[PATCH 2/4] drm/amdgpu: keep copy of VRAM lost counter in job

[Liu, Monk]

Christian, 

Can you elaborate what benefit it brings that keep a counter copy in JOB ? 
Because you know when vram lost happens, the thing real matter is : if the BO in BO_LIST of this job modified/created before this vram lost,
Not that if the job creating timestamp ...

Since kernel side now doesn't implement a way to compare counter per BO (which is too low efficient considering OCL app that have 5000+ BO in BO_LIST), 

if we always consider BO is created/modified since this job's context created (you know many job can refer to one specific BO like texture, which is long time ago modified before JOB creation), than we just need to compare the counter in ctx, 

BR Monk

-----Original Message-----
From: amd-gfx [mailto:amd-gfx-bounces at lists.freedesktop.org] On Behalf Of Christian K?nig
Sent: Monday, October 9, 2017 11:35 PM
To: amd-gfx at lists.freedesktop.org
Subject: [PATCH 2/4] drm/amdgpu: keep copy of VRAM lost counter in job

From: Christian König <christian.koenig at amd.com>

Instead of reading the current counter from fpriv.

Signed-off-by: Christian König <christian.koenig at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c  |  2 ++  drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 13 +++++++------
 3 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 6a8f169..0b8b5c9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1116,6 +1116,7 @@ struct amdgpu_job {
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
 	uint32_t		oa_base, oa_size;
+	uint32_t		vram_lost_counter;
 
 	/* user fence handling */
 	uint64_t		uf_addr;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index ba3e645..fa8502f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -170,6 +170,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 	if (ret)
 		goto free_all_kdata;
 
+	p->job->vram_lost_counter = fpriv->vram_lost_counter;
+
 	if (p->uf_entry.robj)
 		p->job->uf_addr = uf_offset;
 	kfree(chunk_array);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 4510627..c76d17c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -61,6 +61,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
 	(*job)->vm = vm;
 	(*job)->ibs = (void *)&(*job)[1];
 	(*job)->num_ibs = num_ibs;
+	(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
 
 	amdgpu_sync_create(&(*job)->sync);
 	amdgpu_sync_create(&(*job)->dep_sync);
@@ -177,8 +178,8 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)  static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)  {
 	struct dma_fence *fence = NULL;
+	struct amdgpu_device *adev;
 	struct amdgpu_job *job;
-	struct amdgpu_fpriv *fpriv = NULL;
 	int r;
 
 	if (!sched_job) {
@@ -186,17 +187,17 @@ static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)
 		return NULL;
 	}
 	job = to_amdgpu_job(sched_job);
+	adev = job->adev;
 
 	BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL));
 
 	trace_amdgpu_sched_run_job(job);
-	if (job->vm)
-		fpriv = container_of(job->vm, struct amdgpu_fpriv, vm);
 	/* skip ib schedule when vram is lost */
-	if (fpriv && amdgpu_kms_vram_lost(job->adev, fpriv))
+	if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter)) {
 		DRM_ERROR("Skip scheduling IBs!\n");
-	else {
-		r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job, &fence);
+	} else {
+		r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job,
+				       &fence);
 		if (r)
 			DRM_ERROR("Error scheduling IBs (%d)\n", r);
 	}
--
2.7.4

_______________________________________________
amd-gfx mailing list
amd-gfx at lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx