st/dri: Don't expose sRGB formats to clients
authorDaniel Stone <daniels@collabora.com>
Fri, 31 Aug 2018 16:34:03 +0000 (17:34 +0100)
committerDaniel Stone <daniels@collabora.com>
Fri, 31 Aug 2018 17:02:42 +0000 (18:02 +0100)
Though the SARGB8888 format is used internally through its FourCC value,
it is not a real format as defined by drm_fourcc.h; it cannot be used
with KMS or other interfaces expecting drm_fourcc.h format codes.

Ensure we don't advertise it through the dmabuf format/modifier query
interfaces, preventing us from tripping over an assert.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reported-by: Michel Dänzer <michel.daenzer@amd.com>
Fixes: 8c1b9882b2e0 ("egl/dri2: Guard against invalid fourcc formats")
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
src/gallium/state_trackers/dri/dri2.c

index 2ac32205d9ac0cee954a7a724fa9633e88a10a97..c8a484e3926125a4ad10129152c574eeafdd1bf9 100644 (file)
@@ -1485,6 +1485,12 @@ dri2_query_dma_buf_formats(__DRIscreen *_screen, int max, int *formats,
 
    for (i = 0, j = 0; (i < ARRAY_SIZE(fourcc_formats)) &&
          (j < max || max == 0); i++) {
+      /* The sRGB format is not a real FourCC as defined by drm_fourcc.h, so we
+       * must not leak it out to clients.
+       */
+      if (fourcc_formats[i] == __DRI_IMAGE_FOURCC_SARGB8888)
+         continue;
+
       if (pscreen->is_format_supported(pscreen,
                                        fourcc_to_pipe_format(
                                           fourcc_formats[i]),