Skip to contentSkip to conversation

drm/i915/selftests: add some vma_sync

PATCH In review Matthew Auld 4 participants 5 messages Mar 23, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/i915/selftests: add some vma_sync

The subtest shrink_boom was added as a regression test for some missing refcounting on the paging structures, however since the binding is potentially async, setting the vm->fault_attr might apply to the purge vma, and not the intended explode vma. Also it looks like it might also be possible to hit some weird shrinker deadlock where the unbinding of one vma allocates memory by flushing and waiting for its still-pending-bind operation while holding vm->mutex, which will always lands back in the shrinker since we set vm->fault_attr for the selftest.

References: https://gitlab.freedesktop.org/drm/intel/issues/1493

---
 drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +4 -0
index b342bef5e7c9..029406a2a0b3 100644
@@ -951,6 +951,8 @@ static int shrink_boom(struct i915_address_space *vm,
 		if (err)
 			goto err_purge;
 
+		i915_vma_sync(vma);
+
 		/* Should now be ripe for purging */
 		i915_vma_unpin(vma);
 
@@ -974,6 +976,8 @@ static int shrink_boom(struct i915_address_space *vm,
 		if (err)
 			goto err_explode;
 
+		i915_vma_sync(vma);
+
 		i915_vma_unpin(vma);
 
 		i915_gem_object_put(purge);
Signature · 6 lines
2.20.1

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

Quoting Matthew Auld (2020-03-23 11:38:09)

Quoted 4 lines

The subtest shrink_boom was added as a regression test for some missing refcounting on the paging structures, however since the binding is potentially async, setting the vm->fault_attr might apply to the purge vma, and not the intended explode vma.

Hmm. Sounds a fair point, though let's see if that is not an unintended bonus.

Quoted 5 lines

Also it looks like it might also
be possible to hit some weird shrinker deadlock where the unbinding of one vma allocates memory by flushing and waiting for its still-pending-bind operation while holding vm->mutex, which will always lands back in the shrinker since we set vm->fault_attr for the selftest.

However that is a bug we have to handle. And it should be prevented currently by avoiding shrinking active (still being bound) vma, e.g. 6f24e41022f2 ("drm/i915: Avoid recursing onto active vma from the shrinker"). So is that a current observation?
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#2 lore raw
Quoted 16 lines · Chris Wilson

Quoting Matthew Auld (2020-03-23 11:38:09)

Quoted 4 lines

The subtest shrink_boom was added as a regression test for some missing refcounting on the paging structures, however since the binding is potentially async, setting the vm->fault_attr might apply to the purge vma, and not the intended explode vma.

Hmm. Sounds a fair point, though let's see if that is not an unintended bonus.

Quoted 5 lines

Also it looks like it might also
be possible to hit some weird shrinker deadlock where the unbinding of one vma allocates memory by flushing and waiting for its still-pending-bind operation while holding vm->mutex, which will always lands back in the shrinker since we set vm->fault_attr for the selftest.

Quoted 4 lines

However that is a bug we have to handle. And it should be prevented currently by avoiding shrinking active (still being bound) vma, e.g. 6f24e41022f2 ("drm/i915: Avoid recursing onto active vma from the shrinker"). So is that a current observation?

Missed that. Egg-on-face.

Quoted 5 lines

-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw
In reply to Matthew Auld

Quoting Matthew Auld (2020-03-23 11:38:09)

Quoted 10 lines

The subtest shrink_boom was added as a regression test for some missing refcounting on the paging structures, however since the binding is potentially async, setting the vm->fault_attr might apply to the purge vma, and not the intended explode vma. Also it looks like it might also be possible to hit some weird shrinker deadlock where the unbinding of one vma allocates memory by flushing and waiting for its still-pending-bind operation while holding vm->mutex, which will always lands back in the shrinker since we set vm->fault_attr for the selftest.

References: https://gitlab.freedesktop.org/drm/intel/issues/1493

Ah, you picked the wrong deadlock -- that's the gtt eviction recursion. :(
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#4 lore raw

[Intel-gfx] ✗ Fi.CI.BAT: failure for drm/i915/selftests: add some vma_sync

In reply to Matthew Auld
== Series Details ==

== Series Details ==

Series: drm/i915/selftests: add some vma_sync
URL : https://patchwork.freedesktop.org/series/74969/
State : failure

== Summary ==

CI Bug Log - changes from CI_DRM_8178 -> Patchwork_17052
====================================================

Summary
-------

**FAILURE**

Serious unknown changes coming with Patchwork_17052 absolutely need to be verified manually.

If you think the reported changes have nothing to do with the changes introduced in Patchwork_17052, please notify your bug team to allow them to document this new failure mode, which will reduce false positives in CI.

External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/index.html

Possible new issues
-------------------

Here are the unknown changes that may have been introduced in Patchwork_17052:

### IGT changes ###

#### Possible regressions ####

* igt@i915_selftest@live@coherency:
- fi-gdg-551: [PASS][1] -> [DMESG-FAIL][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8178/fi-gdg-551/igt@i915_selftest@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/fi-gdg-551/igt@i915_selftest@[email protected]

Known issues
------------

Here are the changes found in Patchwork_17052 that come from known issues:

### IGT changes ###

#### Issues hit ####

* igt@kms_chamelium@common-hpd-after-suspend:
- fi-icl-u2: [PASS][3] -> [FAIL][4] ([i915#217])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8178/fi-icl-u2/igt@[email protected]
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/fi-icl-u2/igt@[email protected]

#### Possible fixes ####

* igt@i915_selftest@live@gem_contexts:
- fi-cml-s: [DMESG-FAIL][5] ([i915#877]) -> [PASS][6]
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8178/fi-cml-s/igt@i915_selftest@live@gem_contexts.html
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/fi-cml-s/igt@i915_selftest@live@gem_contexts.html

* igt@kms_chamelium@hdmi-hpd-fast:
- fi-kbl-7500u: [FAIL][7] ([i915#323]) -> [PASS][8]
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8178/fi-kbl-7500u/igt@[email protected]
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/fi-kbl-7500u/igt@[email protected]

[i915#217]: https://gitlab.freedesktop.org/drm/intel/issues/217
[i915#323]: https://gitlab.freedesktop.org/drm/intel/issues/323
[i915#877]: https://gitlab.freedesktop.org/drm/intel/issues/877

Participating hosts (50 -> 42)
------------------------------

Additional (1): fi-byt-n2820
Missing (9): fi-kbl-soraka fi-ilk-m540 fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-ctg-p8600 fi-bsw-kefka fi-byt-clapper fi-bdw-samus

Build changes
-------------

* CI: CI-20190529 -> None
* Linux: CI_DRM_8178 -> Patchwork_17052

CI-20190529: 20190529
CI_DRM_8178: 3f4392bf31d28013e860818ba613b598bb227821 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5528: 5dee9128b2aaa77d036163f670f0e0fc15b578ab @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_17052: 8bf02a45845a6710387fc0cab20cb4c79744c8ce @ git://anongit.freedesktop.org/gfx-ci/linux

== Linux commits ==

8bf02a45845a drm/i915/selftests: add some vma_sync

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17052/index.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx