DRM scheduler update
Subsystems (12)
Matthew Brost
maintainerv2 cover DRM scheduler update
Update DRM scheduler in Xe to pending upstream post [1]. Upstream post should be rev7.
Want feedback on introducing Xe scheduler layer. This looks to be required as Xe must implement the message interface rather than the DRM scheduler [2].
Matt
[1] https://patchwork.freedesktop.org/series/121745/
[2] https://patchwork.freedesktop.org/patch/556934/?series=121745&rev=5
Matthew Brost (1):
drm/xe: Update to upstream DRM scheduler code
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 8 +- 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/adreno/adreno_device.c | 4 +- drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +- drivers/gpu/drm/nouveau/nouveau_sched.c | 2 +- drivers/gpu/drm/panfrost/panfrost_job.c | 2 +- drivers/gpu/drm/scheduler/sched_entity.c | 13 +- drivers/gpu/drm/scheduler/sched_main.c | 524 ++++++++++--------- drivers/gpu/drm/v3d/v3d_sched.c | 10 +- drivers/gpu/drm/xe/Makefile | 1 + drivers/gpu/drm/xe/xe_gpu_scheduler.c | 102 ++++ drivers/gpu/drm/xe/xe_gpu_scheduler.h | 75 +++ drivers/gpu/drm/xe/xe_gpu_scheduler_types.h | 58 ++ drivers/gpu/drm/xe/xe_guc_exec_queue_types.h | 8 +- drivers/gpu/drm/xe/xe_guc_submit.c | 142 +++-- drivers/gpu/drm/xe/xe_trace.h | 13 +- include/drm/gpu_scheduler.h | 54 +- 19 files changed, 625 insertions(+), 399 deletions(-) create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler.c create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler.h create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler_types.h
Signature · 1 line
2.34.1
Matthew Brost
maintainerRFC PATCH v2 1/1 drm/xe: Update to upstream DRM scheduler code
The largest change is the message interface has been removed from the DRM scheduler. Xe still needs a message interface so it is implemented in the Xe driver by adding a Xe scheduler layer.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 8 +- 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/adreno/adreno_device.c | 4 +- drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +- drivers/gpu/drm/nouveau/nouveau_sched.c | 2 +- drivers/gpu/drm/panfrost/panfrost_job.c | 2 +- drivers/gpu/drm/scheduler/sched_entity.c | 13 +- drivers/gpu/drm/scheduler/sched_main.c | 524 ++++++++++--------- drivers/gpu/drm/v3d/v3d_sched.c | 10 +- drivers/gpu/drm/xe/Makefile | 1 + drivers/gpu/drm/xe/xe_gpu_scheduler.c | 102 ++++ drivers/gpu/drm/xe/xe_gpu_scheduler.h | 75 +++ drivers/gpu/drm/xe/xe_gpu_scheduler_types.h | 58 ++ drivers/gpu/drm/xe/xe_guc_exec_queue_types.h | 8 +- drivers/gpu/drm/xe/xe_guc_submit.c | 142 +++-- drivers/gpu/drm/xe/xe_trace.h | 13 +- include/drm/gpu_scheduler.h | 54 +- 19 files changed, 625 insertions(+), 399 deletions(-) create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler.c create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler.h create mode 100644 drivers/gpu/drm/xe/xe_gpu_scheduler_types.h
drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +4 -4
@@ -1661,7 +1661,7 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)if (!ring || !ring->sched.ready)continue ; -drm_sched_run_wq_stop (&ring->sched); +drm_sched_wqueue_stop (&ring->sched); }seq_puts (m,"run ib test:\n" );
@@ -1677,7 +1677,7 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)if (!ring || !ring->sched.ready)continue ; -drm_sched_run_wq_start (&ring->sched); +drm_sched_wqueue_start (&ring->sched); }up_write (&adev->reset_domain->sem);
@@ -1915,7 +1915,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) goto pro_end;/* stop the scheduler */ -drm_sched_run_wq_stop (&ring->sched); +drm_sched_wqueue_stop (&ring->sched);/* preempt the IB */ r =amdgpu_ring_preempt_ib (ring);
@@ -1949,7 +1949,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val) failure:/* restart the scheduler */ -drm_sched_run_wq_start (&ring->sched); +drm_sched_wqueue_start (&ring->sched);up_read (&adev->reset_domain->sem);
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1 -1
@@ -2283,7 +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_DEFAULT , +DRM_SCHED_POLICY_UNSET , adev->dev);if (r) { DRM_ERROR("Failed to create scheduler on ring %s.\n",
drivers/gpu/drm/etnaviv/etnaviv_sched.c +1 -1
@@ -137,7 +137,7 @@ 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),DRM_SCHED_POLICY_DEFAULT , +dev_name (gpu->dev),DRM_SCHED_POLICY_UNSET , gpu->dev);if (ret)return ret;
drivers/gpu/drm/lima/lima_sched.c +1 -1
@@ -491,7 +491,7 @@ 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,DRM_SCHED_POLICY_DEFAULT , +NULL , name,DRM_SCHED_POLICY_UNSET , pipe->ldev->dev); }
drivers/gpu/drm/msm/adreno/adreno_device.c +2 -2
@@ -809,7 +809,7 @@ 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; -drm_sched_run_wq_stop (sched); +drm_sched_wqueue_stop (sched); } }
@@ -819,7 +819,7 @@ 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; -drm_sched_run_wq_start (sched); +drm_sched_wqueue_start (sched); } }
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -1
@@ -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, -DRM_SCHED_POLICY_DEFAULT , gpu->dev->dev); +DRM_SCHED_POLICY_UNSET , gpu->dev->dev);if (ret) { goto fail; }
drivers/gpu/drm/nouveau/nouveau_sched.c +1 -1
@@ -438,7 +438,7 @@ 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_SCHED_POLICY_DEFAULT , drm->dev->dev); +DRM_SCHED_POLICY_UNSET , drm->dev->dev); }void nouveau_sched_fini (struct nouveau_drm *drm)
drivers/gpu/drm/panfrost/panfrost_job.c +1 -1
@@ -835,7 +835,7 @@ int panfrost_job_init(struct panfrost_device *pfdev) nentries,0 ,msecs_to_jiffies (JOB_TIMEOUT_MS ), pfdev->reset.wq, -NULL ,"pan_js" ,DRM_SCHED_POLICY_DEFAULT , +NULL ,"pan_js" ,DRM_SCHED_POLICY_UNSET , pfdev->dev);if (ret) {dev_err (pfdev->dev,"Failed to create scheduler: %d." , ret);
drivers/gpu/drm/scheduler/sched_entity.c +7 -6
@@ -33,8 +33,8 @@#define to_drm_sched_job (sched_job) \ container_of((sched_job), struct drm_sched_job, queue_node) -static bool bad_policies (struct drm_gpu_scheduler **sched_list, -unsigned int num_sched_list) +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;
@@ -77,7 +77,7 @@ int drm_sched_entity_init(struct drm_sched_entity *entity, atomic_t *guilty) {if (!(entity && sched_list && (num_sched_list ==0 || sched_list[0 ])) || -bad_policies (sched_list, num_sched_list)) +drm_sched_policy_mismatch (sched_list, num_sched_list))return -EINVAL ;memset (entity,0 ,sizeof (struct drm_sched_entity ));
@@ -91,15 +91,16 @@ 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) {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) -return -EINVAL ; + }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 ; } }
drivers/gpu/drm/scheduler/sched_main.c +268 -256
770 lines not on this page · raw message
drivers/gpu/drm/v3d/v3d_sched.c +5 -5
@@ -391,7 +391,7 @@ 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" ,DRM_SCHED_POLICY_DEFAULT , +NULL ,"v3d_bin" ,DRM_SCHED_POLICY_UNSET , v3d->drm.dev);if (ret)return ret;
@@ -400,7 +400,7 @@ 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 , -ULL ,"v3d_render" ,DRM_SCHED_POLICY_DEFAULT , +ULL ,"v3d_render" ,DRM_SCHED_POLICY_UNSET , v3d->drm.dev);if (ret) goto fail;
@@ -409,7 +409,7 @@ 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" ,DRM_SCHED_POLICY_DEFAULT , +NULL ,"v3d_tfu" ,DRM_SCHED_POLICY_UNSET , v3d->drm.dev);if (ret) goto fail;
@@ -419,7 +419,7 @@ 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" ,DRM_SCHED_POLICY_DEFAULT , +NULL ,"v3d_csd" ,DRM_SCHED_POLICY_UNSET , v3d->drm.dev);if (ret) goto fail;
@@ -429,7 +429,7 @@ v3d_sched_init(struct v3d_dev *v3d) hw_jobs_limit, job_hang_limit,msecs_to_jiffies (hang_limit_ms),NULL ,NULL ,"v3d_cache_clean" , - DRM_SCHED_POLICY_DEFAULT, v3d->drm.dev); + DRM_SCHED_POLICY_UNSET, v3d->drm.dev);if (ret) goto fail; }
drivers/gpu/drm/xe/Makefile +1 -0
@@ -57,6 +57,7 @@ xe-y += xe_bb.o \ xe_exec_queue.o \ xe_force_wake.o \ xe_ggtt.o \ + xe_gpu_scheduler.o \ xe_gt.o \ xe_gt_clock.o \ xe_gt_debugfs.o \
drivers/gpu/drm/xe/xe_gpu_scheduler.c new file +102 -0
102 lines not on this page · raw message
drivers/gpu/drm/xe/xe_gpu_scheduler.h new file +75 -0
75 lines not on this page · raw message
drivers/gpu/drm/xe/xe_gpu_scheduler_types.h new file +58 -0
58 lines not on this page · raw message
drivers/gpu/drm/xe/xe_guc_exec_queue_types.h +4 -4
@@ -9,7 +9,7 @@#include <linux/spinlock.h> #include <linux/workqueue.h> -#include <drm/gpu_scheduler.h> +#include "xe_gpu_scheduler_types.h" struct dma_fence ;struct xe_exec_queue ;
@@ -21,16 +21,16 @@ struct xe_guc_exec_queue {/** @q: Backpointer to parent xe_exec_queue */ struct xe_exec_queue *q;/** @sched: GPU scheduler for this xe_exec_queue */ -struct drm_gpu_scheduler sched; +struct xe_gpu_scheduler sched;/** @entity: Scheduler entity for this xe_exec_queue */ -struct drm_sched_entity entity; +struct xe_sched_entity entity;/** * @static_msgs: Static messages for this xe_exec_queue, used when * a message needs to sent through the GPU scheduler but memory * allocations are not allowed. */ #define MAX_STATIC_MSG_TYPE 3 -struct drm_sched_msg static_msgs[MAX_STATIC_MSG_TYPE ]; +struct xe_sched_msg static_msgs[MAX_STATIC_MSG_TYPE ];/** @lr_tdr: long running TDR worker */ struct work_struct lr_tdr;/** @fini_async: do final fini async from this worker */
drivers/gpu/drm/xe/xe_guc_submit.c +70 -72
440 lines not on this page · raw message
drivers/gpu/drm/xe/xe_trace.h +7 -6
@@ -14,6 +14,7 @@#include "xe_bo_types.h" #include "xe_exec_queue_types.h" +#include "xe_gpu_scheduler_types.h" #include "xe_gt_tlb_invalidation_types.h" #include "xe_gt_types.h" #include "xe_guc_exec_queue_types.h"
@@ -290,8 +291,8 @@ DEFINE_EVENT(xe_sched_job, xe_sched_job_ban,TP_ARGS (job) ); -DECLARE_EVENT_CLASS(drm_sched_msg, - TP_PROTO(struct drm_sched_msg *msg), +DECLARE_EVENT_CLASS(xe_sched_msg, + TP_PROTO(struct xe_sched_msg *msg), TP_ARGS(msg), TP_STRUCT__entry(
@@ -309,13 +310,13 @@ DECLARE_EVENT_CLASS(drm_sched_msg, __entry->opcode) ); -DEFINE_EVENT (drm_sched_msg, drm_sched_msg_add, - TP_PROTO(struct drm_sched_msg *msg), +DEFINE_EVENT (xe_sched_msg, xe_sched_msg_add, + TP_PROTO(struct xe_sched_msg *msg),TP_ARGS (msg) ); -DEFINE_EVENT (drm_sched_msg, drm_sched_msg_recv, - TP_PROTO(struct drm_sched_msg *msg), +DEFINE_EVENT (xe_sched_msg, xe_sched_msg_recv, + TP_PROTO(struct xe_sched_msg *msg),TP_ARGS (msg) );
include/drm/gpu_scheduler.h +16 -38
119 lines not on this page · raw message
Signature · 1 line
2.34.1
Patchwork
CI[Intel-xe] ✓ CI.Patch_applied: success for DRM scheduler update
In reply to Matthew Brost== Series Details ==
== Series Details ==
Series: DRM scheduler update
URL : https://patchwork.freedesktop.org/series/125012/
State : success
== Summary ==
=== Applying kernel patches on branch 'drm-xe-next' with base: === Base commit: 6c0d1e36f drm/xe/dg2: Remove one PCI ID === git am output follows ===
Applying: drm/xe: Update to upstream DRM scheduler code
Patchwork
CI[Intel-xe] ✗ CI.checkpatch: warning for DRM scheduler update
In reply to Matthew Brost== Series Details ==
== Series Details ==
Series: DRM scheduler update
URL : https://patchwork.freedesktop.org/series/125012/
State : warning
== Summary ==
+ KERNEL=/kernel
+ git clone https://gitlab.freedesktop.org/drm/maintainer-tools mt Cloning into 'mt'...
warning: redirecting to https://gitlab.freedesktop.org/drm/maintainer-tools.git/
+ git -C mt rev-list -n1 origin/master 63c2b6b160bca2df6efc7bc4cea6f442097d7854
+ cd /kernel
+ git config --global --add safe.directory /kernel
+ git log -n1
commit cb7c851703f33a012c8d67425369aff28652ad1b
Author: Matthew Brost <[email protected]> Date: Wed Oct 11 16:51:26 2023 -0700
drm/xe: Update to upstream DRM scheduler code
The largest change is the message interface has been removed from the DRM scheduler. Xe still needs a message interface so it is implemented in the Xe driver by adding a Xe scheduler layer.
+ /mt/dim checkpatch 6c0d1e36f310682297a5b07336b0883afc77f66e drm-intel cb7c85170 drm/xe: Update to upstream DRM scheduler code Traceback (most recent call last):
File "scripts/spdxcheck.py", line 6, in <module>
from ply import lex, yacc
ModuleNotFoundError: No module named 'ply'
Traceback (most recent call last):
File "scripts/spdxcheck.py", line 6, in <module>
from ply import lex, yacc
ModuleNotFoundError: No module named 'ply'
Traceback (most recent call last):
File "scripts/spdxcheck.py", line 6, in <module>
from ply import lex, yacc
ModuleNotFoundError: No module named 'ply'
-:180: CHECK:LINE_SPACING: Please don't use multiple blank lines #180: FILE: drivers/gpu/drm/scheduler/sched_entity.c:94:
+
-:260: WARNING:LONG_LINE: line length of 202 exceeds 100 columns
#260: FILE: drivers/gpu/drm/scheduler/sched_main.c:93:
+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).");
-:939: CHECK:BRACES: Blank lines aren't necessary before a close brace '}' #939: FILE: drivers/gpu/drm/scheduler/sched_main.c:1280:
+ }
-:1057: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating? #1057:
new file mode 100644
-:1122: WARNING:UNSPECIFIED_INT: Prefer 'unsigned int' to bare use of 'unsigned' #1122: FILE: drivers/gpu/drm/xe/xe_gpu_scheduler.c:61:
+ uint32_t hw_submission, unsigned hang_limit,
-:1122: CHECK:PREFER_KERNEL_TYPES: Prefer kernel type 'u32' over 'uint32_t' #1122: FILE: drivers/gpu/drm/xe/xe_gpu_scheduler.c:61:
+ uint32_t hw_submission, unsigned hang_limit,
-:1185: WARNING:UNSPECIFIED_INT: Prefer 'unsigned int' to bare use of 'unsigned' #1185: FILE: drivers/gpu/drm/xe/xe_gpu_scheduler.h:16:
+ uint32_t hw_submission, unsigned hang_limit,
-:1185: CHECK:PREFER_KERNEL_TYPES: Prefer kernel type 'u32' over 'uint32_t' #1185: FILE: drivers/gpu/drm/xe/xe_gpu_scheduler.h:16:
+ uint32_t hw_submission, unsigned hang_limit,
-:1593: CHECK:PARENTHESIS_ALIGNMENT: Alignment should match open parenthesis #1593: FILE: drivers/gpu/drm/xe/xe_guc_submit.c:1170:
+ err = xe_sched_init(&ge->sched, &drm_sched_ops, &xe_sched_ops, NULL,
q->lrc[0].ring.size / MAX_JOB_SIZE_BYTES,
total: 0 errors, 4 warnings, 5 checks, 1805 lines checked
Patchwork
CI[Intel-xe] ✗ CI.KUnit: failure for DRM scheduler update
In reply to Matthew Brost== Series Details ==
== Series Details ==
Series: DRM scheduler update
URL : https://patchwork.freedesktop.org/series/125012/
State : failure
== Summary ==
+ trap cleanup EXIT
+ /kernel/tools/testing/kunit/kunit.py run --kunitconfig /kernel/drivers/gpu/drm/xe/.kunitconfig ERROR:root:../drivers/gpu/drm/scheduler/sched_main.c: In function ‘drm_sched_get_cleanup_job’:../drivers/gpu/drm/scheduler/sched_main.c:1044:6: error: implicit declaration of function ‘dma_fence_timestamp’ [-Werror=implicit-function-declaration]
1044 | dma_fence_timestamp(&job->s_fence->finished);
| ^~~~~~~~~~~~~~~~~~~cc1: some warnings being treated as errors
make[7]: *** [../scripts/Makefile.build:243: drivers/gpu/drm/scheduler/sched_main.o] Error 1 make[7]: *** Waiting for unfinished jobs....
make[6]: *** [../scripts/Makefile.build:480: drivers/gpu/drm/scheduler] Error 2 make[6]: *** Waiting for unfinished jobs....
make[5]: *** [../scripts/Makefile.build:480: drivers/gpu/drm] Error 2 make[4]: *** [../scripts/Makefile.build:480: drivers/gpu] Error 2
make[3]: *** [../scripts/Makefile.build:480: drivers] Error 2
make[2]: *** [/kernel/Makefile:1913: .] Error 2
make[1]: *** [/kernel/Makefile:234: __sub-make] Error 2
make: *** [Makefile:234: __sub-make] Error 2
[04:22:10] Configuring KUnit Kernel ...
Generating .config ...
Populating config with:
$ make ARCH=um O=.kunit olddefconfig [04:22:14] Building KUnit Kernel ... Populating config with: $ make ARCH=um O=.kunit olddefconfig Building with: $ make ARCH=um O=.kunit --jobs=48 + cleanup ++ stat -c %u:%g /kernel + chown -R 1003:1003 /kernel