diff options
author | Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com> | 2020-01-21 19:11:03 -0500 |
---|---|---|
committer | Alex Deucher <alexander.deucher@amd.com> | 2020-02-12 16:00:54 -0500 |
commit | 9593f4d6a69b0bc9577a7edb4c79be533957c161 (patch) | |
tree | 4ae1d3632f83c3ea1424cb9ae6a92999191e21ce /drivers/gpu/drm/amd/amdkfd | |
parent | 3c1224c02e995685bb386f259ae2f737eca739f3 (diff) |
drm/amdkfd: refactor runtime pm for baco
So far the kfd driver implemented same routines for runtime and system
wide suspend and resume (s2idle or mem). During system wide suspend the
kfd aquires an atomic lock that prevents any more user processes to
create queues and interact with kfd driver and amd gpu. This mechanism
created problem when amdgpu device is runtime suspended with BACO
enabled. Any application that relies on kfd driver fails to load because
the driver reports a locked kfd device since gpu is runtime suspended.
However, in an ideal case, when gpu is runtime suspended the kfd driver
should be able to:
- auto resume amdgpu driver whenever a client requests compute service
- prevent runtime suspend for amdgpu while kfd is in use
This change refactors the amdgpu and amdkfd drivers to support BACO and
runtime power management.
Reviewed-by: Oak Zeng <oak.zeng@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd')
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_device.c | 29 | ||||
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 1 | ||||
-rw-r--r-- | drivers/gpu/drm/amd/amdkfd/kfd_process.c | 42 |
3 files changed, 58 insertions, 14 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 2a9e40131735..a3d7ec0424cf 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -710,7 +710,7 @@ out: void kgd2kfd_device_exit(struct kfd_dev *kfd) { if (kfd->init_complete) { - kgd2kfd_suspend(kfd); + kgd2kfd_suspend(kfd, false); device_queue_manager_uninit(kfd->dqm); kfd_interrupt_exit(kfd); kfd_topology_remove_device(kfd); @@ -731,7 +731,7 @@ int kgd2kfd_pre_reset(struct kfd_dev *kfd) kfd->dqm->ops.pre_reset(kfd->dqm); - kgd2kfd_suspend(kfd); + kgd2kfd_suspend(kfd, false); kfd_signal_reset_event(kfd); return 0; @@ -765,21 +765,23 @@ bool kfd_is_locked(void) return (atomic_read(&kfd_locked) > 0); } -void kgd2kfd_suspend(struct kfd_dev *kfd) +void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm) { if (!kfd->init_complete) return; - /* For first KFD device suspend all the KFD processes */ - if (atomic_inc_return(&kfd_locked) == 1) - kfd_suspend_all_processes(); + /* for runtime suspend, skip locking kfd */ + if (!run_pm) { + /* For first KFD device suspend all the KFD processes */ + if (atomic_inc_return(&kfd_locked) == 1) + kfd_suspend_all_processes(); + } kfd->dqm->ops.stop(kfd->dqm); - kfd_iommu_suspend(kfd); } -int kgd2kfd_resume(struct kfd_dev *kfd) +int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm) { int ret, count; @@ -790,10 +792,13 @@ int kgd2kfd_resume(struct kfd_dev *kfd) if (ret) return ret; - count = atomic_dec_return(&kfd_locked); - WARN_ONCE(count < 0, "KFD suspend / resume ref. error"); - if (count == 0) - ret = kfd_resume_all_processes(); + /* for runtime resume, skip unlocking kfd */ + if (!run_pm) { + count = atomic_dec_return(&kfd_locked); + WARN_ONCE(count < 0, "KFD suspend / resume ref. error"); + if (count == 0) + ret = kfd_resume_all_processes(); + } return ret; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 9a06eb616cbf..f326e31624d7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -650,6 +650,7 @@ struct kfd_process_device { * function. */ bool already_dequeued; + bool runtime_inuse; /* Is this process/pasid bound to this device? (amd_iommu_bind_pasid) */ enum kfd_pdd_bound bound; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 98dcbb96b2e2..22abdbc6dfd7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -31,6 +31,7 @@ #include <linux/compat.h> #include <linux/mman.h> #include <linux/file.h> +#include <linux/pm_runtime.h> #include "amdgpu_amdkfd.h" #include "amdgpu.h" @@ -527,6 +528,16 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) kfree(pdd->qpd.doorbell_bitmap); idr_destroy(&pdd->alloc_idr); + /* + * before destroying pdd, make sure to report availability + * for auto suspend + */ + if (pdd->runtime_inuse) { + pm_runtime_mark_last_busy(pdd->dev->ddev->dev); + pm_runtime_put_autosuspend(pdd->dev->ddev->dev); + pdd->runtime_inuse = false; + } + kfree(pdd); } } @@ -844,6 +855,7 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, pdd->process = p; pdd->bound = PDD_UNBOUND; pdd->already_dequeued = false; + pdd->runtime_inuse = false; list_add(&pdd->per_device_list, &p->per_device_data); /* Init idr used for memory handle translation */ @@ -933,15 +945,41 @@ struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, return ERR_PTR(-ENOMEM); } + /* + * signal runtime-pm system to auto resume and prevent + * further runtime suspend once device pdd is created until + * pdd is destroyed. + */ + if (!pdd->runtime_inuse) { + err = pm_runtime_get_sync(dev->ddev->dev); + if (err < 0) + return ERR_PTR(err); + } + err = kfd_iommu_bind_process_to_device(pdd); if (err) - return ERR_PTR(err); + goto out; err = kfd_process_device_init_vm(pdd, NULL); if (err) - return ERR_PTR(err); + goto out; + + /* + * make sure that runtime_usage counter is incremented just once + * per pdd + */ + pdd->runtime_inuse = true; return pdd; + +out: + /* balance runpm reference count and exit with error */ + if (!pdd->runtime_inuse) { + pm_runtime_mark_last_busy(dev->ddev->dev); + pm_runtime_put_autosuspend(dev->ddev->dev); + } + + return ERR_PTR(err); } struct kfd_process_device *kfd_get_first_process_device_data( |