<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">I think that the IB tests already set
the ready flag to false when something goes wrong, but for the
ring tests your probably want to double check and maybe generalize
that.<br>
<br>
Christian.<br>
<br>
Am 24.02.20 um 19:15 schrieb Das, Nirmoy:<br>
</div>
<blockquote type="cite" cite="mid:DM5PR12MB23765A9C06DB2B62ECF0B6208BEC0@DM5PR12MB2376.namprd12.prod.outlook.com">
<p style="font-family:Arial;font-size:10pt;color:#0078D7;margin:15pt;" align="Left">
[AMD Official Use Only - Internal Distribution Only]<br>
</p>
<br>
<div>
<div data-ogsc="" style="">
<meta content="text/html; charset=us-ascii" data-ogsc="" style="">
</div>
<div dir="auto" style="color: rgb(33, 33, 33); background-color:
rgb(255, 255, 255); text-align: left;">
Thanks Christian. I will try to send a updated patch soon. </div>
<div id="ms-outlook-mobile-signature" data-ogsc="" style="text-align: left;" dir="auto">
<div><br>
</div>
Get <a href="https://aka.ms/ghei36" moz-do-not-send="true">Outlook
for Android</a></div>
<div id="id-3a39905a-0ef3-4e35-9b95-a5c89c0aa02a" class="ms-outlook-mobile-reference-message" data-ogsc="" style="">
<div style="font-family: sans-serif; font-size: 12pt; color:
rgb(0, 0, 0);"><br>
</div>
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="divRplyFwdMsg"><strong>From:</strong> Koenig,
Christian <a class="moz-txt-link-rfc2396E" href="mailto:Christian.Koenig@amd.com"><Christian.Koenig@amd.com></a><br>
<strong>Sent:</strong> Monday, February 24, 2020, 18:06<br>
<strong>To:</strong> Nirmoy Das<br>
<strong>Cc:</strong> <a class="moz-txt-link-abbreviated" href="mailto:amd-gfx@lists.freedesktop.org">amd-gfx@lists.freedesktop.org</a>; Deucher,
Alexander; Liu, Monk; Li, Dennis; Das, Nirmoy<br>
<strong>Subject:</strong> Re: [RFC PATCH 1/1] drm/amdgpu:
wait for sched to become ready on job submit<br>
</div>
<br>
<meta content="text/html; charset=utf-8">
<div dir="auto">
<div>Hi Nirmoy,<br>
<div class="gmail_extra"><br>
<div class="gmail_quote">Am 24.02.2020 17:48 schrieb
Nirmoy Das <a class="moz-txt-link-rfc2396E" href="mailto:nirmoy.aiemd@gmail.com"><nirmoy.aiemd@gmail.com></a>:<br type="attribution">
<blockquote class="quote" style="margin:0 0 0 .8ex;
border-left:1px #ccc solid; padding-left:1ex">
<div><font size="2"><span style="font-size:11pt">
<div>On reset, amdgpu can set a drm sched's
ready status to false temporarily. drm job<br>
init will fail if all of the drm scheds are
not ready for a HW IP. This patch tries to
make<br>
kernel's internal drm job submit handle,
amdgpu_job_submit() a bit more fault
tolerant.<br>
</div>
</span></font></div>
</blockquote>
</div>
</div>
</div>
<div dir="auto"><br>
</div>
<div dir="auto">I don't think that this approach makes
sense. Since it is a front end property we should rather
stop setting the scheduler ready status to false during
reset.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Instead we should only set it to false when
the ring/IB test fails and we can't bring the ring back to
life again.</div>
<div dir="auto"><br>
</div>
<div dir="auto">Christian.</div>
<div dir="auto"><br>
</div>
<div dir="auto">
<div class="gmail_extra">
<div class="gmail_quote">
<blockquote class="quote" style="margin:0 0 0 .8ex;
border-left:1px #ccc solid; padding-left:1ex">
<div><font size="2"><span style="font-size:11pt">
<div><br>
Signed-off-by: Nirmoy Das
<a class="moz-txt-link-rfc2396E" href="mailto:nirmoy.das@amd.com"><nirmoy.das@amd.com></a><br>
---<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
| 35 +++++++++++++++++++--<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
| 5 +--<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
| 6 ++--<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
| 2 +-<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
| 2 +-<br>
drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
| 2 +-<br>
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
| 2 +-<br>
7 files changed, 43 insertions(+), 11
deletions(-)<br>
<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c<br>
index d42be880a236..0745df80112f 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c<br>
@@ -139,7 +139,38 @@ void
amdgpu_job_free(struct amdgpu_job *job)<br>
kfree(job);<br>
}<br>
<br>
-int amdgpu_job_submit(struct amdgpu_job
*job, struct drm_sched_entity *entity,<br>
+static int amdgpu_job_try_init(struct
amdgpu_device *adev,<br>
+ struct
drm_sched_job *base,<br>
+ struct
drm_sched_entity *entity,<br>
+ void *owner)<br>
+{<br>
+ int r, i;<br>
+<br>
+ r = drm_sched_job_init(base, entity,
owner);<br>
+ if (r == -ENOENT) {<br>
+ /* retry till we come out of
reset phase */<br>
+ while
(!mutex_trylock(&adev->lock_reset))<br>
+ msleep(10);<br>
+ /* retry for a second for
the sched to get ready*/<br>
+ for (i = 0; i < 100; i++)
{<br>
+ msleep(10);<br>
+ r =
drm_sched_job_init(base, entity, owner);<br>
+ if (r == -ENOENT)<br>
+ continue;<br>
+ }<br>
+<br>
+
mutex_unlock(&adev->lock_reset);<br>
+ /* If after all these we
failed to initialize a job<br>
+ * it means the IP is
unrecoverable */<br>
+ if (r == -ENOENT)<br>
+ return -ENODEV;<br>
+ }<br>
+<br>
+ return r;<br>
+}<br>
+<br>
+int amdgpu_job_submit(struct amdgpu_device
*adev,struct amdgpu_job *job,<br>
+ struct
drm_sched_entity *entity,<br>
void *owner, struct
dma_fence **f)<br>
{<br>
enum drm_sched_priority priority;<br>
@@ -149,7 +180,7 @@ int
amdgpu_job_submit(struct amdgpu_job *job,
struct drm_sched_entity *entity,<br>
if (!f)<br>
return -EINVAL;<br>
<br>
- r =
drm_sched_job_init(&job->base,
entity, owner);<br>
+ r = amdgpu_job_try_init(adev,
&job->base, entity, owner);<br>
if (r)<br>
return r;<br>
<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h<br>
index 2e2110dddb76..fed87e96cacc 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h<br>
@@ -70,8 +70,9 @@ int
amdgpu_job_alloc_with_ib(struct
amdgpu_device *adev, unsigned size,<br>
<br>
void amdgpu_job_free_resources(struct
amdgpu_job *job);<br>
void amdgpu_job_free(struct amdgpu_job
*job);<br>
-int amdgpu_job_submit(struct amdgpu_job
*job, struct drm_sched_entity *entity,<br>
- void *owner, struct
dma_fence **f);<br>
+int amdgpu_job_submit(struct amdgpu_device
*adev, struct amdgpu_job *job,<br>
+ struct
drm_sched_entity *entity, void *owner,<br>
+ struct dma_fence **f);<br>
int amdgpu_job_submit_direct(struct
amdgpu_job *job, struct amdgpu_ring *ring,<br>
struct
dma_fence **fence);<br>
<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c<br>
index 660867cf2597..adfde07eb75f 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c<br>
@@ -2066,7 +2066,7 @@ static int
amdgpu_map_buffer(struct ttm_buffer_object
*bo,<br>
if (r)<br>
goto error_free;<br>
<br>
- r = amdgpu_job_submit(job,
