul E. McKenney setting a value of zero disables this feature. 5884f6cc3d85SRik van Riel This feature may be more efficiently disabled 5885f6cc3d85SRik van Riel using the csdlock_debug- kernel parameter. 5886f6cc3d85SRik van Riel 5887f6cc3d85SRik van Riel smp.panic_on_ipistall= [KNL] 5888f6cc3d85SRik van Riel If a csd_lock_timeout extends for more than 5889f6cc3d85SRik van Riel the specified number of milliseconds, panic the 5890f6cc3d85SRik van Riel system. By default, let CSD-lock acquisition 5891e52347bdSJani Nikula take as long as they take. Specifying 300,000 5892e52347bdSJani Nikula for this value provides a 5-minute timeout. 5893e52347bdSJani Nikula 5894e52347bdSJani Nikula smsc-ircc2.nopnp [HW] Don't use PNP to discover SMC devices 5895e52347bdSJani Nikula smsc-ircc2.ircc_cfg= [HW] Device configuration I/O port 5896e52347bdSJani Nikula smsc-ircc2.ircc_sir= [HW] SIR base I/O port 5897e52347bdSJani Nikula smsc-ircc2.ircc_fir= [HW] FIR base I/O port 5898e52347bdSJani Nikula smsc-ircc2.ircc_irq= [HW] IRQ line 5899e52347bdSJani Nikula smsc-ircc2.ircc_dma= [HW] DMA channel 5900e52347bdSJani Nikula smsc-ircc2.ircc_transceiver= [HW] Transceiver type: 5901e52347bdSJani Nikula 0: Toshiba Satellite 1800 (GP data pin select) 590296cb8ae2SJiaxun Yang 1: Fast pin select (default) 5903e52347bdSJani Nikula 2: ATC IRMode 5904e52347bdSJani Nikula 5905e52347bdSJani Nikula smt= [KNL,MIPS,S390] Set the maximum number of threads (logical 5906e52347bdSJani Nikula CPUs) to use per physical CPU on systems capable of 5907e52347bdSJani Nikula symmetric multithreading (SMT). Will be capped to the 5908e52347bdSJani Nikula actual hardware limit. 5909e52347bdSJani Nikula Format: <integer> 5910e52347bdSJani Nikula Default: -1 (no limit) 5911f117955aSGuilherme G. Piccoli 5912e52347bdSJani Nikula softlockup_panic= 5913f117955aSGuilherme G. Piccoli [KNL] Should the soft-lockup detector generate panics. 59140a07bef6SGuilherme G. Piccoli Format: 0 | 1 59150a07bef6SGuilherme G. Piccoli 59160a07bef6SGuilherme G. Piccoli A value of 1 instructs the soft-lockup detector 59170a07bef6SGuilherme G. Piccoli to panic the machine when a soft-lockup occurs. It is 59183ce62385SBorislav Petkov also controlled by the kernel.softlockup_panic sysctl 5919e52347bdSJani Nikula and CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC, which is the 5920e52347bdSJani Nikula respective build-time switch to that functionality. 5921e52347bdSJani Nikula 5922f117955aSGuilherme G. Piccoli softlockup_all_cpu_backtrace= 5923e52347bdSJani Nikula [KNL] Should the soft-lockup detector generate 5924e52347bdSJani Nikula backtraces on all cpus. 59259e1cbedeSMauro Carvalho Chehab Format: 0 | 1 5926e52347bdSJani Nikula 5927d414b401SPawan Gupta sonypi.*= [HW] Sony Programmable I/O Control Device driver 59289c9cd014SJosh Poimboeuf See Documentation/admin-guide/laptops/sonypi.rst 5929d414b401SPawan Gupta 5930d414b401SPawan Gupta spectre_bhi= [X86] Control mitigation of Branch History Injection 5931d414b401SPawan Gupta (BHI) vulnerability. This setting affects the 593239053a34SJosh Poimboeuf deployment of the HW BHI control and the SW BHB 593339053a34SJosh Poimboeuf clearing sequence. 593439053a34SJosh Poimboeuf 5935d414b401SPawan Gupta on - (default) Enable the HW or SW mitigation 5936da285121SDavid Woodhouse as needed. 5937da285121SDavid Woodhouse off - Disable the mitigation. 5938fa1202efSThomas Gleixner 5939fa1202efSThomas Gleixner spectre_v2= [X86] Control mitigation of Spectre variant 2 5940da285121SDavid Woodhouse (indirect branch speculation) vulnerability. 5941fa1202efSThomas Gleixner The default operation protects the kernel from 5942fa1202efSThomas Gleixner user space attacks. 5943fa1202efSThomas Gleixner 5944fa1202efSThomas Gleixner on - unconditionally enable, implies 5945da285121SDavid Woodhouse spectre_v2_user=on 5946da285121SDavid Woodhouse off - unconditionally disable, implies 5947da285121SDavid Woodhouse spectre_v2_user=off 5948da285121SDavid Woodhouse auto - kernel detects whether your CPU model is 5949da285121SDavid Woodhouse vulnerable 5950da285121SDavid Woodhouse 5951da285121SDavid Woodhouse Selecting 'on' will, and 'auto' may, choose a 5952da285121SDavid Woodhouse mitigation method at run time according to the 5953da285121SDavid Woodhouse CPU, the available microcode, the setting of the 5954fa1202efSThomas Gleixner CONFIG_RETPOLINE configuration option, and the 5955fa1202efSThomas Gleixner compiler with which the kernel was built. 5956fa1202efSThomas Gleixner 5957fa1202efSThomas Gleixner Selecting 'on' will also enable the mitigation 5958fa1202efSThomas Gleixner against user space to user space task attacks. 5959fa1202efSThomas Gleixner 5960da285121SDavid Woodhouse Selecting 'off' will disable both the kernel and 5961da285121SDavid Woodhouse the user space protections. 5962da285121SDavid Woodhouse 59635ad3eb11SPeter Zijlstra Specific mitigations can also be selected manually: 59645ad3eb11SPeter Zijlstra 59655ad3eb11SPeter Zijlstra retpoline - replace indirect branches 5966e7862edaSKim Phillips retpoline,generic - Retpolines 5967e7862edaSKim Phillips retpoline,lfence - LFENCE; indirect branch 5968e7862edaSKim Phillips retpoline,amd - alias for retpoline,lfence 59697c693f54SPawan Gupta eibrs - Enhanced/Auto IBRS 5970da285121SDavid Woodhouse eibrs,retpoline - Enhanced/Auto IBRS + Retpolines 5971da285121SDavid Woodhouse eibrs,lfence - Enhanced/Auto IBRS + LFENCE 5972da285121SDavid Woodhouse ibrs - use IBRS to protect kernel 5973da285121SDavid Woodhouse 5974fa1202efSThomas Gleixner Not specifying this option is equivalent to 5975fa1202efSThomas Gleixner spectre_v2=auto. 5976fa1202efSThomas Gleixner 5977fa1202efSThomas Gleixner spectre_v2_user= 5978fa1202efSThomas Gleixner [X86] Control mitigation of Spectre variant 2 5979fa1202efSThomas Gleixner (indirect branch speculation) vulnerability between 5980fa1202efSThomas Gleixner user space tasks 5981fa1202efSThomas Gleixner 5982fa1202efSThomas Gleixner on - Unconditionally enable mitigations. Is 5983fa1202efSThomas Gleixner enforced by spectre_v2=on 5984fa1202efSThomas Gleixner 59857cc765a6SThomas Gleixner off - Unconditionally disable mitigations. Is 59867cc765a6SThomas Gleixner enforced by spectre_v2=off 59877cc765a6SThomas Gleixner 59887cc765a6SThomas Gleixner prctl - Indirect branch speculation is enabled, 59897cc765a6SThomas Gleixner but mitigation can be enabled via prctl 599055a97402SThomas Gleixner per thread. The mitigation control state 599155a97402SThomas Gleixner is inherited on fork. 599255a97402SThomas Gleixner 599355a97402SThomas Gleixner prctl,ibpb 599455a97402SThomas Gleixner - Like "prctl" above, but only STIBP is 599555a97402SThomas Gleixner controlled per thread. IBPB is issued 59966b3e64c2SThomas Gleixner always when switching between different user 59976b3e64c2SThomas Gleixner space processes. 59986b3e64c2SThomas Gleixner 59996b3e64c2SThomas Gleixner seccomp 60006b3e64c2SThomas Gleixner - Same as "prctl" above, but all seccomp 600155a97402SThomas Gleixner threads will enable the mitigation unless 600255a97402SThomas Gleixner they explicitly opt out. 600355a97402SThomas Gleixner 600455a97402SThomas Gleixner seccomp,ibpb 600555a97402SThomas Gleixner - Like "seccomp" above, but only STIBP is 600655a97402SThomas Gleixner controlled per thread. IBPB is issued 6007fa1202efSThomas Gleixner always when switching between different 6008fa1202efSThomas Gleixner user space processes. 60096b3e64c2SThomas Gleixner 60102f46993dSAndrea Arcangeli auto - Kernel selects the mitigation depending on 6011fa1202efSThomas Gleixner the available CPU features and vulnerability. 6012fa1202efSThomas Gleixner 6013fa1202efSThomas Gleixner Default mitigation: "prctl" 6014fa1202efSThomas Gleixner 6015fb3bd914SBorislav Petkov (AMD) Not specifying this option is equivalent to 6016fb3bd914SBorislav Petkov (AMD) spectre_v2_user=auto. 6017fb3bd914SBorislav Petkov (AMD) 6018fb3bd914SBorislav Petkov (AMD) spec_rstack_overflow= 6019fb3bd914SBorislav Petkov (AMD) [X86] Control RAS overflow mitigation on AMD Zen CPUs 6020fb3bd914SBorislav Petkov (AMD) 6021fb3bd914SBorislav Petkov (AMD) off - Disable mitigation 6022fb3bd914SBorislav Petkov (AMD) microcode - Enable microcode mitigation only 6023fb3bd914SBorislav Petkov (AMD) safe-ret - Enable sw-only safe RET mitigation (default) 6024fb3bd914SBorislav Petkov (AMD) ibpb - Enable mitigation by issuing IBPB on 6025fb3bd914SBorislav Petkov (AMD) kernel entry 602624f7fc83SKonrad Rzeszutek Wilk ibpb-vmexit - Issue IBPB only on VMEXIT 602724f7fc83SKonrad Rzeszutek Wilk (cloud-specific mitigation) 602824f7fc83SKonrad Rzeszutek Wilk 602924f7fc83SKonrad Rzeszutek Wilk spec_store_bypass_disable= 603024f7fc83SKonrad Rzeszutek Wilk [HW] Control Speculative Store Bypass (SSB) Disable mitigation 603124f7fc83SKonrad Rzeszutek Wilk (Speculative Store Bypass vulnerability) 603224f7fc83SKonrad Rzeszutek Wilk 603324f7fc83SKonrad Rzeszutek Wilk Certain CPUs are vulnerable to an exploit against a 603424f7fc83SKonrad Rzeszutek Wilk a common industry wide performance optimization known 603524f7fc83SKonrad Rzeszutek Wilk as "Speculative Store Bypass" in which recent stores 603624f7fc83SKonrad Rzeszutek Wilk to the same memory location may not be observed by 603724f7fc83SKonrad Rzeszutek Wilk later loads during speculative execution. The idea 603824f7fc83SKonrad Rzeszutek Wilk is that such stores are unlikely and that they can 603924f7fc83SKonrad Rzeszutek Wilk be detected prior to instruction retirement at the 604024f7fc83SKonrad Rzeszutek Wilk end of a particular speculation execution window. 604124f7fc83SKonrad Rzeszutek Wilk 604224f7fc83SKonrad Rzeszutek Wilk In vulnerable processors, the speculatively forwarded 604324f7fc83SKonrad Rzeszutek Wilk store can be used in a cache side channel attack, for 604424f7fc83SKonrad Rzeszutek Wilk example to read memory to which the attacker does not 604524f7fc83SKonrad Rzeszutek Wilk directly have access (e.g. inside sandboxed code). 604624f7fc83SKonrad Rzeszutek Wilk 60476b4c1360SMichael Ellerman This parameter controls whether the Speculative Store 60486b4c1360SMichael Ellerman Bypass optimization is used. 604924f7fc83SKonrad Rzeszutek Wilk 605024f7fc83SKonrad Rzeszutek Wilk On x86 the options are: 605124f7fc83SKonrad Rzeszutek Wilk 605224f7fc83SKonrad Rzeszutek Wilk on - Unconditionally disable Speculative Store Bypass 6053f21b53b2SKees Cook off - Unconditionally enable Speculative Store Bypass 6054f21b53b2SKees Cook auto - Kernel detects whether the CPU model contains an 6055f21b53b2SKees Cook implementation of Speculative Store Bypass and 6056f21b53b2SKees Cook picks the most appropriate mitigation. If the 6057a73ec77eSThomas Gleixner CPU is not vulnerable, "off" is selected. If the 6058a73ec77eSThomas Gleixner CPU is vulnerable the default mitigation is 6059a73ec77eSThomas Gleixner architecture and Kconfig dependent. See below. 6060a73ec77eSThomas Gleixner prctl - Control Speculative Store Bypass per thread 6061f21b53b2SKees Cook via prctl. Speculative Store Bypass is enabled 6062f21b53b2SKees Cook for a process by default. The state of the control 606324f7fc83SKonrad Rzeszutek Wilk is inherited on fork. 6064f21b53b2SKees Cook seccomp - Same as "prctl" above, but all seccomp threads 60652f46993dSAndrea Arcangeli will disable SSB unless they explicitly opt out. 6066f21b53b2SKees Cook 60676b4c1360SMichael Ellerman Default mitigations: 60686b4c1360SMichael Ellerman X86: "prctl" 60696b4c1360SMichael Ellerman 60706b4c1360SMichael Ellerman On powerpc the options are: 60716b4c1360SMichael Ellerman 60726b4c1360SMichael Ellerman on,auto - On Power8 and Power9 insert a store-forwarding 60736b4c1360SMichael Ellerman barrier on kernel entry and exit. On Power7 60746b4c1360SMichael Ellerman perform a software flush on kernel entry and 60756b4c1360SMichael Ellerman exit. 60766b4c1360SMichael Ellerman off - No action. 60776b4c1360SMichael Ellerman 6078e52347bdSJani Nikula Not specifying this option is equivalent to 6079e52347bdSJani Nikula spec_store_bypass_disable=auto. 6080e52347bdSJani Nikula 6081e52347bdSJani Nikula spia_io_base= [HW,MTD] 6082e52347bdSJani Nikula spia_fio_base= 60836650cdd9SPeter Zijlstra (Intel) spia_pedr= 6084ebca1770SFenghua Yu spia_peddr= 60856650cdd9SPeter Zijlstra (Intel) 60866650cdd9SPeter Zijlstra (Intel) split_lock_detect= 60876650cdd9SPeter Zijlstra (Intel) [X86] Enable split lock detection or bus lock detection 6088ebca1770SFenghua Yu 6089ebca1770SFenghua Yu When enabled (and if hardware support is present), atomic 6090ebca1770SFenghua Yu instructions that access data across cache line 60916650cdd9SPeter Zijlstra (Intel) boundaries will result in an alignment check exception 60926650cdd9SPeter Zijlstra (Intel) for split lock detection or a debug exception for 60936650cdd9SPeter Zijlstra (Intel) bus lock detection. 6094ebca1770SFenghua Yu 60956650cdd9SPeter Zijlstra (Intel) off - not enabled 6096ebca1770SFenghua Yu 6097ebca1770SFenghua Yu warn - the kernel will emit rate-limited warnings 6098ebca1770SFenghua Yu about applications triggering the #AC 6099ebca1770SFenghua Yu exception or the #DB exception. This mode is 6100ebca1770SFenghua Yu the default on CPUs that support split lock 61016650cdd9SPeter Zijlstra (Intel) detection or bus lock detection. Default 61026650cdd9SPeter Zijlstra (Intel) behavior is by #AC if both features are 6103ebca1770SFenghua Yu enabled in hardware. 6104ebca1770SFenghua Yu 6105ebca1770SFenghua Yu fatal - the kernel will send SIGBUS to applications 61066650cdd9SPeter Zijlstra (Intel) that trigger the #AC exception or the #DB 61079d839c28SFenghua Yu exception. Default behavior is by #AC if 61089d839c28SFenghua Yu both features are enabled in hardware. 61099d839c28SFenghua Yu 61109d839c28SFenghua Yu ratelimit:N - 61119d839c28SFenghua Yu Set system wide rate limit to N bus locks 61129d839c28SFenghua Yu per second for bus lock detection. 61139d839c28SFenghua Yu 0 < N <= 1000. 61149d839c28SFenghua Yu 61156650cdd9SPeter Zijlstra (Intel) N/A for split lock detection. 61166650cdd9SPeter Zijlstra (Intel) 61176650cdd9SPeter Zijlstra (Intel) 61186650cdd9SPeter Zijlstra (Intel) If an #AC exception is hit in the kernel or in 61196650cdd9SPeter Zijlstra (Intel) firmware (i.e. not while executing in user mode) 6120ebca1770SFenghua Yu the kernel will oops in either "warn" or "fatal" 6121ebca1770SFenghua Yu mode. 6122ebca1770SFenghua Yu 61237e5b3c26SMark Gross #DB exception for bus lock is triggered only when 61247e5b3c26SMark Gross CPL > 0. 61257e5b3c26SMark Gross 61267e5b3c26SMark Gross srbds= [X86,INTEL] 61277e5b3c26SMark Gross Control the Special Register Buffer Data Sampling 61287e5b3c26SMark Gross (SRBDS) mitigation. 61297e5b3c26SMark Gross 61307e5b3c26SMark Gross Certain CPUs are vulnerable to an MDS-like 61317e5b3c26SMark Gross exploit which can leak bits from the random 61327e5b3c26SMark Gross number generator. 61337e5b3c26SMark Gross 61347e5b3c26SMark Gross By default, this issue is mitigated by 61357e5b3c26SMark Gross microcode. However, the microcode fix can cause 61367e5b3c26SMark Gross the RDRAND and RDSEED instructions to become 61377e5b3c26SMark Gross much slower. Among other effects, this will 61387e5b3c26SMark Gross result in reduced throughput from /dev/urandom. 61397e5b3c26SMark Gross 61407e5b3c26SMark Gross The microcode mitigation can be disabled with 61417e5b3c26SMark Gross the following option: 61427e5b3c26SMark Gross 6143a57ffb3cSPaul E. McKenney off: Disable mitigation and remove 6144a57ffb3cSPaul E. McKenney performance impact to RDRAND and RDSEED 6145a57ffb3cSPaul E. McKenney 6146a57ffb3cSPaul E. McKenney srcutree.big_cpu_lim [KNL] 6147a57ffb3cSPaul E. McKenney Specifies the number of CPUs constituting a 6148a57ffb3cSPaul E. McKenney large system, such that srcu_struct structures 6149a57ffb3cSPaul E. McKenney should immediately allocate an srcu_node array. 6150a57ffb3cSPaul E. McKenney This kernel-boot parameter defaults to 128, 6151a57ffb3cSPaul E. McKenney but takes effect only when the low-order four 6152c69a00a1SPaul E. McKenney bits of srcutree.convert_to_big is equal to 3 6153c69a00a1SPaul E. McKenney (decide at boot). 6154c69a00a1SPaul E. McKenney 6155c69a00a1SPaul E. McKenney srcutree.convert_to_big [KNL] 6156c69a00a1SPaul E. McKenney Specifies under what conditions an SRCU tree 6157c69a00a1SPaul E. McKenney srcu_struct structure will be converted to big 6158c69a00a1SPaul E. McKenney form, that is, with an rcu_node tree: 6159c69a00a1SPaul E. McKenney 6160a57ffb3cSPaul E. McKenney 0: Never. 61619f2e91d9SPaul E. McKenney 1: At init_srcu_struct() time. 6162c69a00a1SPaul E. McKenney 2: When rcutorture decides to. 6163c69a00a1SPaul E. McKenney 3: Decide at boot time (default). 6164c69a00a1SPaul E. McKenney 0x1X: Above plus if high contention. 6165c69a00a1SPaul E. McKenney 6166c69a00a1SPaul E. McKenney Either way, the srcu_node tree will be sized based 6167c350c008SPaul E. McKenney on the actual runtime number of CPUs (nr_cpu_ids) 6168c350c008SPaul E. McKenney instead of the compile-time CONFIG_NR_CPUS. 6169c350c008SPaul E. McKenney 6170c350c008SPaul E. McKenney srcutree.counter_wrap_check [KNL] 6171c350c008SPaul E. McKenney Specifies how frequently to check for 6172c350c008SPaul E. McKenney grace-period sequence counter wrap for the 6173c350c008SPaul E. McKenney srcu_data structure's ->srcu_gp_seq_needed field. 6174c350c008SPaul E. McKenney The greater the number of bits set in this kernel 6175c350c008SPaul E. McKenney parameter, the less frequently counter wrap will 617622607d66SPaul E. McKenney be checked for. Note that the bottom two bits 617722607d66SPaul E. McKenney are ignored. 617822607d66SPaul E. McKenney 617922607d66SPaul E. McKenney srcutree.exp_holdoff [KNL] 618022607d66SPaul E. McKenney Specifies how many nanoseconds must elapse 618122607d66SPaul E. McKenney since the end of the last SRCU grace period for 618222607d66SPaul E. McKenney a given srcu_struct until the next normal SRCU 618322607d66SPaul E. McKenney grace period will be considered for automatic 61844f2bfd94SNeeraj Upadhyay expediting. Set to zero to disable automatic 61854f2bfd94SNeeraj Upadhyay expediting. 61864f2bfd94SNeeraj Upadhyay 61874f2bfd94SNeeraj Upadhyay srcutree.srcu_max_nodelay [KNL] 61884f2bfd94SNeeraj Upadhyay Specifies the number of no-delay instances 61894f2bfd94SNeeraj Upadhyay per jiffy for which the SRCU grace period 61904f2bfd94SNeeraj Upadhyay worker thread will be rescheduled with zero 61914f2bfd94SNeeraj Upadhyay delay. Beyond this limit, worker thread will 61924f2bfd94SNeeraj Upadhyay be rescheduled with a sleep delay of one jiffy. 61934f2bfd94SNeeraj Upadhyay 61944f2bfd94SNeeraj Upadhyay srcutree.srcu_max_nodelay_phase [KNL] 61954f2bfd94SNeeraj Upadhyay Specifies the per-grace-period phase, number of 61964f2bfd94SNeeraj Upadhyay non-sleeping polls of readers. Beyond this limit, 61974f2bfd94SNeeraj Upadhyay grace period worker thread will be rescheduled 61984f2bfd94SNeeraj Upadhyay with a sleep delay of one jiffy, between each 61994f2bfd94SNeeraj Upadhyay rescan of the readers, for a grace period phase. 62004f2bfd94SNeeraj Upadhyay 62014f2bfd94SNeeraj Upadhyay srcutree.srcu_retry_check_delay [KNL] 62029f2e91d9SPaul E. McKenney Specifies number of microseconds of non-sleeping 62039f2e91d9SPaul E. McKenney delay between each non-sleeping poll of readers. 62049f2e91d9SPaul E. McKenney 62059f2e91d9SPaul E. McKenney srcutree.small_contention_lim [KNL] 62069f2e91d9SPaul E. McKenney Specifies the number of update-side contention 62079f2e91d9SPaul E. McKenney events per jiffy will be tolerated before 62089f2e91d9SPaul E. McKenney initiating a conversion of an srcu_struct 62099f2e91d9SPaul E. McKenney structure to big form. Note that the value of 6210a43ae4dfSMarc Zyngier srcutree.convert_to_big must have the 0x10 bit 6211a43ae4dfSMarc Zyngier set for contention-based conversions to occur. 6212a43ae4dfSMarc Zyngier 6213a43ae4dfSMarc Zyngier ssbd= [ARM64,HW] 6214a43ae4dfSMarc Zyngier Speculative Store Bypass Disable control 6215a43ae4dfSMarc Zyngier 6216a43ae4dfSMarc Zyngier On CPUs that are vulnerable to the Speculative 6217a43ae4dfSMarc Zyngier Store Bypass vulnerability and offer a 6218a43ae4dfSMarc Zyngier firmware based mitigation, this parameter 6219a43ae4dfSMarc Zyngier indicates how the mitigation should be used: 6220a43ae4dfSMarc Zyngier 6221a43ae4dfSMarc Zyngier force-on: Unconditionally enable mitigation for 6222a43ae4dfSMarc Zyngier for both kernel and userspace 6223a43ae4dfSMarc Zyngier force-off: Unconditionally disable mitigation for 6224a43ae4dfSMarc Zyngier for both kernel and userspace 6225a43ae4dfSMarc Zyngier kernel: Always enable mitigation in the 6226a43ae4dfSMarc Zyngier kernel, and offer a prctl interface 62271be7107fSHugh Dickins to allow userspace to register its 62281be7107fSHugh Dickins interest in being mitigated too. 62291be7107fSHugh Dickins 62301be7107fSHugh Dickins stack_guard_gap= [MM] 62311be7107fSHugh Dickins override the default stack gap protection. The value 62321be7107fSHugh Dickins is in page units and it defines how many pages prior 62331be7107fSHugh Dickins to (for stacks growing down) resp. after (for stacks 6234e1fdc403SVijayanand Jitta growing up) the main stack are reserved for no other 6235e1fdc403SVijayanand Jitta mapping. Default value is 256 pages. 