public inbox for igt-dev@lists.freedesktop.org
 help / color / mirror / Atom feed
* [Intel-gfx] [PATCH i-g-t] i915/gem_ctx_shared: Prefer explicit domain control
@ 2019-06-11  8:17 Chris Wilson
  2019-06-11  8:50 ` [igt-dev] ✓ Fi.CI.BAT: success for " Patchwork
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Chris Wilson @ 2019-06-11  8:17 UTC (permalink / raw)
  To: intel-gfx; +Cc: igt-dev

Since we are fiddling behind the scenes, we are writing to objects that
are not part of the execbuffer, do not rely on implicit domain
management being able to track the appropriate CPU cache status.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110890
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 tests/i915/gem_ctx_shared.c | 32 ++++++++++++++++++++++----------
 1 file changed, 22 insertions(+), 10 deletions(-)

diff --git a/tests/i915/gem_ctx_shared.c b/tests/i915/gem_ctx_shared.c
index 069964546..ed43e8903 100644
--- a/tests/i915/gem_ctx_shared.c
+++ b/tests/i915/gem_ctx_shared.c
@@ -184,27 +184,30 @@ static void exhaust_shared_gtt(int i915, unsigned int flags)
 static void exec_shared_gtt(int i915, unsigned int ring)
 {
 	const int gen = intel_gen(intel_get_drm_devid(i915));
-	const uint32_t bbe = MI_BATCH_BUFFER_END;
-	struct drm_i915_gem_exec_object2 obj = {
-		.handle = gem_create(i915, 4096)
-	};
+	struct drm_i915_gem_exec_object2 obj = {};
 	struct drm_i915_gem_execbuffer2 execbuf = {
 		.buffers_ptr = to_user_pointer(&obj),
 		.buffer_count = 1,
 		.flags = ring,
 	};
-	uint32_t scratch = obj.handle;
+	uint32_t scratch, *s;
 	uint32_t batch[16];
 	int i;
 
 	gem_require_ring(i915, ring);
 	igt_require(gem_can_store_dword(i915, ring));
 
+	scratch = gem_create(i915, 4096);
+	s = gem_mmap__cpu(i915, scratch, 0, 4096, PROT_WRITE);
+
+	gem_set_domain(i915, scratch, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
+	*s = MI_BATCH_BUFFER_END;
+
 	/* Load object into place in the GTT */
-	gem_write(i915, obj.handle, 0, &bbe, sizeof(bbe));
+	obj.handle = scratch;
 	gem_execbuf(i915, &execbuf);
 
-	/* Presume nothing causes an eviction in the meantime */
+	/* Presume nothing causes an eviction in the meantime! */
 
 	obj.handle = gem_create(i915, 4096);
 
@@ -235,10 +238,19 @@ static void exec_shared_gtt(int i915, unsigned int ring)
 	gem_sync(i915, obj.handle); /* write hazard lies */
 	gem_close(i915, obj.handle);
 
-	gem_read(i915, scratch, 0, batch, sizeof(uint32_t));
-	gem_close(i915, scratch);
+	/*
+	 * If we created the new context with the old GTT, the write
+	 * into the stale location of scratch will have landed in the right
+	 * object. Otherwise, it should read the previous value of
+	 * MI_BATCH_BUFFER_END.
+	 *
+	 * Setting .write = CPU to paper over our write hazard lies above.
+	 */
+	gem_set_domain(i915, scratch, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
+	igt_assert_eq_u32(*s, 0xc0ffee);
 
-	igt_assert_eq_u32(*batch, 0xc0ffee);
+	munmap(s, 4096);
+	gem_close(i915, scratch);
 }
 
 static int nop_sync(int i915, uint32_t ctx, unsigned int ring, int64_t timeout)
-- 
2.20.1

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

^ permalink raw reply related	[flat|nested] 4+ messages in thread
end of thread, other threads:[~2019-06-14 13:18 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-06-11  8:17 [Intel-gfx] [PATCH i-g-t] i915/gem_ctx_shared: Prefer explicit domain control Chris Wilson
2019-06-11  8:50 ` [igt-dev] ✓ Fi.CI.BAT: success for " Patchwork
2019-06-12 10:35 ` [igt-dev] ✓ Fi.CI.IGT: " Patchwork
2019-06-14 13:18 ` [Intel-gfx] [igt-dev] [PATCH i-g-t] " Mika Kuoppala

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox