github-mirror/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-05-27 00:22:00 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity Graph

drm/v3d: Store the active job inside the queue's state

Browse Source 
Instead of storing the queue's active job in four different variables,
store the active job inside the queue's state. This way, it's possible
to access all active jobs using an index based in `enum v3d_queue`.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Melissa Wen <mwen@igalia.com>
Link: https://lore.kernel.org/r/20250826-v3d-queue-lock-v3-2-979efc43e490@igalia.com
Signed-off-by: Maíra Canal <mcanal@igalia.com>
This commit is contained in:
Maíra Canal 2025-08-26 11:18:59 -03:00
parent e5e3d7bff7
commit 0d3768826d
4 changed files with 38 additions and 55 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
drivers/gpu/drm/v3d/v3d_drv.h
View File

@ -58,6 +58,9 @@ struct v3d_queue_state {
/* Stores the GPU stats for this queue in the global context. */
struct v3d_stats stats;
/* Currently active job for this queue */
struct v3d_job *active_job;
};
/* Performance monitor object. The perform lifetime is controlled by userspace
@ -159,11 +162,6 @@ struct v3d_dev {
struct work_struct overflow_mem_work;
struct v3d_bin_job *bin_job;
struct v3d_render_job *render_job;
struct v3d_tfu_job *tfu_job;
struct v3d_csd_job *csd_job;
struct v3d_queue_state queue[V3D_MAX_QUEUES];
/* Spinlock used to synchronize the overflow memory

7
drivers/gpu/drm/v3d/v3d_gem.c
View File

@ -324,6 +324,7 @@ void
v3d_gem_destroy(struct drm_device *dev)
{
struct v3d_dev *v3d = to_v3d_dev(dev);
enum v3d_queue q;
v3d_sched_fini(v3d);
v3d_gemfs_fini(v3d);
@ -331,10 +332,8 @@ v3d_gem_destroy(struct drm_device *dev)
/* Waiting for jobs to finish would need to be done before
* unregistering V3D.
*/
WARN_ON(v3d->bin_job);
WARN_ON(v3d->render_job);
WARN_ON(v3d->tfu_job);
WARN_ON(v3d->csd_job);
for (q = 0; q < V3D_MAX_QUEUES; q++)
WARN_ON(v3d->queue[q].active_job);
drm_mm_takedown(&v3d->mm);

62
drivers/gpu/drm/v3d/v3d_irq.c
View File

@ -42,6 +42,8 @@ v3d_overflow_mem_work(struct work_struct *work)
container_of(work, struct v3d_dev, overflow_mem_work);
struct drm_device *dev = &v3d->drm;
struct v3d_bo *bo = v3d_bo_create(dev, NULL /* XXX: GMP */, 256 * 1024);
struct v3d_queue_state *queue = &v3d->queue[V3D_BIN];
struct v3d_bin_job *bin_job;
struct drm_gem_object *obj;
unsigned long irqflags;
@ -61,13 +63,15 @@ v3d_overflow_mem_work(struct work_struct *work)
* some binner pool anyway.
*/
spin_lock_irqsave(&v3d->job_lock, irqflags);
if (!v3d->bin_job) {
bin_job = (struct v3d_bin_job *)queue->active_job;
if (!bin_job) {
spin_unlock_irqrestore(&v3d->job_lock, irqflags);
goto out;
}
drm_gem_object_get(obj);
list_add_tail(&bo->unref_head, &v3d->bin_job->render->unref_list);
list_add_tail(&bo->unref_head, &bin_job->render->unref_list);
spin_unlock_irqrestore(&v3d->job_lock, irqflags);
v3d_mmu_flush_all(v3d);
@ -79,6 +83,20 @@ v3d_overflow_mem_work(struct work_struct *work)
drm_gem_object_put(obj);
}
static void
v3d_irq_signal_fence(struct v3d_dev *v3d, enum v3d_queue q,
void (*trace_irq)(struct drm_device *, uint64_t))
{
struct v3d_queue_state *queue = &v3d->queue[q];
struct v3d_fence *fence = to_v3d_fence(queue->active_job->irq_fence);
v3d_job_update_stats(queue->active_job, q);
trace_irq(&v3d->drm, fence->seqno);
queue->active_job = NULL;
dma_fence_signal(&fence->base);
}
static irqreturn_t
v3d_irq(int irq, void *arg)
{
@ -102,41 +120,17 @@ v3d_irq(int irq, void *arg)
}
if (intsts & V3D_INT_FLDONE) {
struct v3d_fence *fence =
to_v3d_fence(v3d->bin_job->base.irq_fence);
v3d_job_update_stats(&v3d->bin_job->base, V3D_BIN);
trace_v3d_bcl_irq(&v3d->drm, fence->seqno);
v3d->bin_job = NULL;
dma_fence_signal(&fence->base);
v3d_irq_signal_fence(v3d, V3D_BIN, trace_v3d_bcl_irq);
status = IRQ_HANDLED;
}
if (intsts & V3D_INT_FRDONE) {
struct v3d_fence *fence =
to_v3d_fence(v3d->render_job->base.irq_fence);
v3d_job_update_stats(&v3d->render_job->base, V3D_RENDER);
trace_v3d_rcl_irq(&v3d->drm, fence->seqno);
v3d->render_job = NULL;
dma_fence_signal(&fence->base);
v3d_irq_signal_fence(v3d, V3D_RENDER, trace_v3d_rcl_irq);
status = IRQ_HANDLED;
}
if (intsts & V3D_INT_CSDDONE(v3d->ver)) {
struct v3d_fence *fence =
to_v3d_fence(v3d->csd_job->base.irq_fence);
v3d_job_update_stats(&v3d->csd_job->base, V3D_CSD);
trace_v3d_csd_irq(&v3d->drm, fence->seqno);
v3d->csd_job = NULL;
dma_fence_signal(&fence->base);
v3d_irq_signal_fence(v3d, V3D_CSD, trace_v3d_csd_irq);
status = IRQ_HANDLED;
}
@ -168,15 +162,7 @@ v3d_hub_irq(int irq, void *arg)
V3D_WRITE(V3D_HUB_INT_CLR, intsts);
if (intsts & V3D_HUB_INT_TFUC) {
struct v3d_fence *fence =
to_v3d_fence(v3d->tfu_job->base.irq_fence);
v3d_job_update_stats(&v3d->tfu_job->base, V3D_TFU);
trace_v3d_tfu_irq(&v3d->drm, fence->seqno);
v3d->tfu_job = NULL;
dma_fence_signal(&fence->base);
v3d_irq_signal_fence(v3d, V3D_TFU, trace_v3d_tfu_irq);
status = IRQ_HANDLED;
}

