portal2
portal2 copied to clipboard
ValveSoftware
segfault in new game
Similarly to #281, after the second awakening, the sound freezes and loops. It happens before Wheatly goes into the ceiling, but can vary to seconds after waking to after he's already in the room. I am not having any issues in the first awakening or in the menu.
[ 202.878484] [drm] GPU HANG: ecode 9:0:0x84df7cfc, in portal2_linux [2918], reason: Hang on rcs0, action: reset
[ 202.878485] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[ 202.878486] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[ 202.878486] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[ 202.878487] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[ 202.878487] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[ 202.878498] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 206.879664] asynchronous wait on fence i915:gnome-shell[1836]/1:1f94 timed out
[ 210.814724] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 218.844937] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 226.843273] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 234.841755] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 408.715287] portal2_linux[2918]: segfault at 378 ip 00000000e978fece sp 00000000ff9c4110 error 4 in server.so[e8ef2000+ec4000]
(I waited a few minutes and then killed the process manually.)
I tried updating mesa as #278 suggested, but did not help (although this did fix a similar issue I was having with Portal 1). I also tried updating my kernel with no luck.
Any ideas? Thanks, great game!
Hello @chris13524, please copy your system information from steam (Steam -> Help -> System Information) and put it in a gist, then include a link to the gist in this issue report.
This is most likely the same issue that was tracked at https://github.com/ValveSoftware/Source-1-Games/issues/2418 and you'll need to update to mesa 17.2.7 or 17.3.0 or newer. An alternate workaround is to set MSAA to 2x.
Hi @kisak-valve, and thanks for the quick response. Here's my system information.
I installed the PPA here and updated and upgraded via apt. I then tried to run Portal again, but it crashed again. This in dmesg:
[ 1356.291943] usb 1-1: USB disconnect, device number 2
[ 3014.586306] SGI XFS with ACLs, security attributes, realtime, no debug enabled
[ 3014.592080] JFS: nTxBlock = 8192, nTxLock = 65536
[ 3014.599722] ntfs: driver 2.1.32 [Flags: R/O MODULE].
[ 3014.612022] QNX4 filesystem 0.2.3 registered.
[ 3014.682302] raid6: sse2x1 gen() 11577 MB/s
[ 3014.750313] raid6: sse2x1 xor() 9094 MB/s
[ 3014.818318] raid6: sse2x2 gen() 15215 MB/s
[ 3014.886324] raid6: sse2x2 xor() 10131 MB/s
[ 3014.954325] raid6: sse2x4 gen() 17304 MB/s
[ 3015.022330] raid6: sse2x4 xor() 10840 MB/s
[ 3015.090296] raid6: avx2x1 gen() 23567 MB/s
[ 3015.158336] raid6: avx2x1 xor() 14498 MB/s
[ 3015.226343] raid6: avx2x2 gen() 25317 MB/s
[ 3015.294350] raid6: avx2x2 xor() 14787 MB/s
[ 3015.362342] raid6: avx2x4 gen() 25904 MB/s
[ 3015.430353] raid6: avx2x4 xor() 18429 MB/s
[ 3015.430355] raid6: using algorithm avx2x4 gen() 25904 MB/s
[ 3015.430355] raid6: .... xor() 18429 MB/s, rmw enabled
[ 3015.430356] raid6: using avx2x2 recovery algorithm
[ 3015.432032] xor: automatically using best checksumming function avx
[ 3015.455606] Btrfs loaded, crc32c=crc32c-intel
[ 3227.918844] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3230.030290] asynchronous wait on fence i915:gnome-shell[1836]/1:1169f timed out
[ 3235.917144] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3243.915537] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3251.913978] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3255.113331] asynchronous wait on fence i915:gnome-shell[1836]/1:116a1 timed out
[ 3264.839703] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3265.008762] portal2_linux[12748]: segfault at 0 ip 00000000eaed8ad2 sp 00000000fffeee70 error 4 in client.so[ea06b000+172c000]
I then tired setting my anti-aliasing mode to 2xMSAA in Portal's options and ran through again. Again, I got the looping audio, etc. This from dmesg:
[ 3708.907710] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3716.907793] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3719.979732] asynchronous wait on fence i915:gnome-shell[1836]/1:14dc4 timed out
[ 3724.907822] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3732.907849] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3740.907817] i915 0000:00:02.0: Resetting rcs0 after gpu hang
[ 3741.019671] portal2_linux[12898]: segfault at 378 ip 00000000e994bece sp 00000000ffadecb0 error 4 in server.so[e90ae000+ec4000]
These three files have also appeared in /tmp/dumps/:
- chris13524_log.txt
- crash_20180101220522_1.dmp (base64 encoded)
- crash_20180101221146_1.dmp (base64 encoded)
Thanks for retesting with a current version of mesa. Unfortunately, my minidump reader is rejecting those minidumps after they are converted back from base64 so I can't give any insight from them. You could throw them in an archive and re-attach them if you want.
The gpu resets hint that this is a video driver issue and it could be worthwhile to bring this to the attention of the mesa devs as well if it has not been already.
Thanks for re-attaching the minidumps.
For reference, the minidumps are SIGSEGV in client.so and server.so (both part of Portal 2).
So would those be the two portal2_linux processes? One being a server, the other being the client? Through my testing, they seem to not segfault consistently (although a segfault always happens). Sometimes one segfaults, but I kill them before the other can segfault, and vise-versa. But maybe that's just race condition magic that isn't relevant to this issue...
Are you saying that this could be a Portal 2 issue and not a mesa issue? (could these be causing the video driver errors?) I'm not up to snuff on video drivers, so bear with me ;)
The way I'm reading your log, the gpu is hanging first, then after a non-trivial amount of time, Portal 2 trips over itself and dies. It's not really clear at first glance at if the issue is with Portal 2 or the video driver so it could be either.
Is there anything in addition I could provide to locate a potential issue on Portal 2's side?
Could you provide any guidance as to where to report this video driver bug?
We'll need to wait to hear from the devs if they need more to go on.
There's some info on reporting issues to Mesa at https://www.mesa3d.org/bugs.html.
Created this bug report: https://bugs.freedesktop.org/show_bug.cgi?id=104457