1What: /sys/kernel/debug/habanalabs/hl<n>/addr 2Date: Jan 2019 3KernelVersion: 5.1 4Contact: ogabbay@kernel.org 5Description: Sets the device address to be used for read or write through 6 PCI bar, or the device VA of a host mapped memory to be read or 7 written directly from the host. The latter option is allowed 8 only when the IOMMU is disabled. 9 The acceptable value is a string that starts with "0x" 10 11What: /sys/kernel/debug/habanalabs/hl<n>/clk_gate 12Date: May 2020 13KernelVersion: 5.8 14Contact: ogabbay@kernel.org 15Description: Allow the root user to disable/enable in runtime the clock 16 gating mechanism in Gaudi. Due to how Gaudi is built, the 17 clock gating needs to be disabled in order to access the 18 registers of the TPC and MME engines. This is sometimes needed 19 during debug by the user and hence the user needs this option. 20 The user can supply a bitmask value, each bit represents 21 a different engine to disable/enable its clock gating feature. 22 The bitmask is composed of 20 bits: 23 24 ======= ============ 25 0 - 7 DMA channels 26 8 - 11 MME engines 27 12 - 19 TPC engines 28 ======= ============ 29 30 The bit's location of a specific engine can be determined 31 using (1 << GAUDI_ENGINE_ID_*). GAUDI_ENGINE_ID_* values 32 are defined in uapi habanalabs.h file in enum gaudi_engine_id 33 34What: /sys/kernel/debug/habanalabs/hl<n>/command_buffers 35Date: Jan 2019 36KernelVersion: 5.1 37Contact: ogabbay@kernel.org 38Description: Displays a list with information about the currently allocated 39 command buffers 40 41What: /sys/kernel/debug/habanalabs/hl<n>/command_submission 42Date: Jan 2019 43KernelVersion: 5.1 44Contact: ogabbay@kernel.org 45Description: Displays a list with information about the currently active 46 command submissions 47 48What: /sys/kernel/debug/habanalabs/hl<n>/command_submission_jobs 49Date: Jan 2019 50KernelVersion: 5.1 51Contact: ogabbay@kernel.org 52Description: Displays a list with detailed information about each JOB (CB) of 53 each active command submission 54 55What: /sys/kernel/debug/habanalabs/hl<n>/data32 56Date: Jan 2019 57KernelVersion: 5.1 58Contact: ogabbay@kernel.org 59Description: Allows the root user to read or write directly through the 60 device's PCI bar. Writing to this file generates a write 61 transaction while reading from the file generates a read 62 transaction. This custom interface is needed (instead of using 63 the generic Linux user-space PCI mapping) because the DDR bar 64 is very small compared to the DDR memory and only the driver can 65 move the bar before and after the transaction. 66 67 If the IOMMU is disabled, it also allows the root user to read 68 or write from the host a device VA of a host mapped memory 69 70What: /sys/kernel/debug/habanalabs/hl<n>/data64 71Date: Jan 2020 72KernelVersion: 5.6 73Contact: ogabbay@kernel.org 74Description: Allows the root user to read or write 64 bit data directly 75 through the device's PCI bar. Writing to this file generates a 76 write transaction while reading from the file generates a read 77 transaction. This custom interface is needed (instead of using 78 the generic Linux user-space PCI mapping) because the DDR bar 79 is very small compared to the DDR memory and only the driver can 80 move the bar before and after the transaction. 81 82 If the IOMMU is disabled, it also allows the root user to read 83 or write from the host a device VA of a host mapped memory 84 85What: /sys/kernel/debug/habanalabs/hl<n>/data_dma 86Date: Apr 2021 87KernelVersion: 5.13 88Contact: ogabbay@kernel.org 89Description: Allows the root user to read from the device's internal 90 memory (DRAM/SRAM) through a DMA engine. 91 This property is a binary blob that contains the result of the 92 DMA transfer. 93 This custom interface is needed (instead of using the generic 94 Linux user-space PCI mapping) because the amount of internal 95 memory is huge (>32GB) and reading it via the PCI bar will take 96 a very long time. 97 This interface doesn't support concurrency in the same device. 98 In GAUDI and GOYA, this action can cause undefined behavior 99 in case the it is done while the device is executing user 100 workloads. 101 Only supported on GAUDI at this stage. 102 103What: /sys/kernel/debug/habanalabs/hl<n>/device 104Date: Jan 2019 105KernelVersion: 5.1 106Contact: ogabbay@kernel.org 107Description: Enables the root user to set the device to specific state. 108 Valid values are "disable", "enable", "suspend", "resume". 109 User can read this property to see the valid values 110 111What: /sys/kernel/debug/habanalabs/hl<n>/dma_size 112Date: Apr 2021 113KernelVersion: 5.13 114Contact: ogabbay@kernel.org 115Description: Specify the size of the DMA transaction when using DMA to read 116 from the device's internal memory. The value can not be larger 117 than 128MB. Writing to this value initiates the DMA transfer. 118 When the write is finished, the user can read the "data_dma" 119 blob 120 121What: /sys/kernel/debug/habanalabs/hl<n>/dump_security_violations 122Date: Jan 2021 123KernelVersion: 5.12 124Contact: ogabbay@kernel.org 125Description: Dumps all security violations to dmesg. This will also ack 126 all security violations meanings those violations will not be 127 dumped next time user calls this API 128 129What: /sys/kernel/debug/habanalabs/hl<n>/engines 130Date: Jul 2019 131KernelVersion: 5.3 132Contact: ogabbay@kernel.org 133Description: Displays the status registers values of the device engines and 134 their derived idle status 135 136What: /sys/kernel/debug/habanalabs/hl<n>/i2c_addr 137Date: Jan 2019 138KernelVersion: 5.1 139Contact: ogabbay@kernel.org 140Description: Sets I2C device address for I2C transaction that is generated 141 by the device's CPU 142 143What: /sys/kernel/debug/habanalabs/hl<n>/i2c_bus 144Date: Jan 2019 145KernelVersion: 5.1 146Contact: ogabbay@kernel.org 147Description: Sets I2C bus address for I2C transaction that is generated by 148 the device's CPU 149 150What: /sys/kernel/debug/habanalabs/hl<n>/i2c_data 151Date: Jan 2019 152KernelVersion: 5.1 153Contact: ogabbay@kernel.org 154Description: Triggers an I2C transaction that is generated by the device's 155 CPU. Writing to this file generates a write transaction while 156 reading from the file generates a read transaction 157 158What: /sys/kernel/debug/habanalabs/hl<n>/i2c_reg 159Date: Jan 2019 160KernelVersion: 5.1 161Contact: ogabbay@kernel.org 162Description: Sets I2C register id for I2C transaction that is generated by 163 the device's CPU 164 165What: /sys/kernel/debug/habanalabs/hl<n>/led0 166Date: Jan 2019 167KernelVersion: 5.1 168Contact: ogabbay@kernel.org 169Description: Sets the state of the first S/W led on the device 170 171What: /sys/kernel/debug/habanalabs/hl<n>/led1 172Date: Jan 2019 173KernelVersion: 5.1 174Contact: ogabbay@kernel.org 175Description: Sets the state of the second S/W led on the device 176 177What: /sys/kernel/debug/habanalabs/hl<n>/led2 178Date: Jan 2019 179KernelVersion: 5.1 180Contact: ogabbay@kernel.org 181Description: Sets the state of the third S/W led on the device 182 183What: /sys/kernel/debug/habanalabs/hl<n>/mmu 184Date: Jan 2019 185KernelVersion: 5.1 186Contact: ogabbay@kernel.org 187Description: Displays the hop values and physical address for a given ASID 188 and virtual address. The user should write the ASID and VA into 189 the file and then read the file to get the result. 190 e.g. to display info about VA 0x1000 for ASID 1 you need to do: 191 echo "1 0x1000" > /sys/kernel/debug/habanalabs/hl0/mmu 192 193What: /sys/kernel/debug/habanalabs/hl<n>/mmu_error 194Date: Mar 2021 195KernelVersion: 5.12 196Contact: fkassabri@habana.ai 197Description: Check and display page fault or access violation mmu errors for 198 all MMUs specified in mmu_cap_mask. 199 e.g. to display error info for MMU hw cap bit 9, you need to do: 200 echo "0x200" > /sys/kernel/debug/habanalabs/hl0/mmu_error 201 cat /sys/kernel/debug/habanalabs/hl0/mmu_error 202 203What: /sys/kernel/debug/habanalabs/hl<n>/set_power_state 204Date: Jan 2019 205KernelVersion: 5.1 206Contact: ogabbay@kernel.org 207Description: Sets the PCI power state. Valid values are "1" for D0 and "2" 208 for D3Hot 209 210What: /sys/kernel/debug/habanalabs/hl<n>/skip_reset_on_timeout 211Date: Jun 2021 212KernelVersion: 5.13 213Contact: ynudelman@habana.ai 214Description: Sets the skip reset on timeout option for the device. Value of 215 "0" means device will be reset in case some CS has timed out, 216 otherwise it will not be reset. 217 218What: /sys/kernel/debug/habanalabs/hl<n>/state_dump 219Date: Oct 2021 220KernelVersion: 5.15 221Contact: ynudelman@habana.ai 222Description: Gets the state dump occurring on a CS timeout or failure. 223 State dump is used for debug and is created each time in case of 224 a problem in a CS execution, before reset. 225 Reading from the node returns the newest state dump available. 226 Writing an integer X discards X state dumps, so that the 227 next read would return X+1-st newest state dump. 228 229What: /sys/kernel/debug/habanalabs/hl<n>/stop_on_err 230Date: Mar 2020 231KernelVersion: 5.6 232Contact: ogabbay@kernel.org 233Description: Sets the stop-on_error option for the device engines. Value of 234 "0" is for disable, otherwise enable. 235 236What: /sys/kernel/debug/habanalabs/hl<n>/userptr 237Date: Jan 2019 238KernelVersion: 5.1 239Contact: ogabbay@kernel.org 240Description: Displays a list with information about the currently user 241 pointers (user virtual addresses) that are pinned and mapped 242 to DMA addresses 243 244What: /sys/kernel/debug/habanalabs/hl<n>/userptr_lookup 245Date: Aug 2021 246KernelVersion: 5.15 247Contact: ogabbay@kernel.org 248Description: Allows to search for specific user pointers (user virtual 249 addresses) that are pinned and mapped to DMA addresses, and see 250 their resolution to the specific dma address. 251 252What: /sys/kernel/debug/habanalabs/hl<n>/vm 253Date: Jan 2019 254KernelVersion: 5.1 255Contact: ogabbay@kernel.org 256Description: Displays a list with information about all the active virtual 257 address mappings per ASID and all user mappings of HW blocks 258