1menu "Kernel hacking" 2 3config TRACE_IRQFLAGS_SUPPORT 4 def_bool y 5 6source "lib/Kconfig.debug" 7 8config EARLY_PRINTK_USB 9 bool 10 11config X86_VERBOSE_BOOTUP 12 bool "Enable verbose x86 bootup info messages" 13 default y 14 ---help--- 15 Enables the informational output from the decompression stage 16 (e.g. bzImage) of the boot. If you disable this you will still 17 see errors. Disable this if you want silent bootup. 18 19config EARLY_PRINTK 20 bool "Early printk" if EXPERT 21 default y 22 ---help--- 23 Write kernel log output directly into the VGA buffer or to a serial 24 port. 25 26 This is useful for kernel debugging when your machine crashes very 27 early before the console code is initialized. For normal operation 28 it is not recommended because it looks ugly and doesn't cooperate 29 with klogd/syslogd or the X server. You should normally say N here, 30 unless you want to debug such a crash. 31 32config EARLY_PRINTK_DBGP 33 bool "Early printk via EHCI debug port" 34 depends on EARLY_PRINTK && PCI 35 select EARLY_PRINTK_USB 36 ---help--- 37 Write kernel log output directly into the EHCI debug port. 38 39 This is useful for kernel debugging when your machine crashes very 40 early before the console code is initialized. For normal operation 41 it is not recommended because it looks ugly and doesn't cooperate 42 with klogd/syslogd or the X server. You should normally say N here, 43 unless you want to debug such a crash. You need usb debug device. 44 45config EARLY_PRINTK_EFI 46 bool "Early printk via the EFI framebuffer" 47 depends on EFI && EARLY_PRINTK 48 select FONT_SUPPORT 49 ---help--- 50 Write kernel log output directly into the EFI framebuffer. 51 52 This is useful for kernel debugging when your machine crashes very 53 early before the console code is initialized. 54 55config EARLY_PRINTK_USB_XDBC 56 bool "Early printk via the xHCI debug port" 57 depends on EARLY_PRINTK && PCI 58 select EARLY_PRINTK_USB 59 ---help--- 60 Write kernel log output directly into the xHCI debug port. 61 62 One use for this feature is kernel debugging, for example when your 63 machine crashes very early before the regular console code is 64 initialized. Other uses include simpler, lockless logging instead of 65 a full-blown printk console driver + klogd. 66 67 For normal production environments this is normally not recommended, 68 because it doesn't feed events into klogd/syslogd and doesn't try to 69 print anything on the screen. 70 71 You should normally say N here, unless you want to debug early 72 crashes or need a very simple printk logging facility. 73 74config X86_PTDUMP_CORE 75 def_bool n 76 77config X86_PTDUMP 78 tristate "Export kernel pagetable layout to userspace via debugfs" 79 depends on DEBUG_KERNEL 80 select DEBUG_FS 81 select X86_PTDUMP_CORE 82 ---help--- 83 Say Y here if you want to show the kernel pagetable layout in a 84 debugfs file. This information is only useful for kernel developers 85 who are working in architecture specific areas of the kernel. 86 It is probably not a good idea to enable this feature in a production 87 kernel. 88 If in doubt, say "N" 89 90config EFI_PGT_DUMP 91 bool "Dump the EFI pagetable" 92 depends on EFI 93 select X86_PTDUMP_CORE 94 ---help--- 95 Enable this if you want to dump the EFI page table before 96 enabling virtual mode. This can be used to debug miscellaneous 97 issues with the mapping of the EFI runtime regions into that 98 table. 99 100config DEBUG_WX 101 bool "Warn on W+X mappings at boot" 102 select X86_PTDUMP_CORE 103 ---help--- 104 Generate a warning if any W+X mappings are found at boot. 105 106 This is useful for discovering cases where the kernel is leaving 107 W+X mappings after applying NX, as such mappings are a security risk. 108 109 Look for a message in dmesg output like this: 110 111 x86/mm: Checked W+X mappings: passed, no W+X pages found. 112 113 or like this, if the check failed: 114 115 x86/mm: Checked W+X mappings: FAILED, <N> W+X pages found. 116 117 Note that even if the check fails, your kernel is possibly 118 still fine, as W+X mappings are not a security hole in 119 themselves, what they do is that they make the exploitation 120 of other unfixed kernel bugs easier. 121 122 There is no runtime or memory usage effect of this option 123 once the kernel has booted up - it's a one time check. 124 125 If in doubt, say "Y". 