[PATCH] drm/sched: Use struct for drm_sched_init() params
Christian König
christian.koenig at amd.com
Wed Jan 22 15:06:10 UTC 2025
Am 22.01.25 um 15:48 schrieb Philipp Stanner:
> On Wed, 2025-01-22 at 15:34 +0100, Christian König wrote:
>> Am 22.01.25 um 15:08 schrieb Philipp Stanner:
>>> drm_sched_init() has a great many parameters and upcoming new
>>> functionality for the scheduler might add even more. Generally, the
>>> great number of parameters reduces readability and has already
>>> caused
>>> one missnaming in:
>>>
>>> commit 6f1cacf4eba7 ("drm/nouveau: Improve variable name in
>>> nouveau_sched_init()").
>>>
>>> Introduce a new struct for the scheduler init parameters and port
>>> all
>>> users.
>>>
>>> Signed-off-by: Philipp Stanner <phasta at kernel.org>
>>> ---
>>> Howdy,
>>>
>>> I have a patch-series in the pipe that will add a `flags` argument
>>> to
>>> drm_sched_init(). I thought it would be wise to first rework the
>>> API as
>>> detailed in this patch. It's really a lot of parameters by now, and
>>> I
>>> would expect that it might get more and more over the years for
>>> special
>>> use cases etc.
>>>
>>> Regards,
>>> P.
>>> ---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 21 +++-
>>> drivers/gpu/drm/etnaviv/etnaviv_sched.c | 20 ++-
>>> drivers/gpu/drm/imagination/pvr_queue.c | 21 +++-
>>> drivers/gpu/drm/lima/lima_sched.c | 21 +++-
>>> drivers/gpu/drm/msm/msm_ringbuffer.c | 22 ++--
>>> drivers/gpu/drm/nouveau/nouveau_sched.c | 20 ++-
>>> drivers/gpu/drm/panfrost/panfrost_job.c | 22 ++--
>>> drivers/gpu/drm/panthor/panthor_mmu.c | 18 ++-
>>> drivers/gpu/drm/panthor/panthor_sched.c | 23 ++--
>>> drivers/gpu/drm/scheduler/sched_main.c | 53 +++-----
>>> drivers/gpu/drm/v3d/v3d_sched.c | 135 +++++++++++++++-
>>> -----
>>> drivers/gpu/drm/xe/xe_execlist.c | 20 ++-
>>> drivers/gpu/drm/xe/xe_gpu_scheduler.c | 19 ++-
>>> include/drm/gpu_scheduler.h | 35 +++++-
>>> 14 files changed, 311 insertions(+), 139 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> index cd4fac120834..c1f03eb5f5ea 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>>> @@ -2821,6 +2821,9 @@ static int
>>> amdgpu_device_init_schedulers(struct amdgpu_device *adev)
>>> {
>>> long timeout;
>>> int r, i;
>>> + struct drm_sched_init_params params;
>> Please keep the reverse xmas tree ordering for variable declaration.
>> E.g. long lines first and variables like "i" and "r" last.
> Okay dokay
>
>> Apart from that looks like a good idea to me.
>>
>>
>>> +
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>>
>>> for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
>>> struct amdgpu_ring *ring = adev->rings[i];
>>> @@ -2844,12 +2847,18 @@ static int
>>> amdgpu_device_init_schedulers(struct amdgpu_device *adev)
>>> break;
>>> }
>>>
>>> - r = drm_sched_init(&ring->sched,
>>> &amdgpu_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - ring->num_hw_submission, 0,
>>> - timeout, adev->reset_domain-
>>>> wq,
>>> - ring->sched_score, ring->name,
>>> - adev->dev);
>>> + params.ops = &amdgpu_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = ring->num_hw_submission;
>>> + params.hang_limit = 0;
>> Could we please remove the hang limit as first step to get this awful
>> feature deprecated?
> Remove it? From the struct you mean?
>
> We can mark it as deprecated in the docstring of the new struct. That's
> what you mean, don't you?
No, the function using this parameter already deprecated. What I meant
is to start to completely remove this feature.
The hang_limit basically says how often the scheduler should try to run
a job over and over again before giving up.
And we already agreed that trying the same thing over and over again and
expecting different results is the definition of insanity :)
So my suggestion is to drop the parameter and drop the job as soon as it
caused a timeout.
Regards,
Christian.
>
> P.
>
>> Thanks,
>> Christian.
>>
>>> + params.timeout = timeout;
>>> + params.timeout_wq = adev->reset_domain->wq;
>>> + params.score = ring->sched_score;
>>> + params.name = ring->name;
>>> + params.dev = adev->dev;
>>> +
>>> + r = drm_sched_init(&ring->sched, ¶ms);
>>> if (r) {
>>> DRM_ERROR("Failed to create scheduler on
>>> ring %s.\n",
>>> ring->name);
>>> diff --git a/drivers/gpu/drm/etnaviv/etnaviv_sched.c
>>> b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
>>> index 5b67eda122db..7d8517f1963e 100644
>>> --- a/drivers/gpu/drm/etnaviv/etnaviv_sched.c
>>> +++ b/drivers/gpu/drm/etnaviv/etnaviv_sched.c
>>> @@ -145,12 +145,22 @@ int etnaviv_sched_push_job(struct
>>> etnaviv_gem_submit *submit)
>>> int etnaviv_sched_init(struct etnaviv_gpu *gpu)
>>> {
>>> int ret;
>>> + struct drm_sched_init_params params;
>>>
>>> - ret = drm_sched_init(&gpu->sched, &etnaviv_sched_ops,
>>> NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - etnaviv_hw_jobs_limit,
>>> etnaviv_job_hang_limit,
>>> - msecs_to_jiffies(500), NULL, NULL,
>>> - dev_name(gpu->dev), gpu->dev);
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> + params.ops = &etnaviv_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = etnaviv_hw_jobs_limit;
>>> + params.hang_limit = etnaviv_job_hang_limit;
>>> + params.timeout = msecs_to_jiffies(500);
>>> + params.timeout_wq = NULL; /* Use the system_wq. */
>>> + params.score = NULL;
>>> + params.name = dev_name(gpu->dev);
>>> + params.dev = gpu->dev;
>>> +
>>> + ret = drm_sched_init(&gpu->sched, ¶ms);
>>> if (ret)
>>> return ret;
>>>
>>> diff --git a/drivers/gpu/drm/imagination/pvr_queue.c
>>> b/drivers/gpu/drm/imagination/pvr_queue.c
>>> index c4f08432882b..03a2ce1a88e7 100644
>>> --- a/drivers/gpu/drm/imagination/pvr_queue.c
>>> +++ b/drivers/gpu/drm/imagination/pvr_queue.c
>>> @@ -1211,10 +1211,13 @@ struct pvr_queue *pvr_queue_create(struct
>>> pvr_context *ctx,
>>> };
>>> struct pvr_device *pvr_dev = ctx->pvr_dev;
>>> struct drm_gpu_scheduler *sched;
>>> + struct drm_sched_init_params sched_params;
>>> struct pvr_queue *queue;
>>> int ctx_state_size, err;
>>> void *cpu_map;
>>>
>>> + memset(&sched_params, 0, sizeof(struct
>>> drm_sched_init_params));
>>> +
>>> if (WARN_ON(type >= sizeof(props)))
>>> return ERR_PTR(-EINVAL);
>>>
>>> @@ -1282,12 +1285,18 @@ struct pvr_queue *pvr_queue_create(struct
>>> pvr_context *ctx,
>>>
>>> queue->timeline_ufo.value = cpu_map;
>>>
>>> - err = drm_sched_init(&queue->scheduler,
>>> - &pvr_queue_sched_ops,
>>> - pvr_dev->sched_wq, 1, 64 * 1024, 1,
>>> - msecs_to_jiffies(500),
>>> - pvr_dev->sched_wq, NULL, "pvr-queue",
>>> - pvr_dev->base.dev);
>>> + sched_params.ops = &pvr_queue_sched_ops;
>>> + sched_params.submit_wq = pvr_dev->sched_wq;
>>> + sched_params.num_rqs = 1;
>>> + sched_params.credit_limit = 64 * 1024;
>>> + sched_params.hang_limit = 1;
>>> + sched_params.timeout = msecs_to_jiffies(500);
>>> + sched_params.timeout_wq = pvr_dev->sched_wq;
>>> + sched_params.score = NULL;
>>> + sched_params.name = "pvr-queue";
>>> + sched_params.dev = pvr_dev->base.dev;
>>> +
>>> + err = drm_sched_init(&queue->scheduler, &sched_params);
>>> if (err)
>>> goto err_release_ufo;
>>>
>>> diff --git a/drivers/gpu/drm/lima/lima_sched.c
>>> b/drivers/gpu/drm/lima/lima_sched.c
>>> index b40c90e97d7e..a64c50fb6d1e 100644
>>> --- a/drivers/gpu/drm/lima/lima_sched.c
>>> +++ b/drivers/gpu/drm/lima/lima_sched.c
>>> @@ -513,20 +513,29 @@ static void lima_sched_recover_work(struct
>>> work_struct *work)
>>>
>>> int lima_sched_pipe_init(struct lima_sched_pipe *pipe, const char
>>> *name)
>>> {
>>> + struct drm_sched_init_params params;
>>> unsigned int timeout = lima_sched_timeout_ms > 0 ?
>>> lima_sched_timeout_ms : 10000;
>>>
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> pipe->fence_context = dma_fence_context_alloc(1);
>>> spin_lock_init(&pipe->fence_lock);
>>>
>>> INIT_WORK(&pipe->recover_work, lima_sched_recover_work);
>>>
>>> - return drm_sched_init(&pipe->base, &lima_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - 1,
>>> - lima_job_hang_limit,
>>> - msecs_to_jiffies(timeout), NULL,
>>> - NULL, name, pipe->ldev->dev);
>>> + params.ops = &lima_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = 1;
>>> + params.hang_limit = lima_job_hang_limit;
>>> + params.timeout = msecs_to_jiffies(timeout);
>>> + params.timeout_wq = NULL; /* Use the system_wq. */
>>> + params.score = NULL;
>>> + params.name = name;
>>> + params.dev = pipe->ldev->dev;
>>> +
>>> + return drm_sched_init(&pipe->base, ¶ms);
>>> }
>>>
>>> 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 c803556a8f64..49a2c7422dc6 100644
>>> --- a/drivers/gpu/drm/msm/msm_ringbuffer.c
>>> +++ b/drivers/gpu/drm/msm/msm_ringbuffer.c
>>> @@ -59,11 +59,13 @@ static const struct drm_sched_backend_ops
>>> msm_sched_ops = {
>>> struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu,
>>> int id,
>>> void *memptrs, uint64_t memptrs_iova)
>>> {
>>> + struct drm_sched_init_params params;
>>> struct msm_ringbuffer *ring;
>>> - long sched_timeout;
>>> char name[32];
>>> int ret;
>>>
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> /* We assume everywhere that MSM_GPU_RINGBUFFER_SZ is a
>>> power of 2 */
>>> BUILD_BUG_ON(!is_power_of_2(MSM_GPU_RINGBUFFER_SZ));
>>>
>>> @@ -95,13 +97,19 @@ struct msm_ringbuffer
>>> *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
>>> ring->memptrs = memptrs;
>>> ring->memptrs_iova = memptrs_iova;
>>>
>>> - /* currently managing hangcheck ourselves: */
>>> - sched_timeout = MAX_SCHEDULE_TIMEOUT;
>>> + params.ops = &msm_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = num_hw_submissions;
>>> + params.hang_limit = 0;
>>> + /* currently managing hangcheck ourselves: */
>>> + params.timeout = MAX_SCHEDULE_TIMEOUT;
>>> + params.timeout_wq = NULL; /* Use the system_wq. */
>>> + params.score = NULL;
>>> + params.name = to_msm_bo(ring->bo)->name;
>>> + params.dev = gpu->dev->dev;
>>>
>>> - ret = drm_sched_init(&ring->sched, &msm_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - num_hw_submissions, 0, sched_timeout,
>>> - NULL, NULL, to_msm_bo(ring->bo)-
>>>> name, gpu->dev->dev);
>>> + ret = drm_sched_init(&ring->sched, ¶ms);
>>> if (ret) {
>>> goto fail;
>>> }
>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c
>>> b/drivers/gpu/drm/nouveau/nouveau_sched.c
>>> index 4412f2711fb5..f20c2e612750 100644
>>> --- a/drivers/gpu/drm/nouveau/nouveau_sched.c
>>> +++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
>>> @@ -404,9 +404,11 @@ nouveau_sched_init(struct nouveau_sched
>>> *sched, struct nouveau_drm *drm,
>>> {
>>> struct drm_gpu_scheduler *drm_sched = &sched->base;
>>> struct drm_sched_entity *entity = &sched->entity;
>>> - const long timeout =
>>> msecs_to_jiffies(NOUVEAU_SCHED_JOB_TIMEOUT_MS);
>>> + struct drm_sched_init_params params;
>>> int ret;
>>>
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> if (!wq) {
>>> wq = alloc_workqueue("nouveau_sched_wq_%d", 0,
>>> WQ_MAX_ACTIVE,
>>> current->pid);
>>> @@ -416,10 +418,18 @@ nouveau_sched_init(struct nouveau_sched
>>> *sched, struct nouveau_drm *drm,
>>> sched->wq = wq;
>>> }
>>>
>>> - ret = drm_sched_init(drm_sched, &nouveau_sched_ops, wq,
>>> - NOUVEAU_SCHED_PRIORITY_COUNT,
>>> - credit_limit, 0, timeout,
>>> - NULL, NULL, "nouveau_sched", drm-
>>>> dev->dev);
>>> + params.ops = &nouveau_sched_ops;
>>> + params.submit_wq = wq;
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = credit_limit;
>>> + params.hang_limit = 0;
>>> + params.timeout =
>>> msecs_to_jiffies(NOUVEAU_SCHED_JOB_TIMEOUT_MS);
>>> + params.timeout_wq = NULL; /* Use the system_wq. */
>>> + params.score = NULL;
>>> + params.name = "nouveau_sched";
>>> + params.dev = drm->dev->dev;
>>> +
>>> + ret = drm_sched_init(drm_sched, ¶ms);
>>> if (ret)
>>> goto fail_wq;
>>>
>>> diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c
>>> b/drivers/gpu/drm/panfrost/panfrost_job.c
>>> index 9b8e82fb8bc4..6b509ff446b5 100644
>>> --- a/drivers/gpu/drm/panfrost/panfrost_job.c
>>> +++ b/drivers/gpu/drm/panfrost/panfrost_job.c
>>> @@ -836,10 +836,13 @@ static irqreturn_t
>>> panfrost_job_irq_handler(int irq, void *data)
>>>
>>> int panfrost_job_init(struct panfrost_device *pfdev)
>>> {
>>> + struct drm_sched_init_params params;
>>> struct panfrost_job_slot *js;
>>> unsigned int nentries = 2;
>>> int ret, j;
>>>
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> /* All GPUs have two entries per queue, but without
>>> jobchain
>>> * disambiguation stopping the right job in the close path
>>> is tricky,
>>> * so let's just advertise one entry in that case.
>>> @@ -872,16 +875,21 @@ int panfrost_job_init(struct panfrost_device
>>> *pfdev)
>>> if (!pfdev->reset.wq)
>>> return -ENOMEM;
>>>
>>> + params.ops = &panfrost_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params.credit_limit = nentries;
>>> + params.hang_limit = 0;
>>> + params.timeout = msecs_to_jiffies(JOB_TIMEOUT_MS);
>>> + params.timeout_wq = pfdev->reset.wq;
>>> + params.score = NULL;
>>> + params.name = "pan_js";
>>> + params.dev = pfdev->dev;
>>> +
>>> for (j = 0; j < NUM_JOB_SLOTS; j++) {
>>> js->queue[j].fence_context =
>>> dma_fence_context_alloc(1);
>>>
>>> - ret = drm_sched_init(&js->queue[j].sched,
>>> - &panfrost_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - nentries, 0,
>>> -
>>> msecs_to_jiffies(JOB_TIMEOUT_MS),
>>> - pfdev->reset.wq,
>>> - NULL, "pan_js", pfdev->dev);
>>> + ret = drm_sched_init(&js->queue[j].sched,
>>> ¶ms);
>>> if (ret) {
>>> dev_err(pfdev->dev, "Failed to create
>>> scheduler: %d.", ret);
>>> goto err_sched;
>>> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c
>>> b/drivers/gpu/drm/panthor/panthor_mmu.c
>>> index a49132f3778b..4362442cbfd8 100644
>>> --- a/drivers/gpu/drm/panthor/panthor_mmu.c
>>> +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
>>> @@ -2268,6 +2268,7 @@ panthor_vm_create(struct panthor_device
>>> *ptdev, bool for_mcu,
>>> u64 full_va_range = 1ull << va_bits;
>>> struct drm_gem_object *dummy_gem;
>>> struct drm_gpu_scheduler *sched;
>>> + struct drm_sched_init_params sched_params;
>>> struct io_pgtable_cfg pgtbl_cfg;
>>> u64 mair, min_va, va_range;
>>> struct panthor_vm *vm;
>>> @@ -2284,6 +2285,8 @@ panthor_vm_create(struct panthor_device
>>> *ptdev, bool for_mcu,
>>> goto err_free_vm;
>>> }
>>>
>>> + memset(&sched_params, 0, sizeof(struct
>>> drm_sched_init_params));
>>> +
>>> mutex_init(&vm->heaps.lock);
>>> vm->for_mcu = for_mcu;
>>> vm->ptdev = ptdev;
>>> @@ -2325,11 +2328,18 @@ panthor_vm_create(struct panthor_device
>>> *ptdev, bool for_mcu,
>>> goto err_mm_takedown;
>>> }
>>>
>>> + sched_params.ops = &panthor_vm_bind_ops;
>>> + sched_params.submit_wq = ptdev->mmu->vm.wq;
>>> + sched_params.num_rqs = 1;
>>> + sched_params.credit_limit = 1;
>>> + sched_params.hang_limit = 0;
>>> /* Bind operations are synchronous for now, no timeout
>>> needed. */
>>> - ret = drm_sched_init(&vm->sched, &panthor_vm_bind_ops,
>>> ptdev->mmu->vm.wq,
>>> - 1, 1, 0,
>>> - MAX_SCHEDULE_TIMEOUT, NULL, NULL,
>>> - "panthor-vm-bind", ptdev->base.dev);
>>> + sched_params.timeout = MAX_SCHEDULE_TIMEOUT;
>>> + sched_params.timeout_wq = NULL; /* Use the system_wq. */
>>> + sched_params.score = NULL;
>>> + sched_params.name = "panthor-vm-bind";
>>> + sched_params.dev = ptdev->base.dev;
>>> + ret = drm_sched_init(&vm->sched, &sched_params);
>>> if (ret)
>>> goto err_free_io_pgtable;
>>>
>>> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c
>>> b/drivers/gpu/drm/panthor/panthor_sched.c
>>> index ef4bec7ff9c7..a324346d302f 100644
>>> --- a/drivers/gpu/drm/panthor/panthor_sched.c
>>> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
>>> @@ -3272,6 +3272,7 @@ group_create_queue(struct panthor_group
>>> *group,
>>> const struct drm_panthor_queue_create *args)
>>> {
>>> struct drm_gpu_scheduler *drm_sched;
>>> + struct drm_sched_init_params sched_params;
>>> struct panthor_queue *queue;
>>> int ret;
>>>
>>> @@ -3289,6 +3290,8 @@ group_create_queue(struct panthor_group
>>> *group,
>>> if (!queue)
>>> return ERR_PTR(-ENOMEM);
>>>
>>> + memset(&sched_params, 0, sizeof(struct
>>> drm_sched_init_params));
>>> +
>>> queue->fence_ctx.id = dma_fence_context_alloc(1);
>>> spin_lock_init(&queue->fence_ctx.lock);
>>> INIT_LIST_HEAD(&queue->fence_ctx.in_flight_jobs);
>>> @@ -3341,17 +3344,23 @@ group_create_queue(struct panthor_group
>>> *group,
>>> if (ret)
>>> goto err_free_queue;
>>>
>>> + sched_params.ops = &panthor_queue_sched_ops;
>>> + sched_params.submit_wq = group->ptdev->scheduler->wq;
>>> + sched_params.num_rqs = 1;
>>> /*
>>> - * Credit limit argument tells us the total number of
>>> instructions
>>> + * The credit limit argument tells us the total number of
>>> instructions
>>> * across all CS slots in the ringbuffer, with some jobs
>>> requiring
>>> * twice as many as others, depending on their profiling
>>> status.
>>> */
>>> - ret = drm_sched_init(&queue->scheduler,
>>> &panthor_queue_sched_ops,
>>> - group->ptdev->scheduler->wq, 1,
>>> - args->ringbuf_size / sizeof(u64),
>>> - 0, msecs_to_jiffies(JOB_TIMEOUT_MS),
>>> - group->ptdev->reset.wq,
>>> - NULL, "panthor-queue", group->ptdev-
>>>> base.dev);
>>> + sched_params.credit_limit = args->ringbuf_size /
>>> sizeof(u64);
>>> + sched_params.hang_limit = 0;
>>> + sched_params.timeout = msecs_to_jiffies(JOB_TIMEOUT_MS);
>>> + sched_params.timeout_wq = group->ptdev->reset.wq;
>>> + sched_params.score = NULL;
>>> + sched_params.name = "panthor-queue";
>>> + sched_params.dev = group->ptdev->base.dev;
>>> +
>>> + ret = drm_sched_init(&queue->scheduler, &sched_params);
>>> if (ret)
>>> goto err_free_queue;
>>>
>>> diff --git a/drivers/gpu/drm/scheduler/sched_main.c
>>> b/drivers/gpu/drm/scheduler/sched_main.c
>>> index 57da84908752..27db748a5269 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_main.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_main.c
>>> @@ -1240,40 +1240,25 @@ static void drm_sched_run_job_work(struct
>>> work_struct *w)
>>> * drm_sched_init - Init a gpu scheduler instance
>>> *
>>> * @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
>>> - * @num_rqs: number of runqueues, one for each priority, up to
>>> DRM_SCHED_PRIORITY_COUNT
>>> - * @credit_limit: the number of credits this scheduler can hold
>>> from all jobs
>>> - * @hang_limit: number of times to allow a job to hang before
>>> dropping it
>>> - * @timeout: timeout value in jiffies for the scheduler
>>> - * @timeout_wq: workqueue to use for timeout work. If NULL, the
>>> system_wq is
>>> - * used
>>> - * @score: optional score atomic shared with other schedulers
>>> - * @name: name used for debugging
>>> - * @dev: target &struct device
>>> + * @params: scheduler initialization parameters
>>> *
>>> * Return 0 on success, otherwise error code.
>>> */
>>> int drm_sched_init(struct drm_gpu_scheduler *sched,
>>> - const struct drm_sched_backend_ops *ops,
>>> - struct workqueue_struct *submit_wq,
>>> - u32 num_rqs, u32 credit_limit, unsigned int
>>> hang_limit,
>>> - long timeout, struct workqueue_struct
>>> *timeout_wq,
>>> - atomic_t *score, const char *name, struct
>>> device *dev)
>>> + const struct drm_sched_init_params *params)
>>> {
>>> int i;
>>>
>>> - sched->ops = ops;
>>> - sched->credit_limit = credit_limit;
>>> - sched->name = name;
>>> - sched->timeout = timeout;
>>> - sched->timeout_wq = timeout_wq ? : system_wq;
>>> - sched->hang_limit = hang_limit;
>>> - sched->score = score ? score : &sched->_score;
>>> - sched->dev = dev;
>>> + sched->ops = params->ops;
>>> + sched->credit_limit = params->credit_limit;
>>> + sched->name = params->name;
>>> + sched->timeout = params->timeout;
>>> + sched->timeout_wq = params->timeout_wq ? : system_wq;
>>> + sched->hang_limit = params->hang_limit;
>>> + sched->score = params->score ? params->score : &sched-
>>>> _score;
>>> + sched->dev = params->dev;
>>>
>>> - if (num_rqs > DRM_SCHED_PRIORITY_COUNT) {
>>> + if (params->num_rqs > DRM_SCHED_PRIORITY_COUNT) {
>>> /* This is a gross violation--tell drivers what
>>> the problem is.
>>> */
>>> drm_err(sched, "%s: num_rqs cannot be greater than
>>> DRM_SCHED_PRIORITY_COUNT\n",
>>> @@ -1288,16 +1273,16 @@ int drm_sched_init(struct drm_gpu_scheduler
>>> *sched,
>>> return 0;
>>> }
>>>
>>> - if (submit_wq) {
>>> - sched->submit_wq = submit_wq;
>>> + if (params->submit_wq) {
>>> + sched->submit_wq = params->submit_wq;
>>> sched->own_submit_wq = false;
>>> } else {
>>> #ifdef CONFIG_LOCKDEP
>>> - sched->submit_wq =
>>> alloc_ordered_workqueue_lockdep_map(name,
>>> -
>>> WQ_MEM_RECLAIM,
>>> -
>>> &drm_sched_lockdep_map);
>>> + sched->submit_wq =
>>> alloc_ordered_workqueue_lockdep_map(
>>> + params->name,
>>> WQ_MEM_RECLAIM,
>>> + &drm_sched_lockdep_map);
>>> #else
>>> - sched->submit_wq = alloc_ordered_workqueue(name,
>>> WQ_MEM_RECLAIM);
>>> + sched->submit_wq = alloc_ordered_workqueue(params-
>>>> name, WQ_MEM_RECLAIM);
>>> #endif
>>> if (!sched->submit_wq)
>>> return -ENOMEM;
>>> @@ -1305,11 +1290,11 @@ int drm_sched_init(struct drm_gpu_scheduler
>>> *sched,
>>> sched->own_submit_wq = true;
>>> }
>>>
>>> - sched->sched_rq = kmalloc_array(num_rqs, sizeof(*sched-
>>>> sched_rq),
>>> + sched->sched_rq = kmalloc_array(params->num_rqs,
>>> sizeof(*sched->sched_rq),
>>> GFP_KERNEL | __GFP_ZERO);
>>> if (!sched->sched_rq)
>>> goto Out_check_own;
>>> - sched->num_rqs = num_rqs;
>>> + sched->num_rqs = params->num_rqs;
>>> for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs;
>>> i++) {
>>> sched->sched_rq[i] = kzalloc(sizeof(*sched-
>>>> sched_rq[i]), GFP_KERNEL);
>>> if (!sched->sched_rq[i])
>>> diff --git a/drivers/gpu/drm/v3d/v3d_sched.c
>>> b/drivers/gpu/drm/v3d/v3d_sched.c
>>> index 99ac4995b5a1..716e6d074d87 100644
>>> --- a/drivers/gpu/drm/v3d/v3d_sched.c
>>> +++ b/drivers/gpu/drm/v3d/v3d_sched.c
>>> @@ -814,67 +814,124 @@ static const struct drm_sched_backend_ops
>>> v3d_cpu_sched_ops = {
>>> .free_job = v3d_cpu_job_free
>>> };
>>>
>>> +/*
>>> + * v3d's scheduler instances are all identical, except for ops and
>>> name.
>>> + */
>>> +static void
>>> +v3d_common_sched_init(struct drm_sched_init_params *params, struct
>>> device *dev)
>>> +{
>>> + memset(params, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> + params->submit_wq = NULL; /* Use the system_wq. */
>>> + params->num_rqs = DRM_SCHED_PRIORITY_COUNT;
>>> + params->credit_limit = 1;
>>> + params->hang_limit = 0;
>>> + params->timeout = msecs_to_jiffies(500);
>>> + params->timeout_wq = NULL; /* Use the system_wq. */
>>> + params->score = NULL;
>>> + params->dev = dev;
>>> +}
>>> +
>>> +static int
>>> +v3d_bin_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_bin_sched_ops;
>>> + params.name = "v3d_bin";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_BIN].sched,
>>> ¶ms);
>>> +}
>>> +
>>> +static int
>>> +v3d_render_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_render_sched_ops;
>>> + params.name = "v3d_render";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_RENDER].sched,
>>> ¶ms);
>>> +}
>>> +
>>> +static int
>>> +v3d_tfu_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_tfu_sched_ops;
>>> + params.name = "v3d_tfu";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_TFU].sched,
>>> ¶ms);
>>> +}
>>> +
>>> +static int
>>> +v3d_csd_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_csd_sched_ops;
>>> + params.name = "v3d_csd";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_CSD].sched,
>>> ¶ms);
>>> +}
>>> +
>>> +static int
>>> +v3d_cache_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_cache_clean_sched_ops;
>>> + params.name = "v3d_cache_clean";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_CACHE_CLEAN].sched,
>>> ¶ms);
>>> +}
>>> +
>>> +static int
>>> +v3d_cpu_sched_init(struct v3d_dev *v3d)
>>> +{
>>> + struct drm_sched_init_params params;
>>> +
>>> + v3d_common_sched_init(¶ms, v3d->drm.dev);
>>> + params.ops = &v3d_cpu_sched_ops;
>>> + params.name = "v3d_cpu";
>>> +
>>> + return drm_sched_init(&v3d->queue[V3D_CPU].sched,
>>> ¶ms);
>>> +}
>>> +
>>> int
>>> v3d_sched_init(struct v3d_dev *v3d)
>>> {
>>> - int hw_jobs_limit = 1;
>>> - int job_hang_limit = 0;
>>> - int hang_limit_ms = 500;
>>> int ret;
>>>
>>> - ret = drm_sched_init(&v3d->queue[V3D_BIN].sched,
>>> - &v3d_bin_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - hw_jobs_limit, job_hang_limit,
>>> - msecs_to_jiffies(hang_limit_ms),
>>> NULL,
>>> - NULL, "v3d_bin", v3d->drm.dev);
>>> + ret = v3d_bin_sched_init(v3d);
>>> if (ret)
>>> return ret;
>>>
>>> - ret = drm_sched_init(&v3d->queue[V3D_RENDER].sched,
>>> - &v3d_render_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - hw_jobs_limit, job_hang_limit,
>>> - msecs_to_jiffies(hang_limit_ms),
>>> NULL,
>>> - NULL, "v3d_render", v3d->drm.dev);
>>> + ret = v3d_render_sched_init(v3d);
>>> if (ret)
>>> goto fail;
>>>
>>> - ret = drm_sched_init(&v3d->queue[V3D_TFU].sched,
>>> - &v3d_tfu_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - hw_jobs_limit, job_hang_limit,
>>> - msecs_to_jiffies(hang_limit_ms),
>>> NULL,
>>> - NULL, "v3d_tfu", v3d->drm.dev);
>>> + ret = v3d_tfu_sched_init(v3d);
>>> if (ret)
>>> goto fail;
>>>
>>> if (v3d_has_csd(v3d)) {
>>> - ret = drm_sched_init(&v3d->queue[V3D_CSD].sched,
>>> - &v3d_csd_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - hw_jobs_limit,
>>> job_hang_limit,
>>> -
>>> msecs_to_jiffies(hang_limit_ms), NULL,
>>> - NULL, "v3d_csd", v3d-
>>>> drm.dev);
>>> + ret = v3d_csd_sched_init(v3d);
>>> if (ret)
>>> goto fail;
>>>
>>> - ret = drm_sched_init(&v3d-
>>>> queue[V3D_CACHE_CLEAN].sched,
>>> - &v3d_cache_clean_sched_ops,
>>> NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - hw_jobs_limit,
>>> job_hang_limit,
>>> -
>>> msecs_to_jiffies(hang_limit_ms), NULL,
>>> - NULL, "v3d_cache_clean", v3d-
>>>> drm.dev);
>>> + ret = v3d_cache_sched_init(v3d);
>>> if (ret)
>>> goto fail;
>>> }
>>>
>>> - ret = drm_sched_init(&v3d->queue[V3D_CPU].sched,
>>> - &v3d_cpu_sched_ops, NULL,
>>> - DRM_SCHED_PRIORITY_COUNT,
>>> - 1, job_hang_limit,
>>> - msecs_to_jiffies(hang_limit_ms),
>>> NULL,
>>> - NULL, "v3d_cpu", v3d->drm.dev);
>>> + ret = v3d_cpu_sched_init(v3d);
>>> if (ret)
>>> goto fail;
>>>
>>> diff --git a/drivers/gpu/drm/xe/xe_execlist.c
>>> b/drivers/gpu/drm/xe/xe_execlist.c
>>> index a8c416a48812..7f29b7f04af4 100644
>>> --- a/drivers/gpu/drm/xe/xe_execlist.c
>>> +++ b/drivers/gpu/drm/xe/xe_execlist.c
>>> @@ -332,10 +332,13 @@ static const struct drm_sched_backend_ops
>>> drm_sched_ops = {
>>> static int execlist_exec_queue_init(struct xe_exec_queue *q)
>>> {
>>> struct drm_gpu_scheduler *sched;
>>> + struct drm_sched_init_params params;
>>> struct xe_execlist_exec_queue *exl;
>>> struct xe_device *xe = gt_to_xe(q->gt);
>>> int err;
>>>
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> xe_assert(xe, !xe_device_uc_enabled(xe));
>>>
>>> drm_info(&xe->drm, "Enabling execlist submission (GuC
>>> submission disabled)\n");
>>> @@ -346,11 +349,18 @@ static int execlist_exec_queue_init(struct
>>> xe_exec_queue *q)
>>>
>>> exl->q = q;
>>>
>>> - err = drm_sched_init(&exl->sched, &drm_sched_ops, NULL, 1,
>>> - q->lrc[0]->ring.size /
>>> MAX_JOB_SIZE_BYTES,
>>> - XE_SCHED_HANG_LIMIT,
>>> XE_SCHED_JOB_TIMEOUT,
>>> - NULL, NULL, q->hwe->name,
>>> - gt_to_xe(q->gt)->drm.dev);
>>> + params.ops = &drm_sched_ops;
>>> + params.submit_wq = NULL; /* Use the system_wq. */
>>> + params.num_rqs = 1;
>>> + params.credit_limit = q->lrc[0]->ring.size /
>>> MAX_JOB_SIZE_BYTES;
>>> + params.hang_limit = XE_SCHED_HANG_LIMIT;
>>> + params.timeout = XE_SCHED_JOB_TIMEOUT;
>>> + params.timeout_wq = NULL; /* Use the system_wq. */
>>> + params.score = NULL;
>>> + params.name = q->hwe->name;
>>> + params.dev = gt_to_xe(q->gt)->drm.dev;
>>> +
>>> + err = drm_sched_init(&exl->sched, ¶ms);
>>> if (err)
>>> goto err_free;
>>>
>>> diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.c
>>> b/drivers/gpu/drm/xe/xe_gpu_scheduler.c
>>> index 50361b4638f9..2129fee83f25 100644
>>> --- a/drivers/gpu/drm/xe/xe_gpu_scheduler.c
>>> +++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.c
>>> @@ -63,13 +63,26 @@ int xe_sched_init(struct xe_gpu_scheduler
>>> *sched,
>>> atomic_t *score, const char *name,
>>> struct device *dev)
>>> {
>>> + struct drm_sched_init_params params;
>>> +
>>> sched->ops = xe_ops;
>>> INIT_LIST_HEAD(&sched->msgs);
>>> INIT_WORK(&sched->work_process_msg,
>>> xe_sched_process_msg_work);
>>>
>>> - return drm_sched_init(&sched->base, ops, submit_wq, 1,
>>> hw_submission,
>>> - hang_limit, timeout, timeout_wq,
>>> score, name,
>>> - dev);
>>> + memset(¶ms, 0, sizeof(struct drm_sched_init_params));
>>> +
>>> + params.ops = ops;
>>> + params.submit_wq = submit_wq;
>>> + params.num_rqs = 1;
>>> + params.credit_limit = hw_submission;
>>> + params.hang_limit = hang_limit;
>>> + params.timeout = timeout;
>>> + params.timeout_wq = timeout_wq;
>>> + params.score = score;
>>> + params.name = name;
>>> + params.dev = dev;
>>> +
>>> + return drm_sched_init(&sched->base, ¶ms);
>>> }
>>>
>>> void xe_sched_fini(struct xe_gpu_scheduler *sched)
>>> diff --git a/include/drm/gpu_scheduler.h
>>> b/include/drm/gpu_scheduler.h
>>> index 95e17504e46a..1a834ef43862 100644
>>> --- a/include/drm/gpu_scheduler.h
>>> +++ b/include/drm/gpu_scheduler.h
>>> @@ -553,12 +553,37 @@ struct drm_gpu_scheduler {
>>> struct device *dev;
>>> };
>>>
>>> +/**
>>> + * struct drm_sched_init_params - parameters for initializing a
>>> DRM GPU scheduler
>>> + *
>>> + * @ops: backend operations provided by the driver
>>> + * @submit_wq: workqueue to use for submission. If NULL, an
>>> ordered wq is
>>> + * allocated and used
>>> + * @num_rqs: Number of run-queues. This is at most
>>> DRM_SCHED_PRIORITY_COUNT,
>>> + * as there's usually one run-queue per priority, but
>>> could be less.
>>> + * @credit_limit: the number of credits this scheduler can hold
>>> from all jobs
>>> + * @hang_limit: number of times to allow a job to hang before
>>> dropping it
>>> + * @timeout: timeout value in jiffies for the scheduler
>>> + * @timeout_wq: workqueue to use for timeout work. If NULL, the
>>> system_wq is
>>> + * used
>>> + * @score: optional score atomic shared with other schedulers
>>> + * @name: name used for debugging
>>> + * @dev: associated device. Used for debugging
>>> + */
>>> +struct drm_sched_init_params {
>>> + const struct drm_sched_backend_ops *ops;
>>> + struct workqueue_struct *submit_wq;
>>> + struct workqueue_struct *timeout_wq;
>>> + u32 num_rqs, credit_limit;
>>> + unsigned int hang_limit;
>>> + long timeout;
>>> + atomic_t *score;
>>> + const char *name;
>>> + struct device *dev;
>>> +};
>>> +
>>> int drm_sched_init(struct drm_gpu_scheduler *sched,
>>> - const struct drm_sched_backend_ops *ops,
>>> - struct workqueue_struct *submit_wq,
>>> - u32 num_rqs, u32 credit_limit, unsigned int
>>> hang_limit,
>>> - long timeout, struct workqueue_struct
>>> *timeout_wq,
>>> - atomic_t *score, const char *name, struct
>>> device *dev);
>>> + const struct drm_sched_init_params *params);
>>>
>>> void drm_sched_fini(struct drm_gpu_scheduler *sched);
>>> int drm_sched_job_init(struct drm_sched_job *job,
More information about the Freedreno
mailing list