hasvk: Validate profiles and harden the video session objects.

Assisted-by: Claude

src/intel/vulkan_hasvk/anv_private.h

@@ -87,8 +87,8 @@
 #include "vk_util.h"
 #include "vk_queue.h"
 #include "vk_log.h"
-#include "vk_ycbcr_conversion.h"
 #include "vk_video.h"
+#include "vk_ycbcr_conversion.h"
 
 /* Pre-declarations needed for WSI entrypoints */
 struct wl_surface;
@@ -1071,6 +1071,7 @@ struct anv_device {
     pthread_mutex_t                             mutex;
 
     struct intel_batch_decode_ctx               decoder[ANV_MAX_QUEUE_FAMILIES];
+
     /*
      * When decoding a anv_cmd_buffer, we might need to search for BOs through
      * the cmd_buffer's list.
@@ -2638,8 +2639,8 @@ struct anv_cmd_buffer {
    struct u_trace                               trace;
 
    struct {
-      struct anv_video_session *vid;
+      struct anv_video_session                  *vid;
-      struct anv_video_session_params *params;
+      struct anv_video_session_params           *params;
    } video;
 };
 
@@ -3700,22 +3701,15 @@ struct anv_query_pool {
    struct intel_perf_query_info                 **pass_query;
 };
 
-static inline uint32_t khr_perf_query_preamble_offset(const struct anv_query_pool *pool,
-                                                      uint32_t pass)
-{
-   return pool->pass_size * pass + 8;
-}
-
-struct anv_vid_mem {
-   struct anv_device_memory *mem;
-   VkDeviceSize       offset;
-   VkDeviceSize       size;
-};
-
-#define ANV_VIDEO_MEM_REQS_H264 4
 #define ANV_MB_WIDTH 16
 #define ANV_MB_HEIGHT 16
 
+struct anv_vid_mem {
+   struct anv_device_memory                     *mem;
+   VkDeviceSize                                 offset;
+   VkDeviceSize                                 size;
+};
+
 enum {
    ANV_VID_MEM_H264_INTRA_ROW_STORE,
    ANV_VID_MEM_H264_DEBLOCK_FILTER_ROW_STORE,
@@ -3725,16 +3719,22 @@ enum {
 };
 
 struct anv_video_session {
-   struct vk_video_session vk;
+   struct vk_video_session                      vk;
 
    /* the decoder needs some private memory allocations */
-   struct anv_vid_mem vid_mem[ANV_VID_MEM_H264_MAX];
+   struct anv_vid_mem                           vid_mem[ANV_VID_MEM_H264_MAX];
 };
 
 struct anv_video_session_params {
-   struct vk_video_session_parameters vk;
+   struct vk_video_session_parameters           vk;
 };
 
+static inline uint32_t khr_perf_query_preamble_offset(const struct anv_query_pool *pool,
+                                                      uint32_t pass)
+{
+   return pool->pass_size * pass + 8;
+}
+
 void
 anv_dump_pipe_bits(enum anv_pipe_bits bits);
 
@@ -3864,8 +3864,11 @@ VK_DEFINE_NONDISP_HANDLE_CASTS(anv_sampler, base, VkSampler,
 VK_DEFINE_NONDISP_HANDLE_CASTS(anv_performance_configuration_intel, base,
                                VkPerformanceConfigurationINTEL,
                                VK_OBJECT_TYPE_PERFORMANCE_CONFIGURATION_INTEL)
-VK_DEFINE_NONDISP_HANDLE_CASTS(anv_video_session, vk.base, VkVideoSessionKHR, VK_OBJECT_TYPE_VIDEO_SESSION_KHR)
+VK_DEFINE_NONDISP_HANDLE_CASTS(anv_video_session, vk.base, VkVideoSessionKHR,
-VK_DEFINE_NONDISP_HANDLE_CASTS(anv_video_session_params, vk.base, VkVideoSessionParametersKHR, VK_OBJECT_TYPE_VIDEO_SESSION_PARAMETERS_KHR)
+                               VK_OBJECT_TYPE_VIDEO_SESSION_KHR)
+VK_DEFINE_NONDISP_HANDLE_CASTS(anv_video_session_params, vk.base,
+                               VkVideoSessionParametersKHR,
+                               VK_OBJECT_TYPE_VIDEO_SESSION_PARAMETERS_KHR)
 
 #define anv_genX(devinfo, thing) ({             \
    __typeof(&gfx7_##thing) genX_thing;          \

src/intel/vulkan_hasvk/anv_video.c

@@ -23,18 +23,23 @@
 
 #include "anv_private.h"
 
+#include <stdlib.h>
+#include <string.h>
+#include "vk_common_entrypoints.h"
 #include "vk_video/vulkan_video_codecs_common.h"
+#include "vk_video/vulkan_video_codec_h264std.h"
 
 VkResult
 anv_CreateVideoSessionKHR(VkDevice _device,
-                           const VkVideoSessionCreateInfoKHR *pCreateInfo,
+                          const VkVideoSessionCreateInfoKHR *pCreateInfo,
-                           const VkAllocationCallbacks *pAllocator,
+                          const VkAllocationCallbacks *pAllocator,
-                           VkVideoSessionKHR *pVideoSession)
+                          VkVideoSessionKHR *pVideoSession)
 {
    ANV_FROM_HANDLE(anv_device, device, _device);
 
    struct anv_video_session *vid =
-      vk_alloc2(&device->vk.alloc, pAllocator, sizeof(*vid), 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+      vk_alloc2(&device->vk.alloc, pAllocator, sizeof(*vid), 8,
+                VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
    if (!vid)
       return vk_error(device, VK_ERROR_OUT_OF_HOST_MEMORY);
 
@@ -62,7 +67,10 @@ anv_DestroyVideoSessionKHR(VkDevice _device,
    if (!_session)
       return;
 
-   vk_object_base_finish(&vid->vk.base);
+   /* Otherwise we hit VK_ERROR_DEVICE_LOST for unknown reasons. */
+   vk_common_DeviceWaitIdle(_device);
+
+   vk_video_session_finish(&vid->vk);
    vk_free2(&device->vk.alloc, pAllocator, vid);
 }
 
@@ -101,9 +109,28 @@ anv_DestroyVideoSessionParametersKHR(VkDevice _device,
 
 VkResult
 anv_GetPhysicalDeviceVideoCapabilitiesKHR(VkPhysicalDevice physicalDevice,
-                                           const VkVideoProfileInfoKHR *pVideoProfile,
+                                          const VkVideoProfileInfoKHR
-                                           VkVideoCapabilitiesKHR *pCapabilities)
+                                          *pVideoProfile,
+                                          VkVideoCapabilitiesKHR
+                                          *pCapabilities)
 {
+   if (pVideoProfile->videoCodecOperation != VK_VIDEO_CODEC_OPERATION_DECODE_H264_BIT_KHR) {
+      return VK_ERROR_VIDEO_PROFILE_OPERATION_NOT_SUPPORTED_KHR;
+   }
+
+   const VkVideoDecodeH264ProfileInfoKHR *h264_profile =
+      vk_find_struct_const(pVideoProfile->pNext, VIDEO_DECODE_H264_PROFILE_INFO_KHR);
+   if (h264_profile) {
+      switch (h264_profile->stdProfileIdc) {
+      case STD_VIDEO_H264_PROFILE_IDC_BASELINE:
+      case STD_VIDEO_H264_PROFILE_IDC_MAIN:
+      case STD_VIDEO_H264_PROFILE_IDC_HIGH:
+         break;
+      default:
+         return VK_ERROR_VIDEO_PROFILE_CODEC_NOT_SUPPORTED_KHR;
+      }
+   }
+
    pCapabilities->minBitstreamBufferOffsetAlignment = 32;
    pCapabilities->minBitstreamBufferSizeAlignment = 32;
    pCapabilities->pictureAccessGranularity.width = ANV_MB_WIDTH;
@@ -112,105 +139,150 @@ anv_GetPhysicalDeviceVideoCapabilitiesKHR(VkPhysicalDevice physicalDevice,
    pCapabilities->minCodedExtent.height = ANV_MB_HEIGHT;
    pCapabilities->maxCodedExtent.width = 4096;
    pCapabilities->maxCodedExtent.height = 4096;
-   pCapabilities->flags = VK_VIDEO_CAPABILITY_SEPARATE_REFERENCE_IMAGES_BIT_KHR;
+   pCapabilities->flags =
+      VK_VIDEO_CAPABILITY_SEPARATE_REFERENCE_IMAGES_BIT_KHR;
 
-   struct VkVideoDecodeCapabilitiesKHR *dec_caps = (struct VkVideoDecodeCapabilitiesKHR *)
+   struct VkVideoDecodeCapabilitiesKHR *dec_caps =
+      (struct VkVideoDecodeCapabilitiesKHR *)
       vk_find_struct(pCapabilities->pNext, VIDEO_DECODE_CAPABILITIES_KHR);
    if (dec_caps)
-      dec_caps->flags = VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_COINCIDE_BIT_KHR;
+      dec_caps->flags =
+         VK_VIDEO_DECODE_CAPABILITY_DPB_AND_OUTPUT_COINCIDE_BIT_KHR;
 
-   switch (pVideoProfile->videoCodecOperation) {
+   struct VkVideoDecodeH264CapabilitiesKHR *ext =
-   case VK_VIDEO_CODEC_OPERATION_DECODE_H264_BIT_KHR: {
+      (struct VkVideoDecodeH264CapabilitiesKHR *)
-      struct VkVideoDecodeH264CapabilitiesKHR *ext = (struct VkVideoDecodeH264CapabilitiesKHR *)
+      vk_find_struct(pCapabilities->pNext,
-         vk_find_struct(pCapabilities->pNext, VIDEO_DECODE_H264_CAPABILITIES_KHR);
+                     VIDEO_DECODE_H264_CAPABILITIES_KHR);
-      pCapabilities->maxDpbSlots = 17;
+   pCapabilities->maxDpbSlots = 16;
-      pCapabilities->maxActiveReferencePictures = 16;
+   pCapabilities->maxActiveReferencePictures = 16;
 
+   if (ext) {
       ext->fieldOffsetGranularity.x = 0;
       ext->fieldOffsetGranularity.y = 0;
       ext->maxLevelIdc = 51;
-      strcpy(pCapabilities->stdHeaderVersion.extensionName, VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_EXTENSION_NAME);
-      pCapabilities->stdHeaderVersion.specVersion = VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_SPEC_VERSION;
-      break;
-   }
-   default:
-      break;
    }
+   strcpy(pCapabilities->stdHeaderVersion.extensionName,
+          VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_EXTENSION_NAME);
+   pCapabilities->stdHeaderVersion.specVersion =
+      VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_SPEC_VERSION;
+
    return VK_SUCCESS;
 }
 
 VkResult
 anv_GetPhysicalDeviceVideoFormatPropertiesKHR(VkPhysicalDevice physicalDevice,
-                                               const VkPhysicalDeviceVideoFormatInfoKHR *pVideoFormatInfo,
+                                              const
-                                               uint32_t *pVideoFormatPropertyCount,
+                                              VkPhysicalDeviceVideoFormatInfoKHR
-                                               VkVideoFormatPropertiesKHR *pVideoFormatProperties)
+                                              *pVideoFormatInfo,
+                                              uint32_t
+                                              *pVideoFormatPropertyCount,
+                                              VkVideoFormatPropertiesKHR
+                                              *pVideoFormatProperties)
 {
    *pVideoFormatPropertyCount = 1;
 
    if (!pVideoFormatProperties)
       return VK_SUCCESS;
 
+   VkImageUsageFlags usage_flags = pVideoFormatInfo->imageUsage;
+
+   const VkImageUsageFlags dpb_output_coincide_usage =
+      VK_IMAGE_USAGE_VIDEO_DECODE_DST_BIT_KHR |
+      VK_IMAGE_USAGE_VIDEO_DECODE_DPB_BIT_KHR;
+   if ((usage_flags & dpb_output_coincide_usage) == VK_IMAGE_USAGE_VIDEO_DECODE_DPB_BIT_KHR ||
+       (usage_flags & dpb_output_coincide_usage) == VK_IMAGE_USAGE_VIDEO_DECODE_DST_BIT_KHR)
+      usage_flags |= dpb_output_coincide_usage;
+
+   if (usage_flags & VK_IMAGE_USAGE_VIDEO_DECODE_DST_BIT_KHR)
+      usage_flags |=
+         VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_SAMPLED_BIT;
+
+   pVideoFormatProperties[0].sType =
+      VK_STRUCTURE_TYPE_VIDEO_FORMAT_PROPERTIES_KHR;
+   pVideoFormatProperties[0].pNext = NULL;
    pVideoFormatProperties[0].format = VK_FORMAT_G8_B8R8_2PLANE_420_UNORM;
+   pVideoFormatProperties[0].componentMapping.r =
+      VK_COMPONENT_SWIZZLE_IDENTITY;
+   pVideoFormatProperties[0].componentMapping.g =
+      VK_COMPONENT_SWIZZLE_IDENTITY;
+   pVideoFormatProperties[0].componentMapping.b =
+      VK_COMPONENT_SWIZZLE_IDENTITY;
+   pVideoFormatProperties[0].componentMapping.a =
+      VK_COMPONENT_SWIZZLE_IDENTITY;
+   pVideoFormatProperties[0].imageCreateFlags =
+      VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT | VK_IMAGE_CREATE_EXTENDED_USAGE_BIT
+      | VK_IMAGE_CREATE_ALIAS_BIT;
    pVideoFormatProperties[0].imageType = VK_IMAGE_TYPE_2D;
    pVideoFormatProperties[0].imageTiling = VK_IMAGE_TILING_OPTIMAL;
-   pVideoFormatProperties[0].imageUsageFlags = pVideoFormatInfo->imageUsage;
+   pVideoFormatProperties[0].imageUsageFlags = usage_flags;
+
    return VK_SUCCESS;
 }
 
 static void
 get_h264_video_session_mem_reqs(struct anv_video_session *vid,
                                 VkVideoSessionMemoryRequirementsKHR *mem_reqs,
+                                uint32_t *pVideoSessionMemoryRequirementsCount,
                                 uint32_t memory_types)
 {
-   uint32_t width_in_mb = align(vid->vk.max_coded.width, ANV_MB_WIDTH) / ANV_MB_WIDTH;
+   VK_OUTARRAY_MAKE_TYPED(VkVideoSessionMemoryRequirementsKHR,
-   /* intra row store is width in macroblocks * 64 */
+                          out,
-   mem_reqs[0].memoryBindIndex = ANV_VID_MEM_H264_INTRA_ROW_STORE;
+                          mem_reqs,
-   mem_reqs[0].memoryRequirements.size = width_in_mb * 64;
+                          pVideoSessionMemoryRequirementsCount);
-   mem_reqs[0].memoryRequirements.alignment = 4096;
-   mem_reqs[0].memoryRequirements.memoryTypeBits = memory_types;
 
-   /* deblocking filter row store is width in macroblocks * 64 * 4*/
+   uint32_t width_in_mb =
-   mem_reqs[1].memoryBindIndex = ANV_VID_MEM_H264_DEBLOCK_FILTER_ROW_STORE;
+      align(vid->vk.max_coded.width, ANV_MB_WIDTH) / ANV_MB_WIDTH;
-   mem_reqs[1].memoryRequirements.size = width_in_mb * 64 * 4;
+
-   mem_reqs[1].memoryRequirements.alignment = 4096;
+   /* intra row store is width in macroblocks * 64 */
-   mem_reqs[1].memoryRequirements.memoryTypeBits = memory_types;
+   vk_outarray_append_typed(VkVideoSessionMemoryRequirementsKHR, &out, p) {
+      p->memoryBindIndex = ANV_VID_MEM_H264_INTRA_ROW_STORE;
+      p->memoryRequirements.size = width_in_mb * 64;
+      p->memoryRequirements.alignment = 64;
+      p->memoryRequirements.memoryTypeBits = memory_types;
+   }
+
+   /* deblocking filter row store is width in macroblocks * 64 * 4 */
+   vk_outarray_append_typed(VkVideoSessionMemoryRequirementsKHR, &out, p) {
+      p->memoryBindIndex = ANV_VID_MEM_H264_DEBLOCK_FILTER_ROW_STORE;
+      p->memoryRequirements.size = width_in_mb * 64 * 4;
+      p->memoryRequirements.alignment = 64;
+      p->memoryRequirements.memoryTypeBits = memory_types;
+   }
 
    /* bsd mpc row scratch is width in macroblocks * 64 * 2 */
-   mem_reqs[2].memoryBindIndex = ANV_VID_MEM_H264_BSD_MPC_ROW_SCRATCH;
+   vk_outarray_append_typed(VkVideoSessionMemoryRequirementsKHR, &out, p) {
-   mem_reqs[2].memoryRequirements.size = width_in_mb * 64 * 2;
+      p->memoryBindIndex = ANV_VID_MEM_H264_BSD_MPC_ROW_SCRATCH;
-   mem_reqs[2].memoryRequirements.alignment = 4096;
+      p->memoryRequirements.size = width_in_mb * 64 * 2;
-   mem_reqs[2].memoryRequirements.memoryTypeBits = memory_types;
+      p->memoryRequirements.alignment = 64;
+      p->memoryRequirements.memoryTypeBits = memory_types;
+   }
 
    /* mpr row scratch is width in macroblocks * 64 * 2 */
-   mem_reqs[3].memoryBindIndex = ANV_VID_MEM_H264_MPR_ROW_SCRATCH;
+   vk_outarray_append_typed(VkVideoSessionMemoryRequirementsKHR, &out, p) {
-   mem_reqs[3].memoryRequirements.size = width_in_mb * 64 * 2;
+      p->memoryBindIndex = ANV_VID_MEM_H264_MPR_ROW_SCRATCH;
-   mem_reqs[3].memoryRequirements.alignment = 4096;
+      p->memoryRequirements.size = width_in_mb * 64 * 2;
-   mem_reqs[3].memoryRequirements.memoryTypeBits = memory_types;
+      p->memoryRequirements.alignment = 64;
+      p->memoryRequirements.memoryTypeBits = memory_types;
+   }
 }
 
 VkResult
 anv_GetVideoSessionMemoryRequirementsKHR(VkDevice _device,
                                          VkVideoSessionKHR videoSession,
-                                         uint32_t *pVideoSessionMemoryRequirementsCount,
+                                         uint32_t
-                                         VkVideoSessionMemoryRequirementsKHR *mem_reqs)
+                                         *pVideoSessionMemoryRequirementsCount,
+                                         VkVideoSessionMemoryRequirementsKHR
+                                         *mem_reqs)
 {
    ANV_FROM_HANDLE(anv_device, device, _device);
    ANV_FROM_HANDLE(anv_video_session, vid, videoSession);
 
-   switch (vid->vk.op) {
-   case VK_VIDEO_CODEC_OPERATION_DECODE_H264_BIT_KHR:
-      *pVideoSessionMemoryRequirementsCount = ANV_VIDEO_MEM_REQS_H264;
-      break;
-   default:
-      UNREACHABLE("unknown codec");
-   }
-   if (!mem_reqs)
-      return VK_SUCCESS;
-
    uint32_t memory_types = (1ull << device->physical->memory.type_count) - 1;
    switch (vid->vk.op) {
    case VK_VIDEO_CODEC_OPERATION_DECODE_H264_BIT_KHR:
-      get_h264_video_session_mem_reqs(vid, mem_reqs, memory_types);
+      get_h264_video_session_mem_reqs(vid, mem_reqs,
+                                      pVideoSessionMemoryRequirementsCount,
+                                      memory_types);
       break;
    default:
       UNREACHABLE("unknown codec");
@@ -221,16 +293,17 @@ anv_GetVideoSessionMemoryRequirementsKHR(VkDevice _device,
 
 VkResult
 anv_UpdateVideoSessionParametersKHR(VkDevice _device,
-                                     VkVideoSessionParametersKHR _params,
+                                    VkVideoSessionParametersKHR _params,
-                                     const VkVideoSessionParametersUpdateInfoKHR *pUpdateInfo)
+                                    const
+                                    VkVideoSessionParametersUpdateInfoKHR
+                                    *pUpdateInfo)
 {
    ANV_FROM_HANDLE(anv_video_session_params, params, _params);
    return vk_video_session_parameters_update(&params->vk, pUpdateInfo);
 }
 
 static void
-copy_bind(struct anv_vid_mem *dst,
+copy_bind(struct anv_vid_mem *dst, const VkBindVideoSessionMemoryInfoKHR *src)
-          const VkBindVideoSessionMemoryInfoKHR *src)
 {
    dst->mem = anv_device_memory_from_handle(src->memory);
    dst->offset = src->memoryOffset;
@@ -245,15 +318,19 @@ anv_BindVideoSessionMemoryKHR(VkDevice _device,
 {
    ANV_FROM_HANDLE(anv_video_session, vid, videoSession);
 
-   assert(bind_mem_count == 4);
    switch (vid->vk.op) {
    case VK_VIDEO_CODEC_OPERATION_DECODE_H264_BIT_KHR:
+      /* Validate bind_mem_count and memoryBindIndex values to prevent
+       * out-of-bounds access to vid->vid_mem array. */
       for (unsigned i = 0; i < bind_mem_count; i++) {
+         if (bind_mem[i].memoryBindIndex >= ANV_VID_MEM_H264_MAX) {
+            continue;
+         }
          copy_bind(&vid->vid_mem[bind_mem[i].memoryBindIndex], &bind_mem[i]);
       }
       break;
    default:
-      UNREACHABLE("unknown codec");
+      UNREACHABLE("unknown codec!");
    }
    return VK_SUCCESS;
 }