[Intel-gfx] [PATCH 02/28] drm/i915/selftests: Make the hanging request non-preemptible

[Chris Wilson]

In some of our hangtests, we try to reset an active engine while it is
spinning inside the recursive spinner. However, we also try to flood the
engine with requests that preempt the hang, and so should disable the
preemption to be sure that we reset the right request.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 36 ++++++++++++++------
 1 file changed, 26 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index 4aa4cc917d8b..035f363fb0f8 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -203,12 +203,12 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
 		*batch++ = lower_32_bits(hws_address(hws, rq));
 		*batch++ = upper_32_bits(hws_address(hws, rq));
 		*batch++ = rq->fence.seqno;
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 
 		memset(batch, 0, 1024);
 		batch += 1024 / sizeof(*batch);
 
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 		*batch++ = MI_BATCH_BUFFER_START | 1 << 8 | 1;
 		*batch++ = lower_32_bits(vma->node.start);
 		*batch++ = upper_32_bits(vma->node.start);
@@ -217,12 +217,12 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
 		*batch++ = 0;
 		*batch++ = lower_32_bits(hws_address(hws, rq));
 		*batch++ = rq->fence.seqno;
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 
 		memset(batch, 0, 1024);
 		batch += 1024 / sizeof(*batch);
 
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 		*batch++ = MI_BATCH_BUFFER_START | 1 << 8;
 		*batch++ = lower_32_bits(vma->node.start);
 	} else if (INTEL_GEN(gt->i915) >= 4) {
@@ -230,24 +230,24 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
 		*batch++ = 0;
 		*batch++ = lower_32_bits(hws_address(hws, rq));
 		*batch++ = rq->fence.seqno;
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 
 		memset(batch, 0, 1024);
 		batch += 1024 / sizeof(*batch);
 
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 		*batch++ = MI_BATCH_BUFFER_START | 2 << 6;
 		*batch++ = lower_32_bits(vma->node.start);
 	} else {
 		*batch++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL;
 		*batch++ = lower_32_bits(hws_address(hws, rq));
 		*batch++ = rq->fence.seqno;
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 
 		memset(batch, 0, 1024);
 		batch += 1024 / sizeof(*batch);
 
-		*batch++ = MI_ARB_CHECK;
+		*batch++ = MI_NOOP;
 		*batch++ = MI_BATCH_BUFFER_START | 2 << 6;
 		*batch++ = lower_32_bits(vma->node.start);
 	}
@@ -866,13 +866,29 @@ static int __igt_reset_engines(struct intel_gt *gt,
 			count++;
 
 			if (rq) {
+				if (rq->fence.error != -EIO) {
+					pr_err("i915_reset_engine(%s:%s):"
+					       " failed to reset request %llx:%lld\n",
+					       engine->name, test_name,
+					       rq->fence.context,
+					       rq->fence.seqno);
+					i915_request_put(rq);
+
+					GEM_TRACE_DUMP();
+					intel_gt_set_wedged(gt);
+					err = -EIO;
+					break;
+				}
+
 				if (i915_request_wait(rq, 0, HZ / 5) < 0) {
 					struct drm_printer p =
 						drm_info_printer(gt->i915->drm.dev);
 
 					pr_err("i915_reset_engine(%s:%s):"
-					       " failed to complete request after reset\n",
-					       engine->name, test_name);
+					       " failed to complete request %llx:%lld after reset\n",
+					       engine->name, test_name,
+					       rq->fence.context,
+					       rq->fence.seqno);
 					intel_engine_dump(engine, &p,
 							  "%s\n", engine->name);
 					i915_request_put(rq);
-- 
2.20.1