Hi, I'm using a HD5750 with 3 monitors connected. On at the middle is not rotated, while the left and the right monitors are rotated. That worked fine since yesterday. I did some updates (xf86-video-ati and a xorg-server update) and now I only see dearanged stripes on the left and right monitors (the ones which are rotated). Any ideas? I'll try to bisect now but I don't know which was the version I used before. I'll try to find that out. Hardware: HD 5750 Software: kernel: 2.6.37-rc1-git11 (worked with that one before) mesa: 7.9 xorg-server: 1.9.2.901 xf86-video-ati: git @3455a3b58532ea3ad901a317126968ab6cbb21b7
Created attachment 40381 [details] dmesg output
Created attachment 40382 [details] Xorg logfile
Created attachment 40383 [details] xdm logfile
Created attachment 40384 [details] xrandr --verbose
Thats what bisect says: 3455a3b58532ea3ad901a317126968ab6cbb21b7 is the first bad commit commit 3455a3b58532ea3ad901a317126968ab6cbb21b7 Author: Alex Deucher <alexdeucher@gmail.com> Date: Wed Nov 17 17:32:41 2010 -0500 radeon/kms: fix buffer base alignment for tiling On r6xx+, 2D tiling can require larger than 4k base alignment. :040000 040000 874334c47dbebbc481423faf4fac86f6cffb1911 e2c6083d44a4f393802e4cbec491b317cf1c802b M src
Are there any kind of informations I could provide to help solving this bug? :) It seems to be still there.
*** Bug 31984 has been marked as a duplicate of this bug. ***
Are either you running a compositing manager like compiz? If so which one? Unfortunately, I can't reproduce this with either 2 or 3 monitors with or without compiz.
Does this help? --- a/src/radeon_kms.c +++ b/src/radeon_kms.c @@ -605,7 +605,7 @@ Bool RADEONPreInit_KMS(ScrnInfoPtr pScrn, int flags) if (info->ChipFamily >= CHIP_FAMILY_R600) { /* set default group bytes, overriden by kernel info below */ if (info->ChipFamily >= CHIP_FAMILY_CEDAR) - info->group_bytes = 512; + info->group_bytes = 256;//512; else info->group_bytes = 256; if (info->dri->pKernelDRMVersion->version_minor >= 6) {
I can reproduce it two ways. The first is with a static xorg.conf, in which case the problem manifests at GDM startup and compiz is not running. Alternatively, if I don't use an xorg.conf file, and let gnome+compiz start in landscape mode, then I can trigger the corruption by using xrandr to rotate the display. The common denominator between Felix and myself seems to be multihead setups. I have two displays in portrait hooked up via DVI-0 and DVI-1, Felix has three monitors.
(In reply to comment #10) > I can reproduce it two ways. The first is with a static xorg.conf, in which > case the problem manifests at GDM startup and compiz is not running. > Alternatively, if I don't use an xorg.conf file, and let gnome+compiz start in > landscape mode, then I can trigger the corruption by using xrandr to rotate the > display. > > The common denominator between Felix and myself seems to be multihead setups. I > have two displays in portrait hooked up via DVI-0 and DVI-1, Felix has three > monitors. Ok, I've tried both static configs and manual rotation on juniper, redwood, and cedar using different combinations of 2 and 3 monitors (DVI and displayport) both with and without compiz. Unfortunately, all situations work fine. Does suggested the patch help?
I don't use any kind of compositing manager, just a clean xorg-server :) The patch seems to work for me. Pulled the newest version from the master tree, applied the patch and installed the driver. No screen corruption anymore.
Created attachment 40678 [details] dmesg output
Created attachment 40679 [details] Xorg log
Alex's patch from comment 9 works for me, applied on top of 90f831. Dmesg and xorg log attached.
Fix pushed: e142e55c56d8440d7185b594c696ec9eeb699cda
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.