6236e1fdc403SVijayanand Jitta 6237e1fdc403SVijayanand Jitta stack_depot_disable= [KNL] 6238e1fdc403SVijayanand Jitta Setting this to true through kernel command line will 6239e1fdc403SVijayanand Jitta disable the stack depot thereby saving the static memory 6240e52347bdSJani Nikula consumed by the stack hash table. By default this is set 6241e52347bdSJani Nikula to false. 6242e52347bdSJani Nikula 6243e52347bdSJani Nikula stacktrace [FTRACE] 6244e52347bdSJani Nikula Enabled the stack tracer on boot up. 624525942e5eSRandy Dunlap 6246e52347bdSJani Nikula stacktrace_filter=[function-list] 6247e52347bdSJani Nikula [FTRACE] Limit the functions that the stack tracer 6248e52347bdSJani Nikula will trace at boot up. function-list is a comma-separated 6249e52347bdSJani Nikula list of functions. This list can be changed at run 6250e52347bdSJani Nikula time by the stack_trace_filter file in the debugfs 6251e52347bdSJani Nikula tracing directory. Note, this enables stack tracing 6252e52347bdSJani Nikula and the stacktrace above is not needed. 6253e52347bdSJani Nikula 6254e52347bdSJani Nikula sti= [PARISC,HW] 6255e52347bdSJani Nikula Format: <num> 6256e52347bdSJani Nikula Set the STI (builtin display/keyboard on the HP-PARISC 6257e52347bdSJani Nikula machines) console (graphic card) which should be used 6258e52347bdSJani Nikula as the initial boot-console. 6259e52347bdSJani Nikula See also comment in drivers/video/console/sticore.c. 6260e52347bdSJani Nikula 6261e52347bdSJani Nikula sti_font= [HW] 6262e52347bdSJani Nikula See comment in drivers/video/console/sticore.c. 6263e52347bdSJani Nikula 62643aac3ebeSThomas Gleixner stifb= [HW] 62653aac3ebeSThomas Gleixner Format: bpp:<bpp1>[:<bpp2>[:<bpp3>...]] 62663aac3ebeSThomas Gleixner 62673aac3ebeSThomas Gleixner strict_sas_size= 62683aac3ebeSThomas Gleixner [X86] 62693aac3ebeSThomas Gleixner Format: <bool> 62703aac3ebeSThomas Gleixner Enable or disable strict sigaltstack size checks 62713aac3ebeSThomas Gleixner against the required signal frame size which 62723aac3ebeSThomas Gleixner depends on the supported FPU features. This can 627342551b8dSRandy Dunlap be used to filter out binaries which have 627442551b8dSRandy Dunlap not yet been made aware of AT_MINSIGSTKSZ. 627542551b8dSRandy Dunlap 627642551b8dSRandy Dunlap stress_hpt [PPC] 627742551b8dSRandy Dunlap Limits the number of kernel HPT entries in the hash 627842551b8dSRandy Dunlap page table to increase the rate of hash page table 627942551b8dSRandy Dunlap faults on kernel addresses. 628042551b8dSRandy Dunlap 628142551b8dSRandy Dunlap stress_slb [PPC] 628242551b8dSRandy Dunlap Limits the number of kernel SLB entries, and flushes 6283e52347bdSJani Nikula them frequently to increase the rate of SLB faults 6284e52347bdSJani Nikula on kernel addresses. 6285e52347bdSJani Nikula 6286e52347bdSJani Nikula sunrpc.min_resvport= 6287e52347bdSJani Nikula sunrpc.max_resvport= 6288e52347bdSJani Nikula [NFS,SUNRPC] 6289e52347bdSJani Nikula SunRPC servers often require that client requests 6290e52347bdSJani Nikula originate from a privileged port (i.e. a port in the 6291e52347bdSJani Nikula range 0 < portnr < 1024). 6292e52347bdSJani Nikula An administrator who wishes to reserve some of these 6293e52347bdSJani Nikula ports for other uses may adjust the range that the 6294e52347bdSJani Nikula kernel's sunrpc client considers to be privileged 6295e52347bdSJani Nikula using these two parameters to set the minimum and 6296e52347bdSJani Nikula maximum port values. 6297e52347bdSJani Nikula 6298e52347bdSJani Nikula sunrpc.svc_rpc_per_connection_limit= 6299e52347bdSJani Nikula [NFS,SUNRPC] 6300e52347bdSJani Nikula Limit the number of requests that the server will 6301e52347bdSJani Nikula process in parallel from a single connection. 6302e52347bdSJani Nikula The default value is 0 (no limit). 6303e52347bdSJani Nikula 6304e52347bdSJani Nikula sunrpc.pool_mode= 6305e52347bdSJani Nikula [NFS] 6306e52347bdSJani Nikula Control how the NFS server code allocates CPUs to 6307e52347bdSJani Nikula service thread pools. Depending on how many NICs 6308e52347bdSJani Nikula you have and where their interrupts are bound, this 6309e52347bdSJani Nikula option will affect which CPUs will do NFS serving. 6310e52347bdSJani Nikula Note: this parameter cannot be changed while the 6311e52347bdSJani Nikula NFS server is running. 6312e52347bdSJani Nikula 6313e52347bdSJani Nikula auto the server chooses an appropriate mode 6314e52347bdSJani Nikula automatically using heuristics 6315e52347bdSJani Nikula global a single global pool contains all CPUs 6316e52347bdSJani Nikula percpu one pool for each CPU 6317e52347bdSJani Nikula pernode one pool for each NUMA node (equivalent 6318e52347bdSJani Nikula to global on non-NUMA machines) 6319e52347bdSJani Nikula 6320e52347bdSJani Nikula sunrpc.tcp_slot_table_entries= 6321e52347bdSJani Nikula sunrpc.udp_slot_table_entries= 6322e52347bdSJani Nikula [NFS,SUNRPC] 6323e52347bdSJani Nikula Sets the upper limit on the number of simultaneous 6324e52347bdSJani Nikula RPC calls that can be sent from the client to a 6325e52347bdSJani Nikula server. Increasing these values may allow you to 6326e52347bdSJani Nikula improve throughput, but will also increase the 6327e52347bdSJani Nikula amount of memory reserved for use by the client. 6328e52347bdSJani Nikula 6329e52347bdSJani Nikula suspend.pm_test_delay= 6330e52347bdSJani Nikula [SUSPEND] 6331e52347bdSJani Nikula Sets the number of seconds to remain in a suspend test 6332e52347bdSJani Nikula mode before resuming the system (see 63336a9c930bSRam Pai /sys/power/pm_test). Only available when CONFIG_PM_DEBUG 63346a9c930bSRam Pai is set. Default value is 5. 63356a9c930bSRam Pai 63366a9c930bSRam Pai svm= [PPC] 63376a9c930bSRam Pai Format: { on | off | y | n | 1 | 0 } 6338e52347bdSJani Nikula This parameter controls use of the Protected 633920347fcaSTianyu Lan Execution Facility on pSeries. 6340e52347bdSJani Nikula 634120347fcaSTianyu Lan swiotlb= [ARM,IA-64,PPC,MIPS,X86] 634272311809STianyu Lan Format: { <int> [,<int>] | force | noforce } 634372311809STianyu Lan <int> -- Number of I/O TLB slabs 6344e52347bdSJani Nikula <int> -- Second integer after comma. Number of swiotlb 6345e52347bdSJani Nikula areas with their own lock. Will be rounded up 6346fff5d992SGeert Uytterhoeven to a power of 2. 6347e52347bdSJani Nikula force -- force using of bounce buffers even if they 6348e52347bdSJani Nikula wouldn't be automatically used by the kernel 6349e52347bdSJani Nikula noforce -- Never use bounce buffers (for debugging) 63503db978d4SVlastimil Babka 63513db978d4SVlastimil Babka switches= [HW,M68k] 63523db978d4SVlastimil Babka 63533db978d4SVlastimil Babka sysctl.*= [KNL] 63543db978d4SVlastimil Babka Set a sysctl parameter, right before loading the init 63553db978d4SVlastimil Babka process, as if the value was written to the respective 63563db978d4SVlastimil Babka /proc/sys/... file. Both '.' and '/' are recognized as 63573db978d4SVlastimil Babka separators. Unrecognized parameters and invalid values 63583db978d4SVlastimil Babka are reported in the kernel log. Sysctls registered 6359e52347bdSJani Nikula later by a loaded module cannot be set this way. 6360e52347bdSJani Nikula Example: sysctl.vm.swappiness=40 6361e52347bdSJani Nikula 6362e52347bdSJani Nikula sysrq_always_enabled 6363e52347bdSJani Nikula [KNL] 6364e52347bdSJani Nikula Ignore sysrq setting - this boot parameter will 6365e52347bdSJani Nikula neutralize any effect of /proc/sys/kernel/sysrq. 6366e52347bdSJani Nikula Useful for debugging. 6367e52347bdSJani Nikula 6368e52347bdSJani Nikula tcpmhash_entries= [KNL,NET] 63691cec2cacSMauro Carvalho Chehab Set the number of tcp_metrics_hash slots. 6370e52347bdSJani Nikula Default value is 8192 or 16384 depending on total 6371e52347bdSJani Nikula ram pages. This is used to specify the TCP metrics 6372e52347bdSJani Nikula cache size. See Documentation/networking/ip-sysctl.rst 6373e52347bdSJani Nikula "tcp_no_metrics_save" section for more details. 637459bdbbd5SRandy Dunlap 637559bdbbd5SRandy Dunlap tdfx= [HW,DRM] 6376e52347bdSJani Nikula 6377e52347bdSJani Nikula test_suspend= [SUSPEND] 6378e52347bdSJani Nikula Format: { "mem" | "standby" | "freeze" }[,N] 6379e52347bdSJani Nikula Specify "mem" (for Suspend-to-RAM) or "standby" (for 6380e52347bdSJani Nikula standby suspend) or "freeze" (for suspend type freeze) 6381e52347bdSJani Nikula as the system sleep state during system startup with 6382e52347bdSJani Nikula the optional capability to repeat N number of times. 6383e52347bdSJani Nikula The system is woken from this state using a 6384e52347bdSJani Nikula wakeup-capable RTC alarm. 6385e52347bdSJani Nikula 6386e52347bdSJani Nikula thash_entries= [KNL,NET] 6387e52347bdSJani Nikula Set number of hash buckets for TCP connection 6388e52347bdSJani Nikula 6389e52347bdSJani Nikula thermal.act= [HW,ACPI] 6390e52347bdSJani Nikula -1: disable all active trip points in all thermal zones 6391e52347bdSJani Nikula <degrees C>: override all lowest active trip points 6392e52347bdSJani Nikula 6393e52347bdSJani Nikula thermal.crt= [HW,ACPI] 6394e52347bdSJani Nikula -1: disable all critical trip points in all thermal zones 6395e52347bdSJani Nikula <degrees C>: override all critical trip points 6396e52347bdSJani Nikula 6397e52347bdSJani Nikula thermal.off= [HW,ACPI] 6398e52347bdSJani Nikula 1: disable ACPI thermal control 6399e52347bdSJani Nikula 6400e52347bdSJani Nikula thermal.psv= [HW,ACPI] 6401e52347bdSJani Nikula -1: disable all passive trip points 6402e52347bdSJani Nikula <degrees C>: override all passive trip points to this 6403e52347bdSJani Nikula value 6404e52347bdSJani Nikula 6405e52347bdSJani Nikula thermal.tzp= [HW,ACPI] 6406e52347bdSJani Nikula Specify global default ACPI thermal zone polling rate 6407e52347bdSJani Nikula <deci-seconds>: poll all this frequency 6408e52347bdSJani Nikula 0: no polling (default) 6409e52347bdSJani Nikula 6410e52347bdSJani Nikula threadirqs [KNL] 6411e52347bdSJani Nikula Force threading of all interrupt handlers except those 6412e52347bdSJani Nikula marked explicitly IRQF_NO_THREAD. 6413e52347bdSJani Nikula 6414e52347bdSJani Nikula topology= [S390] 6415e52347bdSJani Nikula Format: {off | on} 6416e52347bdSJani Nikula Specify if the kernel should make use of the cpu 6417e52347bdSJani Nikula topology information if the hardware supports this. 6418e52347bdSJani Nikula The scheduler will make use of this information and 6419e52347bdSJani Nikula e.g. base its process migration decisions on it. 6420e52347bdSJani Nikula Default is on. 6421e52347bdSJani Nikula 6422e52347bdSJani Nikula topology_updates= [KNL, PPC, NUMA] 6423e52347bdSJani Nikula Format: {off} 6424e52347bdSJani Nikula Specify if the kernel should ignore (off) 64258171d3e0SPaul E. McKenney topology updates sent by the hypervisor to this 64268171d3e0SPaul E. McKenney LPAR. 64278171d3e0SPaul E. McKenney 64288171d3e0SPaul E. McKenney torture.disable_onoff_at_boot= [KNL] 64292102ad29SPaul E. McKenney Prevent the CPU-hotplug component of torturing 64302102ad29SPaul E. McKenney until after init has spawned. 64312102ad29SPaul E. McKenney 64322102ad29SPaul E. McKenney torture.ftrace_dump_at_shutdown= [KNL] 64332102ad29SPaul E. McKenney Dump the ftrace buffer at torture-test shutdown, 64342102ad29SPaul E. McKenney even if there were no errors. This can be a 64352102ad29SPaul E. McKenney very costly operation when many torture tests 64368a67a20bSPaul E. McKenney are running concurrently, especially on systems 64378a67a20bSPaul E. McKenney with rotating-rust storage. 64388a67a20bSPaul E. McKenney 64398a67a20bSPaul E. McKenney torture.verbose_sleep_frequency= [KNL] 64408a67a20bSPaul E. McKenney Specifies how many verbose printk()s should be 64418a67a20bSPaul E. McKenney emitted between each sleep. The default of zero 64428a67a20bSPaul E. McKenney disables verbose-printk() sleeping. 64438a67a20bSPaul E. McKenney 6444e52347bdSJani Nikula torture.verbose_sleep_duration= [KNL] 6445e52347bdSJani Nikula Duration of each verbose-printk() sleep in jiffies. 6446e52347bdSJani Nikula 6447e52347bdSJani Nikula tp720= [HW,PS2] 6448e52347bdSJani Nikula 6449e52347bdSJani Nikula tpm_suspend_pcr=[HW,TPM] 6450e52347bdSJani Nikula Format: integer pcr id 6451e52347bdSJani Nikula Specify that at suspend time, the tpm driver 6452e52347bdSJani Nikula should extend the specified pcr with zeros, 6453e52347bdSJani Nikula as a workaround for some chips which fail to 6454e52347bdSJani Nikula flush the last written pcr on TPM_SaveState. 6455bff24699SJarkko Sakkinen This will guarantee that all the other pcrs 6456bff24699SJarkko Sakkinen are saved. 6457bff24699SJarkko Sakkinen 6458bff24699SJarkko Sakkinen tpm_tis.interrupts= [HW,TPM] 6459bff24699SJarkko Sakkinen Enable interrupts for the MMIO based physical layer 6460bff24699SJarkko Sakkinen for the FIFO interface. By default it is set to false 6461bff24699SJarkko Sakkinen (0). For more information about TPM hardware interfaces 6462e52347bdSJani Nikula defined by Trusted Computing Group (TCG) see 6463e52347bdSJani Nikula https://trustedcomputinggroup.org/resource/pc-client-platform-tpm-profile-ptp-specification/ 6464e52347bdSJani Nikula 6465e52347bdSJani Nikula tp_printk [FTRACE] 6466e52347bdSJani Nikula Have the tracepoints sent to printk as well as the 6467e52347bdSJani Nikula tracing ring buffer. This is useful for early boot up 6468e52347bdSJani Nikula where the system hangs or reboots and does not give the 6469e52347bdSJani Nikula option for reading the tracing buffer or performing a 6470e52347bdSJani Nikula ftrace_dump_on_oops. 6471e52347bdSJani Nikula 6472e52347bdSJani Nikula To turn off having tracepoints sent to printk, 6473e52347bdSJani Nikula echo 0 > /proc/sys/kernel/tracepoint_printk 6474f3860136SSteven Rostedt (VMware) Note, echoing 1 into this file without the 6475f3860136SSteven Rostedt (VMware) tracepoint_printk kernel cmdline option has no effect. 6476f3860136SSteven Rostedt (VMware) 6477f3860136SSteven Rostedt (VMware) The tp_printk_stop_on_boot (see below) can also be used 6478e52347bdSJani Nikula to stop the printing of events to console at 6479e52347bdSJani Nikula late_initcall_sync. 6480e52347bdSJani Nikula 6481e52347bdSJani Nikula ** CAUTION ** 6482e52347bdSJani Nikula 6483e52347bdSJani Nikula Having tracepoints sent to printk() and activating high 6484f3860136SSteven Rostedt (VMware) frequency tracepoints such as irq or sched, can cause 6485f3860136SSteven Rostedt (VMware) the system to live lock. 6486f3860136SSteven Rostedt (VMware) 6487f3860136SSteven Rostedt (VMware) tp_printk_stop_on_boot [FTRACE] 6488f3860136SSteven Rostedt (VMware) When tp_printk (above) is set, it can cause a lot of noise 6489f3860136SSteven Rostedt (VMware) on the console. It may be useful to only include the 6490f3860136SSteven Rostedt (VMware) printing of events during boot up, as user space may 6491f3860136SSteven Rostedt (VMware) make the system inoperable. 6492f3860136SSteven Rostedt (VMware) 6493e52347bdSJani Nikula This command line option will stop the printing of events 6494e52347bdSJani Nikula to console at the late_initcall_sync() time frame. 6495e52347bdSJani Nikula 649659bdbbd5SRandy Dunlap trace_buf_size=nn[KMG] 649759bdbbd5SRandy Dunlap [FTRACE] will set tracing buffer size on each cpu. 649859bdbbd5SRandy Dunlap 649959bdbbd5SRandy Dunlap trace_clock= [FTRACE] Set the clock used for tracing events 650059bdbbd5SRandy Dunlap at boot up. 650159bdbbd5SRandy Dunlap local - Use the per CPU time stamp counter 650259bdbbd5SRandy Dunlap (converted into nanoseconds). Fast, but 650359bdbbd5SRandy Dunlap depending on the architecture, may not be 650459bdbbd5SRandy Dunlap in sync between CPUs. 650559bdbbd5SRandy Dunlap global - Event time stamps are synchronize across 650659bdbbd5SRandy Dunlap CPUs. May be slower than the local clock, 650759bdbbd5SRandy Dunlap but better for some race conditions. 650859bdbbd5SRandy Dunlap counter - Simple counting of events (1, 2, ..) 650959bdbbd5SRandy Dunlap note, some counts may be skipped due to the 651059bdbbd5SRandy Dunlap infrastructure grabbing the clock more than 651159bdbbd5SRandy Dunlap once per event. 651259bdbbd5SRandy Dunlap uptime - Use jiffies as the time stamp. 651359bdbbd5SRandy Dunlap perf - Use the same clock that perf uses. 651459bdbbd5SRandy Dunlap mono - Use ktime_get_mono_fast_ns() for time stamps. 651559bdbbd5SRandy Dunlap mono_raw - Use ktime_get_raw_fast_ns() for time 651659bdbbd5SRandy Dunlap stamps. 651759bdbbd5SRandy Dunlap boot - Use ktime_get_boot_fast_ns() for time stamps. 6518e52347bdSJani Nikula Architectures may add more clocks. See 6519e52347bdSJani Nikula Documentation/trace/ftrace.rst for more details. 6520e52347bdSJani Nikula 6521e52347bdSJani Nikula trace_event=[event-list] 6522e52347bdSJani Nikula [FTRACE] Set and start specified trace events in order 6523e52347bdSJani Nikula to facilitate early boot debugging. The event-list is a 6524cb1f98c5SSteven Rostedt (Google) comma-separated list of trace events to enable. See 6525cb1f98c5SSteven Rostedt (Google) also Documentation/trace/events.rst 6526cb1f98c5SSteven Rostedt (Google) 6527cb1f98c5SSteven Rostedt (Google) trace_instance=[instance-info] 6528cb1f98c5SSteven Rostedt (Google) [FTRACE] Create a ring buffer instance early in boot up. 6529cb1f98c5SSteven Rostedt (Google) This will be listed in: 6530c4846480SSteven Rostedt (Google) 6531c4846480SSteven Rostedt (Google) /sys/kernel/tracing/instances 6532c4846480SSteven Rostedt (Google) 6533c4846480SSteven Rostedt (Google) Events can be enabled at the time the instance is created 6534c4846480SSteven Rostedt (Google) via: 6535c4846480SSteven Rostedt (Google) 6536c4846480SSteven Rostedt (Google) trace_instance=<name>,<system1>:<event1>,<system2>:<event2> 6537c4846480SSteven Rostedt (Google) 6538c4846480SSteven Rostedt (Google) Note, the "<system*>:" portion is optional if the event is 6539c4846480SSteven Rostedt (Google) unique. 6540c4846480SSteven Rostedt (Google) 6541c4846480SSteven Rostedt (Google) trace_instance=foo,sched:sched_switch,irq_handler_entry,initcall 6542c4846480SSteven Rostedt (Google) 6543c4846480SSteven Rostedt (Google) will enable the "sched_switch" event (note, the "sched:" is optional, and 6544e52347bdSJani Nikula the same thing would happen if it was left off). The irq_handler_entry 6545e52347bdSJani Nikula event, and all events under the "initcall" system. 6546e52347bdSJani Nikula 6547e52347bdSJani Nikula trace_options=[option-list] 6548e52347bdSJani Nikula [FTRACE] Enable or disable tracer options at boot. 6549e52347bdSJani Nikula The option-list is a comma delimited list of options 65502abfcd29SRoss Zwisler that can be enabled or disabled just as if you were 6551e52347bdSJani Nikula to echo the option name into 6552e52347bdSJani Nikula 6553e52347bdSJani Nikula /sys/kernel/tracing/trace_options 6554e52347bdSJani Nikula 6555e52347bdSJani Nikula For example, to enable stacktrace option (to dump the 6556e52347bdSJani Nikula stack trace of each event), add to the command line: 6557e52347bdSJani Nikula 6558e52347bdSJani Nikula trace_options=stacktrace 6559e52347bdSJani Nikula 6560a01fdc89SSteven Rostedt (Google) See also Documentation/trace/ftrace.rst "trace options" 6561a01fdc89SSteven Rostedt (Google) section. 6562a01fdc89SSteven Rostedt (Google) 6563a01fdc89SSteven Rostedt (Google) trace_trigger=[trigger-list] 6564a01fdc89SSteven Rostedt (Google) [FTRACE] Add a event trigger on specific events. 6565a01fdc89SSteven Rostedt (Google) Set a trigger on top of a specific event, with an optional 6566a01fdc89SSteven Rostedt (Google) filter. 6567a01fdc89SSteven Rostedt (Google) 6568a01fdc89SSteven Rostedt (Google) The format is is "trace_trigger=<event>.<trigger>[ if <filter>],..." 6569a01fdc89SSteven Rostedt (Google) Where more than one trigger may be specified that are comma deliminated. 6570a01fdc89SSteven Rostedt (Google) 6571a01fdc89SSteven Rostedt (Google) For example: 6572a01fdc89SSteven Rostedt (Google) 6573a01fdc89SSteven Rostedt (Google) trace_trigger="sched_switch.stacktrace if prev_state == 2" 6574a01fdc89SSteven Rostedt (Google) 6575a01fdc89SSteven Rostedt (Google) The above will enable the "stacktrace" trigger on the "sched_switch" 6576a01fdc89SSteven Rostedt (Google) event but only trigger it if the "prev_state" of the "sched_switch" 6577a01fdc89SSteven Rostedt (Google) event is "2" (TASK_UNINTERUPTIBLE). 6578a01fdc89SSteven Rostedt (Google) 6579e52347bdSJani Nikula See also "Event triggers" in Documentation/trace/events.rst 6580e52347bdSJani Nikula 6581e52347bdSJani Nikula 6582e52347bdSJani Nikula traceoff_on_warning 65832abfcd29SRoss Zwisler [FTRACE] enable this option to disable tracing when a 6584e52347bdSJani Nikula warning is hit. This turns off "tracing_on". Tracing can 6585e52347bdSJani Nikula be enabled again by echoing '1' into the "tracing_on" 6586e52347bdSJani Nikula file located in /sys/kernel/tracing/ 6587e52347bdSJani Nikula 6588e52347bdSJani Nikula This option is useful, as it disables the trace before 6589e52347bdSJani Nikula the WARNING dump is called, which prevents the trace to 6590e52347bdSJani Nikula be filled with content caused by the warning output. 6591e52347bdSJani Nikula 6592e52347bdSJani Nikula This option can also be set at run time via the sysctl 6593e52347bdSJani Nikula option: kernel/traceoff_on_warning 6594e52347bdSJani Nikula 6595e52347bdSJani Nikula transparent_hugepage= 6596e52347bdSJani Nikula [KNL] 659745c9a74fSMike Rapoport Format: [always|madvise|never] 659845c9a74fSMike Rapoport Can be used to control the default behavior of the system 6599e52347bdSJani Nikula with respect to transparent hugepages. 66005d0682beSSumit Garg See Documentation/admin-guide/mm/transhuge.rst 66015d0682beSSumit Garg for more details. 66025d0682beSSumit Garg 66035d0682beSSumit Garg trusted.source= [KEYS] 66045d0682beSSumit Garg Format: <string> 66055d0682beSSumit Garg This parameter identifies the trust source as a backend 66065d0682beSSumit Garg for trusted keys implementation. Supported trust 6607e9c5048cSAhmad Fatoum sources: 66085d0682beSSumit Garg - "tpm" 66095d0682beSSumit Garg - "tee" 66105d0682beSSumit Garg - "caam" 66115d0682beSSumit Garg If not specified then it defaults to iterating through 66125d0682beSSumit Garg the trust source list starting with TPM and assigns the 6613fcd7c269SAhmad Fatoum first trust source as a backend which is initialized 6614fcd7c269SAhmad Fatoum successfully during iteration. 6615fcd7c269SAhmad Fatoum 6616fcd7c269SAhmad Fatoum trusted.rng= [KEYS] 6617fcd7c269SAhmad Fatoum Format: <string> 6618fcd7c269SAhmad Fatoum The RNG used to generate key material for trusted keys. 6619fcd7c269SAhmad Fatoum Can be one of: 6620fcd7c269SAhmad Fatoum - "kernel" 6621fcd7c269SAhmad Fatoum - the same value as trusted.source: "tpm" or "tee" 6622fcd7c269SAhmad Fatoum - "default" 6623e52347bdSJani Nikula If not specified, "default" is used. In this case, 6624e52347bdSJani Nikula the RNG's choice is left to each individual trust source. 6625e52347bdSJani Nikula 6626e52347bdSJani Nikula tsc= Disable clocksource stability checks for TSC. 6627e52347bdSJani Nikula Format: <string> 6628e52347bdSJani Nikula [x86] reliable: mark tsc clocksource as reliable, this 6629e52347bdSJani Nikula disables clocksource verification at runtime, as well 6630e52347bdSJani Nikula as the stability checks done at bootup. Used to enable 6631e52347bdSJani Nikula high-resolution timer mode on older hardware, and in 6632e52347bdSJani Nikula virtualized environment. 6633e52347bdSJani Nikula [x86] noirqtime: Do not use TSC to do irq accounting. 66346be53520SDou Liyang Used to run time disable IRQ_TIME_ACCOUNTING on any 66356be53520SDou Liyang platforms where RDTSC is slow and this accounting 66366be53520SDou Liyang can add overhead. 66370f0b7e1cSJuri Lelli [x86] unstable: mark the TSC clocksource as unstable, this 66380f0b7e1cSJuri Lelli marks the TSC unconditionally unstable at bootup and 66390f0b7e1cSJuri Lelli avoids any further wobbles once the TSC watchdog notices. 66400f0b7e1cSJuri Lelli [x86] nowatchdog: disable clocksource watchdog. Used 6641a7ec817dSFeng Tang in situations with strict latency requirements (where 6642a7ec817dSFeng Tang interruptions from clocksource watchdog are not 6643a7ec817dSFeng Tang acceptable). 6644a7ec817dSFeng Tang [x86] recalibrate: force recalibration against a HW timer 66450051293cSPaul E. McKenney (HPET or PM timer) on systems whose TSC frequency was 66460051293cSPaul E. McKenney obtained from HW or FW using either an MSR or CPUID(0x15). 66470051293cSPaul E. McKenney Warn if the difference is more than 500 ppm. 66480051293cSPaul E. McKenney [x86] watchdog: Use TSC as the watchdog clocksource with 66490051293cSPaul E. McKenney which to check other HW timers (HPET or PM timer), but 66500051293cSPaul E. McKenney only on systems where TSC has been deemed trustworthy. 6651e52347bdSJani Nikula This will be suppressed by an earlier tsc=nowatchdog and 6652bd35c77eSKrzysztof Piecuch can be overridden by a later tsc=nowatchdog. A console 6653bd35c77eSKrzysztof Piecuch message will flag any such suppression or overriding. 6654bd35c77eSKrzysztof Piecuch 6655bd35c77eSKrzysztof Piecuch tsc_early_khz= [X86] Skip early TSC calibration and use the given 6656bd35c77eSKrzysztof Piecuch value instead. Useful when the early TSC frequency discovery 6657bd35c77eSKrzysztof Piecuch procedure is not reliable, such as on overclocked systems 665895c5824fSPawan Gupta with CPUID.16h support and partial CPUID.15h support. 665995c5824fSPawan Gupta Format: <unsigned int> 666095c5824fSPawan Gupta 666195c5824fSPawan Gupta tsx= [X86] Control Transactional Synchronization 666295c5824fSPawan Gupta Extensions (TSX) feature in Intel processors that 666395c5824fSPawan Gupta support TSX control. 666495c5824fSPawan Gupta 666595c5824fSPawan Gupta This parameter controls the TSX feature. The options are: 666695c5824fSPawan Gupta 666795c5824fSPawan Gupta on - Enable TSX on the system. Although there are 666895c5824fSPawan Gupta mitigations for all known security vulnerabilities, 666995c5824fSPawan Gupta TSX has been known to be an accelerator for 667095c5824fSPawan Gupta several previous speculation-related CVEs, and 667195c5824fSPawan Gupta so there may be unknown security risks associated 667295c5824fSPawan Gupta with leaving it enabled. 667395c5824fSPawan Gupta 667495c5824fSPawan Gupta off - Disable TSX on the system. (Note that this 667595c5824fSPawan Gupta option takes effect only on newer CPUs which are 667695c5824fSPawan Gupta not vulnerable to MDS, i.e., have 667795c5824fSPawan Gupta MSR_IA32_ARCH_CAPABILITIES.MDS_NO=1 and which get 667895c5824fSPawan Gupta the new IA32_TSX_CTRL MSR through a microcode 66797531a359SPawan Gupta update. This new MSR allows for the reliable 66807531a359SPawan Gupta deactivation of the TSX functionality.) 66817531a359SPawan Gupta 668295c5824fSPawan Gupta auto - Disable TSX if X86_BUG_TAA is present, 668395c5824fSPawan Gupta otherwise enable TSX on the system. 668495c5824fSPawan Gupta 668595c5824fSPawan Gupta Not specifying this option is equivalent to tsx=off. 668695c5824fSPawan Gupta 6687a7a248c5SPawan Gupta See Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 6688a7a248c5SPawan Gupta for more details. 6689a7a248c5SPawan Gupta 6690a7a248c5SPawan Gupta tsx_async_abort= [X86,INTEL] Control mitigation for the TSX Async 6691a7a248c5SPawan Gupta Abort (TAA) vulnerability. 6692a7a248c5SPawan Gupta 6693a7a248c5SPawan Gupta Similar to Micro-architectural Data Sampling (MDS) 6694a7a248c5SPawan Gupta certain CPUs that support Transactional 6695a7a248c5SPawan Gupta Synchronization Extensions (TSX) are vulnerable to an 6696a7a248c5SPawan Gupta exploit against CPU internal buffers which can forward 6697a7a248c5SPawan Gupta information to a disclosure gadget under certain 6698a7a248c5SPawan Gupta conditions. 6699a7a248c5SPawan Gupta 6700a7a248c5SPawan Gupta In vulnerable processors, the speculatively forwarded 6701a7a248c5SPawan Gupta data can be used in a cache side channel attack, to 6702a7a248c5SPawan Gupta access data to which the attacker does not have direct 6703a7a248c5SPawan Gupta access. 6704a7a248c5SPawan Gupta 6705a7a248c5SPawan Gupta This parameter controls the TAA mitigation. The 6706a7a248c5SPawan Gupta options are: 6707a7a248c5SPawan Gupta 6708a7a248c5SPawan Gupta full - Enable TAA mitigation on vulnerable CPUs 6709a7a248c5SPawan Gupta if TSX is enabled. 6710a7a248c5SPawan Gupta 6711a7a248c5SPawan Gupta full,nosmt - Enable TAA mitigation and disable SMT on 6712a7a248c5SPawan Gupta vulnerable CPUs. If TSX is disabled, SMT 6713a7a248c5SPawan Gupta is not disabled because CPU is not 671464870ed1SWaiman Long vulnerable to cross-thread TAA attacks. 671564870ed1SWaiman Long off - Unconditionally disable TAA mitigation 671664870ed1SWaiman Long 671764870ed1SWaiman Long On MDS-affected machines, tsx_async_abort=off can be 671864870ed1SWaiman Long prevented by an active MDS mitigation as both vulnerabilities 6719a7a248c5SPawan Gupta are mitigated with the same mechanism so in order to disable 6720a7a248c5SPawan Gupta this mitigation, you need to specify mds=off too. 6721a7a248c5SPawan Gupta 6722a7a248c5SPawan Gupta Not specifying this option is equivalent to 6723a7a248c5SPawan Gupta tsx_async_abort=full. On CPUs which are MDS affected 6724a7a248c5SPawan Gupta and deploy MDS mitigation, TAA mitigation is not 6725a7a248c5SPawan Gupta required and doesn't provide any additional 6726a7a248c5SPawan Gupta mitigation. 6727a7a248c5SPawan Gupta 6728e52347bdSJani Nikula For details see: 6729e52347bdSJani Nikula Documentation/admin-guide/hw-vuln/tsx_async_abort.rst 6730e52347bdSJani Nikula 6731e52347bdSJani Nikula turbografx.map[2|3]= [HW,JOY] 67321752118dSTom Saeger TurboGraFX parallel port interface 6733e52347bdSJani Nikula Format: 6734e52347bdSJani Nikula <port#>,<js1>,<js2>,<js3>,<js4>,<js5>,<js6>,<js7> 6735e52347bdSJani Nikula See also Documentation/input/devices/joystick-parport.rst 6736e52347bdSJani Nikula 6737e52347bdSJani Nikula udbg-immortal [PPC] When debugging early kernel crashes that 6738e52347bdSJani Nikula happen after console_init() and before a proper 6739e52347bdSJani Nikula console driver takes over, this boot options might 6740e52347bdSJani Nikula help "seeing" what's going on. 6741e52347bdSJani Nikula 6742e52347bdSJani Nikula uhash_entries= [KNL,NET] 6743e52347bdSJani Nikula Set number of hash buckets for UDP/UDP-Lite connections 6744e52347bdSJani Nikula 6745e52347bdSJani Nikula uhci-hcd.ignore_oc= 6746e52347bdSJani Nikula [USB] Ignore overcurrent events (default N). 6747e52347bdSJani Nikula Some badly-designed motherboards generate lots of 6748e52347bdSJani Nikula bogus events, for ports that aren't wired to 6749e52347bdSJani Nikula anything. Set this parameter to avoid log spamming. 6750e52347bdSJani Nikula Note that genuine overcurrent events won't be 6751e52347bdSJani Nikula reported either. 6752e52347bdSJani Nikula 675389da5a69SJosh Poimboeuf unknown_nmi_panic 675489da5a69SJosh Poimboeuf [X86] Cause panic on unknown NMI. 675589da5a69SJosh Poimboeuf 675689da5a69SJosh Poimboeuf unwind_debug [X86-64] 675789da5a69SJosh Poimboeuf Enable unwinder debug output. This can be 675889da5a69SJosh Poimboeuf useful for debugging certain unwinder error 6759e52347bdSJani Nikula conditions, including corrupt stacks and 6760e52347bdSJani Nikula bad/missing unwinder metadata. 6761f176638aSAlan Stern 67627bae0432SDmitry Torokhov usbcore.authorized_default= 67637bae0432SDmitry Torokhov [USB] Default USB device authorization: 6764e52347bdSJani Nikula (default -1 = authorized (same as 1), 6765e52347bdSJani Nikula 0 = not authorized, 1 = authorized, 2 = authorized 6766e52347bdSJani Nikula if device connected to internal port) 6767e52347bdSJani Nikula 6768e52347bdSJani Nikula usbcore.autosuspend= 6769e52347bdSJani Nikula [USB] The autosuspend time delay (in seconds) used 6770e52347bdSJani Nikula for newly-detected USB devices (default 2). This 6771e52347bdSJani Nikula is the time required before an idle device will be 6772e52347bdSJani Nikula autosuspended. Devices for which the delay is set 6773e52347bdSJani Nikula to a negative value won't be autosuspended at all. 6774e52347bdSJani Nikula 6775e52347bdSJani Nikula usbcore.usbfs_snoop= 6776e52347bdSJani Nikula [USB] Set to log all usbfs traffic (default 0 = off). 