drm/i915: Use ktime on wait_for
Subsystems (2)
PATCH 1/1 drm/i915: Use ktime on wait_for
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a wait_for loop, we return timeout errorneously much earlier than what the real wallclock would say.
We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
References: https://lkml.org/lkml/2018/4/18/798 [1]
--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_drv.h +2 -2
@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + const ktime_t end__ = ktime_add_ns(ktime_get_raw(), 1000ll * (US)); \ long wait__ = (Wmin);/* recommended min for usleep is 10 us */ \int ret__; \might_sleep (); \for (;;) { \ -bool expired__ =time_after (jiffies, timeout__); \ +const bool expired__ =ktime_after (ktime_get_raw (), end__); \OP ; \if (COND ) { \ ret__ =0 ; \
Signature · 6 lines
2.14.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 24 lines · Mika Kuoppala
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_drv.h +0
-0
drivers/gpu/drm/i915/intel_drv.h line 52
Quoted 5 lines
drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + const ktime_t end__ = ktime_add_ns(ktime_get_raw(), 1000ll * (US)); \
Is ktime_get_raw() monotonic? Thomas suggested ktime_get()
Quoted 9 lines
drivers/gpu/drm/i915/intel_drv.h +1
-1long wait__ = (Wmin);/* recommended min for usleep is 10 us */ \int ret__; \might_sleep (); \for (;;) { \ -bool expired__ =time_after (jiffies, timeout__); \ +const bool expired__ =ktime_after (ktime_get_raw (), end__); \OP ; \if (COND ) { \ ret__ =0 ; \
Signature · 7 lines
Thanks, Sagar _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoting Sagar Arun Kamble (2018-04-20 11:23:50)
Quoted 22 lines · Mika Kuoppala
Quoted 19 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Quoted 11 lines
drivers/gpu/drm/i915/intel_drv.h line 52
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + const ktime_t end__ = ktime_add_ns(ktime_get_raw(), 1000ll * (US)); \Is
ktime_get_raw()monotonic? Thomas suggestedktime_get()
It proclaims to be monotonic, without the clock drift calibration. For the milliseconds we should be sleeping at most, I hope that is immaterial.
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 33 lines · Chris Wilson
Quoting Sagar Arun Kamble (2018-04-20 11:23:50)
Quoted 21 lines · Mika Kuoppala
Quoted 19 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)Quoted 11 lines
drivers/gpu/drm/i915/intel_drv.h line 52
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + const ktime_t end__ = ktime_add_ns(ktime_get_raw(), 1000ll * (US)); \Is
ktime_get_raw()monotonic? Thomas suggestedktime_get()
Quoted 3 lines
It proclaims to be monotonic, without the clock drift calibration. For the milliseconds we should be sleeping at most, I hope that is immaterial.
Yes. I remembered from Imre's comment[1] that raw clock can jump and will not be calibrated. If jumps are are not > jiffies we should be good get_raw then.
[1] https://lists.freedesktop.org/archives/dri-devel/2012-October/028878.html
Quoted 1 line
-Chris
Signature · 7 lines
Thanks, Sagar _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 35 lines · Chris Wilson
Quoting Sagar Arun Kamble (2018-04-20 11:23:50)
Quoted 22 lines · Mika Kuoppala
Quoted 19 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)Quoted 11 lines
drivers/gpu/drm/i915/intel_drv.h line 52
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +1
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + const ktime_t end__ = ktime_add_ns(ktime_get_raw(), 1000ll * (US)); \Is
ktime_get_raw()monotonic? Thomas suggestedktime_get()
Quoted 3 lines
It proclaims to be monotonic, without the clock drift calibration. For the milliseconds we should be sleeping at most, I hope that is immaterial.
Agreed. And even if we would be affected by drift, we would want to extend the end a little instead of accepting a drift calibration while we are waiting.
Here is how the code size is affected:
add/remove: 0/0 grow/shrink: 5/29 up/down: 77/-1099 (-1022)
Function old new delta __intel_wait_for_register_fw 462 496 +34 swsci 576 596 +20 skl_pcode_request 488 507 +19 intel_hdcp_auth 3806 3809 +3 __intel_wait_for_register 436 437 +1 hsw_power_well_disable 548 539 -9 i915_gem_idle_work_handler 608 593 -15 guc_fw_xfer 875 854 -21 bxt_ddi_pll_disable 324 303 -21 chv_set_cdclk 357 334 -23 vlv_wm_get_hw_state 2474 2450 -24 gmbus_wait 564 538 -26 cnl_cdclk_pll_enable 286 259 -27 cnl_cdclk_pll_disable 247 220 -27 hsw_enable_pc8 1699 1671 -28 bdw_set_cdclk 1006 977 -29 wait_for_pipe_scanline_moving 388 358 -30 intel_enable_dsi_pll 1060 1028 -32 i915_gem_wait_for_idle 348 314 -34 chv_set_memory_dvfs 258 223 -35 vlv_set_power_well 321 282 -39 intel_hdmi_hdcp_check_link 485 446 -39 chv_set_pipe_power_well.constprop 359 320 -39 g33_do_reset 216 175 -41 vlv_wait_for_pw_status 210 168 -42 intel_engines_park 348 306 -42 vlv_set_cdclk 711 662 -49 lspcon_wait_mode 265 211 -54 lpt_init_pch_refclk 1542 1488 -54 g4x_do_reset 595 541 -54 bxt_ddi_pll_enable 2407 2353 -54 ironlake_crtc_enable 3279 3211 -68 i915_do_reset 443 375 -68 intel_guc_send_ct 1691 1616 -75
Total: Before=1232834, After=1231812, chg -0.08%
-Mika
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoting Mika Kuoppala (2018-04-20 10:54:26)
Quoted 9 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Quoted 6 lines
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>
The atomic variant is already jiffie-less (since it has to work in irq-off contexts). Maybe a bit tricky to suggest that the callers know if jiffie incremens are accurate or not.
What is not clear from the link is whether our wait_for() is running across suspend, or whether it is just jiffie recalibration some time during resume that breaks.
Quoted 12 lines
drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_drv.h +0
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \
drivers/gpu/drm/i915/intel_drv.h line 60
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +2
-1+const ktime_t end__ =ktime_add_ns (ktime_get_raw (),1000ll * (US )); \long wait__ = (Wmin);/* recommended min for usleep is 10 us */ \int ret__; \might_sleep (); \for (;;) { \ -bool expired__ =time_after (jiffies, timeout__); \ +const bool expired__ =ktime_after (ktime_get_raw (), end__); \OP ; \if (COND ) { \ ret__ =0 ; \
Nevertheless, the patch is ok and I don't have too much objection to adding another tsc (at best, hpet at worst!) read around every mmio+sleep, plus expanding the code for the function calls. Out of curiosity what is the size delta? How many wait_for() do we have left that we need to convert to a function call rather than macro expansion?
Cc stable?
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 17 lines · Chris Wilson
Quoting Mika Kuoppala (2018-04-20 10:54:26)
Quoted 9 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Quoted 6 lines
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>
Quoted 7 lines
The atomic variant is already jiffie-less (since it has to work in irq-off contexts). Maybe a bit tricky to suggest that the callers know if jiffie incremens are accurate or not.
What is not clear from the link is whether our
wait_for()is running across suspend, or whether it is just jiffie recalibration some time during resume that breaks.
The wait_for starts on the resume path, so the jump shouldn't be related to any of the timekeeping adjustments across suspend/resume (happening already during syscore resume). It looks like a delayed LAPIC timer interrupt on that GLK system, trying to get more details on that with irqsoff ftracing.
