Summary: | x11-base/xorg-server-1.9.2 segmentation fault with x11-drivers/xf86-video-ati | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Thomas Wabner <waffel> |
Component: | Current packages | Assignee: | Gentoo X packagers <x11> |
Status: | RESOLVED OBSOLETE | ||
Severity: | normal | CC: | amne, cheepeero, dennis, divided.mind, j.habenicht, jan.narovec, kripton, zioalex |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | Follow http://www.gentoo.org/doc/en/xorg-config.xml to ensure that KMS is enabled (and vesafb/radeonfb etc. are disabled) before reporting here | ||
Package list: | Runtime testing required: | --- | |
Attachments: |
Xorg.0.log logfile from xserver start
xorg.log containing the segfault rs690 kms gallium for both r600 and r300, causes flickering interimittently |
Description
Thomas Wabner
2011-01-01 13:23:29 UTC
Created attachment 258568 [details]
Xorg.0.log logfile from xserver start
This is the actual logfile from the xorg-server startup try
Please attach dmesg. Does this happen with both KMS and UMS? If you use KMS and build radeon as module, you need to ensure that it loads before X is started. Also you could see whether the crash happens both with classic and gallium selected in eselect mesa. ok, I found a solution (for me). Thanks for the hint! If I set via 'eselect mesa set r300 classic' it works. If I set r600 and/or sw to gallium it works. Only if I set r300 to gallium I get the segmentation fault again (r600 and sw settings playing no role). The radeon driver is loaded as a kernel module (per default with "autoload modules"). I can attach dmesg, if you want for a working configuration and for a non working configuration. After the upgrade, I was getting segmentation violations too. Thomas' workaround worked (switching R300 to classic). (BTW, unless I'm mixing version numbers, this is a R600. May the error be gallium accessing a non-existing R300?) I didn't see any warning about this, but I'll double-check the emerge logs. Same problem here. But the workaround above doesn't work for me. I got X running with disabling glx in the Module-section of my xorg.conf. The reporter's issue is probably only happening in the combination of UMS and gallium. He has a Mobility Radeon X1400 which is driven by r300, switching r600 will not have any effect. (In reply to comment #6) > The reporter's issue is probably only happening in the combination of UMS and > gallium. He has a Mobility Radeon X1400 which is driven by r300, switching r600 > will not have any effect. > I have stumbled into the same mesa+gallium+ati+xorg-1.9.2 bug after upgrading an older system. The eselect mesa set r300 classic worked around the crash. My adapter is Radon X1600 (RV530). It is my firm opinion users of xf86-video-ati should get a big warning when they upgrade to this server/driver/mesa version, which proves poorly tested for these configurations.(In reply to comment #4) > After the upgrade, I was getting segmentation violations too. > > Thomas' workaround worked (switching R300 to classic). (BTW, unless I'm mixing > version numbers, this is a R600. May the error be gallium accessing a > non-existing R300?) > > I didn't see any warning about this, but I'll double-check the emerge logs. > Sorry about the mess in the above post. With a bad internet connection, I pasted my previous post after reloading the page and ... Sorry :( I just managed to crash the server while using it in KDE4, and it continues to crash as long as I hover my mouse over the taskbar (generating window previes) or type into combo boxes (generating dropdowns and such). This only happens from time to time, but certainly after 5-10 minutes. I guess 2D effects will have to do for now :) This bug solved the same issue I was having (Comment #3). I have a Radeon Mobility X1400 using the ati drivers, Gentoo amd64 system w/ KDE from layman. Unlike some users in other bugs who report OpenGL as the problem, I have 3D effects working just fine with OpenGL, tested using Desktop Cube effect. Some say switching to XRender works well too. Advanced settings include OpenGL composite type, Keep Window Thumbnails as "Always (breaks minimization)", functionality checks disabled, direct rendering and vsync enabled, with fastest texture filter. Thanks for the solution! =) I'm seeing this as well. Comment #3 allowed me to start up X but before disabling GLX it kept crashing during various unrelated operations in different applications. Radeon X1650 Pro, KDE 4.4.5 from portage. So when someone attaches his dmesg from when the crash happens (using KMS+gallium, or UMS+classic), maybe the issue can be isolated further. Created attachment 258803 [details]
xorg.log containing the segfault
I'm not entirely sure which lines of the following dmesg output were actually caused by the crash as X restarted instantly afterwards but here it goes:
mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
mtrr: type mismatch for d0000000,10000000 old: write-back new: write-combining
[drm] Setting GART location based on new memory map
[drm] Loading R500 Microcode
[drm] Num pipes: 1
[drm] writeback test succeeded in 1 usecs
[drm] Num pipes: 1
Re-Enabling GLX didn't cause X to crash but re-enabling KDE compositing effects crashed X instantly.
> video=vesafb:mtrr:3,ywrap video=0x317 radeon.modeset=0 nomodeset
Please do not use fb drivers such as vesafb or uvesafb with modern X, this does not work. When kernel modesetting is disabled, make sure that eselect mesa is set to classic.
(In reply to comment #14) > > video=vesafb:mtrr:3,ywrap video=0x317 radeon.modeset=0 nomodeset > > Please do not use fb drivers such as vesafb or uvesafb with modern X, this does > not work. When kernel modesetting is disabled, make sure that eselect mesa is > set to classic. > Removing fb didn't help. I've since enabled kms and switched mesa to use gallium and X starts up and runs just fine. So the situation for me is: - ums+gallium: segfault on X start - ums+classic: segfault with KDE compositing effects - kms+gallium: working so far Hope that helps... This bug may be the same as reported in bugreport #347721. I've also experienced segfaults with the newest xorg-server (KDE4 with compositing effects, radeon, UMS + classic mesa, amd64, vanilla kernel 2.6.36.2). (In reply to comment #15) > - ums+gallium: segfault on X start > - ums+classic: segfault with KDE compositing effects > - kms+gallium: working so far > > Hope that helps... > Yes, thanks a lot! I have almost an hour of stable uptime of X :) Though the screen flickers from time to time, it's usable. i had the same issue and the resolution is the same but my setup is a bit different than everyone else. i have an onboard rs690 mobo chipset using the radeon x1250 onboard vid card. i'm using the xf86-video-ati-6.13.2 drivers which i can't seem to specify kms/ums (what are these acronyms anyway?). x segfaulted when mesa's r300 was set to gallium. i followed the suggestions and i've been stable (gnome + compiz in all its fun) ever since w/ r300 set to classic. here's my current mesa list: i915 (Intel 915, 945) i965 (Intel 965, G/Q3x, G/Q4x) r300 (Radeon R300-R500) [1] classic * [2] gallium r600 (Radeon R600-R700, Evergreen) [1] classic * [2] gallium sw (Software renderer) [1] classic [2] gallium * i followed the status whiteboard to set kms and disable the older *fb drivers. now the console just stays blank whenever i boot into the kernel compiled with those params. however, after a bit of time, x starts up and i can login using x, but if i try to switch to a console tty, my computer seemingly freezes (at least the display since i can still ssh and make my keyboard lights change w/ the various lock keys). this is a dmesg grep of such a boot: $ dmesg | grep -i -e radeon -e ati -e fail -e error -e edid Using ACPI (MADT) for SMP configuration information Allocating PCI resources starting at cff00000 (gap: cff00000:10100000) Subtract (51 early reservations) Preemptable hierarchical RCU implementation. Fast TSC calibration failed TSC: using HPET reference calibration Calibrating delay loop (skipped), value calculated using timer frequency.. 6013.63 BogoMIPS (lpj=3006819) PCI: Using configuration type 1 for base access ata3: softreset failed (device not ready) ata3.00: HPA detected: current 976771055, native 976773168 ata1: softreset failed (device not ready) ata1.00: HPA detected: current 145223999, native 145226112 pata_atiixp 0000:00:14.1: PCI INT A -> GSI 16 (level, low) -> IRQ 16 scsi4 : pata_atiixp scsi5 : pata_atiixp [drm] radeon defaulting to kernel modesetting. [drm] radeon kernel modesetting enabled. radeon 0000:01:05.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18 ATOM BIOS: ATI radeon 0000:01:05.0: VRAM: 128M 0xD0000000 - 0xD7FFFFFF (128M used) radeon 0000:01:05.0: GTT: 512M 0xA0000000 - 0xBFFFFFFF [drm] radeon: irq initialized. [drm] radeon: 128M of VRAM memory ready [drm] radeon: 512M of GTT memory ready. [drm] radeon: 1 quad pipes, 1 z pipes initialized. [drm] radeon: ring at 0x00000000A0000000 [drm] radeon: ib pool ready. failed to evaluate ATIF got AE_BAD_PARAMETER [drm] Radeon Display Connectors fb0: radeondrmfb frame buffer device [drm] Initialized radeon 2.6.0 20080528 for 0000:01:05.0 on minor 0 [drm:drm_mode_getfb] *ERROR* invalid framebuffer id it seems the fb for the console can't find where it's going. i still have X going to classic instead of gallium yet, but i'll try it out and report back if things work with it like that. Please, someone who sees this crash with KMS+gallium, KMS+classic or UMS+classic attach his full dmesg. If the console goes blank after radeon KMS initializes, this could be due to missing CONFIG_FRAMEBUFFER_CONSOLE in the kernel (or if built as module, fbcon not loaded). Created attachment 258994 [details]
rs690 kms gallium for both r600 and r300, causes flickering interimittently
so i tested out gallium for mesa r300 and r600. here's my lsmod showing the kms being in effect:
$ lsmod | grep -e drm -e radeon -e fb -e ati
cpufreq_conservative 7772 4
radeon 725092 2
snd_hda_codec_atihdmi 2171 1
ttm 43370 1 radeon
drm_kms_helper 20825 1 radeon
drm 145346 5 radeon,ttm,drm_kms_helper
fb 30454 2 radeon,drm_kms_helper
fbdev 665 1 fb
snd_hda_codec 48990 3 snd_hda_codec_atihdmi,snd_hda_codec_realtek,snd_hda_intel
i2c_algo_bit 4183 1 radeon
cfbcopyarea 2857 1 radeon
cfbimgblt 1881 1 radeon
pata_atiixp 3481 0
cfbfillrect 2917 1 radeon
i managed to restart my X after setting the mesa stuff to gallium (blind typing after "switching" to a console... it never switched, but i had a hunch that my typing would still work in the tty and it did) and i've attached the xorg log of this session.
the screen flickers intermittently like this, so i'll be switching back to classic for r300.
Intermittent blanking is a separate issue. Try booting with drm_kms_helper.poll=0 kernel parameter. thanks chi-thanh. loading fbcon solved my console issue, and the blanking was fixed by just having the FB_MODE_HELPERS compiled in. i didn't even have to change the poll to 0 as you suggested, but might try that b/c i'm getting a tiny bit of flicker only when i switch desktops using compiz's cube effect. anyway, i'm running kms+gallium pretty stable now. will upload a dmesg if it crashes. After recompiled the kernel and using only KMS (following the suggested documentation) AND switching all to gallium it works without problems (also with compiz). I think the ebuild needs a big warning if UMS and gallium is used (or switch automaticly to classic with ums). With KMS and gallium all thinks are fine. following instructions below : i915 (Intel 915, 945) i965 (Intel 965, G/Q3x, G/Q4x) r300 (Radeon R300-R500) [1] classic * [2] gallium r600 (Radeon R600-R700, Evergreen) [1] classic * [2] gallium sw (Software renderer) [1] classic [2] gallium * Solved the issue (changed r300 from gallium to classic) although the performance is better then the old xorg (2D) glxgears shows weird deviations (on one of the wheels is a bar attached that spins with the wheel) did not test 3D sufficient to draw any conclusions. Gentoo PPC Ibook G4 Dear all, I had encouter the same problem before but I found a different solution to it. According to this page: http://www.x.org/wiki/radeonBuildHowTo , I recomplie the kernel as the instructions and reboot, the problem solved. My packages are: x11-base/xorg-server-1.9.4 x11-drivers/xf86-video-ati-6.14.0 media-libs/mesa-7.9.1 x11-libs/libdrm-2.4.23 sys-kernel/gentoo-sources-2.6.36-r5 And the results: OpenGL renderer string: Gallium 0.4 on softpipe May this be helpful. I worked around the crash by disabling Kde’s opengl Windows Effects and installed compiz-fusion to have transparences and effects (I’m interested in the former). But I would like to solve this issue. I’m using gentoo-sources-2.6.37 and stable X related packages (xf86-video-ati). I'm getting the same crashes with R200 (so it seems it isn't a gallium issue, as there's no gallium for R200). X starts up all right, but when using OpenGL window effects (kwin4) it segfaults after a few minutes. Versions: x11-base/xorg-server-1.9.5:0::gentoo media-libs/mesa-7.10.3:0::gentoo x11-drivers/xf86-video-ati-6.14.1:0::gentoo Kernel: 2.6.38-gentoo-r6 (I use UMS. I will try to setup KMS and check if it still crashes.) I can confirm that switching to KMS/DRI2 fixes the issue on my system. At least it has been running for over an hour, with fairly heavy OpenGL effects usage, and is stable thus far. KMS brakes suspend/resume stuff. Nobody else has resume issue? It's best to report this upstream. Marking as obsolete as the original problem appeared only with UMS which is no longer supported in the latest xf86-video-ati. |