* [PATCH] drm/amdkfd: Check if there are kfd porcesses using adev by kfd_processes_count
@ 2026-04-24 19:11 Xiaogang.Chen
2026-04-29 23:04 ` Kuehling, Felix
0 siblings, 1 reply; 2+ messages in thread
From: Xiaogang.Chen @ 2026-04-24 19:11 UTC (permalink / raw)
To: amd-gfx; +Cc: felix.kuehling, Xiaogang Chen
From: Xiaogang Chen <xiaogang.chen@amd.com>
Fixes: 6cca686dfce7("drm/amdkfd: kfd driver supports hot unplug/replug
amdgpu devices")
During gpu hot-unplug need check if there are kfd porcesses still using the
being removed gpu before clean resources of the device. Current driver checks
if kfd_processes_table is empty. kfd processes are not terminated after
removed from kfd_processes_table immediately. They are still alive and may
access the device until kfd_process_wq work queue got ran.
Check kfd->kfd_processes_count value that is updated after kfd process got
uninitialized when its ref becomes zero.
Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
---
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 33 +------------------------
1 file changed, 1 insertion(+), 32 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 9a66ee661e57..71fd3b1d0b3a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1737,37 +1737,6 @@ bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entr
return false;
}
-/* check if there is kfd process still uses adev */
-static bool kgd2kfd_check_device_idle(struct amdgpu_device *adev)
-{
- struct kfd_process *p;
- struct hlist_node *p_temp;
- unsigned int temp;
- struct kfd_node *dev;
-
- mutex_lock(&kfd_processes_mutex);
-
- if (hash_empty(kfd_processes_table)) {
- mutex_unlock(&kfd_processes_mutex);
- return true;
- }
-
- /* check if there is device still use adev */
- hash_for_each_safe(kfd_processes_table, temp, p_temp, p, kfd_processes) {
- for (int i = 0; i < p->n_pdds; i++) {
- dev = p->pdds[i]->dev;
- if (dev->adev == adev) {
- mutex_unlock(&kfd_processes_mutex);
- return false;
- }
- }
- }
-
- mutex_unlock(&kfd_processes_mutex);
-
- return true;
-}
-
/** kgd2kfd_teardown_processes - gracefully tear down existing
* kfd processes that use adev
*
@@ -1800,7 +1769,7 @@ void kgd2kfd_teardown_processes(struct amdgpu_device *adev)
mutex_unlock(&kfd_processes_mutex);
/* wait all kfd processes use adev terminate */
- while (!kgd2kfd_check_device_idle(adev))
+ while (!!atomic_read(&adev->kfd.dev->kfd_processes_count))
cond_resched();
}
--
2.34.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] drm/amdkfd: Check if there are kfd porcesses using adev by kfd_processes_count
2026-04-24 19:11 [PATCH] drm/amdkfd: Check if there are kfd porcesses using adev by kfd_processes_count Xiaogang.Chen
@ 2026-04-29 23:04 ` Kuehling, Felix
0 siblings, 0 replies; 2+ messages in thread
From: Kuehling, Felix @ 2026-04-29 23:04 UTC (permalink / raw)
To: Xiaogang.Chen, amd-gfx
On 2026-04-24 15:11, Xiaogang.Chen wrote:
> From: Xiaogang Chen <xiaogang.chen@amd.com>
>
> Fixes: 6cca686dfce7("drm/amdkfd: kfd driver supports hot unplug/replug
> amdgpu devices")
>
> During gpu hot-unplug need check if there are kfd porcesses still using the
> being removed gpu before clean resources of the device. Current driver checks
> if kfd_processes_table is empty. kfd processes are not terminated after
> removed from kfd_processes_table immediately. They are still alive and may
> access the device until kfd_process_wq work queue got ran.
>
> Check kfd->kfd_processes_count value that is updated after kfd process got
> uninitialized when its ref becomes zero.
>
> Signed-off-by: Xiaogang Chen <xiaogang.chen@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_device.c | 33 +------------------------
> 1 file changed, 1 insertion(+), 32 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> index 9a66ee661e57..71fd3b1d0b3a 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> @@ -1737,37 +1737,6 @@ bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entr
> return false;
> }
>
> -/* check if there is kfd process still uses adev */
> -static bool kgd2kfd_check_device_idle(struct amdgpu_device *adev)
> -{
> - struct kfd_process *p;
> - struct hlist_node *p_temp;
> - unsigned int temp;
> - struct kfd_node *dev;
> -
> - mutex_lock(&kfd_processes_mutex);
> -
> - if (hash_empty(kfd_processes_table)) {
> - mutex_unlock(&kfd_processes_mutex);
> - return true;
> - }
> -
> - /* check if there is device still use adev */
> - hash_for_each_safe(kfd_processes_table, temp, p_temp, p, kfd_processes) {
> - for (int i = 0; i < p->n_pdds; i++) {
> - dev = p->pdds[i]->dev;
> - if (dev->adev == adev) {
> - mutex_unlock(&kfd_processes_mutex);
> - return false;
> - }
> - }
> - }
> -
> - mutex_unlock(&kfd_processes_mutex);
> -
> - return true;
> -}
> -
> /** kgd2kfd_teardown_processes - gracefully tear down existing
> * kfd processes that use adev
> *
> @@ -1800,7 +1769,7 @@ void kgd2kfd_teardown_processes(struct amdgpu_device *adev)
> mutex_unlock(&kfd_processes_mutex);
>
> /* wait all kfd processes use adev terminate */
> - while (!kgd2kfd_check_device_idle(adev))
> + while (!!atomic_read(&adev->kfd.dev->kfd_processes_count))
> cond_resched();
> }
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-04-29 23:04 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-04-24 19:11 [PATCH] drm/amdkfd: Check if there are kfd porcesses using adev by kfd_processes_count Xiaogang.Chen
2026-04-29 23:04 ` Kuehling, Felix
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox