drm/msm/rd: support for 64b iova

For backwards compat, the rd format puts the high 32b after the size
field in the GPUADDR packet.

Signed-off-by: Rob Clark <robdclark@gmail.com>

diff --git a/drivers/gpu/drm/msm/msm_rd.c b/drivers/gpu/drm/msm/msm_rd.c
index 8487f46..6607456 100644 (file)
--- a/drivers/gpu/drm/msm/msm_rd.c
+++ b/drivers/gpu/drm/msm/msm_rd.c
@@ -289,7 +289,7 @@ void msm_rd_debugfs_cleanup(struct drm_minor *minor)
 
 static void snapshot_buf(struct msm_rd_state *rd,
                struct msm_gem_submit *submit, int idx,
-               uint32_t iova, uint32_t size)
+               uint64_t iova, uint32_t size)
 {
        struct msm_gem_object *obj = submit->bos[idx].obj;
        const char *buf;
@@ -306,7 +306,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
        }
 
        rd_write_section(rd, RD_GPUADDR,
-                       (uint32_t[2]){ iova, size }, 8);
+                       (uint32_t[3]){ iova, size, iova >> 32 }, 12);
        rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
        msm_gem_put_vaddr_locked(&obj->base);