| Summary: | x11-base/xorg-server with EXA makes firefox (and others) crash with BadAlloc | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Alexandre Ghisoli <alex> |
| Component: | Current packages | Assignee: | Gentoo X packagers <x11> |
| Status: | RESOLVED FIXED | ||
| Severity: | major | CC: | dkarasik, remi, sven.koehler, tienchi |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
|
Description
Alexandre Ghisoli
2008-03-18 15:57:09 UTC
In fact, the two workaround are OR'ed. I mean stay with x11-drivers/xf86-video-i810-2.2.1 and the xorg.conf fixe or use the -2.1.1. Sorry for the confusion Same bug appear on tested 2.2.99.901
Chip :
Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
Subsystem: IBM Unknown device 0557
Flags: bus master, fast devsel, latency 0, IRQ 16
Memory at e0000000 (32-bit, prefetchable) [size=128M]
Memory at d0000000 (32-bit, non-prefetchable) [size=512K]
I/O ports at 1800 [size=8]
Capabilities: [d0] Power Management version 1
Yeah, I can reproduce this bug on my i855GM too with firefox-bin-3_beta5. This is more likely a firefox bug. Firefox likes to think of video memory as a pool of unlimited memory where it can allocate all the pixmaps it wants. In theory, firefox should check for BadAlloc when using the Xlib but obviously, it doesn't. Maybe it's a different bug, but since I really don't know how to debug Firefox, I can't really say for sure. I do use firefox daily, but it is one of the apps that cause a lot of problems on X11. I'll keep tracking this bug, because I've hit it in the past, but on the side of the Intel driver there's really nothing more to be done, unless a stack trace or something else could prove that the driver is doing something wrong. Thanks (In reply to comment #3) > This is more likely a firefox bug. Firefox likes to think of video memory as a > pool of unlimited memory where it can allocate all the pixmaps it wants. In > theory, firefox should check for BadAlloc when using the Xlib but obviously, it > doesn't. I really suspect something in the i810 driver, because changing the rendering method gives better results (XAA vs EXA) and it *was* working pre -2.2.0 series. And it seems that other apps can be impacted by this error (I cannot stay without browser so i've not see any other errors right now) Xv as errors reported as well : https://bugs.freedesktop.org/show_bug.cgi?id=2772 and wine : https://bugs.freedesktop.org/show_bug.cgi?id=13779 > I'll keep tracking this bug, because I've hit it in the past, but on the side > of the Intel driver there's really nothing more to be done, unless a stack > trace or something else could prove that the driver is doing something wrong. Hum, I dont know how to get a full trace here, since it die out of firefox and we should debug the driver itsef. It still crash on i810 2.2.99.902 as well. Working ok if using XAA (not EXA). (In reply to comment #4) > I really suspect something in the i810 driver, because changing the rendering > method gives better results (XAA vs EXA) and it *was* working pre -2.2.0 > series. Pre 2.2.0, EXA had to be enabled by default, that could explain that issue. And before 2.2.0, EXA didn't really do much in the Intel driver. > And it seems that other apps can be impacted by this error (I cannot stay > without browser so i've not see any other errors right now) > > Xv as errors reported as well : > https://bugs.freedesktop.org/show_bug.cgi?id=2772 > > and wine : > https://bugs.freedesktop.org/show_bug.cgi?id=13779 Yep, these do look similar. I'd say those apps try to allocate a biiig pixmap and the driver returns BadAlloc. > Hum, I dont know how to get a full trace here, since it die out of firefox and > we should debug the driver itsef. True, but we should know what Firefox is really trying to do, that would be a big help for upstream devs. Could you try using EXA but also disabling the Render extension? Adding this should be enough : Section "Extensions" Option "RENDER" "false" EndSection My other advice is to open a bug on freedesktop's bugzilla with this nifty url that crashes firefox instantly, that's a huge help for debugging and I'm sure upstream devs will be able to identify the issue quickly (maybe not fix it, but at least know what's happening) Upstream devs will probably ask you to read this document http://intellinuxgraphics.org/how_to_report_bug.html If you do file a bug upstream, please paste the URL here so I can track it too. Thanks a lot for not giving up :) (In reply to comment #6) > My other advice is to open a bug on freedesktop's bugzilla with this nifty url > that crashes firefox instantly, that's a huge help for debugging and I'm sure > upstream devs will be able to identify the issue quickly (maybe not fix it, but > at least know what's happening) > > Upstream devs will probably ask you to read this document > http://intellinuxgraphics.org/how_to_report_bug.html > > If you do file a bug upstream, please paste the URL here so I can track it too. > > Thanks a lot for not giving up :) > I'm not able to get a meaningfull gdb for the FireFox crash (even a break on gdk_x_error), it crash without BT information (yes, I've compiled FF with debug symbols). It seems to be related to exa itself, not i810 driver. It seems to be fixed in xorg commit : http://cgit.freedesktop.org/xorg/xserver/commit/?id=bc2d516f16d94c805b4dfa8e5b9eef40ff0cbe98 More info on the bug here. https://bugs.launchpad.net/ubuntu/+source/xorg-server/+bug/205599 The URL pointed out in comment from Tawsi (2008-03-26) crash my FireFox session 100% of the time. So, the xorg git commit seems to have fixed it (not tested right now). There is a discussion on the ML about : http://lists.freedesktop.org/archives/xorg/2007-June/025579.html And the debian bug : http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=471782 It seems to be a quite common bug, but I dont know how to check witch versions on Xorg are affected, and how to test it. Thanks for your support, Donnie, The fix for this bug applies on top of xorg-server, can I apply to 1.4.0.90 (creating a -r4 in the process) ? Thanks *** Bug 212799 has been marked as a duplicate of this bug. *** (In reply to comment #8) > Donnie, > > The fix for this bug applies on top of xorg-server, can I apply to 1.4.0.90 > (creating a -r4 in the process) ? Just putting what I said on IRC here as well ... a number of bugs assigned to x11@ have the Inclusion keyword. They should all be evaluated for inclusion. One option is applying a patch to the latest git of the 1.4 branch, and then cherrypicking other patches as needed. I pulled this patch into my local 1.4 branch and requested that upstream take it. Fixed with xorg-server 1.4.1 and newer. Thanks |
