http://bugs.winehq.org/show_bug.cgi?id=30307
Bug #: 30307 Summary: Memory Allocation severely limited when running wine executable via condor scheduler Product: Wine Version: 1.4-rc6 Platform: x86-64 OS/Version: Linux Status: UNCONFIRMED Severity: blocker Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: jbherman@gmail.com Classification: Unclassified
memory allocation problem: symptom: ~300MB allocation denied on an ubuntu linux box with 8GB running wine1.4 (also tested wine1.5).
I've created a simplified test case taking advantage of the memory test program contributed here (http://bugs.winehq.org/attachment.cgi?id=32366). I compile it via:
#i586-mingw32msvc-gcc -O0 test-mem.c -o test-mem.exe
thereby creating a windows executable that runs under wine. The test program successfully allocates memory (malloc) up to 1408 megabytes, which (while disappointing) is completely expected and about as much as I can get in 32bit windows.
THE PROBLEM: run the same executable via the condor scheduler (http://research.cs.wisc.edu/condor/) and the largest allocation is 448MB!! I have extensively investigated the issue and posted on the condor-users list, to absolutely no avail.
I need to provide my 32-bit windows program with as much memory as possible. WHAT TO DO? any thoughts/help appreciated.
thanks, jason
http://bugs.winehq.org/show_bug.cgi?id=30307
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |testcase Severity|blocker |normal
--- Comment #1 from Austin English austinenglish@gmail.com 2012-03-29 16:02:09 CDT --- Not a blocker.
http://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #2 from jbherman@gmail.com 2012-03-29 17:20:02 CDT --- the plot thickens....
I just tested the same job with a different scheduler (SGE - sun grid engine) and got the EXACT same result. maximum allocation of 448 MB!!!
so it is not specific to condor. perhaps I'll try a different flavor of linux next.
j
http://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #3 from jbherman@gmail.com 2012-03-29 20:47:37 CDT --- i should mention: the same C code compiled for linux doesn't hit any memory allocation limits whether run on the command line or via any scheduler.
So this issue occurs with ANY SCHEDULED WINE JOB.
http://bugs.winehq.org/show_bug.cgi?id=30307
Vitaliy Margolen vitaliy-bugzilla@kievinfo.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|normal |minor
--- Comment #4 from Vitaliy Margolen vitaliy-bugzilla@kievinfo.com 2012-03-29 21:32:36 CDT --- http://bugs.winehq.org/page.cgi?id=fields.html#importance
Problem seems to me external to Wine. Programs work just fine with plain Wine and standard system configuration. That's why it's minor.
http://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #5 from jbherman@gmail.com 2012-03-29 23:20:07 CDT --- after spending the greater part of a week investigating this issue, i can say most certainly it's not strictly external to wine. it's clearly an issue of wine interacting with the environment. The operating system has no limits on memory allocation and the problem in reproducible on multiple schedulers. My best guess is that some aspect of the environment influences the way wine loads libraries into memory in such a way that fragments the available space into small pieces or loads duplicate or unnecessary libraries. There is just no evidence that linux OS is actually refusing to provide memory.
this suggests debug messages regarding failed memory allocation would be invaluable, but I haven't found a way to produce or find them. ( debugging produced gigs of logs)
In my experience interaction issues can be some of the hardest to pin down.
From reading wine forums and bug dialogues it seems like this kind of issue is
more common with wine because of the nature of the beast OS+wine+app in the context of attempting to replicate windows, not just get it right. tough task!
i'm not experienced with filing bugs. it was blocking my work so i filed it as a blocker. In my opinion it's not minor because running on the scheduler is essential for my application as is ram >1GB. but whatever, major, minor i don't mind....
workaround: i think i have a way around the problem. Instead of just running the scheduled job, I schedule a script which ssh's right back to the machine and runs the job. This ssh session is apparently identical (in whatever respect is impacting wine) to the interactive shell, thus i get up to ~1.5GB ram. ugly, but easy enough and it seems to be working in early tests..
j
http://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #6 from Alexandre Julliard julliard@winehq.org 2012-03-30 03:56:44 CDT --- Most likely some library gets loaded in the middle of the address space. Check /proc/pid/maps to see the address space layout. Requiring huge contiguous blocks is not a good idea anyway.
http://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #7 from jbherman@gmail.com 2012-03-30 11:57:22 CDT --- thanks alexandre for the reply.
that is what i was suspecting. however the key here is: how could the manner in which the executable is invoked (in this case by a scheduler) affect what libraries are loaded?
regarding large blocks of memory. surprised to hear that in this 'big data' day and age. I analyze multi-GB datasets all the time, requiring they be in-memory.
i looked at the /proc/pid/maps but will try again and post.
next challenge is to provide my 32-bit program the full 2GB memory allocation it could receive in a 64-bit windows. possible?
thanks, jason
https://bugs.winehq.org/show_bug.cgi?id=30307
--- Comment #8 from Austin English austinenglish@gmail.com --- This is your friendly reminder that there has been no bug activity for 2 years. Is this still an issue in current (1.7.16 or newer) wine?
https://bugs.winehq.org/show_bug.cgi?id=30307
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution|--- |ABANDONED
--- Comment #9 from Austin English austinenglish@gmail.com --- (In reply to Austin English from comment #8)
This is your friendly reminder that there has been no bug activity for 2 years. Is this still an issue in current (1.7.16 or newer) wine?
Abandoned.
https://bugs.winehq.org/show_bug.cgi?id=30307
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #10 from Austin English austinenglish@gmail.com --- Closing.