radeonsi: set a per-buffer flag that disables inter-process sharing (v4)

For lower overhead in the CS ioctl.
Winsys allocators are not used with interprocess-sharable resources.

v2: It shouldn't crash anymore, but the kernel will reject the new flag.
v3 (christian): Rename the flag, avoid sending those buffers in the BO list.
v4 (christian): Remove setting the kernel flag for now

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

src/gallium/drivers/radeon/r600_buffer_common.c

@@ -167,6 +167,13 @@ void r600_init_resource_fields(struct r600_common_screen *rscreen,
 			 RADEON_FLAG_GTT_WC;
 	}
 
+	/* Only displayable single-sample textures can be shared between
+	 * processes. */
+	if (res->b.b.target == PIPE_BUFFER ||
+	    res->b.b.nr_samples >= 2 ||
+	    rtex->surface.micro_tile_mode != RADEON_MICRO_MODE_DISPLAY)
+		res->flags |= RADEON_FLAG_NO_INTERPROCESS_SHARING;
+
 	/* If VRAM is just stolen system memory, allow both VRAM and
 	 * GTT, whichever has free space. If a buffer is evicted from
 	 * VRAM to GTT, it will stay there.

src/gallium/drivers/radeon/radeon_winsys.h

@@ -54,6 +54,7 @@ enum radeon_bo_flag { /* bitfield */
     RADEON_FLAG_NO_CPU_ACCESS = (1 << 1),
     RADEON_FLAG_NO_SUBALLOC =   (1 << 2),
     RADEON_FLAG_SPARSE =        (1 << 3),
+    RADEON_FLAG_NO_INTERPROCESS_SHARING = (1 << 4),
 };
 
 enum radeon_bo_usage { /* bitfield */
@@ -661,14 +662,19 @@ static inline unsigned radeon_flags_from_heap(enum radeon_heap heap)
 {
     switch (heap) {
     case RADEON_HEAP_VRAM_NO_CPU_ACCESS:
-        return RADEON_FLAG_GTT_WC | RADEON_FLAG_NO_CPU_ACCESS;
+        return RADEON_FLAG_GTT_WC |
+               RADEON_FLAG_NO_CPU_ACCESS |
+               RADEON_FLAG_NO_INTERPROCESS_SHARING;
+
     case RADEON_HEAP_VRAM:
     case RADEON_HEAP_VRAM_GTT:
     case RADEON_HEAP_GTT_WC:
-        return RADEON_FLAG_GTT_WC;
+        return RADEON_FLAG_GTT_WC |
+               RADEON_FLAG_NO_INTERPROCESS_SHARING;
+
     case RADEON_HEAP_GTT:
     default:
-        return 0;
+        return RADEON_FLAG_NO_INTERPROCESS_SHARING;
     }
 }
 
@@ -700,8 +706,14 @@ static inline int radeon_get_heap_index(enum radeon_bo_domain domain,
     /* NO_CPU_ACCESS implies VRAM only. */
     assert(!(flags & RADEON_FLAG_NO_CPU_ACCESS) || domain == RADEON_DOMAIN_VRAM);
 
+    /* Resources with interprocess sharing don't use any winsys allocators. */
+    if (!(flags & RADEON_FLAG_NO_INTERPROCESS_SHARING))
+        return -1;
+
     /* Unsupported flags: NO_SUBALLOC, SPARSE. */
-    if (flags & ~(RADEON_FLAG_GTT_WC | RADEON_FLAG_NO_CPU_ACCESS))
+    if (flags & ~(RADEON_FLAG_GTT_WC |
+                  RADEON_FLAG_NO_CPU_ACCESS |
+                  RADEON_FLAG_NO_INTERPROCESS_SHARING))
         return -1;
 
     switch (domain) {

src/gallium/winsys/amdgpu/drm/amdgpu_bo.c

@@ -1138,7 +1138,7 @@ amdgpu_bo_create(struct radeon_winsys *rws,
 {
    struct amdgpu_winsys *ws = amdgpu_winsys(rws);
    struct amdgpu_winsys_bo *bo;
-   unsigned usage = 0, pb_cache_bucket;
+   unsigned usage = 0, pb_cache_bucket = 0;
 
    /* VRAM implies WC. This is not optional. */
    assert(!(domain & RADEON_DOMAIN_VRAM) || flags & RADEON_FLAG_GTT_WC);
@@ -1193,19 +1193,23 @@ no_slab:
    size = align64(size, ws->info.gart_page_size);
    alignment = align(alignment, ws->info.gart_page_size);
 
-   int heap = radeon_get_heap_index(domain, flags);
+   bool use_reusable_pool = flags & RADEON_FLAG_NO_INTERPROCESS_SHARING;
-   assert(heap >= 0 && heap < RADEON_MAX_CACHED_HEAPS);
-   usage = 1 << heap; /* Only set one usage bit for each heap. */
 
-   pb_cache_bucket = radeon_get_pb_cache_bucket_index(heap);
+   if (use_reusable_pool) {
-   assert(pb_cache_bucket < ARRAY_SIZE(ws->bo_cache.buckets));
+       int heap = radeon_get_heap_index(domain, flags);
+       assert(heap >= 0 && heap < RADEON_MAX_CACHED_HEAPS);
+       usage = 1 << heap; /* Only set one usage bit for each heap. */
 
-   /* Get a buffer from the cache. */
+       pb_cache_bucket = radeon_get_pb_cache_bucket_index(heap);
-   bo = (struct amdgpu_winsys_bo*)
+       assert(pb_cache_bucket < ARRAY_SIZE(ws->bo_cache.buckets));
-        pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment, usage,
+
-                                pb_cache_bucket);
+       /* Get a buffer from the cache. */
-   if (bo)
+       bo = (struct amdgpu_winsys_bo*)
-      return &bo->base;
+            pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment, usage,
+                                    pb_cache_bucket);
+       if (bo)
+          return &bo->base;
+   }
 
    /* Create a new one. */
    bo = amdgpu_create_bo(ws, size, alignment, usage, domain, flags,
@@ -1220,7 +1224,7 @@ no_slab:
          return NULL;
    }
 
-   bo->u.real.use_reusable_pool = true;
+   bo->u.real.use_reusable_pool = use_reusable_pool;
    return &bo->base;
 }
 

src/gallium/winsys/radeon/drm/radeon_drm_bo.c

@@ -914,7 +914,7 @@ radeon_winsys_bo_create(struct radeon_winsys *rws,
 {
     struct radeon_drm_winsys *ws = radeon_drm_winsys(rws);
     struct radeon_bo *bo;
-    unsigned usage = 0, pb_cache_bucket;
+    unsigned usage = 0, pb_cache_bucket = 0;
 
     assert(!(flags & RADEON_FLAG_SPARSE)); /* not supported */
 
@@ -969,17 +969,22 @@ no_slab:
     size = align(size, ws->info.gart_page_size);
     alignment = align(alignment, ws->info.gart_page_size);
 
-    int heap = radeon_get_heap_index(domain, flags);
+    bool use_reusable_pool = flags & RADEON_FLAG_NO_INTERPROCESS_SHARING;
-    assert(heap >= 0 && heap < RADEON_MAX_CACHED_HEAPS);
-    usage = 1 << heap; /* Only set one usage bit for each heap. */
 
-    pb_cache_bucket = radeon_get_pb_cache_bucket_index(heap);
+    /* Shared resources don't use cached heaps. */
-    assert(pb_cache_bucket < ARRAY_SIZE(ws->bo_cache.buckets));
+    if (use_reusable_pool) {
+        int heap = radeon_get_heap_index(domain, flags);
+        assert(heap >= 0 && heap < RADEON_MAX_CACHED_HEAPS);
+        usage = 1 << heap; /* Only set one usage bit for each heap. */
 
-    bo = radeon_bo(pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment,
+        pb_cache_bucket = radeon_get_pb_cache_bucket_index(heap);
-                                           usage, pb_cache_bucket));
+        assert(pb_cache_bucket < ARRAY_SIZE(ws->bo_cache.buckets));
-    if (bo)
+
-        return &bo->base;
+        bo = radeon_bo(pb_cache_reclaim_buffer(&ws->bo_cache, size, alignment,
+                                               usage, pb_cache_bucket));
+        if (bo)
+            return &bo->base;
+    }
 
     bo = radeon_create_bo(ws, size, alignment, usage, domain, flags,
                           pb_cache_bucket);
@@ -994,7 +999,7 @@ no_slab:
             return NULL;
     }
 
-    bo->u.real.use_reusable_pool = true;
+    bo->u.real.use_reusable_pool = use_reusable_pool;
 
     mtx_lock(&ws->bo_handles_mutex);
     util_hash_table_set(ws->bo_handles, (void*)(uintptr_t)bo->handle, bo);