[PATCH] drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[PATCH] drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Lionel Landwerlin]

On Gen8+ this register is not priviledged and we want to use it in
Mesa to implement a feature required by GPA called Null Hardware. The
idea is to have the command parser turn 3DPRIMITIVE/GPGPU_WALKER into
NOOPs.

This patch just whitelists the bits that we need and that are
currently not used by the kernel.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
---
 drivers/gpu/drm/i915/i915_cmd_parser.c | 8 ++++++++
 drivers/gpu/drm/i915/i915_reg.h        | 3 +++
 2 files changed, 11 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 95478db9998b..1db6447460eb 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -534,6 +534,14 @@ struct drm_i915_reg_descriptor {
 	{ .addr = _reg ## _UDW(idx) }
 
 static const struct drm_i915_reg_descriptor gen7_render_regs[] = {
+	REG32(INSTPM,
+	      .mask = ~((INSTPM_3D_STATE_INSTRUCTION_DISABLE |
+			 INSTPM_3D_RENDERING_INSTRUCTION_DISABLE |
+			 INSTPM_3D_MEDIA_INSTRUCTION_DISABLE) << 16 |
+			(INSTPM_3D_STATE_INSTRUCTION_DISABLE |
+			 INSTPM_3D_RENDERING_INSTRUCTION_DISABLE |
+			 INSTPM_3D_MEDIA_INSTRUCTION_DISABLE)),
+	      .value = 0),
 	REG64(GPGPU_THREADS_DISPATCHED),
 	REG64(HS_INVOCATION_COUNT),
 	REG64(DS_INVOCATION_COUNT),
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 196a0eb79272..2db9b6a177d9 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -2531,6 +2531,9 @@ enum i915_power_well_id {
 #define   INSTPM_FORCE_ORDERING				(1<<7) /* GEN6+ */
 #define   INSTPM_TLB_INVALIDATE	(1<<9)
 #define   INSTPM_SYNC_FLUSH	(1<<5)
+#define   INSTPM_3D_MEDIA_INSTRUCTION_DISABLE     (1<<3) /* GEN6+ */
+#define   INSTPM_3D_RENDERING_INSTRUCTION_DISABLE (1<<2) /* GEN6+ */
+#define   INSTPM_3D_STATE_INSTRUCTION_DISABLE     (1<<1) /* GEN6+ */
 #define ACTHD	        _MMIO(0x20c8)
 #define MEM_MODE	_MMIO(0x20cc)
 #define   MEM_DISPLAY_B_TRICKLE_FEED_DISABLE (1<<3) /* 830 only */
-- 
2.17.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Lionel Landwerlin]

On 18/05/18 15:26, Lionel Landwerlin wrote:
> On Gen8+ this register is not priviledged and we want to use it in
> Mesa to implement a feature required by GPA called Null Hardware. The
> idea is to have the command parser turn 3DPRIMITIVE/GPGPU_WALKER into
> NOOPs.
>
> This patch just whitelists the bits that we need and that are
> currently not used by the kernel.
>

One thing I don't really know is whether is should be considered an 
issue with the current command parser and therefore be backported as a fix.
It would certainly make things easier because we can't really detect the 
behavior from userspace.

-
Lionel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Chris Wilson]

Quoting Lionel Landwerlin (2018-05-18 15:29:21)
> On 18/05/18 15:26, Lionel Landwerlin wrote:
> > On Gen8+ this register is not priviledged and we want to use it in
> > Mesa to implement a feature required by GPA called Null Hardware. The
> > idea is to have the command parser turn 3DPRIMITIVE/GPGPU_WALKER into
> > NOOPs.
> >
> > This patch just whitelists the bits that we need and that are
> > currently not used by the kernel.
> >
> 
> One thing I don't really know is whether is should be considered an 
> issue with the current command parser and therefore be backported as a fix.
> It would certainly make things easier because we can't really detect the 
> behavior from userspace.

You can always bump the cmdparser version. The cmdparser doesn't no-op,
it just rejects if a command is outside the whitelist, right? It would
be detectable...

The key question is whether INSTPM is context saved on gen7. A test in
gem_exec_parse to confirm we allow INSTPM in the command parser and that
one context does not leak to another would be useful.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✗ Fi.CI.CHECKPATCH: warning for drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Patchwork]

== Series Details ==

Series: drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits
URL   : https://patchwork.freedesktop.org/series/43420/
State : warning

== Summary ==

$ dim checkpatch origin/drm-tip
2fac9bc00108 drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits
-:44: CHECK:SPACING: spaces preferred around that '<<' (ctx:VxV)
#44: FILE: drivers/gpu/drm/i915/i915_reg.h:2534:
+#define   INSTPM_3D_MEDIA_INSTRUCTION_DISABLE     (1<<3) /* GEN6+ */
                                                     ^

-:45: CHECK:SPACING: spaces preferred around that '<<' (ctx:VxV)
#45: FILE: drivers/gpu/drm/i915/i915_reg.h:2535:
+#define   INSTPM_3D_RENDERING_INSTRUCTION_DISABLE (1<<2) /* GEN6+ */
                                                     ^

-:46: CHECK:SPACING: spaces preferred around that '<<' (ctx:VxV)
#46: FILE: drivers/gpu/drm/i915/i915_reg.h:2536:
+#define   INSTPM_3D_STATE_INSTRUCTION_DISABLE     (1<<1) /* GEN6+ */
                                                     ^

total: 0 errors, 0 warnings, 3 checks, 23 lines checked

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✓ Fi.CI.BAT: success for drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Patchwork]

== Series Details ==

Series: drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits
URL   : https://patchwork.freedesktop.org/series/43420/
State : success

== Summary ==

= CI Bug Log - changes from CI_DRM_4208 -> Patchwork_9052 =

== Summary - WARNING ==

  Minor unknown changes coming with Patchwork_9052 need to be verified
  manually.
  
  If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_9052, 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/43420/revisions/1/mbox/

== Possible new issues ==

  Here are the unknown changes that may have been introduced in Patchwork_9052:

  === IGT changes ===

    ==== Warnings ====

    igt@gem_exec_gttfill@basic:
      fi-pnv-d510:        SKIP -> PASS

    
== Known issues ==

  Here are the changes found in Patchwork_9052 that come from known issues:

  === IGT changes ===

    ==== Issues hit ====

    igt@kms_flip@basic-flip-vs-wf_vblank:
      fi-cnl-psr:         PASS -> FAIL (fdo#100368)

    igt@kms_pipe_crc_basic@nonblocking-crc-pipe-c-frame-sequence:
      fi-cnl-psr:         PASS -> FAIL (fdo#103481)

    
    ==== Possible fixes ====

    igt@kms_pipe_crc_basic@nonblocking-crc-pipe-b-frame-sequence:
      fi-skl-6770hq:      FAIL (fdo#103481) -> PASS

    
  fdo#100368 https://bugs.freedesktop.org/show_bug.cgi?id=100368
  fdo#103481 https://bugs.freedesktop.org/show_bug.cgi?id=103481


== Participating hosts (43 -> 39) ==

  Missing    (4): fi-ilk-m540 fi-byt-squawks fi-bsw-cyan fi-skl-6700hq 


== Build changes ==

    * Linux: CI_DRM_4208 -> Patchwork_9052

  CI_DRM_4208: a25e1590436f9d962fa1833a706a7c09bd83c3ea @ git://anongit.freedesktop.org/gfx-ci/linux
  IGT_4487: eccae1360d6d01e73c6af2bd97122cef708207ef @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
  Patchwork_9052: 2fac9bc001084e7f06ee4611d83ac86594757b6b @ git://anongit.freedesktop.org/gfx-ci/linux
  piglit_4487: 6ab75f7eb5e1dccbb773e1739beeb2d7cbd6ad0d @ git://anongit.freedesktop.org/piglit


== Linux commits ==

2fac9bc00108 drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_9052/issues.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Lionel Landwerlin]

On 18/05/18 15:40, Chris Wilson wrote:
> Quoting Lionel Landwerlin (2018-05-18 15:29:21)
>> On 18/05/18 15:26, Lionel Landwerlin wrote:
>>> On Gen8+ this register is not priviledged and we want to use it in
>>> Mesa to implement a feature required by GPA called Null Hardware. The
>>> idea is to have the command parser turn 3DPRIMITIVE/GPGPU_WALKER into
>>> NOOPs.
>>>
>>> This patch just whitelists the bits that we need and that are
>>> currently not used by the kernel.
>>>
>> One thing I don't really know is whether is should be considered an
>> issue with the current command parser and therefore be backported as a fix.
>> It would certainly make things easier because we can't really detect the
>> behavior from userspace.
> You can always bump the cmdparser version. The cmdparser doesn't no-op,
> it just rejects if a command is outside the whitelist, right? It would
> be detectable...

Ah right!

>
> The key question is whether INSTPM is context saved on gen7.

Documentation says it is. I'm not sure I trust it though :(

>   A test in
> gem_exec_parse to confirm we allow INSTPM in the command parser and that
> one context does not leak to another would be useful.

Sure.

> -Chris
>

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✓ Fi.CI.IGT: success for drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits

[Patchwork]

== Series Details ==

Series: drm/i915/cmdparser: Whitelist INSTPM instruction parsing disable bits
URL   : https://patchwork.freedesktop.org/series/43420/
State : success

== Summary ==

= CI Bug Log - changes from CI_DRM_4208_full -> Patchwork_9052_full =

== Summary - WARNING ==

  Minor unknown changes coming with Patchwork_9052_full need to be verified
  manually.
  
  If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_9052_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/43420/revisions/1/mbox/

== Possible new issues ==

  Here are the unknown changes that may have been introduced in Patchwork_9052_full:

  === IGT changes ===

    ==== Warnings ====

    igt@gem_exec_schedule@deep-vebox:
      shard-kbl:          PASS -> SKIP

    
== Known issues ==

  Here are the changes found in Patchwork_9052_full that come from known issues:

  === IGT changes ===

    ==== Issues hit ====

    igt@gem_exec_schedule@deep-bsd1:
      shard-kbl:          SKIP -> DMESG-WARN (fdo#103558, fdo#103313, fdo#105602)

    igt@kms_flip@2x-flip-vs-absolute-wf_vblank:
      shard-hsw:          PASS -> FAIL (fdo#103928)

    igt@kms_flip@2x-modeset-vs-vblank-race-interruptible:
      shard-hsw:          PASS -> FAIL (fdo#103060)

    igt@kms_flip@flip-vs-blocking-wf-vblank:
      shard-hsw:          PASS -> FAIL (fdo#100368)

    igt@kms_flip_tiling@flip-x-tiled:
      shard-glk:          PASS -> FAIL (fdo#104724, fdo#103822)

    igt@kms_frontbuffer_tracking@fbc-2p-primscrn-shrfb-pgflip-blt:
      shard-glk:          PASS -> FAIL (fdo#104724, fdo#103167)

    igt@kms_frontbuffer_tracking@fbc-rgb565-draw-pwrite:
      shard-apl:          PASS -> DMESG-FAIL (fdo#103558, fdo#105602)

    igt@kms_pipe_crc_basic@hang-read-crc-pipe-c:
      shard-apl:          PASS -> DMESG-WARN (fdo#103558, fdo#105602) +5

    igt@pm_rpm@legacy-planes:
      shard-kbl:          PASS -> DMESG-WARN (fdo#103558, fdo#103313)

    
    ==== Possible fixes ====

    igt@kms_atomic_transition@1x-modeset-transitions-nonblocking-fencing:
      shard-glk:          FAIL (fdo#105703) -> PASS

    igt@kms_flip@2x-wf_vblank-ts-check-interruptible:
      shard-glk:          FAIL (fdo#100368) -> PASS

    igt@kms_frontbuffer_tracking@fbc-2p-primscrn-pri-shrfb-draw-mmap-cpu:
      shard-glk:          FAIL (fdo#104724, fdo#103167) -> PASS

    igt@kms_setmode@basic:
      shard-apl:          FAIL (fdo#99912) -> PASS

    
  fdo#100368 https://bugs.freedesktop.org/show_bug.cgi?id=100368
  fdo#103060 https://bugs.freedesktop.org/show_bug.cgi?id=103060
  fdo#103167 https://bugs.freedesktop.org/show_bug.cgi?id=103167
  fdo#103313 https://bugs.freedesktop.org/show_bug.cgi?id=103313
  fdo#103558 https://bugs.freedesktop.org/show_bug.cgi?id=103558
  fdo#103822 https://bugs.freedesktop.org/show_bug.cgi?id=103822
  fdo#103928 https://bugs.freedesktop.org/show_bug.cgi?id=103928
  fdo#104724 https://bugs.freedesktop.org/show_bug.cgi?id=104724
  fdo#105602 https://bugs.freedesktop.org/show_bug.cgi?id=105602
  fdo#105703 https://bugs.freedesktop.org/show_bug.cgi?id=105703
  fdo#99912 https://bugs.freedesktop.org/show_bug.cgi?id=99912


== Participating hosts (9 -> 9) ==

  No changes in participating hosts


== Build changes ==

    * Linux: CI_DRM_4208 -> Patchwork_9052

  CI_DRM_4208: a25e1590436f9d962fa1833a706a7c09bd83c3ea @ git://anongit.freedesktop.org/gfx-ci/linux
  IGT_4487: eccae1360d6d01e73c6af2bd97122cef708207ef @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
  Patchwork_9052: 2fac9bc001084e7f06ee4611d83ac86594757b6b @ git://anongit.freedesktop.org/gfx-ci/linux
  piglit_4487: 6ab75f7eb5e1dccbb773e1739beeb2d7cbd6ad0d @ git://anongit.freedesktop.org/piglit

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_9052/shards.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx