winsys/amdgpu: Retrieve WC flags from imported buffers.

Otherwise reading from an imported mapped GTT+WC linear texture
is painfully slow.

Sadly no radeon winsys implementation, as I don't know a suitable
kernel driver operation.

Hit this  in vaGetImage with an image imported from minigbm (which
we are switching to allocate WC for SCANOUT images).

Cc: <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
Acked-by: Pierre-Eric Pelloux-Prayer <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4542>

4 files changed, 25 insertions, 0 deletions

diff --git a/src/gallium/drivers/radeon/radeon_winsys.h b/src/gallium/drivers/radeon/radeon_winsys.h
index d02370af671..a2293837763 100644
--- a/src/gallium/drivers/radeon/radeon_winsys.h
+++ b/src/gallium/drivers/radeon/radeon_winsys.h
@@ -444,6 +444,14 @@ struct radeon_winsys {
     */
    enum radeon_bo_domain (*buffer_get_initial_domain)(struct pb_buffer *buf);
 
+   /**
+    * Query the flags used for creation of this buffer.
+    *
+    * Note that for imported buffer this may be lossy since not all flags
+    * are passed 1:1.
+    */
+   enum radeon_bo_flag (*buffer_get_flags)(struct pb_buffer *buf);
+
    /**************************************************************************
     * Command submission.
     *
diff --git a/src/gallium/drivers/radeonsi/si_texture.c b/src/gallium/drivers/radeonsi/si_texture.c
index 4f7744a887d..3b10fb0636b 100644
--- a/src/gallium/drivers/radeonsi/si_texture.c
+++ b/src/gallium/drivers/radeonsi/si_texture.c
@@ -1303,6 +1303,8 @@ static struct si_texture *si_texture_create_object(struct pipe_screen *screen,
          resource->vram_usage = resource->bo_size;
       else if (resource->domains & RADEON_DOMAIN_GTT)
          resource->gart_usage = resource->bo_size;
+      if (sscreen->ws->buffer_get_flags)
+         resource->flags = sscreen->ws->buffer_get_flags(resource->buf);
    }
 
    if (tex->cmask_buffer) {
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
index 8a821c8fd4f..e5a9278ce23 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.c
@@ -148,6 +148,12 @@ static enum radeon_bo_domain amdgpu_bo_get_initial_domain(
    return ((struct amdgpu_winsys_bo*)buf)->initial_domain;
 }
 
+static enum radeon_bo_flag amdgpu_bo_get_flags(
+      struct pb_buffer *buf)
+{
+   return ((struct amdgpu_winsys_bo*)buf)->flags;
+}
+
 static void amdgpu_bo_remove_fences(struct amdgpu_winsys_bo *bo)
 {
    for (unsigned i = 0; i < bo->num_fences; ++i)
@@ -559,6 +565,7 @@ static struct amdgpu_winsys_bo *amdgpu_create_bo(struct amdgpu_winsys *ws,
    bo->va = va;
    bo->u.real.va_handle = va_handle;
    bo->initial_domain = initial_domain;
+   bo->flags = flags;
    bo->unique_id = __sync_fetch_and_add(&ws->next_bo_unique_id, 1);
 
    if (initial_domain & RADEON_DOMAIN_VRAM)
@@ -1435,6 +1442,7 @@ static struct pb_buffer *amdgpu_bo_from_handle(struct radeon_winsys *rws,
    amdgpu_va_handle va_handle = NULL;
    struct amdgpu_bo_info info = {0};
    enum radeon_bo_domain initial = 0;
+   enum radeon_bo_flag flags = 0;
    int r;
 
    switch (whandle->type) {
@@ -1495,6 +1503,10 @@ static struct pb_buffer *amdgpu_bo_from_handle(struct radeon_winsys *rws,
       initial |= RADEON_DOMAIN_VRAM;
    if (info.preferred_heap & AMDGPU_GEM_DOMAIN_GTT)
       initial |= RADEON_DOMAIN_GTT;
+   if (info.alloc_flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)
+      flags |= RADEON_FLAG_NO_CPU_ACCESS;
+   if (info.alloc_flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
+      flags |= RADEON_FLAG_GTT_WC;
 
    /* Initialize the structure. */
    simple_mtx_init(&bo->lock, mtx_plain);
@@ -1507,6 +1519,7 @@ static struct pb_buffer *amdgpu_bo_from_handle(struct radeon_winsys *rws,
    bo->va = va;
    bo->u.real.va_handle = va_handle;
    bo->initial_domain = initial;
+   bo->flags = flags;
    bo->unique_id = __sync_fetch_and_add(&ws->next_bo_unique_id, 1);
    bo->is_shared = true;
 
@@ -1705,4 +1718,5 @@ void amdgpu_bo_init_functions(struct amdgpu_screen_winsys *ws)
    ws->base.buffer_commit = amdgpu_bo_sparse_commit;
    ws->base.buffer_get_virtual_address = amdgpu_bo_get_va;
    ws->base.buffer_get_initial_domain = amdgpu_bo_get_initial_domain;
+   ws->base.buffer_get_flags = amdgpu_bo_get_flags;
 }
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h
index fa8aabbeede..9f37437cd6e 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h
@@ -96,6 +96,7 @@ struct amdgpu_winsys_bo {
    uint32_t unique_id;
    uint64_t va;
    enum radeon_bo_domain initial_domain;
+   enum radeon_bo_flag flags;
 
    /* how many command streams is this bo referenced in? */
    int num_cs_references;