AMDGPU DebugFS¶
The amdgpu driver provides a number of debugfs files to aid in debuggingissues in the driver. These are usually found in/sys/kernel/debug/dri/<num>.
DebugFS Files¶
amdgpu_benchmark¶
Run benchmarks using the DMA engine the driver uses for GPU memory paging.Write a number to the file to run the test. The results are written to thekernel log. VRAM is on device memory (dGPUs) or carve out (APUs) and GTT(Graphics Translation Tables) is system memory that is accessible by the GPU.The following tests are available:
1: simple test, VRAM to GTT and GTT to VRAM
2: simple test, VRAM to VRAM
3: GTT to VRAM, buffer size sweep, powers of 2
4: VRAM to GTT, buffer size sweep, powers of 2
5: VRAM to VRAM, buffer size sweep, powers of 2
6: GTT to VRAM, buffer size sweep, common display sizes
7: VRAM to GTT, buffer size sweep, common display sizes
8: VRAM to VRAM, buffer size sweep, common display sizes
amdgpu_test_ib¶
Read this file to run simple IB (Indirect Buffer) tests on all kernel managedrings. IBs are command buffers usually generated by userspace applicationswhich are submitted to the kernel for execution on an particular GPU engine.This just runs the simple IB tests included in the kernel. These testsare engine specific and verify that IB submission works.
amdgpu_discovery¶
Provides raw access to the IP discovery binary provided by the GPU. Read thisfile to access the raw binary. This is useful for verifying the contents ofthe IP discovery table. It is chip specific.
amdgpu_vbios¶
Provides raw access to the ROM binary image from the GPU. Read this file toaccess the raw binary. This is useful for verifying the contents of thevideo BIOS ROM. It is board specific.
amdgpu_evict_gtt¶
Evict all buffers from the GTT memory pool. Read this file to evict allbuffers from this pool.
amdgpu_evict_vram¶
Evict all buffers from the VRAM memory pool. Read this file to evict allbuffers from this pool.
amdgpu_gpu_recover¶
Trigger a GPU reset. Read this file to trigger reset the entire GPU.All work currently running on the GPU will be lost.
amdgpu_ring_<name>¶
Provides read access to the kernel managed ring buffers for each ring <name>.These are useful for debugging problems on a particular ring. The ring bufferis how the CPU sends commands to the GPU. The CPU writes commands into thebuffer and then asks the GPU engine to process it. This is the raw binarycontents of the ring buffer. Use a tool like UMR to decode the rings into humanreadable form.
amdgpu_mqd_<name>¶
Provides read access to the kernel managed MQD (Memory Queue Descriptor) forring <name> managed by the kernel driver. MQDs define the features of the ringand are used to store the ring’s state when it is not connected to hardware.The driver writes the requested ring features and metadata (GPU addresses ofthe ring itself and associated buffers) to the MQD and the firmware uses the MQDto populate the hardware when the ring is mapped to a hardware slot. Onlyavailable on engines which use MQDs. This provides access to the raw MQDbinary.
amdgpu_error_<name>¶
Provides an interface to set an error code on the dma fences associated withring <name>. The error code specified is propagated to all fences associatedwith the ring. Use this to inject a fence error into a ring.
amdgpu_pm_info¶
Provides human readable information about the power management featuresand state of the GPU. This includes current GFX clock, Memory clock,voltages, average SoC power, temperature, GFX load, Memory load, SMUfeature mask, VCN power state, clock and power gating features.
amdgpu_firmware_info¶
Lists the firmware versions for all firmwares used by the GPU. Onlyentries with a non-0 version are valid. If the version is 0, the firmwareis not valid for the GPU.
amdgpu_fence_info¶
Shows the last signalled and emitted fence sequence numbers for eachkernel driver managed ring. Fences are associated with submissionsto the engine. Emitted fences have been submitted to the ringand signalled fences have been signalled by the GPU. Rings with alarger emitted fence value have outstanding work that is still beingprocessed by the engine that owns that ring. When the emitted andsignalled fence values are equal, the ring is idle.
amdgpu_gem_info¶
Lists all of the PIDs using the GPU and the GPU buffers that they haveallocated. This lists the buffer size, pool (VRAM, GTT, etc.), and bufferattributes (CPU access required, CPU cache attributes, etc.).
amdgpu_vm_info¶
Lists all of the PIDs using the GPU and the GPU buffers that they haveallocated as well as the status of those buffers relative to that process’GPU virtual address space (e.g., evicted, idle, invalidated, etc.).
amdgpu_sa_info¶
Prints out all of the suballocations (sa) by the suballocation manager in thekernel driver. Prints the GPU address, size, and fence info associatedwith each suballocation. The suballocations are used internally withinthe kernel driver for various things.
amdgpu_<pool>_mm¶
Prints TTM information about the memory pool <pool>.
amdgpu_vram¶
Provides direct access to VRAM. Used by tools like UMR to inspectobjects in VRAM.
amdgpu_iomem¶
Provides direct access to GTT memory. Used by tools like UMR to inspectGTT memory.
amdgpu_regs_*¶
Provides direct access to various register apertures on the GPU. Usedby tools like UMR to access GPU registers.
amdgpu_regs2¶
Provides an IOCTL interface used by UMR for interacting with GPU registers.
amdgpu_sensors¶
Provides an interface to query GPU power metrics (temperature, averagepower, etc.). Used by tools like UMR to query GPU power metrics.
amdgpu_gca_config¶
Provides an interface to query GPU details (Graphics/Compute Array config,PCI config, GPU family, etc.). Used by tools like UMR to query GPU details.
amdgpu_wave¶
Used to query GFX/compute wave information from the hardware. Used by toolslike UMR to query GFX/compute wave information.
amdgpu_gpr¶
Used to query GFX/compute GPR (General Purpose Register) information from thehardware. Used by tools like UMR to query GPRs when debugging shaders.
amdgpu_gprwave¶
Provides an IOCTL interface used by UMR for interacting with shader waves.
amdgpu_fw_attestation¶
Provides an interface for reading back firmware attestation records.