http://bugs.winehq.org/show_bug.cgi?id=25015
Summary: Pando downloader hangs Product: Wine Version: 1.3.6 Platform: x86 OS/Version: Linux Status: NEW Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: dank@kegel.com
Running the downloader for a game that uses the Pando peer to peer system, I noticed the UI had hang, the app was still using lots of CPU, the downloaded file didn't seem to be growing (it was at 1GB out of 4GB), and the console showed
err:winediag:FILE_CreateFile Too many open files, ulimit -n probably needs to be increased
lsof showed that PMB.exe has two listening sockets, 25 ESTABLISHED sockets, 96 pipes, one UDP socket, 7 UNIX sockets, 97 regular files, 3 directories, and one device, for a grand total of 233. 233 is way below ulimit -n (1024). Perhaps the app had closed lots of peer connections in the hour after the warning? Which makes it sound like there isn't a bad leak. No idea what it's doing with all those pipes, though.
It's possible Pando just needs a higher ulimit. Or perhaps it can operate fine with this "low" ulimit, and the hang is not related to running out of fds.
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #1 from Dan Kegel dank@kegel.com 2010-11-03 04:54:15 CDT --- I killed and restarted it, and it happily continued; right from the start, it used 96 pipes (all in one process!), 75 to 93 established tcp sessions (varying rapidly), 57 UDP sockets (connected to lots of different peers, plus three nonconnected), and 116 other fds, mostly local files.
So it was definitely happier after restart.
http://bugs.winehq.org/show_bug.cgi?id=25015
Dan Kegel dank@kegel.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |Installer
--- Comment #2 from Dan Kegel dank@kegel.com 2010-11-03 04:57:47 CDT --- If it turns out to really be a ulimit problem, see https://bugs.launchpad.net/ubuntu/+bug/663090 for a request to raise the default ulimit on ubuntu (and why it's hard), and go vote for http://brainstorm.ubuntu.com/idea/26178/
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #3 from Alexandre Julliard julliard@winehq.org 2010-11-03 09:53:06 CDT --- It's the wineserver that's running out of fds, not the client. No evidence of Wine bug though.(In reply to comment #2)
If it turns out to really be a ulimit problem, see https://bugs.launchpad.net/ubuntu/+bug/663090 for a request to raise the default ulimit on ubuntu (and why it's hard),
It shouldn't be hard to raise the hard limit.
And note that it's the wineserver that's running out of fds, so checking the handles open in the client is not very meaningful.
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #4 from Dan Kegel dank@kegel.com 2010-11-03 16:51:43 CDT --- Raising the hard limit probably requires purging the ubuntu codebase of all unsafe uses of select().
Some wrapper that raised the hard limit just for wineserver, now, that'd be easier.
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #5 from Alexandre Julliard julliard@winehq.org 2010-11-04 04:28:33 CDT --- (In reply to comment #4)
Raising the hard limit probably requires purging the ubuntu codebase of all unsafe uses of select().
No, apps would still be limited by the soft limit unless they explicitly raise it (like Wine does).
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #6 from Dan Kegel dank@kegel.com 2010-11-05 05:11:32 CDT --- That's the first time I've see soft limits be useful, I'd forgotten all about them. Thanks! Patch sent, http://www.winehq.org/pipermail/wine-patches/2010-November/095464.html
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #7 from Dan Kegel dank@kegel.com 2011-02-10 23:53:32 CST --- Patch to wine not needed (although this is arguably still a wine problem since wine doubled the number of fd's it uses a while ago).
See http://wiki.winehq.org/FAQ#head-80aeef30021fc0bae9ba2254f9ce6e249577ea93 for workaround until the distros raise the hard ulimit by default.
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #8 from Dan Kegel dank@kegel.com 2011-06-17 10:22:04 CDT --- Aw, peachy. I'm on the new ubuntu with the hard limit of 4096, and an overnight run of lotro's pmb-based installer still hangs, this time after downloading 2.3GB out of 10 or so. The app still redraws properly, and says "4 hours 12 minutes remaining", but does not respond when you press on the 'stop' or 'pause' buttons. The log shows
err:winediag:FILE_CreateFile Too many open files, ulimit -n probably needs to be increased err:seh:setup_exception_record stack overflow 924 bytes in thread 0009 eip 7bc3d2ef esp 00a70f94 stack 0xa70000-0xa71000-0xc70000
Several hours after the hang started, "lsof | grep wineser | wc -l" shows 4030. "lsof | grep PMB | wc -l" shows 229.
Killing, then running the app again lets it resume.
You can watch the action with the command watch "lsof -n > lsof.txt && grep PMB < lsof.txt | wc -l && grep lotrohigh < lsof.txt | wc -l && grep wineser < lsof.txt | wc -l"
In a few minutes, I watched this climb to 368 1794 1821
More investigation to follow...
http://bugs.winehq.org/show_bug.cgi?id=25015
Pierre Etchemaite pe-winehq@concept-micro.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |pe-winehq@concept-micro.com
--- Comment #9 from Pierre Etchemaite pe-winehq@concept-micro.com 2013-03-15 18:08:59 CDT --- Problem also experienced with Pando Media Booster with Wine 1.5.24: - the PMB.exe process opens new sockets rapidly, but the number of opened fds at any moment is ok (I often see between 150 and 200 open descriptors) - on the other hand the number of descriptors opened by wineserver.exe slowly increases, on average, by 1 per second. It's not strictly monotonous, but it very obviously increases
Running lsof on wineserver's PID, the majority of the list is made of
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME ... wineserve 30654 petchema 1677u IPv4 12564509 0t0 TCP localhost:50467->localhost:60524 (CLOSE_WAIT) wineserve 30654 petchema 1678u IPv4 12564511 0t0 TCP localhost:52319->localhost:60524 (CLOSE_WAIT) wineserve 30654 petchema 1679u IPv4 12564513 0t0 TCP localhost:36610->localhost:60524 (CLOSE_WAIT) wineserve 30654 petchema 1680u IPv4 12563918 0t0 TCP localhost:55995->localhost:60524 (CLOSE_WAIT) ... etc. lines
http://bugs.winehq.org/show_bug.cgi?id=25015
lizhenbo litimetal@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |litimetal@gmail.com
--- Comment #10 from lizhenbo litimetal@gmail.com 2013-06-23 07:51:54 CDT --- I faced a similar problem in Wine 1.6-rc3 But I can't make sure that if it's the same bug. Could you please offer your terminal output? Thanks.
http://bugs.winehq.org/show_bug.cgi?id=25015
--- Comment #11 from Pierre Etchemaite pe-winehq@concept-micro.com 2013-08-16 18:34:24 CDT --- ... fixme:ieframe:handle_navigation_error Navigate to error page fixme:mshtml:nsChannel_IsNoCacheResponse (0xcf5bb58)->(0x10ef29f) err:winediag:FILE_CreateFile Too many open files, ulimit -n probably needs to be increased err:seh:raise_exception Unhandled exception code c0000005 flags 0 addr 0x7bc87c78
http://bugs.winehq.org/show_bug.cgi?id=25015
Jarkko K jarkko_korpi@hotmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jarkko_korpi@hotmail.com
--- Comment #12 from Jarkko K jarkko_korpi@hotmail.com --- Seems to be duplicate of
http://bugs.winehq.org/show_bug.cgi?id=22152
http://bugs.winehq.org/show_bug.cgi?id=25015
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |focht@gmx.net Summary|Pando downloader hangs |Pando Media Booster (PMB) | |downloader hangs after | |certain time, hitting Linux | |file descriptor ulimit | |(lotro)
--- Comment #13 from Anastasius Focht focht@gmx.net --- Hello Jarkko,
no, it's a different issue.
You guys could check with Windows Sysinternals 'Process Explorer' on Windows (handle count) if the behaviour of excessive socket/file object usage is the same there.
Anyway, PMB is no longer relevant because many publishers dropped/removed this piece of garbage. It's now considered malware hence any further investigation is a waste of time.
Regards
http://bugs.winehq.org/show_bug.cgi?id=25015
Brandon Corujo haku08879@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |haku08879@gmail.com
https://bugs.winehq.org/show_bug.cgi?id=25015
Anastasius Focht focht@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |ABANDONED
--- Comment #14 from Anastasius Focht focht@gmx.net --- Hello folks,
resolving 'ABANDONED' here. Reopen if you can still find a game installer that uses PMB technology and is able to download payloads from CDNs.
https://en.wikipedia.org/wiki/Pando_Media_Booster
--- quote --- These conflicts ceased to be reported in 2013, when Pando shut down its servers and ceased business. But after 2014, when the Pando Media Browser was hijacked, unsuspecting persons who installed a prompted update instead had their internet browsers hijacked and the "Sweet Page" browser virus installed. --- quote ---
Regards
https://bugs.winehq.org/show_bug.cgi?id=25015
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #15 from Austin English austinenglish@gmail.com --- Closing.
-
wine-bugs＠winehq.org -
WineHQ Bugzilla