6777e52347bdSJani Nikula 6778e52347bdSJani Nikula usbcore.usbfs_snoop_max= 6779e52347bdSJani Nikula [USB] Maximum number of bytes to snoop in each URB 6780e52347bdSJani Nikula (default = 65536). 6781e52347bdSJani Nikula 6782e52347bdSJani Nikula usbcore.blinkenlights= 6783e52347bdSJani Nikula [USB] Set to cycle leds on hubs (default 0 = off). 67843155f4f4SAlan Stern 6785e52347bdSJani Nikula usbcore.old_scheme_first= 6786e52347bdSJani Nikula [USB] Start with the old device initialization 6787e52347bdSJani Nikula scheme (default 0 = off). 6788e52347bdSJani Nikula 6789e52347bdSJani Nikula usbcore.usbfs_memory_mb= 6790e52347bdSJani Nikula [USB] Memory limit (in MB) for buffers allocated by 6791e52347bdSJani Nikula usbfs (default = 16, 0 = max = 2047). 6792e52347bdSJani Nikula 6793e52347bdSJani Nikula usbcore.use_both_schemes= 6794e52347bdSJani Nikula [USB] Try the other device initialization scheme 6795e52347bdSJani Nikula if the first one fails (default 1 = enabled). 6796e52347bdSJani Nikula 6797e52347bdSJani Nikula usbcore.initial_descriptor_timeout= 6798e52347bdSJani Nikula [USB] Specifies timeout for the initial 64-byte 6799e52347bdSJani Nikula USB_REQ_GET_DESCRIPTOR request in milliseconds 6800e52347bdSJani Nikula (default 5000 = 5.0 seconds). 6801027bd6caSKai-Heng Feng 6802027bd6caSKai-Heng Feng usbcore.nousb [USB] Disable the USB subsystem 6803027bd6caSKai-Heng Feng 6804027bd6caSKai-Heng Feng usbcore.quirks= 6805027bd6caSKai-Heng Feng [USB] A list of quirk entries to augment the built-in 6806027bd6caSKai-Heng Feng usb core quirk list. List entries are separated by 6807027bd6caSKai-Heng Feng commas. Each entry has the form 6808027bd6caSKai-Heng Feng VendorID:ProductID:Flags. The IDs are 4-digit hex 6809027bd6caSKai-Heng Feng numbers and Flags is a set of letters. Each letter 6810027bd6caSKai-Heng Feng will change the built-in quirk; setting it if it is 6811027bd6caSKai-Heng Feng clear and clearing it if it is set. The letters have 6812027bd6caSKai-Heng Feng the following meanings: 6813027bd6caSKai-Heng Feng a = USB_QUIRK_STRING_FETCH_255 (string 6814027bd6caSKai-Heng Feng descriptors must not be fetched using 6815027bd6caSKai-Heng Feng a 255-byte read); 6816027bd6caSKai-Heng Feng b = USB_QUIRK_RESET_RESUME (device can't resume 6817027bd6caSKai-Heng Feng correctly so reset it instead); 6818027bd6caSKai-Heng Feng c = USB_QUIRK_NO_SET_INTF (device can't handle 6819027bd6caSKai-Heng Feng Set-Interface requests); 6820027bd6caSKai-Heng Feng d = USB_QUIRK_CONFIG_INTF_STRINGS (device can't 6821027bd6caSKai-Heng Feng handle its Configuration or Interface 6822027bd6caSKai-Heng Feng strings); 6823027bd6caSKai-Heng Feng e = USB_QUIRK_RESET (device can't be reset 6824027bd6caSKai-Heng Feng (e.g morph devices), don't use reset); 6825027bd6caSKai-Heng Feng f = USB_QUIRK_HONOR_BNUMINTERFACES (device has 6826027bd6caSKai-Heng Feng more interface descriptions than the 6827027bd6caSKai-Heng Feng bNumInterfaces count, and can't handle 6828027bd6caSKai-Heng Feng talking to these interfaces); 6829027bd6caSKai-Heng Feng g = USB_QUIRK_DELAY_INIT (device needs a pause 6830027bd6caSKai-Heng Feng during initialization, after we read 6831027bd6caSKai-Heng Feng the device descriptor); 6832027bd6caSKai-Heng Feng h = USB_QUIRK_LINEAR_UFRAME_INTR_BINTERVAL (For 6833027bd6caSKai-Heng Feng high speed and super speed interrupt 6834027bd6caSKai-Heng Feng endpoints, the USB 2.0 and USB 3.0 spec 6835027bd6caSKai-Heng Feng require the interval in microframes (1 6836027bd6caSKai-Heng Feng microframe = 125 microseconds) to be 6837027bd6caSKai-Heng Feng calculated as interval = 2 ^ 6838027bd6caSKai-Heng Feng (bInterval-1). 6839027bd6caSKai-Heng Feng Devices with this quirk report their 6840027bd6caSKai-Heng Feng bInterval as the result of this 6841027bd6caSKai-Heng Feng calculation instead of the exponent 6842027bd6caSKai-Heng Feng variable used in the calculation); 6843027bd6caSKai-Heng Feng i = USB_QUIRK_DEVICE_QUALIFIER (device can't 6844027bd6caSKai-Heng Feng handle device_qualifier descriptor 6845027bd6caSKai-Heng Feng requests); 6846027bd6caSKai-Heng Feng j = USB_QUIRK_IGNORE_REMOTE_WAKEUP (device 6847027bd6caSKai-Heng Feng generates spurious wakeup, ignore 6848027bd6caSKai-Heng Feng remote wakeup capability); 6849027bd6caSKai-Heng Feng k = USB_QUIRK_NO_LPM (device can't handle Link 6850027bd6caSKai-Heng Feng Power Management); 6851027bd6caSKai-Heng Feng l = USB_QUIRK_LINEAR_FRAME_INTR_BINTERVAL 6852027bd6caSKai-Heng Feng (Device reports its bInterval as linear 6853027bd6caSKai-Heng Feng frames instead of the USB 2.0 68544d8d5a39SKai-Heng Feng calculation); 68554d8d5a39SKai-Heng Feng m = USB_QUIRK_DISCONNECT_SUSPEND (Device needs 68564d8d5a39SKai-Heng Feng to be disconnected before suspend to 6857781f0766SKai-Heng Feng prevent spurious wakeup); 6858781f0766SKai-Heng Feng n = USB_QUIRK_DELAY_CTRL_MSG (Device needs a 68593adcbec4SHardik Gajjar pause after every control message); 68603adcbec4SHardik Gajjar o = USB_QUIRK_HUB_SLOW_RESET (Hub needs extra 68613adcbec4SHardik Gajjar delay after resetting its port); 6862027bd6caSKai-Heng Feng p = USB_QUIRK_SHORT_SET_ADDRESS_REQ_TIMEOUT 6863027bd6caSKai-Heng Feng (Reduce timeout of the SET_ADDRESS 6864e52347bdSJani Nikula request from 5000 ms to 500 ms); 6865e52347bdSJani Nikula Example: quirks=0781:5580:bk,0a5c:5834:gij 6866e52347bdSJani Nikula 6867933bfe4dSTobias Jakobi usbhid.mousepoll= 6868933bfe4dSTobias Jakobi [USBHID] The interval which mice are to be polled at. 6869933bfe4dSTobias Jakobi 68702ddc8e2dSFilip Alac usbhid.jspoll= 68712ddc8e2dSFilip Alac [USBHID] The interval which joysticks are to be polled at. 68722ddc8e2dSFilip Alac 6873e52347bdSJani Nikula usbhid.kbpoll= 6874e52347bdSJani Nikula [USBHID] The interval which keyboards are to be polled at. 6875e52347bdSJani Nikula 6876e52347bdSJani Nikula usb-storage.delay_use= 6877e52347bdSJani Nikula [UMS] The delay in seconds before a new device is 6878e52347bdSJani Nikula scanned for Logical Units (default 1). 6879e52347bdSJani Nikula 6880e52347bdSJani Nikula usb-storage.quirks= 6881e52347bdSJani Nikula [UMS] A list of quirks entries to supplement or 6882e52347bdSJani Nikula override the built-in unusual_devs list. List 6883e52347bdSJani Nikula entries are separated by commas. Each entry has 6884e52347bdSJani Nikula the form VID:PID:Flags where VID and PID are Vendor 6885e52347bdSJani Nikula and Product ID values (4-digit hex numbers) and 688665cc8bf9SOliver Neukum Flags is a set of characters, each corresponding 6887e52347bdSJani Nikula to a common usb-storage quirk flag as follows: 688865cc8bf9SOliver Neukum a = SANE_SENSE (collect more than 18 bytes 6889e52347bdSJani Nikula of sense data, not on uas); 6890e52347bdSJani Nikula b = BAD_SENSE (don't collect more than 18 6891e52347bdSJani Nikula bytes of sense data, not on uas); 689265cc8bf9SOliver Neukum c = FIX_CAPACITY (decrease the reported 6893e52347bdSJani Nikula device capacity by one sector); 6894e52347bdSJani Nikula d = NO_READ_DISC_INFO (don't use 6895e52347bdSJani Nikula READ_DISC_INFO command, not on uas); 6896e52347bdSJani Nikula e = NO_READ_CAPACITY_16 (don't use 6897e52347bdSJani Nikula READ_CAPACITY_16 command); 6898e52347bdSJani Nikula f = NO_REPORT_OPCODES (don't use report opcodes 6899e52347bdSJani Nikula command, uas only); 6900e52347bdSJani Nikula g = MAX_SECTORS_240 (don't transfer more than 6901e52347bdSJani Nikula 240 sectors at a time, uas only); 6902e52347bdSJani Nikula h = CAPACITY_HEURISTICS (decrease the 6903e52347bdSJani Nikula reported device capacity by one 6904e52347bdSJani Nikula sector if the number is odd); 6905e52347bdSJani Nikula i = IGNORE_DEVICE (don't bind to this 69068010622cSOliver Neukum device); 6907e52347bdSJani Nikula j = NO_REPORT_LUNS (don't use report luns 690865cc8bf9SOliver Neukum command, uas only); 6909e52347bdSJani Nikula k = NO_SAME (do not use WRITE_SAME, uas only) 691065cc8bf9SOliver Neukum l = NOT_LOCKABLE (don't try to lock and 691165cc8bf9SOliver Neukum unlock ejectable media, not on uas); 6912e52347bdSJani Nikula m = MAX_SECTORS_64 (don't transfer more 691365cc8bf9SOliver Neukum than 64 sectors = 32 KB at a time, 6914e52347bdSJani Nikula not on uas); 691565cc8bf9SOliver Neukum n = INITIAL_READ10 (force a retry of the 6916e52347bdSJani Nikula initial READ(10) command, not on uas); 691765cc8bf9SOliver Neukum o = CAPACITY_OK (accept the capacity 6918e52347bdSJani Nikula reported by the device, not on uas); 691965cc8bf9SOliver Neukum p = WRITE_CACHE (the device cache is ON 6920e52347bdSJani Nikula by default, not on uas); 6921e52347bdSJani Nikula r = IGNORE_RESIDUE (the device reports 6922e52347bdSJani Nikula bogus residue values, not on uas); 6923e52347bdSJani Nikula s = SINGLE_LUN (the device has only one 6924e52347bdSJani Nikula Logical Unit); 6925e52347bdSJani Nikula t = NO_ATA_1X (don't allow ATA(12) and ATA(16) 6926e52347bdSJani Nikula commands, uas only); 6927e52347bdSJani Nikula u = IGNORE_UAS (don't bind to the uas driver); 692865cc8bf9SOliver Neukum w = NO_WP_DETECT (don't test whether the 692965cc8bf9SOliver Neukum medium is write-protected). 6930e52347bdSJani Nikula y = ALWAYS_SYNC (issue a SYNCHRONIZE_CACHE 6931e52347bdSJani Nikula even if the device claims no cache, 6932e52347bdSJani Nikula not on uas) 6933e52347bdSJani Nikula Example: quirks=0419:aaf5:rl,0421:0433:rc 6934e52347bdSJani Nikula 6935e52347bdSJani Nikula user_debug= [KNL,ARM] 6936e52347bdSJani Nikula Format: <int> 6937e52347bdSJani Nikula See arch/arm/Kconfig.debug help text. 6938e52347bdSJani Nikula 1 - undefined instruction events 6939e52347bdSJani Nikula 2 - system calls 6940e52347bdSJani Nikula 4 - invalid data aborts 6941e52347bdSJani Nikula 8 - SIGSEGV faults 6942e52347bdSJani Nikula 16 - SIGBUS faults 6943e52347bdSJani Nikula Example: user_debug=31 6944e52347bdSJani Nikula 6945e52347bdSJani Nikula userpte= 6946e52347bdSJani Nikula [X86] Flags controlling user PTE allocations. 6947e52347bdSJani Nikula 6948e52347bdSJani Nikula nohigh = do not allocate PTE pages in 694959bdbbd5SRandy Dunlap HIGHMEM regardless of setting 6950e52347bdSJani Nikula of CONFIG_HIGHPTE. 6951e52347bdSJani Nikula 6952e52347bdSJani Nikula vdso= [X86,SH,SPARC] 6953e52347bdSJani Nikula On X86_32, this is an alias for vdso32=. Otherwise: 6954e52347bdSJani Nikula 6955e52347bdSJani Nikula vdso=1: enable VDSO (the default) 6956e52347bdSJani Nikula vdso=0: disable VDSO mapping 6957e52347bdSJani Nikula 6958e52347bdSJani Nikula vdso32= [X86] Control the 32-bit vDSO 6959e52347bdSJani Nikula vdso32=1: enable 32-bit VDSO 6960e52347bdSJani Nikula vdso32=0 or vdso32=2: disable 32-bit VDSO 6961e52347bdSJani Nikula 6962e52347bdSJani Nikula See the help text for CONFIG_COMPAT_VDSO for more 6963e52347bdSJani Nikula details. If CONFIG_COMPAT_VDSO is set, the default is 6964e52347bdSJani Nikula vdso32=0; otherwise, the default is vdso32=1. 6965e52347bdSJani Nikula 6966e52347bdSJani Nikula For compatibility with older kernels, vdso32=2 is an 6967e52347bdSJani Nikula alias for vdso32=0. 6968e52347bdSJani Nikula 6969e52347bdSJani Nikula Try vdso32=0 if you encounter an error that says: 6970e52347bdSJani Nikula dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed! 6971e52347bdSJani Nikula 6972e52347bdSJani Nikula vector= [IA-64,SMP] 6973ab42b818SMauro Carvalho Chehab vector=percpu: enable percpu vector domain 6974e52347bdSJani Nikula 697559bdbbd5SRandy Dunlap video= [FB] Frame buffer configuration 697659bdbbd5SRandy Dunlap See Documentation/fb/modedb.rst. 6977e52347bdSJani Nikula 6978e52347bdSJani Nikula video.brightness_switch_enabled= [ACPI] 6979e52347bdSJani Nikula Format: [0|1] 698059bdbbd5SRandy Dunlap If set to 1, on receiving an ACPI notify event 6981e52347bdSJani Nikula generated by hotkey, video driver will adjust brightness 6982e52347bdSJani Nikula level and then send out the event to user space through 6983e52347bdSJani Nikula the allocated input device. If set to 0, video driver 6984e52347bdSJani Nikula will only send out the event without touching backlight 6985e52347bdSJani Nikula brightness level. 6986e52347bdSJani Nikula default: 1 6987e52347bdSJani Nikula 6988e52347bdSJani Nikula virtio_mmio.device= 6989e52347bdSJani Nikula [VMMIO] Memory mapped virtio (platform) device. 6990e52347bdSJani Nikula 6991e52347bdSJani Nikula <size>@<baseaddr>:<irq>[:<id>] 6992e52347bdSJani Nikula where: 6993e52347bdSJani Nikula <size> := size (can use standard suffixes 6994e52347bdSJani Nikula like K, M and G) 6995e52347bdSJani Nikula <baseaddr> := physical base address 6996e52347bdSJani Nikula <irq> := interrupt number (as passed to 6997e52347bdSJani Nikula request_irq()) 6998e52347bdSJani Nikula <id> := (optional) platform device id 6999e52347bdSJani Nikula example: 7000e52347bdSJani Nikula virtio_mmio.device=1K@0x100b0000:48:7 7001e52347bdSJani Nikula 7002ff61f079SJonathan Corbet Can be used multiple times for multiple devices. 70034f4cfa6cSMauro Carvalho Chehab 7004e52347bdSJani Nikula vga= [BOOT,X86-32] Select a particular video mode 7005e52347bdSJani Nikula See Documentation/arch/x86/boot.rst and 7006e52347bdSJani Nikula Documentation/admin-guide/svga.rst. 7007e52347bdSJani Nikula Use vga=ask for menu. 7008f682a97aSAlexander Duyck This is actually a boot loader parameter; the value is 7009f682a97aSAlexander Duyck passed to the kernel using a special protocol. 7010f682a97aSAlexander Duyck 7011f682a97aSAlexander Duyck vm_debug[=options] [KNL] Available with CONFIG_DEBUG_VM=y. 7012f682a97aSAlexander Duyck May slow down system boot speed, especially when 7013f682a97aSAlexander Duyck enabled on systems with a large amount of memory. 7014f682a97aSAlexander Duyck All options are enabled by default, and this 7015f682a97aSAlexander Duyck interface is meant to allow for selectively 7016f682a97aSAlexander Duyck enabling or disabling specific virtual memory 7017f682a97aSAlexander Duyck debugging features. 7018f682a97aSAlexander Duyck 7019f682a97aSAlexander Duyck Available options are: 7020e52347bdSJani Nikula P Enable page structure init time poisoning 7021e52347bdSJani Nikula - Disable all of the above options 7022e52347bdSJani Nikula 7023e52347bdSJani Nikula vmalloc=nn[KMG] [KNL,BOOT] Forces the vmalloc area to have an exact 7024e52347bdSJani Nikula size of <nn>. This can be used to increase the 7025e52347bdSJani Nikula minimum size (128MB on x86). It can also be used to 70263f429842SHeiko Carstens decrease the size and leave more room for directly 70273f429842SHeiko Carstens mapped kernel RAM. 70283f429842SHeiko Carstens 70293f429842SHeiko Carstens vmcp_cma=nn[MG] [KNL,S390] 7030e52347bdSJani Nikula Sets the memory size reserved for contiguous memory 7031e52347bdSJani Nikula allocations for the vmcp device driver. 7032e52347bdSJani Nikula 7033e52347bdSJani Nikula vmhalt= [KNL,S390] Perform z/VM CP command after system halt. 7034e52347bdSJani Nikula Format: <command> 7035e52347bdSJani Nikula 7036e52347bdSJani Nikula vmpanic= [KNL,S390] Perform z/VM CP command after kernel panic. 7037e52347bdSJani Nikula Format: <command> 7038e52347bdSJani Nikula 7039e52347bdSJani Nikula vmpoff= [KNL,S390] Perform z/VM CP command after power off. 7040e52347bdSJani Nikula Format: <command> 7041e52347bdSJani Nikula 7042e52347bdSJani Nikula vsyscall= [X86-64] 7043e52347bdSJani Nikula Controls the behavior of vsyscalls (i.e. calls to 7044e52347bdSJani Nikula fixed addresses of 0xffffffffff600x00 from legacy 7045e52347bdSJani Nikula code). Most statically-linked binaries and older 7046e52347bdSJani Nikula versions of glibc use these calls. Because these 704717f0669cSSohil Mehta functions are at fixed addresses, they make nice 704817f0669cSSohil Mehta targets for exploits that can control RIP. 704917f0669cSSohil Mehta 7050e52347bdSJani Nikula emulate Vsyscalls turn into traps and are emulated 705117f0669cSSohil Mehta reasonably safely. The vsyscall page is 7052bd49e16eSAndy Lutomirski readable. 7053bd49e16eSAndy Lutomirski 7054e52347bdSJani Nikula xonly [default] Vsyscalls turn into traps and are 7055e52347bdSJani Nikula emulated reasonably safely. The vsyscall 7056e52347bdSJani Nikula page is not readable. 7057e52347bdSJani Nikula 7058e52347bdSJani Nikula none Vsyscalls don't work at all. This makes 7059e52347bdSJani Nikula them quite hard to use for exploits but 7060e52347bdSJani Nikula might break your system. 7061e52347bdSJani Nikula 7062e52347bdSJani Nikula vt.color= [VT] Default text color. 7063e52347bdSJani Nikula Format: 0xYX, X = foreground, Y = background. 7064e52347bdSJani Nikula Default: 0x07 = light gray on black. 7065e52347bdSJani Nikula 7066e52347bdSJani Nikula vt.cur_default= [VT] Default cursor shape. 7067e52347bdSJani Nikula Format: 0xCCBBAA, where AA, BB, and CC are the same as 7068e52347bdSJani Nikula the parameters of the <Esc>[?A;B;Cc escape sequence; 7069e52347bdSJani Nikula see VGA-softcursor.txt. Default: 2 = underline. 7070e52347bdSJani Nikula 7071e52347bdSJani Nikula vt.default_blu= [VT] 7072e52347bdSJani Nikula Format: <blue0>,<blue1>,<blue2>,...,<blue15> 7073e52347bdSJani Nikula Change the default blue palette of the console. 7074e52347bdSJani Nikula This is a 16-member array composed of values 7075e52347bdSJani Nikula ranging from 0-255. 7076e52347bdSJani Nikula 7077e52347bdSJani Nikula vt.default_grn= [VT] 7078e52347bdSJani Nikula Format: <green0>,<green1>,<green2>,...,<green15> 7079e52347bdSJani Nikula Change the default green palette of the console. 7080e52347bdSJani Nikula This is a 16-member array composed of values 7081e52347bdSJani Nikula ranging from 0-255. 7082e52347bdSJani Nikula 7083e52347bdSJani Nikula vt.default_red= [VT] 7084e52347bdSJani Nikula Format: <red0>,<red1>,<red2>,...,<red15> 7085e52347bdSJani Nikula Change the default red palette of the console. 7086e52347bdSJani Nikula This is a 16-member array composed of values 7087e52347bdSJani Nikula ranging from 0-255. 7088e52347bdSJani Nikula 7089e52347bdSJani Nikula vt.default_utf8= 7090e52347bdSJani Nikula [VT] 7091e52347bdSJani Nikula Format=<0|1> 7092e52347bdSJani Nikula Set system-wide default UTF-8 mode for all tty's. 7093e52347bdSJani Nikula Default is 1, i.e. UTF-8 mode is enabled for all 7094e52347bdSJani Nikula newly opened terminals. 7095e52347bdSJani Nikula 7096e52347bdSJani Nikula vt.global_cursor_default= 7097e52347bdSJani Nikula [VT] 7098e52347bdSJani Nikula Format=<-1|0|1> 7099e52347bdSJani Nikula Set system-wide default for whether a cursor 7100e52347bdSJani Nikula is shown on new VTs. Default is -1, 7101e52347bdSJani Nikula i.e. cursors will be created by default unless 7102e52347bdSJani Nikula overridden by individual drivers. 0 will hide 7103e52347bdSJani Nikula cursors, 1 will display them. 7104e52347bdSJani Nikula 7105e52347bdSJani Nikula vt.italic= [VT] Default color for italic text; 0-15. 7106e52347bdSJani Nikula Default: 2 = green. 7107e52347bdSJani Nikula 7108e52347bdSJani Nikula vt.underline= [VT] Default color for underlined text; 0-15. 7109cc2a2d19SMauro Carvalho Chehab Default: 3 = cyan. 7110e52347bdSJani Nikula 7111e52347bdSJani Nikula watchdog timers [HW,WDT] For information on watchdog timers, 7112e52347bdSJani Nikula see Documentation/watchdog/watchdog-parameters.rst 711311295055SLaurence Oberman or other driver-specific files in the 711411295055SLaurence Oberman Documentation/watchdog/ directory. 711511295055SLaurence Oberman 711611295055SLaurence Oberman watchdog_thresh= 711711295055SLaurence Oberman [KNL] 711811295055SLaurence Oberman Set the hard lockup detector stall duration 711911295055SLaurence Oberman threshold in seconds. The soft lockup detector 712011295055SLaurence Oberman threshold is set to twice the value. A value of 0 7121ace3c549Stiozhang disables both lockup detectors. Default is 10 7122ace3c549Stiozhang seconds. 7123ace3c549Stiozhang 7124ace3c549Stiozhang workqueue.unbound_cpus= 7125ace3c549Stiozhang [KNL,SMP] Specify to constrain one or some CPUs 7126ace3c549Stiozhang to use in unbound workqueues. 7127ace3c549Stiozhang Format: <cpu-list> 7128e52347bdSJani Nikula By default, all online CPUs are available for 7129e52347bdSJani Nikula unbound workqueues. 7130e52347bdSJani Nikula 7131e52347bdSJani Nikula workqueue.watchdog_thresh= 7132e52347bdSJani Nikula If CONFIG_WQ_WATCHDOG is configured, workqueue can 7133e52347bdSJani Nikula warn stall conditions and dump internal state to 7134e52347bdSJani Nikula help debugging. 0 disables workqueue stall 7135e52347bdSJani Nikula detection; otherwise, it's the stall threshold 7136e52347bdSJani Nikula duration in seconds. The default value is 30 and 7137616db877STejun Heo it can be updated at runtime by writing to the 7138616db877STejun Heo corresponding sysfs file. 7139616db877STejun Heo 7140616db877STejun Heo workqueue.cpu_intensive_thresh_us= 7141616db877STejun Heo Per-cpu work items which run for longer than this 7142616db877STejun Heo threshold are automatically considered CPU intensive 7143616db877STejun Heo and excluded from concurrency management to prevent 714463638450STejun Heo them from noticeably delaying other per-cpu work 714563638450STejun Heo items. Default is 10000 (10ms). 714663638450STejun Heo 714763638450STejun Heo If CONFIG_WQ_CPU_INTENSIVE_REPORT is set, the kernel 714863638450STejun Heo will report the work functions which violate this 7149e52347bdSJani Nikula threshold repeatedly. They are likely good 7150e52347bdSJani Nikula candidates for using WQ_UNBOUND workqueues instead. 