16
drivers/gpu/drm/v3d/v3d_sched.c
View File

@ -232,7 +232,7 @@ static struct dma_fence *v3d_bin_job_run(struct drm_sched_job *sched_job)
if (unlikely(job->base.base.s_fence->finished.error)) {
spin_lock_irqsave(&v3d->job_lock, irqflags);
v3d->bin_job = NULL;
v3d->queue[V3D_BIN].active_job = NULL;
spin_unlock_irqrestore(&v3d->job_lock, irqflags);
return NULL;
}
@ -241,7 +241,7 @@ static struct dma_fence *v3d_bin_job_run(struct drm_sched_job *sched_job)
* v3d_overflow_mem_work().
*/
spin_lock_irqsave(&v3d->job_lock, irqflags);
v3d->bin_job = job;
v3d->queue[V3D_BIN].active_job = &job->base;
/* Clear out the overflow allocation, so we don't
* reuse the overflow attached to a previous job.
*/
@ -290,11 +290,11 @@ static struct dma_fence *v3d_render_job_run(struct drm_sched_job *sched_job)
struct dma_fence *fence;
if (unlikely(job->base.base.s_fence->finished.error)) {
v3d->render_job = NULL;
v3d->queue[V3D_RENDER].active_job = NULL;
return NULL;
}
v3d->render_job = job;
v3d->queue[V3D_RENDER].active_job = &job->base;
/* Can we avoid this flush? We need to be careful of
* scheduling, though -- imagine job0 rendering to texture and
@ -338,11 +338,11 @@ v3d_tfu_job_run(struct drm_sched_job *sched_job)
struct dma_fence *fence;
if (unlikely(job->base.base.s_fence->finished.error)) {
v3d->tfu_job = NULL;
v3d->queue[V3D_TFU].active_job = NULL;
return NULL;
}
v3d->tfu_job = job;
v3d->queue[V3D_TFU].active_job = &job->base;
fence = v3d_fence_create(v3d, V3D_TFU);
if (IS_ERR(fence))
@ -386,11 +386,11 @@ v3d_csd_job_run(struct drm_sched_job *sched_job)
int i, csd_cfg0_reg;
if (unlikely(job->base.base.s_fence->finished.error)) {
v3d->csd_job = NULL;
v3d->queue[V3D_CSD].active_job = NULL;
return NULL;
}
v3d->csd_job = job;
v3d->queue[V3D_CSD].active_job = &job->base;
v3d_invalidate_caches(v3d);