Mukul posted a patch for this already.
"drm/amdgpu: Fix module unload hang with RAS enabled"
Thanks,
Lijo
On 1/24/2024 9:09 AM, YiPeng Chai wrote:
> The following is the error message:
> [ 484.495995] task:rmmod state:D stack: 0 pid: 2195 ppid: 2194 flags:0x00004002
> [ 484.496000] Call Trace:
> [ 484.496002] <TASK>
> [ 484.496007] __schedule+0xaf8/0x1870
> [ 484.496015] ? update_load_avg+0x74/0x7b0
> [ 484.496021] schedule+0x58/0xc0
> [ 484.496022] schedule_timeout+0x276/0x480
> [ 484.496024] ? ttwu_do_activate+0x9f/0x570
> [ 484.496028] wait_for_completion+0x8b/0x130
> [ 484.496030] kthread_stop+0x71/0x1a0
> [ 484.496037] amdgpu_ras_pre_fini+0x5b/0xe0 [amdgpu]
> [ 484.496202] amdgpu_device_fini_hw+0x165/0x4fc [amdgpu]
> [ 484.496406] ? blocking_notifier_chain_unregister+0x56/0xb0
> [ 484.496409] amdgpu_driver_unload_kms+0x51/0x60 [amdgpu]
> [ 484.496522] amdgpu_pci_remove+0x3b/0x70 [amdgpu]
> [ 484.496627] pci_device_remove+0x39/0xa0
> [ 484.496631] device_remove+0x46/0x70
> [ 484.496634] device_release_driver_internal+0xcd/0x140
> [ 484.496636] driver_detach+0x4a/0x90
> [ 484.496638] bus_remove_driver+0x6c/0xf0
> [ 484.496641] driver_unregister+0x31/0x70
> [ 484.496643] pci_unregister_driver+0x40/0x90
> [ 484.496647] amdgpu_exit+0x15/0x22b [amdgpu]
> [ 484.496849] __x64_sys_delete_module+0x14a/0x260
> [ 484.496853] ? syscall_exit_to_user_mode+0x26/0x40
> [ 484.496856] ? __x64_sys_close+0x12/0x40
> [ 484.496860] do_syscall_64+0x5c/0x80
> [ 484.496861] ? __x64_sys_read+0x1a/0x20
> [ 484.496863] ? do_syscall_64+0x69/0x80
> [ 484.496864] ? syscall_exit_to_user_mode+0x26/0x40
> [ 484.496866] ? do_syscall_64+0x69/0x80
> [ 484.496866] ? exc_page_fault+0x87/0x170
> [ 484.496868] ? asm_exc_page_fault+0x8/0x30
> [ 484.496871] entry_SYSCALL_64_after_hwframe+0x44/0xae
>
> Signed-off-by: YiPeng Chai <YiPeng.Chai@xxxxxxx>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index 8c1cb3ec2762..768a98f4bd22 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -2674,7 +2674,11 @@ static int amdgpu_ras_page_retirement_thread(void *param)
> while (!kthread_should_stop()) {
>
> wait_event_interruptible(con->page_retirement_wq,
> - atomic_read(&con->page_retirement_req_cnt));
> + atomic_read(&con->page_retirement_req_cnt) ||
> + kthread_should_stop());
> +
> + if (kthread_should_stop())
> + break;
>
> dev_info(adev->dev, "Start processing page retirement. request:%d\n",
> atomic_read(&con->page_retirement_req_cnt));
