Bugzilla – Bug 153566
*instant* hang when changing video modes
Last modified: 2006-03-15 16:39:26 UTC
I first noticed this when booting the 10.1 beta 5 CD. I would say that approx. 2/3 of the time (60-70%) I get a system that is completely hung. After some experimentation I have found that it seems to center around switching video modes - specifically, /after/ the "Loading kernel" stage when the video mode switches from the graphical grub menu to text mode (which usually switches right back to framebuffer), the machine hangs hard. This is a machine which has been absolutely rock solid with 9.1 through 10.0 Additionally, I notice that I can get the machine to hang /at will/, by switching the "font" - (this is hard to explain so bear with me). I have a Dell laptop which has a "font" button which switches between full screen and a mode that has a black border around it - this is done entirely in hardware as far as I can tell and has worked for /all/ versions of Linux (SuSE 9.1 through 10.0, Fedora core, Ubuntu, Debian, etc...). After the fourth or fifth time booting tonight, and finding that it hards immediately after it switches to the just-loaded kernel (about 2 in 3 times), I also discovered that I can get it to hang by pressing function and 'font'. While strictly speaking I don't care specifically about this mode of operation, it seems like a pretty strong coincidence to be unrelated. I'll try to get you as much info as I can, but as yet, this is all I've got. It's a Dell with an NVidia GeForce 2 "go" - I'll try to get hwinfo in a moment. This seems to be to be a very serious bug.
This might be a kernel isse. Attach 500 lines of your syslog and try booting with ACPI=off - does it still happen then?
Unfortunately, using ACPI=off is /worse/. The machine hangs at boot time from time to time. Other times it works OK. When it does boot OK, changing video modes does not help. As for syslog, I can't even get that far. I'll try to attach a hwinfo.txt and a photo.
Created attachment 70561 [details] hung due to ACPI
Created attachment 70562 [details] hwinfo all
Does booting in textmode help (vga=normal). This sounds like buggy graphics hardware to me. Hint: Install the system in text mode and with a bit of luck you can still use the framebuffer if you set it up to use exactly the same resolution and color depth as you X11 uses, in most cases, this helps. Please add `hwinfo --gfx' and attach /var/log/Xorg.0.log, you can check it for (ww) and (ee) lines.
Would having no vga parameter at all be the same as vga=normal? Related to a different bug I had to remove the vga parameter, it was implicated in other problems. I'll try to add hwinfo --gfx tonight. You can get the /var/log/Xorg.0.log here: https://bugzilla.novell.com/attachment.cgi?id=70239
No that's not the same, as default it will use a framebuffer mode, whereas vga=normal sets up ASCII-mode.
Oh! I find that surprising, but that may help to explain a number of things. I'm glad to have learned this, I'll give this a shot tonight and update this bug (and probably one other) with the results. Thanks!!
Created attachment 70720 [details] hwinfo --gfx
Also, not even the 'magic sys-rq (had to enable it)' works when hung. It does work when not hung (just so you know I tested it). I hit the function-font key and this hangs *instantly*. I ran a 'find /' and hit function-font - instant hang. Crazy!
You mean that booting without the framebuffer also results in a hang? At which point exactly?
OK, I think I've gone and confused the situation. Let me try to restate: I've been trying the SUSE 10.1 betas, starting with beta3. When I first booted the beta5 CD, approximately 2/3 of the time it would hang. Since then, the rate has dropped to about 1/5. I think I just got unlucky at first, but I have literally installed 10.1 beta5 on this machine 15-20 times now. Anyway, I happened at the same time to notice that the 'Font' button ("Function" and "F7" on my laptop's keyboard) /also/ hangs the machine in a similar (identical?) fashion. The function+f7 combo changes the "font" used - it's really hard to describe what it does, but basically either there is a big black border around the entire LCD and the font is smaller, or there isn't. It's a BIOS thing. In all previous Linux-based operating systems (and I've run a bunch of them, although I *use* exclusively SUSE), this wasn't a problem. In SUSE10.1 it appears to be - I suspect a kernel and BIOS interaction that is funky. Also related may be bug 152101 - I'll give their proposal (in comment #41) a try - largely, the SUSE betas have been good - I had quite a few problems with beta4 as compared to beta3, but beta5 is an overall improvement. Anyway, I've got another 21 minutes of install time left before I can give all of these bugs a once-over to see if anything has changed. I'll write more in this bug when I know more - I hope this helps. I'd still like to know what causes the hang, even though it's frequency is quite low now. I am unable to explain why initially my results were so poor, but the first 5 or 6 times I booted it would hang on most of them.
Unfortunately, removing bootsplash and using vga=0x317 does not solve the problem. vga=normal does not help either (also with bootsplash removed).
OK, I've tried beta6. The asynchronous problem appears largely gone (still not sure though), however "Function+F7[Font]" still locks the system instantly.
Is the system stable in runlevel 3? If this does not help, this could be a kernel problem. You should start by removing all non-essential kernel-modules from the running system and check it for it's stability. If it works, test each additional module you load for its functionality until you found the offending one.
I have been able to extend my indication of a stable system as originally indicated in comment #14 - beta6 is much more stable for me (rock solid thus far), although the "Function+F7" bug still exists. Unless you care about the "Function+F7" problem I think that beta6 seems to have solved the problem. I found a page on google that indicated something about others experiencing a similar problem - the "cause" was that the "Font" changing takes place /in the BIOS/. I'll note that just to be sure I booted CD1 of SUSE 10.0 (boxed set) into rescue mode and Function+F7 works perfectly. Perhaps the newer kernel is more aggressive about memory locations that earlier kernels avoided (BIOS memory locations)?
What machine is this? Some IBM? As a fact Beta7 is out now, you migth want to try it with this release.
The machine is a Dell Insprion 2650 laptop. I'll give beta7 a try as soon as I can.
As per: http://lists.opensuse.org/archive/opensuse-announce/2006-Mar/0001.html "Note that there's no Beta7. For internal consistency we will just skip that one." I'll have to wait until beta8 (Mar 17).
On top of a fresh beta6 install, I downloaded and installed the kernel from 'factory': kernel-default-2.6.16_rc5_git9-3 and I no longer get hangs or lockups or anything weird (yet) of any kind!!!!! The function+f7 button doesn't work anymore, which is kinda strange, but that is much better than a hang. THANK YOU!
Forgot to mark as ASSIGNED.
You might consider filing a seperate report for the Fn+F7-problem. Let's consider this problem solved. If you experience problems of that sort again, please reopen this one.