1973591 - Crash in [@ wgpu_core::storage::Storage<T>::remove]

Status: RESOLVED FIXED

(Core :: Graphics: WebGPU, defect)

Description

[Erich Gubler [:ErichDonGubler] (he/him)]

Crash report: https://crash-stats.mozilla.org/report/index/c6466e41-6635-4309-90c5-b38f10250622

MOZ_CRASH Reason:

assertion `left == right` failed
  left: 1
 right: 3

Top relevant frames:

10 	wgpu_core::device::global::<impl wgpu_core::global::Global>::buffer_drop 	third_party/rust/wgpu-core/src/device/global.rs:285 	cfi
11 	mozilla::webgpu::WebGPUParent::RecvSwapChainDrop(mozilla::layers::RemoteTextureOwnerId const&, unsigned int, unsigned long long) 	dom/webgpu/ipc/WebGPUParent.cpp:1632 	cfi
12 	mozilla::webgpu::ReadbackPresentCallback(unsigned char*, mozilla::webgpu::ffi::WGPUBufferMapAsyncStatus)::$_0::operator()() const 	dom/webgpu/ipc/WebGPUParent.cpp:1084 	inlined
12 	mozilla::ScopeExit<mozilla::webgpu::ReadbackPresentCallback(unsigned char*, mozilla::webgpu::ffi::WGPUBufferMapAsyncStatus)::$_0>::~ScopeExit() 	mfbt/ScopeExit.h:106 	inlined
12 	mozilla::ScopeExit<mozilla::webgpu::ReadbackPresentCallback(unsigned char*, mozilla::webgpu::ffi::WGPUBufferMapAsyncStatus)::$_0>::~ScopeExit() 	mfbt/ScopeExit.h:104 	inlined
12 	mozilla::webgpu::ReadbackPresentCallback(unsigned char*, mozilla::webgpu::ffi::WGPUBufferMapAsyncStatus) 	dom/webgpu/ipc/WebGPUParent.cpp:1166 	cfi
13 	<alloc::boxed::Box<F, A> as core::ops::function::FnOnce<Args>>::call_once 	/builds/worker/fetches/rustc/lib/rustlib/src/rust/library/alloc/src/boxed.rs:1976 	inlined
13 	wgpu_core::device::UserClosures::fire 	third_party/rust/wgpu-core/src/device/mod.rs:182 	cfi
14 	wgpu_core::device::global::<impl wgpu_core::global::Global>::poll_all_devices 	third_party/rust/wgpu-core/src/device/global.rs:2048 	cfi
15 	wgpu_server_poll_all_devices 	gfx/wgpu_bindings/src/server.rs:187 	cfi

Comment 1

[Teodor Tanasoaia [:teoxoy]]

There were also a few instances of "Cannot remove a vacant resource" in command_encoder_drop, these have been fixed by Bug 1981225.

The remaining are epoch mismatches in ReadbackPresentCallback which I think is a regression from D252687. SwapChainDrop might not immediately drop the readback buffers in the parent actor, meanwhile the child actor will try to reuse the IDs.

Comment 2

[Teodor Tanasoaia [:teoxoy]]

Attachment: Bug 1973591 - Avoid freeing swapchain buffer IDs too early. r=#webgpu-reviewers

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/ec8a6a54a047
https://hg.mozilla.org/integration/autoland/rev/acca32fb0d6b
Avoid freeing swapchain buffer IDs too early. r=webgpu-reviewers,aleiserson

Comment 4

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/acca32fb0d6b