Skip to contentSkip to conversation

drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern

PATCH Superseded Rafael Antognolli 3 participants 5 messages Jan 26, 2018 – Jan 27, 2018

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern

This workaround should prevent a bug that can be hit on a context restore. To avoid the issue, we must emit a PIPE_CONTROL with CS stall (0x7a000004 0x00100000 0x00000000 0x00000000) followed by 12DW's of NOOP(0x0) in the indirect context batch buffer, to ensure the engine is idle prior to programming 3DSTATE_SAMPLE_PATTERN.

References: HSD#1939868

v2:
- More descriptive changelog and comments.
- Fix math when counting dwords.

---
 drivers/gpu/drm/i915/intel_lrc.c | 34 +++++++++++++++++++++++++++++++++-
 1 file changed, 33 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_lrc.c +33 -1
index 51e61b04a555..46c130d106d7 100644
@@ -1366,6 +1366,36 @@ static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
 	return batch;
 }
 
+static u32 *gen10_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+	int i;
+
+	/*
+	 * WaPipeControlBefore3DStateSamplePattern: cnl
+	 *
+	 * Ensure the engine is idle prior to programming a
+	 * 3DSTATE_SAMPLE_PATTERN during a context restore.
+	 */
+	batch = gen8_emit_pipe_control(batch,
+				       PIPE_CONTROL_CS_STALL,
+				       0);
+	/*
+	 * WaPipeControlBefore3DStateSamplePattern says we need 4 dwords for
+	 * the PIPE_CONTROL followed by 12 dwords of 0x0, so 16 dwords in
+	 * total. Since gen8_emit_pipe_control() already advances the batch by
+	 * 6 dwords, we advance the other 10 here.
+	 */
+	for (i = 0; i < 10; i++)
+		*batch++ = MI_NOOP;
+
+	/* Pad to end of cacheline */
+	while ((unsigned long)batch % CACHELINE_BYTES)
+		*batch++ = MI_NOOP;
+
+	return batch;
+}
+
+
 #define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
 
 static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
@@ -1419,7 +1449,9 @@ static int intel_init_workaround_bb(struct intel_engine_cs *engine)
 
 	switch (INTEL_GEN(engine->i915)) {
 	case 10:
-		return 0;
+		wa_bb_fn[0] = gen10_init_indirectctx_bb;
+		wa_bb_fn[1] = NULL;
+		break;
 	case 9:
 		wa_bb_fn[0] = gen9_init_indirectctx_bb;
 		wa_bb_fn[1] = NULL;
Signature · 6 lines
2.14.3

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Acked-by on v2 1/1
#1 lore raw

Quoting Rafael Antognolli (2018-01-26 23:05:24)

Quoted 37 lines

This workaround should prevent a bug that can be hit on a context restore. To avoid the issue, we must emit a PIPE_CONTROL with CS stall (0x7a000004 0x00100000 0x00000000 0x00000000) followed by 12DW's of NOOP(0x0) in the indirect context batch buffer, to ensure the engine is idle prior to programming 3DSTATE_SAMPLE_PATTERN.

References: HSD#1939868

v2:
- More descriptive changelog and comments.
- Fix math when counting dwords.

---
 drivers/gpu/drm/i915/intel_lrc.c | 34 +++++++++++++++++++++++++++++++++-
 1 file changed, 33 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_lrc.c +11 -0
index 51e61b04a555..46c130d106d7 100644
@@ -1366,6 +1366,36 @@ static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
        return batch;
 }
 
+static u32 *gen10_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+       int i;
+
+       /*
+        * WaPipeControlBefore3DStateSamplePattern: cnl
+        *
+        * Ensure the engine is idle prior to programming a
+        * 3DSTATE_SAMPLE_PATTERN during a context restore.
+        */
+       batch = gen8_emit_pipe_control(batch,

drivers/gpu/drm/i915/intel_lrc.c static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch) · line 1389

