freedreno/a5xx: align height to GMEM

Similar to the way width/pitch alignment works, it seems like we need to
do similar for height.  Otherwise the BLIT from system memory to GMEM
can over-fetch beyond the end of the buffer, triggering a fault.

I'm not sure if there is a better solution yet.  Possibly we could fall
back to pre-a5xx style DRAW packets for cases where BLIT might over-
fetch.  (We in theory have that problem already with rendering to higher
mipmap levels, although fortunately those tend to use GMEM bypass.)

This fixes issues reported with glamor.

Reported-by: don.harbin@linaro.org
Cc: 17.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
This commit is contained in:
Rob Clark 2017-09-14 15:03:11 -04:00
parent 146c2b7c28
commit 16ac70bdcf
1 changed files with 5 additions and 1 deletions

View File

@ -704,8 +704,9 @@ static uint32_t
setup_slices(struct fd_resource *rsc, uint32_t alignment, enum pipe_format format)
{
struct pipe_resource *prsc = &rsc->base.b;
struct fd_screen *screen = fd_screen(prsc->screen);
enum util_format_layout layout = util_format_description(format)->layout;
uint32_t pitchalign = fd_screen(prsc->screen)->gmem_alignw;
uint32_t pitchalign = screen->gmem_alignw;
uint32_t level, size = 0;
uint32_t width = prsc->width0;
uint32_t height = prsc->height0;
@ -715,6 +716,9 @@ setup_slices(struct fd_resource *rsc, uint32_t alignment, enum pipe_format forma
*/
uint32_t layers_in_level = rsc->layer_first ? 1 : prsc->array_size;
if (is_a5xx(screen) && (rsc->base.b.target >= PIPE_TEXTURE_2D))
height = align(height, screen->gmem_alignh);
for (level = 0; level <= prsc->last_level; level++) {
struct fd_resource_slice *slice = fd_resource_slice(rsc, level);
uint32_t blocks;