https://bugs.winehq.org/show_bug.cgi?id=42462
Bug ID: 42462 Summary: Severe hang in several places in FFXIV Product: Wine-staging Version: 2.1 Hardware: x86 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown Assignee: wine-bugs@winehq.org Reporter: spryor@redhat.com CC: erich.e.hoover@wine-staging.com, michael@fds-team.de, sebastian@fds-team.de Distribution: ---
After playing FFXIV for a certain amount of time, or in certain locations (I haven't noticed a common shader that seems to trigger this) the game will experience a severe hang. When running perf top to examine what's going on, this line stands out:
Overhead Shared Object Symbol 9.22% wined3d-csmt.dll.so [.] 0x0000000000048362
It seems to vary from around 7-15% when not experiencing the issue, to around 50%. The jump in overhead accompanies an almost total lockup of the desktop, if the game manages to recover, it usually accompanies graphical glitches not present before the lockup, and will usually crash not too long after recovery if gameplay proceeds (which I would hazard a guess is due to framebuffer corruption)
Variant: Fedora 25, fully updated as of 2/16/2017 Desktop: KDE Processor: i7-7700k GPU: Nvidia GTX 1080 (with proprietary driver 378.13) WINE: wine-2.1-1.fc25.x86_64 (current from the fedora repos as of 2/16/2017)
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Distribution|--- |Fedora
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |spryor@redhat.com
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #1 from Sean Pryor spryor@redhat.com --- So this gets weirder...
I closed Chrome, and everything unfroze
There seems to be some bizarre interaction between WINE and Chrome
According to top, the system was still over 50% idle, but the nvidia GUI showed 100% GPU utilization, so it appears to be something locking up the GPU
https://bugs.winehq.org/show_bug.cgi?id=42462
winetest@luukku.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |winetest@luukku.com
--- Comment #2 from winetest@luukku.com --- Are you sure it's not chrome issue? I think it has a setting that it's open even it's closed. Try running without chrome open at all.
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #3 from Sean Pryor spryor@redhat.com --- Yeah, killing all chrome processes does cause it to unfreeze, but it only freezes the system when running FFXIV and chrome at the same time. It's kind of a pain to not be able to run both, and the overhead from perf showing wined3d-csmt.dll.so does point to it originating somewhere in WINE
Is there an easy way to turn that address back into a function name?
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #4 from Sean Pryor spryor@redhat.com --- After doing some additional testing, I was able to trigger it with no chrome processes running by enabling their HDR setting and running around a mildly populated area for a minute or two
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #5 from Sebastian Lackner sebastian@fds-team.de --- (In reply to Sean Pryor from comment #4)
After doing some additional testing, I was able to trigger it with no chrome processes running by enabling their HDR setting and running around a mildly populated area for a minute or two
It all sounds like you are running out of graphic resources. Could you please test if this issue can also be reproduced without CSMT? And is there any terminal output when the issue occurs (for example, messages related to running out of adapter memory or OpenGL errors)?
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #6 from Sean Pryor spryor@redhat.com --- Ok, after doing that, I did notice what is probably the problem. The slowdown does seem to be caused by memory pressure, but not where I had thought it would be. GPU memory seems to be at about 50% utilization
I ran top, and noticed that it quickly ate up all the system memory and swap with HDR enabled (32Gb ram and 16Gb swap)
Running with the default debug settings from playonlinux also revealed some messages that are probably important (attached is a snippet)
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #7 from Sean Pryor spryor@redhat.com --- Created attachment 57347 --> https://bugs.winehq.org/attachment.cgi?id=57347 Snippet from playonlinux.log
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #8 from Sebastian Lackner sebastian@fds-team.de --- (In reply to Sean Pryor from comment #6)
I ran top, and noticed that it quickly ate up all the system memory and swap with HDR enabled (32Gb ram and 16Gb swap)
The log confirms that system memory is the problem, but unfortunately doesn't give any reason. Could you please check if the issue is related to CSMT, or if it also occurs without using experimental features. You can temporarily disable it in the winecfg Staging tab.
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #9 from Sean Pryor spryor@redhat.com --- Oh, sorry yes, that last test I performed where it ate all of the system memory was with CSMT disabled
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Hardware|x86 |x86-64 Component|-unknown |-unknown Product|Wine-staging |Wine
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #10 from Sean Pryor spryor@redhat.com --- An additional note, I was able to reproduce it without CSMT when toggling occlusion culling (it gives that message about wrong thread in the logs). Toggling it on then off again causes it to almost immediately eat the entirety of ram. I'm doubling back on the other tests to make sure it is the same problem.
Also, it's worth mentioning, the graphical glitches in the game without CSMT are legion
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #11 from Sean Pryor spryor@redhat.com --- Ok, confirmed it's all the same issue
So far here are the cases tested: + CSMT + HDR + occlusion culling = super fast memory leak
+ CSMT + HDR - occlusion culling = slower memory leak (and I got this setup to crash with a segfault a fair bit)
+ CSMT - HDR - occlusion culling = (original problem) This one took a lot longer to occur, and I had to go between several heavily populated areas to trigger it. I eventually found a place I could spin my camera next to a wall and trigger it quickly. After double checking, this one definitely took the longest to occur, but still did happen (For anyone else testing, go to Quarrymill, South Shroud, stand near the bottom of the stairs next to the Levemete, over towards the signpost, and pan your camera against the building the Levemete is standing in front of)
- CSMT + HDR + occlusion culling = memory leak, and message about wrong thread in the log (probably no surprises that this one causes problems)
- CSMT + HDR - occlusion culling = slower memory leak, though it didn't crash
- CSMT - HDR + occlusion culling = slower than with HDR, but was still able to trigger it
- CSMT - HDR - occlusion culling = slower than with HDR, but was still able to trigger it
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Summary|Severe hang in several |Memory leak in several |places in FFXIV |places in FFXIV
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED
--- Comment #12 from Sean Pryor spryor@redhat.com --- As of 2.4, I am no longer able to reproduce this bug
https://bugs.winehq.org/show_bug.cgi?id=42462
Alexandre Julliard julliard@winehq.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #13 from Alexandre Julliard julliard@winehq.org --- Closing bugs fixed in 2.7.
https://bugs.winehq.org/show_bug.cgi?id=42462
Les De Ridder winehq@lesderid.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |winehq@lesderid.net
--- Comment #14 from Les De Ridder winehq@lesderid.net --- I'm still experiencing this issue in 2.5 (with CSMT).
The average time between memory leaks seems to be dependent on both the GPU model and the graphics settings. For my friend, it will almost always start leaking aggressively in heavily populated areas. For me, the leaks occur seemingly randomly, not necessarily in heavily populated areas.
The game still has graphical glitches before the hang or when it can recover from the hang (e.g. flickering or invisible UI, missing textures).
Could this be reopened, or should I do some more testing on 2.6/2.7 (with/without CSMT) first?
https://bugs.winehq.org/show_bug.cgi?id=42462
fate junk@daijobu-fansub.fr changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |junk@daijobu-fansub.fr
--- Comment #15 from fate junk@daijobu-fansub.fr --- I tested with wine 2.7 and the bug is still present. I tested with/without HDR and with/without occlusion culling. In all configuration I managed to reproduce this bug at the same place than Sean Pryor. I just spin the camera and the memory leak begins. I have the same messages than Sean Pryor in log file (err:d3d:resource_init Failed to allocate system memory). I only tested without CSMT
Variant: Archlinux Desktop: KDE Processor: i5-4570 GPU: Nvidia GTX 660 (with proprietary driver 378.13) WINE: wine-2.7
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Resolution|FIXED |--- Status|CLOSED |UNCONFIRMED
--- Comment #16 from Sean Pryor spryor@redhat.com --- If we're still experiencing it, probably a good idea to reopen to track it. Unfortunately I can no longer reproduce it. For the people still experiencing it, are there any other flags we can set to help the devs locate where we're leaking memory?
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #17 from Sean Pryor spryor@redhat.com --- Yep, can confirm this has started to reappear. It's a lot slower than it used to be, but I can watch the memory usage slowly creep up as I play
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #18 from Les De Ridder winehq@lesderid.net --- (In reply to Sean Pryor from comment #17)
Yep, can confirm this has started to reappear. It's a lot slower than it used to be, but I can watch the memory usage slowly creep up as I play
Are you sure it was resolved in an earlier version? If so, maybe we can figure out which commit broke it again (if it wasn't the game itself).
Do you remember a version (and settings) where the bug was not present?
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #19 from Sean Pryor spryor@redhat.com --- I recall not being able to reproduce it with the same methods in 2.4, I'll attempt to test that and provide feedback, though I have heard others suffer from it on that release as well
https://bugs.winehq.org/show_bug.cgi?id=42462
Sean Pryor spryor@redhat.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|2.1 |2.10
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #20 from Sean Pryor spryor@redhat.com --- Ok, I was indeed able to reproduce it on 2.4, so it looks like my original report of it being fixed in that release is incorrect
That being said, there have been improvements, I go much longer between memory leaks now and additionally I don't hit the problems with black textures when a crash is imminent
So, what should I do next to try to track down the source of the leaks? I can probably leave it on overnight in a big city and see if it's killed my system when I return since running it under valgrind or something would make it nigh unplayable
What options do I have in terms of helping track it down
https://bugs.winehq.org/show_bug.cgi?id=42462
joaopa jeremielapuree@yahoo.fr changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jeremielapuree@yahoo.fr
--- Comment #21 from joaopa jeremielapuree@yahoo.fr --- Does the bug still occur with current wine(4.0-rc1)?
https://bugs.winehq.org/show_bug.cgi?id=42462
--- Comment #22 from joaopa jeremielapuree@yahoo.fr --- No news from the OP since 3 year. No download available. Can an administrator close this bug as ABANDONED?
https://bugs.winehq.org/show_bug.cgi?id=42462
Olivier F. R. Dierick o.dierick@piezo-forte.be changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |o.dierick@piezo-forte.be
--- Comment #23 from Olivier F. R. Dierick o.dierick@piezo-forte.be --- (In reply to joaopa from comment #22)
No news from the OP since 3 year. No download available. Can an administrator close this bug as ABANDONED?
Hello,
What about the free trial?
https://freetrial.finalfantasyxiv.com/gb/
https://store.steampowered.com/app/312060/FINAL_FANTASY_XIV_Online_Free_Tria...
Regards.
https://bugs.winehq.org/show_bug.cgi?id=42462
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution|--- |ABANDONED
--- Comment #24 from Austin English austinenglish@gmail.com --- (In reply to Olivier F. R. Dierick from comment #23)
(In reply to joaopa from comment #22)
No news from the OP since 3 year. No download available. Can an administrator close this bug as ABANDONED?
Hello,
What about the free trial?
https://freetrial.finalfantasyxiv.com/gb/
https://store.steampowered.com/app/312060/ FINAL_FANTASY_XIV_Online_Free_Trial/
Regards.
Well, it's been two years since then and no one has tried, so I think it's fair to close.
If someone wants to test, feel free to reopen/open a new issue.
https://bugs.winehq.org/show_bug.cgi?id=42462
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #25 from Austin English austinenglish@gmail.com --- Closing.