Commit cf1bee5d authored by Zhigang Luo's avatar Zhigang Luo Committed by Zheng Zucheng
Browse files

amd/amdkfd: sync all devices to wait all processes being evicted 

mainline inclusion
from mainline-v6.10-rc1
commit dfb15c4ab58658aaa6161b546e7eb852ae7cc132
category: bugfix
bugzilla: https://gitee.com/src-openeuler/kernel/issues/I9U8NU
CVE: CVE-2024-36949

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=dfb15c4ab58658aaa6161b546e7eb852ae7cc132



--------------------------------

If there are more than one device doing reset in parallel, the first
device will call kfd_suspend_all_processes() to evict all processes
on all devices, this call takes time to finish. other device will
start reset and recover without waiting. if the process has not been
evicted before doing recover, it will be restored, then caused page
fault.

Signed-off-by: default avatarZhigang Luo <Zhigang.Luo@amd.com>
Reviewed-by: default avatarFelix Kuehling <felix.kuehling@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>

Conflicts:
     drivers/gpu/drm/amd/amdkfd/kfd_device.c
[Some contexts different. No functional impact.]
Signed-off-by: default avatarZheng Zucheng <zhengzucheng@huawei.com>
parent 14e98c86

drivers/gpu/drm/amd/amdkfd/kfd_device.c

+6 −10
Original line number Diff line number Diff line
@@ -860,18 +860,16 @@ bool kfd_is_locked(void) 


void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)

{

	int count;

	if (!kfd->init_complete)

		return;



	/* for runtime suspend, skip locking kfd */

	if (!run_pm) {

		mutex_lock(&kfd_processes_mutex);

		count = ++kfd_locked;

		mutex_unlock(&kfd_processes_mutex);

		/* For first KFD device suspend all the KFD processes */

		if (count == 1)

		if (++kfd_locked == 1)

			kfd_suspend_all_processes();

		mutex_unlock(&kfd_processes_mutex);

	}



	kfd->dqm->ops.stop(kfd->dqm);
@@ -880,7 +878,7 @@ void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm) 


int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)

{

	int ret, count;

	int ret;



	if (!kfd->init_complete)

		return 0;
@@ -892,12 +890,10 @@ int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm) 
	/* for runtime resume, skip unlocking kfd */

	if (!run_pm) {

		mutex_lock(&kfd_processes_mutex);

		count = --kfd_locked;

		mutex_unlock(&kfd_processes_mutex);



		WARN_ONCE(count < 0, "KFD suspend / resume ref. error");

		if (count == 0)

		if (--kfd_locked == 0)

			ret = kfd_resume_all_processes();

		WARN_ONCE(kfd_locked < 0, "KFD suspend / resume ref. error");

		mutex_unlock(&kfd_processes_mutex);

	}



	return ret;