drm/nvc0/fb: fix crash when different mutex is used to protect same list
[~shefty/rdma-dev.git] / drivers / gpu / drm / nouveau / core / subdev / fb / nvc0.c
index 306bdf121452e6e9da8d6ddeb7ba6210ddeb0bde..7606ed15b6faf55ea531e6637cc86976b24012c2 100644 (file)
@@ -145,14 +145,14 @@ nvc0_fb_vram_new(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin,
        mem->memtype = type;
        mem->size = size;
 
-       mutex_lock(&mm->mutex);
+       mutex_lock(&pfb->base.mutex);
        do {
                if (back)
                        ret = nouveau_mm_tail(mm, 1, size, ncmin, align, &r);
                else
                        ret = nouveau_mm_head(mm, 1, size, ncmin, align, &r);
                if (ret) {
-                       mutex_unlock(&mm->mutex);
+                       mutex_unlock(&pfb->base.mutex);
                        pfb->ram.put(pfb, &mem);
                        return ret;
                }
@@ -160,7 +160,7 @@ nvc0_fb_vram_new(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin,
                list_add_tail(&r->rl_entry, &mem->regions);
                size -= r->length;
        } while (size);
-       mutex_unlock(&mm->mutex);
+       mutex_unlock(&pfb->base.mutex);
 
        r = list_first_entry(&mem->regions, struct nouveau_mm_node, rl_entry);
        mem->offset = (u64)r->offset << 12;