&adev->mman.entity,<br>
+ r = amdgpu_job_submit(adev, job,
&adev->mman.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, &fence);<br>
if (r)<br>
goto error_free;<br>
@@ -2137,7 +2137,7 @@ int
amdgpu_copy_buffer(struct amdgpu_ring *ring,
uint64_t src_offset,<br>
if (direct_submit)<br>
r =
amdgpu_job_submit_direct(job, ring, fence);<br>
else<br>
- r = amdgpu_job_submit(job,
&adev->mman.entity,<br>
+ r = amdgpu_job_submit(adev,
job, &adev->mman.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, fence);<br>
if (r)<br>
goto error_free;<br>
@@ -2231,7 +2231,7 @@ int
amdgpu_fill_buffer(struct amdgpu_bo *bo,<br>
<br>
amdgpu_ring_pad_ib(ring,
&job->ibs[0]);<br>
WARN_ON(job->ibs[0].length_dw
> num_dw);<br>
- r = amdgpu_job_submit(job,
&adev->mman.entity,<br>
+ r = amdgpu_job_submit(adev, job,
&adev->mman.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, fence);<br>
if (r)<br>
goto error_free;<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c<br>
index 5fd32ad1c575..8ff97b24914e 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c<br>
@@ -1104,7 +1104,7 @@ static int
amdgpu_uvd_send_msg(struct amdgpu_ring
*ring, struct amdgpu_bo *bo,<br>
if (r)<br>
goto err_free;<br>
<br>
- r = amdgpu_job_submit(job,
&adev->uvd.entity,<br>
+ r = amdgpu_job_submit(adev,
job, &adev->uvd.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, &f);<br>
if (r)<br>
goto err_free;<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c<br>
index 59ddba137946..e721d3367783 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c<br>
@@ -554,7 +554,7 @@ static int
amdgpu_vce_get_destroy_msg(struct
amdgpu_ring *ring, uint32_t handle,<br>
if (direct)<br>
r =
amdgpu_job_submit_direct(job, ring, &f);<br>
else<br>
- r = amdgpu_job_submit(job,
&ring->adev->vce.entity,<br>
+ r =
amdgpu_job_submit(ring->adev, job,
&ring->adev->vce.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, &f);<br>
if (r)<br>
goto err;<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c<br>
index 4cc7881f438c..b536962c22d9 100644<br>
---
a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c<br>
+++
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c<br>
@@ -100,7 +100,7 @@ static int
amdgpu_vm_sdma_commit(struct
amdgpu_vm_update_params *p,<br>
WARN_ON(ib->length_dw == 0);<br>
amdgpu_ring_pad_ib(ring, ib);<br>
WARN_ON(ib->length_dw >
p->num_dw_left);<br>
- r = amdgpu_job_submit(p->job,
entity, AMDGPU_FENCE_OWNER_VM, &f);<br>
+ r = amdgpu_job_submit(p->adev,
p->job, entity, AMDGPU_FENCE_OWNER_VM,
&f);<br>
if (r)<br>
goto error;<br>
<br>
diff --git
a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c<br>
index 9775eca6fe43..a4aaa2a1f878 100644<br>
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c<br>
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c<br>
@@ -377,7 +377,7 @@ static void
gmc_v10_0_flush_gpu_tlb(struct amdgpu_device
*adev, uint32_t vmid,<br>
job->vm_needs_flush = true;<br>
job->ibs->ptr[job->ibs->length_dw++]
= ring->funcs->nop;<br>
amdgpu_ring_pad_ib(ring,
&job->ibs[0]);<br>
- r = amdgpu_job_submit(job,
&adev->mman.entity,<br>
+ r = amdgpu_job_submit(adev, job,
&adev->mman.entity,<br>
AMDGPU_FENCE_OWNER_UNDEFINED, &fence);<br>
if (r)<br>
goto error_submit;<br>
-- <br>
2.25.0<br>
<br>
</div>
</span></font></div>
</blockquote>
</div>
<br>
</div>
</div>
</div>
<br>
</div>
</div>
</blockquote>
<br>
</body>
</html>