drm/i915/gt: Flush gen3 relocs harder, again

gen3 does not fully flush MI stores to memory on MI_FLUSH, such that a
subsequent read from e.g. the sampler can bypass the store and read the
stale value from memory. This is a serious issue when we are using MI
stores to rewrite the batches for relocation, as it means that the batch
is reading from random user/kernel memory. While it is particularly
sensitive [and detectable] for relocations, reading stale data at any
time is a worry.

Having started with a small number of delaying stores and doubling until
no more incoherency was seen over a few hours (with and without
background memory pressure), 32 was the magic number.

Note that it definitely doesn't fix the issue, merely adds a long delay
between requests, sufficient to mostly hide the problem, enough to raise
the mtbf to several hours. This is merely a stop gap.

v2: Follow more closer with the gen5 w/a and include some
post-invalidate flushes as well.

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2018
References: a889580c ("drm/i915: Flush GPU relocs harder for gen3")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200612123949.7093-1-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/gt/gen2_engine_cs.c

@@ -13,28 +13,25 @@
 
 int gen2_emit_flush(struct i915_request *rq, u32 mode)
 {
-	unsigned int num_store_dw;
+	unsigned int num_store_dw = 12;
 	u32 cmd, *cs;
 
 	cmd = MI_FLUSH;
-	num_store_dw = 0;
 	if (mode & EMIT_INVALIDATE)
 		cmd |= MI_READ_FLUSH;
-	if (mode & EMIT_FLUSH)
-		num_store_dw = 4;
 
-	cs = intel_ring_begin(rq, 2 + 3 * num_store_dw);
+	cs = intel_ring_begin(rq, 2 + 4 * num_store_dw);
 	if (IS_ERR(cs))
 		return PTR_ERR(cs);
 
 	*cs++ = cmd;
 	while (num_store_dw--) {
-		*cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL;
-		*cs++ = intel_gt_scratch_offset(rq->engine->gt,
-						INTEL_GT_SCRATCH_FIELD_DEFAULT);
+		*cs++ = MI_STORE_DWORD_INDEX;
+		*cs++ = I915_GEM_HWS_SCRATCH * sizeof(u32);
 		*cs++ = 0;
+		*cs++ = MI_FLUSH | MI_NO_WRITE_FLUSH;
 	}
-	*cs++ = MI_FLUSH | MI_NO_WRITE_FLUSH;
+	*cs++ = cmd;
 
 	intel_ring_advance(rq, cs);
 
@@ -142,38 +139,21 @@ int gen4_emit_flush_vcs(struct i915_request *rq, u32 mode)
 	return 0;
 }
 
-u32 *gen3_emit_breadcrumb(struct i915_request *rq, u32 *cs)
+static u32 *__gen2_emit_breadcrumb(struct i915_request *rq, u32 *cs,
+				   int flush, int post)
 {
 	GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma);
 	GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR);
 
 	*cs++ = MI_FLUSH;
 
-	*cs++ = MI_STORE_DWORD_INDEX;
-	*cs++ = I915_GEM_HWS_SEQNO_ADDR;
-	*cs++ = rq->fence.seqno;
-
-	*cs++ = MI_USER_INTERRUPT;
-	*cs++ = MI_NOOP;
-
-	rq->tail = intel_ring_offset(rq, cs);
-	assert_ring_tail_valid(rq->ring, rq->tail);
-
-	return cs;
-}
-
-#define GEN5_WA_STORES 8 /* must be at least 1! */
-u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs)
-{
-	int i;
-
-	GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma);
-	GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR);
-
-	*cs++ = MI_FLUSH;
+	while (flush--) {
+		*cs++ = MI_STORE_DWORD_INDEX;
+		*cs++ = I915_GEM_HWS_SCRATCH * sizeof(u32);
+		*cs++ = rq->fence.seqno;
+	}
 
-	BUILD_BUG_ON(GEN5_WA_STORES < 1);
-	for (i = 0; i < GEN5_WA_STORES; i++) {
+	while (post--) {
 		*cs++ = MI_STORE_DWORD_INDEX;
 		*cs++ = I915_GEM_HWS_SEQNO_ADDR;
 		*cs++ = rq->fence.seqno;
@@ -186,7 +166,16 @@ u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs)
 
 	return cs;
 }
-#undef GEN5_WA_STORES
+
+u32 *gen3_emit_breadcrumb(struct i915_request *rq, u32 *cs)
+{
+	return __gen2_emit_breadcrumb(rq, cs, 16, 8);
+}
+
+u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs)
+{
+	return __gen2_emit_breadcrumb(rq, cs, 8, 8);
+}
 
 /* Just userspace ABI convention to limit the wa batch bo to a resonable size */
 #define I830_BATCH_LIMIT SZ_256K