Skip to contentSkip to conversation

drm/i915: avoid concurrent writes to aux_inv

PATCH In review [email protected] 3 participants 4 messages Mar 18, 2022

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

PATCH 1/1 drm/i915: avoid concurrent writes to aux_inv

From: Fei Yang <[email protected]>

GPU hangs have been observed when multiple engines write to the same aux_inv register at the same time. To avoid this each engine should only invalidate its own auxiliary table. The function gen12_emit_flush_xcs() currently invalidate the auxiliary table for all engines because the rq->engine is not necessarily the engine eventually carrying out the request, and potentially the engine could even be a virtual one (with engine->instance being -1). With the MMIO remap feature, we can actually set bit 17 of MI_LRI instruction and let the hardware to figure out the local aux_inv register at runtime to avoid invalidating auxiliary table for all engines.

Bspec: 45728

Cc Stuart Summers <[email protected]>
Cc Tvrtko Ursulin <[email protected]>
---
 drivers/gpu/drm/i915/gt/gen8_engine_cs.c     | 42 +++++---------------
 drivers/gpu/drm/i915/gt/intel_gpu_commands.h |  1 +
 2 files changed, 11 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +10 -32
index 36148887c699..d440c5dfb6b7 100644
@@ -165,30 +165,6 @@ static u32 preparser_disable(bool state)
 	return MI_ARB_CHECK | 1 << 8 | state;
 }
 
-static i915_reg_t aux_inv_reg(const struct intel_engine_cs *engine)
-{
-	static const i915_reg_t vd[] = {
-		GEN12_VD0_AUX_NV,
-		GEN12_VD1_AUX_NV,
-		GEN12_VD2_AUX_NV,
-		GEN12_VD3_AUX_NV,
-	};
-
-	static const i915_reg_t ve[] = {
-		GEN12_VE0_AUX_NV,
-		GEN12_VE1_AUX_NV,
-	};
-
-	if (engine->class == VIDEO_DECODE_CLASS)
-		return vd[engine->instance];
-
-	if (engine->class == VIDEO_ENHANCEMENT_CLASS)
-		return ve[engine->instance];
-
-	GEM_BUG_ON("unknown aux_inv reg\n");
-	return INVALID_MMIO_REG;
-}
-
 static u32 *gen12_emit_aux_table_inv(const i915_reg_t inv_reg, u32 *cs)
 {
 	*cs++ = MI_LOAD_REGISTER_IMM(1);
@@ -296,7 +272,7 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
 		if (!HAS_FLAT_CCS(rq->engine->i915)) {
 			aux_inv = rq->engine->mask & ~BIT(BCS0);
 			if (aux_inv)
-				cmd += 2 * hweight32(aux_inv) + 2;
+				cmd += 4;
 		}
 	}
 
@@ -329,14 +305,16 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
 	*cs++ = 0; /* value */
 
 	if (aux_inv) { /* hsdes: 1809175790 */
-		struct intel_engine_cs *engine;
-		unsigned int tmp;
-
-		*cs++ = MI_LOAD_REGISTER_IMM(hweight32(aux_inv));
-		for_each_engine_masked(engine, rq->engine->gt, aux_inv, tmp) {
-			*cs++ = i915_mmio_reg_offset(aux_inv_reg(engine));
-			*cs++ = AUX_INV;
+		*cs++ = MI_LOAD_REGISTER_IMM(1) | MI_LRI_MMIO_REMAP_EN;
+		if (rq->engine->class == VIDEO_DECODE_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VD0_AUX_NV);
+		} else if (rq->engine->class == VIDEO_ENHANCEMENT_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VE0_AUX_NV);
+		} else {
+			GEM_BUG_ON("unknown aux_inv reg\n");
+			*cs++ = i915_mmio_reg_offset(INVALID_MMIO_REG);
 		}
+		*cs++ = AUX_INV;
 		*cs++ = MI_NOOP;
 	}
 
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +1 -0
index d112ffd56418..2d150eec5c65 100644
@@ -144,6 +144,7 @@
 #define MI_LOAD_REGISTER_IMM(x)	MI_INSTR(0x22, 2*(x)-1)
 /* Gen11+. addr = base + (ctx_restore ? offset & GENMASK(12,2) : offset) */
 #define   MI_LRI_LRM_CS_MMIO		REG_BIT(19)
+#define   MI_LRI_MMIO_REMAP_EN		(1 << 17)
 #define   MI_LRI_FORCE_POSTED		(1<<12)
 #define MI_LOAD_REGISTER_IMM_MAX_REGS (126)
 #define MI_STORE_REGISTER_MEM        MI_INSTR(0x24, 1)
Signature · 1 line
2.25.1
#1 lore raw

[Intel-gfx] ✗ Fi.CI.BAT: failure for drm/i915: avoid concurrent writes to aux_inv (rev7)

== Series Details ==

== Series Details ==

Series: drm/i915: avoid concurrent writes to aux_inv (rev7)
URL : https://patchwork.freedesktop.org/series/100772/
State : failure

== Summary ==

CI Bug Log - changes from CI_DRM_11380 -> Patchwork_22606
====================================================

Summary
-------

**FAILURE**

Serious unknown changes coming with Patchwork_22606 absolutely need to be verified manually.

If you think the reported changes have nothing to do with the changes introduced in Patchwork_22606, 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_22606/index.html

Participating hosts (47 -> 45)
------------------------------

Additional (3): bat-adlm-1 bat-jsl-2 fi-pnv-d510
Missing (5): shard-tglu bat-dg1-5 fi-bsw-cyan shard-rkl fi-bdw-samus

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

Here are the unknown changes that may have been introduced in Patchwork_22606:

### IGT changes ###

#### Possible regressions ####

* igt@kms_cursor_legacy@basic-flip-before-cursor-atomic:
- fi-tgl-1115g4: [PASS][1] -> [DMESG-WARN][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_11380/fi-tgl-1115g4/igt@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-tgl-1115g4/igt@[email protected]

#### Suppressed ####

The following results come from untrusted machines, tests, or statuses. They do not affect the overall result.

* igt@i915_selftest@live@hangcheck:
- {bat-jsl-2}: NOTRUN -> [INCOMPLETE][3]
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/bat-jsl-2/igt@i915_selftest@[email protected]

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

Here are the changes found in Patchwork_22606 that come from known issues:

### IGT changes ###

#### Issues hit ####

* igt@gem_huc_copy@huc-copy:
- fi-pnv-d510: NOTRUN -> [SKIP][4] ([fdo#109271]) +57 similar issues
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-pnv-d510/igt@[email protected]

* igt@i915_selftest@live@hangcheck:
- bat-dg1-6: [PASS][5] -> [DMESG-FAIL][6] ([i915#4494] / [i915#4957])
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_11380/bat-dg1-6/igt@i915_selftest@[email protected]
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/bat-dg1-6/igt@i915_selftest@[email protected]
- fi-bdw-5557u: NOTRUN -> [INCOMPLETE][7] ([i915#3921])
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-bdw-5557u/igt@i915_selftest@[email protected]

* igt@kms_chamelium@vga-edid-read:
- fi-bdw-5557u: NOTRUN -> [SKIP][8] ([fdo#109271] / [fdo#111827]) +8 similar issues
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-bdw-5557u/igt@[email protected]

* igt@kms_pipe_crc_basic@compare-crc-sanitycheck-pipe-c:
- fi-pnv-d510: NOTRUN -> [SKIP][9] ([fdo#109271] / [i915#5341])
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-pnv-d510/igt@[email protected]
- fi-cfl-8109u: [PASS][10] -> [DMESG-WARN][11] ([i915#295] / [i915#5341])
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_11380/fi-cfl-8109u/igt@[email protected]
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-cfl-8109u/igt@[email protected]

* igt@kms_pipe_crc_basic@read-crc-pipe-b:
- fi-cfl-8109u: [PASS][12] -> [DMESG-WARN][13] ([i915#295]) +11 similar issues
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_11380/fi-cfl-8109u/igt@[email protected]
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-cfl-8109u/igt@[email protected]

* igt@kms_setmode@basic-clone-single-crtc:
- fi-bdw-5557u: NOTRUN -> [SKIP][14] ([fdo#109271]) +14 similar issues
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/fi-bdw-5557u/igt@[email protected]

#### Possible fixes ####

* igt@i915_pm_rpm@module-reload:
- {bat-rpls-2}: [DMESG-WARN][15] ([i915#4391]) -> [PASS][16] +1 similar issue
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_11380/bat-rpls-2/igt@[email protected]
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/bat-rpls-2/igt@[email protected]

{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).

[fdo#109271]: https://bugs.freedesktop.org/show_bug.cgi?id=109271
[fdo#109285]: https://bugs.freedesktop.org/show_bug.cgi?id=109285
[fdo#111825]: https://bugs.freedesktop.org/show_bug.cgi?id=111825
[fdo#111827]: https://bugs.freedesktop.org/show_bug.cgi?id=111827
[i915#1845]: https://gitlab.freedesktop.org/drm/intel/issues/1845
[i915#2190]: https://gitlab.freedesktop.org/drm/intel/issues/2190
[i915#2582]: https://gitlab.freedesktop.org/drm/intel/issues/2582
[i915#295]: https://gitlab.freedesktop.org/drm/intel/issues/295
[i915#3282]: https://gitlab.freedesktop.org/drm/intel/issues/3282
[i915#3301]: https://gitlab.freedesktop.org/drm/intel/issues/3301
[i915#3555]: https://gitlab.freedesktop.org/drm/intel/issues/3555
[i915#3576]: https://gitlab.freedesktop.org/drm/intel/issues/3576
[i915#3637]: https://gitlab.freedesktop.org/drm/intel/issues/3637
[i915#3921]: https://gitlab.freedesktop.org/drm/intel/issues/3921
[i915#4103]: https://gitlab.freedesktop.org/drm/intel/issues/4103
[i915#4391]: https://gitlab.freedesktop.org/drm/intel/issues/4391
[i915#4494]: https://gitlab.freedesktop.org/drm/intel/issues/4494
[i915#4613]: https://gitlab.freedesktop.org/drm/intel/issues/4613
[i915#4957]: https://gitlab.freedesktop.org/drm/intel/issues/4957
[i915#4983]: https://gitlab.freedesktop.org/drm/intel/issues/4983
[i915#533]: https://gitlab.freedesktop.org/drm/intel/issues/533
[i915#5338]: https://gitlab.freedesktop.org/drm/intel/issues/5338
[i915#5339]: https://gitlab.freedesktop.org/drm/intel/issues/5339
[i915#5341]: https://gitlab.freedesktop.org/drm/intel/issues/5341
[i915#5342]: https://gitlab.freedesktop.org/drm/intel/issues/5342
[i915#5361]: https://gitlab.freedesktop.org/drm/intel/issues/5361

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

* Linux: CI_DRM_11380 -> Patchwork_22606

CI-20190529: 20190529
CI_DRM_11380: fe83949cd4316608ea785fc376b6ed444224adad @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_6385: f3df40281d93d5a63ee98fa30e90852d780673c9 @ https://gitlab.freedesktop.org/drm/igt-gpu-tools.git
Patchwork_22606: 439959df00adab483693620266652e0ba6a2ee0a @ git://anongit.freedesktop.org/gfx-ci/linux

== Linux commits ==

439959df00ad drm/i915: avoid concurrent writes to aux_inv

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_22606/index.html

#2 lore raw
In reply to [email protected]
Quoted 71 lines · On 18/03/2022 05:26, [email protected] wrote:

From: Fei Yang <[email protected]>

GPU hangs have been observed when multiple engines write to the same aux_inv register at the same time. To avoid this each engine should only invalidate its own auxiliary table. The function gen12_emit_flush_xcs() currently invalidate the auxiliary table for all engines because the rq->engine is not necessarily the engine eventually carrying out the request, and potentially the engine could even be a virtual one (with engine->instance being -1). With the MMIO remap feature, we can actually set bit 17 of MI_LRI instruction and let the hardware to figure out the local aux_inv register at runtime to avoid invalidating auxiliary table for all engines.

Bspec: 45728

Cc Stuart Summers <[email protected]>
Cc Tvrtko Ursulin <[email protected]>
---
  drivers/gpu/drm/i915/gt/gen8_engine_cs.c     | 42 +++++---------------
  drivers/gpu/drm/i915/gt/intel_gpu_commands.h |  1 +
  2 files changed, 11 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -25
index 36148887c699..d440c5dfb6b7 100644
@@ -165,30 +165,6 @@ static u32 preparser_disable(bool state)
  	return MI_ARB_CHECK | 1 << 8 | state;
  }
 
-static i915_reg_t aux_inv_reg(const struct intel_engine_cs *engine)
-{
-	static const i915_reg_t vd[] = {
-		GEN12_VD0_AUX_NV,
-		GEN12_VD1_AUX_NV,
-		GEN12_VD2_AUX_NV,
-		GEN12_VD3_AUX_NV,
-	};
-
-	static const i915_reg_t ve[] = {
-		GEN12_VE0_AUX_NV,
-		GEN12_VE1_AUX_NV,
-	};
-
-	if (engine->class == VIDEO_DECODE_CLASS)
-		return vd[engine->instance];
-
-	if (engine->class == VIDEO_ENHANCEMENT_CLASS)
-		return ve[engine->instance];
-
-	GEM_BUG_ON("unknown aux_inv reg\n");
-	return INVALID_MMIO_REG;
-}
-
  static u32 *gen12_emit_aux_table_inv(const i915_reg_t inv_reg, u32 *cs)
  {
  	*cs++ = MI_LOAD_REGISTER_IMM(1);
@@ -296,7 +272,7 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
  		if (!HAS_FLAT_CCS(rq->engine->i915)) {
  			aux_inv = rq->engine->mask & ~BIT(BCS0);
  			if (aux_inv)
-				cmd += 2 * hweight32(aux_inv) + 2;
+				cmd += 4;
  		}
  	}
 
@@ -329,14 +305,16 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
  	*cs++ = 0; /* value */

drivers/gpu/drm/i915/gt/gen8_engine_cs.c int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode) · line 308

Quoted 10 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -7
 
  	if (aux_inv) { /* hsdes: 1809175790 */
-		struct intel_engine_cs *engine;
-		unsigned int tmp;
-
-		*cs++ = MI_LOAD_REGISTER_IMM(hweight32(aux_inv));
-		for_each_engine_masked(engine, rq->engine->gt, aux_inv, tmp) {
-			*cs++ = i915_mmio_reg_offset(aux_inv_reg(engine));
-			*cs++ = AUX_INV;
+		*cs++ = MI_LOAD_REGISTER_IMM(1) | MI_LRI_MMIO_REMAP_EN;

Cool, I didn't know this exists. First Bspec link I found did not mention these register, but second did. That one however (29545) has a worrying "removed by" tag which seems to point to a story suggesting the remapping table might be gone on machines with flat ccs?! Could you double check please?

drivers/gpu/drm/i915/gt/gen8_engine_cs.c int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode) · line 315

Quoted 7 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +7 -0
+		if (rq->engine->class == VIDEO_DECODE_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VD0_AUX_NV);
+		} else if (rq->engine->class == VIDEO_ENHANCEMENT_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VE0_AUX_NV);
+		} else {
+			GEM_BUG_ON("unknown aux_inv reg\n");
+			*cs++ = i915_mmio_reg_offset(INVALID_MMIO_REG);

I'd consider not emitting the LRI if we don't know what to put in unless there is some hidden point to do it?

Quoted 10 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -0
  		}
+		*cs++ = AUX_INV;
  		*cs++ = MI_NOOP;
  	}
 
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +0 -0
index d112ffd56418..2d150eec5c65 100644

drivers/gpu/drm/i915/gt/intel_gpu_commands.h line 148

Quoted 5 lines
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +1 -0
@@ -144,6 +144,7 @@
  #define MI_LOAD_REGISTER_IMM(x)	MI_INSTR(0x22, 2*(x)-1)
  /* Gen11+. addr = base + (ctx_restore ? offset & GENMASK(12,2) : offset) */
  #define   MI_LRI_LRM_CS_MMIO		REG_BIT(19)
+#define   MI_LRI_MMIO_REMAP_EN		(1 << 17)
  #define   MI_LRI_FORCE_POSTED		(1<<12)

Passing observation - three bits, three flavours of expressing them, sigh...

Regards,

Tvrtko

Quoted 2 lines
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +0 -0
  #define MI_LOAD_REGISTER_IMM_MAX_REGS (126)
  #define MI_STORE_REGISTER_MEM        MI_INSTR(0x24, 1)
author
#3 lore raw
Quoted 15 lines
Quoted 15 lines
  static u32 *gen12_emit_aux_table_inv(const i915_reg_t inv_reg, u32 *cs)
  {
  	*cs++ = MI_LOAD_REGISTER_IMM(1);
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -1
@@ -296,7 +272,7 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
  		if (!HAS_FLAT_CCS(rq->engine->i915)) {
  			aux_inv = rq->engine->mask & ~BIT(BCS0);
  			if (aux_inv)
-				cmd += 2 * hweight32(aux_inv) + 2;
+				cmd += 4;
  		}
  	}
 
@@ -329,14 +305,16 @@ int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode)
  	*cs++ = 0; /* value */
 
Quoted 13 lines

drivers/gpu/drm/i915/gt/gen8_engine_cs.c int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode) · line 308

Quoted 9 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -7
  	if (aux_inv) { /* hsdes: 1809175790 */
-		struct intel_engine_cs *engine;
-		unsigned int tmp;
-
-		*cs++ = MI_LOAD_REGISTER_IMM(hweight32(aux_inv));
-		for_each_engine_masked(engine, rq->engine->gt, aux_inv, tmp) {
-			*cs++ = i915_mmio_reg_offset(aux_inv_reg(engine));
-			*cs++ = AUX_INV;
+		*cs++ = MI_LOAD_REGISTER_IMM(1) | MI_LRI_MMIO_REMAP_EN;

Cool, I didn't know this exists. First Bspec link I found did not mention these register, but second did. That one however (29545) has a worrying "removed by" tag which seems to point to a story suggesting the remapping table might be gone on machines with flat ccs?! Could you double check please?

The variable aux_inv is set only if (!HAS_FLAT_CCS(rq->engine->i915)).

Quoted 9 lines

drivers/gpu/drm/i915/gt/gen8_engine_cs.c int gen12_emit_flush_xcs(struct i915_request *rq, u32 mode) · line 315

Quoted 7 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +7 -0
+		if (rq->engine->class == VIDEO_DECODE_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VD0_AUX_NV);
+		} else if (rq->engine->class == VIDEO_ENHANCEMENT_CLASS) {
+			*cs++ = i915_mmio_reg_offset(GEN12_VE0_AUX_NV);
+		} else {
+			GEM_BUG_ON("unknown aux_inv reg\n");
+			*cs++ = i915_mmio_reg_offset(INVALID_MMIO_REG);

I'd consider not emitting the LRI if we don't know what to put in unless there is some hidden point to do it?

That's true. I was following the original logic flow here. I think it would be better to check for engine class before setting the variable aux_inv.

Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/i915/gt/gen8_engine_cs.c +1 -0
  		}
+		*cs++ = AUX_INV;
  		*cs++ = MI_NOOP;
  	}
 
Quoted 13 lines

drivers/gpu/drm/i915/gt/intel_gpu_commands.h line 148

Quoted 11 lines
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +1 -0
index d112ffd56418..2d150eec5c65 100644
@@ -144,6 +144,7 @@
  #define MI_LOAD_REGISTER_IMM(x)	MI_INSTR(0x22, 2*(x)-1)
  /* Gen11+. addr = base + (ctx_restore ? offset & GENMASK(12,2) : offset) */
  #define   MI_LRI_LRM_CS_MMIO		REG_BIT(19)
+#define   MI_LRI_MMIO_REMAP_EN		(1 << 17)
  #define   MI_LRI_FORCE_POSTED		(1<<12)

Passing observation - three bits, three flavours of expressing them, sigh...

Haha, REG_BIT(17) it is. The other one causes a CHECK:SPACING, but don't want to touch that in this patch.

Quoted 2 lines
Quoted 2 lines
drivers/gpu/drm/i915/gt/intel_gpu_commands.h +0 -0
  #define MI_LOAD_REGISTER_IMM_MAX_REGS (126)
  #define MI_STORE_REGISTER_MEM        MI_INSTR(0x24, 1)