126 127config DOUBLEFAULT 128 default y 129 bool "Enable doublefault exception handler" if EXPERT 130 ---help--- 131 This option allows trapping of rare doublefault exceptions that 132 would otherwise cause a system to silently reboot. Disabling this 133 option saves about 4k and might cause you much additional grey 134 hair. 135 136config DEBUG_TLBFLUSH 137 bool "Set upper limit of TLB entries to flush one-by-one" 138 depends on DEBUG_KERNEL 139 ---help--- 140 141 X86-only for now. 142 143 This option allows the user to tune the amount of TLB entries the 144 kernel flushes one-by-one instead of doing a full TLB flush. In 145 certain situations, the former is cheaper. This is controlled by the 146 tlb_flushall_shift knob under /sys/kernel/debug/x86. If you set it 147 to -1, the code flushes the whole TLB unconditionally. Otherwise, 148 for positive values of it, the kernel will use single TLB entry 149 invalidating instructions according to the following formula: 150 151 flush_entries <= active_tlb_entries / 2^tlb_flushall_shift 152 153 If in doubt, say "N". 154 155config IOMMU_DEBUG 156 bool "Enable IOMMU debugging" 157 depends on GART_IOMMU && DEBUG_KERNEL 158 depends on X86_64 159 ---help--- 160 Force the IOMMU to on even when you have less than 4GB of 161 memory and add debugging code. On overflow always panic. And 162 allow to enable IOMMU leak tracing. Can be disabled at boot 163 time with iommu=noforce. This will also enable scatter gather 164 list merging. Currently not recommended for production 165 code. When you use it make sure you have a big enough 166 IOMMU/AGP aperture. Most of the options enabled by this can 167 be set more finegrained using the iommu= command line 168 options. See Documentation/x86/x86_64/boot-options.txt for more 169 details. 170 171config IOMMU_STRESS 172 bool "Enable IOMMU stress-test mode" 173 ---help--- 174 This option disables various optimizations in IOMMU related 175 code to do real stress testing of the IOMMU code. This option 176 will cause a performance drop and should only be enabled for 177 testing. 178 179config IOMMU_LEAK 180 bool "IOMMU leak tracing" 181 depends on IOMMU_DEBUG && DMA_API_DEBUG 182 ---help--- 183 Add a simple leak tracer to the IOMMU code. This is useful when you 184 are debugging a buggy device driver that leaks IOMMU mappings. 185 186config HAVE_MMIOTRACE_SUPPORT 187 def_bool y 188 189config X86_DECODER_SELFTEST 190 bool "x86 instruction decoder selftest" 191 depends on DEBUG_KERNEL && KPROBES 192 depends on !COMPILE_TEST 193 ---help--- 194 Perform x86 instruction decoder selftests at build time. 195 This option is useful for checking the sanity of x86 instruction 196 decoder code. 197 If unsure, say "N". 198 199# 200# IO delay types: 201# 202 203config IO_DELAY_TYPE_0X80 204 int 205 default "0" 206 207config IO_DELAY_TYPE_0XED 208 int 209 default "1" 210 211config IO_DELAY_TYPE_UDELAY 212 int 213 default "2" 214 215config IO_DELAY_TYPE_NONE 216 int 217 default "3" 218 219choice 220 prompt "IO delay type" 221 default IO_DELAY_0X80 222 223config IO_DELAY_0X80 224 bool "port 0x80 based port-IO delay [recommended]" 225 ---help--- 226 This is the traditional Linux IO delay used for in/out_p. 227 It is the most tested hence safest selection here. 228 229config IO_DELAY_0XED 230 bool "port 0xed based port-IO delay" 231 ---help--- 232 Use port 0xed as the IO delay. This frees up port 0x80 which is 233 often used as a hardware-debug port. 234 235config IO_DELAY_UDELAY 236 bool "udelay based port-IO delay" 237 ---help--- 238 Use udelay(2) as the IO delay method. This provides the delay 239 while not having any side-effect on the IO port space. 240 241config IO_DELAY_NONE 242 bool "no port-IO delay" 243 ---help--- 244 No port-IO delay. Will break on old boxes that require port-IO 245 delay for certain operations. Should work on most new machines. 246 247endchoice 248 249if IO_DELAY_0X80 250config DEFAULT_IO_DELAY_TYPE 251 int 252 default IO_DELAY_TYPE_0X80 253endif 254 255if IO_DELAY_0XED 256config DEFAULT_IO_DELAY_TYPE 257 int 258 default IO_DELAY_TYPE_0XED 259endif 260 261if IO_DELAY_UDELAY 262config DEFAULT_IO_DELAY_TYPE 263 int 264 default IO_DELAY_TYPE_UDELAY 265endif 266 267if IO_DELAY_NONE 268config DEFAULT_IO_DELAY_TYPE 269 int 270 default IO_DELAY_TYPE_NONE 271endif 272 273config DEBUG_BOOT_PARAMS 274 bool "Debug boot parameters" 275 depends on DEBUG_KERNEL 276 depends on DEBUG_FS 277 ---help--- 278 This option will cause struct boot_params to be exported via debugfs. 279 280config CPA_DEBUG 281 bool "CPA self-test code" 282 depends on DEBUG_KERNEL 283 ---help--- 284 Do change_page_attr() self-tests every 30 seconds. 285 286config OPTIMIZE_INLINING 287 bool "Allow gcc to uninline functions marked 'inline'" 288 ---help--- 289 This option determines if the kernel forces gcc to inline the functions 290 developers have marked 'inline'. Doing so takes away freedom from gcc to 291 do what it thinks is best, which is desirable for the gcc 3.x series of 292 compilers. The gcc 4.x series have a rewritten inlining algorithm and 293 enabling this option will generate a smaller kernel there. Hopefully 294 this algorithm is so good that allowing gcc 4.x and above to make the 295 decision will become the default in the future. Until then this option 296 is there to test gcc for this. 297 298 If unsure, say N. 299 300config DEBUG_ENTRY 301 bool "Debug low-level entry code" 302 depends on DEBUG_KERNEL 303 ---help--- 304 This option enables sanity checks in x86's low-level entry code. 305 Some of these sanity checks may slow down kernel entries and 306 exits or otherwise impact performance. 307 308 If unsure, say N. 309 310config DEBUG_NMI_SELFTEST 311 bool "NMI Selftest" 312 depends on DEBUG_KERNEL && X86_LOCAL_APIC 313 ---help--- 314 Enabling this option turns on a quick NMI selftest to verify 315 that the NMI behaves correctly. 316 317 This might help diagnose strange hangs that rely on NMI to 318 function properly. 319 320 If unsure, say N. 321 322config DEBUG_IMR_SELFTEST 323 bool "Isolated Memory Region self test" 324 default n 325 depends on INTEL_IMR 326 ---help--- 327 This option enables automated sanity testing of the IMR code. 328 Some simple tests are run to verify IMR bounds checking, alignment 329 and overlapping. This option is really only useful if you are 330 debugging an IMR memory map or are modifying the IMR code and want to 331 test your changes. 332 333 If unsure say N here. 334 335config X86_DEBUG_FPU 336 bool "Debug the x86 FPU code" 337 depends on DEBUG_KERNEL 338 default y 339 ---help--- 340 If this option is enabled then there will be extra sanity 341 checks and (boot time) debug printouts added to the kernel. 342 This debugging adds some small amount of runtime overhead 343 to the kernel. 344 345 If unsure, say N. 346 347config PUNIT_ATOM_DEBUG 348 tristate "ATOM Punit debug driver" 349 depends on PCI 350 select DEBUG_FS 351 select IOSF_MBI 352 ---help--- 353 This is a debug driver, which gets the power states 354 of all Punit North Complex devices. The power states of 355 each device is exposed as part of the debugfs interface. 356 The current power state can be read from 357 /sys/kernel/debug/punit_atom/dev_power_state 358 359choice 360 prompt "Choose kernel unwinder" 361 default FRAME_POINTER_UNWINDER 362 ---help--- 363 This determines which method will be used for unwinding kernel stack 364 traces for panics, oopses, bugs, warnings, perf, /proc/<pid>/stack, 365 livepatch, lockdep, and more. 366 367config FRAME_POINTER_UNWINDER 368 bool "Frame pointer unwinder" 369 select FRAME_POINTER 370 ---help--- 371 This option enables the frame pointer unwinder for unwinding kernel 372 stack traces. 373 374 The unwinder itself is fast and it uses less RAM than the ORC 375 unwinder, but the kernel text size will grow by ~3% and the kernel's 376 overall performance will degrade by roughly 5-10%. 377 378 This option is recommended if you want to use the livepatch 379 consistency model, as this is currently the only way to get a 380 reliable stack trace (CONFIG_HAVE_RELIABLE_STACKTRACE). 381 382config ORC_UNWINDER 383 bool "ORC unwinder" 384 depends on X86_64 385 select STACK_VALIDATION 386 ---help--- 387 This option enables the ORC (Oops Rewind Capability) unwinder for 388 unwinding kernel stack traces. It uses a custom data format which is 389 a simplified version of the DWARF Call Frame Information standard. 390 391 This unwinder is more accurate across interrupt entry frames than the 392 frame pointer unwinder. It also enables a 5-10% performance 393 improvement across the entire kernel compared to frame pointers. 394 395 Enabling this option will increase the kernel's runtime memory usage 396 by roughly 2-4MB, depending on your kernel config. 397 398config GUESS_UNWINDER 399 bool "Guess unwinder" 400 depends on EXPERT 401 ---help--- 402 This option enables the "guess" unwinder for unwinding kernel stack 403 traces. It scans the stack and reports every kernel text address it 404 finds. Some of the addresses it reports may be incorrect. 405 406 While this option often produces false positives, it can still be 407 useful in many cases. Unlike the other unwinders, it has no runtime 408 overhead. 409 410endchoice 411 412config FRAME_POINTER 413 depends on !ORC_UNWINDER && !GUESS_UNWINDER 414 bool 415 416endmenu 417