FS#16674 - [kernel26] Radeon Boot Error
Attached to Project:
Arch Linux
Opened by Mr. K. (KitchM) - Thursday, 15 October 2009, 18:27 GMT
Last edited by Andreas Radke (AndyRTR) - Thursday, 31 December 2009, 10:19 GMT
Opened by Mr. K. (KitchM) - Thursday, 15 October 2009, 18:27 GMT
Last edited by Andreas Radke (AndyRTR) - Thursday, 31 December 2009, 10:19 GMT
|
Details
Description:
After an upgrade of the system today, the display of the OS stops working properly during the boot process (after getting to the UDEV section), and the display simply shows odd patterns of red lines on the screen. I edited the first (default) GRUB menu line, and selected to edit the second section line of that by adding radeon.modeset=0 to the end of it. This worked the first time I tried it, and I could even get into the Xfce DE. However, it never worked again and I have no usable display at all; it only shows the same sort of red patterns each time I boot. I cannot give any further information as I cannot get to a CLI. Additional info: * package version(s) * config and/or log files etc. Steps to reproduce: Boot computer. Watch screen. |
This task depends upon
The boot process stops at "Starting UDEV Daemon [Busy]"
Selecting a separate kernel (possibly without KMS enabled) boots with no problem.
Card: Radeon 9600 (RV350)
Kernel log attached (call trace)
root (hd0,0)
kernel/vmlinuz26 root=/dev/disk/by-uuid/c882892e-14d0-495b-a9cb-66ad1a678f5f ro
initrd /kernel26.img
Pretty standard stuff.
I tried the grub editing (which I wrote about in the wiki), but the options given in the news item's wiki link do not seem to work. If there was something to try to get it to work as it did the first time, I would like to put that in. I did try the nomodeset, but got nothing.
I wonder why the radeon.modeset=0 worked once, but not again. Isn't that significant to you programmers?
If I understand correctly, the alternative is to boot from the ISO cd, try to mount the file system found on the hard drive, start nano with the rc.conf file and add something. The referenced intel_apg doesn't seem right to me with my ATI Radeon-based card.
I will try to see what I can do, but programming is not my area of expertise.
It might be possible to set disablemodules=radeon (or whatever the actual module is named) on the grub line. It should boot, but won't give you X, I imagine.
Hmm, when do these "red lines" appear? Before X, or after?
I tried disablemodules=radeon as you suggested. I did get a different pattern of garbage on the screen, but nothing else.
Just for fun I listened for the hard drive to stop accessing, typed in root, Enter, and then the password, Enter. Then I tried startx but got no change of screen. I heard disk access for the amount of time I would expect it to take to get to the DE. Then I had to do Ctrl-Alt-Del to reboot. Everything sounds like it is working, but I can't see anything until I reboot. There is no doubt that something in the upgrade (most likely KMS) caused the video problem.
More later. (I've got to find that darn DVD. Now where did I put it.......)
BTW, found the DVD; now I've just got to figure out how to mount my hard drive so as to be able to edit the files on it. (Damn man page......)
I just wish the darn thing worked the first time. The lost productivity is adding up. I guess I don't understand why it was necessary to screw around with an ATI setup that was working. There is an implied trust that is broken when an automatic "upgrade" that is performed causes a working system to become unusable. (Once I get this figured out, I'll be very interested in finding the individual responsible for this royal f**k-up. He and I will have a little talk.)
I ran into the same problem trying to install Archlinux for the first time on my old PC.
Booting on the netinstall CD --> no problem
Archlinux installation --> no problem
First boot --> Black screen a few seconds after UDEV loading
I rebooted countless times in the install cd, mounting my root partition, editing /etc/rc.sysinit, and echoing some stuff..
Finaly, adding a big "sleep 30" after the udev section makes my system to hang before displaying other things, so the problem effectively seems to come from udev loading.
And I found this bug, and adding "radeon.modeset=0" to the kernel boot command line fixes my problem.
My ATI card from lspci : Radeon R350 (Radeon 9800 Pro)
Is there any thing/trace I can give to you to help you fix the problem ?
Marc
The important parts of rc.conf:
MOD_AUTOLOAD="yes"
MODULES=(fuse)
USELVM="no"
DAEMONS=(syslog-ng network netfs @crond alsa @openntpd hal avahi-daemon avahi-dnsconfd dhcpcd webmin !stbd samba)
More later.
Nothing in mkinitcpio.conf either.
There appears to be no way to get a non-KMS mode as an alternative. So, yes, that must also be broken.
If you have any other requests for information, I can get to it now; so let me know.
In the news item I read this:
"Experimental kernel modesetting (KMS) for the radeon driver is now enabled."
I found this in the wiki:
"Note: For now, only Radeons up to R5xx (X1xxx) support KMS. Support for later Radeon cards will be added in Linux kernel 2.6.32."
(I'm beginning to think that isn't correct.)
IMHO, someone put an experimental function into the normal upgrade process. Experimental stuff is supposed to go into the [testing] repo, so that we can avoid it automatically. Further, this appears to <u>not</u> be tested on all ATI cards. Knowing that there are enough problems with the many permutations of ATI and nVidia cards, these things should always be optional at the very least. At the best, coders should put into the upgrade process a chipset identification routine, and a user choice of questionable "features" before compilation of the new kernel.
Does anyone know how to compile the old kernel from an install CD?
Bottom line, the video had never given any problem before. Exactly after the "upgrade" and one (1) reboot, I got the aforementioned result. Nothing seems to bring it back. I am working on restoring the old kernel and hopeful of getting back to a usable computer.