7151e52347bdSJani Nikula 7152e52347bdSJani Nikula workqueue.power_efficient 7153e52347bdSJani Nikula Per-cpu workqueues are generally preferred because 7154e52347bdSJani Nikula they show better performance thanks to cache 7155e52347bdSJani Nikula locality; unfortunately, per-cpu workqueues tend to 7156e52347bdSJani Nikula be more power hungry than unbound workqueues. 7157e52347bdSJani Nikula 7158e52347bdSJani Nikula Enabling this makes the per-cpu workqueues which 7159e52347bdSJani Nikula were observed to contribute significantly to power 7160e52347bdSJani Nikula consumption unbound, leading to measurably lower 7161e52347bdSJani Nikula power usage at the cost of small performance 7162e52347bdSJani Nikula overhead. 7163e52347bdSJani Nikula 716463c5484eSTejun Heo The default value of this parameter is determined by 716563c5484eSTejun Heo the config option CONFIG_WQ_POWER_EFFICIENT_DEFAULT. 716663c5484eSTejun Heo 716763c5484eSTejun Heo workqueue.default_affinity_scope= 716863c5484eSTejun Heo Select the default affinity scope to use for unbound 716963c5484eSTejun Heo workqueues. Can be one of "cpu", "smt", "cache", 717063c5484eSTejun Heo "numa" and "system". Default is "cache". For more 7171523a301eSTejun Heo information, see the Affinity Scopes section in 7172523a301eSTejun Heo Documentation/core-api/workqueue.rst. 7173523a301eSTejun Heo 7174523a301eSTejun Heo This can be changed after boot by writing to the 717563c5484eSTejun Heo matching /sys/module/workqueue/parameters file. All 7176e52347bdSJani Nikula workqueues with the "default" affinity scope will be 7177e52347bdSJani Nikula updated accordignly. 7178e52347bdSJani Nikula 7179e52347bdSJani Nikula workqueue.debug_force_rr_cpu 7180e52347bdSJani Nikula Workqueue used to implicitly guarantee that work 7181e52347bdSJani Nikula items queued without explicit CPU specified are put 7182e52347bdSJani Nikula on the local CPU. This guarantee is no longer true 7183e52347bdSJani Nikula and while local CPU is still preferred work items 7184e52347bdSJani Nikula may be put on foreign CPUs. This debug option 7185e52347bdSJani Nikula forces round-robin CPU selection to flush out 7186e52347bdSJani Nikula usages which depend on the now broken guarantee. 718716c52e50SHuacai Chen When enabled, memory and cache locality will be 718816c52e50SHuacai Chen impacted. 718916c52e50SHuacai Chen 719016c52e50SHuacai Chen writecombine= [LOONGARCH] Control the MAT (Memory Access Type) of 719116c52e50SHuacai Chen ioremap_wc(). 719216c52e50SHuacai Chen 7193e52347bdSJani Nikula on - Enable writecombine, use WUC for ioremap_wc() 7194e52347bdSJani Nikula off - Disable writecombine, use SUC for ioremap_wc() 7195e52347bdSJani Nikula 7196e52347bdSJani Nikula x2apic_phys [X86-64,APIC] Use x2apic physical mode instead of 7197e52347bdSJani Nikula default x2apic cluster mode on platforms 7198e52347bdSJani Nikula supporting x2apic. 7199e52347bdSJani Nikula 7200e52347bdSJani Nikula xen_512gb_limit [KNL,X86-64,XEN] 7201e52347bdSJani Nikula Restricts the kernel running paravirtualized under Xen 7202e52347bdSJani Nikula to use only up to 512 GB of RAM. The reason to do so is 7203e52347bdSJani Nikula crash analysis tools and Xen tools for doing domain 7204e52347bdSJani Nikula save/restore/migration must be enabled to handle larger 7205e52347bdSJani Nikula domains. 7206e52347bdSJani Nikula 7207e52347bdSJani Nikula xen_emul_unplug= [HW,X86,XEN] 7208e52347bdSJani Nikula Unplug Xen emulated devices 7209e52347bdSJani Nikula Format: [unplug0,][unplug1] 7210e52347bdSJani Nikula ide-disks -- unplug primary master IDE devices 7211e52347bdSJani Nikula aux-ide-disks -- unplug non-primary-master IDE devices 7212e52347bdSJani Nikula nics -- unplug network devices 7213e52347bdSJani Nikula all -- unplug all emulated devices (NICs and IDE disks) 7214e52347bdSJani Nikula unnecessary -- unplugging emulated devices is 7215e52347bdSJani Nikula unnecessary even if the host did not respond to 7216c6875f3aSBoris Ostrovsky the unplug protocol 7217c6875f3aSBoris Ostrovsky never -- do not unplug even if version check succeeds 7218c6875f3aSBoris Ostrovsky 7219c6875f3aSBoris Ostrovsky xen_legacy_crash [X86,XEN] 72203fac3734SJuergen Gross Crash from Xen panic notifier, without executing late 72213fac3734SJuergen Gross panic() code such as dumping handler. 72223fac3734SJuergen Gross 72233fac3734SJuergen Gross xen_msr_safe= [X86,XEN] 72243fac3734SJuergen Gross Format: <bool> 72253fac3734SJuergen Gross Select whether to always use non-faulting (safe) MSR 7226e52347bdSJani Nikula access functions when running as Xen PV guest. The 72279a3c05e6SZhenzhong Duan default value is controlled by CONFIG_XEN_PV_MSR_SAFE. 72289a3c05e6SZhenzhong Duan 72299a3c05e6SZhenzhong Duan xen_nopvspin [X86,XEN] 7230e52347bdSJani Nikula Disables the qspinlock slowpath using Xen PV optimizations. 7231e52347bdSJani Nikula This parameter is obsoleted by "nopvspin" parameter, which 7232e52347bdSJani Nikula has equivalent effect for XEN platform. 7233e52347bdSJani Nikula 7234b39b0497SZhenzhong Duan xen_nopv [X86] 7235b39b0497SZhenzhong Duan Disables the PV optimizations forcing the HVM guest to 7236e52347bdSJani Nikula run as generic HVM guest with no PV drivers. 7237b36b0fe9SDavid Woodhouse This option is obsoleted by the "nopv" option, which 7238b36b0fe9SDavid Woodhouse has equivalent effect for XEN platform. 7239b36b0fe9SDavid Woodhouse 7240b36b0fe9SDavid Woodhouse xen_no_vector_callback 7241197ecb38SMarek Marczykowski-Górecki [KNL,X86,XEN] Disable the vector callback for Xen 7242197ecb38SMarek Marczykowski-Górecki event channel interrupts. 7243197ecb38SMarek Marczykowski-Górecki 7244197ecb38SMarek Marczykowski-Górecki xen_scrub_pages= [XEN] 7245197ecb38SMarek Marczykowski-Górecki Boolean option to control scrubbing pages before giving them back 7246197ecb38SMarek Marczykowski-Górecki to Xen, for use by other domains. Can be also changed at runtime 72472ec16bc0SRyan Thibodeaux with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. 72482ec16bc0SRyan Thibodeaux Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. 72492ec16bc0SRyan Thibodeaux 72502ec16bc0SRyan Thibodeaux xen_timer_slop= [X86-64,XEN] 72512ec16bc0SRyan Thibodeaux Set the timer slop (in nanoseconds) for the virtual Xen 72522ec16bc0SRyan Thibodeaux timers (default is 100000). This adjusts the minimum 72532ec16bc0SRyan Thibodeaux delta of virtualized Xen timers, where lower values 725440fdea02SJuergen Gross improve timer resolution at the expense of processing 725540fdea02SJuergen Gross more timer interrupts. 725640fdea02SJuergen Gross 725740fdea02SJuergen Gross xen.balloon_boot_timeout= [XEN] 725840fdea02SJuergen Gross The time (in seconds) to wait before giving up to boot 725940fdea02SJuergen Gross in case initial ballooning fails to free enough memory. 726040fdea02SJuergen Gross Applies only when running as HVM or PVH guest and 7261e99502f7SJuergen Gross started with less memory configured than allowed at 7262e99502f7SJuergen Gross max. Default is 180. 7263e99502f7SJuergen Gross 7264e99502f7SJuergen Gross xen.event_eoi_delay= [XEN] 7265e99502f7SJuergen Gross How long to delay EOI handling in case of event 7266e99502f7SJuergen Gross storms (jiffies). Default is 10. 7267e99502f7SJuergen Gross 7268e99502f7SJuergen Gross xen.event_loop_timeout= [XEN] 72691a89c1dcSJuergen Gross After which time (jiffies) the event handling loop 72701a89c1dcSJuergen Gross should start to delay EOI handling. Default is 2. 72711a89c1dcSJuergen Gross 72721a89c1dcSJuergen Gross xen.fifo_events= [XEN] 72731a89c1dcSJuergen Gross Boolean parameter to disable using fifo event handling 72741a89c1dcSJuergen Gross even if available. Normally fifo event handling is 72751a89c1dcSJuergen Gross preferred over the 2-level event handling, as it is 7276e52347bdSJani Nikula fairer and the number of possible event channels is 7277e52347bdSJani Nikula much higher. Default is on (use fifo events). 7278e52347bdSJani Nikula 7279c0addc9aSLaurentiu Tudor xirc2ps_cs= [NET,PCMCIA] 7280ba45cff6SMichael Neuling Format: 7281ba45cff6SMichael Neuling <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] 7282ba45cff6SMichael Neuling 7283ba45cff6SMichael Neuling xive= [PPC] 7284ba45cff6SMichael Neuling By default on POWER9 and above, the kernel will 7285ba45cff6SMichael Neuling natively use the XIVE interrupt controller. This option 7286ba45cff6SMichael Neuling allows the fallback firmware mode to be used: 7287ba45cff6SMichael Neuling 7288ba45cff6SMichael Neuling off Fallback to firmware control of XIVE interrupt 7289c21ee04fSCédric Le Goater controller on both pseries and powernv 7290c21ee04fSCédric Le Goater platforms. Only useful on POWER9 and above. 7291c21ee04fSCédric Le Goater 7292c21ee04fSCédric Le Goater xive.store-eoi=off [PPC] 7293c21ee04fSCédric Le Goater By default on POWER10 and above, the kernel will use 7294c21ee04fSCédric Le Goater stores for EOI handling when the XIVE interrupt mode 7295c0addc9aSLaurentiu Tudor is active. This option allows the XIVE driver to use 7296c0addc9aSLaurentiu Tudor loads instead, as on POWER9. 7297c0addc9aSLaurentiu Tudor 7298c0addc9aSLaurentiu Tudor xhci-hcd.quirks [USB,KNL] 72996278f55bSGustavo Romero A hex value specifying bitmask with supplemental xhci 73006278f55bSGustavo Romero host controller quirks. Meaning of each bit can be 73016278f55bSGustavo Romero consulted in header drivers/usb/host/xhci.h. 73026278f55bSGustavo Romero 73036278f55bSGustavo Romero xmon [PPC] 73046278f55bSGustavo Romero Format: { early | on | rw | ro | off } 73056278f55bSGustavo Romero Controls if xmon debugger is enabled. Default is off. 73066278f55bSGustavo Romero Passing only "xmon" is equivalent to "xmon=early". 73076278f55bSGustavo Romero early Call xmon as early as possible on boot; xmon 73086278f55bSGustavo Romero debugger is called from setup_arch(). 73096278f55bSGustavo Romero on xmon debugger hooks will be installed so xmon 73106278f55bSGustavo Romero is only called on a kernel crash. Default mode, 73116278f55bSGustavo Romero i.e. either "ro" or "rw" mode, is controlled 73126278f55bSGustavo Romero with CONFIG_XMON_DEFAULT_RO_MODE. 73136278f55bSGustavo Romero rw xmon debugger hooks will be installed so xmon 73146278f55bSGustavo Romero is called only on a kernel crash, mode is write, 73156278f55bSGustavo Romero meaning SPR registers, memory and, other data 73166278f55bSGustavo Romero can be written using xmon commands. 73176278f55bSGustavo Romero ro same as "rw" option above but SPR registers, 73181056d314SPerry Yuan memory, and other data can't be written using 7319 xmon commands. 7320 off xmon is disabled. 7321 7322