drm/amdgpu: Fix the gpu recover deadlock issue in abnormal situations
Subsystems (2)
PATCH 1/1 drm/amdgpu: Fix the gpu recover deadlock issue in abnormal situations
rma occurred in the RAS records exceed threshold device. When the device was performing gpu recover, the reset domain lock was not released, resulting in kernel panic
[ 630.141619] INFO: task umc_page_retire:9472 blocked for more than 122 seconds. [ 630.157663] Tainted: G OE 6.9.0-0_fbk6_brcmrdma11_125_gfecec9d12677 #1 [ 630.176213] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 630.193591] task:umc_page_retire state:D stack:0 pid:9472 tgid:9472 ppid:2 flags:0x00004000 [ 630.214251] Call Trace: [ 630.219673] <TASK> [ 630.224326] __schedule+0x589/0xad0 [ 630.232070] schedule+0x67/0xb0 [ 630.239038] schedule_timeout+0x22/0x100 [ 630.247740] ? __smp_call_single_queue+0x36/0x100 [ 630.258195] ? wake_up_process+0x7df/0x1190 [ 630.267477] ? wait_for_common+0xa4/0x150 [ 630.276371] wait_for_common+0x99/0x150 [ 630.284881] __flush_work.llvm.6727378282878825097+0x20a/0x290 [ 630.297826] ? rcu_work_rcufn+0x20/0x20 [ 630.306338] amdgpu_ras_page_retirement_thread+0x469/0x4e0 [amdgpu] [ 630.320344] ? amdgpu_ras_do_recovery+0x5f0/0x5f0 [amdgpu] [ 630.332587] kthread+0xdd/0x120 [ 630.339552] ? __sched_group_set_shares+0x160/0x160 [ 630.350373] ret_from_fork+0x2f/0x40 [ 630.358303] ? __sched_group_set_shares+0x160/0x160 [ 630.369128] ret_from_fork_asm+0x11/0x20 [ 630.377831] </TASK> [ 630.382796] INFO: task kworker/u1536:1:10591 blocked for more than 123 seconds. [ 630.399018] Tainted: G OE 6.9.0-0_fbk6_brcmrdma11_125_gfecec9d12677 #1 [ 630.417554] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 630.434935] task:kworker/u1536:1 state:D stack:0 pid:10591 tgid:10591 ppid:2 flags:0x00004000 [ 630.455594] Workqueue: amdgpu-reset-hive amdgpu_ras_do_recovery [amdgpu] [ 630.470537] Call Trace: [ 630.475958] <TASK> [ 630.480609] __schedule+0x589/0xad0 [ 630.488347] schedule+0x67/0xb0 [ 630.495313] schedule_preempt_disabled+0xa/0x10 [ 630.505362] rwsem_down_write_slowpath+0x2ba/0x510 [ 630.515994] down_write+0x2b/0x30 [ 630.523346] amdgpu_device_halt_activities+0xef/0x400 [amdgpu] [ 630.536363] amdgpu_device_gpu_recover+0x124/0x230 [amdgpu] [ 630.548797] amdgpu_ras_do_recovery+0x5af/0x5f0 [amdgpu] [ 630.560653] process_scheduled_works+0x184/0x370 [ 630.570900] worker_thread+0xc6/0x3f0 [ 630.579022] ? __ipv6_chk_addr_and_flags.llvm.7715710786076949193+0x160/0x160 [ 630.594857] ? __ipv6_chk_addr_and_flags.llvm.7715710786076949193+0x160/0x160 [ 630.610693] kthread+0xdd/0x120 [ 630.617660] ? __sched_group_set_shares+0x160/0x160 [ 630.628483] ret_from_fork+0x2f/0x40 [ 630.636413] ? __sched_group_set_shares+0x160/0x160 [ 630.647232] ret_from_fork_asm+0x11/0x20
Signed-off-by Ce Sun <[email protected]>
--- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 ----- 1 file changed, 5 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -5
@@ -6305,7 +6305,6 @@ static void amdgpu_device_gpu_resume(struct amdgpu_device *adev, tmp_adev = list_first_entry(device_list, struct amdgpu_device, reset_list);amdgpu_device_unlock_reset_domain (tmp_adev->reset_domain); - }
@@ -6396,12 +6395,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, } r =amdgpu_device_asic_reset (adev, &device_list, reset_context); -if (r) - goto end_reset; skip_hw_reset: r =amdgpu_device_sched_resume (&device_list, reset_context, job_signaled); -if (r) - goto end_reset; skip_sched_resume:amdgpu_device_gpu_resume (adev, &device_list, need_emergency_restart); end_reset:
Signature · 1 line
2.34.1