Now everything is back like it was before the disastrous kernel change. I realize that I must not have understood the Arch definition of "cutting edge". It must be "buggy and in need of testing by users". I should probably count myself lucky in that I was able to solve the problem by myself, and that I only lost one week of computer usage.
Last try: report you rc.conf, modprobe.conf, mkinitcpio.conf and kernel append line from grub. make sure you rebuild your initrd after changes you do. When something goes wrong report exactly the error messages you can get from screen, dmesg, everything.log.
If you are not interested I'll close this one soon.
I believe I was extremely clear in how to solve the problem. I described how I had to downgrade to the previous kernel. Those are the steps to solve the problem. Did you not read my post?
Of course I can't use the latest kernel since IT DOESN'T WORK RIGHT!!!!!
Of course one needs to get the card to work in non-kms mode; duh. You haven't been able to help with that yet; which is the point of posting the bug report.
Better feedback? Did you not read the post? There was no command line. I documented everything I had to do to get to the place where I could read the files requested. When I was finally able, I clearly listed the important content. Also, please note that no one asked for modprobe.conf before. And immediately when asked, here it is:
#
# /etc/modprobe.d/modprobe.conf (for v2.6 kernels)
#
Hope that helps a lot.
Error messages?
screen - I don't know how to read red lines. Maybe you do. If so, you're better than me.
dmesg - see attached
everything - see attached
(Please note again that the last two are only available now that I did all the work to get the system working again.)
I suggest this only be closed when the problem is found by the people who made the changes to the latest kernel and there is a published fix. Until such a time is reached, this is obviously not fixed and remains a reproduceable bug.
If you don't want to help, that's fine. But don't ever imply that I didn't do all I was asked. In fact I went much further to research and check with others who had similar problems. When a user cannot get to a usable screen, the very first thing should have been to give clear steps to boot into a usable display, but you choose not to do so. How else would some be able to give proper feedback? That is so obvious that it shouldn't have to be explained.
So, IMHO, we can either be a part of the problem or a part of the solution. Being a part of the solution is offering useful steps to get out of the problem. This I have done. What about you?
everything.log (0 KiB)
Please put radeon.modeset=0 into your /etc/modprobe.d/modprobe.conf, make sure the same is in your grub append line or try there with "nomodeset". Clean mkinitcpio.conf and rc.conf. pacman -S kernel26. See also http://wiki.archlinux.org/index.php/ATI#Kernel_Mode_Setting_Troubleshooting - so far you are the only one not getting 2.6.31.4 kernel booting.
When the kernel loads in non-kms mode passing the initrd shouldn't be hard. If it keeps hanging at udev loading modules that's can usually be fixed with loading the desired apg module before radeon module.
And please post logs from failing runs not from working old kernel. You should be able to find the log in everything.log even after a hardfreeze when klogd was already up at that state. Or try pressing STRG+PageUp to go back if possible.
Second, if the setting is in grub, why does it also have to be in modprobe.conf, or vice versa? Also, I want to know why you think that setting will work now when it didn't before.
Third, what do you mean by "Clean mkinitcpio.conf and rc.conf"?
Fourth, do you next want me to run pacman -S kernel26 from the CLI?
Fifth, to be clear, when I pointed out that I had done the research, please know that I have already gone over the applicable part of the linked wiki page a number of times. And your statement that I am the only appears to be an unwarranted assumption without any foundation in reality. The fact is that I immediately knew what caused the problem, while many others would not recognise that. Perhaps they would not respond either. Also, I posted, when others may not have known what to do. Further, there are many other associated problems with this kernel upgrade which are listed right here in the bugs listings. IMHO, I think there is a lot more than video issues here, and many more people than might be sampled in the bugs listings. So, I hope you can understand my misgivings regarding the statements made.
Sorry, but you can understand how we are beginning to question some of the things we are told. Your suggested steps entail a lot of work whether or not successful, and the system works right now just fine with regard to the video. (I don't yet experience any other issue.)
Anyway, I am not sure how one guarantees that the kernel does load without the KMS mode. And as I pointed out, there is evidence to indicate that the system does not hang, but rather is struck with an unusable video. As to the logs, I don't think you understand how difficult what you are asking is to do. Hell, I don't even know what "STRG" stands for.
2) mkinicpio can build modules into the initrd. you need to specify the file modprobe.conf to include in mkinitcpio.conf. so you can modify the behavior in early "mode", if the radeon module should be included or not and what option you want to pass to the module.
3) remove the radeon module from the module line or blackmask it if you don't want it to be inluced in intrd or loaded at udev call
4) when you change settings for the initrd you need to rebuild it. this can be done with mkinitcpio -p kernelname or pacman -S kernelname
5) Closing this bug. KMS and non KMS mode both work for me and anybody else. The bugtracker is not for personal support. If you can't manage to solve it with the hints I gave you please ask in our support forum and irc channels. I'm sure you will find out what you are doing wrong.
Thanks for your help. I appreciate people who aren't afraid to stand up and point out that the emperor has no clothes.
Rather than that, can you try git master and see if the problem appears for you? I can only replicate it inconsistently.
So there is no way I will go thru that problem again. I'll wait until the new kernel is available in the normal way. I figure that more delay is beter, because I can then allow them (whoever they are) more time for testing. Like many users, I'm kind of tired being a guinea pig. So I'll be watching this problem closely.
Thanks again.
Yahooooooo!!!!!!
Of course there is no way in hell that any of us peons will be able to find where the change was made and exactly what it was. But I do know one thing for sure; it works now and the version changed. Discuss it amongst yourselves.
The bottom line? It works for me. Mark this one as "Solved".
Big thanks to whoever figured it out!