Quoted 10 lines
drivers/gpu/drm/i915/intel_lrc.c +10 -0
+                                      PIPE_CONTROL_CS_STALL,
+                                      0);
+       /*
+        * WaPipeControlBefore3DStateSamplePattern says we need 4 dwords for
+        * the PIPE_CONTROL followed by 12 dwords of 0x0, so 16 dwords in
+        * total. Since gen8_emit_pipe_control() already advances the batch by
+        * 6 dwords, we advance the other 10 here.
+        */
+       for (i = 0; i < 10; i++)
+               *batch++ = MI_NOOP;

If the spec says emit 12 nops, emit 12 nops. Otherwise if we add another w/a here it may break this magic.

I am not going to ask what the magic nops do, it just gives a 72 cycle delay in the CS parser. My guess is that they are trying to isolate a cacheline, in which case they are expecting the pipecontrol to be cacheline aligned. (I would get that clarified and add an assert to document such a requirement.) Or they just mean that the indirect ctx is cacheline aligned and the nops are just the normal padding and nothing to do with the w/a.

s/10/12/ here and tweak the comment to remove the assumed cacheline tail, but please see if someone can clarify if the nops are indeed part of the w/a or just the normal filler.
With s/10/12/,

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

#2 lore raw
Quoted 43 lines · Chris Wilson

Quoting Rafael Antognolli (2018-01-26 23:05:24)

Quoted 42 lines

This workaround should prevent a bug that can be hit on a context restore. To avoid the issue, we must emit a PIPE_CONTROL with CS stall (0x7a000004 0x00100000 0x00000000 0x00000000) followed by 12DW's of NOOP(0x0) in the indirect context batch buffer, to ensure the engine is idle prior to programming 3DSTATE_SAMPLE_PATTERN.

References: HSD#1939868

v2:
- More descriptive changelog and comments.
- Fix math when counting dwords.

---
 drivers/gpu/drm/i915/intel_lrc.c | 34 +++++++++++++++++++++++++++++++++-
 1 file changed, 33 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_lrc.c +16 -0
index 51e61b04a555..46c130d106d7 100644
@@ -1366,6 +1366,36 @@ static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
        return batch;
 }
 
+static u32 *gen10_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+       int i;
+
+       /*
+        * WaPipeControlBefore3DStateSamplePattern: cnl
+        *
+        * Ensure the engine is idle prior to programming a
+        * 3DSTATE_SAMPLE_PATTERN during a context restore.
+        */
+       batch = gen8_emit_pipe_control(batch,
+                                      PIPE_CONTROL_CS_STALL,
+                                      0);
+       /*
+        * WaPipeControlBefore3DStateSamplePattern says we need 4 dwords for
+        * the PIPE_CONTROL followed by 12 dwords of 0x0, so 16 dwords in
Quoted 8 lines

drivers/gpu/drm/i915/intel_lrc.c static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch) · line 1389

Quoted 5 lines
drivers/gpu/drm/i915/intel_lrc.c +5 -0
+        * total. Since gen8_emit_pipe_control() already advances the batch by
+        * 6 dwords, we advance the other 10 here.
+        */
+       for (i = 0; i < 10; i++)
+               *batch++ = MI_NOOP;

If the spec says emit 12 nops, emit 12 nops. Otherwise if we add another w/a here it may break this magic.

So, in the bspec, there's the following text:

"The address above needs to be a GGTT address and contain a pipe control with CS stall(0x7a000004 0x00100000 0x00000000 0x00000000)followed by 12DW’s of NOOP(0x00000000)"

while on an internal ticket pointed by it, it is:

"The address above needs to be a GGTT address.. It would contain the following value: 7A000004 00100000 followed by 14 DW’s of zero."

That's why I think it is really 16 DW's in total. And gen8_emit_pipe_control() already emits 6 of them. How are we supposed to emit 12 after that?

I don't know if they expect the pipe control to be cacheline aligned, but I can ask and try to find out.

Quoted 8 lines

I am not going to ask what the magic nops do, it just gives a 72 cycle delay in the CS parser. My guess is that they are trying to isolate a cacheline, in which case they are expecting the pipecontrol to be cacheline aligned. (I would get that clarified and add an assert to document such a requirement.) Or they just mean that the indirect ctx is cacheline aligned and the nops are just the normal padding and nothing to do with the w/a.

Quoted 6 lines

s/10/12/ here and tweak the comment to remove the assumed cacheline tail, but please see if someone can clarify if the nops are indeed part of the w/a or just the normal filler.
With s/10/12/,

-Chris

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

#3 lore raw

✓ Fi.CI.BAT: success for drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern (rev2)

In reply to Rafael Antognolli
== Series Details ==

== Series Details ==

Series: drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern (rev2)
URL : https://patchwork.freedesktop.org/series/37148/
State : success

== Summary ==

Series 37148v2 drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern
https://patchwork.freedesktop.org/api/1.0/series/37148/revisions/2/mbox/

Test debugfs_test:

        Subgroup read_all_entries:
                dmesg-fail -> DMESG-WARN (fi-elk-e7500) fdo#103989
Test kms_pipe_crc_basic:
        Subgroup suspend-read-crc-pipe-b:
                pass       -> INCOMPLETE (fi-snb-2520m) fdo#103713

fdo#103989 https://bugs.freedesktop.org/show_bug.cgi?id=103989
fdo#103713 https://bugs.freedesktop.org/show_bug.cgi?id=103713

fi-bdw-5557u     total:288  pass:267  dwarn:0   dfail:0   fail:0   skip:21  time:422s
fi-bdw-gvtdvm    total:288  pass:264  dwarn:0   dfail:0   fail:0   skip:24  time:426s
fi-blb-e6850     total:288  pass:223  dwarn:1   dfail:0   fail:0   skip:64  time:371s
fi-bsw-n3050     total:288  pass:242  dwarn:0   dfail:0   fail:0   skip:46  time:490s
fi-bwr-2160      total:288  pass:183  dwarn:0   dfail:0   fail:0   skip:105 time:281s
fi-bxt-dsi       total:288  pass:258  dwarn:0   dfail:0   fail:0   skip:30  time:480s
fi-bxt-j4205     total:288  pass:259  dwarn:0   dfail:0   fail:0   skip:29  time:488s
fi-byt-j1900     total:288  pass:253  dwarn:0   dfail:0   fail:0   skip:35  time:470s
fi-byt-n2820     total:288  pass:249  dwarn:0   dfail:0   fail:0   skip:39  time:465s
fi-cfl-s2        total:288  pass:262  dwarn:0   dfail:0   fail:0   skip:26  time:572s
fi-cnl-y2        total:288  pass:261  dwarn:0   dfail:0   fail:0   skip:27  time:530s
fi-elk-e7500     total:224  pass:168  dwarn:10  dfail:0   fail:0   skip:45 
fi-gdg-551       total:288  pass:179  dwarn:0   dfail:0   fail:1   skip:108 time:294s
fi-glk-1         total:288  pass:260  dwarn:0   dfail:0   fail:0   skip:28  time:516s
fi-hsw-4770      total:288  pass:261  dwarn:0   dfail:0   fail:0   skip:27  time:392s
fi-hsw-4770r     total:288  pass:261  dwarn:0   dfail:0   fail:0   skip:27  time:401s
fi-ilk-650       total:288  pass:228  dwarn:0   dfail:0   fail:0   skip:60  time:423s
fi-ivb-3520m     total:288  pass:259  dwarn:0   dfail:0   fail:0   skip:29  time:457s
fi-ivb-3770      total:288  pass:255  dwarn:0   dfail:0   fail:0   skip:33  time:414s
fi-kbl-7500u     total:288  pass:263  dwarn:1   dfail:0   fail:0   skip:24  time:459s
fi-kbl-7560u     total:288  pass:269  dwarn:0   dfail:0   fail:0   skip:19  time:502s
fi-kbl-7567u     total:288  pass:268  dwarn:0   dfail:0   fail:0   skip:20  time:456s
fi-kbl-r         total:288  pass:261  dwarn:0   dfail:0   fail:0   skip:27  time:501s
fi-pnv-d510      total:288  pass:222  dwarn:1   dfail:0   fail:0   skip:65  time:578s
fi-skl-6260u     total:288  pass:268  dwarn:0   dfail:0   fail:0   skip:20  time:433s
fi-skl-6600u     total:288  pass:261  dwarn:0   dfail:0   fail:0   skip:27  time:510s
fi-skl-6700hq    total:288  pass:262  dwarn:0   dfail:0   fail:0   skip:26  time:533s
fi-skl-6700k2    total:288  pass:264  dwarn:0   dfail:0   fail:0   skip:24  time:498s
fi-skl-6770hq    total:288  pass:268  dwarn:0   dfail:0   fail:0   skip:20  time:483s
fi-skl-guc       total:288  pass:260  dwarn:0   dfail:0   fail:0   skip:28  time:418s
fi-skl-gvtdvm    total:288  pass:265  dwarn:0   dfail:0   fail:0   skip:23  time:431s
fi-snb-2520m     total:245  pass:211  dwarn:0   dfail:0   fail:0   skip:33 
fi-snb-2600      total:288  pass:248  dwarn:0   dfail:0   fail:0   skip:40  time:401s

Blacklisted hosts:
fi-glk-dsi total:288 pass:258 dwarn:0 dfail:0 fail:0 skip:30 time:480s

59275f1cec1d31adab39ddb6ab948519ac8ddffb drm-tip: 2018y-01m-26d-13h-05m-14s UTC integration manifest 7b8180832a9f drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_7796/issues.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#4 lore raw

✓ Fi.CI.IGT: success for drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern (rev2)

In reply to Rafael Antognolli
== Series Details ==

== Series Details ==

Series: drm/i915/cnl: WaPipeControlBefore3DStateSamplePattern (rev2)
URL : https://patchwork.freedesktop.org/series/37148/
State : success

== Summary ==

Warning: bzip CI_DRM_3686/shard-glkb6/results14.json.bz2 wasn't in correct JSON format Test kms_flip:

        Subgroup plain-flip-fb-recreate-interruptible:
                pass       -> FAIL       (shard-hsw) fdo#100368
Test kms_sysfs_edid_timing:
                warn       -> PASS       (shard-apl) fdo#100047
Test kms_mmap_write_crc:
                skip       -> PASS       (shard-apl)
Test kms_frontbuffer_tracking:
        Subgroup fbc-1p-offscren-pri-shrfb-draw-blt:
                fail       -> PASS       (shard-snb) fdo#101623
Test perf:
        Subgroup enable-disable:
                fail       -> PASS       (shard-apl) fdo#103715

fdo#100368 https://bugs.freedesktop.org/show_bug.cgi?id=100368
fdo#100047 https://bugs.freedesktop.org/show_bug.cgi?id=100047
fdo#101623 https://bugs.freedesktop.org/show_bug.cgi?id=101623
fdo#103715 https://bugs.freedesktop.org/show_bug.cgi?id=103715

shard-apl        total:2678 pass:1668 dwarn:1   dfail:0   fail:20  skip:989 time:11995s
shard-hsw        total:2838 pass:1735 dwarn:1   dfail:0   fail:11  skip:1090 time:12054s
shard-snb        total:2838 pass:1330 dwarn:1   dfail:0   fail:10  skip:1497 time:6658s

Blacklisted hosts:
shard-kbl total:2825 pass:1854 dwarn:5 dfail:0 fail:22 skip:943 time:9539s

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_7796/shards.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx