Hang at logo screen in Neverness to Everness (NTE) when enabling VKD3D descriptor heap #11
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Compatibility Report
System Information
I confirm:
Symptoms
When launching Neverness to Everness (NTE / 异环) using dwproton with VKD3D's new
VK_EXT_descriptor_heapfeature enabled via environment variables, the game completely hangs at the initial logo screen.It freezes for about 5 minutes before the system/game pops up a "frozen/not responding" window prompting me to force quit the application.
Reproduction
WINENTSYNC=1(tested without this makes no difference)WINEDEBUG=+ntoskrnl,+wdfldr,+seh,+loaddllVKD3D_CONFIG=descriptor_heapJust a quick update / log highlight:
I was looking at the log right around the time the 5-minute hang starts, and it
consistently stops at these specific lines right after the descriptor heap is
created:
info: Creating resource descriptor heap (8192 kB)
...
trace:seh:sigsys_handler SIGSYS, rax 0xffff, rip 0x600000000005.
trace:seh:install_bpf Seccomp filters already installed.
This SIGSYS trap (rax 0xffff) seems to trigger both when NTEUpdate.exe and
CrashClientReporter.exe are invoked. Just leaving this specific error signature
here in case it helps narrow down if it's a vkd3d memory allocation issue or
something tripping the anti-cheat/seccomp!
This is not really a problem, just syscall emulation logging, you can ignore it.
As for the descriptor heap issue, the version in dwproton is likely kinda old and missing all the latest updates. You should try replacing the vkd3d-proton files manually, I will eventually make a release based on latest proton-cachyos after the game drops this weekend.
vkd3d-bratan from cachyos-proton-10.0-20260407,latest commit from vkd3d-proton
github.com/HansKristian-Work/vkd3d-proton@2889f1c8tested and failed same way as before
https://github.com/HansKristian-Work/vkd3d-proton/issues/2914
caused by this issue
https://github.com/HansKristian-Work/vkd3d-proton/issues/2914
https://github.com/NVIDIA/open-gpu-kernel-modules/issues/1097
as long as having VKD3D_CONFIG=no_upload_hvv,descriptor_heap
this will happen
fedora kernel: NVRM: Xid (PCI:0000:01:00): 109, pid=45891, name=HTGame.exe, channel 0x0000000f, errorString CTX SWITCH TIMEOUT, Info 0x2c012
this might due to 595.58.03 driver internal issue
i manually backport some commits from main to descriptor-heap-rebase and the compiled vkd3d with heap make it to the login screen (one on protoncachyos failed even before login screen , just after logo appears) and fail just about to show the 3d world.
https://github.com/sunnyyangyangyang/vkd3d-proton/tree/descriptor-heap-rebase
https://forums.developer.nvidia.com/t/xid109-ctx-switch-timeout-driver-crashes-in-many-applications/283722
😭
WhyAreAllTheseTaken
4月 19 日
As a last resort, I tried using the nouveau drivers. With this, Vintage Story fully freezes but I do get a somewhat more detailed error:
【 796.199239】 nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:9 gfid:0 level:2 type:109 scope:1 part:233 fault_addr:0000000000000000 fault_type:00000000
【 796.199263】 nouveau 0000:01:00.0: fifo:000000:0008:0009:[Vintagestory【5865】] errored - disabling channel
【 796.199278】 nouveau 0000:01:00.0: Vintagestory【5865】: channel 9 killed!
【 796.206853】 nouveau 0000:01:00.0: Vintagestory【5865】: error fencing pushbuf: -19
【 796.206910】 nouveau 0000:01:00.0: Vintagestory【5865】: error fencing pushbuf: -19
【 796.206934】 nouveau 0000:01:00.0: Vintagestory【5865】: error fencing pushbuf: -1
https://forums.developer.nvidia.com/t/xid109-ctx-switch-timeout-driver-crashes-in-many-applications/283722/69
This might caused by the GSP firmware 595.58.03 accidentally calculate the wrong pointer address under certain conditions, this illegal address caused the driver to reset.