Write after Write Hazard Reported Incorrectly when using Dynamic Rendering Local Read?

[jmoguill]

Environment:

• OS: Windows 11
• GPU and driver version: RTX 4070 TI
• SDK or header version if building from repo: 1.3.290
• Options enabled (synchronization, best practices, etc.):

Describe the Issue

Validation layer reports a write after write hazard when issuing a call to pipeline barrier, when the destination access mask is a read-only state....

The app issues a call to pipeline barrier where the dst access mask is transfer_read (a read only state), and the validation layer is reporting a write after write hazard?

imageMemoryBarrier: image: A336760000000093 oldLayout: VK_IMAGE_LAYOUT_RENDERING_LOCAL_READ_KHR newLayout: VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL srcStageMask: VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT | dstStageMask: VK_PIPELINE_STAGE_TRANSFER_BIT | srcAccessMask: VK_ACCESS_INPUT_ATTACHMENT_READ_BIT | dstAccessMask: VK_ACCESS_TRANSFER_READ_BIT |

Validation Error: [ SYNC-HAZARD-WRITE-AFTER-WRITE ] Object 0: handle = 0xa336760000000093, name = test_fr_0, type = VK_OBJECT_TYPE_IMAGE; | MessageID = 0x5c0ec5d6 | vkCmdPipelineBarrier(): Hazard WRITE_AFTER_WRITE for image barrier 0 VkImage 0xa336760000000093[test_fr_0]. Access info (usage: SYNC_IMAGE_LAYOUT_TRANSITION, prior_usage: SYNC_COLOR_ATTACHMENT_OUTPUT_COLOR_ATTACHMENT_WRITE, write_barriers: 0, command: vkCmdEndRenderingKHR, seq_no: 7, reset_no: 3).

A clear and concise description of what the bug is.

Expected behavior

Write-after-write hazard should not be reported?

A clear and concise description of what you expected to happen.

Valid Usage ID
If applicable, please include the validation messages encountered leading up to the issue

Additional context

code or terminal output

# callstacks, crashes, etc.
# EX:
Validation Error: [ VUID-vkCmdDrawMultiEXT-colorAttachmentCount-06188 ] Object 0: handle = 0x3d47e60 ...

[artem-lunarg]

I think everything related to VK_IMAGE_LAYOUT_RENDERING_LOCAL_READ_KHR is not supported by syncval. It is in the same category as #8703.

It should be enough to have single #7776 ticket to track progress on VK_IMAGE_LAYOUT_RENDERING_LOCAL_READ_KHR.
It makes sense to have a separate version of this issue for syncval, since it will require separate testing.

[jmoguill]

@artem-lunarg thanks for clarifying