When I have a background set with fbcon_decor the system freezes in certain moments. I noticed it always freezes when I run a kernel's `make menuconfig` when it's supposed to draw the menu. The screen is cleared and only the bottom line is blue. It also freezes when compiling a package, though I can't say when. But when it does, the bottom line is an exact copy of the previous. "Line" as in a horizontal line of symbols on the screen. I tested various setting and I noticed that only the background needs to be set for this to happen. By "freezes" I mean that it doesn't respond to any input, SysRq calls included. ATI Radeon r300, KMS, 1280x1024, 24bpp Reproducible: Always Steps to Reproduce:
Fails only with 24bit colour depth. On 1024x768 and 1280x1024 it crashes. On 800x600 and 640x480 there's image corruption but it works. 16bit and 32bit works ok. Should I take this upstream?
Copying spock, as he is upstream. Anything different with gentoo-sources-2.6.36?
Created attachment 253303 [details] dmesg With .36 it still crashes, but the rest of the system still work. I could toggle LEDs on the keyboard, and I could log in via SSH and remotely use the PC. So I finally got some kernel output. \(^o^)/ Emerge halted right after it would output ebuild info and `poweroff` didn't work; `poweroff -f` did though.
Oh, and my LCD display didn't go into standby mode.
Does not happen in 2.6.37.
Sorry, I set wrong bpp, it still malfunctions.
Same on r600 here. It hangs after modprobe radeon in initrd, SysRQ+W/T shows that fbcondecor_helper is waiting for mutex. If there is no modprobe radeon in initrd, then udev will try to load it, resulting in frozen system (but it will timeout after a 60 seconds or could be "resumed" by SysRQ+I). Removing fbcondecor_helper from initrd and from system resolves this issue.
Created attachment 274297 [details] fbcondecor_help backtrace from sysrq+w This could be slightly different issue, by the way, but looks like very similar.
BTW, slight addition: this is for 2.6.39. Sorry, hadn't seen version from bug.
Created attachment 274831 [details] 800x600-24 output I also experience the issue with initrd. The issue I bugged still exists in .39. Here's how the output for 800x600-24 or lower resolution looks like.
Created attachment 274833 [details] 1280x1024-24 On 1024x768-24 or higher it stops after the first line. That black space on the right of the blue line has |||| symbol like the ones visible on the previous shot. If you ask me, it looks like the driver has wrong char/bpp length for -24 or something. On lower resolutions it doesn't crash so the issue might be that on higher ones it go so out-of-bounds it tries to write RO memory and the driver is killed or something.
Still valid with more recent kernels?
Created attachment 316037 [details] kernel message (3.4.0) Yes, it still happens on 3.4.0 (with video=#x#-24 as said earlier). It seems something goes wrong when decorations are set, since it works okay when there are none, and fails after setting them to on with fbcondecor_ctl.
I switched the graphics card to a WinFast GeForce 6200 and now when I set video=1280x1024-24, then the screen had no colours, was uhh... how to put it... divided like 3:2 with a copy on the right and condensed (also with no decorations). My LCD has broken, so I'm using an older CRT now, so maybe the monitor can't support this mode? In any case... Is 'video=<width>x<height>-24' a valid value in the first place? If no one else had this problem then I cannot rule out the possibily of a hardware failure on my Radeon here, though.
(In reply to comment #14) > In any case... Is 'video=<width>x<height>-24' a valid value in the first > place? http://www.kernel.org/doc/Documentation/fb/modedb.txt video=<driver>:<xres>x<yres>[-<bpp>][@refresh] Maybe that "<driver>:" part is required, which you can assume by the syntax in this documentation (it has no optional brackets around it).
(In reply to Tom Wijsman (TomWij) from comment #15) > (In reply to comment #14) > > In any case... Is 'video=<width>x<height>-24' a valid value in the first > > place? > > http://www.kernel.org/doc/Documentation/fb/modedb.txt > > video=<driver>:<xres>x<yres>[-<bpp>][@refresh] > > Maybe that "<driver>:" part is required, which you can assume by the syntax > in this documentation (it has no optional brackets around it). dottomi, did you try it?
(In reply to Pacho Ramos from comment #16) > (In reply to Tom Wijsman (TomWij) from comment #15) > > (In reply to comment #14) > > > In any case... Is 'video=<width>x<height>-24' a valid value in the first > > > place? > > > > http://www.kernel.org/doc/Documentation/fb/modedb.txt > > > > video=<driver>:<xres>x<yres>[-<bpp>][@refresh] > > > > Maybe that "<driver>:" part is required, which you can assume by the syntax > > in this documentation (it has no optional brackets around it). > > dottomi, did you try it?