* [PATCH v6 0/7] DRM scheduler changes for Xe
@ 2023-10-17 15:09 Matthew Brost
2023-10-17 15:09 ` [PATCH v6 1/7] drm/sched: Add drm_sched_wqueue_* helpers Matthew Brost
` (7 more replies)
0 siblings, 8 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
As a prerequisite to merging the new Intel Xe DRM driver [1] [2], we
have been asked to merge our common DRM scheduler patches first.
This a continuation of a RFC [3] with all comments addressed, ready for
a full review, and hopefully in state which can merged in the near
future. More details of this series can found in the cover letter of the
RFC [3].
These changes have been tested with the Xe driver. Based on drm-tip branch.
A follow up series will be posted to address some of dakr requets for
kernel doc changes.
v2:
- Break run job, free job, and process message in own work items
- This might break other drivers as run job and free job now can run in
parallel, can fix up if needed
v3:
- Include missing patch 'drm/sched: Add drm_sched_submit_* helpers'
- Fix issue with setting timestamp to early
- Don't dequeue jobs for single entity after calling entity fini
- Flush pending jobs on entity fini
- Add documentation for entity teardown
- Add Matthew Brost to maintainers of DRM scheduler
v4:
- Drop message interface
- Drop 'Flush pending jobs on entity fini'
- Drop 'Add documentation for entity teardown'
- Address all feedback
v5:
- Address Luben's feedback
- Drop starting TDR after calling run_job()
- Drop adding Matthew Brost to maintainers of DRM scheduler
v6:
- Address Luben's feedback
- Include base commit
Matt
[1] https://gitlab.freedesktop.org/drm/xe/kernel
[2] https://patchwork.freedesktop.org/series/112188/
[3] https://patchwork.freedesktop.org/series/116055/
Matthew Brost (7):
drm/sched: Add drm_sched_wqueue_* helpers
drm/sched: Convert drm scheduler to use a work queue rather than
kthread
drm/sched: Move schedule policy to scheduler
drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
drm/sched: Split free_job into own work item
drm/sched: Add drm_sched_start_timeout_unlocked helper
drm/sched: Add a helper to queue TDR immediately
.../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 15 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +-
drivers/gpu/drm/etnaviv/etnaviv_sched.c | 5 +-
drivers/gpu/drm/lima/lima_sched.c | 5 +-
drivers/gpu/drm/msm/adreno/adreno_device.c | 6 +-
drivers/gpu/drm/msm/msm_ringbuffer.c | 7 +-
drivers/gpu/drm/nouveau/nouveau_sched.c | 5 +-
drivers/gpu/drm/panfrost/panfrost_job.c | 5 +-
drivers/gpu/drm/scheduler/sched_entity.c | 85 ++-
drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
drivers/gpu/drm/scheduler/sched_main.c | 507 ++++++++++++------
drivers/gpu/drm/v3d/v3d_sched.c | 25 +-
include/drm/gpu_scheduler.h | 48 +-
14 files changed, 498 insertions(+), 234 deletions(-)
base-commit: 201c8a7bd1f3f415920a2df4b8a8817e973f42fe
--
2.34.1
^ permalink raw reply [flat|nested] 22+ messages in thread
* [PATCH v6 1/7] drm/sched: Add drm_sched_wqueue_* helpers
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-17 15:09 ` [PATCH v6 2/7] drm/sched: Convert drm scheduler to use a work queue rather than kthread Matthew Brost
` (6 subsequent siblings)
7 siblings, 0 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
Add scheduler wqueue ready, stop, and start helpers to hide the
implementation details of the scheduler from the drivers.
v2:
- s/sched_wqueue/sched_wqueue (Luben)
- Remove the extra white line after the return-statement (Luben)
- update drm_sched_wqueue_ready comment (Luben)
Cc: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
---
.../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 15 +++----
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12 +++---
drivers/gpu/drm/msm/adreno/adreno_device.c | 6 ++-
drivers/gpu/drm/scheduler/sched_main.c | 39 ++++++++++++++++++-
include/drm/gpu_scheduler.h | 3 ++
6 files changed, 59 insertions(+), 18 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 625db444df1c..10d56979fe3b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -290,7 +290,7 @@ static int suspend_resume_compute_scheduler(struct amdgpu_device *adev, bool sus
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
- if (!(ring && ring->sched.thread))
+ if (!(ring && drm_sched_wqueue_ready(&ring->sched)))
continue;
/* stop secheduler and drain ring. */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index a4faea4fa0b5..a4c0bb358db7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -1659,9 +1659,9 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
- kthread_park(ring->sched.thread);
+ drm_sched_wqueue_stop(&ring->sched);
}
seq_puts(m, "run ib test:\n");
@@ -1675,9 +1675,9 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
- kthread_unpark(ring->sched.thread);
+ drm_sched_wqueue_start(&ring->sched);
}
up_write(&adev->reset_domain->sem);
@@ -1897,7 +1897,8 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
ring = adev->rings[val];
- if (!ring || !ring->funcs->preempt_ib || !ring->sched.thread)
+ if (!ring || !ring->funcs->preempt_ib ||
+ !drm_sched_wqueue_ready(&ring->sched))
return -EINVAL;
/* the last preemption failed */
@@ -1915,7 +1916,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
goto pro_end;
/* stop the scheduler */
- kthread_park(ring->sched.thread);
+ drm_sched_wqueue_stop(&ring->sched);
/* preempt the IB */
r = amdgpu_ring_preempt_ib(ring);
@@ -1949,7 +1950,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
failure:
/* restart the scheduler */
- kthread_unpark(ring->sched.thread);
+ drm_sched_wqueue_start(&ring->sched);
up_read(&adev->reset_domain->sem);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 2b8356699f23..b1aafe815f28 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4588,7 +4588,7 @@ bool amdgpu_device_has_job_running(struct amdgpu_device *adev)
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
spin_lock(&ring->sched.job_list_lock);
@@ -4727,7 +4727,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
/* Clear job fence from fence drv to avoid force_completion
@@ -5266,7 +5266,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = tmp_adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
drm_sched_stop(&ring->sched, job ? &job->base : NULL);
@@ -5341,7 +5341,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = tmp_adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
drm_sched_start(&ring->sched, true);
@@ -5667,7 +5667,7 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
drm_sched_stop(&ring->sched, NULL);
@@ -5795,7 +5795,7 @@ void amdgpu_pci_resume(struct pci_dev *pdev)
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
- if (!ring || !ring->sched.thread)
+ if (!ring || !drm_sched_wqueue_ready(&ring->sched))
continue;
drm_sched_start(&ring->sched, true);
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index fa527935ffd4..8fa9ce3746b6 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -809,7 +809,8 @@ static void suspend_scheduler(struct msm_gpu *gpu)
*/
for (i = 0; i < gpu->nr_rings; i++) {
struct drm_gpu_scheduler *sched = &gpu->rb[i]->sched;
- kthread_park(sched->thread);
+
+ drm_sched_wqueue_stop(sched);
}
}
@@ -819,7 +820,8 @@ static void resume_scheduler(struct msm_gpu *gpu)
for (i = 0; i < gpu->nr_rings; i++) {
struct drm_gpu_scheduler *sched = &gpu->rb[i]->sched;
- kthread_unpark(sched->thread);
+
+ drm_sched_wqueue_start(sched);
}
}
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 5a3a622fc672..6f2f7dd4ba0a 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -439,7 +439,7 @@ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad)
{
struct drm_sched_job *s_job, *tmp;
- kthread_park(sched->thread);
+ drm_sched_wqueue_stop(sched);
/*
* Reinsert back the bad job here - now it's safe as
@@ -552,7 +552,7 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
spin_unlock(&sched->job_list_lock);
}
- kthread_unpark(sched->thread);
+ drm_sched_wqueue_start(sched);
}
EXPORT_SYMBOL(drm_sched_start);
@@ -1206,3 +1206,38 @@ void drm_sched_increase_karma(struct drm_sched_job *bad)
}
}
EXPORT_SYMBOL(drm_sched_increase_karma);
+
+/**
+ * drm_sched_wqueue_ready - Is the scheduler ready for submission
+ *
+ * @sched: scheduler instance
+ *
+ * Returns true if submission is ready
+ */
+bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched)
+{
+ return !!sched->thread;
+}
+EXPORT_SYMBOL(drm_sched_wqueue_ready);
+
+/**
+ * drm_sched_wqueue_stop - stop scheduler submission
+ *
+ * @sched: scheduler instance
+ */
+void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
+{
+ kthread_park(sched->thread);
+}
+EXPORT_SYMBOL(drm_sched_wqueue_stop);
+
+/**
+ * drm_sched_wqueue_start - start scheduler submission
+ *
+ * @sched: scheduler instance
+ */
+void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
+{
+ kthread_unpark(sched->thread);
+}
+EXPORT_SYMBOL(drm_sched_wqueue_start);
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index f9544d9b670d..38578fe74573 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -550,6 +550,9 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
void drm_sched_job_cleanup(struct drm_sched_job *job);
void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched);
+bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched);
+void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched);
+void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched);
void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad);
void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery);
void drm_sched_resubmit_jobs(struct drm_gpu_scheduler *sched);
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 2/7] drm/sched: Convert drm scheduler to use a work queue rather than kthread
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
2023-10-17 15:09 ` [PATCH v6 1/7] drm/sched: Add drm_sched_wqueue_* helpers Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-17 15:09 ` [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler Matthew Brost
` (5 subsequent siblings)
7 siblings, 0 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
In Xe, the new Intel GPU driver, a choice has made to have a 1 to 1
mapping between a drm_gpu_scheduler and drm_sched_entity. At first this
seems a bit odd but let us explain the reasoning below.
1. In Xe the submission order from multiple drm_sched_entity is not
guaranteed to be the same completion even if targeting the same hardware
engine. This is because in Xe we have a firmware scheduler, the GuC,
which allowed to reorder, timeslice, and preempt submissions. If a using
shared drm_gpu_scheduler across multiple drm_sched_entity, the TDR falls
apart as the TDR expects submission order == completion order. Using a
dedicated drm_gpu_scheduler per drm_sched_entity solve this problem.
2. In Xe submissions are done via programming a ring buffer (circular
buffer), a drm_gpu_scheduler provides a limit on number of jobs, if the
limit of number jobs is set to RING_SIZE / MAX_SIZE_PER_JOB we get flow
control on the ring for free.
A problem with this design is currently a drm_gpu_scheduler uses a
kthread for submission / job cleanup. This doesn't scale if a large
number of drm_gpu_scheduler are used. To work around the scaling issue,
use a worker rather than kthread for submission / job cleanup.
v2:
- (Rob Clark) Fix msm build
- Pass in run work queue
v3:
- (Boris) don't have loop in worker
v4:
- (Tvrtko) break out submit ready, stop, start helpers into own patch
v5:
- (Boris) default to ordered work queue
v6:
- (Luben / checkpatch) fix alignment in msm_ringbuffer.c
- (Luben) s/drm_sched_submit_queue/drm_sched_wqueue_enqueue
- (Luben) Update comment for drm_sched_wqueue_enqueue
- (Luben) Positive check for submit_wq in drm_sched_init
- (Luben) s/alloc_submit_wq/own_submit_wq
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
drivers/gpu/drm/etnaviv/etnaviv_sched.c | 2 +-
drivers/gpu/drm/lima/lima_sched.c | 2 +-
drivers/gpu/drm/msm/msm_ringbuffer.c | 7 +-
drivers/gpu/drm/nouveau/nouveau_sched.c | 2 +-
drivers/gpu/drm/panfrost/panfrost_job.c | 2 +-
drivers/gpu/drm/scheduler/sched_main.c | 118 ++++++++++-----------
drivers/gpu/drm/v3d/v3d_sched.c | 10 +-
include/drm/gpu_scheduler.h | 14 ++-
9 files changed, 82 insertions(+), 77 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index b1aafe815f28..b54c4d771104 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2279,7 +2279,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
break;
}
- r = drm_sched_init(&ring->sched, &amdgpu_sched_ops,
+ r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, NULL,
ring->num_hw_submission, 0,
timeout, adev->reset_domain->wq,
ring->sched_score, ring->name,
diff --git a/drivers/gpu/drm/etnaviv/etnaviv_sched.c b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
index 345fec6cb1a4..618a804ddc34 100644
--- a/drivers/gpu/drm/etnaviv/etnaviv_sched.c
+++ b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
@@ -134,7 +134,7 @@ int etnaviv_sched_init(struct etnaviv_gpu *gpu)
{
int ret;
- ret = drm_sched_init(&gpu->sched, &etnaviv_sched_ops,
+ ret = drm_sched_init(&gpu->sched, &etnaviv_sched_ops, NULL,
etnaviv_hw_jobs_limit, etnaviv_job_hang_limit,
msecs_to_jiffies(500), NULL, NULL,
dev_name(gpu->dev), gpu->dev);
diff --git a/drivers/gpu/drm/lima/lima_sched.c b/drivers/gpu/drm/lima/lima_sched.c
index ffd91a5ee299..8d858aed0e56 100644
--- a/drivers/gpu/drm/lima/lima_sched.c
+++ b/drivers/gpu/drm/lima/lima_sched.c
@@ -488,7 +488,7 @@ int lima_sched_pipe_init(struct lima_sched_pipe *pipe, const char *name)
INIT_WORK(&pipe->recover_work, lima_sched_recover_work);
- return drm_sched_init(&pipe->base, &lima_sched_ops, 1,
+ return drm_sched_init(&pipe->base, &lima_sched_ops, NULL, 1,
lima_job_hang_limit,
msecs_to_jiffies(timeout), NULL,
NULL, name, pipe->ldev->dev);
diff --git a/drivers/gpu/drm/msm/msm_ringbuffer.c b/drivers/gpu/drm/msm/msm_ringbuffer.c
index 40c0bc35a44c..1097f8e93d6b 100644
--- a/drivers/gpu/drm/msm/msm_ringbuffer.c
+++ b/drivers/gpu/drm/msm/msm_ringbuffer.c
@@ -94,9 +94,10 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
/* currently managing hangcheck ourselves: */
sched_timeout = MAX_SCHEDULE_TIMEOUT;
- ret = drm_sched_init(&ring->sched, &msm_sched_ops,
- num_hw_submissions, 0, sched_timeout,
- NULL, NULL, to_msm_bo(ring->bo)->name, gpu->dev->dev);
+ ret = drm_sched_init(&ring->sched, &msm_sched_ops, NULL,
+ num_hw_submissions, 0, sched_timeout,
+ NULL, NULL, to_msm_bo(ring->bo)->name,
+ gpu->dev->dev);
if (ret) {
goto fail;
}
diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c b/drivers/gpu/drm/nouveau/nouveau_sched.c
index 3b7ea5221226..4c959dec42b3 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sched.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
@@ -435,7 +435,7 @@ int nouveau_sched_init(struct nouveau_drm *drm)
if (!drm->sched_wq)
return -ENOMEM;
- return drm_sched_init(sched, &nouveau_sched_ops,
+ return drm_sched_init(sched, &nouveau_sched_ops, NULL,
NOUVEAU_SCHED_HW_SUBMISSIONS, 0, job_hang_limit,
NULL, NULL, "nouveau_sched", drm->dev->dev);
}
diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
index fb16de2d0420..934b7b930c76 100644
--- a/drivers/gpu/drm/panfrost/panfrost_job.c
+++ b/drivers/gpu/drm/panfrost/panfrost_job.c
@@ -852,7 +852,7 @@ int panfrost_job_init(struct panfrost_device *pfdev)
js->queue[j].fence_context = dma_fence_context_alloc(1);
ret = drm_sched_init(&js->queue[j].sched,
- &panfrost_sched_ops,
+ &panfrost_sched_ops, NULL,
nentries, 0,
msecs_to_jiffies(JOB_TIMEOUT_MS),
pfdev->reset.wq,
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 6f2f7dd4ba0a..8b1d52cff1e9 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -48,7 +48,6 @@
* through the jobs entity pointer.
*/
-#include <linux/kthread.h>
#include <linux/wait.h>
#include <linux/sched.h>
#include <linux/completion.h>
@@ -256,6 +255,16 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
return rb ? rb_entry(rb, struct drm_sched_entity, rb_tree_node) : NULL;
}
+/**
+ * drm_sched_wqueue_enqueue - enqueue scheduler submission
+ * @sched: scheduler instance
+ */
+static void drm_sched_wqueue_enqueue(struct drm_gpu_scheduler *sched)
+{
+ if (!READ_ONCE(sched->pause_submit))
+ queue_work(sched->submit_wq, &sched->work_submit);
+}
+
/**
* drm_sched_job_done - complete a job
* @s_job: pointer to the job which is done
@@ -275,7 +284,7 @@ static void drm_sched_job_done(struct drm_sched_job *s_job, int result)
dma_fence_get(&s_fence->finished);
drm_sched_fence_finished(s_fence, result);
dma_fence_put(&s_fence->finished);
- wake_up_interruptible(&sched->wake_up_worker);
+ drm_sched_wqueue_enqueue(sched);
}
/**
@@ -868,7 +877,7 @@ static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched)
{
if (drm_sched_can_queue(sched))
- wake_up_interruptible(&sched->wake_up_worker);
+ drm_sched_wqueue_enqueue(sched);
}
/**
@@ -978,61 +987,42 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list,
}
EXPORT_SYMBOL(drm_sched_pick_best);
-/**
- * drm_sched_blocked - check if the scheduler is blocked
- *
- * @sched: scheduler instance
- *
- * Returns true if blocked, otherwise false.
- */
-static bool drm_sched_blocked(struct drm_gpu_scheduler *sched)
-{
- if (kthread_should_park()) {
- kthread_parkme();
- return true;
- }
-
- return false;
-}
-
/**
* drm_sched_main - main scheduler thread
*
* @param: scheduler instance
- *
- * Returns 0.
*/
-static int drm_sched_main(void *param)
+static void drm_sched_main(struct work_struct *w)
{
- struct drm_gpu_scheduler *sched = (struct drm_gpu_scheduler *)param;
+ struct drm_gpu_scheduler *sched =
+ container_of(w, struct drm_gpu_scheduler, work_submit);
+ struct drm_sched_entity *entity;
+ struct drm_sched_job *cleanup_job;
int r;
- sched_set_fifo_low(current);
+ if (READ_ONCE(sched->pause_submit))
+ return;
- while (!kthread_should_stop()) {
- struct drm_sched_entity *entity = NULL;
- struct drm_sched_fence *s_fence;
- struct drm_sched_job *sched_job;
- struct dma_fence *fence;
- struct drm_sched_job *cleanup_job = NULL;
+ cleanup_job = drm_sched_get_cleanup_job(sched);
+ entity = drm_sched_select_entity(sched);
- wait_event_interruptible(sched->wake_up_worker,
- (cleanup_job = drm_sched_get_cleanup_job(sched)) ||
- (!drm_sched_blocked(sched) &&
- (entity = drm_sched_select_entity(sched))) ||
- kthread_should_stop());
+ if (!entity && !cleanup_job)
+ return; /* No more work */
- if (cleanup_job)
- sched->ops->free_job(cleanup_job);
+ if (cleanup_job)
+ sched->ops->free_job(cleanup_job);
- if (!entity)
- continue;
+ if (entity) {
+ struct dma_fence *fence;
+ struct drm_sched_fence *s_fence;
+ struct drm_sched_job *sched_job;
sched_job = drm_sched_entity_pop_job(entity);
-
if (!sched_job) {
complete_all(&entity->entity_idle);
- continue;
+ if (!cleanup_job)
+ return; /* No more work */
+ goto again;
}
s_fence = sched_job->s_fence;
@@ -1063,7 +1053,9 @@ static int drm_sched_main(void *param)
wake_up(&sched->job_scheduled);
}
- return 0;
+
+again:
+ drm_sched_wqueue_enqueue(sched);
}
/**
@@ -1071,6 +1063,8 @@ static int drm_sched_main(void *param)
*
* @sched: scheduler instance
* @ops: backend operations for this scheduler
+ * @submit_wq: workqueue to use for submission. If NULL, an ordered wq is
+ * allocated and used
* @hw_submission: number of hw submissions that can be in flight
* @hang_limit: number of times to allow a job to hang before dropping it
* @timeout: timeout value in jiffies for the scheduler
@@ -1084,14 +1078,25 @@ static int drm_sched_main(void *param)
*/
int drm_sched_init(struct drm_gpu_scheduler *sched,
const struct drm_sched_backend_ops *ops,
+ struct workqueue_struct *submit_wq,
unsigned hw_submission, unsigned hang_limit,
long timeout, struct workqueue_struct *timeout_wq,
atomic_t *score, const char *name, struct device *dev)
{
- int i, ret;
+ int i;
sched->ops = ops;
sched->hw_submission_limit = hw_submission;
sched->name = name;
+ if (submit_wq) {
+ sched->submit_wq = submit_wq;
+ sched->own_submit_wq = false;
+ } else {
+ sched->submit_wq = alloc_ordered_workqueue(name, 0);
+ if (!sched->submit_wq)
+ return -ENOMEM;
+
+ sched->own_submit_wq = true;
+ }
sched->timeout = timeout;
sched->timeout_wq = timeout_wq ? : system_wq;
sched->hang_limit = hang_limit;
@@ -1100,23 +1105,15 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
drm_sched_rq_init(sched, &sched->sched_rq[i]);
- init_waitqueue_head(&sched->wake_up_worker);
init_waitqueue_head(&sched->job_scheduled);
INIT_LIST_HEAD(&sched->pending_list);
spin_lock_init(&sched->job_list_lock);
atomic_set(&sched->hw_rq_count, 0);
INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
+ INIT_WORK(&sched->work_submit, drm_sched_main);
atomic_set(&sched->_score, 0);
atomic64_set(&sched->job_id_count, 0);
-
- /* Each scheduler will run on a seperate kernel thread */
- sched->thread = kthread_run(drm_sched_main, sched, sched->name);
- if (IS_ERR(sched->thread)) {
- ret = PTR_ERR(sched->thread);
- sched->thread = NULL;
- DRM_DEV_ERROR(sched->dev, "Failed to create scheduler for %s.\n", name);
- return ret;
- }
+ sched->pause_submit = false;
sched->ready = true;
return 0;
@@ -1135,8 +1132,7 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
struct drm_sched_entity *s_entity;
int i;
- if (sched->thread)
- kthread_stop(sched->thread);
+ drm_sched_wqueue_stop(sched);
for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
struct drm_sched_rq *rq = &sched->sched_rq[i];
@@ -1159,6 +1155,8 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
/* Confirm no work left behind accessing device structures */
cancel_delayed_work_sync(&sched->work_tdr);
+ if (sched->own_submit_wq)
+ destroy_workqueue(sched->submit_wq);
sched->ready = false;
}
EXPORT_SYMBOL(drm_sched_fini);
@@ -1216,7 +1214,7 @@ EXPORT_SYMBOL(drm_sched_increase_karma);
*/
bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched)
{
- return !!sched->thread;
+ return sched->ready;
}
EXPORT_SYMBOL(drm_sched_wqueue_ready);
@@ -1227,7 +1225,8 @@ EXPORT_SYMBOL(drm_sched_wqueue_ready);
*/
void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
{
- kthread_park(sched->thread);
+ WRITE_ONCE(sched->pause_submit, true);
+ cancel_work_sync(&sched->work_submit);
}
EXPORT_SYMBOL(drm_sched_wqueue_stop);
@@ -1238,6 +1237,7 @@ EXPORT_SYMBOL(drm_sched_wqueue_stop);
*/
void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
{
- kthread_unpark(sched->thread);
+ WRITE_ONCE(sched->pause_submit, false);
+ queue_work(sched->submit_wq, &sched->work_submit);
}
EXPORT_SYMBOL(drm_sched_wqueue_start);
diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
index 06238e6d7f5c..38e092ea41e6 100644
--- a/drivers/gpu/drm/v3d/v3d_sched.c
+++ b/drivers/gpu/drm/v3d/v3d_sched.c
@@ -388,7 +388,7 @@ v3d_sched_init(struct v3d_dev *v3d)
int ret;
ret = drm_sched_init(&v3d->queue[V3D_BIN].sched,
- &v3d_bin_sched_ops,
+ &v3d_bin_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
NULL, "v3d_bin", v3d->drm.dev);
@@ -396,7 +396,7 @@ v3d_sched_init(struct v3d_dev *v3d)
return ret;
ret = drm_sched_init(&v3d->queue[V3D_RENDER].sched,
- &v3d_render_sched_ops,
+ &v3d_render_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
NULL, "v3d_render", v3d->drm.dev);
@@ -404,7 +404,7 @@ v3d_sched_init(struct v3d_dev *v3d)
goto fail;
ret = drm_sched_init(&v3d->queue[V3D_TFU].sched,
- &v3d_tfu_sched_ops,
+ &v3d_tfu_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
NULL, "v3d_tfu", v3d->drm.dev);
@@ -413,7 +413,7 @@ v3d_sched_init(struct v3d_dev *v3d)
if (v3d_has_csd(v3d)) {
ret = drm_sched_init(&v3d->queue[V3D_CSD].sched,
- &v3d_csd_sched_ops,
+ &v3d_csd_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
NULL, "v3d_csd", v3d->drm.dev);
@@ -421,7 +421,7 @@ v3d_sched_init(struct v3d_dev *v3d)
goto fail;
ret = drm_sched_init(&v3d->queue[V3D_CACHE_CLEAN].sched,
- &v3d_cache_clean_sched_ops,
+ &v3d_cache_clean_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
NULL, "v3d_cache_clean", v3d->drm.dev);
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index 38578fe74573..211bd3cdabdc 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -473,17 +473,16 @@ struct drm_sched_backend_ops {
* @timeout: the time after which a job is removed from the scheduler.
* @name: name of the ring for which this scheduler is being used.
* @sched_rq: priority wise array of run queues.
- * @wake_up_worker: the wait queue on which the scheduler sleeps until a job
- * is ready to be scheduled.
* @job_scheduled: once @drm_sched_entity_do_release is called the scheduler
* waits on this wait queue until all the scheduled jobs are
* finished.
* @hw_rq_count: the number of jobs currently in the hardware queue.
* @job_id_count: used to assign unique id to the each job.
+ * @submit_wq: workqueue used to queue @work_submit
* @timeout_wq: workqueue used to queue @work_tdr
+ * @work_submit: schedules jobs and cleans up entities
* @work_tdr: schedules a delayed call to @drm_sched_job_timedout after the
* timeout interval is over.
- * @thread: the kthread on which the scheduler which run.
* @pending_list: the list of jobs which are currently in the job queue.
* @job_list_lock: lock to protect the pending_list.
* @hang_limit: once the hangs by a job crosses this limit then it is marked
@@ -492,6 +491,8 @@ struct drm_sched_backend_ops {
* @_score: score used when the driver doesn't provide one
* @ready: marks if the underlying HW is ready to work
* @free_guilty: A hit to time out handler to free the guilty job.
+ * @pause_submit: pause queuing of @work_submit on @submit_wq
+ * @own_submit_wq: scheduler owns allocation of @submit_wq
* @dev: system &struct device
*
* One scheduler is implemented for each hardware ring.
@@ -502,13 +503,13 @@ struct drm_gpu_scheduler {
long timeout;
const char *name;
struct drm_sched_rq sched_rq[DRM_SCHED_PRIORITY_COUNT];
- wait_queue_head_t wake_up_worker;
wait_queue_head_t job_scheduled;
atomic_t hw_rq_count;
atomic64_t job_id_count;
+ struct workqueue_struct *submit_wq;
struct workqueue_struct *timeout_wq;
+ struct work_struct work_submit;
struct delayed_work work_tdr;
- struct task_struct *thread;
struct list_head pending_list;
spinlock_t job_list_lock;
int hang_limit;
@@ -516,11 +517,14 @@ struct drm_gpu_scheduler {
atomic_t _score;
bool ready;
bool free_guilty;
+ bool pause_submit;
+ bool own_submit_wq;
struct device *dev;
};
int drm_sched_init(struct drm_gpu_scheduler *sched,
const struct drm_sched_backend_ops *ops,
+ struct workqueue_struct *submit_wq,
uint32_t hw_submission, unsigned hang_limit,
long timeout, struct workqueue_struct *timeout_wq,
atomic_t *score, const char *name, struct device *dev);
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
2023-10-17 15:09 ` [PATCH v6 1/7] drm/sched: Add drm_sched_wqueue_* helpers Matthew Brost
2023-10-17 15:09 ` [PATCH v6 2/7] drm/sched: Convert drm scheduler to use a work queue rather than kthread Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-24 3:34 ` Luben Tuikov
2023-10-17 15:09 ` [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy Matthew Brost
` (4 subsequent siblings)
7 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
Rather than a global modparam for scheduling policy, move the scheduling
policy to scheduler so user can control each scheduler policy.
v2:
- s/DRM_SCHED_POLICY_MAX/DRM_SCHED_POLICY_COUNT (Luben)
- Only include policy in scheduler (Luben)
v3:
- use a ternary operator as opposed to an if-control (Luben)
- s/DRM_SCHED_POLICY_DEFAULT/DRM_SCHED_POLICY_UNSET/ (Luben)
- s/default_drm_sched_policy/drm_sched_policy_default/ (Luben)
- Update commit message (Boris)
- Fix v3d build (CI)
- s/bad_policies/drm_sched_policy_mismatch/ (Luben)
- Don't update modparam doc (Luben)
v4:
- Fix alignment in msm_ringbuffer_new (Luben / checkpatch)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 +
drivers/gpu/drm/etnaviv/etnaviv_sched.c | 3 ++-
drivers/gpu/drm/lima/lima_sched.c | 3 ++-
drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +-
drivers/gpu/drm/nouveau/nouveau_sched.c | 3 ++-
drivers/gpu/drm/panfrost/panfrost_job.c | 3 ++-
drivers/gpu/drm/scheduler/sched_entity.c | 24 ++++++++++++++++++----
drivers/gpu/drm/scheduler/sched_main.c | 19 ++++++++++++-----
drivers/gpu/drm/v3d/v3d_sched.c | 15 +++++++++-----
include/drm/gpu_scheduler.h | 20 ++++++++++++------
10 files changed, 68 insertions(+), 25 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index b54c4d771104..e4e6f91450a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2283,6 +2283,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
ring->num_hw_submission, 0,
timeout, adev->reset_domain->wq,
ring->sched_score, ring->name,
+ DRM_SCHED_POLICY_UNSET,
adev->dev);
if (r) {
DRM_ERROR("Failed to create scheduler on ring %s.\n",
diff --git a/drivers/gpu/drm/etnaviv/etnaviv_sched.c b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
index 618a804ddc34..15b0e2f1abe5 100644
--- a/drivers/gpu/drm/etnaviv/etnaviv_sched.c
+++ b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
@@ -137,7 +137,8 @@ int etnaviv_sched_init(struct etnaviv_gpu *gpu)
ret = drm_sched_init(&gpu->sched, &etnaviv_sched_ops, NULL,
etnaviv_hw_jobs_limit, etnaviv_job_hang_limit,
msecs_to_jiffies(500), NULL, NULL,
- dev_name(gpu->dev), gpu->dev);
+ dev_name(gpu->dev), DRM_SCHED_POLICY_UNSET,
+ gpu->dev);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/lima/lima_sched.c b/drivers/gpu/drm/lima/lima_sched.c
index 8d858aed0e56..50c2075228aa 100644
--- a/drivers/gpu/drm/lima/lima_sched.c
+++ b/drivers/gpu/drm/lima/lima_sched.c
@@ -491,7 +491,8 @@ int lima_sched_pipe_init(struct lima_sched_pipe *pipe, const char *name)
return drm_sched_init(&pipe->base, &lima_sched_ops, NULL, 1,
lima_job_hang_limit,
msecs_to_jiffies(timeout), NULL,
- NULL, name, pipe->ldev->dev);
+ NULL, name, DRM_SCHED_POLICY_UNSET,
+ pipe->ldev->dev);
}
void lima_sched_pipe_fini(struct lima_sched_pipe *pipe)
diff --git a/drivers/gpu/drm/msm/msm_ringbuffer.c b/drivers/gpu/drm/msm/msm_ringbuffer.c
index 1097f8e93d6b..173ad2f17c50 100644
--- a/drivers/gpu/drm/msm/msm_ringbuffer.c
+++ b/drivers/gpu/drm/msm/msm_ringbuffer.c
@@ -97,7 +97,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
ret = drm_sched_init(&ring->sched, &msm_sched_ops, NULL,
num_hw_submissions, 0, sched_timeout,
NULL, NULL, to_msm_bo(ring->bo)->name,
- gpu->dev->dev);
+ DRM_SCHED_POLICY_UNSET, gpu->dev->dev);
if (ret) {
goto fail;
}
diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c b/drivers/gpu/drm/nouveau/nouveau_sched.c
index 4c959dec42b3..c4e09d2e77f9 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sched.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
@@ -437,7 +437,8 @@ int nouveau_sched_init(struct nouveau_drm *drm)
return drm_sched_init(sched, &nouveau_sched_ops, NULL,
NOUVEAU_SCHED_HW_SUBMISSIONS, 0, job_hang_limit,
- NULL, NULL, "nouveau_sched", drm->dev->dev);
+ NULL, NULL, "nouveau_sched",
+ DRM_SCHED_POLICY_UNSET, drm->dev->dev);
}
void nouveau_sched_fini(struct nouveau_drm *drm)
diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
index 934b7b930c76..95330ff402ba 100644
--- a/drivers/gpu/drm/panfrost/panfrost_job.c
+++ b/drivers/gpu/drm/panfrost/panfrost_job.c
@@ -856,7 +856,8 @@ int panfrost_job_init(struct panfrost_device *pfdev)
nentries, 0,
msecs_to_jiffies(JOB_TIMEOUT_MS),
pfdev->reset.wq,
- NULL, "pan_js", pfdev->dev);
+ NULL, "pan_js", DRM_SCHED_POLICY_UNSET,
+ pfdev->dev);
if (ret) {
dev_err(pfdev->dev, "Failed to create scheduler: %d.", ret);
goto err_sched;
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index a42763e1429d..cf42e2265d64 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -33,6 +33,20 @@
#define to_drm_sched_job(sched_job) \
container_of((sched_job), struct drm_sched_job, queue_node)
+static bool drm_sched_policy_mismatch(struct drm_gpu_scheduler **sched_list,
+ unsigned int num_sched_list)
+{
+ enum drm_sched_policy sched_policy = sched_list[0]->sched_policy;
+ unsigned int i;
+
+ /* All schedule policies must match */
+ for (i = 1; i < num_sched_list; ++i)
+ if (sched_policy != sched_list[i]->sched_policy)
+ return true;
+
+ return false;
+}
+
/**
* drm_sched_entity_init - Init a context entity used by scheduler when
* submit to HW ring.
@@ -62,7 +76,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
unsigned int num_sched_list,
atomic_t *guilty)
{
- if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])))
+ if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])) ||
+ drm_sched_policy_mismatch(sched_list, num_sched_list))
return -EINVAL;
memset(entity, 0, sizeof(struct drm_sched_entity));
@@ -486,7 +501,7 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
* Update the entity's location in the min heap according to
* the timestamp of the next job, if any.
*/
- if (drm_sched_policy == DRM_SCHED_POLICY_FIFO) {
+ if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
struct drm_sched_job *next;
next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
@@ -558,7 +573,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
{
struct drm_sched_entity *entity = sched_job->entity;
- bool first;
+ bool first, fifo = entity->rq->sched->sched_policy ==
+ DRM_SCHED_POLICY_FIFO;
ktime_t submit_ts;
trace_drm_sched_job(sched_job, entity);
@@ -587,7 +603,7 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
drm_sched_rq_add_entity(entity->rq, entity);
spin_unlock(&entity->rq_lock);
- if (drm_sched_policy == DRM_SCHED_POLICY_FIFO)
+ if (fifo)
drm_sched_rq_update_fifo(entity, submit_ts);
drm_sched_wakeup_if_can_queue(entity->rq->sched);
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 8b1d52cff1e9..150e5330f0fa 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -66,14 +66,14 @@
#define to_drm_sched_job(sched_job) \
container_of((sched_job), struct drm_sched_job, queue_node)
-int drm_sched_policy = DRM_SCHED_POLICY_FIFO;
+int drm_sched_policy_default = DRM_SCHED_POLICY_FIFO;
/**
* DOC: sched_policy (int)
* Used to override default entities scheduling policy in a run queue.
*/
MODULE_PARM_DESC(sched_policy, "Specify the scheduling policy for entities on a run-queue, " __stringify(DRM_SCHED_POLICY_RR) " = Round Robin, " __stringify(DRM_SCHED_POLICY_FIFO) " = FIFO (default).");
-module_param_named(sched_policy, drm_sched_policy, int, 0444);
+module_param_named(sched_policy, drm_sched_policy_default, int, 0444);
static __always_inline bool drm_sched_entity_compare_before(struct rb_node *a,
const struct rb_node *b)
@@ -177,7 +177,7 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
if (rq->current_entity == entity)
rq->current_entity = NULL;
- if (drm_sched_policy == DRM_SCHED_POLICY_FIFO)
+ if (rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO)
drm_sched_rq_remove_fifo_locked(entity);
spin_unlock(&rq->lock);
@@ -898,7 +898,7 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
/* Kernel run queue has higher priority than normal run queue*/
for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
- entity = drm_sched_policy == DRM_SCHED_POLICY_FIFO ?
+ entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
if (entity)
@@ -1072,6 +1072,7 @@ static void drm_sched_main(struct work_struct *w)
* used
* @score: optional score atomic shared with other schedulers
* @name: name used for debugging
+ * @sched_policy: schedule policy
* @dev: target &struct device
*
* Return 0 on success, otherwise error code.
@@ -1081,9 +1082,15 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
struct workqueue_struct *submit_wq,
unsigned hw_submission, unsigned hang_limit,
long timeout, struct workqueue_struct *timeout_wq,
- atomic_t *score, const char *name, struct device *dev)
+ atomic_t *score, const char *name,
+ enum drm_sched_policy sched_policy,
+ struct device *dev)
{
int i;
+
+ if (sched_policy >= DRM_SCHED_POLICY_COUNT)
+ return -EINVAL;
+
sched->ops = ops;
sched->hw_submission_limit = hw_submission;
sched->name = name;
@@ -1102,6 +1109,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
sched->hang_limit = hang_limit;
sched->score = score ? score : &sched->_score;
sched->dev = dev;
+ sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
+ drm_sched_policy_default : sched_policy;
for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
drm_sched_rq_init(sched, &sched->sched_rq[i]);
diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
index 38e092ea41e6..dec89c5b8cb1 100644
--- a/drivers/gpu/drm/v3d/v3d_sched.c
+++ b/drivers/gpu/drm/v3d/v3d_sched.c
@@ -391,7 +391,8 @@ v3d_sched_init(struct v3d_dev *v3d)
&v3d_bin_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
- NULL, "v3d_bin", v3d->drm.dev);
+ NULL, "v3d_bin", DRM_SCHED_POLICY_UNSET,
+ v3d->drm.dev);
if (ret)
return ret;
@@ -399,7 +400,8 @@ v3d_sched_init(struct v3d_dev *v3d)
&v3d_render_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
- NULL, "v3d_render", v3d->drm.dev);
+ NULL, "v3d_render", DRM_SCHED_POLICY_UNSET,
+ v3d->drm.dev);
if (ret)
goto fail;
@@ -407,7 +409,8 @@ v3d_sched_init(struct v3d_dev *v3d)
&v3d_tfu_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
- NULL, "v3d_tfu", v3d->drm.dev);
+ NULL, "v3d_tfu", DRM_SCHED_POLICY_UNSET,
+ v3d->drm.dev);
if (ret)
goto fail;
@@ -416,7 +419,8 @@ v3d_sched_init(struct v3d_dev *v3d)
&v3d_csd_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
- NULL, "v3d_csd", v3d->drm.dev);
+ NULL, "v3d_csd", DRM_SCHED_POLICY_UNSET,
+ v3d->drm.dev);
if (ret)
goto fail;
@@ -424,7 +428,8 @@ v3d_sched_init(struct v3d_dev *v3d)
&v3d_cache_clean_sched_ops, NULL,
hw_jobs_limit, job_hang_limit,
msecs_to_jiffies(hang_limit_ms), NULL,
- NULL, "v3d_cache_clean", v3d->drm.dev);
+ NULL, "v3d_cache_clean",
+ DRM_SCHED_POLICY_UNSET, v3d->drm.dev);
if (ret)
goto fail;
}
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index 211bd3cdabdc..320f0a720486 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -72,11 +72,15 @@ enum drm_sched_priority {
DRM_SCHED_PRIORITY_UNSET = -2
};
-/* Used to chose between FIFO and RR jobs scheduling */
-extern int drm_sched_policy;
-
-#define DRM_SCHED_POLICY_RR 0
-#define DRM_SCHED_POLICY_FIFO 1
+/* Used to chose default scheduling policy*/
+extern int default_drm_sched_policy;
+
+enum drm_sched_policy {
+ DRM_SCHED_POLICY_UNSET,
+ DRM_SCHED_POLICY_RR,
+ DRM_SCHED_POLICY_FIFO,
+ DRM_SCHED_POLICY_COUNT,
+};
/**
* struct drm_sched_entity - A wrapper around a job queue (typically
@@ -489,6 +493,7 @@ struct drm_sched_backend_ops {
* guilty and it will no longer be considered for scheduling.
* @score: score to help loadbalancer pick a idle sched
* @_score: score used when the driver doesn't provide one
+ * @sched_policy: Schedule policy for scheduler
* @ready: marks if the underlying HW is ready to work
* @free_guilty: A hit to time out handler to free the guilty job.
* @pause_submit: pause queuing of @work_submit on @submit_wq
@@ -515,6 +520,7 @@ struct drm_gpu_scheduler {
int hang_limit;
atomic_t *score;
atomic_t _score;
+ enum drm_sched_policy sched_policy;
bool ready;
bool free_guilty;
bool pause_submit;
@@ -527,7 +533,9 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
struct workqueue_struct *submit_wq,
uint32_t hw_submission, unsigned hang_limit,
long timeout, struct workqueue_struct *timeout_wq,
- atomic_t *score, const char *name, struct device *dev);
+ atomic_t *score, const char *name,
+ enum drm_sched_policy sched_policy,
+ struct device *dev);
void drm_sched_fini(struct drm_gpu_scheduler *sched);
int drm_sched_job_init(struct drm_sched_job *job,
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
` (2 preceding siblings ...)
2023-10-17 15:09 ` [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-24 3:50 ` Luben Tuikov
2023-10-17 15:09 ` [PATCH v6 5/7] drm/sched: Split free_job into own work item Matthew Brost
` (3 subsequent siblings)
7 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
DRM_SCHED_POLICY_SINGLE_ENTITY creates a 1 to 1 relationship between
scheduler and entity. No priorities or run queue used in this mode.
Intended for devices with firmware schedulers.
v2:
- Drop sched / rq union (Luben)
v3:
- Don't pick entity if stopped in drm_sched_select_entity (Danilo)
v4:
- Rework if statement in drm_sched_entity_init (Luben)
- Update comment for drm_sched_entity_to_scheduler (Luben)
- Reword a few things in DOC comment (Luben)
- Do not check sched policy in for statement (Luben)
v5:
- Fix extra blank lines (Luben / Checkpatch)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
drivers/gpu/drm/scheduler/sched_entity.c | 69 +++++++++++++++----
drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
drivers/gpu/drm/scheduler/sched_main.c | 87 ++++++++++++++++++------
include/drm/gpu_scheduler.h | 8 +++
4 files changed, 130 insertions(+), 36 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index cf42e2265d64..940f63dd6965 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -83,6 +83,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
memset(entity, 0, sizeof(struct drm_sched_entity));
INIT_LIST_HEAD(&entity->list);
entity->rq = NULL;
+ entity->single_sched = NULL;
entity->guilty = guilty;
entity->num_sched_list = num_sched_list;
entity->priority = priority;
@@ -90,8 +91,17 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
RCU_INIT_POINTER(entity->last_scheduled, NULL);
RB_CLEAR_NODE(&entity->rb_tree_node);
- if(num_sched_list)
- entity->rq = &sched_list[0]->sched_rq[entity->priority];
+ if (num_sched_list) {
+ if (sched_list[0]->sched_policy !=
+ DRM_SCHED_POLICY_SINGLE_ENTITY) {
+ entity->rq = &sched_list[0]->sched_rq[entity->priority];
+ } else if (num_sched_list == 1 && !sched_list[0]->single_entity) {
+ sched_list[0]->single_entity = entity;
+ entity->single_sched = sched_list[0];
+ } else {
+ return -EINVAL;
+ }
+ }
init_completion(&entity->entity_idle);
@@ -124,7 +134,8 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
struct drm_gpu_scheduler **sched_list,
unsigned int num_sched_list)
{
- WARN_ON(!num_sched_list || !sched_list);
+ WARN_ON(!num_sched_list || !sched_list ||
+ !!entity->single_sched);
entity->sched_list = sched_list;
entity->num_sched_list = num_sched_list;
@@ -231,13 +242,15 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
{
struct drm_sched_job *job;
struct dma_fence *prev;
+ bool single_entity = !!entity->single_sched;
- if (!entity->rq)
+ if (!entity->rq && !single_entity)
return;
spin_lock(&entity->rq_lock);
entity->stopped = true;
- drm_sched_rq_remove_entity(entity->rq, entity);
+ if (!single_entity)
+ drm_sched_rq_remove_entity(entity->rq, entity);
spin_unlock(&entity->rq_lock);
/* Make sure this entity is not used by the scheduler at the moment */
@@ -259,6 +272,20 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
dma_fence_put(prev);
}
+/**
+ * drm_sched_entity_to_scheduler - Schedule entity to GPU scheduler
+ * @entity: scheduler entity
+ *
+ * Returns GPU scheduler for the entity
+ */
+struct drm_gpu_scheduler *
+drm_sched_entity_to_scheduler(struct drm_sched_entity *entity)
+{
+ bool single_entity = !!entity->single_sched;
+
+ return single_entity ? entity->single_sched : entity->rq->sched;
+}
+
/**
* drm_sched_entity_flush - Flush a context entity
*
@@ -276,11 +303,12 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
struct drm_gpu_scheduler *sched;
struct task_struct *last_user;
long ret = timeout;
+ bool single_entity = !!entity->single_sched;
- if (!entity->rq)
+ if (!entity->rq && !single_entity)
return 0;
- sched = entity->rq->sched;
+ sched = drm_sched_entity_to_scheduler(entity);
/**
* The client will not queue more IBs during this fini, consume existing
* queued IBs or discard them on SIGKILL
@@ -373,7 +401,7 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
container_of(cb, struct drm_sched_entity, cb);
drm_sched_entity_clear_dep(f, cb);
- drm_sched_wakeup_if_can_queue(entity->rq->sched);
+ drm_sched_wakeup_if_can_queue(drm_sched_entity_to_scheduler(entity));
}
/**
@@ -387,6 +415,8 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
enum drm_sched_priority priority)
{
+ WARN_ON(!!entity->single_sched);
+
spin_lock(&entity->rq_lock);
entity->priority = priority;
spin_unlock(&entity->rq_lock);
@@ -399,7 +429,7 @@ EXPORT_SYMBOL(drm_sched_entity_set_priority);
*/
static bool drm_sched_entity_add_dependency_cb(struct drm_sched_entity *entity)
{
- struct drm_gpu_scheduler *sched = entity->rq->sched;
+ struct drm_gpu_scheduler *sched = drm_sched_entity_to_scheduler(entity);
struct dma_fence *fence = entity->dependency;
struct drm_sched_fence *s_fence;
@@ -501,7 +531,8 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
* Update the entity's location in the min heap according to
* the timestamp of the next job, if any.
*/
- if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
+ if (drm_sched_entity_to_scheduler(entity)->sched_policy ==
+ DRM_SCHED_POLICY_FIFO) {
struct drm_sched_job *next;
next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
@@ -524,6 +555,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
struct drm_gpu_scheduler *sched;
struct drm_sched_rq *rq;
+ WARN_ON(!!entity->single_sched);
+
/* single possible engine and already selected */
if (!entity->sched_list)
return;
@@ -573,12 +606,13 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
{
struct drm_sched_entity *entity = sched_job->entity;
- bool first, fifo = entity->rq->sched->sched_policy ==
- DRM_SCHED_POLICY_FIFO;
+ bool single_entity = !!entity->single_sched;
+ bool first;
ktime_t submit_ts;
trace_drm_sched_job(sched_job, entity);
- atomic_inc(entity->rq->sched->score);
+ if (!single_entity)
+ atomic_inc(entity->rq->sched->score);
WRITE_ONCE(entity->last_user, current->group_leader);
/*
@@ -591,6 +625,10 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
/* first job wakes up scheduler */
if (first) {
+ struct drm_gpu_scheduler *sched =
+ drm_sched_entity_to_scheduler(entity);
+ bool fifo = sched->sched_policy == DRM_SCHED_POLICY_FIFO;
+
/* Add the entity to the run queue */
spin_lock(&entity->rq_lock);
if (entity->stopped) {
@@ -600,13 +638,14 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
return;
}
- drm_sched_rq_add_entity(entity->rq, entity);
+ if (!single_entity)
+ drm_sched_rq_add_entity(entity->rq, entity);
spin_unlock(&entity->rq_lock);
if (fifo)
drm_sched_rq_update_fifo(entity, submit_ts);
- drm_sched_wakeup_if_can_queue(entity->rq->sched);
+ drm_sched_wakeup_if_can_queue(sched);
}
}
EXPORT_SYMBOL(drm_sched_entity_push_job);
diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c
index 06cedfe4b486..f6b926f5e188 100644
--- a/drivers/gpu/drm/scheduler/sched_fence.c
+++ b/drivers/gpu/drm/scheduler/sched_fence.c
@@ -225,7 +225,7 @@ void drm_sched_fence_init(struct drm_sched_fence *fence,
{
unsigned seq;
- fence->sched = entity->rq->sched;
+ fence->sched = drm_sched_entity_to_scheduler(entity);
seq = atomic_inc_return(&entity->fence_seq);
dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled,
&fence->lock, entity->fence_context, seq);
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 150e5330f0fa..273e0fbc4eab 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -32,7 +32,8 @@
* backend operations to the scheduler like submitting a job to hardware run queue,
* returning the dependencies of a job etc.
*
- * The organisation of the scheduler is the following:
+ * For scheduling policies DRM_SCHED_POLICY_RR and DRM_SCHED_POLICY_FIFO, the
+ * scheduler organization is:
*
* 1. Each hw run queue has one scheduler
* 2. Each scheduler has multiple run queues with different priorities
@@ -43,7 +44,24 @@
*
* The jobs in a entity are always scheduled in the order that they were pushed.
*
- * Note that once a job was taken from the entities queue and pushed to the
+ * For DRM_SCHED_POLICY_SINGLE_ENTITY, the organization of the scheduler is:
+ *
+ * 1. One to one relationship between scheduler and entity
+ * 2. No priorities implemented per scheduler (single job queue)
+ * 3. No run queues in scheduler rather jobs are directly dequeued from entity
+ * 4. The entity maintains a queue of jobs that will be scheduled to the
+ * hardware
+ *
+ * The jobs in a entity are always scheduled in the order that they were pushed
+ * regardless of scheduling policy. Single-entity scheduling is essentially a
+ * FIFO for jobs.
+ *
+ * A policy of DRM_SCHED_POLICY_RR or DRM_SCHED_POLICY_FIFO is expected to be
+ * used when the kernel-mode driver is scheduling directly to the hardware while
+ * a scheduling policy of DRM_SCHED_POLICY_SINGLE_ENTITY is expected to be used
+ * when there is a firmware scheduler.
+ *
+ * Note that once a job is taken from the entities queue and pushed to the
* hardware, i.e. the pending queue, the entity must not be referenced anymore
* through the jobs entity pointer.
*/
@@ -96,6 +114,8 @@ static inline void drm_sched_rq_remove_fifo_locked(struct drm_sched_entity *enti
void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
{
+ WARN_ON(!!entity->single_sched);
+
/*
* Both locks need to be grabbed, one to protect from entity->rq change
* for entity from within concurrent drm_sched_entity_select_rq and the
@@ -126,6 +146,8 @@ void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
struct drm_sched_rq *rq)
{
+ WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
+
spin_lock_init(&rq->lock);
INIT_LIST_HEAD(&rq->entities);
rq->rb_tree_root = RB_ROOT_CACHED;
@@ -144,6 +166,8 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
struct drm_sched_entity *entity)
{
+ WARN_ON(!!entity->single_sched);
+
if (!list_empty(&entity->list))
return;
@@ -166,6 +190,8 @@ void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
struct drm_sched_entity *entity)
{
+ WARN_ON(!!entity->single_sched);
+
if (list_empty(&entity->list))
return;
@@ -641,7 +667,7 @@ int drm_sched_job_init(struct drm_sched_job *job,
struct drm_sched_entity *entity,
void *owner)
{
- if (!entity->rq)
+ if (!entity->rq && !entity->single_sched)
return -ENOENT;
job->entity = entity;
@@ -674,13 +700,16 @@ void drm_sched_job_arm(struct drm_sched_job *job)
{
struct drm_gpu_scheduler *sched;
struct drm_sched_entity *entity = job->entity;
+ bool single_entity = !!entity->single_sched;
BUG_ON(!entity);
- drm_sched_entity_select_rq(entity);
- sched = entity->rq->sched;
+ if (!single_entity)
+ drm_sched_entity_select_rq(entity);
+ sched = drm_sched_entity_to_scheduler(entity);
job->sched = sched;
- job->s_priority = entity->rq - sched->sched_rq;
+ if (!single_entity)
+ job->s_priority = entity->rq - sched->sched_rq;
job->id = atomic64_inc_return(&sched->job_id_count);
drm_sched_fence_init(job->s_fence, job->entity);
@@ -896,6 +925,14 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
if (!drm_sched_can_queue(sched))
return NULL;
+ if (sched->single_entity) {
+ if (!READ_ONCE(sched->single_entity->stopped) &&
+ drm_sched_entity_is_ready(sched->single_entity))
+ return sched->single_entity;
+
+ return NULL;
+ }
+
/* Kernel run queue has higher priority than normal run queue*/
for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
@@ -1092,6 +1129,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
return -EINVAL;
sched->ops = ops;
+ sched->single_entity = NULL;
sched->hw_submission_limit = hw_submission;
sched->name = name;
if (submit_wq) {
@@ -1111,8 +1149,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
sched->dev = dev;
sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
drm_sched_policy_default : sched_policy;
- for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
- drm_sched_rq_init(sched, &sched->sched_rq[i]);
+ if (sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
+ for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
+ drm_sched_rq_init(sched, &sched->sched_rq[i]);
+ }
init_waitqueue_head(&sched->job_scheduled);
INIT_LIST_HEAD(&sched->pending_list);
@@ -1143,19 +1183,24 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
drm_sched_wqueue_stop(sched);
- for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
- struct drm_sched_rq *rq = &sched->sched_rq[i];
-
- spin_lock(&rq->lock);
- list_for_each_entry(s_entity, &rq->entities, list)
- /*
- * Prevents reinsertion and marks job_queue as idle,
- * it will removed from rq in drm_sched_entity_fini
- * eventually
- */
- s_entity->stopped = true;
- spin_unlock(&rq->lock);
+ if (sched->single_entity) {
+ spin_lock(&sched->single_entity->rq_lock);
+ sched->single_entity->stopped = true;
+ spin_unlock(&sched->single_entity->rq_lock);
+ } else if (sched->sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
+ for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
+ struct drm_sched_rq *rq = &sched->sched_rq[i];
+ spin_lock(&rq->lock);
+ list_for_each_entry(s_entity, &rq->entities, list)
+ /*
+ * Prevents reinsertion and marks job_queue as idle,
+ * it will removed from rq in drm_sched_entity_fini
+ * eventually
+ */
+ s_entity->stopped = true;
+ spin_unlock(&rq->lock);
+ }
}
/* Wakeup everyone stuck in drm_sched_entity_flush for this scheduler */
@@ -1186,6 +1231,8 @@ void drm_sched_increase_karma(struct drm_sched_job *bad)
struct drm_sched_entity *entity;
struct drm_gpu_scheduler *sched = bad->sched;
+ WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
+
/* don't change @bad's karma if it's from KERNEL RQ,
* because sometimes GPU hang would cause kernel jobs (like VM updating jobs)
* corrupt but keep in mind that kernel jobs always considered good.
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index 320f0a720486..e67b53c3546b 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -79,6 +79,7 @@ enum drm_sched_policy {
DRM_SCHED_POLICY_UNSET,
DRM_SCHED_POLICY_RR,
DRM_SCHED_POLICY_FIFO,
+ DRM_SCHED_POLICY_SINGLE_ENTITY,
DRM_SCHED_POLICY_COUNT,
};
@@ -112,6 +113,9 @@ struct drm_sched_entity {
*/
struct drm_sched_rq *rq;
+ /** @single_sched: Single scheduler */
+ struct drm_gpu_scheduler *single_sched;
+
/**
* @sched_list:
*
@@ -473,6 +477,7 @@ struct drm_sched_backend_ops {
* struct drm_gpu_scheduler - scheduler instance-specific data
*
* @ops: backend operations provided by the driver.
+ * @single_entity: Single entity for the scheduler
* @hw_submission_limit: the max size of the hardware queue.
* @timeout: the time after which a job is removed from the scheduler.
* @name: name of the ring for which this scheduler is being used.
@@ -504,6 +509,7 @@ struct drm_sched_backend_ops {
*/
struct drm_gpu_scheduler {
const struct drm_sched_backend_ops *ops;
+ struct drm_sched_entity *single_entity;
uint32_t hw_submission_limit;
long timeout;
const char *name;
@@ -587,6 +593,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
struct drm_gpu_scheduler **sched_list,
unsigned int num_sched_list,
atomic_t *guilty);
+struct drm_gpu_scheduler *
+drm_sched_entity_to_scheduler(struct drm_sched_entity *entity);
long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
void drm_sched_entity_fini(struct drm_sched_entity *entity);
void drm_sched_entity_destroy(struct drm_sched_entity *entity);
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
` (3 preceding siblings ...)
2023-10-17 15:09 ` [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-19 1:25 ` Luben Tuikov
2023-10-23 12:16 ` Boris Brezillon
2023-10-17 15:09 ` [PATCH v6 6/7] drm/sched: Add drm_sched_start_timeout_unlocked helper Matthew Brost
` (2 subsequent siblings)
7 siblings, 2 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
Rather than call free_job and run_job in same work item have a dedicated
work item for each. This aligns with the design and intended use of work
queues.
v2:
- Test for DMA_FENCE_FLAG_TIMESTAMP_BIT before setting
timestamp in free_job() work item (Danilo)
v3:
- Drop forward dec of drm_sched_select_entity (Boris)
- Return in drm_sched_run_job_work if entity NULL (Boris)
v4:
- Replace dequeue with peek and invert logic (Luben)
- Wrap to 100 lines (Luben)
- Update comments for *_queue / *_queue_if_ready functions (Luben)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
drivers/gpu/drm/scheduler/sched_main.c | 287 +++++++++++++++----------
include/drm/gpu_scheduler.h | 8 +-
2 files changed, 178 insertions(+), 117 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 273e0fbc4eab..b1b8d9f96da5 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -213,11 +213,12 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
* drm_sched_rq_select_entity_rr - Select an entity which could provide a job to run
*
* @rq: scheduler run queue to check.
+ * @peek: Just find, don't set to current.
*
* Try to find a ready entity, returns NULL if none found.
*/
static struct drm_sched_entity *
-drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
+drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq, bool peek)
{
struct drm_sched_entity *entity;
@@ -227,8 +228,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
if (entity) {
list_for_each_entry_continue(entity, &rq->entities, list) {
if (drm_sched_entity_is_ready(entity)) {
- rq->current_entity = entity;
- reinit_completion(&entity->entity_idle);
+ if (!peek) {
+ rq->current_entity = entity;
+ reinit_completion(&entity->entity_idle);
+ }
spin_unlock(&rq->lock);
return entity;
}
@@ -238,8 +241,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
list_for_each_entry(entity, &rq->entities, list) {
if (drm_sched_entity_is_ready(entity)) {
- rq->current_entity = entity;
- reinit_completion(&entity->entity_idle);
+ if (!peek) {
+ rq->current_entity = entity;
+ reinit_completion(&entity->entity_idle);
+ }
spin_unlock(&rq->lock);
return entity;
}
@@ -257,11 +262,12 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
* drm_sched_rq_select_entity_fifo - Select an entity which provides a job to run
*
* @rq: scheduler run queue to check.
+ * @peek: Just find, don't set to current.
*
* Find oldest waiting ready entity, returns NULL if none found.
*/
static struct drm_sched_entity *
-drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
+drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq, bool peek)
{
struct rb_node *rb;
@@ -271,8 +277,10 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
entity = rb_entry(rb, struct drm_sched_entity, rb_tree_node);
if (drm_sched_entity_is_ready(entity)) {
- rq->current_entity = entity;
- reinit_completion(&entity->entity_idle);
+ if (!peek) {
+ rq->current_entity = entity;
+ reinit_completion(&entity->entity_idle);
+ }
break;
}
}
@@ -282,13 +290,98 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
}
/**
- * drm_sched_wqueue_enqueue - enqueue scheduler submission
+ * drm_sched_run_job_queue - enqueue run-job work
+ * @sched: scheduler instance
+ */
+static void drm_sched_run_job_queue(struct drm_gpu_scheduler *sched)
+{
+ if (!READ_ONCE(sched->pause_submit))
+ queue_work(sched->submit_wq, &sched->work_run_job);
+}
+
+/**
+ * drm_sched_can_queue -- Can we queue more to the hardware?
+ * @sched: scheduler instance
+ *
+ * Return true if we can push more jobs to the hw, otherwise false.
+ */
+static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
+{
+ return atomic_read(&sched->hw_rq_count) <
+ sched->hw_submission_limit;
+}
+
+/**
+ * drm_sched_select_entity - Select next entity to process
+ *
+ * @sched: scheduler instance
+ * @peek: Just find, don't set to current.
+ *
+ * Returns the entity to process or NULL if none are found.
+ */
+static struct drm_sched_entity *
+drm_sched_select_entity(struct drm_gpu_scheduler *sched, bool peek)
+{
+ struct drm_sched_entity *entity;
+ int i;
+
+ if (!drm_sched_can_queue(sched))
+ return NULL;
+
+ if (sched->single_entity) {
+ if (!READ_ONCE(sched->single_entity->stopped) &&
+ drm_sched_entity_is_ready(sched->single_entity))
+ return sched->single_entity;
+
+ return NULL;
+ }
+
+ /* Kernel run queue has higher priority than normal run queue*/
+ for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
+ entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
+ drm_sched_rq_select_entity_fifo(&sched->sched_rq[i], peek) :
+ drm_sched_rq_select_entity_rr(&sched->sched_rq[i], peek);
+ if (entity)
+ break;
+ }
+
+ return entity;
+}
+
+/**
+ * drm_sched_run_job_queue_if_ready - enqueue run-job work if ready
+ * @sched: scheduler instance
+ */
+static void drm_sched_run_job_queue_if_ready(struct drm_gpu_scheduler *sched)
+{
+ if (drm_sched_select_entity(sched, false))
+ drm_sched_run_job_queue(sched);
+}
+
+/**
+ * drm_sched_free_job_queue - enqueue free-job work
* @sched: scheduler instance
*/
-static void drm_sched_wqueue_enqueue(struct drm_gpu_scheduler *sched)
+static void drm_sched_free_job_queue(struct drm_gpu_scheduler *sched)
{
if (!READ_ONCE(sched->pause_submit))
- queue_work(sched->submit_wq, &sched->work_submit);
+ queue_work(sched->submit_wq, &sched->work_free_job);
+}
+
+/**
+ * drm_sched_free_job_queue_if_ready - enqueue free-job work if ready
+ * @sched: scheduler instance
+ */
+static void drm_sched_free_job_queue_if_ready(struct drm_gpu_scheduler *sched)
+{
+ struct drm_sched_job *job;
+
+ spin_lock(&sched->job_list_lock);
+ job = list_first_entry_or_null(&sched->pending_list,
+ struct drm_sched_job, list);
+ if (job && dma_fence_is_signaled(&job->s_fence->finished))
+ drm_sched_free_job_queue(sched);
+ spin_unlock(&sched->job_list_lock);
}
/**
@@ -310,7 +403,7 @@ static void drm_sched_job_done(struct drm_sched_job *s_job, int result)
dma_fence_get(&s_fence->finished);
drm_sched_fence_finished(s_fence, result);
dma_fence_put(&s_fence->finished);
- drm_sched_wqueue_enqueue(sched);
+ drm_sched_free_job_queue(sched);
}
/**
@@ -576,7 +669,7 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
drm_sched_job_done(s_job, fence->error);
else if (r)
DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
- r);
+ r);
} else
drm_sched_job_done(s_job, -ECANCELED);
}
@@ -885,18 +978,6 @@ void drm_sched_job_cleanup(struct drm_sched_job *job)
}
EXPORT_SYMBOL(drm_sched_job_cleanup);
-/**
- * drm_sched_can_queue -- Can we queue more to the hardware?
- * @sched: scheduler instance
- *
- * Return true if we can push more jobs to the hw, otherwise false.
- */
-static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
-{
- return atomic_read(&sched->hw_rq_count) <
- sched->hw_submission_limit;
-}
-
/**
* drm_sched_wakeup_if_can_queue - Wake up the scheduler
* @sched: scheduler instance
@@ -906,43 +987,7 @@ static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched)
{
if (drm_sched_can_queue(sched))
- drm_sched_wqueue_enqueue(sched);
-}
-
-/**
- * drm_sched_select_entity - Select next entity to process
- *
- * @sched: scheduler instance
- *
- * Returns the entity to process or NULL if none are found.
- */
-static struct drm_sched_entity *
-drm_sched_select_entity(struct drm_gpu_scheduler *sched)
-{
- struct drm_sched_entity *entity;
- int i;
-
- if (!drm_sched_can_queue(sched))
- return NULL;
-
- if (sched->single_entity) {
- if (!READ_ONCE(sched->single_entity->stopped) &&
- drm_sched_entity_is_ready(sched->single_entity))
- return sched->single_entity;
-
- return NULL;
- }
-
- /* Kernel run queue has higher priority than normal run queue*/
- for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
- entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
- drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
- drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
- if (entity)
- break;
- }
-
- return entity;
+ drm_sched_run_job_queue(sched);
}
/**
@@ -974,8 +1019,10 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
typeof(*next), list);
if (next) {
- next->s_fence->scheduled.timestamp =
- dma_fence_timestamp(&job->s_fence->finished);
+ if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
+ &next->s_fence->scheduled.flags))
+ next->s_fence->scheduled.timestamp =
+ dma_fence_timestamp(&job->s_fence->finished);
/* start TO timer for next job */
drm_sched_start_timeout(sched);
}
@@ -1025,74 +1072,83 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list,
EXPORT_SYMBOL(drm_sched_pick_best);
/**
- * drm_sched_main - main scheduler thread
+ * drm_sched_free_job_work - worker to call free_job
*
- * @param: scheduler instance
+ * @w: free job work
*/
-static void drm_sched_main(struct work_struct *w)
+static void drm_sched_free_job_work(struct work_struct *w)
{
struct drm_gpu_scheduler *sched =
- container_of(w, struct drm_gpu_scheduler, work_submit);
- struct drm_sched_entity *entity;
+ container_of(w, struct drm_gpu_scheduler, work_free_job);
struct drm_sched_job *cleanup_job;
- int r;
if (READ_ONCE(sched->pause_submit))
return;
cleanup_job = drm_sched_get_cleanup_job(sched);
- entity = drm_sched_select_entity(sched);
-
- if (!entity && !cleanup_job)
- return; /* No more work */
-
- if (cleanup_job)
+ if (cleanup_job) {
sched->ops->free_job(cleanup_job);
- if (entity) {
- struct dma_fence *fence;
- struct drm_sched_fence *s_fence;
- struct drm_sched_job *sched_job;
-
- sched_job = drm_sched_entity_pop_job(entity);
- if (!sched_job) {
- complete_all(&entity->entity_idle);
- if (!cleanup_job)
- return; /* No more work */
- goto again;
- }
+ drm_sched_free_job_queue_if_ready(sched);
+ drm_sched_run_job_queue_if_ready(sched);
+ }
+}
- s_fence = sched_job->s_fence;
+/**
+ * drm_sched_run_job_work - worker to call run_job
+ *
+ * @w: run job work
+ */
+static void drm_sched_run_job_work(struct work_struct *w)
+{
+ struct drm_gpu_scheduler *sched =
+ container_of(w, struct drm_gpu_scheduler, work_run_job);
+ struct drm_sched_entity *entity;
+ struct dma_fence *fence;
+ struct drm_sched_fence *s_fence;
+ struct drm_sched_job *sched_job;
+ int r;
- atomic_inc(&sched->hw_rq_count);
- drm_sched_job_begin(sched_job);
+ if (READ_ONCE(sched->pause_submit))
+ return;
+
+ entity = drm_sched_select_entity(sched, true);
+ if (!entity)
+ return;
- trace_drm_run_job(sched_job, entity);
- fence = sched->ops->run_job(sched_job);
+ sched_job = drm_sched_entity_pop_job(entity);
+ if (!sched_job) {
complete_all(&entity->entity_idle);
- drm_sched_fence_scheduled(s_fence, fence);
+ return; /* No more work */
+ }
- if (!IS_ERR_OR_NULL(fence)) {
- /* Drop for original kref_init of the fence */
- dma_fence_put(fence);
+ s_fence = sched_job->s_fence;
- r = dma_fence_add_callback(fence, &sched_job->cb,
- drm_sched_job_done_cb);
- if (r == -ENOENT)
- drm_sched_job_done(sched_job, fence->error);
- else if (r)
- DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
- r);
- } else {
- drm_sched_job_done(sched_job, IS_ERR(fence) ?
- PTR_ERR(fence) : 0);
- }
+ atomic_inc(&sched->hw_rq_count);
+ drm_sched_job_begin(sched_job);
- wake_up(&sched->job_scheduled);
+ trace_drm_run_job(sched_job, entity);
+ fence = sched->ops->run_job(sched_job);
+ complete_all(&entity->entity_idle);
+ drm_sched_fence_scheduled(s_fence, fence);
+
+ if (!IS_ERR_OR_NULL(fence)) {
+ /* Drop for original kref_init of the fence */
+ dma_fence_put(fence);
+
+ r = dma_fence_add_callback(fence, &sched_job->cb,
+ drm_sched_job_done_cb);
+ if (r == -ENOENT)
+ drm_sched_job_done(sched_job, fence->error);
+ else if (r)
+ DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
+ } else {
+ drm_sched_job_done(sched_job, IS_ERR(fence) ?
+ PTR_ERR(fence) : 0);
}
-again:
- drm_sched_wqueue_enqueue(sched);
+ wake_up(&sched->job_scheduled);
+ drm_sched_run_job_queue_if_ready(sched);
}
/**
@@ -1159,7 +1215,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
spin_lock_init(&sched->job_list_lock);
atomic_set(&sched->hw_rq_count, 0);
INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
- INIT_WORK(&sched->work_submit, drm_sched_main);
+ INIT_WORK(&sched->work_run_job, drm_sched_run_job_work);
+ INIT_WORK(&sched->work_free_job, drm_sched_free_job_work);
atomic_set(&sched->_score, 0);
atomic64_set(&sched->job_id_count, 0);
sched->pause_submit = false;
@@ -1282,7 +1339,8 @@ EXPORT_SYMBOL(drm_sched_wqueue_ready);
void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
{
WRITE_ONCE(sched->pause_submit, true);
- cancel_work_sync(&sched->work_submit);
+ cancel_work_sync(&sched->work_run_job);
+ cancel_work_sync(&sched->work_free_job);
}
EXPORT_SYMBOL(drm_sched_wqueue_stop);
@@ -1294,6 +1352,7 @@ EXPORT_SYMBOL(drm_sched_wqueue_stop);
void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
{
WRITE_ONCE(sched->pause_submit, false);
- queue_work(sched->submit_wq, &sched->work_submit);
+ queue_work(sched->submit_wq, &sched->work_run_job);
+ queue_work(sched->submit_wq, &sched->work_free_job);
}
EXPORT_SYMBOL(drm_sched_wqueue_start);
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index e67b53c3546b..625ffe040de3 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -487,9 +487,10 @@ struct drm_sched_backend_ops {
* finished.
* @hw_rq_count: the number of jobs currently in the hardware queue.
* @job_id_count: used to assign unique id to the each job.
- * @submit_wq: workqueue used to queue @work_submit
+ * @submit_wq: workqueue used to queue @work_run_job and @work_free_job
* @timeout_wq: workqueue used to queue @work_tdr
- * @work_submit: schedules jobs and cleans up entities
+ * @work_run_job: schedules jobs
+ * @work_free_job: cleans up jobs
* @work_tdr: schedules a delayed call to @drm_sched_job_timedout after the
* timeout interval is over.
* @pending_list: the list of jobs which are currently in the job queue.
@@ -519,7 +520,8 @@ struct drm_gpu_scheduler {
atomic64_t job_id_count;
struct workqueue_struct *submit_wq;
struct workqueue_struct *timeout_wq;
- struct work_struct work_submit;
+ struct work_struct work_run_job;
+ struct work_struct work_free_job;
struct delayed_work work_tdr;
struct list_head pending_list;
spinlock_t job_list_lock;
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 6/7] drm/sched: Add drm_sched_start_timeout_unlocked helper
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
` (4 preceding siblings ...)
2023-10-17 15:09 ` [PATCH v6 5/7] drm/sched: Split free_job into own work item Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-17 15:09 ` [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately Matthew Brost
2023-10-18 8:24 ` [PATCH v6 0/7] DRM scheduler changes for Xe Daniel Vetter
7 siblings, 0 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
Also add a lockdep assert to drm_sched_start_timeout.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
---
drivers/gpu/drm/scheduler/sched_main.c | 23 +++++++++++++----------
1 file changed, 13 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index b1b8d9f96da5..ff2cfad54342 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -427,11 +427,20 @@ static void drm_sched_job_done_cb(struct dma_fence *f, struct dma_fence_cb *cb)
*/
static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched)
{
+ lockdep_assert_held(&sched->job_list_lock);
+
if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
!list_empty(&sched->pending_list))
queue_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout);
}
+static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched)
+{
+ spin_lock(&sched->job_list_lock);
+ drm_sched_start_timeout(sched);
+ spin_unlock(&sched->job_list_lock);
+}
+
/**
* drm_sched_fault - immediately start timeout handler
*
@@ -544,11 +553,8 @@ static void drm_sched_job_timedout(struct work_struct *work)
spin_unlock(&sched->job_list_lock);
}
- if (status != DRM_GPU_SCHED_STAT_ENODEV) {
- spin_lock(&sched->job_list_lock);
- drm_sched_start_timeout(sched);
- spin_unlock(&sched->job_list_lock);
- }
+ if (status != DRM_GPU_SCHED_STAT_ENODEV)
+ drm_sched_start_timeout_unlocked(sched);
}
/**
@@ -674,11 +680,8 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
drm_sched_job_done(s_job, -ECANCELED);
}
- if (full_recovery) {
- spin_lock(&sched->job_list_lock);
- drm_sched_start_timeout(sched);
- spin_unlock(&sched->job_list_lock);
- }
+ if (full_recovery)
+ drm_sched_start_timeout_unlocked(sched);
drm_sched_wqueue_start(sched);
}
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
` (5 preceding siblings ...)
2023-10-17 15:09 ` [PATCH v6 6/7] drm/sched: Add drm_sched_start_timeout_unlocked helper Matthew Brost
@ 2023-10-17 15:09 ` Matthew Brost
2023-10-18 16:18 ` Luben Tuikov
2023-10-18 8:24 ` [PATCH v6 0/7] DRM scheduler changes for Xe Daniel Vetter
7 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-17 15:09 UTC (permalink / raw)
To: dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, Matthew Brost, lina, sarah.walker,
ketil.johnsen, Liviu.Dudau, mcanal, luben.tuikov, dakr,
donald.robson, boris.brezillon, christian.koenig, faith.ekstrand
Add a helper whereby a driver can invoke TDR immediately.
v2:
- Drop timeout args, rename function, use mod delayed work (Luben)
v3:
- s/XE/Xe (Luben)
- present tense in commit message (Luben)
- Adjust comment for drm_sched_tdr_queue_imm (Luben)
v4:
- Adjust commit message (Luben)
Cc: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
drivers/gpu/drm/scheduler/sched_main.c | 18 +++++++++++++++++-
include/drm/gpu_scheduler.h | 1 +
2 files changed, 18 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index ff2cfad54342..563a4f0f6956 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -431,7 +431,7 @@ static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched)
if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
!list_empty(&sched->pending_list))
- queue_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout);
+ mod_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout);
}
static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched)
@@ -441,6 +441,22 @@ static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched)
spin_unlock(&sched->job_list_lock);
}
+/**
+ * drm_sched_tdr_queue_imm: - immediately start job timeout handler
+ *
+ * @sched: scheduler for which the timeout handling should be started.
+ *
+ * Start timeout handling immediately for the named scheduler.
+ */
+void drm_sched_tdr_queue_imm(struct drm_gpu_scheduler *sched)
+{
+ spin_lock(&sched->job_list_lock);
+ sched->timeout = 0;
+ drm_sched_start_timeout(sched);
+ spin_unlock(&sched->job_list_lock);
+}
+EXPORT_SYMBOL(drm_sched_tdr_queue_imm);
+
/**
* drm_sched_fault - immediately start timeout handler
*
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index 625ffe040de3..998b32b8d212 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -568,6 +568,7 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
struct drm_gpu_scheduler **sched_list,
unsigned int num_sched_list);
+void drm_sched_tdr_queue_imm(struct drm_gpu_scheduler *sched);
void drm_sched_job_cleanup(struct drm_sched_job *job);
void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched);
bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched);
--
2.34.1
^ permalink raw reply related [flat|nested] 22+ messages in thread
* Re: [PATCH v6 0/7] DRM scheduler changes for Xe
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
` (6 preceding siblings ...)
2023-10-17 15:09 ` [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately Matthew Brost
@ 2023-10-18 8:24 ` Daniel Vetter
7 siblings, 0 replies; 22+ messages in thread
From: Daniel Vetter @ 2023-10-18 8:24 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, lina, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, dri-devel, christian.koenig, luben.tuikov,
dakr, donald.robson, boris.brezillon, intel-xe, faith.ekstrand
On Tue, Oct 17, 2023 at 08:09:51AM -0700, Matthew Brost wrote:
> As a prerequisite to merging the new Intel Xe DRM driver [1] [2], we
> have been asked to merge our common DRM scheduler patches first.
>
> This a continuation of a RFC [3] with all comments addressed, ready for
> a full review, and hopefully in state which can merged in the near
> future. More details of this series can found in the cover letter of the
> RFC [3].
>
> These changes have been tested with the Xe driver. Based on drm-tip branch.
>
> A follow up series will be posted to address some of dakr requets for
> kernel doc changes.
I think it'd be really good to include the doc improved in this series.
drm/sched is one of the least documented parts of drm, and has some of the
trickiest sharp corners. We really should be documenting these as
consensus around how things work and as features get added.
I think it would also be really awesome if someone goes through all the
interfaces and adds docs for what we already have. That's probably for
later, but I think really needed to make sure the docs are somewhat
consistent (things like the api includ scaffolding, or having the code
related pieces all pushed to the .c files as DOC: comments and all these
things otherwise tend to end up a bit more chaotic than needed).
Cheers, Sima
>
> v2:
> - Break run job, free job, and process message in own work items
> - This might break other drivers as run job and free job now can run in
> parallel, can fix up if needed
>
> v3:
> - Include missing patch 'drm/sched: Add drm_sched_submit_* helpers'
> - Fix issue with setting timestamp to early
> - Don't dequeue jobs for single entity after calling entity fini
> - Flush pending jobs on entity fini
> - Add documentation for entity teardown
> - Add Matthew Brost to maintainers of DRM scheduler
>
> v4:
> - Drop message interface
> - Drop 'Flush pending jobs on entity fini'
> - Drop 'Add documentation for entity teardown'
> - Address all feedback
>
> v5:
> - Address Luben's feedback
> - Drop starting TDR after calling run_job()
> - Drop adding Matthew Brost to maintainers of DRM scheduler
>
> v6:
> - Address Luben's feedback
> - Include base commit
>
> Matt
>
> [1] https://gitlab.freedesktop.org/drm/xe/kernel
> [2] https://patchwork.freedesktop.org/series/112188/
> [3] https://patchwork.freedesktop.org/series/116055/
>
>
> Matthew Brost (7):
> drm/sched: Add drm_sched_wqueue_* helpers
> drm/sched: Convert drm scheduler to use a work queue rather than
> kthread
> drm/sched: Move schedule policy to scheduler
> drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
> drm/sched: Split free_job into own work item
> drm/sched: Add drm_sched_start_timeout_unlocked helper
> drm/sched: Add a helper to queue TDR immediately
>
> .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 15 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +-
> drivers/gpu/drm/etnaviv/etnaviv_sched.c | 5 +-
> drivers/gpu/drm/lima/lima_sched.c | 5 +-
> drivers/gpu/drm/msm/adreno/adreno_device.c | 6 +-
> drivers/gpu/drm/msm/msm_ringbuffer.c | 7 +-
> drivers/gpu/drm/nouveau/nouveau_sched.c | 5 +-
> drivers/gpu/drm/panfrost/panfrost_job.c | 5 +-
> drivers/gpu/drm/scheduler/sched_entity.c | 85 ++-
> drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
> drivers/gpu/drm/scheduler/sched_main.c | 507 ++++++++++++------
> drivers/gpu/drm/v3d/v3d_sched.c | 25 +-
> include/drm/gpu_scheduler.h | 48 +-
> 14 files changed, 498 insertions(+), 234 deletions(-)
>
>
> base-commit: 201c8a7bd1f3f415920a2df4b8a8817e973f42fe
> --
> 2.34.1
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately
2023-10-17 15:09 ` [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately Matthew Brost
@ 2023-10-18 16:18 ` Luben Tuikov
0 siblings, 0 replies; 22+ messages in thread
From: Luben Tuikov @ 2023-10-18 16:18 UTC (permalink / raw)
To: Matthew Brost, dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, boris.brezillon, dakr, donald.robson, lina,
christian.koenig, faith.ekstrand
On 2023-10-17 11:09, Matthew Brost wrote:
> Add a helper whereby a driver can invoke TDR immediately.
>
> v2:
> - Drop timeout args, rename function, use mod delayed work (Luben)
> v3:
> - s/XE/Xe (Luben)
> - present tense in commit message (Luben)
> - Adjust comment for drm_sched_tdr_queue_imm (Luben)
> v4:
> - Adjust commit message (Luben)
>
> Cc: Luben Tuikov <luben.tuikov@amd.com>
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Yeah, this patch is very good now--thanks for updating it.
Regards,
Luben
> ---
> drivers/gpu/drm/scheduler/sched_main.c | 18 +++++++++++++++++-
> include/drm/gpu_scheduler.h | 1 +
> 2 files changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index ff2cfad54342..563a4f0f6956 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -431,7 +431,7 @@ static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched)
>
> if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
> !list_empty(&sched->pending_list))
> - queue_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout);
> + mod_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout);
> }
>
> static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched)
> @@ -441,6 +441,22 @@ static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched)
> spin_unlock(&sched->job_list_lock);
> }
>
> +/**
> + * drm_sched_tdr_queue_imm: - immediately start job timeout handler
> + *
> + * @sched: scheduler for which the timeout handling should be started.
> + *
> + * Start timeout handling immediately for the named scheduler.
> + */
> +void drm_sched_tdr_queue_imm(struct drm_gpu_scheduler *sched)
> +{
> + spin_lock(&sched->job_list_lock);
> + sched->timeout = 0;
> + drm_sched_start_timeout(sched);
> + spin_unlock(&sched->job_list_lock);
> +}
> +EXPORT_SYMBOL(drm_sched_tdr_queue_imm);
> +
> /**
> * drm_sched_fault - immediately start timeout handler
> *
> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> index 625ffe040de3..998b32b8d212 100644
> --- a/include/drm/gpu_scheduler.h
> +++ b/include/drm/gpu_scheduler.h
> @@ -568,6 +568,7 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
> struct drm_gpu_scheduler **sched_list,
> unsigned int num_sched_list);
>
> +void drm_sched_tdr_queue_imm(struct drm_gpu_scheduler *sched);
> void drm_sched_job_cleanup(struct drm_sched_job *job);
> void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched);
> bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched);
--
Regards,
Luben
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-17 15:09 ` [PATCH v6 5/7] drm/sched: Split free_job into own work item Matthew Brost
@ 2023-10-19 1:25 ` Luben Tuikov
2023-10-19 16:55 ` Matthew Brost
2023-10-23 12:16 ` Boris Brezillon
1 sibling, 1 reply; 22+ messages in thread
From: Luben Tuikov @ 2023-10-19 1:25 UTC (permalink / raw)
To: Matthew Brost, dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, boris.brezillon, dakr, donald.robson, lina,
christian.koenig, faith.ekstrand
Hi,
On 2023-10-17 11:09, Matthew Brost wrote:
> Rather than call free_job and run_job in same work item have a dedicated
> work item for each. This aligns with the design and intended use of work
> queues.
>
> v2:
> - Test for DMA_FENCE_FLAG_TIMESTAMP_BIT before setting
> timestamp in free_job() work item (Danilo)
> v3:
> - Drop forward dec of drm_sched_select_entity (Boris)
> - Return in drm_sched_run_job_work if entity NULL (Boris)
> v4:
> - Replace dequeue with peek and invert logic (Luben)
> - Wrap to 100 lines (Luben)
> - Update comments for *_queue / *_queue_if_ready functions (Luben)
>
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> ---
> drivers/gpu/drm/scheduler/sched_main.c | 287 +++++++++++++++----------
> include/drm/gpu_scheduler.h | 8 +-
> 2 files changed, 178 insertions(+), 117 deletions(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 273e0fbc4eab..b1b8d9f96da5 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -213,11 +213,12 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
> * drm_sched_rq_select_entity_rr - Select an entity which could provide a job to run
> *
> * @rq: scheduler run queue to check.
> + * @peek: Just find, don't set to current.
The "peek" rename is good--thanks!
> *
> * Try to find a ready entity, returns NULL if none found.
> */
> static struct drm_sched_entity *
> -drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> +drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq, bool peek)
> {
> struct drm_sched_entity *entity;
>
> @@ -227,8 +228,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> if (entity) {
> list_for_each_entry_continue(entity, &rq->entities, list) {
> if (drm_sched_entity_is_ready(entity)) {
> - rq->current_entity = entity;
> - reinit_completion(&entity->entity_idle);
> + if (!peek) {
> + rq->current_entity = entity;
> + reinit_completion(&entity->entity_idle);
> + }
> spin_unlock(&rq->lock);
> return entity;
> }
> @@ -238,8 +241,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> list_for_each_entry(entity, &rq->entities, list) {
>
> if (drm_sched_entity_is_ready(entity)) {
> - rq->current_entity = entity;
> - reinit_completion(&entity->entity_idle);
> + if (!peek) {
> + rq->current_entity = entity;
> + reinit_completion(&entity->entity_idle);
> + }
> spin_unlock(&rq->lock);
> return entity;
> }
> @@ -257,11 +262,12 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> * drm_sched_rq_select_entity_fifo - Select an entity which provides a job to run
> *
> * @rq: scheduler run queue to check.
> + * @peek: Just find, don't set to current.
> *
> * Find oldest waiting ready entity, returns NULL if none found.
> */
> static struct drm_sched_entity *
> -drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
> +drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq, bool peek)
> {
> struct rb_node *rb;
>
> @@ -271,8 +277,10 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
>
> entity = rb_entry(rb, struct drm_sched_entity, rb_tree_node);
> if (drm_sched_entity_is_ready(entity)) {
> - rq->current_entity = entity;
> - reinit_completion(&entity->entity_idle);
> + if (!peek) {
> + rq->current_entity = entity;
> + reinit_completion(&entity->entity_idle);
> + }
> break;
> }
> }
> @@ -282,13 +290,98 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
> }
>
> /**
> - * drm_sched_wqueue_enqueue - enqueue scheduler submission
> + * drm_sched_run_job_queue - enqueue run-job work
Hmm... so this removes the change from patch 1 in this series, and as such
obviates patch 1.
Do you want to go with "run_job_queue" and the names introduced here?
In this case, we can have that in patch 1 instead, and this patch
would only split the "free job" into its own work item.
> + * @sched: scheduler instance
> + */
> +static void drm_sched_run_job_queue(struct drm_gpu_scheduler *sched)
> +{
> + if (!READ_ONCE(sched->pause_submit))
> + queue_work(sched->submit_wq, &sched->work_run_job);
> +}
> +
> +/**
> + * drm_sched_can_queue -- Can we queue more to the hardware?
> + * @sched: scheduler instance
> + *
> + * Return true if we can push more jobs to the hw, otherwise false.
> + */
> +static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> +{
> + return atomic_read(&sched->hw_rq_count) <
> + sched->hw_submission_limit;
> +}
> +
> +/**
> + * drm_sched_select_entity - Select next entity to process
> + *
> + * @sched: scheduler instance
> + * @peek: Just find, don't set to current.
> + *
> + * Returns the entity to process or NULL if none are found.
> + */
> +static struct drm_sched_entity *
> +drm_sched_select_entity(struct drm_gpu_scheduler *sched, bool peek)
> +{
> + struct drm_sched_entity *entity;
> + int i;
> +
> + if (!drm_sched_can_queue(sched))
> + return NULL;
> +
> + if (sched->single_entity) {
> + if (!READ_ONCE(sched->single_entity->stopped) &&
> + drm_sched_entity_is_ready(sched->single_entity))
> + return sched->single_entity;
> +
> + return NULL;
> + }
> +
> + /* Kernel run queue has higher priority than normal run queue*/
> + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> + entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> + drm_sched_rq_select_entity_fifo(&sched->sched_rq[i], peek) :
> + drm_sched_rq_select_entity_rr(&sched->sched_rq[i], peek);
> + if (entity)
> + break;
> + }
> +
> + return entity;
> +}
Can you shed some light on why you need the "peek" capability, i.e. to just
see the entity without actually arming it?
In a single-entity scheduling, surely peeking at the single entity of
a scheduler, can be done easier.
I'm asking as none of these functions were intended as a peek-only/-capable
solution, and if you need it, this shows that the infrastructure lacks
something which you need.
It's not designed for "peek", as you can see above: it gets passed-through
the function stack until it reaches core functionality to be used in logic.
(It just makes the code too complicated with too many special cases, "if's"
and if we keep doing this, it would become very hard to understand,
maintain, and contribute to in a few years.)
> +
> +/**
> + * drm_sched_run_job_queue_if_ready - enqueue run-job work if ready
> + * @sched: scheduler instance
> + */
"if ready" here makes perfect sense, but in the "free job" case,
it should be "if done". See below.
> +static void drm_sched_run_job_queue_if_ready(struct drm_gpu_scheduler *sched)
> +{
> + if (drm_sched_select_entity(sched, false))
> + drm_sched_run_job_queue(sched);
> +}
> +
> +/**
> + * drm_sched_free_job_queue - enqueue free-job work
> * @sched: scheduler instance
> */
> -static void drm_sched_wqueue_enqueue(struct drm_gpu_scheduler *sched)
> +static void drm_sched_free_job_queue(struct drm_gpu_scheduler *sched)
> {
> if (!READ_ONCE(sched->pause_submit))
> - queue_work(sched->submit_wq, &sched->work_submit);
> + queue_work(sched->submit_wq, &sched->work_free_job);
> +}
> +
> +/**
> + * drm_sched_free_job_queue_if_ready - enqueue free-job work if ready
This should be "if done". Please change this to "if done".
> + * @sched: scheduler instance
> + */
> +static void drm_sched_free_job_queue_if_ready(struct drm_gpu_scheduler *sched)
This should be "if_done". Please change this to "if_done".
A "job" is _done_ when it's on the pending list and its fence has been signalled,
and we free a job only when it's done, not "ready".
> +{
> + struct drm_sched_job *job;
> +
> + spin_lock(&sched->job_list_lock);
> + job = list_first_entry_or_null(&sched->pending_list,
> + struct drm_sched_job, list);
> + if (job && dma_fence_is_signaled(&job->s_fence->finished))
> + drm_sched_free_job_queue(sched);
> + spin_unlock(&sched->job_list_lock);
> }
>
> /**
> @@ -310,7 +403,7 @@ static void drm_sched_job_done(struct drm_sched_job *s_job, int result)
> dma_fence_get(&s_fence->finished);
> drm_sched_fence_finished(s_fence, result);
> dma_fence_put(&s_fence->finished);
> - drm_sched_wqueue_enqueue(sched);
> + drm_sched_free_job_queue(sched);
> }
>
> /**
> @@ -576,7 +669,7 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
> drm_sched_job_done(s_job, fence->error);
> else if (r)
> DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> - r);
> + r);
> } else
> drm_sched_job_done(s_job, -ECANCELED);
> }
> @@ -885,18 +978,6 @@ void drm_sched_job_cleanup(struct drm_sched_job *job)
> }
> EXPORT_SYMBOL(drm_sched_job_cleanup);
>
> -/**
> - * drm_sched_can_queue -- Can we queue more to the hardware?
> - * @sched: scheduler instance
> - *
> - * Return true if we can push more jobs to the hw, otherwise false.
> - */
> -static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> -{
> - return atomic_read(&sched->hw_rq_count) <
> - sched->hw_submission_limit;
> -}
> -
> /**
> * drm_sched_wakeup_if_can_queue - Wake up the scheduler
> * @sched: scheduler instance
> @@ -906,43 +987,7 @@ static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched)
> {
> if (drm_sched_can_queue(sched))
> - drm_sched_wqueue_enqueue(sched);
> -}
> -
> -/**
> - * drm_sched_select_entity - Select next entity to process
> - *
> - * @sched: scheduler instance
> - *
> - * Returns the entity to process or NULL if none are found.
> - */
> -static struct drm_sched_entity *
> -drm_sched_select_entity(struct drm_gpu_scheduler *sched)
> -{
> - struct drm_sched_entity *entity;
> - int i;
> -
> - if (!drm_sched_can_queue(sched))
> - return NULL;
> -
> - if (sched->single_entity) {
> - if (!READ_ONCE(sched->single_entity->stopped) &&
> - drm_sched_entity_is_ready(sched->single_entity))
> - return sched->single_entity;
> -
> - return NULL;
> - }
> -
> - /* Kernel run queue has higher priority than normal run queue*/
> - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> - entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> - drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
> - drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
> - if (entity)
> - break;
> - }
> -
> - return entity;
> + drm_sched_run_job_queue(sched);
> }
>
> /**
> @@ -974,8 +1019,10 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
> typeof(*next), list);
>
> if (next) {
> - next->s_fence->scheduled.timestamp =
> - dma_fence_timestamp(&job->s_fence->finished);
> + if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
> + &next->s_fence->scheduled.flags))
> + next->s_fence->scheduled.timestamp =
> + dma_fence_timestamp(&job->s_fence->finished);
> /* start TO timer for next job */
> drm_sched_start_timeout(sched);
> }
> @@ -1025,74 +1072,83 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list,
> EXPORT_SYMBOL(drm_sched_pick_best);
>
> /**
> - * drm_sched_main - main scheduler thread
> + * drm_sched_free_job_work - worker to call free_job
> *
> - * @param: scheduler instance
> + * @w: free job work
> */
> -static void drm_sched_main(struct work_struct *w)
> +static void drm_sched_free_job_work(struct work_struct *w)
> {
> struct drm_gpu_scheduler *sched =
> - container_of(w, struct drm_gpu_scheduler, work_submit);
> - struct drm_sched_entity *entity;
> + container_of(w, struct drm_gpu_scheduler, work_free_job);
> struct drm_sched_job *cleanup_job;
> - int r;
>
> if (READ_ONCE(sched->pause_submit))
> return;
>
> cleanup_job = drm_sched_get_cleanup_job(sched);
> - entity = drm_sched_select_entity(sched);
> -
> - if (!entity && !cleanup_job)
> - return; /* No more work */
> -
> - if (cleanup_job)
> + if (cleanup_job) {
> sched->ops->free_job(cleanup_job);
>
> - if (entity) {
> - struct dma_fence *fence;
> - struct drm_sched_fence *s_fence;
> - struct drm_sched_job *sched_job;
> -
> - sched_job = drm_sched_entity_pop_job(entity);
> - if (!sched_job) {
> - complete_all(&entity->entity_idle);
> - if (!cleanup_job)
> - return; /* No more work */
> - goto again;
> - }
> + drm_sched_free_job_queue_if_ready(sched);
> + drm_sched_run_job_queue_if_ready(sched);
> + }
> +}
>
> - s_fence = sched_job->s_fence;
> +/**
> + * drm_sched_run_job_work - worker to call run_job
> + *
> + * @w: run job work
> + */
> +static void drm_sched_run_job_work(struct work_struct *w)
> +{
> + struct drm_gpu_scheduler *sched =
> + container_of(w, struct drm_gpu_scheduler, work_run_job);
> + struct drm_sched_entity *entity;
> + struct dma_fence *fence;
> + struct drm_sched_fence *s_fence;
> + struct drm_sched_job *sched_job;
> + int r;
>
> - atomic_inc(&sched->hw_rq_count);
> - drm_sched_job_begin(sched_job);
> + if (READ_ONCE(sched->pause_submit))
> + return;
> +
> + entity = drm_sched_select_entity(sched, true);
> + if (!entity)
> + return;
>
> - trace_drm_run_job(sched_job, entity);
> - fence = sched->ops->run_job(sched_job);
> + sched_job = drm_sched_entity_pop_job(entity);
> + if (!sched_job) {
> complete_all(&entity->entity_idle);
> - drm_sched_fence_scheduled(s_fence, fence);
> + return; /* No more work */
> + }
>
> - if (!IS_ERR_OR_NULL(fence)) {
> - /* Drop for original kref_init of the fence */
> - dma_fence_put(fence);
> + s_fence = sched_job->s_fence;
>
> - r = dma_fence_add_callback(fence, &sched_job->cb,
> - drm_sched_job_done_cb);
> - if (r == -ENOENT)
> - drm_sched_job_done(sched_job, fence->error);
> - else if (r)
> - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> - r);
> - } else {
> - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> - PTR_ERR(fence) : 0);
> - }
> + atomic_inc(&sched->hw_rq_count);
> + drm_sched_job_begin(sched_job);
>
> - wake_up(&sched->job_scheduled);
> + trace_drm_run_job(sched_job, entity);
> + fence = sched->ops->run_job(sched_job);
> + complete_all(&entity->entity_idle);
> + drm_sched_fence_scheduled(s_fence, fence);
> +
> + if (!IS_ERR_OR_NULL(fence)) {
> + /* Drop for original kref_init of the fence */
> + dma_fence_put(fence);
> +
> + r = dma_fence_add_callback(fence, &sched_job->cb,
> + drm_sched_job_done_cb);
> + if (r == -ENOENT)
> + drm_sched_job_done(sched_job, fence->error);
> + else if (r)
> + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> + } else {
> + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> + PTR_ERR(fence) : 0);
> }
>
> -again:
> - drm_sched_wqueue_enqueue(sched);
> + wake_up(&sched->job_scheduled);
> + drm_sched_run_job_queue_if_ready(sched);
> }
>
> /**
> @@ -1159,7 +1215,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> spin_lock_init(&sched->job_list_lock);
> atomic_set(&sched->hw_rq_count, 0);
> INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
> - INIT_WORK(&sched->work_submit, drm_sched_main);
> + INIT_WORK(&sched->work_run_job, drm_sched_run_job_work);
> + INIT_WORK(&sched->work_free_job, drm_sched_free_job_work);
> atomic_set(&sched->_score, 0);
> atomic64_set(&sched->job_id_count, 0);
> sched->pause_submit = false;
> @@ -1282,7 +1339,8 @@ EXPORT_SYMBOL(drm_sched_wqueue_ready);
> void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
> {
> WRITE_ONCE(sched->pause_submit, true);
> - cancel_work_sync(&sched->work_submit);
> + cancel_work_sync(&sched->work_run_job);
> + cancel_work_sync(&sched->work_free_job);
> }
> EXPORT_SYMBOL(drm_sched_wqueue_stop);
>
> @@ -1294,6 +1352,7 @@ EXPORT_SYMBOL(drm_sched_wqueue_stop);
> void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
> {
> WRITE_ONCE(sched->pause_submit, false);
> - queue_work(sched->submit_wq, &sched->work_submit);
> + queue_work(sched->submit_wq, &sched->work_run_job);
> + queue_work(sched->submit_wq, &sched->work_free_job);
> }
> EXPORT_SYMBOL(drm_sched_wqueue_start);
> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> index e67b53c3546b..625ffe040de3 100644
> --- a/include/drm/gpu_scheduler.h
> +++ b/include/drm/gpu_scheduler.h
> @@ -487,9 +487,10 @@ struct drm_sched_backend_ops {
> * finished.
> * @hw_rq_count: the number of jobs currently in the hardware queue.
> * @job_id_count: used to assign unique id to the each job.
> - * @submit_wq: workqueue used to queue @work_submit
> + * @submit_wq: workqueue used to queue @work_run_job and @work_free_job
> * @timeout_wq: workqueue used to queue @work_tdr
> - * @work_submit: schedules jobs and cleans up entities
> + * @work_run_job: schedules jobs
Perhaps a more descriptive explanation could be had,
@work_run_job: work which calls run_job op of each scheduler.
> + * @work_free_job: cleans up jobs
> * @work_tdr: schedules a delayed call to @drm_sched_job_timedout after the
> * timeout interval is over.
> * @pending_list: the list of jobs which are currently in the job queue.
> @@ -519,7 +520,8 @@ struct drm_gpu_scheduler {
> atomic64_t job_id_count;
> struct workqueue_struct *submit_wq;
> struct workqueue_struct *timeout_wq;
> - struct work_struct work_submit;
> + struct work_struct work_run_job;
> + struct work_struct work_free_job;
> struct delayed_work work_tdr;
> struct list_head pending_list;
> spinlock_t job_list_lock;
--
Regards,
Luben
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-19 1:25 ` Luben Tuikov
@ 2023-10-19 16:55 ` Matthew Brost
2023-10-22 1:27 ` Luben Tuikov
0 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-19 16:55 UTC (permalink / raw)
To: Luben Tuikov
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, dri-devel, christian.koenig,
boris.brezillon, dakr, donald.robson, lina, intel-xe,
faith.ekstrand
On Wed, Oct 18, 2023 at 09:25:36PM -0400, Luben Tuikov wrote:
> Hi,
>
> On 2023-10-17 11:09, Matthew Brost wrote:
> > Rather than call free_job and run_job in same work item have a dedicated
> > work item for each. This aligns with the design and intended use of work
> > queues.
> >
> > v2:
> > - Test for DMA_FENCE_FLAG_TIMESTAMP_BIT before setting
> > timestamp in free_job() work item (Danilo)
> > v3:
> > - Drop forward dec of drm_sched_select_entity (Boris)
> > - Return in drm_sched_run_job_work if entity NULL (Boris)
> > v4:
> > - Replace dequeue with peek and invert logic (Luben)
> > - Wrap to 100 lines (Luben)
> > - Update comments for *_queue / *_queue_if_ready functions (Luben)
> >
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > ---
> > drivers/gpu/drm/scheduler/sched_main.c | 287 +++++++++++++++----------
> > include/drm/gpu_scheduler.h | 8 +-
> > 2 files changed, 178 insertions(+), 117 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> > index 273e0fbc4eab..b1b8d9f96da5 100644
> > --- a/drivers/gpu/drm/scheduler/sched_main.c
> > +++ b/drivers/gpu/drm/scheduler/sched_main.c
> > @@ -213,11 +213,12 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
> > * drm_sched_rq_select_entity_rr - Select an entity which could provide a job to run
> > *
> > * @rq: scheduler run queue to check.
> > + * @peek: Just find, don't set to current.
>
> The "peek" rename is good--thanks!
>
> > *
> > * Try to find a ready entity, returns NULL if none found.
> > */
> > static struct drm_sched_entity *
> > -drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> > +drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq, bool peek)
> > {
> > struct drm_sched_entity *entity;
> >
> > @@ -227,8 +228,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> > if (entity) {
> > list_for_each_entry_continue(entity, &rq->entities, list) {
> > if (drm_sched_entity_is_ready(entity)) {
> > - rq->current_entity = entity;
> > - reinit_completion(&entity->entity_idle);
> > + if (!peek) {
> > + rq->current_entity = entity;
> > + reinit_completion(&entity->entity_idle);
> > + }
> > spin_unlock(&rq->lock);
> > return entity;
> > }
> > @@ -238,8 +241,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> > list_for_each_entry(entity, &rq->entities, list) {
> >
> > if (drm_sched_entity_is_ready(entity)) {
> > - rq->current_entity = entity;
> > - reinit_completion(&entity->entity_idle);
> > + if (!peek) {
> > + rq->current_entity = entity;
> > + reinit_completion(&entity->entity_idle);
> > + }
> > spin_unlock(&rq->lock);
> > return entity;
> > }
> > @@ -257,11 +262,12 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
> > * drm_sched_rq_select_entity_fifo - Select an entity which provides a job to run
> > *
> > * @rq: scheduler run queue to check.
> > + * @peek: Just find, don't set to current.
> > *
> > * Find oldest waiting ready entity, returns NULL if none found.
> > */
> > static struct drm_sched_entity *
> > -drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
> > +drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq, bool peek)
> > {
> > struct rb_node *rb;
> >
> > @@ -271,8 +277,10 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
> >
> > entity = rb_entry(rb, struct drm_sched_entity, rb_tree_node);
> > if (drm_sched_entity_is_ready(entity)) {
> > - rq->current_entity = entity;
> > - reinit_completion(&entity->entity_idle);
> > + if (!peek) {
> > + rq->current_entity = entity;
> > + reinit_completion(&entity->entity_idle);
> > + }
> > break;
> > }
> > }
> > @@ -282,13 +290,98 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
> > }
> >
> > /**
> > - * drm_sched_wqueue_enqueue - enqueue scheduler submission
> > + * drm_sched_run_job_queue - enqueue run-job work
>
> Hmm... so this removes the change from patch 1 in this series, and as such
> obviates patch 1.
>
> Do you want to go with "run_job_queue" and the names introduced here?
>
Yes, I like the run_job_queue name here.
> In this case, we can have that in patch 1 instead, and this patch
> would only split the "free job" into its own work item.
>
Sure, so s/drm_sched_wqueue_enqueue/drm_sched_run_job_queue in patch #1?
> > + * @sched: scheduler instance
> > + */
> > +static void drm_sched_run_job_queue(struct drm_gpu_scheduler *sched)
> > +{
> > + if (!READ_ONCE(sched->pause_submit))
> > + queue_work(sched->submit_wq, &sched->work_run_job);
> > +}
> > +
> > +/**
> > + * drm_sched_can_queue -- Can we queue more to the hardware?
> > + * @sched: scheduler instance
> > + *
> > + * Return true if we can push more jobs to the hw, otherwise false.
> > + */
> > +static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> > +{
> > + return atomic_read(&sched->hw_rq_count) <
> > + sched->hw_submission_limit;
> > +}
> > +
> > +/**
> > + * drm_sched_select_entity - Select next entity to process
> > + *
> > + * @sched: scheduler instance
> > + * @peek: Just find, don't set to current.
> > + *
> > + * Returns the entity to process or NULL if none are found.
> > + */
> > +static struct drm_sched_entity *
> > +drm_sched_select_entity(struct drm_gpu_scheduler *sched, bool peek)
> > +{
> > + struct drm_sched_entity *entity;
> > + int i;
> > +
> > + if (!drm_sched_can_queue(sched))
> > + return NULL;
> > +
> > + if (sched->single_entity) {
> > + if (!READ_ONCE(sched->single_entity->stopped) &&
> > + drm_sched_entity_is_ready(sched->single_entity))
> > + return sched->single_entity;
> > +
> > + return NULL;
> > + }
> > +
> > + /* Kernel run queue has higher priority than normal run queue*/
> > + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> > + entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> > + drm_sched_rq_select_entity_fifo(&sched->sched_rq[i], peek) :
> > + drm_sched_rq_select_entity_rr(&sched->sched_rq[i], peek);
> > + if (entity)
> > + break;
> > + }
> > +
> > + return entity;
> > +}
>
> Can you shed some light on why you need the "peek" capability, i.e. to just
> see the entity without actually arming it?
>
> In a single-entity scheduling, surely peeking at the single entity of
> a scheduler, can be done easier.
>
> I'm asking as none of these functions were intended as a peek-only/-capable
> solution, and if you need it, this shows that the infrastructure lacks
> something which you need.
>
> It's not designed for "peek", as you can see above: it gets passed-through
> the function stack until it reaches core functionality to be used in logic.
>
> (It just makes the code too complicated with too many special cases, "if's"
> and if we keep doing this, it would become very hard to understand,
> maintain, and contribute to in a few years.)
>
This question made me realize that the callers of this function inverted
the peek arguments. Will fix in next rev.
The peek is need for drm_sched_run_job_queue_if_ready which checks if a
job is ready before queuing the job run worker. This is called from the
free job worker (hw submission count decrement makes a new job ready) or
from the run job worker (another job in queue).
If we want to blindly queue the job in these cases then this can be
removed.
Or alternatively we could remove the peek argument and blindly reinit
the idle when selecting entity. I think this fine if I am reading the
code correctly. If you agree, I'd lean towards this option.
> > +
> > +/**
> > + * drm_sched_run_job_queue_if_ready - enqueue run-job work if ready
> > + * @sched: scheduler instance
> > + */
>
> "if ready" here makes perfect sense, but in the "free job" case,
> it should be "if done". See below.
>
Yes, agree that if done for job makes more sense.
> > +static void drm_sched_run_job_queue_if_ready(struct drm_gpu_scheduler *sched)
> > +{
> > + if (drm_sched_select_entity(sched, false))
> > + drm_sched_run_job_queue(sched);
> > +}
> > +
> > +/**
> > + * drm_sched_free_job_queue - enqueue free-job work
> > * @sched: scheduler instance
> > */
> > -static void drm_sched_wqueue_enqueue(struct drm_gpu_scheduler *sched)
> > +static void drm_sched_free_job_queue(struct drm_gpu_scheduler *sched)
> > {
> > if (!READ_ONCE(sched->pause_submit))
> > - queue_work(sched->submit_wq, &sched->work_submit);
> > + queue_work(sched->submit_wq, &sched->work_free_job);
> > +}
> > +
> > +/**
> > + * drm_sched_free_job_queue_if_ready - enqueue free-job work if ready
>
> This should be "if done". Please change this to "if done".
>
> > + * @sched: scheduler instance
> > + */
> > +static void drm_sched_free_job_queue_if_ready(struct drm_gpu_scheduler *sched)
>
> This should be "if_done". Please change this to "if_done".
>
> A "job" is _done_ when it's on the pending list and its fence has been signalled,
> and we free a job only when it's done, not "ready".
>
Agree on all of this.
> > +{
> > + struct drm_sched_job *job;
> > +
> > + spin_lock(&sched->job_list_lock);
> > + job = list_first_entry_or_null(&sched->pending_list,
> > + struct drm_sched_job, list);
> > + if (job && dma_fence_is_signaled(&job->s_fence->finished))
> > + drm_sched_free_job_queue(sched);
> > + spin_unlock(&sched->job_list_lock);
> > }
> >
> > /**
> > @@ -310,7 +403,7 @@ static void drm_sched_job_done(struct drm_sched_job *s_job, int result)
> > dma_fence_get(&s_fence->finished);
> > drm_sched_fence_finished(s_fence, result);
> > dma_fence_put(&s_fence->finished);
> > - drm_sched_wqueue_enqueue(sched);
> > + drm_sched_free_job_queue(sched);
> > }
> >
> > /**
> > @@ -576,7 +669,7 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
> > drm_sched_job_done(s_job, fence->error);
> > else if (r)
> > DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > - r);
> > + r);
> > } else
> > drm_sched_job_done(s_job, -ECANCELED);
> > }
> > @@ -885,18 +978,6 @@ void drm_sched_job_cleanup(struct drm_sched_job *job)
> > }
> > EXPORT_SYMBOL(drm_sched_job_cleanup);
> >
> > -/**
> > - * drm_sched_can_queue -- Can we queue more to the hardware?
> > - * @sched: scheduler instance
> > - *
> > - * Return true if we can push more jobs to the hw, otherwise false.
> > - */
> > -static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> > -{
> > - return atomic_read(&sched->hw_rq_count) <
> > - sched->hw_submission_limit;
> > -}
> > -
> > /**
> > * drm_sched_wakeup_if_can_queue - Wake up the scheduler
> > * @sched: scheduler instance
> > @@ -906,43 +987,7 @@ static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
> > void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched)
> > {
> > if (drm_sched_can_queue(sched))
> > - drm_sched_wqueue_enqueue(sched);
> > -}
> > -
> > -/**
> > - * drm_sched_select_entity - Select next entity to process
> > - *
> > - * @sched: scheduler instance
> > - *
> > - * Returns the entity to process or NULL if none are found.
> > - */
> > -static struct drm_sched_entity *
> > -drm_sched_select_entity(struct drm_gpu_scheduler *sched)
> > -{
> > - struct drm_sched_entity *entity;
> > - int i;
> > -
> > - if (!drm_sched_can_queue(sched))
> > - return NULL;
> > -
> > - if (sched->single_entity) {
> > - if (!READ_ONCE(sched->single_entity->stopped) &&
> > - drm_sched_entity_is_ready(sched->single_entity))
> > - return sched->single_entity;
> > -
> > - return NULL;
> > - }
> > -
> > - /* Kernel run queue has higher priority than normal run queue*/
> > - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> > - entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> > - drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
> > - drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
> > - if (entity)
> > - break;
> > - }
> > -
> > - return entity;
> > + drm_sched_run_job_queue(sched);
> > }
> >
> > /**
> > @@ -974,8 +1019,10 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
> > typeof(*next), list);
> >
> > if (next) {
> > - next->s_fence->scheduled.timestamp =
> > - dma_fence_timestamp(&job->s_fence->finished);
> > + if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
> > + &next->s_fence->scheduled.flags))
> > + next->s_fence->scheduled.timestamp =
> > + dma_fence_timestamp(&job->s_fence->finished);
> > /* start TO timer for next job */
> > drm_sched_start_timeout(sched);
> > }
> > @@ -1025,74 +1072,83 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list,
> > EXPORT_SYMBOL(drm_sched_pick_best);
> >
> > /**
> > - * drm_sched_main - main scheduler thread
> > + * drm_sched_free_job_work - worker to call free_job
> > *
> > - * @param: scheduler instance
> > + * @w: free job work
> > */
> > -static void drm_sched_main(struct work_struct *w)
> > +static void drm_sched_free_job_work(struct work_struct *w)
> > {
> > struct drm_gpu_scheduler *sched =
> > - container_of(w, struct drm_gpu_scheduler, work_submit);
> > - struct drm_sched_entity *entity;
> > + container_of(w, struct drm_gpu_scheduler, work_free_job);
> > struct drm_sched_job *cleanup_job;
> > - int r;
> >
> > if (READ_ONCE(sched->pause_submit))
> > return;
> >
> > cleanup_job = drm_sched_get_cleanup_job(sched);
> > - entity = drm_sched_select_entity(sched);
> > -
> > - if (!entity && !cleanup_job)
> > - return; /* No more work */
> > -
> > - if (cleanup_job)
> > + if (cleanup_job) {
> > sched->ops->free_job(cleanup_job);
> >
> > - if (entity) {
> > - struct dma_fence *fence;
> > - struct drm_sched_fence *s_fence;
> > - struct drm_sched_job *sched_job;
> > -
> > - sched_job = drm_sched_entity_pop_job(entity);
> > - if (!sched_job) {
> > - complete_all(&entity->entity_idle);
> > - if (!cleanup_job)
> > - return; /* No more work */
> > - goto again;
> > - }
> > + drm_sched_free_job_queue_if_ready(sched);
> > + drm_sched_run_job_queue_if_ready(sched);
> > + }
> > +}
> >
> > - s_fence = sched_job->s_fence;
> > +/**
> > + * drm_sched_run_job_work - worker to call run_job
> > + *
> > + * @w: run job work
> > + */
> > +static void drm_sched_run_job_work(struct work_struct *w)
> > +{
> > + struct drm_gpu_scheduler *sched =
> > + container_of(w, struct drm_gpu_scheduler, work_run_job);
> > + struct drm_sched_entity *entity;
> > + struct dma_fence *fence;
> > + struct drm_sched_fence *s_fence;
> > + struct drm_sched_job *sched_job;
> > + int r;
> >
> > - atomic_inc(&sched->hw_rq_count);
> > - drm_sched_job_begin(sched_job);
> > + if (READ_ONCE(sched->pause_submit))
> > + return;
> > +
> > + entity = drm_sched_select_entity(sched, true);
> > + if (!entity)
> > + return;
> >
> > - trace_drm_run_job(sched_job, entity);
> > - fence = sched->ops->run_job(sched_job);
> > + sched_job = drm_sched_entity_pop_job(entity);
> > + if (!sched_job) {
> > complete_all(&entity->entity_idle);
> > - drm_sched_fence_scheduled(s_fence, fence);
> > + return; /* No more work */
> > + }
> >
> > - if (!IS_ERR_OR_NULL(fence)) {
> > - /* Drop for original kref_init of the fence */
> > - dma_fence_put(fence);
> > + s_fence = sched_job->s_fence;
> >
> > - r = dma_fence_add_callback(fence, &sched_job->cb,
> > - drm_sched_job_done_cb);
> > - if (r == -ENOENT)
> > - drm_sched_job_done(sched_job, fence->error);
> > - else if (r)
> > - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > - r);
> > - } else {
> > - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > - PTR_ERR(fence) : 0);
> > - }
> > + atomic_inc(&sched->hw_rq_count);
> > + drm_sched_job_begin(sched_job);
> >
> > - wake_up(&sched->job_scheduled);
> > + trace_drm_run_job(sched_job, entity);
> > + fence = sched->ops->run_job(sched_job);
> > + complete_all(&entity->entity_idle);
> > + drm_sched_fence_scheduled(s_fence, fence);
> > +
> > + if (!IS_ERR_OR_NULL(fence)) {
> > + /* Drop for original kref_init of the fence */
> > + dma_fence_put(fence);
> > +
> > + r = dma_fence_add_callback(fence, &sched_job->cb,
> > + drm_sched_job_done_cb);
> > + if (r == -ENOENT)
> > + drm_sched_job_done(sched_job, fence->error);
> > + else if (r)
> > + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> > + } else {
> > + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > + PTR_ERR(fence) : 0);
> > }
> >
> > -again:
> > - drm_sched_wqueue_enqueue(sched);
> > + wake_up(&sched->job_scheduled);
> > + drm_sched_run_job_queue_if_ready(sched);
> > }
> >
> > /**
> > @@ -1159,7 +1215,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> > spin_lock_init(&sched->job_list_lock);
> > atomic_set(&sched->hw_rq_count, 0);
> > INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
> > - INIT_WORK(&sched->work_submit, drm_sched_main);
> > + INIT_WORK(&sched->work_run_job, drm_sched_run_job_work);
> > + INIT_WORK(&sched->work_free_job, drm_sched_free_job_work);
> > atomic_set(&sched->_score, 0);
> > atomic64_set(&sched->job_id_count, 0);
> > sched->pause_submit = false;
> > @@ -1282,7 +1339,8 @@ EXPORT_SYMBOL(drm_sched_wqueue_ready);
> > void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
> > {
> > WRITE_ONCE(sched->pause_submit, true);
> > - cancel_work_sync(&sched->work_submit);
> > + cancel_work_sync(&sched->work_run_job);
> > + cancel_work_sync(&sched->work_free_job);
> > }
> > EXPORT_SYMBOL(drm_sched_wqueue_stop);
> >
> > @@ -1294,6 +1352,7 @@ EXPORT_SYMBOL(drm_sched_wqueue_stop);
> > void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
> > {
> > WRITE_ONCE(sched->pause_submit, false);
> > - queue_work(sched->submit_wq, &sched->work_submit);
> > + queue_work(sched->submit_wq, &sched->work_run_job);
> > + queue_work(sched->submit_wq, &sched->work_free_job);
> > }
> > EXPORT_SYMBOL(drm_sched_wqueue_start);
> > diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> > index e67b53c3546b..625ffe040de3 100644
> > --- a/include/drm/gpu_scheduler.h
> > +++ b/include/drm/gpu_scheduler.h
> > @@ -487,9 +487,10 @@ struct drm_sched_backend_ops {
> > * finished.
> > * @hw_rq_count: the number of jobs currently in the hardware queue.
> > * @job_id_count: used to assign unique id to the each job.
> > - * @submit_wq: workqueue used to queue @work_submit
> > + * @submit_wq: workqueue used to queue @work_run_job and @work_free_job
> > * @timeout_wq: workqueue used to queue @work_tdr
> > - * @work_submit: schedules jobs and cleans up entities
> > + * @work_run_job: schedules jobs
>
> Perhaps a more descriptive explanation could be had,
>
> @work_run_job: work which calls run_job op of each scheduler.
>
Got it.
> > + * @work_free_job: cleans up jobs
> > * @work_tdr: schedules a delayed call to @drm_sched_job_timedout after the
> > * timeout interval is over.
> > * @pending_list: the list of jobs which are currently in the job queue.
> > @@ -519,7 +520,8 @@ struct drm_gpu_scheduler {
> > atomic64_t job_id_count;
> > struct workqueue_struct *submit_wq;
> > struct workqueue_struct *timeout_wq;
> > - struct work_struct work_submit;
> > + struct work_struct work_run_job;
> > + struct work_struct work_free_job;
> > struct delayed_work work_tdr;
> > struct list_head pending_list;
> > spinlock_t job_list_lock;
>
> --
Thanks for the review.
Matt
> Regards,
> Luben
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-19 16:55 ` Matthew Brost
@ 2023-10-22 1:27 ` Luben Tuikov
0 siblings, 0 replies; 22+ messages in thread
From: Luben Tuikov @ 2023-10-22 1:27 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, dri-devel, christian.koenig,
boris.brezillon, dakr, donald.robson, lina, intel-xe,
faith.ekstrand
Hi,
On 2023-10-19 12:55, Matthew Brost wrote:
> On Wed, Oct 18, 2023 at 09:25:36PM -0400, Luben Tuikov wrote:
>> Hi,
>>
>> On 2023-10-17 11:09, Matthew Brost wrote:
>>> Rather than call free_job and run_job in same work item have a dedicated
>>> work item for each. This aligns with the design and intended use of work
>>> queues.
>>>
>>> v2:
>>> - Test for DMA_FENCE_FLAG_TIMESTAMP_BIT before setting
>>> timestamp in free_job() work item (Danilo)
>>> v3:
>>> - Drop forward dec of drm_sched_select_entity (Boris)
>>> - Return in drm_sched_run_job_work if entity NULL (Boris)
>>> v4:
>>> - Replace dequeue with peek and invert logic (Luben)
>>> - Wrap to 100 lines (Luben)
>>> - Update comments for *_queue / *_queue_if_ready functions (Luben)
>>>
>>> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
>>> ---
>>> drivers/gpu/drm/scheduler/sched_main.c | 287 +++++++++++++++----------
>>> include/drm/gpu_scheduler.h | 8 +-
>>> 2 files changed, 178 insertions(+), 117 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
>>> index 273e0fbc4eab..b1b8d9f96da5 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_main.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_main.c
>>> @@ -213,11 +213,12 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
>>> * drm_sched_rq_select_entity_rr - Select an entity which could provide a job to run
>>> *
>>> * @rq: scheduler run queue to check.
>>> + * @peek: Just find, don't set to current.
>>
>> The "peek" rename is good--thanks!
>>
>>> *
>>> * Try to find a ready entity, returns NULL if none found.
>>> */
>>> static struct drm_sched_entity *
>>> -drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
>>> +drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq, bool peek)
>>> {
>>> struct drm_sched_entity *entity;
>>>
>>> @@ -227,8 +228,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
>>> if (entity) {
>>> list_for_each_entry_continue(entity, &rq->entities, list) {
>>> if (drm_sched_entity_is_ready(entity)) {
>>> - rq->current_entity = entity;
>>> - reinit_completion(&entity->entity_idle);
>>> + if (!peek) {
>>> + rq->current_entity = entity;
>>> + reinit_completion(&entity->entity_idle);
>>> + }
>>> spin_unlock(&rq->lock);
>>> return entity;
>>> }
>>> @@ -238,8 +241,10 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
>>> list_for_each_entry(entity, &rq->entities, list) {
>>>
>>> if (drm_sched_entity_is_ready(entity)) {
>>> - rq->current_entity = entity;
>>> - reinit_completion(&entity->entity_idle);
>>> + if (!peek) {
>>> + rq->current_entity = entity;
>>> + reinit_completion(&entity->entity_idle);
>>> + }
>>> spin_unlock(&rq->lock);
>>> return entity;
>>> }
>>> @@ -257,11 +262,12 @@ drm_sched_rq_select_entity_rr(struct drm_sched_rq *rq)
>>> * drm_sched_rq_select_entity_fifo - Select an entity which provides a job to run
>>> *
>>> * @rq: scheduler run queue to check.
>>> + * @peek: Just find, don't set to current.
>>> *
>>> * Find oldest waiting ready entity, returns NULL if none found.
>>> */
>>> static struct drm_sched_entity *
>>> -drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
>>> +drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq, bool peek)
>>> {
>>> struct rb_node *rb;
>>>
>>> @@ -271,8 +277,10 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
>>>
>>> entity = rb_entry(rb, struct drm_sched_entity, rb_tree_node);
>>> if (drm_sched_entity_is_ready(entity)) {
>>> - rq->current_entity = entity;
>>> - reinit_completion(&entity->entity_idle);
>>> + if (!peek) {
>>> + rq->current_entity = entity;
>>> + reinit_completion(&entity->entity_idle);
>>> + }
>>> break;
>>> }
>>> }
>>> @@ -282,13 +290,98 @@ drm_sched_rq_select_entity_fifo(struct drm_sched_rq *rq)
>>> }
>>>
>>> /**
>>> - * drm_sched_wqueue_enqueue - enqueue scheduler submission
>>> + * drm_sched_run_job_queue - enqueue run-job work
>>
>> Hmm... so this removes the change from patch 1 in this series, and as such
>> obviates patch 1.
>>
>> Do you want to go with "run_job_queue" and the names introduced here?
>>
>
> Yes, I like the run_job_queue name here.
>
>> In this case, we can have that in patch 1 instead, and this patch
>> would only split the "free job" into its own work item.
>>
>
> Sure, so s/drm_sched_wqueue_enqueue/drm_sched_run_job_queue in patch #1?
Yes. That'll create less thrashing about here and essentialize the changes
this patch is doing here (i.e. minimize the number of changes).
I think "run_job_queue" is a good enough name, and please add lucid comments,
so that it's easy to understand--kinda of like you're telling a story. Thanks!
Generally, patch series shouldn't introduce a change (like in patch #1) which
the same patch series further changes yet again to something else (like in this
patch here). It should change one thing only once.
>
>>> + * @sched: scheduler instance
>>> + */
>>> +static void drm_sched_run_job_queue(struct drm_gpu_scheduler *sched)
>>> +{
>>> + if (!READ_ONCE(sched->pause_submit))
>>> + queue_work(sched->submit_wq, &sched->work_run_job);
>>> +}
>>> +
>>> +/**
>>> + * drm_sched_can_queue -- Can we queue more to the hardware?
>>> + * @sched: scheduler instance
>>> + *
>>> + * Return true if we can push more jobs to the hw, otherwise false.
>>> + */
>>> +static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
>>> +{
>>> + return atomic_read(&sched->hw_rq_count) <
>>> + sched->hw_submission_limit;
>>> +}
>>> +
>>> +/**
>>> + * drm_sched_select_entity - Select next entity to process
>>> + *
>>> + * @sched: scheduler instance
>>> + * @peek: Just find, don't set to current.
>>> + *
>>> + * Returns the entity to process or NULL if none are found.
>>> + */
>>> +static struct drm_sched_entity *
>>> +drm_sched_select_entity(struct drm_gpu_scheduler *sched, bool peek)
>>> +{
>>> + struct drm_sched_entity *entity;
>>> + int i;
>>> +
>>> + if (!drm_sched_can_queue(sched))
>>> + return NULL;
>>> +
>>> + if (sched->single_entity) {
>>> + if (!READ_ONCE(sched->single_entity->stopped) &&
>>> + drm_sched_entity_is_ready(sched->single_entity))
>>> + return sched->single_entity;
>>> +
>>> + return NULL;
>>> + }
>>> +
>>> + /* Kernel run queue has higher priority than normal run queue*/
>>> + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
>>> + entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
>>> + drm_sched_rq_select_entity_fifo(&sched->sched_rq[i], peek) :
>>> + drm_sched_rq_select_entity_rr(&sched->sched_rq[i], peek);
>>> + if (entity)
>>> + break;
>>> + }
>>> +
>>> + return entity;
>>> +}
>>
>> Can you shed some light on why you need the "peek" capability, i.e. to just
>> see the entity without actually arming it?
>>
>> In a single-entity scheduling, surely peeking at the single entity of
>> a scheduler, can be done easier.
>>
>> I'm asking as none of these functions were intended as a peek-only/-capable
>> solution, and if you need it, this shows that the infrastructure lacks
>> something which you need.
>>
>> It's not designed for "peek", as you can see above: it gets passed-through
>> the function stack until it reaches core functionality to be used in logic.
>>
>> (It just makes the code too complicated with too many special cases, "if's"
>> and if we keep doing this, it would become very hard to understand,
>> maintain, and contribute to in a few years.)
>>
>
> This question made me realize that the callers of this function inverted
> the peek arguments. Will fix in next rev.
>
> The peek is need for drm_sched_run_job_queue_if_ready which checks if a
> job is ready before queuing the job run worker. This is called from the
> free job worker (hw submission count decrement makes a new job ready) or
> from the run job worker (another job in queue).
Yes, we kind of want "queue if ready" to be "atomic", instead of thrashing
about like,
if run_if_ready(peek),
j = run_if_ready(!peek),
run_job(j).
to become
j = run_if_ready(void).
if (j) run_job(j).
> If we want to blindly queue the job in these cases then this can be
> removed.
Yes, it's what we're doing now.
> Or alternatively we could remove the peek argument and blindly reinit
> the idle when selecting entity. I think this fine if I am reading the
> code correctly. If you agree, I'd lean towards this option.
Yes please--it's what we have now, so no need to change it. select_entity()
really just gives back the entity whose jobs we can (should) run next.
So, yes, let's leave it as is, without a "peek". Thanks!
>
>>> +
>>> +/**
>>> + * drm_sched_run_job_queue_if_ready - enqueue run-job work if ready
>>> + * @sched: scheduler instance
>>> + */
>>
>> "if ready" here makes perfect sense, but in the "free job" case,
>> it should be "if done". See below.
>>
>
> Yes, agree that if done for job makes more sense.
>
>>> +static void drm_sched_run_job_queue_if_ready(struct drm_gpu_scheduler *sched)
>>> +{
>>> + if (drm_sched_select_entity(sched, false))
>>> + drm_sched_run_job_queue(sched);
>>> +}
>>> +
>>> +/**
>>> + * drm_sched_free_job_queue - enqueue free-job work
>>> * @sched: scheduler instance
>>> */
>>> -static void drm_sched_wqueue_enqueue(struct drm_gpu_scheduler *sched)
>>> +static void drm_sched_free_job_queue(struct drm_gpu_scheduler *sched)
>>> {
>>> if (!READ_ONCE(sched->pause_submit))
>>> - queue_work(sched->submit_wq, &sched->work_submit);
>>> + queue_work(sched->submit_wq, &sched->work_free_job);
>>> +}
>>> +
>>> +/**
>>> + * drm_sched_free_job_queue_if_ready - enqueue free-job work if ready
>>
>> This should be "if done". Please change this to "if done".
>>
>>> + * @sched: scheduler instance
>>> + */
>>> +static void drm_sched_free_job_queue_if_ready(struct drm_gpu_scheduler *sched)
>>
>> This should be "if_done". Please change this to "if_done".
>>
>> A "job" is _done_ when it's on the pending list and its fence has been signalled,
>> and we free a job only when it's done, not "ready".
>>
>
> Agree on all of this.
Great! :-)
Regards,
Luben
>
>>> +{
>>> + struct drm_sched_job *job;
>>> +
>>> + spin_lock(&sched->job_list_lock);
>>> + job = list_first_entry_or_null(&sched->pending_list,
>>> + struct drm_sched_job, list);
>>> + if (job && dma_fence_is_signaled(&job->s_fence->finished))
>>> + drm_sched_free_job_queue(sched);
>>> + spin_unlock(&sched->job_list_lock);
>>> }
>>>
>>> /**
>>> @@ -310,7 +403,7 @@ static void drm_sched_job_done(struct drm_sched_job *s_job, int result)
>>> dma_fence_get(&s_fence->finished);
>>> drm_sched_fence_finished(s_fence, result);
>>> dma_fence_put(&s_fence->finished);
>>> - drm_sched_wqueue_enqueue(sched);
>>> + drm_sched_free_job_queue(sched);
>>> }
>>>
>>> /**
>>> @@ -576,7 +669,7 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
>>> drm_sched_job_done(s_job, fence->error);
>>> else if (r)
>>> DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
>>> - r);
>>> + r);
>>> } else
>>> drm_sched_job_done(s_job, -ECANCELED);
>>> }
>>> @@ -885,18 +978,6 @@ void drm_sched_job_cleanup(struct drm_sched_job *job)
>>> }
>>> EXPORT_SYMBOL(drm_sched_job_cleanup);
>>>
>>> -/**
>>> - * drm_sched_can_queue -- Can we queue more to the hardware?
>>> - * @sched: scheduler instance
>>> - *
>>> - * Return true if we can push more jobs to the hw, otherwise false.
>>> - */
>>> -static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
>>> -{
>>> - return atomic_read(&sched->hw_rq_count) <
>>> - sched->hw_submission_limit;
>>> -}
>>> -
>>> /**
>>> * drm_sched_wakeup_if_can_queue - Wake up the scheduler
>>> * @sched: scheduler instance
>>> @@ -906,43 +987,7 @@ static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched)
>>> void drm_sched_wakeup_if_can_queue(struct drm_gpu_scheduler *sched)
>>> {
>>> if (drm_sched_can_queue(sched))
>>> - drm_sched_wqueue_enqueue(sched);
>>> -}
>>> -
>>> -/**
>>> - * drm_sched_select_entity - Select next entity to process
>>> - *
>>> - * @sched: scheduler instance
>>> - *
>>> - * Returns the entity to process or NULL if none are found.
>>> - */
>>> -static struct drm_sched_entity *
>>> -drm_sched_select_entity(struct drm_gpu_scheduler *sched)
>>> -{
>>> - struct drm_sched_entity *entity;
>>> - int i;
>>> -
>>> - if (!drm_sched_can_queue(sched))
>>> - return NULL;
>>> -
>>> - if (sched->single_entity) {
>>> - if (!READ_ONCE(sched->single_entity->stopped) &&
>>> - drm_sched_entity_is_ready(sched->single_entity))
>>> - return sched->single_entity;
>>> -
>>> - return NULL;
>>> - }
>>> -
>>> - /* Kernel run queue has higher priority than normal run queue*/
>>> - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
>>> - entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
>>> - drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
>>> - drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
>>> - if (entity)
>>> - break;
>>> - }
>>> -
>>> - return entity;
>>> + drm_sched_run_job_queue(sched);
>>> }
>>>
>>> /**
>>> @@ -974,8 +1019,10 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
>>> typeof(*next), list);
>>>
>>> if (next) {
>>> - next->s_fence->scheduled.timestamp =
>>> - dma_fence_timestamp(&job->s_fence->finished);
>>> + if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
>>> + &next->s_fence->scheduled.flags))
>>> + next->s_fence->scheduled.timestamp =
>>> + dma_fence_timestamp(&job->s_fence->finished);
>>> /* start TO timer for next job */
>>> drm_sched_start_timeout(sched);
>>> }
>>> @@ -1025,74 +1072,83 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list,
>>> EXPORT_SYMBOL(drm_sched_pick_best);
>>>
>>> /**
>>> - * drm_sched_main - main scheduler thread
>>> + * drm_sched_free_job_work - worker to call free_job
>>> *
>>> - * @param: scheduler instance
>>> + * @w: free job work
>>> */
>>> -static void drm_sched_main(struct work_struct *w)
>>> +static void drm_sched_free_job_work(struct work_struct *w)
>>> {
>>> struct drm_gpu_scheduler *sched =
>>> - container_of(w, struct drm_gpu_scheduler, work_submit);
>>> - struct drm_sched_entity *entity;
>>> + container_of(w, struct drm_gpu_scheduler, work_free_job);
>>> struct drm_sched_job *cleanup_job;
>>> - int r;
>>>
>>> if (READ_ONCE(sched->pause_submit))
>>> return;
>>>
>>> cleanup_job = drm_sched_get_cleanup_job(sched);
>>> - entity = drm_sched_select_entity(sched);
>>> -
>>> - if (!entity && !cleanup_job)
>>> - return; /* No more work */
>>> -
>>> - if (cleanup_job)
>>> + if (cleanup_job) {
>>> sched->ops->free_job(cleanup_job);
>>>
>>> - if (entity) {
>>> - struct dma_fence *fence;
>>> - struct drm_sched_fence *s_fence;
>>> - struct drm_sched_job *sched_job;
>>> -
>>> - sched_job = drm_sched_entity_pop_job(entity);
>>> - if (!sched_job) {
>>> - complete_all(&entity->entity_idle);
>>> - if (!cleanup_job)
>>> - return; /* No more work */
>>> - goto again;
>>> - }
>>> + drm_sched_free_job_queue_if_ready(sched);
>>> + drm_sched_run_job_queue_if_ready(sched);
>>> + }
>>> +}
>>>
>>> - s_fence = sched_job->s_fence;
>>> +/**
>>> + * drm_sched_run_job_work - worker to call run_job
>>> + *
>>> + * @w: run job work
>>> + */
>>> +static void drm_sched_run_job_work(struct work_struct *w)
>>> +{
>>> + struct drm_gpu_scheduler *sched =
>>> + container_of(w, struct drm_gpu_scheduler, work_run_job);
>>> + struct drm_sched_entity *entity;
>>> + struct dma_fence *fence;
>>> + struct drm_sched_fence *s_fence;
>>> + struct drm_sched_job *sched_job;
>>> + int r;
>>>
>>> - atomic_inc(&sched->hw_rq_count);
>>> - drm_sched_job_begin(sched_job);
>>> + if (READ_ONCE(sched->pause_submit))
>>> + return;
>>> +
>>> + entity = drm_sched_select_entity(sched, true);
>>> + if (!entity)
>>> + return;
>>>
>>> - trace_drm_run_job(sched_job, entity);
>>> - fence = sched->ops->run_job(sched_job);
>>> + sched_job = drm_sched_entity_pop_job(entity);
>>> + if (!sched_job) {
>>> complete_all(&entity->entity_idle);
>>> - drm_sched_fence_scheduled(s_fence, fence);
>>> + return; /* No more work */
>>> + }
>>>
>>> - if (!IS_ERR_OR_NULL(fence)) {
>>> - /* Drop for original kref_init of the fence */
>>> - dma_fence_put(fence);
>>> + s_fence = sched_job->s_fence;
>>>
>>> - r = dma_fence_add_callback(fence, &sched_job->cb,
>>> - drm_sched_job_done_cb);
>>> - if (r == -ENOENT)
>>> - drm_sched_job_done(sched_job, fence->error);
>>> - else if (r)
>>> - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
>>> - r);
>>> - } else {
>>> - drm_sched_job_done(sched_job, IS_ERR(fence) ?
>>> - PTR_ERR(fence) : 0);
>>> - }
>>> + atomic_inc(&sched->hw_rq_count);
>>> + drm_sched_job_begin(sched_job);
>>>
>>> - wake_up(&sched->job_scheduled);
>>> + trace_drm_run_job(sched_job, entity);
>>> + fence = sched->ops->run_job(sched_job);
>>> + complete_all(&entity->entity_idle);
>>> + drm_sched_fence_scheduled(s_fence, fence);
>>> +
>>> + if (!IS_ERR_OR_NULL(fence)) {
>>> + /* Drop for original kref_init of the fence */
>>> + dma_fence_put(fence);
>>> +
>>> + r = dma_fence_add_callback(fence, &sched_job->cb,
>>> + drm_sched_job_done_cb);
>>> + if (r == -ENOENT)
>>> + drm_sched_job_done(sched_job, fence->error);
>>> + else if (r)
>>> + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
>>> + } else {
>>> + drm_sched_job_done(sched_job, IS_ERR(fence) ?
>>> + PTR_ERR(fence) : 0);
>>> }
>>>
>>> -again:
>>> - drm_sched_wqueue_enqueue(sched);
>>> + wake_up(&sched->job_scheduled);
>>> + drm_sched_run_job_queue_if_ready(sched);
>>> }
>>>
>>> /**
>>> @@ -1159,7 +1215,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
>>> spin_lock_init(&sched->job_list_lock);
>>> atomic_set(&sched->hw_rq_count, 0);
>>> INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
>>> - INIT_WORK(&sched->work_submit, drm_sched_main);
>>> + INIT_WORK(&sched->work_run_job, drm_sched_run_job_work);
>>> + INIT_WORK(&sched->work_free_job, drm_sched_free_job_work);
>>> atomic_set(&sched->_score, 0);
>>> atomic64_set(&sched->job_id_count, 0);
>>> sched->pause_submit = false;
>>> @@ -1282,7 +1339,8 @@ EXPORT_SYMBOL(drm_sched_wqueue_ready);
>>> void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched)
>>> {
>>> WRITE_ONCE(sched->pause_submit, true);
>>> - cancel_work_sync(&sched->work_submit);
>>> + cancel_work_sync(&sched->work_run_job);
>>> + cancel_work_sync(&sched->work_free_job);
>>> }
>>> EXPORT_SYMBOL(drm_sched_wqueue_stop);
>>>
>>> @@ -1294,6 +1352,7 @@ EXPORT_SYMBOL(drm_sched_wqueue_stop);
>>> void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched)
>>> {
>>> WRITE_ONCE(sched->pause_submit, false);
>>> - queue_work(sched->submit_wq, &sched->work_submit);
>>> + queue_work(sched->submit_wq, &sched->work_run_job);
>>> + queue_work(sched->submit_wq, &sched->work_free_job);
>>> }
>>> EXPORT_SYMBOL(drm_sched_wqueue_start);
>>> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
>>> index e67b53c3546b..625ffe040de3 100644
>>> --- a/include/drm/gpu_scheduler.h
>>> +++ b/include/drm/gpu_scheduler.h
>>> @@ -487,9 +487,10 @@ struct drm_sched_backend_ops {
>>> * finished.
>>> * @hw_rq_count: the number of jobs currently in the hardware queue.
>>> * @job_id_count: used to assign unique id to the each job.
>>> - * @submit_wq: workqueue used to queue @work_submit
>>> + * @submit_wq: workqueue used to queue @work_run_job and @work_free_job
>>> * @timeout_wq: workqueue used to queue @work_tdr
>>> - * @work_submit: schedules jobs and cleans up entities
>>> + * @work_run_job: schedules jobs
>>
>> Perhaps a more descriptive explanation could be had,
>>
>> @work_run_job: work which calls run_job op of each scheduler.
>>
>
> Got it.
>
>>> + * @work_free_job: cleans up jobs
>>> * @work_tdr: schedules a delayed call to @drm_sched_job_timedout after the
>>> * timeout interval is over.
>>> * @pending_list: the list of jobs which are currently in the job queue.
>>> @@ -519,7 +520,8 @@ struct drm_gpu_scheduler {
>>> atomic64_t job_id_count;
>>> struct workqueue_struct *submit_wq;
>>> struct workqueue_struct *timeout_wq;
>>> - struct work_struct work_submit;
>>> + struct work_struct work_run_job;
>>> + struct work_struct work_free_job;
>>> struct delayed_work work_tdr;
>>> struct list_head pending_list;
>>> spinlock_t job_list_lock;
>>
>> --
>
> Thanks for the review.
>
> Matt
>
>> Regards,
>> Luben
>>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-17 15:09 ` [PATCH v6 5/7] drm/sched: Split free_job into own work item Matthew Brost
2023-10-19 1:25 ` Luben Tuikov
@ 2023-10-23 12:16 ` Boris Brezillon
2023-10-23 12:39 ` Boris Brezillon
1 sibling, 1 reply; 22+ messages in thread
From: Boris Brezillon @ 2023-10-23 12:16 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, dri-devel, christian.koenig, luben.tuikov,
dakr, donald.robson, lina, intel-xe, faith.ekstrand
Hi,
On Tue, 17 Oct 2023 08:09:56 -0700
Matthew Brost <matthew.brost@intel.com> wrote:
> +static void drm_sched_run_job_work(struct work_struct *w)
> +{
> + struct drm_gpu_scheduler *sched =
> + container_of(w, struct drm_gpu_scheduler, work_run_job);
> + struct drm_sched_entity *entity;
> + struct dma_fence *fence;
> + struct drm_sched_fence *s_fence;
> + struct drm_sched_job *sched_job;
> + int r;
>
> - atomic_inc(&sched->hw_rq_count);
> - drm_sched_job_begin(sched_job);
> + if (READ_ONCE(sched->pause_submit))
> + return;
> +
> + entity = drm_sched_select_entity(sched, true);
> + if (!entity)
> + return;
>
> - trace_drm_run_job(sched_job, entity);
> - fence = sched->ops->run_job(sched_job);
> + sched_job = drm_sched_entity_pop_job(entity);
> + if (!sched_job) {
> complete_all(&entity->entity_idle);
> - drm_sched_fence_scheduled(s_fence, fence);
> + return; /* No more work */
> + }
>
> - if (!IS_ERR_OR_NULL(fence)) {
> - /* Drop for original kref_init of the fence */
> - dma_fence_put(fence);
> + s_fence = sched_job->s_fence;
>
> - r = dma_fence_add_callback(fence, &sched_job->cb,
> - drm_sched_job_done_cb);
> - if (r == -ENOENT)
> - drm_sched_job_done(sched_job, fence->error);
> - else if (r)
> - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> - r);
> - } else {
> - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> - PTR_ERR(fence) : 0);
> - }
> + atomic_inc(&sched->hw_rq_count);
> + drm_sched_job_begin(sched_job);
>
> - wake_up(&sched->job_scheduled);
> + trace_drm_run_job(sched_job, entity);
> + fence = sched->ops->run_job(sched_job);
> + complete_all(&entity->entity_idle);
> + drm_sched_fence_scheduled(s_fence, fence);
> +
> + if (!IS_ERR_OR_NULL(fence)) {
> + /* Drop for original kref_init of the fence */
> + dma_fence_put(fence);
> +
> + r = dma_fence_add_callback(fence, &sched_job->cb,
> + drm_sched_job_done_cb);
> + if (r == -ENOENT)
> + drm_sched_job_done(sched_job, fence->error);
> + else if (r)
> + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> + } else {
> + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> + PTR_ERR(fence) : 0);
> }
Just ran into a race condition when using a non-ordered workqueue
for drm_sched:
thread A thread B
drm_sched_run_job_work()
drm_sched_job_begin()
// inserts jobA in pending_list
drm_sched_free_job_work()
drm_sched_get_cleanup_job()
// check first job in pending list
// if s_fence->parent == NULL, consider
// for cleanup
->free_job(jobA)
drm_sched_job_cleanup()
// set sched_job->s_fence = NULL
->run_job()
drm_sched_fence_scheduled()
// sched_job->s_fence->parent = parent_fence
// BOOM => NULL pointer deref
For now, I'll just use a dedicated ordered wq, but if we claim
multi-threaded workqueues are supported, this is probably worth fixing.
I know there's been some discussions about when the timeout should be
started, and the job insertion in the pending_list is kinda related.
If we want this insertion to happen before ->run_job() is called, we
need a way to flag when a job is inserted, but not fully submitted yet.
Regards,
Boris
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-23 12:16 ` Boris Brezillon
@ 2023-10-23 12:39 ` Boris Brezillon
2023-10-23 13:54 ` Matthew Brost
0 siblings, 1 reply; 22+ messages in thread
From: Boris Brezillon @ 2023-10-23 12:39 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, dri-devel, christian.koenig, luben.tuikov,
dakr, donald.robson, lina, intel-xe, faith.ekstrand
On Mon, 23 Oct 2023 14:16:06 +0200
Boris Brezillon <boris.brezillon@collabora.com> wrote:
> Hi,
>
> On Tue, 17 Oct 2023 08:09:56 -0700
> Matthew Brost <matthew.brost@intel.com> wrote:
>
> > +static void drm_sched_run_job_work(struct work_struct *w)
> > +{
> > + struct drm_gpu_scheduler *sched =
> > + container_of(w, struct drm_gpu_scheduler, work_run_job);
> > + struct drm_sched_entity *entity;
> > + struct dma_fence *fence;
> > + struct drm_sched_fence *s_fence;
> > + struct drm_sched_job *sched_job;
> > + int r;
> >
> > - atomic_inc(&sched->hw_rq_count);
> > - drm_sched_job_begin(sched_job);
> > + if (READ_ONCE(sched->pause_submit))
> > + return;
> > +
> > + entity = drm_sched_select_entity(sched, true);
> > + if (!entity)
> > + return;
> >
> > - trace_drm_run_job(sched_job, entity);
> > - fence = sched->ops->run_job(sched_job);
> > + sched_job = drm_sched_entity_pop_job(entity);
> > + if (!sched_job) {
> > complete_all(&entity->entity_idle);
> > - drm_sched_fence_scheduled(s_fence, fence);
> > + return; /* No more work */
> > + }
> >
> > - if (!IS_ERR_OR_NULL(fence)) {
> > - /* Drop for original kref_init of the fence */
> > - dma_fence_put(fence);
> > + s_fence = sched_job->s_fence;
> >
> > - r = dma_fence_add_callback(fence, &sched_job->cb,
> > - drm_sched_job_done_cb);
> > - if (r == -ENOENT)
> > - drm_sched_job_done(sched_job, fence->error);
> > - else if (r)
> > - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > - r);
> > - } else {
> > - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > - PTR_ERR(fence) : 0);
> > - }
> > + atomic_inc(&sched->hw_rq_count);
> > + drm_sched_job_begin(sched_job);
> >
> > - wake_up(&sched->job_scheduled);
> > + trace_drm_run_job(sched_job, entity);
> > + fence = sched->ops->run_job(sched_job);
> > + complete_all(&entity->entity_idle);
> > + drm_sched_fence_scheduled(s_fence, fence);
> > +
> > + if (!IS_ERR_OR_NULL(fence)) {
> > + /* Drop for original kref_init of the fence */
> > + dma_fence_put(fence);
> > +
> > + r = dma_fence_add_callback(fence, &sched_job->cb,
> > + drm_sched_job_done_cb);
> > + if (r == -ENOENT)
> > + drm_sched_job_done(sched_job, fence->error);
> > + else if (r)
> > + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> > + } else {
> > + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > + PTR_ERR(fence) : 0);
> > }
>
> Just ran into a race condition when using a non-ordered workqueue
> for drm_sched:
>
> thread A thread B
>
> drm_sched_run_job_work()
> drm_sched_job_begin()
> // inserts jobA in pending_list
>
> drm_sched_free_job_work()
> drm_sched_get_cleanup_job()
> // check first job in pending list
> // if s_fence->parent == NULL, consider
> // for cleanup
> ->free_job(jobA)
> drm_sched_job_cleanup()
> // set sched_job->s_fence = NULL
>
> ->run_job()
> drm_sched_fence_scheduled()
Correction: the NULL pointer deref happens in drm_sched_job_done()
(when the driver returns an error directly) not in
drm_sched_fence_scheduled(), but the problem remains the same.
> // sched_job->s_fence->parent = parent_fence
> // BOOM => NULL pointer deref
>
> For now, I'll just use a dedicated ordered wq, but if we claim
> multi-threaded workqueues are supported, this is probably worth fixing.
> I know there's been some discussions about when the timeout should be
> started, and the job insertion in the pending_list is kinda related.
> If we want this insertion to happen before ->run_job() is called, we
> need a way to flag when a job is inserted, but not fully submitted yet.
>
> Regards,
>
> Boris
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-23 12:39 ` Boris Brezillon
@ 2023-10-23 13:54 ` Matthew Brost
2023-10-23 14:22 ` Boris Brezillon
0 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-23 13:54 UTC (permalink / raw)
To: Boris Brezillon
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen, lina,
mcanal, Liviu.Dudau, dri-devel, intel-xe, luben.tuikov, dakr,
donald.robson, christian.koenig, faith.ekstrand
On Mon, Oct 23, 2023 at 02:39:37PM +0200, Boris Brezillon wrote:
> On Mon, 23 Oct 2023 14:16:06 +0200
> Boris Brezillon <boris.brezillon@collabora.com> wrote:
>
> > Hi,
> >
> > On Tue, 17 Oct 2023 08:09:56 -0700
> > Matthew Brost <matthew.brost@intel.com> wrote:
> >
> > > +static void drm_sched_run_job_work(struct work_struct *w)
> > > +{
> > > + struct drm_gpu_scheduler *sched =
> > > + container_of(w, struct drm_gpu_scheduler, work_run_job);
> > > + struct drm_sched_entity *entity;
> > > + struct dma_fence *fence;
> > > + struct drm_sched_fence *s_fence;
> > > + struct drm_sched_job *sched_job;
> > > + int r;
> > >
> > > - atomic_inc(&sched->hw_rq_count);
> > > - drm_sched_job_begin(sched_job);
> > > + if (READ_ONCE(sched->pause_submit))
> > > + return;
> > > +
> > > + entity = drm_sched_select_entity(sched, true);
> > > + if (!entity)
> > > + return;
> > >
> > > - trace_drm_run_job(sched_job, entity);
> > > - fence = sched->ops->run_job(sched_job);
> > > + sched_job = drm_sched_entity_pop_job(entity);
> > > + if (!sched_job) {
> > > complete_all(&entity->entity_idle);
> > > - drm_sched_fence_scheduled(s_fence, fence);
> > > + return; /* No more work */
> > > + }
> > >
> > > - if (!IS_ERR_OR_NULL(fence)) {
> > > - /* Drop for original kref_init of the fence */
> > > - dma_fence_put(fence);
> > > + s_fence = sched_job->s_fence;
> > >
> > > - r = dma_fence_add_callback(fence, &sched_job->cb,
> > > - drm_sched_job_done_cb);
> > > - if (r == -ENOENT)
> > > - drm_sched_job_done(sched_job, fence->error);
> > > - else if (r)
> > > - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > > - r);
> > > - } else {
> > > - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > - PTR_ERR(fence) : 0);
> > > - }
> > > + atomic_inc(&sched->hw_rq_count);
> > > + drm_sched_job_begin(sched_job);
> > >
> > > - wake_up(&sched->job_scheduled);
> > > + trace_drm_run_job(sched_job, entity);
> > > + fence = sched->ops->run_job(sched_job);
> > > + complete_all(&entity->entity_idle);
> > > + drm_sched_fence_scheduled(s_fence, fence);
> > > +
> > > + if (!IS_ERR_OR_NULL(fence)) {
> > > + /* Drop for original kref_init of the fence */
> > > + dma_fence_put(fence);
> > > +
> > > + r = dma_fence_add_callback(fence, &sched_job->cb,
> > > + drm_sched_job_done_cb);
> > > + if (r == -ENOENT)
> > > + drm_sched_job_done(sched_job, fence->error);
> > > + else if (r)
> > > + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> > > + } else {
> > > + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > + PTR_ERR(fence) : 0);
> > > }
> >
> > Just ran into a race condition when using a non-ordered workqueue
> > for drm_sched:
> >
> > thread A thread B
> >
> > drm_sched_run_job_work()
> > drm_sched_job_begin()
> > // inserts jobA in pending_list
> >
> > drm_sched_free_job_work()
> > drm_sched_get_cleanup_job()
> > // check first job in pending list
> > // if s_fence->parent == NULL, consider
> > // for cleanup
> > ->free_job(jobA)
> > drm_sched_job_cleanup()
> > // set sched_job->s_fence = NULL
> >
> > ->run_job()
> > drm_sched_fence_scheduled()
>
> Correction: the NULL pointer deref happens in drm_sched_job_done()
> (when the driver returns an error directly) not in
> drm_sched_fence_scheduled(), but the problem remains the same.
>
>
Trying to understand this. I don't see how drm_sched_get_cleanup_job can
return a job until dma_fence_is_signaled(&job->s_fence->finished) is
true. That fence is no signaled until drm_sched_fence_finished(s_fence,
result); called in drm_sched_job_done().
What am I missing here?
Matt
> > // sched_job->s_fence->parent = parent_fence
> > // BOOM => NULL pointer deref
> >
> > For now, I'll just use a dedicated ordered wq, but if we claim
> > multi-threaded workqueues are supported, this is probably worth fixing.
> > I know there's been some discussions about when the timeout should be
> > started, and the job insertion in the pending_list is kinda related.
> > If we want this insertion to happen before ->run_job() is called, we
> > need a way to flag when a job is inserted, but not fully submitted yet.
> >
> > Regards,
> >
> > Boris
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-23 13:54 ` Matthew Brost
@ 2023-10-23 14:22 ` Boris Brezillon
2023-10-23 22:03 ` Matthew Brost
0 siblings, 1 reply; 22+ messages in thread
From: Boris Brezillon @ 2023-10-23 14:22 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen, lina,
mcanal, Liviu.Dudau, dri-devel, intel-xe, luben.tuikov, dakr,
donald.robson, christian.koenig, faith.ekstrand
On Mon, 23 Oct 2023 13:54:13 +0000
Matthew Brost <matthew.brost@intel.com> wrote:
> On Mon, Oct 23, 2023 at 02:39:37PM +0200, Boris Brezillon wrote:
> > On Mon, 23 Oct 2023 14:16:06 +0200
> > Boris Brezillon <boris.brezillon@collabora.com> wrote:
> >
> > > Hi,
> > >
> > > On Tue, 17 Oct 2023 08:09:56 -0700
> > > Matthew Brost <matthew.brost@intel.com> wrote:
> > >
> > > > +static void drm_sched_run_job_work(struct work_struct *w)
> > > > +{
> > > > + struct drm_gpu_scheduler *sched =
> > > > + container_of(w, struct drm_gpu_scheduler, work_run_job);
> > > > + struct drm_sched_entity *entity;
> > > > + struct dma_fence *fence;
> > > > + struct drm_sched_fence *s_fence;
> > > > + struct drm_sched_job *sched_job;
> > > > + int r;
> > > >
> > > > - atomic_inc(&sched->hw_rq_count);
> > > > - drm_sched_job_begin(sched_job);
> > > > + if (READ_ONCE(sched->pause_submit))
> > > > + return;
> > > > +
> > > > + entity = drm_sched_select_entity(sched, true);
> > > > + if (!entity)
> > > > + return;
> > > >
> > > > - trace_drm_run_job(sched_job, entity);
> > > > - fence = sched->ops->run_job(sched_job);
> > > > + sched_job = drm_sched_entity_pop_job(entity);
> > > > + if (!sched_job) {
> > > > complete_all(&entity->entity_idle);
> > > > - drm_sched_fence_scheduled(s_fence, fence);
> > > > + return; /* No more work */
> > > > + }
> > > >
> > > > - if (!IS_ERR_OR_NULL(fence)) {
> > > > - /* Drop for original kref_init of the fence */
> > > > - dma_fence_put(fence);
> > > > + s_fence = sched_job->s_fence;
> > > >
> > > > - r = dma_fence_add_callback(fence, &sched_job->cb,
> > > > - drm_sched_job_done_cb);
> > > > - if (r == -ENOENT)
> > > > - drm_sched_job_done(sched_job, fence->error);
> > > > - else if (r)
> > > > - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > > > - r);
> > > > - } else {
> > > > - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > > - PTR_ERR(fence) : 0);
> > > > - }
> > > > + atomic_inc(&sched->hw_rq_count);
> > > > + drm_sched_job_begin(sched_job);
> > > >
> > > > - wake_up(&sched->job_scheduled);
> > > > + trace_drm_run_job(sched_job, entity);
> > > > + fence = sched->ops->run_job(sched_job);
> > > > + complete_all(&entity->entity_idle);
> > > > + drm_sched_fence_scheduled(s_fence, fence);
> > > > +
> > > > + if (!IS_ERR_OR_NULL(fence)) {
> > > > + /* Drop for original kref_init of the fence */
> > > > + dma_fence_put(fence);
> > > > +
> > > > + r = dma_fence_add_callback(fence, &sched_job->cb,
> > > > + drm_sched_job_done_cb);
> > > > + if (r == -ENOENT)
> > > > + drm_sched_job_done(sched_job, fence->error);
> > > > + else if (r)
> > > > + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> > > > + } else {
> > > > + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > > + PTR_ERR(fence) : 0);
> > > > }
> > >
> > > Just ran into a race condition when using a non-ordered workqueue
> > > for drm_sched:
> > >
> > > thread A thread B
> > >
> > > drm_sched_run_job_work()
> > > drm_sched_job_begin()
> > > // inserts jobA in pending_list
> > >
> > > drm_sched_free_job_work()
> > > drm_sched_get_cleanup_job()
> > > // check first job in pending list
> > > // if s_fence->parent == NULL, consider
> > > // for cleanup
> > > ->free_job(jobA)
> > > drm_sched_job_cleanup()
> > > // set sched_job->s_fence = NULL
> > >
> > > ->run_job()
> > > drm_sched_fence_scheduled()
> >
> > Correction: the NULL pointer deref happens in drm_sched_job_done()
> > (when the driver returns an error directly) not in
> > drm_sched_fence_scheduled(), but the problem remains the same.
> >
> >
>
> Trying to understand this. I don't see how drm_sched_get_cleanup_job can
> return a job until dma_fence_is_signaled(&job->s_fence->finished) is
> true. That fence is no signaled until drm_sched_fence_finished(s_fence,
> result); called in drm_sched_job_done().
>
> What am I missing here?
Uh, my bad. I was trying to backport panthor (and all its deps) to a 6.1
kernel, and the condition in drm_sched_get_cleanup_job() is different
there:
if (job && (!job->s_fence->parent ||
dma_fence_is_signaled(job->s_fence->parent))) {
// pick this job for cleanup
}
Sorry for the noise.
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 5/7] drm/sched: Split free_job into own work item
2023-10-23 14:22 ` Boris Brezillon
@ 2023-10-23 22:03 ` Matthew Brost
0 siblings, 0 replies; 22+ messages in thread
From: Matthew Brost @ 2023-10-23 22:03 UTC (permalink / raw)
To: Boris Brezillon
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen, lina,
mcanal, Liviu.Dudau, dri-devel, intel-xe, luben.tuikov, dakr,
donald.robson, christian.koenig, faith.ekstrand
On Mon, Oct 23, 2023 at 04:22:16PM +0200, Boris Brezillon wrote:
> On Mon, 23 Oct 2023 13:54:13 +0000
> Matthew Brost <matthew.brost@intel.com> wrote:
>
> > On Mon, Oct 23, 2023 at 02:39:37PM +0200, Boris Brezillon wrote:
> > > On Mon, 23 Oct 2023 14:16:06 +0200
> > > Boris Brezillon <boris.brezillon@collabora.com> wrote:
> > >
> > > > Hi,
> > > >
> > > > On Tue, 17 Oct 2023 08:09:56 -0700
> > > > Matthew Brost <matthew.brost@intel.com> wrote:
> > > >
> > > > > +static void drm_sched_run_job_work(struct work_struct *w)
> > > > > +{
> > > > > + struct drm_gpu_scheduler *sched =
> > > > > + container_of(w, struct drm_gpu_scheduler, work_run_job);
> > > > > + struct drm_sched_entity *entity;
> > > > > + struct dma_fence *fence;
> > > > > + struct drm_sched_fence *s_fence;
> > > > > + struct drm_sched_job *sched_job;
> > > > > + int r;
> > > > >
> > > > > - atomic_inc(&sched->hw_rq_count);
> > > > > - drm_sched_job_begin(sched_job);
> > > > > + if (READ_ONCE(sched->pause_submit))
> > > > > + return;
> > > > > +
> > > > > + entity = drm_sched_select_entity(sched, true);
> > > > > + if (!entity)
> > > > > + return;
> > > > >
> > > > > - trace_drm_run_job(sched_job, entity);
> > > > > - fence = sched->ops->run_job(sched_job);
> > > > > + sched_job = drm_sched_entity_pop_job(entity);
> > > > > + if (!sched_job) {
> > > > > complete_all(&entity->entity_idle);
> > > > > - drm_sched_fence_scheduled(s_fence, fence);
> > > > > + return; /* No more work */
> > > > > + }
> > > > >
> > > > > - if (!IS_ERR_OR_NULL(fence)) {
> > > > > - /* Drop for original kref_init of the fence */
> > > > > - dma_fence_put(fence);
> > > > > + s_fence = sched_job->s_fence;
> > > > >
> > > > > - r = dma_fence_add_callback(fence, &sched_job->cb,
> > > > > - drm_sched_job_done_cb);
> > > > > - if (r == -ENOENT)
> > > > > - drm_sched_job_done(sched_job, fence->error);
> > > > > - else if (r)
> > > > > - DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n",
> > > > > - r);
> > > > > - } else {
> > > > > - drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > > > - PTR_ERR(fence) : 0);
> > > > > - }
> > > > > + atomic_inc(&sched->hw_rq_count);
> > > > > + drm_sched_job_begin(sched_job);
> > > > >
> > > > > - wake_up(&sched->job_scheduled);
> > > > > + trace_drm_run_job(sched_job, entity);
> > > > > + fence = sched->ops->run_job(sched_job);
> > > > > + complete_all(&entity->entity_idle);
> > > > > + drm_sched_fence_scheduled(s_fence, fence);
> > > > > +
> > > > > + if (!IS_ERR_OR_NULL(fence)) {
> > > > > + /* Drop for original kref_init of the fence */
> > > > > + dma_fence_put(fence);
> > > > > +
> > > > > + r = dma_fence_add_callback(fence, &sched_job->cb,
> > > > > + drm_sched_job_done_cb);
> > > > > + if (r == -ENOENT)
> > > > > + drm_sched_job_done(sched_job, fence->error);
> > > > > + else if (r)
> > > > > + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r);
> > > > > + } else {
> > > > > + drm_sched_job_done(sched_job, IS_ERR(fence) ?
> > > > > + PTR_ERR(fence) : 0);
> > > > > }
> > > >
> > > > Just ran into a race condition when using a non-ordered workqueue
> > > > for drm_sched:
> > > >
> > > > thread A thread B
> > > >
> > > > drm_sched_run_job_work()
> > > > drm_sched_job_begin()
> > > > // inserts jobA in pending_list
> > > >
> > > > drm_sched_free_job_work()
> > > > drm_sched_get_cleanup_job()
> > > > // check first job in pending list
> > > > // if s_fence->parent == NULL, consider
> > > > // for cleanup
> > > > ->free_job(jobA)
> > > > drm_sched_job_cleanup()
> > > > // set sched_job->s_fence = NULL
> > > >
> > > > ->run_job()
> > > > drm_sched_fence_scheduled()
> > >
> > > Correction: the NULL pointer deref happens in drm_sched_job_done()
> > > (when the driver returns an error directly) not in
> > > drm_sched_fence_scheduled(), but the problem remains the same.
> > >
> > >
> >
> > Trying to understand this. I don't see how drm_sched_get_cleanup_job can
> > return a job until dma_fence_is_signaled(&job->s_fence->finished) is
> > true. That fence is no signaled until drm_sched_fence_finished(s_fence,
> > result); called in drm_sched_job_done().
> >
> > What am I missing here?
>
> Uh, my bad. I was trying to backport panthor (and all its deps) to a 6.1
> kernel, and the condition in drm_sched_get_cleanup_job() is different
> there:
>
> if (job && (!job->s_fence->parent ||
> dma_fence_is_signaled(job->s_fence->parent))) {
> // pick this job for cleanup
> }
>
> Sorry for the noise.
No problem, glad this is resolved.
Matt
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler
2023-10-17 15:09 ` [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler Matthew Brost
@ 2023-10-24 3:34 ` Luben Tuikov
0 siblings, 0 replies; 22+ messages in thread
From: Luben Tuikov @ 2023-10-24 3:34 UTC (permalink / raw)
To: Matthew Brost, dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, boris.brezillon, dakr, donald.robson, lina,
christian.koenig, faith.ekstrand
On 2023-10-17 11:09, Matthew Brost wrote:
> Rather than a global modparam for scheduling policy, move the scheduling
> policy to scheduler so user can control each scheduler policy.
>
> v2:
> - s/DRM_SCHED_POLICY_MAX/DRM_SCHED_POLICY_COUNT (Luben)
> - Only include policy in scheduler (Luben)
> v3:
> - use a ternary operator as opposed to an if-control (Luben)
> - s/DRM_SCHED_POLICY_DEFAULT/DRM_SCHED_POLICY_UNSET/ (Luben)
> - s/default_drm_sched_policy/drm_sched_policy_default/ (Luben)
> - Update commit message (Boris)
> - Fix v3d build (CI)
> - s/bad_policies/drm_sched_policy_mismatch/ (Luben)
> - Don't update modparam doc (Luben)
> v4:
> - Fix alignment in msm_ringbuffer_new (Luben / checkpatch)
>
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 +
> drivers/gpu/drm/etnaviv/etnaviv_sched.c | 3 ++-
> drivers/gpu/drm/lima/lima_sched.c | 3 ++-
> drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +-
> drivers/gpu/drm/nouveau/nouveau_sched.c | 3 ++-
> drivers/gpu/drm/panfrost/panfrost_job.c | 3 ++-
> drivers/gpu/drm/scheduler/sched_entity.c | 24 ++++++++++++++++++----
> drivers/gpu/drm/scheduler/sched_main.c | 19 ++++++++++++-----
> drivers/gpu/drm/v3d/v3d_sched.c | 15 +++++++++-----
> include/drm/gpu_scheduler.h | 20 ++++++++++++------
> 10 files changed, 68 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index b54c4d771104..e4e6f91450a4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -2283,6 +2283,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
> ring->num_hw_submission, 0,
> timeout, adev->reset_domain->wq,
> ring->sched_score, ring->name,
> + DRM_SCHED_POLICY_UNSET,
> adev->dev);
I think we should drop this patch.
Drivers shouldn't be able to select their own policy, when there's a kernel parameter,
which says what the scheduling policy should be. Imagine you're a user setting
the policy at kernel command line, only to learn that some driver has decided
to ignore (programmed, mind you) your choice and set whatever it decides. Plus,
this opens the Pandora box for other drivers to do this, and it's not a good
software engineering direction.
For the 1-1-1 case in S-R-E (sched-rq-entity) which Xe is using, DRM scheduling policy is
irrelevant. We can show this by showing that,
a) In the case of RR, we always sample the only entity there, namely the single
entity in the single run-queue.
b) In the case of FIFO, we always pick the only node in the RB tree, namely the single
entity in the single run-queue.
So whether it is RR or FIFO, for the 1-1-1 case, it always works as expected, and
the actual selection policy (scheduling policy) is irrelevant. This is a good design.
However, what prevents the Xe driver from doing this cleanly, is this ghastly
array of "priorities" in the scheduler struct,
struct drm_gpu_scheduler {
...
struct drm_sched_rq sched_rq[DRM_SCHED_PRIORITY_COUNT];
...
which is very speculative and ambitious... Why should the scheduler have
a set, constant, number of "priorities" and no more or less? Who said that
those are _the_only_ options, and why? It makes for a very rigid design
which doesn't yield well to novel and varied implementations, and that's
not a sign of a good design.
With the "[PATCH] drm/sched: Convert the GPU scheduler to variable number
of run-queues" (https://lore.kernel.org/r/20231023032251.164775-1-luben.tuikov@amd.com)
drivers can specify the number of run-queues in the S-R-E relationship.
Note, that in the S-R-E relationship, the driver already controls
the number of E's. With the patch above, drivers will also control the number of R's,
and I think that will have much more flexible implications for drivers to play with
rather than keeping the R constant.
The idea is that the Xe driver would specify that they're using a scheduler
with a single R, at drm_sched_init() and then attach a single E to that,
leaving much of the rest of the code alone. This generalizes (leaves alone) the rest of
the code, which is a good design.
There's a bug in the amdgpu code which when using dynamic rq, it oopses, because
it was using the scheduler without ever calling drm_sched_init()--the only reason
amdgpu was getting away with it was because the array was statically defined. I've posted
a patch for that. (https://lore.kernel.org/r/20231023032344.164925-1-luben.tuikov@amd.com)
(There was also another bug where amdgpu did sched_rq[-2], and the fix is already in,
fa8391ad68c167.)
Regards,
Luben
> if (r) {
> DRM_ERROR("Failed to create scheduler on ring %s.\n",
> diff --git a/drivers/gpu/drm/etnaviv/etnaviv_sched.c b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
> index 618a804ddc34..15b0e2f1abe5 100644
> --- a/drivers/gpu/drm/etnaviv/etnaviv_sched.c
> +++ b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
> @@ -137,7 +137,8 @@ int etnaviv_sched_init(struct etnaviv_gpu *gpu)
> ret = drm_sched_init(&gpu->sched, &etnaviv_sched_ops, NULL,
> etnaviv_hw_jobs_limit, etnaviv_job_hang_limit,
> msecs_to_jiffies(500), NULL, NULL,
> - dev_name(gpu->dev), gpu->dev);
> + dev_name(gpu->dev), DRM_SCHED_POLICY_UNSET,
> + gpu->dev);
> if (ret)
> return ret;
>
> diff --git a/drivers/gpu/drm/lima/lima_sched.c b/drivers/gpu/drm/lima/lima_sched.c
> index 8d858aed0e56..50c2075228aa 100644
> --- a/drivers/gpu/drm/lima/lima_sched.c
> +++ b/drivers/gpu/drm/lima/lima_sched.c
> @@ -491,7 +491,8 @@ int lima_sched_pipe_init(struct lima_sched_pipe *pipe, const char *name)
> return drm_sched_init(&pipe->base, &lima_sched_ops, NULL, 1,
> lima_job_hang_limit,
> msecs_to_jiffies(timeout), NULL,
> - NULL, name, pipe->ldev->dev);
> + NULL, name, DRM_SCHED_POLICY_UNSET,
> + pipe->ldev->dev);
> }
>
> void lima_sched_pipe_fini(struct lima_sched_pipe *pipe)
> diff --git a/drivers/gpu/drm/msm/msm_ringbuffer.c b/drivers/gpu/drm/msm/msm_ringbuffer.c
> index 1097f8e93d6b..173ad2f17c50 100644
> --- a/drivers/gpu/drm/msm/msm_ringbuffer.c
> +++ b/drivers/gpu/drm/msm/msm_ringbuffer.c
> @@ -97,7 +97,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
> ret = drm_sched_init(&ring->sched, &msm_sched_ops, NULL,
> num_hw_submissions, 0, sched_timeout,
> NULL, NULL, to_msm_bo(ring->bo)->name,
> - gpu->dev->dev);
> + DRM_SCHED_POLICY_UNSET, gpu->dev->dev);
> if (ret) {
> goto fail;
> }
> diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c b/drivers/gpu/drm/nouveau/nouveau_sched.c
> index 4c959dec42b3..c4e09d2e77f9 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_sched.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
> @@ -437,7 +437,8 @@ int nouveau_sched_init(struct nouveau_drm *drm)
>
> return drm_sched_init(sched, &nouveau_sched_ops, NULL,
> NOUVEAU_SCHED_HW_SUBMISSIONS, 0, job_hang_limit,
> - NULL, NULL, "nouveau_sched", drm->dev->dev);
> + NULL, NULL, "nouveau_sched",
> + DRM_SCHED_POLICY_UNSET, drm->dev->dev);
> }
>
> void nouveau_sched_fini(struct nouveau_drm *drm)
> diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
> index 934b7b930c76..95330ff402ba 100644
> --- a/drivers/gpu/drm/panfrost/panfrost_job.c
> +++ b/drivers/gpu/drm/panfrost/panfrost_job.c
> @@ -856,7 +856,8 @@ int panfrost_job_init(struct panfrost_device *pfdev)
> nentries, 0,
> msecs_to_jiffies(JOB_TIMEOUT_MS),
> pfdev->reset.wq,
> - NULL, "pan_js", pfdev->dev);
> + NULL, "pan_js", DRM_SCHED_POLICY_UNSET,
> + pfdev->dev);
> if (ret) {
> dev_err(pfdev->dev, "Failed to create scheduler: %d.", ret);
> goto err_sched;
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index a42763e1429d..cf42e2265d64 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -33,6 +33,20 @@
> #define to_drm_sched_job(sched_job) \
> container_of((sched_job), struct drm_sched_job, queue_node)
>
> +static bool drm_sched_policy_mismatch(struct drm_gpu_scheduler **sched_list,
> + unsigned int num_sched_list)
> +{
> + enum drm_sched_policy sched_policy = sched_list[0]->sched_policy;
> + unsigned int i;
> +
> + /* All schedule policies must match */
> + for (i = 1; i < num_sched_list; ++i)
> + if (sched_policy != sched_list[i]->sched_policy)
> + return true;
> +
> + return false;
> +}
> +
> /**
> * drm_sched_entity_init - Init a context entity used by scheduler when
> * submit to HW ring.
> @@ -62,7 +76,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> unsigned int num_sched_list,
> atomic_t *guilty)
> {
> - if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])))
> + if (!(entity && sched_list && (num_sched_list == 0 || sched_list[0])) ||
> + drm_sched_policy_mismatch(sched_list, num_sched_list))
> return -EINVAL;
>
> memset(entity, 0, sizeof(struct drm_sched_entity));
> @@ -486,7 +501,7 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
> * Update the entity's location in the min heap according to
> * the timestamp of the next job, if any.
> */
> - if (drm_sched_policy == DRM_SCHED_POLICY_FIFO) {
> + if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
> struct drm_sched_job *next;
>
> next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
> @@ -558,7 +573,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> {
> struct drm_sched_entity *entity = sched_job->entity;
> - bool first;
> + bool first, fifo = entity->rq->sched->sched_policy ==
> + DRM_SCHED_POLICY_FIFO;
> ktime_t submit_ts;
>
> trace_drm_sched_job(sched_job, entity);
> @@ -587,7 +603,7 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> drm_sched_rq_add_entity(entity->rq, entity);
> spin_unlock(&entity->rq_lock);
>
> - if (drm_sched_policy == DRM_SCHED_POLICY_FIFO)
> + if (fifo)
> drm_sched_rq_update_fifo(entity, submit_ts);
>
> drm_sched_wakeup_if_can_queue(entity->rq->sched);
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 8b1d52cff1e9..150e5330f0fa 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -66,14 +66,14 @@
> #define to_drm_sched_job(sched_job) \
> container_of((sched_job), struct drm_sched_job, queue_node)
>
> -int drm_sched_policy = DRM_SCHED_POLICY_FIFO;
> +int drm_sched_policy_default = DRM_SCHED_POLICY_FIFO;
>
> /**
> * DOC: sched_policy (int)
> * Used to override default entities scheduling policy in a run queue.
> */
> MODULE_PARM_DESC(sched_policy, "Specify the scheduling policy for entities on a run-queue, " __stringify(DRM_SCHED_POLICY_RR) " = Round Robin, " __stringify(DRM_SCHED_POLICY_FIFO) " = FIFO (default).");
> -module_param_named(sched_policy, drm_sched_policy, int, 0444);
> +module_param_named(sched_policy, drm_sched_policy_default, int, 0444);
>
> static __always_inline bool drm_sched_entity_compare_before(struct rb_node *a,
> const struct rb_node *b)
> @@ -177,7 +177,7 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
> if (rq->current_entity == entity)
> rq->current_entity = NULL;
>
> - if (drm_sched_policy == DRM_SCHED_POLICY_FIFO)
> + if (rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO)
> drm_sched_rq_remove_fifo_locked(entity);
>
> spin_unlock(&rq->lock);
> @@ -898,7 +898,7 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
>
> /* Kernel run queue has higher priority than normal run queue*/
> for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> - entity = drm_sched_policy == DRM_SCHED_POLICY_FIFO ?
> + entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> drm_sched_rq_select_entity_fifo(&sched->sched_rq[i]) :
> drm_sched_rq_select_entity_rr(&sched->sched_rq[i]);
> if (entity)
> @@ -1072,6 +1072,7 @@ static void drm_sched_main(struct work_struct *w)
> * used
> * @score: optional score atomic shared with other schedulers
> * @name: name used for debugging
> + * @sched_policy: schedule policy
> * @dev: target &struct device
> *
> * Return 0 on success, otherwise error code.
> @@ -1081,9 +1082,15 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> struct workqueue_struct *submit_wq,
> unsigned hw_submission, unsigned hang_limit,
> long timeout, struct workqueue_struct *timeout_wq,
> - atomic_t *score, const char *name, struct device *dev)
> + atomic_t *score, const char *name,
> + enum drm_sched_policy sched_policy,
> + struct device *dev)
> {
> int i;
> +
> + if (sched_policy >= DRM_SCHED_POLICY_COUNT)
> + return -EINVAL;
> +
> sched->ops = ops;
> sched->hw_submission_limit = hw_submission;
> sched->name = name;
> @@ -1102,6 +1109,8 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> sched->hang_limit = hang_limit;
> sched->score = score ? score : &sched->_score;
> sched->dev = dev;
> + sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
> + drm_sched_policy_default : sched_policy;
> for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
> drm_sched_rq_init(sched, &sched->sched_rq[i]);
>
> diff --git a/drivers/gpu/drm/v3d/v3d_sched.c b/drivers/gpu/drm/v3d/v3d_sched.c
> index 38e092ea41e6..dec89c5b8cb1 100644
> --- a/drivers/gpu/drm/v3d/v3d_sched.c
> +++ b/drivers/gpu/drm/v3d/v3d_sched.c
> @@ -391,7 +391,8 @@ v3d_sched_init(struct v3d_dev *v3d)
> &v3d_bin_sched_ops, NULL,
> hw_jobs_limit, job_hang_limit,
> msecs_to_jiffies(hang_limit_ms), NULL,
> - NULL, "v3d_bin", v3d->drm.dev);
> + NULL, "v3d_bin", DRM_SCHED_POLICY_UNSET,
> + v3d->drm.dev);
> if (ret)
> return ret;
>
> @@ -399,7 +400,8 @@ v3d_sched_init(struct v3d_dev *v3d)
> &v3d_render_sched_ops, NULL,
> hw_jobs_limit, job_hang_limit,
> msecs_to_jiffies(hang_limit_ms), NULL,
> - NULL, "v3d_render", v3d->drm.dev);
> + NULL, "v3d_render", DRM_SCHED_POLICY_UNSET,
> + v3d->drm.dev);
> if (ret)
> goto fail;
>
> @@ -407,7 +409,8 @@ v3d_sched_init(struct v3d_dev *v3d)
> &v3d_tfu_sched_ops, NULL,
> hw_jobs_limit, job_hang_limit,
> msecs_to_jiffies(hang_limit_ms), NULL,
> - NULL, "v3d_tfu", v3d->drm.dev);
> + NULL, "v3d_tfu", DRM_SCHED_POLICY_UNSET,
> + v3d->drm.dev);
> if (ret)
> goto fail;
>
> @@ -416,7 +419,8 @@ v3d_sched_init(struct v3d_dev *v3d)
> &v3d_csd_sched_ops, NULL,
> hw_jobs_limit, job_hang_limit,
> msecs_to_jiffies(hang_limit_ms), NULL,
> - NULL, "v3d_csd", v3d->drm.dev);
> + NULL, "v3d_csd", DRM_SCHED_POLICY_UNSET,
> + v3d->drm.dev);
> if (ret)
> goto fail;
>
> @@ -424,7 +428,8 @@ v3d_sched_init(struct v3d_dev *v3d)
> &v3d_cache_clean_sched_ops, NULL,
> hw_jobs_limit, job_hang_limit,
> msecs_to_jiffies(hang_limit_ms), NULL,
> - NULL, "v3d_cache_clean", v3d->drm.dev);
> + NULL, "v3d_cache_clean",
> + DRM_SCHED_POLICY_UNSET, v3d->drm.dev);
> if (ret)
> goto fail;
> }
> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> index 211bd3cdabdc..320f0a720486 100644
> --- a/include/drm/gpu_scheduler.h
> +++ b/include/drm/gpu_scheduler.h
> @@ -72,11 +72,15 @@ enum drm_sched_priority {
> DRM_SCHED_PRIORITY_UNSET = -2
> };
>
> -/* Used to chose between FIFO and RR jobs scheduling */
> -extern int drm_sched_policy;
> -
> -#define DRM_SCHED_POLICY_RR 0
> -#define DRM_SCHED_POLICY_FIFO 1
> +/* Used to chose default scheduling policy*/
> +extern int default_drm_sched_policy;
> +
> +enum drm_sched_policy {
> + DRM_SCHED_POLICY_UNSET,
> + DRM_SCHED_POLICY_RR,
> + DRM_SCHED_POLICY_FIFO,
> + DRM_SCHED_POLICY_COUNT,
> +};
>
> /**
> * struct drm_sched_entity - A wrapper around a job queue (typically
> @@ -489,6 +493,7 @@ struct drm_sched_backend_ops {
> * guilty and it will no longer be considered for scheduling.
> * @score: score to help loadbalancer pick a idle sched
> * @_score: score used when the driver doesn't provide one
> + * @sched_policy: Schedule policy for scheduler
> * @ready: marks if the underlying HW is ready to work
> * @free_guilty: A hit to time out handler to free the guilty job.
> * @pause_submit: pause queuing of @work_submit on @submit_wq
> @@ -515,6 +520,7 @@ struct drm_gpu_scheduler {
> int hang_limit;
> atomic_t *score;
> atomic_t _score;
> + enum drm_sched_policy sched_policy;
> bool ready;
> bool free_guilty;
> bool pause_submit;
> @@ -527,7 +533,9 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> struct workqueue_struct *submit_wq,
> uint32_t hw_submission, unsigned hang_limit,
> long timeout, struct workqueue_struct *timeout_wq,
> - atomic_t *score, const char *name, struct device *dev);
> + atomic_t *score, const char *name,
> + enum drm_sched_policy sched_policy,
> + struct device *dev);
>
> void drm_sched_fini(struct drm_gpu_scheduler *sched);
> int drm_sched_job_init(struct drm_sched_job *job,
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
2023-10-17 15:09 ` [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy Matthew Brost
@ 2023-10-24 3:50 ` Luben Tuikov
2023-10-25 15:13 ` Matthew Brost
0 siblings, 1 reply; 22+ messages in thread
From: Luben Tuikov @ 2023-10-24 3:50 UTC (permalink / raw)
To: Matthew Brost, dri-devel, intel-xe
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen,
Liviu.Dudau, mcanal, boris.brezillon, dakr, donald.robson, lina,
christian.koenig, faith.ekstrand
Hi,
On 2023-10-17 11:09, Matthew Brost wrote:
> DRM_SCHED_POLICY_SINGLE_ENTITY creates a 1 to 1 relationship between
> scheduler and entity. No priorities or run queue used in this mode.
> Intended for devices with firmware schedulers.
>
> v2:
> - Drop sched / rq union (Luben)
> v3:
> - Don't pick entity if stopped in drm_sched_select_entity (Danilo)
> v4:
> - Rework if statement in drm_sched_entity_init (Luben)
> - Update comment for drm_sched_entity_to_scheduler (Luben)
> - Reword a few things in DOC comment (Luben)
> - Do not check sched policy in for statement (Luben)
> v5:
> - Fix extra blank lines (Luben / Checkpatch)
>
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> ---
> drivers/gpu/drm/scheduler/sched_entity.c | 69 +++++++++++++++----
> drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
> drivers/gpu/drm/scheduler/sched_main.c | 87 ++++++++++++++++++------
> include/drm/gpu_scheduler.h | 8 +++
> 4 files changed, 130 insertions(+), 36 deletions(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index cf42e2265d64..940f63dd6965 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -83,6 +83,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> memset(entity, 0, sizeof(struct drm_sched_entity));
> INIT_LIST_HEAD(&entity->list);
> entity->rq = NULL;
> + entity->single_sched = NULL;
> entity->guilty = guilty;
> entity->num_sched_list = num_sched_list;
> entity->priority = priority;
> @@ -90,8 +91,17 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> RCU_INIT_POINTER(entity->last_scheduled, NULL);
> RB_CLEAR_NODE(&entity->rb_tree_node);
>
> - if(num_sched_list)
> - entity->rq = &sched_list[0]->sched_rq[entity->priority];
> + if (num_sched_list) {
> + if (sched_list[0]->sched_policy !=
> + DRM_SCHED_POLICY_SINGLE_ENTITY) {
> + entity->rq = &sched_list[0]->sched_rq[entity->priority];
> + } else if (num_sched_list == 1 && !sched_list[0]->single_entity) {
> + sched_list[0]->single_entity = entity;
> + entity->single_sched = sched_list[0];
To simplify the rest of the GPU scheduler design and generalize the logic,
we can do
entity->rq = sched_list[0]->sched_rq[entity->priority];
where the "priority" is 0, thus having a single rq.
We shouldn't splice scheduler and entity, but rather make it so that
we can set the number of rqs to 1, thus resulting in a single rq.
(https://lore.kernel.org/r/20231023032251.164775-1-luben.tuikov@amd.com)
> + } else {
> + return -EINVAL;
> + }
> + }
>
> init_completion(&entity->entity_idle);
>
> @@ -124,7 +134,8 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
> struct drm_gpu_scheduler **sched_list,
> unsigned int num_sched_list)
> {
> - WARN_ON(!num_sched_list || !sched_list);
> + WARN_ON(!num_sched_list || !sched_list ||
> + !!entity->single_sched);
We wouldn't need this check.
>
> entity->sched_list = sched_list;
> entity->num_sched_list = num_sched_list;
> @@ -231,13 +242,15 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
> {
> struct drm_sched_job *job;
> struct dma_fence *prev;
> + bool single_entity = !!entity->single_sched;
>
> - if (!entity->rq)
> + if (!entity->rq && !single_entity)
> return;
>
> spin_lock(&entity->rq_lock);
> entity->stopped = true;
> - drm_sched_rq_remove_entity(entity->rq, entity);
> + if (!single_entity)
> + drm_sched_rq_remove_entity(entity->rq, entity);
> spin_unlock(&entity->rq_lock);
We should be able to carry on the existing infrastructure when
having num_rqs = 1, with dynamic rqs. So we shouldn't warrant
any changes here.
>
> /* Make sure this entity is not used by the scheduler at the moment */
> @@ -259,6 +272,20 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
> dma_fence_put(prev);
> }
>
> +/**
> + * drm_sched_entity_to_scheduler - Schedule entity to GPU scheduler
> + * @entity: scheduler entity
> + *
> + * Returns GPU scheduler for the entity
> + */
> +struct drm_gpu_scheduler *
> +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity)
> +{
> + bool single_entity = !!entity->single_sched;
> +
> + return single_entity ? entity->single_sched : entity->rq->sched;
It would be "entity->rq->sched" for any and all cases which simplifies things.
> +}
> +
> /**
> * drm_sched_entity_flush - Flush a context entity
> *
> @@ -276,11 +303,12 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
> struct drm_gpu_scheduler *sched;
> struct task_struct *last_user;
> long ret = timeout;
> + bool single_entity = !!entity->single_sched;
>
> - if (!entity->rq)
> + if (!entity->rq && !single_entity)
> return 0;
>
> - sched = entity->rq->sched;
> + sched = drm_sched_entity_to_scheduler(entity);
> /**
> * The client will not queue more IBs during this fini, consume existing
> * queued IBs or discard them on SIGKILL
> @@ -373,7 +401,7 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
> container_of(cb, struct drm_sched_entity, cb);
>
> drm_sched_entity_clear_dep(f, cb);
> - drm_sched_wakeup_if_can_queue(entity->rq->sched);
> + drm_sched_wakeup_if_can_queue(drm_sched_entity_to_scheduler(entity));
> }
We can carry on that too, without changes. The good part of that is that
we'll inherit all the fence work and checking for free.
>
> /**
> @@ -387,6 +415,8 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
> void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
> enum drm_sched_priority priority)
> {
> + WARN_ON(!!entity->single_sched);
> +
> spin_lock(&entity->rq_lock);
> entity->priority = priority;
> spin_unlock(&entity->rq_lock);
> @@ -399,7 +429,7 @@ EXPORT_SYMBOL(drm_sched_entity_set_priority);
> */
> static bool drm_sched_entity_add_dependency_cb(struct drm_sched_entity *entity)
> {
> - struct drm_gpu_scheduler *sched = entity->rq->sched;
> + struct drm_gpu_scheduler *sched = drm_sched_entity_to_scheduler(entity);
> struct dma_fence *fence = entity->dependency;
> struct drm_sched_fence *s_fence;
>
> @@ -501,7 +531,8 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
> * Update the entity's location in the min heap according to
> * the timestamp of the next job, if any.
> */
> - if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
> + if (drm_sched_entity_to_scheduler(entity)->sched_policy ==
> + DRM_SCHED_POLICY_FIFO) {
> struct drm_sched_job *next;
>
> next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
> @@ -524,6 +555,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> struct drm_gpu_scheduler *sched;
> struct drm_sched_rq *rq;
>
> + WARN_ON(!!entity->single_sched);
> +
> /* single possible engine and already selected */
> if (!entity->sched_list)
> return;
> @@ -573,12 +606,13 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> {
> struct drm_sched_entity *entity = sched_job->entity;
> - bool first, fifo = entity->rq->sched->sched_policy ==
> - DRM_SCHED_POLICY_FIFO;
> + bool single_entity = !!entity->single_sched;
> + bool first;
> ktime_t submit_ts;
>
> trace_drm_sched_job(sched_job, entity);
> - atomic_inc(entity->rq->sched->score);
> + if (!single_entity)
> + atomic_inc(entity->rq->sched->score);
> WRITE_ONCE(entity->last_user, current->group_leader);
>
> /*
> @@ -591,6 +625,10 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
>
> /* first job wakes up scheduler */
> if (first) {
> + struct drm_gpu_scheduler *sched =
> + drm_sched_entity_to_scheduler(entity);
> + bool fifo = sched->sched_policy == DRM_SCHED_POLICY_FIFO;
> +
> /* Add the entity to the run queue */
> spin_lock(&entity->rq_lock);
> if (entity->stopped) {
> @@ -600,13 +638,14 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> return;
> }
>
> - drm_sched_rq_add_entity(entity->rq, entity);
> + if (!single_entity)
> + drm_sched_rq_add_entity(entity->rq, entity);
> spin_unlock(&entity->rq_lock);
>
> if (fifo)
> drm_sched_rq_update_fifo(entity, submit_ts);
>
> - drm_sched_wakeup_if_can_queue(entity->rq->sched);
> + drm_sched_wakeup_if_can_queue(sched);
> }
> }
> EXPORT_SYMBOL(drm_sched_entity_push_job);
> diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c
> index 06cedfe4b486..f6b926f5e188 100644
> --- a/drivers/gpu/drm/scheduler/sched_fence.c
> +++ b/drivers/gpu/drm/scheduler/sched_fence.c
> @@ -225,7 +225,7 @@ void drm_sched_fence_init(struct drm_sched_fence *fence,
> {
> unsigned seq;
>
> - fence->sched = entity->rq->sched;
> + fence->sched = drm_sched_entity_to_scheduler(entity);
> seq = atomic_inc_return(&entity->fence_seq);
> dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled,
> &fence->lock, entity->fence_context, seq);
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 150e5330f0fa..273e0fbc4eab 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -32,7 +32,8 @@
> * backend operations to the scheduler like submitting a job to hardware run queue,
> * returning the dependencies of a job etc.
> *
> - * The organisation of the scheduler is the following:
> + * For scheduling policies DRM_SCHED_POLICY_RR and DRM_SCHED_POLICY_FIFO, the
> + * scheduler organization is:
> *
> * 1. Each hw run queue has one scheduler
> * 2. Each scheduler has multiple run queues with different priorities
> @@ -43,7 +44,24 @@
> *
> * The jobs in a entity are always scheduled in the order that they were pushed.
> *
> - * Note that once a job was taken from the entities queue and pushed to the
> + * For DRM_SCHED_POLICY_SINGLE_ENTITY, the organization of the scheduler is:
> + *
> + * 1. One to one relationship between scheduler and entity
> + * 2. No priorities implemented per scheduler (single job queue)
> + * 3. No run queues in scheduler rather jobs are directly dequeued from entity
> + * 4. The entity maintains a queue of jobs that will be scheduled to the
> + * hardware
> + *
> + * The jobs in a entity are always scheduled in the order that they were pushed
> + * regardless of scheduling policy. Single-entity scheduling is essentially a
> + * FIFO for jobs.
> + *
> + * A policy of DRM_SCHED_POLICY_RR or DRM_SCHED_POLICY_FIFO is expected to be
> + * used when the kernel-mode driver is scheduling directly to the hardware while
> + * a scheduling policy of DRM_SCHED_POLICY_SINGLE_ENTITY is expected to be used
> + * when there is a firmware scheduler.
> + *
> + * Note that once a job is taken from the entities queue and pushed to the
> * hardware, i.e. the pending queue, the entity must not be referenced anymore
> * through the jobs entity pointer.
> */
> @@ -96,6 +114,8 @@ static inline void drm_sched_rq_remove_fifo_locked(struct drm_sched_entity *enti
>
> void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
> {
> + WARN_ON(!!entity->single_sched);
> +
> /*
> * Both locks need to be grabbed, one to protect from entity->rq change
> * for entity from within concurrent drm_sched_entity_select_rq and the
> @@ -126,6 +146,8 @@ void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
> static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
> struct drm_sched_rq *rq)
> {
> + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
> +
> spin_lock_init(&rq->lock);
> INIT_LIST_HEAD(&rq->entities);
> rq->rb_tree_root = RB_ROOT_CACHED;
> @@ -144,6 +166,8 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
> void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
> struct drm_sched_entity *entity)
> {
> + WARN_ON(!!entity->single_sched);
> +
> if (!list_empty(&entity->list))
> return;
>
> @@ -166,6 +190,8 @@ void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
> void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
> struct drm_sched_entity *entity)
> {
> + WARN_ON(!!entity->single_sched);
> +
> if (list_empty(&entity->list))
> return;
>
> @@ -641,7 +667,7 @@ int drm_sched_job_init(struct drm_sched_job *job,
> struct drm_sched_entity *entity,
> void *owner)
> {
> - if (!entity->rq)
> + if (!entity->rq && !entity->single_sched)
> return -ENOENT;
>
> job->entity = entity;
> @@ -674,13 +700,16 @@ void drm_sched_job_arm(struct drm_sched_job *job)
> {
> struct drm_gpu_scheduler *sched;
> struct drm_sched_entity *entity = job->entity;
> + bool single_entity = !!entity->single_sched;
>
> BUG_ON(!entity);
> - drm_sched_entity_select_rq(entity);
> - sched = entity->rq->sched;
> + if (!single_entity)
> + drm_sched_entity_select_rq(entity);
> + sched = drm_sched_entity_to_scheduler(entity);
>
> job->sched = sched;
> - job->s_priority = entity->rq - sched->sched_rq;
> + if (!single_entity)
> + job->s_priority = entity->rq - sched->sched_rq;
> job->id = atomic64_inc_return(&sched->job_id_count);
>
> drm_sched_fence_init(job->s_fence, job->entity);
> @@ -896,6 +925,14 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
> if (!drm_sched_can_queue(sched))
> return NULL;
>
> + if (sched->single_entity) {
> + if (!READ_ONCE(sched->single_entity->stopped) &&
> + drm_sched_entity_is_ready(sched->single_entity))
> + return sched->single_entity;
> +
> + return NULL;
> + }
> +
> /* Kernel run queue has higher priority than normal run queue*/
> for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> @@ -1092,6 +1129,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> return -EINVAL;
>
> sched->ops = ops;
> + sched->single_entity = NULL;
> sched->hw_submission_limit = hw_submission;
> sched->name = name;
> if (submit_wq) {
> @@ -1111,8 +1149,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> sched->dev = dev;
> sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
> drm_sched_policy_default : sched_policy;
> - for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
> - drm_sched_rq_init(sched, &sched->sched_rq[i]);
> + if (sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
> + for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
> + drm_sched_rq_init(sched, &sched->sched_rq[i]);
> + }
With dynamic rq, no changes would be necessary here--we just go over the single rq.
>
> init_waitqueue_head(&sched->job_scheduled);
> INIT_LIST_HEAD(&sched->pending_list);
> @@ -1143,19 +1183,24 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
>
> drm_sched_wqueue_stop(sched);
>
> - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> - struct drm_sched_rq *rq = &sched->sched_rq[i];
> -
> - spin_lock(&rq->lock);
> - list_for_each_entry(s_entity, &rq->entities, list)
> - /*
> - * Prevents reinsertion and marks job_queue as idle,
> - * it will removed from rq in drm_sched_entity_fini
> - * eventually
> - */
> - s_entity->stopped = true;
> - spin_unlock(&rq->lock);
> + if (sched->single_entity) {
> + spin_lock(&sched->single_entity->rq_lock);
> + sched->single_entity->stopped = true;
> + spin_unlock(&sched->single_entity->rq_lock);
> + } else if (sched->sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
> + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> + struct drm_sched_rq *rq = &sched->sched_rq[i];
>
> + spin_lock(&rq->lock);
> + list_for_each_entry(s_entity, &rq->entities, list)
> + /*
> + * Prevents reinsertion and marks job_queue as idle,
> + * it will removed from rq in drm_sched_entity_fini
> + * eventually
> + */
> + s_entity->stopped = true;
> + spin_unlock(&rq->lock);
> + }
> }
Same here--we can keep the loop intact, we have a single rq and we just process it.
>
> /* Wakeup everyone stuck in drm_sched_entity_flush for this scheduler */
> @@ -1186,6 +1231,8 @@ void drm_sched_increase_karma(struct drm_sched_job *bad)
> struct drm_sched_entity *entity;
> struct drm_gpu_scheduler *sched = bad->sched;
>
> + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
> +
> /* don't change @bad's karma if it's from KERNEL RQ,
> * because sometimes GPU hang would cause kernel jobs (like VM updating jobs)
> * corrupt but keep in mind that kernel jobs always considered good.
> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> index 320f0a720486..e67b53c3546b 100644
> --- a/include/drm/gpu_scheduler.h
> +++ b/include/drm/gpu_scheduler.h
> @@ -79,6 +79,7 @@ enum drm_sched_policy {
> DRM_SCHED_POLICY_UNSET,
> DRM_SCHED_POLICY_RR,
> DRM_SCHED_POLICY_FIFO,
> + DRM_SCHED_POLICY_SINGLE_ENTITY,
> DRM_SCHED_POLICY_COUNT,
> };
>
> @@ -112,6 +113,9 @@ struct drm_sched_entity {
> */
> struct drm_sched_rq *rq;
>
> + /** @single_sched: Single scheduler */
> + struct drm_gpu_scheduler *single_sched;
> +
> /**
> * @sched_list:
> *
> @@ -473,6 +477,7 @@ struct drm_sched_backend_ops {
> * struct drm_gpu_scheduler - scheduler instance-specific data
> *
> * @ops: backend operations provided by the driver.
> + * @single_entity: Single entity for the scheduler
> * @hw_submission_limit: the max size of the hardware queue.
> * @timeout: the time after which a job is removed from the scheduler.
> * @name: name of the ring for which this scheduler is being used.
> @@ -504,6 +509,7 @@ struct drm_sched_backend_ops {
> */
> struct drm_gpu_scheduler {
> const struct drm_sched_backend_ops *ops;
> + struct drm_sched_entity *single_entity;
Right, as I mentioned above, we shouldn't splice up between a scheduler
and an entity sometimes, and other times scheduler to rq to entity--it just
creates too much thrashing about in the code, with too many ifs, conditions,
etc. Simpler is better--parameterization of the number of rqs.
Instead, let's get the dynamic rqs in, and this way we can keep much
of the code the same, inheriting fence work and checks, and so on,
without much changes.
Regards,
Luben
> uint32_t hw_submission_limit;
> long timeout;
> const char *name;
> @@ -587,6 +593,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> struct drm_gpu_scheduler **sched_list,
> unsigned int num_sched_list,
> atomic_t *guilty);
> +struct drm_gpu_scheduler *
> +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity);
> long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
> void drm_sched_entity_fini(struct drm_sched_entity *entity);
> void drm_sched_entity_destroy(struct drm_sched_entity *entity);
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
2023-10-24 3:50 ` Luben Tuikov
@ 2023-10-25 15:13 ` Matthew Brost
2023-10-25 18:39 ` Luben Tuikov
0 siblings, 1 reply; 22+ messages in thread
From: Matthew Brost @ 2023-10-25 15:13 UTC (permalink / raw)
To: Luben Tuikov
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen, lina,
mcanal, Liviu.Dudau, dri-devel, christian.koenig,
boris.brezillon, dakr, donald.robson, intel-xe, faith.ekstrand
On Mon, Oct 23, 2023 at 11:50:26PM -0400, Luben Tuikov wrote:
> Hi,
>
> On 2023-10-17 11:09, Matthew Brost wrote:
> > DRM_SCHED_POLICY_SINGLE_ENTITY creates a 1 to 1 relationship between
> > scheduler and entity. No priorities or run queue used in this mode.
> > Intended for devices with firmware schedulers.
> >
> > v2:
> > - Drop sched / rq union (Luben)
> > v3:
> > - Don't pick entity if stopped in drm_sched_select_entity (Danilo)
> > v4:
> > - Rework if statement in drm_sched_entity_init (Luben)
> > - Update comment for drm_sched_entity_to_scheduler (Luben)
> > - Reword a few things in DOC comment (Luben)
> > - Do not check sched policy in for statement (Luben)
> > v5:
> > - Fix extra blank lines (Luben / Checkpatch)
> >
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > ---
> > drivers/gpu/drm/scheduler/sched_entity.c | 69 +++++++++++++++----
> > drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
> > drivers/gpu/drm/scheduler/sched_main.c | 87 ++++++++++++++++++------
> > include/drm/gpu_scheduler.h | 8 +++
> > 4 files changed, 130 insertions(+), 36 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> > index cf42e2265d64..940f63dd6965 100644
> > --- a/drivers/gpu/drm/scheduler/sched_entity.c
> > +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> > @@ -83,6 +83,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> > memset(entity, 0, sizeof(struct drm_sched_entity));
> > INIT_LIST_HEAD(&entity->list);
> > entity->rq = NULL;
> > + entity->single_sched = NULL;
> > entity->guilty = guilty;
> > entity->num_sched_list = num_sched_list;
> > entity->priority = priority;
> > @@ -90,8 +91,17 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> > RCU_INIT_POINTER(entity->last_scheduled, NULL);
> > RB_CLEAR_NODE(&entity->rb_tree_node);
> >
> > - if(num_sched_list)
> > - entity->rq = &sched_list[0]->sched_rq[entity->priority];
> > + if (num_sched_list) {
> > + if (sched_list[0]->sched_policy !=
> > + DRM_SCHED_POLICY_SINGLE_ENTITY) {
> > + entity->rq = &sched_list[0]->sched_rq[entity->priority];
> > + } else if (num_sched_list == 1 && !sched_list[0]->single_entity) {
> > + sched_list[0]->single_entity = entity;
> > + entity->single_sched = sched_list[0];
>
> To simplify the rest of the GPU scheduler design and generalize the logic,
> we can do
> entity->rq = sched_list[0]->sched_rq[entity->priority];
> where the "priority" is 0, thus having a single rq.
>
> We shouldn't splice scheduler and entity, but rather make it so that
> we can set the number of rqs to 1, thus resulting in a single rq.
>
> (https://lore.kernel.org/r/20231023032251.164775-1-luben.tuikov@amd.com)
>
I pulled out this patch [1] + previous one [2] and included your [3] to
test this in Xe [4].
It seems to work with just one rq per scheduler. We can go with this
approach in feel like this is the route. My next post will include your
patch [3] if we agree.
Matt
[1] https://patchwork.freedesktop.org/patch/563094/?series=121745&rev=8
[2] https://patchwork.freedesktop.org/patch/563093/?series=121745&rev=8
[3] https://patchwork.freedesktop.org/patch/563817/?series=125433&rev=1
[4] https://patchwork.freedesktop.org/series/125540/
> > + } else {
> > + return -EINVAL;
> > + }
> > + }
> >
> > init_completion(&entity->entity_idle);
> >
> > @@ -124,7 +134,8 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
> > struct drm_gpu_scheduler **sched_list,
> > unsigned int num_sched_list)
> > {
> > - WARN_ON(!num_sched_list || !sched_list);
> > + WARN_ON(!num_sched_list || !sched_list ||
> > + !!entity->single_sched);
>
> We wouldn't need this check.
>
> >
> > entity->sched_list = sched_list;
> > entity->num_sched_list = num_sched_list;
> > @@ -231,13 +242,15 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
> > {
> > struct drm_sched_job *job;
> > struct dma_fence *prev;
> > + bool single_entity = !!entity->single_sched;
> >
> > - if (!entity->rq)
> > + if (!entity->rq && !single_entity)
> > return;
> >
> > spin_lock(&entity->rq_lock);
> > entity->stopped = true;
> > - drm_sched_rq_remove_entity(entity->rq, entity);
> > + if (!single_entity)
> > + drm_sched_rq_remove_entity(entity->rq, entity);
> > spin_unlock(&entity->rq_lock);
>
> We should be able to carry on the existing infrastructure when
> having num_rqs = 1, with dynamic rqs. So we shouldn't warrant
> any changes here.
>
> >
> > /* Make sure this entity is not used by the scheduler at the moment */
> > @@ -259,6 +272,20 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
> > dma_fence_put(prev);
> > }
> >
> > +/**
> > + * drm_sched_entity_to_scheduler - Schedule entity to GPU scheduler
> > + * @entity: scheduler entity
> > + *
> > + * Returns GPU scheduler for the entity
> > + */
> > +struct drm_gpu_scheduler *
> > +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity)
> > +{
> > + bool single_entity = !!entity->single_sched;
> > +
> > + return single_entity ? entity->single_sched : entity->rq->sched;
>
> It would be "entity->rq->sched" for any and all cases which simplifies things.
>
> > +}
> > +
> > /**
> > * drm_sched_entity_flush - Flush a context entity
> > *
> > @@ -276,11 +303,12 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
> > struct drm_gpu_scheduler *sched;
> > struct task_struct *last_user;
> > long ret = timeout;
> > + bool single_entity = !!entity->single_sched;
> >
> > - if (!entity->rq)
> > + if (!entity->rq && !single_entity)
> > return 0;
> >
> > - sched = entity->rq->sched;
> > + sched = drm_sched_entity_to_scheduler(entity);
> > /**
> > * The client will not queue more IBs during this fini, consume existing
> > * queued IBs or discard them on SIGKILL
> > @@ -373,7 +401,7 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
> > container_of(cb, struct drm_sched_entity, cb);
> >
> > drm_sched_entity_clear_dep(f, cb);
> > - drm_sched_wakeup_if_can_queue(entity->rq->sched);
> > + drm_sched_wakeup_if_can_queue(drm_sched_entity_to_scheduler(entity));
> > }
>
> We can carry on that too, without changes. The good part of that is that
> we'll inherit all the fence work and checking for free.
>
> >
> > /**
> > @@ -387,6 +415,8 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
> > void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
> > enum drm_sched_priority priority)
> > {
> > + WARN_ON(!!entity->single_sched);
> > +
> > spin_lock(&entity->rq_lock);
> > entity->priority = priority;
> > spin_unlock(&entity->rq_lock);
> > @@ -399,7 +429,7 @@ EXPORT_SYMBOL(drm_sched_entity_set_priority);
> > */
> > static bool drm_sched_entity_add_dependency_cb(struct drm_sched_entity *entity)
> > {
> > - struct drm_gpu_scheduler *sched = entity->rq->sched;
> > + struct drm_gpu_scheduler *sched = drm_sched_entity_to_scheduler(entity);
> > struct dma_fence *fence = entity->dependency;
> > struct drm_sched_fence *s_fence;
> >
> > @@ -501,7 +531,8 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
> > * Update the entity's location in the min heap according to
> > * the timestamp of the next job, if any.
> > */
> > - if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
> > + if (drm_sched_entity_to_scheduler(entity)->sched_policy ==
> > + DRM_SCHED_POLICY_FIFO) {
> > struct drm_sched_job *next;
> >
> > next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
> > @@ -524,6 +555,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> > struct drm_gpu_scheduler *sched;
> > struct drm_sched_rq *rq;
> >
> > + WARN_ON(!!entity->single_sched);
> > +
> > /* single possible engine and already selected */
> > if (!entity->sched_list)
> > return;
> > @@ -573,12 +606,13 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> > void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> > {
> > struct drm_sched_entity *entity = sched_job->entity;
> > - bool first, fifo = entity->rq->sched->sched_policy ==
> > - DRM_SCHED_POLICY_FIFO;
> > + bool single_entity = !!entity->single_sched;
> > + bool first;
> > ktime_t submit_ts;
> >
> > trace_drm_sched_job(sched_job, entity);
> > - atomic_inc(entity->rq->sched->score);
> > + if (!single_entity)
> > + atomic_inc(entity->rq->sched->score);
> > WRITE_ONCE(entity->last_user, current->group_leader);
> >
> > /*
> > @@ -591,6 +625,10 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> >
> > /* first job wakes up scheduler */
> > if (first) {
> > + struct drm_gpu_scheduler *sched =
> > + drm_sched_entity_to_scheduler(entity);
> > + bool fifo = sched->sched_policy == DRM_SCHED_POLICY_FIFO;
> > +
> > /* Add the entity to the run queue */
> > spin_lock(&entity->rq_lock);
> > if (entity->stopped) {
> > @@ -600,13 +638,14 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
> > return;
> > }
> >
> > - drm_sched_rq_add_entity(entity->rq, entity);
> > + if (!single_entity)
> > + drm_sched_rq_add_entity(entity->rq, entity);
> > spin_unlock(&entity->rq_lock);
> >
> > if (fifo)
> > drm_sched_rq_update_fifo(entity, submit_ts);
> >
> > - drm_sched_wakeup_if_can_queue(entity->rq->sched);
> > + drm_sched_wakeup_if_can_queue(sched);
> > }
> > }
> > EXPORT_SYMBOL(drm_sched_entity_push_job);
> > diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c
> > index 06cedfe4b486..f6b926f5e188 100644
> > --- a/drivers/gpu/drm/scheduler/sched_fence.c
> > +++ b/drivers/gpu/drm/scheduler/sched_fence.c
> > @@ -225,7 +225,7 @@ void drm_sched_fence_init(struct drm_sched_fence *fence,
> > {
> > unsigned seq;
> >
> > - fence->sched = entity->rq->sched;
> > + fence->sched = drm_sched_entity_to_scheduler(entity);
> > seq = atomic_inc_return(&entity->fence_seq);
> > dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled,
> > &fence->lock, entity->fence_context, seq);
> > diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> > index 150e5330f0fa..273e0fbc4eab 100644
> > --- a/drivers/gpu/drm/scheduler/sched_main.c
> > +++ b/drivers/gpu/drm/scheduler/sched_main.c
> > @@ -32,7 +32,8 @@
> > * backend operations to the scheduler like submitting a job to hardware run queue,
> > * returning the dependencies of a job etc.
> > *
> > - * The organisation of the scheduler is the following:
> > + * For scheduling policies DRM_SCHED_POLICY_RR and DRM_SCHED_POLICY_FIFO, the
> > + * scheduler organization is:
> > *
> > * 1. Each hw run queue has one scheduler
> > * 2. Each scheduler has multiple run queues with different priorities
> > @@ -43,7 +44,24 @@
> > *
> > * The jobs in a entity are always scheduled in the order that they were pushed.
> > *
> > - * Note that once a job was taken from the entities queue and pushed to the
> > + * For DRM_SCHED_POLICY_SINGLE_ENTITY, the organization of the scheduler is:
> > + *
> > + * 1. One to one relationship between scheduler and entity
> > + * 2. No priorities implemented per scheduler (single job queue)
> > + * 3. No run queues in scheduler rather jobs are directly dequeued from entity
> > + * 4. The entity maintains a queue of jobs that will be scheduled to the
> > + * hardware
> > + *
> > + * The jobs in a entity are always scheduled in the order that they were pushed
> > + * regardless of scheduling policy. Single-entity scheduling is essentially a
> > + * FIFO for jobs.
> > + *
> > + * A policy of DRM_SCHED_POLICY_RR or DRM_SCHED_POLICY_FIFO is expected to be
> > + * used when the kernel-mode driver is scheduling directly to the hardware while
> > + * a scheduling policy of DRM_SCHED_POLICY_SINGLE_ENTITY is expected to be used
> > + * when there is a firmware scheduler.
> > + *
> > + * Note that once a job is taken from the entities queue and pushed to the
> > * hardware, i.e. the pending queue, the entity must not be referenced anymore
> > * through the jobs entity pointer.
> > */
> > @@ -96,6 +114,8 @@ static inline void drm_sched_rq_remove_fifo_locked(struct drm_sched_entity *enti
> >
> > void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
> > {
> > + WARN_ON(!!entity->single_sched);
> > +
> > /*
> > * Both locks need to be grabbed, one to protect from entity->rq change
> > * for entity from within concurrent drm_sched_entity_select_rq and the
> > @@ -126,6 +146,8 @@ void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
> > static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
> > struct drm_sched_rq *rq)
> > {
> > + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
> > +
> > spin_lock_init(&rq->lock);
> > INIT_LIST_HEAD(&rq->entities);
> > rq->rb_tree_root = RB_ROOT_CACHED;
> > @@ -144,6 +166,8 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
> > void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
> > struct drm_sched_entity *entity)
> > {
> > + WARN_ON(!!entity->single_sched);
> > +
> > if (!list_empty(&entity->list))
> > return;
> >
> > @@ -166,6 +190,8 @@ void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
> > void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
> > struct drm_sched_entity *entity)
> > {
> > + WARN_ON(!!entity->single_sched);
> > +
> > if (list_empty(&entity->list))
> > return;
> >
> > @@ -641,7 +667,7 @@ int drm_sched_job_init(struct drm_sched_job *job,
> > struct drm_sched_entity *entity,
> > void *owner)
> > {
> > - if (!entity->rq)
> > + if (!entity->rq && !entity->single_sched)
> > return -ENOENT;
> >
> > job->entity = entity;
> > @@ -674,13 +700,16 @@ void drm_sched_job_arm(struct drm_sched_job *job)
> > {
> > struct drm_gpu_scheduler *sched;
> > struct drm_sched_entity *entity = job->entity;
> > + bool single_entity = !!entity->single_sched;
> >
> > BUG_ON(!entity);
> > - drm_sched_entity_select_rq(entity);
> > - sched = entity->rq->sched;
> > + if (!single_entity)
> > + drm_sched_entity_select_rq(entity);
> > + sched = drm_sched_entity_to_scheduler(entity);
> >
> > job->sched = sched;
> > - job->s_priority = entity->rq - sched->sched_rq;
> > + if (!single_entity)
> > + job->s_priority = entity->rq - sched->sched_rq;
> > job->id = atomic64_inc_return(&sched->job_id_count);
> >
> > drm_sched_fence_init(job->s_fence, job->entity);
> > @@ -896,6 +925,14 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
> > if (!drm_sched_can_queue(sched))
> > return NULL;
> >
> > + if (sched->single_entity) {
> > + if (!READ_ONCE(sched->single_entity->stopped) &&
> > + drm_sched_entity_is_ready(sched->single_entity))
> > + return sched->single_entity;
> > +
> > + return NULL;
> > + }
> > +
> > /* Kernel run queue has higher priority than normal run queue*/
> > for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> > entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
> > @@ -1092,6 +1129,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> > return -EINVAL;
> >
> > sched->ops = ops;
> > + sched->single_entity = NULL;
> > sched->hw_submission_limit = hw_submission;
> > sched->name = name;
> > if (submit_wq) {
> > @@ -1111,8 +1149,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
> > sched->dev = dev;
> > sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
> > drm_sched_policy_default : sched_policy;
> > - for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
> > - drm_sched_rq_init(sched, &sched->sched_rq[i]);
> > + if (sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
> > + for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
> > + drm_sched_rq_init(sched, &sched->sched_rq[i]);
> > + }
>
> With dynamic rq, no changes would be necessary here--we just go over the single rq.
>
> >
> > init_waitqueue_head(&sched->job_scheduled);
> > INIT_LIST_HEAD(&sched->pending_list);
> > @@ -1143,19 +1183,24 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
> >
> > drm_sched_wqueue_stop(sched);
> >
> > - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> > - struct drm_sched_rq *rq = &sched->sched_rq[i];
> > -
> > - spin_lock(&rq->lock);
> > - list_for_each_entry(s_entity, &rq->entities, list)
> > - /*
> > - * Prevents reinsertion and marks job_queue as idle,
> > - * it will removed from rq in drm_sched_entity_fini
> > - * eventually
> > - */
> > - s_entity->stopped = true;
> > - spin_unlock(&rq->lock);
> > + if (sched->single_entity) {
> > + spin_lock(&sched->single_entity->rq_lock);
> > + sched->single_entity->stopped = true;
> > + spin_unlock(&sched->single_entity->rq_lock);
> > + } else if (sched->sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
> > + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
> > + struct drm_sched_rq *rq = &sched->sched_rq[i];
> >
> > + spin_lock(&rq->lock);
> > + list_for_each_entry(s_entity, &rq->entities, list)
> > + /*
> > + * Prevents reinsertion and marks job_queue as idle,
> > + * it will removed from rq in drm_sched_entity_fini
> > + * eventually
> > + */
> > + s_entity->stopped = true;
> > + spin_unlock(&rq->lock);
> > + }
> > }
>
> Same here--we can keep the loop intact, we have a single rq and we just process it.
>
> >
> > /* Wakeup everyone stuck in drm_sched_entity_flush for this scheduler */
> > @@ -1186,6 +1231,8 @@ void drm_sched_increase_karma(struct drm_sched_job *bad)
> > struct drm_sched_entity *entity;
> > struct drm_gpu_scheduler *sched = bad->sched;
> >
> > + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
> > +
> > /* don't change @bad's karma if it's from KERNEL RQ,
> > * because sometimes GPU hang would cause kernel jobs (like VM updating jobs)
> > * corrupt but keep in mind that kernel jobs always considered good.
> > diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> > index 320f0a720486..e67b53c3546b 100644
> > --- a/include/drm/gpu_scheduler.h
> > +++ b/include/drm/gpu_scheduler.h
> > @@ -79,6 +79,7 @@ enum drm_sched_policy {
> > DRM_SCHED_POLICY_UNSET,
> > DRM_SCHED_POLICY_RR,
> > DRM_SCHED_POLICY_FIFO,
> > + DRM_SCHED_POLICY_SINGLE_ENTITY,
> > DRM_SCHED_POLICY_COUNT,
> > };
> >
> > @@ -112,6 +113,9 @@ struct drm_sched_entity {
> > */
> > struct drm_sched_rq *rq;
> >
> > + /** @single_sched: Single scheduler */
> > + struct drm_gpu_scheduler *single_sched;
> > +
> > /**
> > * @sched_list:
> > *
> > @@ -473,6 +477,7 @@ struct drm_sched_backend_ops {
> > * struct drm_gpu_scheduler - scheduler instance-specific data
> > *
> > * @ops: backend operations provided by the driver.
> > + * @single_entity: Single entity for the scheduler
> > * @hw_submission_limit: the max size of the hardware queue.
> > * @timeout: the time after which a job is removed from the scheduler.
> > * @name: name of the ring for which this scheduler is being used.
> > @@ -504,6 +509,7 @@ struct drm_sched_backend_ops {
> > */
> > struct drm_gpu_scheduler {
> > const struct drm_sched_backend_ops *ops;
> > + struct drm_sched_entity *single_entity;
>
> Right, as I mentioned above, we shouldn't splice up between a scheduler
> and an entity sometimes, and other times scheduler to rq to entity--it just
> creates too much thrashing about in the code, with too many ifs, conditions,
> etc. Simpler is better--parameterization of the number of rqs.
>
> Instead, let's get the dynamic rqs in, and this way we can keep much
> of the code the same, inheriting fence work and checks, and so on,
> without much changes.
>
> Regards,
> Luben
>
> > uint32_t hw_submission_limit;
> > long timeout;
> > const char *name;
> > @@ -587,6 +593,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
> > struct drm_gpu_scheduler **sched_list,
> > unsigned int num_sched_list,
> > atomic_t *guilty);
> > +struct drm_gpu_scheduler *
> > +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity);
> > long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
> > void drm_sched_entity_fini(struct drm_sched_entity *entity);
> > void drm_sched_entity_destroy(struct drm_sched_entity *entity);
>
^ permalink raw reply [flat|nested] 22+ messages in thread
* Re: [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy
2023-10-25 15:13 ` Matthew Brost
@ 2023-10-25 18:39 ` Luben Tuikov
0 siblings, 0 replies; 22+ messages in thread
From: Luben Tuikov @ 2023-10-25 18:39 UTC (permalink / raw)
To: Matthew Brost
Cc: robdclark, thomas.hellstrom, sarah.walker, ketil.johnsen, lina,
mcanal, Liviu.Dudau, dri-devel, christian.koenig,
boris.brezillon, dakr, donald.robson, intel-xe, faith.ekstrand
Hi Matt,
On 2023-10-25 11:13, Matthew Brost wrote:
> On Mon, Oct 23, 2023 at 11:50:26PM -0400, Luben Tuikov wrote:
>> Hi,
>>
>> On 2023-10-17 11:09, Matthew Brost wrote:
>>> DRM_SCHED_POLICY_SINGLE_ENTITY creates a 1 to 1 relationship between
>>> scheduler and entity. No priorities or run queue used in this mode.
>>> Intended for devices with firmware schedulers.
>>>
>>> v2:
>>> - Drop sched / rq union (Luben)
>>> v3:
>>> - Don't pick entity if stopped in drm_sched_select_entity (Danilo)
>>> v4:
>>> - Rework if statement in drm_sched_entity_init (Luben)
>>> - Update comment for drm_sched_entity_to_scheduler (Luben)
>>> - Reword a few things in DOC comment (Luben)
>>> - Do not check sched policy in for statement (Luben)
>>> v5:
>>> - Fix extra blank lines (Luben / Checkpatch)
>>>
>>> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
>>> ---
>>> drivers/gpu/drm/scheduler/sched_entity.c | 69 +++++++++++++++----
>>> drivers/gpu/drm/scheduler/sched_fence.c | 2 +-
>>> drivers/gpu/drm/scheduler/sched_main.c | 87 ++++++++++++++++++------
>>> include/drm/gpu_scheduler.h | 8 +++
>>> 4 files changed, 130 insertions(+), 36 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
>>> index cf42e2265d64..940f63dd6965 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>>> @@ -83,6 +83,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>>> memset(entity, 0, sizeof(struct drm_sched_entity));
>>> INIT_LIST_HEAD(&entity->list);
>>> entity->rq = NULL;
>>> + entity->single_sched = NULL;
>>> entity->guilty = guilty;
>>> entity->num_sched_list = num_sched_list;
>>> entity->priority = priority;
>>> @@ -90,8 +91,17 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>>> RCU_INIT_POINTER(entity->last_scheduled, NULL);
>>> RB_CLEAR_NODE(&entity->rb_tree_node);
>>>
>>> - if(num_sched_list)
>>> - entity->rq = &sched_list[0]->sched_rq[entity->priority];
>>> + if (num_sched_list) {
>>> + if (sched_list[0]->sched_policy !=
>>> + DRM_SCHED_POLICY_SINGLE_ENTITY) {
>>> + entity->rq = &sched_list[0]->sched_rq[entity->priority];
>>> + } else if (num_sched_list == 1 && !sched_list[0]->single_entity) {
>>> + sched_list[0]->single_entity = entity;
>>> + entity->single_sched = sched_list[0];
>>
>> To simplify the rest of the GPU scheduler design and generalize the logic,
>> we can do
>> entity->rq = sched_list[0]->sched_rq[entity->priority];
>> where the "priority" is 0, thus having a single rq.
>>
>> We shouldn't splice scheduler and entity, but rather make it so that
>> we can set the number of rqs to 1, thus resulting in a single rq.
>>
>> (https://lore.kernel.org/r/20231023032251.164775-1-luben.tuikov@amd.com)
>>
>
> I pulled out this patch [1] + previous one [2] and included your [3] to
> test this in Xe [4].
>
> It seems to work with just one rq per scheduler. We can go with this
> approach in feel like this is the route. My next post will include your
> patch [3] if we agree.
Yeah, this is good. Thanks!
I feel that the sched_rq[] static array should've been a dynamic one
from the outset--one of the hallmarks of a flexible scheduler. Then
let each driver announce how many priorities they care about. A scheduler
shouldn't be as rigid as to force drivers to care and/or support so and
so many run-queues (priorities).
For a good code design, we want to allow for more driver implementations
with minimal (or no) DRM changes--all accommodated by drm_sched_init(), yet,
accommodating a varied behaviour, and having the sched_rq be dynamic and let
each driver announce how many run-queues it wants is the minimal change we
can do now (and easiest), with best outcome for Xe and new drivers.
Regards,
Luben
>
> Matt
>
> [1] https://patchwork.freedesktop.org/patch/563094/?series=121745&rev=8
> [2] https://patchwork.freedesktop.org/patch/563093/?series=121745&rev=8
> [3] https://patchwork.freedesktop.org/patch/563817/?series=125433&rev=1
> [4] https://patchwork.freedesktop.org/series/125540/
>
>>> + } else {
>>> + return -EINVAL;
>>> + }
>>> + }
>>>
>>> init_completion(&entity->entity_idle);
>>>
>>> @@ -124,7 +134,8 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
>>> struct drm_gpu_scheduler **sched_list,
>>> unsigned int num_sched_list)
>>> {
>>> - WARN_ON(!num_sched_list || !sched_list);
>>> + WARN_ON(!num_sched_list || !sched_list ||
>>> + !!entity->single_sched);
>>
>> We wouldn't need this check.
>>
>>>
>>> entity->sched_list = sched_list;
>>> entity->num_sched_list = num_sched_list;
>>> @@ -231,13 +242,15 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
>>> {
>>> struct drm_sched_job *job;
>>> struct dma_fence *prev;
>>> + bool single_entity = !!entity->single_sched;
>>>
>>> - if (!entity->rq)
>>> + if (!entity->rq && !single_entity)
>>> return;
>>>
>>> spin_lock(&entity->rq_lock);
>>> entity->stopped = true;
>>> - drm_sched_rq_remove_entity(entity->rq, entity);
>>> + if (!single_entity)
>>> + drm_sched_rq_remove_entity(entity->rq, entity);
>>> spin_unlock(&entity->rq_lock);
>>
>> We should be able to carry on the existing infrastructure when
>> having num_rqs = 1, with dynamic rqs. So we shouldn't warrant
>> any changes here.
>>
>>>
>>> /* Make sure this entity is not used by the scheduler at the moment */
>>> @@ -259,6 +272,20 @@ static void drm_sched_entity_kill(struct drm_sched_entity *entity)
>>> dma_fence_put(prev);
>>> }
>>>
>>> +/**
>>> + * drm_sched_entity_to_scheduler - Schedule entity to GPU scheduler
>>> + * @entity: scheduler entity
>>> + *
>>> + * Returns GPU scheduler for the entity
>>> + */
>>> +struct drm_gpu_scheduler *
>>> +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity)
>>> +{
>>> + bool single_entity = !!entity->single_sched;
>>> +
>>> + return single_entity ? entity->single_sched : entity->rq->sched;
>>
>> It would be "entity->rq->sched" for any and all cases which simplifies things.
>>
>>> +}
>>> +
>>> /**
>>> * drm_sched_entity_flush - Flush a context entity
>>> *
>>> @@ -276,11 +303,12 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>>> struct drm_gpu_scheduler *sched;
>>> struct task_struct *last_user;
>>> long ret = timeout;
>>> + bool single_entity = !!entity->single_sched;
>>>
>>> - if (!entity->rq)
>>> + if (!entity->rq && !single_entity)
>>> return 0;
>>>
>>> - sched = entity->rq->sched;
>>> + sched = drm_sched_entity_to_scheduler(entity);
>>> /**
>>> * The client will not queue more IBs during this fini, consume existing
>>> * queued IBs or discard them on SIGKILL
>>> @@ -373,7 +401,7 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
>>> container_of(cb, struct drm_sched_entity, cb);
>>>
>>> drm_sched_entity_clear_dep(f, cb);
>>> - drm_sched_wakeup_if_can_queue(entity->rq->sched);
>>> + drm_sched_wakeup_if_can_queue(drm_sched_entity_to_scheduler(entity));
>>> }
>>
>> We can carry on that too, without changes. The good part of that is that
>> we'll inherit all the fence work and checking for free.
>>
>>>
>>> /**
>>> @@ -387,6 +415,8 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
>>> void drm_sched_entity_set_priority(struct drm_sched_entity *entity,
>>> enum drm_sched_priority priority)
>>> {
>>> + WARN_ON(!!entity->single_sched);
>>> +
>>> spin_lock(&entity->rq_lock);
>>> entity->priority = priority;
>>> spin_unlock(&entity->rq_lock);
>>> @@ -399,7 +429,7 @@ EXPORT_SYMBOL(drm_sched_entity_set_priority);
>>> */
>>> static bool drm_sched_entity_add_dependency_cb(struct drm_sched_entity *entity)
>>> {
>>> - struct drm_gpu_scheduler *sched = entity->rq->sched;
>>> + struct drm_gpu_scheduler *sched = drm_sched_entity_to_scheduler(entity);
>>> struct dma_fence *fence = entity->dependency;
>>> struct drm_sched_fence *s_fence;
>>>
>>> @@ -501,7 +531,8 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
>>> * Update the entity's location in the min heap according to
>>> * the timestamp of the next job, if any.
>>> */
>>> - if (entity->rq->sched->sched_policy == DRM_SCHED_POLICY_FIFO) {
>>> + if (drm_sched_entity_to_scheduler(entity)->sched_policy ==
>>> + DRM_SCHED_POLICY_FIFO) {
>>> struct drm_sched_job *next;
>>>
>>> next = to_drm_sched_job(spsc_queue_peek(&entity->job_queue));
>>> @@ -524,6 +555,8 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
>>> struct drm_gpu_scheduler *sched;
>>> struct drm_sched_rq *rq;
>>>
>>> + WARN_ON(!!entity->single_sched);
>>> +
>>> /* single possible engine and already selected */
>>> if (!entity->sched_list)
>>> return;
>>> @@ -573,12 +606,13 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
>>> void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
>>> {
>>> struct drm_sched_entity *entity = sched_job->entity;
>>> - bool first, fifo = entity->rq->sched->sched_policy ==
>>> - DRM_SCHED_POLICY_FIFO;
>>> + bool single_entity = !!entity->single_sched;
>>> + bool first;
>>> ktime_t submit_ts;
>>>
>>> trace_drm_sched_job(sched_job, entity);
>>> - atomic_inc(entity->rq->sched->score);
>>> + if (!single_entity)
>>> + atomic_inc(entity->rq->sched->score);
>>> WRITE_ONCE(entity->last_user, current->group_leader);
>>>
>>> /*
>>> @@ -591,6 +625,10 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
>>>
>>> /* first job wakes up scheduler */
>>> if (first) {
>>> + struct drm_gpu_scheduler *sched =
>>> + drm_sched_entity_to_scheduler(entity);
>>> + bool fifo = sched->sched_policy == DRM_SCHED_POLICY_FIFO;
>>> +
>>> /* Add the entity to the run queue */
>>> spin_lock(&entity->rq_lock);
>>> if (entity->stopped) {
>>> @@ -600,13 +638,14 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
>>> return;
>>> }
>>>
>>> - drm_sched_rq_add_entity(entity->rq, entity);
>>> + if (!single_entity)
>>> + drm_sched_rq_add_entity(entity->rq, entity);
>>> spin_unlock(&entity->rq_lock);
>>>
>>> if (fifo)
>>> drm_sched_rq_update_fifo(entity, submit_ts);
>>>
>>> - drm_sched_wakeup_if_can_queue(entity->rq->sched);
>>> + drm_sched_wakeup_if_can_queue(sched);
>>> }
>>> }
>>> EXPORT_SYMBOL(drm_sched_entity_push_job);
>>> diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c
>>> index 06cedfe4b486..f6b926f5e188 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_fence.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_fence.c
>>> @@ -225,7 +225,7 @@ void drm_sched_fence_init(struct drm_sched_fence *fence,
>>> {
>>> unsigned seq;
>>>
>>> - fence->sched = entity->rq->sched;
>>> + fence->sched = drm_sched_entity_to_scheduler(entity);
>>> seq = atomic_inc_return(&entity->fence_seq);
>>> dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled,
>>> &fence->lock, entity->fence_context, seq);
>>> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
>>> index 150e5330f0fa..273e0fbc4eab 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_main.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_main.c
>>> @@ -32,7 +32,8 @@
>>> * backend operations to the scheduler like submitting a job to hardware run queue,
>>> * returning the dependencies of a job etc.
>>> *
>>> - * The organisation of the scheduler is the following:
>>> + * For scheduling policies DRM_SCHED_POLICY_RR and DRM_SCHED_POLICY_FIFO, the
>>> + * scheduler organization is:
>>> *
>>> * 1. Each hw run queue has one scheduler
>>> * 2. Each scheduler has multiple run queues with different priorities
>>> @@ -43,7 +44,24 @@
>>> *
>>> * The jobs in a entity are always scheduled in the order that they were pushed.
>>> *
>>> - * Note that once a job was taken from the entities queue and pushed to the
>>> + * For DRM_SCHED_POLICY_SINGLE_ENTITY, the organization of the scheduler is:
>>> + *
>>> + * 1. One to one relationship between scheduler and entity
>>> + * 2. No priorities implemented per scheduler (single job queue)
>>> + * 3. No run queues in scheduler rather jobs are directly dequeued from entity
>>> + * 4. The entity maintains a queue of jobs that will be scheduled to the
>>> + * hardware
>>> + *
>>> + * The jobs in a entity are always scheduled in the order that they were pushed
>>> + * regardless of scheduling policy. Single-entity scheduling is essentially a
>>> + * FIFO for jobs.
>>> + *
>>> + * A policy of DRM_SCHED_POLICY_RR or DRM_SCHED_POLICY_FIFO is expected to be
>>> + * used when the kernel-mode driver is scheduling directly to the hardware while
>>> + * a scheduling policy of DRM_SCHED_POLICY_SINGLE_ENTITY is expected to be used
>>> + * when there is a firmware scheduler.
>>> + *
>>> + * Note that once a job is taken from the entities queue and pushed to the
>>> * hardware, i.e. the pending queue, the entity must not be referenced anymore
>>> * through the jobs entity pointer.
>>> */
>>> @@ -96,6 +114,8 @@ static inline void drm_sched_rq_remove_fifo_locked(struct drm_sched_entity *enti
>>>
>>> void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
>>> {
>>> + WARN_ON(!!entity->single_sched);
>>> +
>>> /*
>>> * Both locks need to be grabbed, one to protect from entity->rq change
>>> * for entity from within concurrent drm_sched_entity_select_rq and the
>>> @@ -126,6 +146,8 @@ void drm_sched_rq_update_fifo(struct drm_sched_entity *entity, ktime_t ts)
>>> static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
>>> struct drm_sched_rq *rq)
>>> {
>>> + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
>>> +
>>> spin_lock_init(&rq->lock);
>>> INIT_LIST_HEAD(&rq->entities);
>>> rq->rb_tree_root = RB_ROOT_CACHED;
>>> @@ -144,6 +166,8 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched,
>>> void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
>>> struct drm_sched_entity *entity)
>>> {
>>> + WARN_ON(!!entity->single_sched);
>>> +
>>> if (!list_empty(&entity->list))
>>> return;
>>>
>>> @@ -166,6 +190,8 @@ void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
>>> void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
>>> struct drm_sched_entity *entity)
>>> {
>>> + WARN_ON(!!entity->single_sched);
>>> +
>>> if (list_empty(&entity->list))
>>> return;
>>>
>>> @@ -641,7 +667,7 @@ int drm_sched_job_init(struct drm_sched_job *job,
>>> struct drm_sched_entity *entity,
>>> void *owner)
>>> {
>>> - if (!entity->rq)
>>> + if (!entity->rq && !entity->single_sched)
>>> return -ENOENT;
>>>
>>> job->entity = entity;
>>> @@ -674,13 +700,16 @@ void drm_sched_job_arm(struct drm_sched_job *job)
>>> {
>>> struct drm_gpu_scheduler *sched;
>>> struct drm_sched_entity *entity = job->entity;
>>> + bool single_entity = !!entity->single_sched;
>>>
>>> BUG_ON(!entity);
>>> - drm_sched_entity_select_rq(entity);
>>> - sched = entity->rq->sched;
>>> + if (!single_entity)
>>> + drm_sched_entity_select_rq(entity);
>>> + sched = drm_sched_entity_to_scheduler(entity);
>>>
>>> job->sched = sched;
>>> - job->s_priority = entity->rq - sched->sched_rq;
>>> + if (!single_entity)
>>> + job->s_priority = entity->rq - sched->sched_rq;
>>> job->id = atomic64_inc_return(&sched->job_id_count);
>>>
>>> drm_sched_fence_init(job->s_fence, job->entity);
>>> @@ -896,6 +925,14 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
>>> if (!drm_sched_can_queue(sched))
>>> return NULL;
>>>
>>> + if (sched->single_entity) {
>>> + if (!READ_ONCE(sched->single_entity->stopped) &&
>>> + drm_sched_entity_is_ready(sched->single_entity))
>>> + return sched->single_entity;
>>> +
>>> + return NULL;
>>> + }
>>> +
>>> /* Kernel run queue has higher priority than normal run queue*/
>>> for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
>>> entity = sched->sched_policy == DRM_SCHED_POLICY_FIFO ?
>>> @@ -1092,6 +1129,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
>>> return -EINVAL;
>>>
>>> sched->ops = ops;
>>> + sched->single_entity = NULL;
>>> sched->hw_submission_limit = hw_submission;
>>> sched->name = name;
>>> if (submit_wq) {
>>> @@ -1111,8 +1149,10 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
>>> sched->dev = dev;
>>> sched->sched_policy = sched_policy == DRM_SCHED_POLICY_UNSET ?
>>> drm_sched_policy_default : sched_policy;
>>> - for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
>>> - drm_sched_rq_init(sched, &sched->sched_rq[i]);
>>> + if (sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
>>> + for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++)
>>> + drm_sched_rq_init(sched, &sched->sched_rq[i]);
>>> + }
>>
>> With dynamic rq, no changes would be necessary here--we just go over the single rq.
>>
>>>
>>> init_waitqueue_head(&sched->job_scheduled);
>>> INIT_LIST_HEAD(&sched->pending_list);
>>> @@ -1143,19 +1183,24 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched)
>>>
>>> drm_sched_wqueue_stop(sched);
>>>
>>> - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
>>> - struct drm_sched_rq *rq = &sched->sched_rq[i];
>>> -
>>> - spin_lock(&rq->lock);
>>> - list_for_each_entry(s_entity, &rq->entities, list)
>>> - /*
>>> - * Prevents reinsertion and marks job_queue as idle,
>>> - * it will removed from rq in drm_sched_entity_fini
>>> - * eventually
>>> - */
>>> - s_entity->stopped = true;
>>> - spin_unlock(&rq->lock);
>>> + if (sched->single_entity) {
>>> + spin_lock(&sched->single_entity->rq_lock);
>>> + sched->single_entity->stopped = true;
>>> + spin_unlock(&sched->single_entity->rq_lock);
>>> + } else if (sched->sched_policy != DRM_SCHED_POLICY_SINGLE_ENTITY) {
>>> + for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
>>> + struct drm_sched_rq *rq = &sched->sched_rq[i];
>>>
>>> + spin_lock(&rq->lock);
>>> + list_for_each_entry(s_entity, &rq->entities, list)
>>> + /*
>>> + * Prevents reinsertion and marks job_queue as idle,
>>> + * it will removed from rq in drm_sched_entity_fini
>>> + * eventually
>>> + */
>>> + s_entity->stopped = true;
>>> + spin_unlock(&rq->lock);
>>> + }
>>> }
>>
>> Same here--we can keep the loop intact, we have a single rq and we just process it.
>>
>>>
>>> /* Wakeup everyone stuck in drm_sched_entity_flush for this scheduler */
>>> @@ -1186,6 +1231,8 @@ void drm_sched_increase_karma(struct drm_sched_job *bad)
>>> struct drm_sched_entity *entity;
>>> struct drm_gpu_scheduler *sched = bad->sched;
>>>
>>> + WARN_ON(sched->sched_policy == DRM_SCHED_POLICY_SINGLE_ENTITY);
>>> +
>>> /* don't change @bad's karma if it's from KERNEL RQ,
>>> * because sometimes GPU hang would cause kernel jobs (like VM updating jobs)
>>> * corrupt but keep in mind that kernel jobs always considered good.
>>> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
>>> index 320f0a720486..e67b53c3546b 100644
>>> --- a/include/drm/gpu_scheduler.h
>>> +++ b/include/drm/gpu_scheduler.h
>>> @@ -79,6 +79,7 @@ enum drm_sched_policy {
>>> DRM_SCHED_POLICY_UNSET,
>>> DRM_SCHED_POLICY_RR,
>>> DRM_SCHED_POLICY_FIFO,
>>> + DRM_SCHED_POLICY_SINGLE_ENTITY,
>>> DRM_SCHED_POLICY_COUNT,
>>> };
>>>
>>> @@ -112,6 +113,9 @@ struct drm_sched_entity {
>>> */
>>> struct drm_sched_rq *rq;
>>>
>>> + /** @single_sched: Single scheduler */
>>> + struct drm_gpu_scheduler *single_sched;
>>> +
>>> /**
>>> * @sched_list:
>>> *
>>> @@ -473,6 +477,7 @@ struct drm_sched_backend_ops {
>>> * struct drm_gpu_scheduler - scheduler instance-specific data
>>> *
>>> * @ops: backend operations provided by the driver.
>>> + * @single_entity: Single entity for the scheduler
>>> * @hw_submission_limit: the max size of the hardware queue.
>>> * @timeout: the time after which a job is removed from the scheduler.
>>> * @name: name of the ring for which this scheduler is being used.
>>> @@ -504,6 +509,7 @@ struct drm_sched_backend_ops {
>>> */
>>> struct drm_gpu_scheduler {
>>> const struct drm_sched_backend_ops *ops;
>>> + struct drm_sched_entity *single_entity;
>>
>> Right, as I mentioned above, we shouldn't splice up between a scheduler
>> and an entity sometimes, and other times scheduler to rq to entity--it just
>> creates too much thrashing about in the code, with too many ifs, conditions,
>> etc. Simpler is better--parameterization of the number of rqs.
>>
>> Instead, let's get the dynamic rqs in, and this way we can keep much
>> of the code the same, inheriting fence work and checks, and so on,
>> without much changes.
>>
>> Regards,
>> Luben
>>
>>> uint32_t hw_submission_limit;
>>> long timeout;
>>> const char *name;
>>> @@ -587,6 +593,8 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>>> struct drm_gpu_scheduler **sched_list,
>>> unsigned int num_sched_list,
>>> atomic_t *guilty);
>>> +struct drm_gpu_scheduler *
>>> +drm_sched_entity_to_scheduler(struct drm_sched_entity *entity);
>>> long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout);
>>> void drm_sched_entity_fini(struct drm_sched_entity *entity);
>>> void drm_sched_entity_destroy(struct drm_sched_entity *entity);
>>
^ permalink raw reply [flat|nested] 22+ messages in thread
end of thread, other threads:[~2023-10-25 18:39 UTC | newest]
Thread overview: 22+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-10-17 15:09 [PATCH v6 0/7] DRM scheduler changes for Xe Matthew Brost
2023-10-17 15:09 ` [PATCH v6 1/7] drm/sched: Add drm_sched_wqueue_* helpers Matthew Brost
2023-10-17 15:09 ` [PATCH v6 2/7] drm/sched: Convert drm scheduler to use a work queue rather than kthread Matthew Brost
2023-10-17 15:09 ` [PATCH v6 3/7] drm/sched: Move schedule policy to scheduler Matthew Brost
2023-10-24 3:34 ` Luben Tuikov
2023-10-17 15:09 ` [PATCH v6 4/7] drm/sched: Add DRM_SCHED_POLICY_SINGLE_ENTITY scheduling policy Matthew Brost
2023-10-24 3:50 ` Luben Tuikov
2023-10-25 15:13 ` Matthew Brost
2023-10-25 18:39 ` Luben Tuikov
2023-10-17 15:09 ` [PATCH v6 5/7] drm/sched: Split free_job into own work item Matthew Brost
2023-10-19 1:25 ` Luben Tuikov
2023-10-19 16:55 ` Matthew Brost
2023-10-22 1:27 ` Luben Tuikov
2023-10-23 12:16 ` Boris Brezillon
2023-10-23 12:39 ` Boris Brezillon
2023-10-23 13:54 ` Matthew Brost
2023-10-23 14:22 ` Boris Brezillon
2023-10-23 22:03 ` Matthew Brost
2023-10-17 15:09 ` [PATCH v6 6/7] drm/sched: Add drm_sched_start_timeout_unlocked helper Matthew Brost
2023-10-17 15:09 ` [PATCH v6 7/7] drm/sched: Add a helper to queue TDR immediately Matthew Brost
2023-10-18 16:18 ` Luben Tuikov
2023-10-18 8:24 ` [PATCH v6 0/7] DRM scheduler changes for Xe Daniel Vetter
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).