Re: [PATCH] drm/scheduler: Fix drm_sched_entity_set_priority()

Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxx> · Mon, 22 Jul 2024 15:43:54 +0100

On 22/07/2024 15:06, Christian König wrote:
Am 22.07.24 um 15:52 schrieb Tvrtko Ursulin:

On 19/07/2024 16:18, Christian König wrote:
Am 19.07.24 um 15:02 schrieb Christian König:
Am 19.07.24 um 11:47 schrieb Tvrtko Ursulin:
From: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxx>

Long time ago in commit b3ac17667f11 ("drm/scheduler: rework entity
creation") a change was made which prevented priority changes for 
entities
with only one assigned scheduler.

The commit reduced drm_sched_entity_set_priority() to simply update 
the
entities priority, but the run queue selection logic in
drm_sched_entity_select_rq() was never able to actually change the
originally assigned run queue.

In practice that only affected amdgpu, being the only driver which 
can do
dynamic priority changes. And that appears was attempted to be 
rectified
there in 2316a86bde49 ("drm/amdgpu: change hw sched list on ctx 
priority
override").

A few unresolved problems however were that this only fixed
drm_sched_entity_set_priority() *if* 
drm_sched_entity_modify_sched() was
called first. That was not documented anywhere.

Secondly, this only works if drm_sched_entity_modify_sched() is 
actually
called, which in amdgpu's case today is true only for gfx and compute.
Priority changes for other engines with only one scheduler 
assigned, such
as jpeg and video decode will still not work.

Note that this was also noticed in 981b04d96856 ("drm/sched: 
improve docs
around drm_sched_entity").

Completely different set of non-obvious confusion was that whereas
drm_sched_entity_init() was not keeping the passed in list of 
schedulers
(courtesy of 8c23056bdc7a ("drm/scheduler: do not keep a copy of sched
list")), drm_sched_entity_modify_sched() was disagreeing with that and
would simply assign the single item list.

That incosistency appears to be semi-silently fixed in ac4eb83ab255
("drm/sched: select new rq even if there is only one v3").

What was also not documented is why it was important to not keep the
list of schedulers when there is only one. I suspect it could have
something to do with the fact the passed in array is on stack for many
callers with just one scheduler. With more than one scheduler 
amdgpu is
the only caller, and there the container is not on the stack. 
Keeping a
stack backed list in the entity would obviously be undefined behaviour
*if* the list was kept.

Amdgpu however did only stop passing in stack backed container for 
the more
than one scheduler case in 977f7e1068be ("drm/amdgpu: allocate 
entities on
demand"). Until then I suspect dereferencing freed stack from
drm_sched_entity_select_rq() was still present.

In order to untangle all that and fix priority changes this patch is
bringing back the entity owned container for storing the passed in
scheduler list.

Please don't. That makes the mess just more horrible.

The background of not keeping the array is to intentionally prevent 
the priority override from working.

The bug is rather that adding drm_sched_entity_modify_sched() messed 
this up.

To give more background: Amdgpu has two different ways of handling 
priority:
1. The priority in the DRM scheduler.
2. Different HW rings with different priorities.

Your analysis is correct that drm_sched_entity_init() initially 
dropped the scheduler list to avoid using a stack allocated list, and 
that functionality is still used in amdgpu_ctx_init_entity() for 
example.

Setting the scheduler priority was basically just a workaround 
because we didn't had the hw priorities at that time. Since that is 
no longer the case I suggest to just completely drop the 
drm_sched_entity_set_priority() function instead.

Removing drm_sched_entity_set_priority() is one thing, but we also 
need to clear up the sched_list container ownership issue. It is 
neither documented, nor robustly handled in the code. The "num_scheds 
== 1" special casing throughout IMHO has to go too.

I disagree. Keeping the scheduler list in the entity is only useful for 
load balancing.

As long as only one scheduler is provided and we don't load balance the 
entity doesn't needs the scheduler list in the first place.

Once set_priority is removed then it indeed it doesn't. But even when it 
is removed it needs documenting who owns the passed in container. Today 
drivers are okay to pass a stack array when it is one element, but if 
they did it with more than one they would be in for a nasty surprise.

Another thing if you want to get rid of frontend priority handling is 
to stop configuring scheduler instances with DRM_SCHED_PRIORITY_COUNT 
priority levels, to avoid wasting memory on pointless run queues.

I would rather like to completely drop the RR with the runlists 
altogether and keep only the FIFO approach around. This way priority can 
be implemented by boosting the score of submissions by a certain degree.

You mean larger refactoring of the scheduler removing the 1:N between 
drm_sched and drm_sched_rq?

And final thing is to check whether the locking in 
drm_sched_entity_modify_sched() is okay. Because according to kerneldoc:

 * Note that this must be called under the same common lock for 
@entity as
 * drm_sched_job_arm() and drm_sched_entity_push_job(), or the driver 
needs to
 * guarantee through some other means that this is never called while 
new jobs
 * can be pushed to @entity.

I don't see that is the case. Priority override is under 
amdgpu_ctx_mgr->lock, while job arm and push appear not. I also cannot 
spot anything else preventing amdgpu_sched_ioctl() running in parallel 
to everything else.

Yeah that's certainly incorrect as well. drm_sched_entity_modify_sched() 
should grab entity->rq_lock instead.

Ah cool. Well not cool, but good problem has been identified. Are you 
going to work in it or should I? Once we agree the correct fix for all 
this I am happy to write more patches if you are too busy.

In general scheduler priorities were meant to be used for things like 
kernel queues which would always have higher priority than user space 
submissions and using them for userspace turned out to be not such a 
good idea.

Out of curiousity what were the problems? I cannot think of anything 
fundamental since there are priorities at the backend level after all, 
just fewer levels.

A higher level queue can starve lower level queues to the point that 
they never get a chance to get anything running.

Oh that.. well I call that implementation details. :) Because nowhere in 
the uapi is I think guaranteed execution ordering needs to be strictly 
in descending priority. This potentially goes back to what you said 
above that a potential larger rewrite might be beneficial. Implementing 
some smarter scheduling. Although the issue will be it is just frontend 
scheduling after all. So a bit questionable to invest in making it too 
smart.

I think this goes more back to what I was suggesting during early xe 
days, that potentially drm scheduler should be split into dependency 
handling part and the scheduler part. Drivers with 1:1 entity:scheduler 
and full hardware/firmware scheduling do not really need neither fifo or rr.

This basically means that userspace gets a chance to submit infinity 
fences with all the bad consequences.


I mean one problem unrelated to this discussion is this:

void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
{
    struct dma_fence *fence;
    struct drm_gpu_scheduler *sched;
    struct drm_sched_rq *rq;

    /* queue non-empty, stay on the same engine */
    if (spsc_queue_count(&entity->job_queue))
        return;

Which makes it look like the entity with a constant trickle of jobs 
will never actually manage to change it's run queue. Neither today, 
nor after the potential drm_sched_entity_set_priority() removal.

That's intentional and based on how the scheduler load balancing works.

I see that it is intentional but if it can silently prevent priority 
changes (even hw priority) it is not very solid. Unless I am missing 
something here.

Regards,

Tvrtko


Regards,
Christian.


Regards,

Tvrtko

Regards,
Christian.


Container is now owned by the entity and the pointers are
owned by the drivers. List of schedulers is always kept including 
for the
one scheduler case.

The patch can therefore also removes the single scheduler special 
case,
which means that priority changes should now work (be able to 
change the
selected run-queue) for all drivers and engines. In other words
drm_sched_entity_set_priority() should now just work for all cases.

To enable maintaining its own container some API calls needed to 
grow a
capability for returning success/failure, which is a change which
percolates mostly through amdgpu source.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxx>
Fixes: b3ac17667f11 ("drm/scheduler: rework entity creation")
References: 8c23056bdc7a ("drm/scheduler: do not keep a copy of 
sched list")
References: 977f7e1068be ("drm/amdgpu: allocate entities on demand")
References: 2316a86bde49 ("drm/amdgpu: change hw sched list on ctx 
priority override")
References: ac4eb83ab255 ("drm/sched: select new rq even if there 
is only one v3")
References: 981b04d96856 ("drm/sched: improve docs around 
drm_sched_entity")
Cc: Christian König <christian.koenig@xxxxxxx>
Cc: Alex Deucher <alexander.deucher@xxxxxxx>
Cc: Luben Tuikov <ltuikov89@xxxxxxxxx>
Cc: Matthew Brost <matthew.brost@xxxxxxxxx>
Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
Cc: amd-gfx@xxxxxxxxxxxxxxxxxxxxx
Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx
Cc: <stable@xxxxxxxxxxxxxxx> # v5.6+
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c   | 31 +++++---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h   |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c | 13 +--
  drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c   |  3 +-
  drivers/gpu/drm/scheduler/sched_entity.c  | 96 
++++++++++++++++-------
  include/drm/gpu_scheduler.h               | 16 ++--
  6 files changed, 100 insertions(+), 61 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index 5cb33ac99f70..387247f8307e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -802,15 +802,15 @@ struct dma_fence *amdgpu_ctx_get_fence(struct 
amdgpu_ctx *ctx,
      return fence;
  }
  -static void amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx,
-                       struct amdgpu_ctx_entity *aentity,
-                       int hw_ip,
-                       int32_t priority)
+static int amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx,
+                      struct amdgpu_ctx_entity *aentity,
+                      int hw_ip,
+                      int32_t priority)
  {
      struct amdgpu_device *adev = ctx->mgr->adev;
-    unsigned int hw_prio;
      struct drm_gpu_scheduler **scheds = NULL;
-    unsigned num_scheds;
+    unsigned int hw_prio, num_scheds;
+    int ret = 0;
        /* set sw priority */
      drm_sched_entity_set_priority(&aentity->entity,
@@ -822,16 +822,18 @@ static void 
amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx,
          hw_prio = array_index_nospec(hw_prio, AMDGPU_RING_PRIO_MAX);
          scheds = adev->gpu_sched[hw_ip][hw_prio].sched;
          num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;
- drm_sched_entity_modify_sched(&aentity->entity, scheds,
-                          num_scheds);
+        ret = drm_sched_entity_modify_sched(&aentity->entity, scheds,
+                            num_scheds);
      }
+
+    return ret;
  }
  -void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
-                  int32_t priority)
+int amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, int32_t 
priority)
  {
      int32_t ctx_prio;
      unsigned i, j;
+    int ret;
        ctx->override_priority = priority;
  @@ -842,10 +844,15 @@ void amdgpu_ctx_priority_override(struct 
amdgpu_ctx *ctx,
              if (!ctx->entities[i][j])
                  continue;
  -            amdgpu_ctx_set_entity_priority(ctx, 
ctx->entities[i][j],
-                               i, ctx_prio);
+            ret = amdgpu_ctx_set_entity_priority(ctx,
+                                 ctx->entities[i][j],
+                                 i, ctx_prio);
+            if (ret)
+                return ret;
          }
      }
+
+    return 0;
  }
    int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
index 85376baaa92f..835661515e33 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
@@ -82,7 +82,7 @@ struct dma_fence *amdgpu_ctx_get_fence(struct 
amdgpu_ctx *ctx,
                         struct drm_sched_entity *entity,
                         uint64_t seq);
  bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio);
-void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, int32_t 
ctx_prio);
+int amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, int32_t 
ctx_prio);
    int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
               struct drm_file *filp);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
index 863b2a34b2d6..944edb7f00a2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
@@ -54,12 +54,15 @@ static int 
amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
        mgr = &fpriv->ctx_mgr;
      mutex_lock(&mgr->lock);
-    idr_for_each_entry(&mgr->ctx_handles, ctx, id)
-        amdgpu_ctx_priority_override(ctx, priority);
+    idr_for_each_entry(&mgr->ctx_handles, ctx, id) {
+        r = amdgpu_ctx_priority_override(ctx, priority);
+        if (r)
+            break;
+    }
      mutex_unlock(&mgr->lock);
        fdput(f);
-    return 0;
+    return r;
  }
    static int amdgpu_sched_context_priority_override(struct 
amdgpu_device *adev,
@@ -88,11 +91,11 @@ static int 
amdgpu_sched_context_priority_override(struct amdgpu_device *adev,
          return -EINVAL;
      }
  -    amdgpu_ctx_priority_override(ctx, priority);
+    r = amdgpu_ctx_priority_override(ctx, priority);
      amdgpu_ctx_put(ctx);
      fdput(f);
  -    return 0;
+    return r;
  }
    int amdgpu_sched_ioctl(struct drm_device *dev, void *data,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c 
b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
index 81fb99729f37..2453decc73c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
@@ -1362,8 +1362,7 @@ static int vcn_v4_0_5_limit_sched(struct 
amdgpu_cs_parser *p,
        scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_ENC]
          [AMDGPU_RING_PRIO_0].sched;
-    drm_sched_entity_modify_sched(job->base.entity, scheds, 1);
-    return 0;
+    return drm_sched_entity_modify_sched(job->base.entity, scheds, 
1);
  }
    static int vcn_v4_0_5_dec_msg(struct amdgpu_cs_parser *p, 
struct amdgpu_job *job,
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c 
b/drivers/gpu/drm/scheduler/sched_entity.c
index 58c8161289fe..cb5cc65f461d 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -45,7 +45,12 @@
   * @guilty: atomic_t set to 1 when a job on this queue
   *          is found to be guilty causing a timeout
   *
- * Note that the &sched_list must have at least one element to 
schedule the entity.
+ * Note that the &sched_list must have at least one element to 
schedule the
+ * entity.
+ *
+ * The individual drm_gpu_scheduler pointers have borrow 
semantics, ie.
+ * they must remain valid during entities lifetime, while the 
containing
+ * array can be freed after this call returns.
   *
   * For changing @priority later on at runtime see
   * drm_sched_entity_set_priority(). For changing the set of 
schedulers
@@ -69,27 +74,24 @@ int drm_sched_entity_init(struct 
drm_sched_entity *entity,
      INIT_LIST_HEAD(&entity->list);
      entity->rq = NULL;
      entity->guilty = guilty;
-    entity->num_sched_list = num_sched_list;
      entity->priority = priority;
-    /*
-     * It's perfectly valid to initialize an entity without having 
a valid
-     * scheduler attached. It's just not valid to use the 
scheduler before it
-     * is initialized itself.
-     */
-    entity->sched_list = num_sched_list > 1 ? sched_list : NULL;
      RCU_INIT_POINTER(entity->last_scheduled, NULL);
      RB_CLEAR_NODE(&entity->rb_tree_node);
  -    if (num_sched_list && !sched_list[0]->sched_rq) {
-        /* Since every entry covered by num_sched_list
-         * should be non-NULL and therefore we warn drivers
-         * not to do this and to fix their DRM calling order.
-         */
-        pr_warn("%s: called with uninitialized scheduler\n", 
__func__);
-    } else if (num_sched_list) {
-        /* The "priority" of an entity cannot exceed the number of 
run-queues of a
-         * scheduler. Protect against num_rqs being 0, by 
converting to signed. Choose
-         * the lowest priority available.
+    if (num_sched_list) {
+        int ret;
+
+        ret = drm_sched_entity_modify_sched(entity,
+                            sched_list,
+                            num_sched_list);
+        if (ret)
+            return ret;
+
+        /*
+         * The "priority" of an entity cannot exceed the number of
+         * run-queues of a scheduler. Protect against num_rqs 
being 0,
+         * by converting to signed. Choose the lowest priority
+         * available.
           */
          if (entity->priority >= sched_list[0]->num_rqs) {
              drm_err(sched_list[0], "entity with out-of-bounds 
priority:%u num_rqs:%u\n",
@@ -122,19 +124,58 @@ EXPORT_SYMBOL(drm_sched_entity_init);
   *         existing entity->sched_list
   * @num_sched_list: number of drm sched in sched_list
   *
+ * The individual drm_gpu_scheduler pointers have borrow 
semantics, ie.
+ * they must remain valid during entities lifetime, while the 
containing
+ * array can be freed after this call returns.
+ *
   * Note that this must be called under the same common lock for 
@entity as
   * drm_sched_job_arm() and drm_sched_entity_push_job(), or the 
driver needs to
   * guarantee through some other means that this is never called 
while new jobs
   * can be pushed to @entity.
+ *
+ * Returns zero on success and a negative error code on failure.
   */
-void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
-                    struct drm_gpu_scheduler **sched_list,
-                    unsigned int num_sched_list)
+int 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);
+    struct drm_gpu_scheduler **new, **old;
+    unsigned int i;
  -    entity->sched_list = sched_list;
+    if (!(entity && sched_list && (num_sched_list == 0 || 
sched_list[0])))
+        return -EINVAL;
+
+    /*
+     * It's perfectly valid to initialize an entity without having 
a valid
+     * scheduler attached. It's just not valid to use the 
scheduler before
+     * it is initialized itself.
+     *
+     * Since every entry covered by num_sched_list should be 
non-NULL and
+     * therefore we warn drivers not to do this and to fix their 
DRM calling
+     * order.
+     */
+    for (i = 0; i < num_sched_list; i++) {
+        if (!sched_list[i]) {
+            pr_warn("%s: called with uninitialized scheduler %u!\n",
+                __func__, i);
+            return -EINVAL;
+        }
+    }
+
+    new = kmemdup_array(sched_list,
+                num_sched_list,
+                sizeof(*sched_list),
+                GFP_KERNEL);
+    if (!new)
+        return -ENOMEM;
+
+    old = entity->sched_list;
+    entity->sched_list = new;
      entity->num_sched_list = num_sched_list;
+
+    kfree(old);
+
+    return 0;
  }
  EXPORT_SYMBOL(drm_sched_entity_modify_sched);
  @@ -341,6 +382,8 @@ void drm_sched_entity_fini(struct 
drm_sched_entity *entity)
dma_fence_put(rcu_dereference_check(entity->last_scheduled, true));
      RCU_INIT_POINTER(entity->last_scheduled, NULL);
+
+    kfree(entity->sched_list);
  }
  EXPORT_SYMBOL(drm_sched_entity_fini);
  @@ -531,10 +574,6 @@ void drm_sched_entity_select_rq(struct 
drm_sched_entity *entity)
      struct drm_gpu_scheduler *sched;
      struct drm_sched_rq *rq;
  -    /* single possible engine and already selected */
-    if (!entity->sched_list)
-        return;
-
      /* queue non-empty, stay on the same engine */
      if (spsc_queue_count(&entity->job_queue))
          return;
@@ -561,9 +600,6 @@ void drm_sched_entity_select_rq(struct 
drm_sched_entity *entity)
          entity->rq = rq;
      }
      spin_unlock(&entity->rq_lock);
-
-    if (entity->num_sched_list == 1)
-        entity->sched_list = NULL;
  }
    /**
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index 5acc64954a88..09e1d063a5c0 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -110,18 +110,12 @@ struct drm_sched_entity {
      /**
       * @sched_list:
       *
-     * A list of schedulers (struct drm_gpu_scheduler). Jobs from 
this entity can
-     * be scheduled on any scheduler on this list.
+     * A list of schedulers (struct drm_gpu_scheduler). Jobs from 
this
+     * entity can be scheduled on any scheduler on this list.
       *
       * This can be modified by calling 
drm_sched_entity_modify_sched().
       * Locking is entirely up to the driver, see the above 
function for more
       * details.
-     *
-     * This will be set to NULL if &num_sched_list equals 1 and 
@rq has been
-     * set already.
-     *
-     * FIXME: This means priority changes through
-     * drm_sched_entity_set_priority() will be lost henceforth in 
this case.
       */
      struct drm_gpu_scheduler        **sched_list;
  @@ -568,9 +562,9 @@ int 
drm_sched_job_add_implicit_dependencies(struct drm_sched_job *job,
                          bool write);
    -void drm_sched_entity_modify_sched(struct drm_sched_entity 
*entity,
-                    struct drm_gpu_scheduler **sched_list,
-                                   unsigned int num_sched_list);
+int 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);









