Valdis@alchemy.mcs.clarkson.edu (Valdis Kletnieks) (07/06/88)
Has anybody seen this, or (hopefully) have a fix? Setup: Sun3/260 with FPA, hi-res monitor running SunOS 3.5 Problem: The X11R2 Xsun server keeps sucking down more and more memory. It starts off at around 1204K, and keeps growing. Eventually, whatever malloc (or whatever) it's doing fails, taking out Xsun. At this point, the keyboard/mouse is hung big time, and requires a reboot to reclaim. The failure depends how soon it runs out of swap space, and it can have grown to anywhere between 4M and 9M in size by that point. It does this even if I am not starting up new programs. All that is usually running is 3-4 xterms, 3 xperfmons, xbiff and xclock, but the actual programs running don't seem to matter. The only other cure besides rebooting seems to be to close down X when I leave each night, and fire it up in the morning, and pray it grows slow enough to survive the workday. The next time I get bitten, I'll run adb on the dump, and maybe find what's killing it. Of course, chances are that it won't fail on the same allocate that's causing the problem (which in fact is almost for sure a failure to *DE*-allocate something)... Valdis Kletnieks Sr. Systems Programmer
RWS@ZERMATT.LCS.MIT.EDU (Robert Scheifler) (07/06/88)
Yup, we've had various people complain about various servers eating up memory. We took a stab at trying to use the MPROF facility (from Berkeley) described at the recent Usenix, to try and figure out what the problem was, but we haven't been able to make the server run correctly when MPROF is linked in. If anyone has suggestions of how to get this to work, or better tools to use, we're interested in hearing about it.
avr@mtgzz.att.com (XMRP50000[jcm]-a.v.reed) (07/06/88)
In article <1144@sun.soe.clarkson.edu>, Valdis@alchemy.mcs.clarkson.edu (Valdis Kletnieks) writes:
< Has anybody seen this, or (hopefully) have a fix?
<
< Setup: Sun3/260 with FPA, hi-res monitor running SunOS 3.5
<
< Problem: The X11R2 Xsun server keeps sucking down more and more memory.
< It starts off at around 1204K, and keeps growing. Eventually,
< whatever malloc (or whatever) it's doing fails, taking out Xsun.
< At this point, the keyboard/mouse is hung big time, and requires a
< reboot to reclaim. The failure depends how soon it runs out of swap space,
< and it can have grown to anywhere between 4M and 9M in size by that point.
<
< It does this even if I am not starting up new programs. All that is usually
< running is 3-4 xterms, 3 xperfmons, xbiff and xclock, but the actual
< programs running don't seem to matter.
<
< The only other cure besides rebooting seems to be to close down X when I
< leave each night, and fire it up in the morning, and pray it grows slow
< enough to survive the workday.
My 3/60 with a standard B&W buffer does the same thing but more slowly -
it takes several days to run out of 8 Meg. I'd be grateful for a fix.
Adam Reed (mtgzz!avr)
warner%s3snorkel@S3DAWN.ARPA (Ken Warner) (07/07/88)
I have a Sun 3/50 and have not seen this problem!? I have been running the same Xsun for almost two weeks now and the size right now is 1056K including text and 496K for data. I am running the Purdue speed-ups with all appropriate patches. Ken Warner
Valdis@alchemy.mcs.clarkson.edu (Valdis Kletnieks) (07/07/88)
More details: I installed all the patches off expo.lcs.mit.edu, plus the Purdue mods. It's still doing it. However, it seems to be doing it only when I am running an X10 application through 'x10tox11'. My guess (gurus please correct me) is that either 'x10tox11' or the main X10 program I run (xrn 3.10) is doing an XcreateWindow with some sort of event_mask specifying some event, but then doing a XCheckMaskEvent that excludes the event. End result is that all the unconsumed events pile up in the server..... Off to go wade through the code (as soon as I get dug out from under all the OTHER stuff I have to do... :-) Valdis Kletnieks P.S. The Purdue mods *do* make a big speed difference. I'm tempted to go install 'gcc' just to see how much more I can squeeze out.. :-)
turner@daisy.UUCP (D'arc Angel) (07/08/88)
We found two memory leaks in the Sun server for the 386i, one we plugged and one that we are still tracking down. I'm at home and it's pre-coffee so I don't have the details in front of me. Send me email if you want details. -- Don't you think that Mahitabel is a bit too toujour gay? - Archie ...{decwrl|ucbvax}!imagen!atari!daisy!turner (James M. Turner) Daisy Systems, 700 E. Middlefield Rd, P.O. Box 7006, Mountain View CA 94039-7006. (415)960-0123