From Bugzilla Helper: User-Agent: Mozilla/5.0 (compatible; Konqueror/3.5; Linux) KHTML/3.5.4 (like Gecko) Description of problem: Have been experiencing frequent crashes of Firefox (my box and others I administer). I usually always can get it to crash when visiting a Flash related site (like any under the www.adobe.com site) but even then it might take a few tries to get it to crash. Firefox will also crash when not even visiting a Flash related site and it has been so random as when I have returned to work in the morning, Firefox has crashed for no apparent reason. When running through the gdb debugger, most of the backtraces point to a flash (?) related problem (ie "Flash_EnforceLocalSecurity"), whether a visiting a Flash site or not. I am attaching a full backtrace of a crash. I have seen this behavior on Profiles with no Extensions or Themes installed and on a newly setup ("clean") Profile. Some users see crashes more sporadically then others, and of course, different individuals will have different preference settings etc. but the consistent thing is that all the machines were imaged from the same source image so the basic firefox/flash setup should be the same, and further, none of these issues were happening in RH4 Firefox and Flash (with current Profiles). Version-Release number of selected component (if applicable): firefox-1.5.0.12-7.el5 kernel-2.6.18-53.1.4.el5 flash-plugin-9.0.115.0-1.el5 How reproducible: Sometimes Steps to Reproduce: 1.run Firefox 2. visit a Flash related site or, 3. do nothing specific and eventually Firefox will crash Actual Results: Firefox crashes. Expected Results: Firefox doesn't crash. Additional info: When running directly from the command line, Firefox gives no errors. Checks of system logs and SE Linux troubleshooter shows no errors.
Created attachment 292842 [details] backtrace of Firefox crash
Created attachment 292846 [details] another backtrace of Firefox crashing Did not take long for Firefox to crash again, this time no Flash page were directly involved but was clicking around in another page.
For official Red Hat Enterprise Linux support, please log into the Red Hat support website at http://www.redhat.com/support and file a support ticket, or alternatively contact Red Hat Global Support Services at 1-888-RED-HAT1 to speak directly with a support associate and escalate an issue.
I have submitted a support request.
OK, let me know when/if your problem will be escalated to the bugzilla.
Hi, I just thought I'd pass this along. I posted some information regarding a well-known (non-flash related, as far as I can tell) firefox crash on RHEL5/5.1 in bug 244530 (comment #2). Some of the info at the links might be useful in fixing it.
Andrew, thanks for the input and the links. I did not see anything in those links to try that I haven't already tried.... and still, Firefox crashes frequently enough to make it rather irritating. At least visiting those sites confirmed that this is not an isolated problem to our environment. Jonathan
The javascript diagnosis on the first link in comment 2 of bug 244530 looks like it might be useful in isolating the problem. Hopefully it'll help. For me the problem happens rarely but, when it does, I am usually saving something from the browser (save page as, save image as, save link as etc). If it crashes it will happen as soon as the save window opens. When I restore session (firefox 2), everything works normally. Have you tried running the firefox 3 beta? I noticed on a few bug reporting sites that there are claims it is more stable under RHEL. Andrew
I rarely do "saves" from the browser so I cant speak to that issue; my (my "users") crashes have come from a variety of actions and non-actions. In fact, it just crashed on my first attempt to write this by simply clicking around in the RHN site for some info. Due to our environment and the fact we pay Redhat (for licenses, support, etc) we generally stick to the mainline RH distributed releases, such as 1.5xxx for firefox. Frankly, I would think a downlevel "stable" version should be more reliable then an uplevel beta. This is compoundly confusing when we see no problems with firefox 1.5x crashes on RHEL 4 (only RHEL 5). RH did release a critical patch for Firefox today... but since my system has already crashed since installing it, I can not be too optimistic that it has fixed the crash issues we are seeing. thanks Jonathan
For me ... *everyone* should see the following: 1) Install the flash-plugin 2) launch firefox and go to http://www.youtube.com/ ( or any flash site) 2) goto www.google.com , then www.math.wichita.edu, then www.fedoraproject.org, and then www.redhat.com ( * you really only need to go to 4 places without flash and don't hit the forward or back button ) 3) now click on File -> Open File Result: Firefox will crash. Some Crash Features: a) Any attempt to open a new dialog box ( add a new book mark, a webpage popup, the browser complaing about a font, etc ) will crash firefox. b) You have to visit 4 or more web pages. 1, 2 or 3 and you will not get a crash. c) The java plugin shows the exact same bug. d) You can prevent the crash by using File -> Open File and just canceling immediately upon starting firefox. Then you don't get any crash. I suppose the last thing (click on File -> Open File -> Cancel) with every firefox start is a "workaround" to allow you to browse the web. Personally I've tried compiling seamonkey and firefox 2 and I still get the same problem. I'm pretty sure a variant of it is in Fedora as well. If I had to guess the bug in in a mixing of firefox plugins starting threads, firefox threads for dialogs, and the kernel's handeling of said threads. - mark
Hi Jonathan, I'm sorry to hear the update hasn't fixed your crashes. I'll try it out when it comes out in the CentOS repositories. In case you do decide you want to experiment with FF3 I wrote out a set of instructions in the second message here: http://www.nixadmins.net/node/313#comment-522 They are for FF2 but will work with very minor modification (all 2's -> 3's) for FF3. This will keep your existing FF1.5xx intact (in fact, if you skip steps 9 your users wont even know that FF3 is installed. That way you could test it out on one machine with just one account.) I'm not sure it will help but thought I'd pass it along in case you want to try it. If you'd like to correspond further please feel free to e-mail me. Oh, one last thing. Could you e-mail me the link to the latest patch for FF1.5 patch from RedHat? I'd like to read up on the fixes. P.S.- I just saw Mark's message here. This type of crashing on opening a save dialog is exactly what I've been seeing. I wonder if one can write a script that runs firefox and automatically "saves as and cancels" on startup :) Mark, thanks for posting this. I think that will be useful in diagnosing for a solution. Cheers, Andrew
Andrew: here is the link to the errate detail for latest FF patch: https://rhn.redhat.com/rhn/errata/details/Details.do?eid=6625 if you can't view it, I can send separately via direct email. Mark: Tried your crash procedure but did not crash. To me, goes to show how obscure this can be... So our latest theory here is... on (some of) our systems we may be bumping up againest the maximum on the memory and the kernel kills firefox. This could happen at the point where something in is done in Firefox that pushes past the memory limit that the RH5 kernel "tolerates". This could be compounded if FF 1.5 has memory leaks/problems; whereas FF 3.x is being compeletely rearchitected... which may explain why it performs better in some situations. We'll probably try some tests here and turn on full kernel logging to see if we catch something. this memory max out theory could explain why, at least at our location, some users see way more crashes of unknown nature then others... thanks Jonathan
Hi All, Jonathan, thanks for the link. It redirects me to the rhn login but I managed to log in with an old rhn account which appears to still be active so I can view the advisory. Mark's method produces the crash for me on FF2. It gives the same exit code (echo $? yields 139) that I've always been getting for my crashes. (flash-blockers must be disabled as the actual flash needs to run). I also tried it with firefox 3beta2 and it did not crash. Thanks, Andrew
Google is my friend. :) Seems this is a rather well seen bug in the firefox+plugin community. One fix is at ... http://bugs.activestate.com/show_bug.cgi?id=64037 It basically just adds a check for any empty callback. Like the author ( Todd Whiteman ) states at their bugzilla: "works for the case where there is no colormap, but will also stop legitimate callbacks that may not be using any callback data" For me the 3 line fix stopped all of my known crash methods. - mark
Thanks for finding this. I'm guessing you are talking about the file at http://bugs.activestate.com/attachment.cgi?id=7330 I have to admit my ignorance. How do I use this? Thanks, Andrew
In case anyone else comes here looking for a potential fix to Firefox crashes, the crash described in comment number 10 can be easily fixed. Please see the thread at: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=11589&viewmode=flat&order=ASC&start=20 It seems to be mainly a gtk2 issue. Look around comment #30 onwards in that thread.
We issued https://rhn.redhat.com/errata/RHBA-2008-0147.html to fix some of these crashes, but we seem to have missed one, which is being tracked at: https://bugzilla.redhat.com/show_bug.cgi?id=433823 *** This bug has been marked as a duplicate of 433823 ***