[Bug 105720] screen corruption using SNA and TearFree on Intel GeminiLake

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=105720

--- Comment #36 from Chris Wilson <chris at chris-wilson.co.uk> ---
So, we simplified the wait_for_shadow() to always do a fullscreen copy onto a
new buffer and by first filling the buffer with a checkerboard we confirmed
that the copy was not occurring.

By using #define DEBUG_NO_BLT 1 and comparing with Option "AccelMethod" "blt",
we were able to conclude this affects the render copy; i.e. gen9_render.c seems
completely suspect.

That the copy fails entirely suggests some sort of pixel kill, or that the
sampler read hits 0. The latter has happened before, when the sample offset was
larger the supported table size.

But there is another possibility, that we fail to submit the sampler table,
which is tucked in at the end of the batch.

Currently we are investigating this angle with

diff --git a/src/sna/sna.h b/src/sna/sna.h
index 6fe1f0d2..af61e4b7 100644
--- a/src/sna/sna.h
+++ b/src/sna/sna.h
@@ -80,7 +80,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "compiler.h"
 #include "debug.h"

-#define DEBUG_NO_BLT 0
+#define DEBUG_NO_BLT 1

 #define DEBUG_FLUSH_BATCH 0

and so far, all seems ok.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20180405/b9a680e3/attachment.html>