amdgpu_device_gpu_recover(9)
π amdgpu_device_gpu_recover β kernel API manual, bestie edition β¨
okay so you're looking at amdgpu_device_gpu_recover(9), which is the kernel API doc for the "my GPU is being dramatic, please fix it" function. she's giving main character energy in drivers/gpu/drm/amd/amdgpu and honestly? iconic.
π NAME π
amdgpu_device_gpu_recover β reset the asic and recover scheduler πΈ
π¦ SYNOPSIS π¦
int amdgpu_device_gpu_recover (struct amdgpu_device *adev , struct
amdgpu_job *job , struct amdgpu_reset_context *reset_context );
like, three args, one int coming back, we love a clean signature fr fr π
β¨ ARGUMENTS β¨
adevβ amdgpu_device pointer π this is your GPU's whole personality filejobβ which job trigger hang π the lil guy that got us into this messreset_contextβ amdgpu reset context pointer π«Ά all the context for the reset vibe
π DESCRIPTION π
Attempt to reset the GPU if it has hung (all asics). π« β¨ she'll attempt to do a soft-reset or a full-reset and then reinitialize the Asic. Returns 0 for success or an error on failure. periodt.
ngl, "if it has hung" is doing a lot of heavy lifting here β this whole function is basically the GPU's "have you tried turning it off and on again" era π
π§ SEE ALSO π§
Kernel file ./drivers/gpu/drm/amd/amdgpu/amdgpu_device.c β go read the source, iykyk π
and now the full friend group, no cap, every single one:
pcie_replay_count(9)board_info(9)uma/carveout_options(9)uma/carveout(9)amdgpu_device_supports_px(9)amdgpu_device_supports_boco(9)amdgpu_device_supports_baco(9)amdgpu_device_supports_smart_shift(9)amdgpu_device_mm_access(9)amdgpu_device_aper_access(9)amdgpu_device_vram_access(9)amdgpu_device_get_rev_id(9)amdgpu_device_asic_init(9)amdgpu_device_mem_scratch_init(9)amdgpu_device_mem_scratch_fini(9)amdgpu_device_program_register_sequence(9)amdgpu_device_pci_config_reset(9)amdgpu_device_pci_reset(9)amdgpu_device_resize_fb_bar(9)amdgpu_device_need_post(9)amdgpu_device_should_use_aspm(9)amdgpu_device_vga_set_decode(9)amdgpu_device_check_block_size(9)amdgpu_device_check_vm_size(9)amdgpu_device_check_arguments(9)amdgpu_switcheroo_set_state(9)amdgpu_switcheroo_can_switch(9)amdgpu_device_enable_virtual_display(9)amdgpu_device_parse_gpu_info_fw(9)amdgpu_device_ip_early_init(9)amdgpu_device_ip_init(9)amdgpu_device_fill_reset_magic(9)amdgpu_device_check_vram_lost(9)amdgpu_device_set_cg_state(9)amdgpu_device_ip_late_init(9)amdgpu_device_smu_fini_early(9)amdgpu_device_ip_fini(9)amdgpu_device_delayed_init_work_handler(9)amdgpu_device_ip_suspend_phase1(9)amdgpu_device_ip_suspend_phase2(9)amdgpu_device_ip_suspend(9)amdgpu_device_ip_resume_phase1(9)amdgpu_device_ip_resume_phase2(9)amdgpu_device_ip_resume_phase3(9)amdgpu_device_ip_resume(9)amdgpu_device_detect_sriov_bios(9)amdgpu_device_asic_has_dc_support(9)amdgpu_device_has_dc_support(9)amdgpu_device_check_iommu_direct_map(9)amdgpu_device_check_iommu_remap(9)amdgpu_device_init(9)amdgpu_device_fini_hw(9)amdgpu_device_evict_resources(9)amdgpu_device_pm_notifier(9)amdgpu_device_prepare(9)amdgpu_device_complete(9)amdgpu_device_suspend(9)amdgpu_device_resume(9)amdgpu_device_reset_sriov(9)amdgpu_device_has_job_running(9)amdgpu_device_should_recover_gpu(9)amdgpu_device_partner_bandwidth(9)amdgpu_device_gpu_bandwidth(9)amdgpu_device_get_pcie_info(9)amdgpu_device_is_peer_accessible(9)amdgpu_pci_error_detected(9)amdgpu_pci_mmio_enabled(9)amdgpu_pci_slot_reset(9)amdgpu_pci_resume(9)amdgpu_device_halt(9)amdgpu_device_get_gang(9)amdgpu_device_switch_gang(9)amdgpu_device_enforce_isolation(9)
π sign-off π
that's the whole page bestie π source + version vibes come from ./drivers/gpu/drm/amd/amdgpu, dated October 2026, and the header/footer both just wanna remind you this is amdgpu_device_gpu_recover(9) β we been knew π
β¨ stay slay, go reset that asic π¦π
π¦ where this page comes from
- nixpkgs unstable: linux-manual 7.3.1
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on : a language model's rewrite of the original, under its terms