Quoted 30 lines
Quoted 13 lines
drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_drv.h +0
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \drivers/gpu/drm/i915/intel_drv.h line 60
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +2
-1+const ktime_t end__ =ktime_add_ns (ktime_get_raw (),1000ll * (US )); \long wait__ = (Wmin);/* recommended min for usleep is 10 us */ \int ret__; \might_sleep (); \for (;;) { \ -bool expired__ =time_after (jiffies, timeout__); \ +const bool expired__ =ktime_after (ktime_get_raw (), end__); \OP ; \if (COND ) { \ ret__ =0 ; \Nevertheless, the patch is ok and I don't have too much objection to adding another tsc (at best, hpet at worst!) read around every mmio+sleep, plus expanding the code for the function calls. Out of curiosity what is the size delta? How many
wait_for()do we have left that we need to convert to a function call rather than macro expansion?
Quoted 8 lines
Reviewed-by Chris Wilson <[email protected]>Cc stable?
-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
Quoted 26 lines · Imre Deak
Quoted 17 lines · Chris Wilson
Quoting Mika Kuoppala (2018-04-20 10:54:26)
Quoted 9 lines
We use jiffies to determine when wait expires. However Imre did find out that jiffies can and will do a >1 increments on certain situations [1]. When this happens in a
wait_forloop, we return timeout errorneously much earlier than what the real wallclock would say.We can't afford our waits to timeout prematurely. Discard jiffies and change to ktime to detect timeouts.
Quoted 6 lines
Reported-by Imre Deak <[email protected]>References: https://lkml.org/lkml/2018/4/18/798 [1]
Cc Imre Deak <[email protected]>Cc Chris Wilson <[email protected]>Cc Ville Syrjälä <[email protected]>Signed-off-by Mika Kuoppala <[email protected]>Quoted 7 lines
The atomic variant is already jiffie-less (since it has to work in irq-off contexts). Maybe a bit tricky to suggest that the callers know if jiffie incremens are accurate or not.
What is not clear from the link is whether our
wait_for()is running across suspend, or whether it is just jiffie recalibration some time during resume that breaks.
Quoted 5 lines
The
wait_forstarts on the resume path, so the jump shouldn't be related to any of the timekeeping adjustments across suspend/resume (happening already during syscore resume). It looks like a delayed LAPIC timer interrupt on that GLK system, trying to get more details on that with irqsoff ftracing.
Both patches pushed. Thanks for the report and reviews.
-Mika
Quoted 31 lines
Quoted 31 lines
Quoted 14 lines
drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/intel_drv.h +0
-1@@ -49,12 +49,12 @@* check the condition before the timeout. */ #define __wait_for (OP ,COND ,US , Wmin, Wmax) ({ \ - unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \drivers/gpu/drm/i915/intel_drv.h line 60
Quoted 10 lines
drivers/gpu/drm/i915/intel_drv.h +2
-1+const ktime_t end__ =ktime_add_ns (ktime_get_raw (),1000ll * (US )); \long wait__ = (Wmin);/* recommended min for usleep is 10 us */ \int ret__; \might_sleep (); \for (;;) { \ -bool expired__ =time_after (jiffies, timeout__); \ +const bool expired__ =ktime_after (ktime_get_raw (), end__); \OP ; \if (COND ) { \ ret__ =0 ; \Nevertheless, the patch is ok and I don't have too much objection to adding another tsc (at best, hpet at worst!) read around every mmio+sleep, plus expanding the code for the function calls. Out of curiosity what is the size delta? How many
wait_for()do we have left that we need to convert to a function call rather than macro expansion?
Quoted 8 lines
Quoted 8 lines
Reviewed-by Chris Wilson <[email protected]>Cc stable?
-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
Patchwork
CI✓ Fi.CI.BAT: success for drm/i915: Use ktime on wait_for
In reply to Mika Kuoppala== Series Details ==
== Series Details ==
Series: drm/i915: Use ktime on wait_for
URL : https://patchwork.freedesktop.org/series/42023/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4072 -> Patchwork_8762 =
== Summary - SUCCESS ==
No regressions found.
External URL: https://patchwork.freedesktop.org/api/1.0/series/42023/revisions/1/mbox/
== Known issues ==
Here are the changes found in Patchwork_8762 that come from known issues:
=== IGT changes ===
==== Issues hit ====
igt@gem_mmap_gtt@basic-small-bo-tiledx:
fi-gdg-551: PASS -> FAIL (fdo#102575)fdo#102575 https://bugs.freedesktop.org/show_bug.cgi?id=102575
== Participating hosts (33 -> 31) ==
Missing (2): fi-ilk-m540 fi-skl-6700hq
== Build changes ==
* Linux: CI_DRM_4072 -> Patchwork_8762
CI_DRM_4072: b35e59e5c6a9cae11d5183d2bf9c5c99ceedbc7c @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4442: 8168bb65d5e64d4df4e5d847d448bab2d2825d73 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_8762: 62a2b14393160309a7e197bd6b3131a56edd01ab @ git://anongit.freedesktop.org/gfx-ci/linux
piglit_4442: e60d247eb359f044caf0c09904da14e39d7adca1 @ git://anongit.freedesktop.org/piglit
== Linux commits ==
62a2b1439316 drm/i915: Use ktime on wait_for
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_8762/issues.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Patchwork
CI✓ Fi.CI.IGT: success for drm/i915: Use ktime on wait_for
In reply to Mika Kuoppala== Series Details ==
== Series Details ==
Series: drm/i915: Use ktime on wait_for
URL : https://patchwork.freedesktop.org/series/42023/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4072_full -> Patchwork_8762_full =
== Summary - WARNING ==
Minor unknown changes coming with Patchwork_8762_full need to be verified manually.
If you think the reported changes have nothing to do with the changes introduced in Patchwork_8762_full, please notify your bug team to allow them to document this new failure mode, which will reduce false positives in CI.
External URL: https://patchwork.freedesktop.org/api/1.0/series/42023/revisions/1/mbox/
== Possible new issues ==
Here are the unknown changes that may have been introduced in Patchwork_8762_full:
=== IGT changes ===
==== Warnings ====
igt@gem_exec_schedule@deep-blt:
shard-kbl: PASS -> SKIP
igt@gem_exec_schedule@deep-bsd1:
shard-kbl: SKIP -> PASS +1
igt@kms_pipe_crc_basic@nonblocking-crc-pipe-c-frame-sequence:
shard-glk: PASS -> SKIP +88
igt@kms_universal_plane@universal-plane-pipe-b-sanity:
shard-glk: SKIP -> PASS +102== Known issues ==
Here are the changes found in Patchwork_8762_full that come from known issues:
=== IGT changes ===
==== Issues hit ====
igt@gem_ppgtt@blt-vs-render-ctxn:
shard-kbl: PASS -> INCOMPLETE (fdo#106023, fdo#103665)
==== Possible fixes ====
igt@kms_flip@dpms-vs-vblank-race-interruptible:
shard-glk: FAIL (fdo#103060) -> PASS
igt@kms_flip@plain-flip-fb-recreate:
shard-hsw: FAIL (fdo#100368) -> PASS
shard-glk: FAIL (fdo#100368) -> SKIP
igt@perf@blocking:
shard-hsw: FAIL (fdo#102252) -> PASS fdo#100368 https://bugs.freedesktop.org/show_bug.cgi?id=100368
fdo#102252 https://bugs.freedesktop.org/show_bug.cgi?id=102252
fdo#103060 https://bugs.freedesktop.org/show_bug.cgi?id=103060
fdo#103665 https://bugs.freedesktop.org/show_bug.cgi?id=103665
fdo#106023 https://bugs.freedesktop.org/show_bug.cgi?id=106023
== Participating hosts (6 -> 5) ==
Missing (1): shard-glkb
== Build changes ==
* Linux: CI_DRM_4072 -> Patchwork_8762
CI_DRM_4072: b35e59e5c6a9cae11d5183d2bf9c5c99ceedbc7c @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4442: 8168bb65d5e64d4df4e5d847d448bab2d2825d73 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_8762: 62a2b14393160309a7e197bd6b3131a56edd01ab @ git://anongit.freedesktop.org/gfx-ci/linux
piglit_4442: e60d247eb359f044caf0c09904da14e39d7adca1 @ git://anongit.freedesktop.org/piglit
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_8762/shards.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx