From: Andrey Grodzovsky <andrey.grodzovsky@amd.com> To: "Christian König" <ckoenig.leichtzumerken@gmail.com>, dri-devel@lists.freedesktop.org, amd-gfx@lists.freedesktop.org Cc: horace.chen@amd.com, christian.koenig@amd.com, Monk.Liu@amd.com Subject: Re: [RFC 4/6] drm/amdgpu: Serialize non TDR gpu recovery with TDRs Date: Mon, 20 Dec 2021 17:17:01 -0500 [thread overview] Message-ID: <bfecd7e4-6848-e2ee-4e80-f394403af08d@amd.com> (raw) In-Reply-To: <c143e561-d149-6680-0b89-2cda78ea1d51@gmail.com> On 2021-12-20 2:20 a.m., Christian König wrote: > Am 17.12.21 um 23:27 schrieb Andrey Grodzovsky: >> Use reset domain wq also for non TDR gpu recovery trigers >> such as sysfs and RAS. We must serialize all possible >> GPU recoveries to gurantee no concurrency there. >> For TDR call the original recovery function directly since >> it's already executed from within the wq. For others just >> use a wrapper to qeueue work and wait on it to finish. >> >> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> >> --- >> drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ >> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 33 +++++++++++++++++++++- >> drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 2 +- >> 3 files changed, 35 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> b/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> index b5ff76aae7e0..8e96b9a14452 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> @@ -1296,6 +1296,8 @@ bool amdgpu_device_has_job_running(struct >> amdgpu_device *adev); >> bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev); >> int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> struct amdgpu_job* job); >> +int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev, >> + struct amdgpu_job *job); >> void amdgpu_device_pci_config_reset(struct amdgpu_device *adev); >> int amdgpu_device_pci_reset(struct amdgpu_device *adev); >> bool amdgpu_device_need_post(struct amdgpu_device *adev); >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> index b595e6d699b5..55cd67b9ede2 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> @@ -4979,7 +4979,7 @@ static void amdgpu_device_recheck_guilty_jobs( >> * Returns 0 for success or an error on failure. >> */ >> -int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> +int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev, >> struct amdgpu_job *job) >> { >> struct list_head device_list, *device_list_handle = NULL; >> @@ -5236,6 +5236,37 @@ int amdgpu_device_gpu_recover(struct >> amdgpu_device *adev, >> return r; >> } >> +struct recover_work_struct { > > Please add an amdgpu_ prefix to the name. > >> + struct work_struct base; >> + struct amdgpu_device *adev; >> + struct amdgpu_job *job; >> + int ret; >> +}; >> + >> +static void amdgpu_device_queue_gpu_recover_work(struct work_struct >> *work) >> +{ >> + struct recover_work_struct *recover_work = container_of(work, >> struct recover_work_struct, base); >> + >> + recover_work->ret = >> amdgpu_device_gpu_recover_imp(recover_work->adev, recover_work->job); >> +} >> +/* >> + * Serialize gpu recover into reset domain single threaded wq >> + */ >> +int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> + struct amdgpu_job *job) >> +{ >> + struct recover_work_struct work = {.adev = adev, .job = job}; >> + >> + INIT_WORK(&work.base, amdgpu_device_queue_gpu_recover_work); >> + >> + if (!queue_work(adev->reset_domain.wq, &work.base)) >> + return -EAGAIN; >> + >> + flush_work(&work.base); >> + >> + return work.ret; >> +} > > Maybe that should be part of the scheduler code? Not really sure, just > an idea. Seems to me that since the reset domain is almost always above a single scheduler granularity then it wouldn't feet very well there. Andrey > > Apart from that looks good to me, > Christian. > >> + >> /** >> * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot >> * >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> index bfc47bea23db..38c9fd7b7ad4 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> @@ -63,7 +63,7 @@ static enum drm_gpu_sched_stat >> amdgpu_job_timedout(struct drm_sched_job *s_job) >> ti.process_name, ti.tgid, ti.task_name, ti.pid); >> if (amdgpu_device_should_recover_gpu(ring->adev)) { >> - amdgpu_device_gpu_recover(ring->adev, job); >> + amdgpu_device_gpu_recover_imp(ring->adev, job); >> } else { >> drm_sched_suspend_timeout(&ring->sched); >> if (amdgpu_sriov_vf(adev)) >
WARNING: multiple messages have this Message-ID (diff)
From: Andrey Grodzovsky <andrey.grodzovsky@amd.com> To: "Christian König" <ckoenig.leichtzumerken@gmail.com>, dri-devel@lists.freedesktop.org, amd-gfx@lists.freedesktop.org Cc: daniel@ffwll.ch, horace.chen@amd.com, christian.koenig@amd.com, Monk.Liu@amd.com Subject: Re: [RFC 4/6] drm/amdgpu: Serialize non TDR gpu recovery with TDRs Date: Mon, 20 Dec 2021 17:17:01 -0500 [thread overview] Message-ID: <bfecd7e4-6848-e2ee-4e80-f394403af08d@amd.com> (raw) In-Reply-To: <c143e561-d149-6680-0b89-2cda78ea1d51@gmail.com> On 2021-12-20 2:20 a.m., Christian König wrote: > Am 17.12.21 um 23:27 schrieb Andrey Grodzovsky: >> Use reset domain wq also for non TDR gpu recovery trigers >> such as sysfs and RAS. We must serialize all possible >> GPU recoveries to gurantee no concurrency there. >> For TDR call the original recovery function directly since >> it's already executed from within the wq. For others just >> use a wrapper to qeueue work and wait on it to finish. >> >> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> >> --- >> drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ >> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 33 +++++++++++++++++++++- >> drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 2 +- >> 3 files changed, 35 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> b/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> index b5ff76aae7e0..8e96b9a14452 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h >> @@ -1296,6 +1296,8 @@ bool amdgpu_device_has_job_running(struct >> amdgpu_device *adev); >> bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev); >> int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> struct amdgpu_job* job); >> +int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev, >> + struct amdgpu_job *job); >> void amdgpu_device_pci_config_reset(struct amdgpu_device *adev); >> int amdgpu_device_pci_reset(struct amdgpu_device *adev); >> bool amdgpu_device_need_post(struct amdgpu_device *adev); >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> index b595e6d699b5..55cd67b9ede2 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c >> @@ -4979,7 +4979,7 @@ static void amdgpu_device_recheck_guilty_jobs( >> * Returns 0 for success or an error on failure. >> */ >> -int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> +int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev, >> struct amdgpu_job *job) >> { >> struct list_head device_list, *device_list_handle = NULL; >> @@ -5236,6 +5236,37 @@ int amdgpu_device_gpu_recover(struct >> amdgpu_device *adev, >> return r; >> } >> +struct recover_work_struct { > > Please add an amdgpu_ prefix to the name. > >> + struct work_struct base; >> + struct amdgpu_device *adev; >> + struct amdgpu_job *job; >> + int ret; >> +}; >> + >> +static void amdgpu_device_queue_gpu_recover_work(struct work_struct >> *work) >> +{ >> + struct recover_work_struct *recover_work = container_of(work, >> struct recover_work_struct, base); >> + >> + recover_work->ret = >> amdgpu_device_gpu_recover_imp(recover_work->adev, recover_work->job); >> +} >> +/* >> + * Serialize gpu recover into reset domain single threaded wq >> + */ >> +int amdgpu_device_gpu_recover(struct amdgpu_device *adev, >> + struct amdgpu_job *job) >> +{ >> + struct recover_work_struct work = {.adev = adev, .job = job}; >> + >> + INIT_WORK(&work.base, amdgpu_device_queue_gpu_recover_work); >> + >> + if (!queue_work(adev->reset_domain.wq, &work.base)) >> + return -EAGAIN; >> + >> + flush_work(&work.base); >> + >> + return work.ret; >> +} > > Maybe that should be part of the scheduler code? Not really sure, just > an idea. Seems to me that since the reset domain is almost always above a single scheduler granularity then it wouldn't feet very well there. Andrey > > Apart from that looks good to me, > Christian. > >> + >> /** >> * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot >> * >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> index bfc47bea23db..38c9fd7b7ad4 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c >> @@ -63,7 +63,7 @@ static enum drm_gpu_sched_stat >> amdgpu_job_timedout(struct drm_sched_job *s_job) >> ti.process_name, ti.tgid, ti.task_name, ti.pid); >> if (amdgpu_device_should_recover_gpu(ring->adev)) { >> - amdgpu_device_gpu_recover(ring->adev, job); >> + amdgpu_device_gpu_recover_imp(ring->adev, job); >> } else { >> drm_sched_suspend_timeout(&ring->sched); >> if (amdgpu_sriov_vf(adev)) >
next prev parent reply other threads:[~2021-12-20 22:17 UTC|newest] Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-12-17 22:27 [RFC 0/6] Define and use reset domain for GPU recovery in amdgpu Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-17 22:27 ` [RFC 1/6] drm/amdgpu: Init GPU reset single threaded wq Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-17 22:27 ` [RFC 2/6] drm/amdgpu: Move scheduler init to after XGMI is ready Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-20 7:16 ` Christian König 2021-12-20 7:16 ` Christian König 2021-12-20 21:51 ` Andrey Grodzovsky 2021-12-20 21:51 ` Andrey Grodzovsky 2021-12-21 7:05 ` Christian König 2021-12-21 7:05 ` Christian König 2021-12-17 22:27 ` [RFC 3/6] drm/amdgpu: Fix crash on modprobe Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-20 7:17 ` Christian König 2021-12-20 7:17 ` Christian König 2021-12-20 19:22 ` Andrey Grodzovsky 2021-12-20 19:22 ` Andrey Grodzovsky 2021-12-21 7:02 ` Christian König 2021-12-21 7:02 ` Christian König 2021-12-21 16:03 ` Andrey Grodzovsky 2021-12-21 16:03 ` Andrey Grodzovsky 2021-12-22 7:50 ` Christian König 2021-12-22 7:50 ` Christian König 2021-12-17 22:27 ` [RFC 4/6] drm/amdgpu: Serialize non TDR gpu recovery with TDRs Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-20 7:20 ` Christian König 2021-12-20 7:20 ` Christian König 2021-12-20 22:17 ` Andrey Grodzovsky [this message] 2021-12-20 22:17 ` Andrey Grodzovsky 2021-12-21 7:59 ` Christian König 2021-12-21 7:59 ` Christian König 2021-12-21 16:10 ` Andrey Grodzovsky 2021-12-21 16:10 ` Andrey Grodzovsky 2021-12-17 22:27 ` [RFC 5/6] drm/amdgpu: Drop hive->in_reset Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-17 22:27 ` [RFC 6/6] drm/amdgpu: Drop concurrent GPU reset protection for device Andrey Grodzovsky 2021-12-17 22:27 ` Andrey Grodzovsky 2021-12-20 7:25 ` [RFC 0/6] Define and use reset domain for GPU recovery in amdgpu Christian König 2021-12-20 7:25 ` Christian König 2021-12-20 9:43 ` Daniel Vetter 2021-12-20 9:43 ` Daniel Vetter 2021-12-20 17:06 ` Liu, Shaoyun 2021-12-20 17:06 ` Liu, Shaoyun 2021-12-20 19:11 ` Andrey Grodzovsky 2021-12-20 19:11 ` Andrey Grodzovsky
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=bfecd7e4-6848-e2ee-4e80-f394403af08d@amd.com \ --to=andrey.grodzovsky@amd.com \ --cc=Monk.Liu@amd.com \ --cc=amd-gfx@lists.freedesktop.org \ --cc=christian.koenig@amd.com \ --cc=ckoenig.leichtzumerken@gmail.com \ --cc=dri-devel@lists.freedesktop.org \ --cc=horace.chen@amd.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.