From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.2-2r1 i686; en-US; rv:0.9.1) Gecko/20010608 Description of problem: Kernel 2.4.3-12 released by Red Hat, compiled with ReiserFS into the kernel and disabled ReiserFS internal checking, running on a root ReiserFS system. When system is shut down, it hangs at the stage "Unmounting file systems:". Pressing Ctrl-Alt-Del here gives a warning: "shutdown: warning: cannot open /var/run/shutdown.pid". Problem did not exist on kernel 2.4.2-2. How reproducible: Always Steps to Reproduce: 1.Your system should be running ReiserFS on root partition (see http://cambuca.ldhs.cetuc.puc-rio.br/ ) 2.Install kernel-source-2.4.3-12 3.make oldconfig using configs/kernel-2.4.2-i686.config 4.make menuconfig and edit File Systems, change ReiserFS to compile into the kernel instead of as a module. Disable internal code checking for ReiserFS 5.Compile kernel and install kernel. 6.Reboot into new kernel 2.4.3-12 7.Log in as root 8.Run 'shutdown -h now' Actual Results: When you reach 'Unmounting file systems:' on shutdown, it will hang. Press "Ctrl-Alt-Del". Message "shutdown: warning: cannot open /var/run/shutdown.pid" appears. Expected Results: It should have shut down cleanly. The problem was not present when I compiled kernel 2.4.2-2 with ReiserFS support and internal checking disabled, for the exact same system. Additional info: The kernel was compiled with the steps according to my own documentation at http://www.pobox.com/~tzd/linux/gateway/3350.html
Oops one typo. Steps to reproduce, #3 should say: make oldconfig using configs/kernel-2.4.3-i686.config instead of kernel-2.4.2-i686.config. -- tzd
investigating
Created attachment 21726 [details] reiserfs patch
Could you try applying the previously attached patch to your kernelsource and see if that makes it work ?
Yes, the patch (id 21726) works. I can shut down cleanly now. Do you mind explaining what the patch does (I don't know how to program)? Thanks!
It removes a bugfix that was applied to our kernel for a race on module-unload, however the bugfix was not tested for the "reiserfs is not a module" case. It seems it deadlocked in that case due to the locking changes in the original bugfix.
Thanks for the explanation. Is it safe for me to continue using the patch?
Yes.
Great! And thanks for the speedy resolution!
I had this exact same problem. I seem to be having an issue recompiling the 2.4.3-12 from the source RPM. Using the binary RPM of the 2.4.3-12 kernel supplied by RHAT, the aic7xxx_mod driver is listed in my /etc/modules.conf, as the default aic7xxx driver to use. When I recompiled this kernel from the source RPM, and I try to make an initrd image, it says it cannot find the aic7xxx_mod file. After looking through my /li/modules/2.4.3 tree, I saw that only aic7xxx and aic7xxx_old are compiled. Which one should I use to boot with Justin Gibbs' aic7xxx driver? I'm guessing the aic7xxx.o file, but I'm not sure. Why wouldn't the aic7xxx_mod driver compile from the source?
aic7xxx.o is the new driver. Justin Gibb's driver got renamed in Linus' / Alan's kernels, in the RPM build we rename the driver to it's original "2.4.2-2" name to make sure our kernel rpm keeps working for people using this driver.