Skip to contentSkip to conversation

drm/i915/selftests: Add a GuC doorbells selftest

PATCH Superseded Michel Thierry 5 participants 10 messages Nov 15, 2017 – Nov 16, 2017

Subsystems (2)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/3 drm/i915/selftests: Add a GuC doorbells selftest

The first test aims to check guc_init_doorbell_hw, changing the existing guc clients and doorbells state before calling it.

The second test tries to create as many clients as it is currently possible (currently limited to max number of doorbells) and exercise the doorbell alloc/dealloc code.

Since our usage mode require very few clients/doorbells, this code has been exercised very lightly and it's good to have a simple test for it.

As reference, this test already helped identify the bug fixed by commit 7f1ea2ac3017 ("drm/i915/guc: Fix doorbell id selection").

v2: Extend number of clients; check for client allocation failure when number of doorbells is exceeded; validate client properties; reuse guc_init_doorbell_hw (Chris).

v3: guc_init_doorbell_hw test added per Chris suggestion.

v4: Try to explain why guc_init_doorbell_hw exist and comment some details in the subtest.

Cc Michal Wajdeczko <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>
---
 drivers/gpu/drm/i915/i915_guc_submission.c         |   4 +
 .../gpu/drm/i915/selftests/i915_live_selftests.h   |   1 +
 drivers/gpu/drm/i915/selftests/intel_guc.c         | 362 +++++++++++++++++++++
 3 files changed, 367 insertions(+)
 create mode 100644 drivers/gpu/drm/i915/selftests/intel_guc.c
drivers/gpu/drm/i915/i915_guc_submission.c +4 -0
index 0ba2fc04fe9c..5d6576e01a91 100644
@@ -1464,3 +1464,7 @@ void i915_guc_submission_disable(struct drm_i915_private *dev_priv)
 
 	guc_clients_destroy(guc);
 }
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/intel_guc.c"
+#endif
drivers/gpu/drm/i915/selftests/i915_live_selftests.h +1 -0
index d7dd98a6acad..088f45bc6199 100644
@@ -20,3 +20,4 @@ selftest(evict, i915_gem_evict_live_selftests)
 selftest(hugepages, i915_gem_huge_page_live_selftests)
 selftest(contexts, i915_gem_context_live_selftests)
 selftest(hangcheck, intel_hangcheck_live_selftests)
+selftest(guc, intel_guc_live_selftest)
drivers/gpu/drm/i915/selftests/intel_guc.c new file +362 -0

362 lines not on this page · raw message

Signature · 6 lines
2.15.0

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

RFC 2/3 drm/i915/guc: Omit guc_init_doorbell_hw during driver load

During driver load we create the GuC clients and allocate their doorbells just before executing guc_init_doorbell_hw; but since we just created these doorbells, how can they be out of sync?
This code has had more than enough refactoring (2 more still in progress) so I would not be surprised if calling guc_init_doorbell_hw made sense at some point, but not anymore.

The resume path is different, in this case the driver doesn't recreate clients, and it is still reasonable to validate/reallocate the doorbells in order to confirm that they still belong to the clients.

And probably guc_init_doorbell_hw is no longer the right name, but I'll leave that to someone else.

Cc Sagar Arun Kamble <[email protected]>
Cc Sujaritha Sundaresan <[email protected]>
Cc Michal Wajdeczko <[email protected]>
Cc Oscar Mateo <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>
---
 drivers/gpu/drm/i915/i915_guc_submission.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_guc_submission.c +4 -4
index 5d6576e01a91..d6762ca42cf1 100644
@@ -1424,16 +1424,16 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv)
 	} else {
 		guc_reset_wq(guc->execbuf_client);
 		guc_reset_wq(guc->preempt_client);
+
+		err = guc_init_doorbell_hw(guc);
+		if (err)
+			goto err_free_clients;
 	}
 
 	err = intel_guc_sample_forcewake(guc);
 	if (err)
 		goto err_free_clients;
 
-	err = guc_init_doorbell_hw(guc);
-	if (err)
-		goto err_free_clients;
-
 	/* Take over from manual control of ELSP (execlists) */
 	guc_interrupts_capture(dev_priv);
 
Signature · 6 lines
2.15.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Acked-by on v1 2/3
#2 lore raw

Re: [RFC 2/3] drm/i915/guc: Omit guc_init_doorbell_hw during driver load

Quoted 7 lines · Michel Thierry

During driver load we create the GuC clients and allocate their doorbells just before executing guc_init_doorbell_hw; but since we just created these doorbells, how can they be out of sync?
This code has had more than enough refactoring (2 more still in progress) so I would not be surprised if calling guc_init_doorbell_hw made sense at some point, but not anymore.

I think the idea was to clean up the unallocated doorbells on takeover to be covered in case the previous occupant of the GPU didn't release them when leaving the HW. We do a full gpu reset during i915 load now in i915_gem_sanitize so the doorbell HW should be cleaned up by that, but there is still a possible issue when i915.reset=0. However, with reset=0 this wouldn't be the only thing not sanitized and the only bad consequence would be extra irqs to GuC (which would be ignored), so I don't think it is worth worrying about that case.

Daniele

Quoted 34 lines

The resume path is different, in this case the driver doesn't recreate clients, and it is still reasonable to validate/reallocate the doorbells in order to confirm that they still belong to the clients.

And probably guc_init_doorbell_hw is no longer the right name, but I'll leave that to someone else.

Cc Sagar Arun Kamble <[email protected]>
Cc Sujaritha Sundaresan <[email protected]>
Cc Michal Wajdeczko <[email protected]>
Cc Oscar Mateo <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>
---
  drivers/gpu/drm/i915/i915_guc_submission.c | 8 ++++----
  1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_guc_submission.c +4 -0
index 5d6576e01a91..d6762ca42cf1 100644
@@ -1424,16 +1424,16 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv)
  	} else {
  		guc_reset_wq(guc->execbuf_client);
  		guc_reset_wq(guc->preempt_client);
+
+		err = guc_init_doorbell_hw(guc);
+		if (err)
+			goto err_free_clients;
  	}
 
  	err = intel_guc_sample_forcewake(guc);

drivers/gpu/drm/i915/i915_guc_submission.c int i915_guc_submission_enable(struct drm_i915_private *dev_priv) · line 1439

Quoted 10 lines
drivers/gpu/drm/i915/i915_guc_submission.c +0 -4
  	if (err)
  		goto err_free_clients;
 
-	err = guc_init_doorbell_hw(guc);
-	if (err)
-		goto err_free_clients;
-
  	/* Take over from manual control of ELSP (execlists) */
  	guc_interrupts_capture(dev_priv);
 

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

Acked-by on v1 2/3
#3 lore raw

Re: [RFC 2/3] drm/i915/guc: Omit guc_init_doorbell_hw during driver load

In reply to Michel Thierry
Quoted 10 lines · Michel Thierry

During driver load we create the GuC clients and allocate their doorbells just before executing guc_init_doorbell_hw; but since we just created these doorbells, how can they be out of sync?
This code has had more than enough refactoring (2 more still in progress) so I would not be surprised if calling guc_init_doorbell_hw made sense at some point, but not anymore.

The resume path is different, in this case the driver doesn't recreate clients, and it is still reasonable to validate/reallocate the doorbells in order to confirm that they still belong to the clients.

Planning to change this in upcoming series (allocate doorbells on resume when not needing uc_init_hw) and then we can do away with this validation. Another problem I see is, this is time consuming and leads to increase in the resume time (we also sanitize on resume hence this is unnecessary for all unused doorbells)

Quoted 4 lines

And probably guc_init_doorbell_hw is no longer the right name, but I'll leave that to someone else.

Change looks good to me.

Quoted 25 lines
Cc Sagar Arun Kamble <[email protected]>
Cc Sujaritha Sundaresan <[email protected]>
Cc Michal Wajdeczko <[email protected]>
Cc Oscar Mateo <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>
---
  drivers/gpu/drm/i915/i915_guc_submission.c | 8 ++++----
  1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/i915_guc_submission.c +4 -0
index 5d6576e01a91..d6762ca42cf1 100644
@@ -1424,16 +1424,16 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv)
  	} else {
  		guc_reset_wq(guc->execbuf_client);
  		guc_reset_wq(guc->preempt_client);
+
+		err = guc_init_doorbell_hw(guc);
+		if (err)
+			goto err_free_clients;
  	}
 
  	err = intel_guc_sample_forcewake(guc);

drivers/gpu/drm/i915/i915_guc_submission.c int i915_guc_submission_enable(struct drm_i915_private *dev_priv) · line 1439

Quoted 10 lines
drivers/gpu/drm/i915/i915_guc_submission.c +0 -4
  	if (err)
  		goto err_free_clients;
 
-	err = guc_init_doorbell_hw(guc);
-	if (err)
-		goto err_free_clients;
-
  	/* Take over from manual control of ELSP (execlists) */
  	guc_interrupts_capture(dev_priv);
 

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

#4 lore raw

PATCH 3/3 HAX enable GuC submission for CI

In reply to Michel Thierry

From: Michal Wajdeczko <[email protected]>

Also revert ("drm/i915/guc: Assert that we switch between known ggtt->invalidate functions")

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 8 ++------
 drivers/gpu/drm/i915/i915_params.h  | 4 ++--
 2 files changed, 4 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -6
index 3c3a699436c9..b56e785d2c42 100644
@@ -3590,17 +3590,13 @@ int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv)
 
 void i915_ggtt_enable_guc(struct drm_i915_private *i915)
 {
-	GEM_BUG_ON(i915->ggtt.invalidate != gen6_ggtt_invalidate);
-
 	i915->ggtt.invalidate = guc_ggtt_invalidate;
 }
 
 void i915_ggtt_disable_guc(struct drm_i915_private *i915)
 {
-	/* We should only be called after i915_ggtt_enable_guc() */
-	GEM_BUG_ON(i915->ggtt.invalidate != guc_ggtt_invalidate);
-
-	i915->ggtt.invalidate = gen6_ggtt_invalidate;
+	if (i915->ggtt.invalidate == guc_ggtt_invalidate)
+		i915->ggtt.invalidate = gen6_ggtt_invalidate;
 }
 
 void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv)
drivers/gpu/drm/i915/i915_params.h +2 -2
index c7292268ed43..c38cef07b9fe 100644
@@ -44,8 +44,8 @@
 	param(int, disable_power_well, -1) \
 	param(int, enable_ips, 1) \
 	param(int, invert_brightness, 0) \
-	param(int, enable_guc_loading, 0) \
-	param(int, enable_guc_submission, 0) \
+	param(int, enable_guc_loading, 1) \
+	param(int, enable_guc_submission, 1) \
 	param(int, guc_log_level, -1) \
 	param(char *, guc_firmware_path, NULL) \
 	param(char *, huc_firmware_path, NULL) \
Signature · 6 lines
2.15.0

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

✓ Fi.CI.BAT: success for series starting with [v4,1/3] drm/i915/selftests: Add a GuC doorbells selftest

In reply to Michel Thierry
== Series Details ==

== Series Details ==

Series: series starting with [v4,1/3] drm/i915/selftests: Add a GuC doorbells selftest
URL : https://patchwork.freedesktop.org/series/33896/
State : success

== Summary ==

Series 33896v1 series starting with [v4,1/3] drm/i915/selftests: Add a GuC doorbells selftest
https://patchwork.freedesktop.org/api/1.0/series/33896/revisions/1/mbox/

Test gem_exec_reloc:

        Subgroup basic-gtt-active:
                fail       -> PASS       (fi-gdg-551) fdo#102582 +1
Test gem_ringfill:
        Subgroup basic-default-hang:
                pass       -> DMESG-WARN (fi-pnv-d510) fdo#101600

fdo#102582 https://bugs.freedesktop.org/show_bug.cgi?id=102582
fdo#101600 https://bugs.freedesktop.org/show_bug.cgi?id=101600

fi-bdw-5557u     total:289  pass:268  dwarn:0   dfail:0   fail:0   skip:21  time:443s
fi-bdw-gvtdvm    total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:462s
fi-blb-e6850     total:289  pass:223  dwarn:1   dfail:0   fail:0   skip:65  time:383s
fi-bsw-n3050     total:289  pass:243  dwarn:0   dfail:0   fail:0   skip:46  time:540s
fi-bwr-2160      total:289  pass:183  dwarn:0   dfail:0   fail:0   skip:106 time:279s
fi-bxt-dsi       total:289  pass:259  dwarn:0   dfail:0   fail:0   skip:30  time:511s
fi-bxt-j4205     total:289  pass:260  dwarn:0   dfail:0   fail:0   skip:29  time:512s
fi-byt-j1900     total:289  pass:254  dwarn:0   dfail:0   fail:0   skip:35  time:497s
fi-byt-n2820     total:289  pass:250  dwarn:0   dfail:0   fail:0   skip:39  time:496s
fi-elk-e7500     total:289  pass:229  dwarn:0   dfail:0   fail:0   skip:60  time:434s
fi-gdg-551       total:289  pass:178  dwarn:1   dfail:0   fail:1   skip:109 time:264s
fi-glk-1         total:289  pass:261  dwarn:0   dfail:0   fail:0   skip:28  time:539s
fi-hsw-4770      total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:428s
fi-hsw-4770r     total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:443s
fi-ilk-650       total:289  pass:228  dwarn:0   dfail:0   fail:0   skip:61  time:426s
fi-ivb-3520m     total:289  pass:260  dwarn:0   dfail:0   fail:0   skip:29  time:488s
fi-ivb-3770      total:289  pass:260  dwarn:0   dfail:0   fail:0   skip:29  time:466s
fi-kbl-7500u     total:289  pass:264  dwarn:1   dfail:0   fail:0   skip:24  time:483s
fi-kbl-7560u     total:289  pass:270  dwarn:0   dfail:0   fail:0   skip:19  time:529s
fi-kbl-7567u     total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:474s
fi-kbl-r         total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:535s
fi-pnv-d510      total:289  pass:222  dwarn:1   dfail:0   fail:0   skip:66  time:574s
fi-skl-6260u     total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:454s
fi-skl-6600u     total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:545s
fi-skl-6700hq    total:289  pass:263  dwarn:0   dfail:0   fail:0   skip:26  time:571s
fi-skl-6700k     total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:518s
fi-skl-6770hq    total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:504s
fi-skl-gvtdvm    total:289  pass:266  dwarn:0   dfail:0   fail:0   skip:23  time:458s
fi-snb-2520m     total:289  pass:250  dwarn:0   dfail:0   fail:0   skip:39  time:560s
fi-snb-2600      total:289  pass:249  dwarn:0   dfail:0   fail:0   skip:40  time:418s

Blacklisted hosts:
fi-cfl-s total:289 pass:254 dwarn:3 dfail:0 fail:0 skip:32 time:528s
fi-cnl-y failed to connect after reboot

bd5e410d0f036c80ae80e075592ee6dd90c660a8 drm-tip: 2017y-11m-15d-17h-22m-19s UTC integration manifest 9cf074373888 HAX enable GuC submission for CI
8edc9934b18d drm/i915/guc: Omit guc_init_doorbell_hw during driver load
b21f9aa34707 drm/i915/selftests: Add a GuC doorbells selftest

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_7149/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#6 lore raw

✗ Fi.CI.IGT: warning for series starting with [v4,1/3] drm/i915/selftests: Add a GuC doorbells selftest

In reply to Michel Thierry
== Series Details ==

== Series Details ==

Series: series starting with [v4,1/3] drm/i915/selftests: Add a GuC doorbells selftest
URL : https://patchwork.freedesktop.org/series/33896/
State : warning

== Summary ==

Test kms_frontbuffer_tracking:

        Subgroup fbc-1p-primscrn-pri-indfb-draw-pwrite:
                pass       -> SKIP       (shard-hsw)
Test drv_selftest:
        Subgroup mock_sanitycheck:
                dmesg-warn -> PASS       (shard-hsw) fdo#103719

fdo#103719 https://bugs.freedesktop.org/show_bug.cgi?id=103719

shard-hsw total:2585 pass:1471 dwarn:4 dfail:1 fail:9 skip:1100 time:9508s
Blacklisted hosts:

shard-apl        total:2585 pass:1617 dwarn:5   dfail:2   fail:23  skip:938 time:13553s
shard-kbl        total:2472 pass:1623 dwarn:7   dfail:2   fail:25  skip:812 time:10105s
shard-snb        total:2585 pass:1260 dwarn:2   dfail:1   fail:11  skip:1311 time:8083s

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_7149/shards.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reviewed-by on v4 1/3
#7 lore raw
In reply to Michel Thierry

Quoting Michel Thierry (2017-11-15 18:30:27)

Quoted 19 lines

The first test aims to check guc_init_doorbell_hw, changing the existing guc clients and doorbells state before calling it.

The second test tries to create as many clients as it is currently possible (currently limited to max number of doorbells) and exercise the doorbell alloc/dealloc code.

Since our usage mode require very few clients/doorbells, this code has been exercised very lightly and it's good to have a simple test for it.

As reference, this test already helped identify the bug fixed by commit 7f1ea2ac3017 ("drm/i915/guc: Fix doorbell id selection").

v2: Extend number of clients; check for client allocation failure when number of doorbells is exceeded; validate client properties; reuse guc_init_doorbell_hw (Chris).

v3: guc_init_doorbell_hw test added per Chris suggestion.

Quoted 7 lines

v4: Try to explain why guc_init_doorbell_hw exist and comment some details in the subtest.

Cc Michal Wajdeczko <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>

I think this is a good start. The biggest wishlist item I have for this is to mock it and so always run it (as well as then checking the hw interaction in live selftests).

drivers/gpu/drm/i915/selftests/intel_guc.c line 104

Quoted 8 lines
drivers/gpu/drm/i915/selftests/intel_guc.c new file +8 -0
+static int igt_guc_init_doorbell_hw(void *args)
+{
+       struct drm_i915_private *dev_priv = args;
+       struct intel_guc *guc;
+       DECLARE_BITMAP(db_bitmap_bk, GUC_NUM_DOORBELLS);
+       int i, err = 0;
+
+       pr_info("GuC init_doorbell_hw selftest\n");

Just prior to this there is a pr_debug("igt_guc_init_doorbell_hw\n"); Hence why I feel these entry pr_info are redundant; enable debugging ;) -Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#8 lore raw
In reply to Michel Thierry

Quoting Michel Thierry (2017-11-15 18:30:27)

Quoted 19 lines

The first test aims to check guc_init_doorbell_hw, changing the existing guc clients and doorbells state before calling it.

The second test tries to create as many clients as it is currently possible (currently limited to max number of doorbells) and exercise the doorbell alloc/dealloc code.

Since our usage mode require very few clients/doorbells, this code has been exercised very lightly and it's good to have a simple test for it.

As reference, this test already helped identify the bug fixed by commit 7f1ea2ac3017 ("drm/i915/guc: Fix doorbell id selection").

v2: Extend number of clients; check for client allocation failure when number of doorbells is exceeded; validate client properties; reuse guc_init_doorbell_hw (Chris).

v3: guc_init_doorbell_hw test added per Chris suggestion.

Quoted 7 lines

v4: Try to explain why guc_init_doorbell_hw exist and comment some details in the subtest.

Cc Michal Wajdeczko <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>

Rebase, resend with r-bs and lets land this monster. -Chris
_______________________________________________
Intel-gfx mailing list [email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#9 lore raw
Quoted 23 lines · Chris Wilson

Quoting Michel Thierry (2017-11-15 18:30:27)

Quoted 22 lines

The first test aims to check guc_init_doorbell_hw, changing the existing guc clients and doorbells state before calling it.

The second test tries to create as many clients as it is currently possible (currently limited to max number of doorbells) and exercise the doorbell alloc/dealloc code.

Since our usage mode require very few clients/doorbells, this code has been exercised very lightly and it's good to have a simple test for it.

As reference, this test already helped identify the bug fixed by commit 7f1ea2ac3017 ("drm/i915/guc: Fix doorbell id selection").

v2: Extend number of clients; check for client allocation failure when number of doorbells is exceeded; validate client properties; reuse guc_init_doorbell_hw (Chris).

v3: guc_init_doorbell_hw test added per Chris suggestion.

v4: Try to explain why guc_init_doorbell_hw exist and comment some details in the subtest.

Quoted 6 lines
Quoted 4 lines
Cc Michal Wajdeczko <[email protected]>
Cc Daniele Ceraolo Spurio <[email protected]>
Cc Chris Wilson <[email protected]>

Rebase, resend with r-bs and lets land this monster.

Thanks, and I will think about ways to mock this.
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx