From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.12; Mac_PowerPC) Description of problem: Kernel kernel-smp-2.4.9-6, delivered recently causes a kernel panic on a Dell PoweEdge 2550 w/ RAID card. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Install kernel 2. Boot system 3. Actual Results: Panic Expected Results: Normal boot Additional info: Switching back to the stock kernel boots fine.
Another customer john.emerson reported: 7.1 installs and runs, but the current kernel patch in the mirror (2.4.9) hoses the machine. On reboot after the updates there are errors like /lib/aacraid.o: unresolved symbol ERROR: /bin/insmod exited abnormally ... Kernel panic: No init found.
Question: you did upgrade mkinitrd and the other packages required for installing the kernel first, right ?
In John Emerson's case, he did a --force --nodeps on the install, which is probably why his broke. In Hunter's case, he got the new kernel via up2date.
Apparently it is a bug in up2date with the dependency list... Another user states: "I think the real issue is that the filesystem rpm in noarch needed to be installed with other things in i386." Is it possible that up2date didn't install the correct packages?
I'll try to see if I can hunt down a similar machine in our lab to check what is going on. Without testing this on similar hardware it is hard for me to say whether up2date needs to be smarter aout anything.
I have tried to install this on 2 PowerEdge 2550's (Dual PIII 1GHz, 1GB RAM, RAID 1). I installed all patches (kernel last always) from ftp.redhat.com from the i386, i686 and noarch directories. Everytime I boot to the 2.4.9-6 kernel smp or uniprocessor the machine kernel panics and dumps a log file into /var/log. I can forward this to anyone who would like to take a look.
the kernel requires the correct filesystem rpm? or filutils? [alikins@sludge i386]$ rpm -qp --requires kernel-2.4.9-6.i386.rpm rpmlib(VersionedDependencies) <= 3.0.3-1 fileutils modutils >= 2.4.6 initscripts >= 5.83 mkinitrd >= 3.2.2 /bin/sh /bin/sh /bin/sh /bin/sh rpmlib(PayloadFilesHavePrefix) <= 4.0-1 rpmlib(CompressedFileNames) <= 3.0.4-1 That kernel doesnt seem to have any dep on "filesystem". The filesystem package just provides "filesystem". So, just updating "kernel" will not have brought in the "filesystem" package as best I can tell.
One issue that might be related to the comment by edg.edu on 10-26... If you installed the i386 and the i686 "kernel" packages, and didnt run "mkinitrd" manually, you only get one initrd created (the initrd isnt arch specific). If those modules are not arch independent, that may be causing the problem. dunno if that is really the case, but something to look into.
mkinitrd, which has the real dependency on the /initrd directory, depends on filesystem
kernel modules are very architecture specific; i386 ones won't work with the i686 kernel
>mkinitrd, which has the real dependency on the /initrd directory, depends on >filesystem true. And even if nothing had the dep, up2date would find the new filesystem package as an available update as well.
I only used the kernel's in i686 and I have tried the uni-processor, the smp and the enterprise and they all have the same results. I also ran mkinitrd by hand.
edg.edu: but you DID upgrade the required filesystem package too, did you? and you do have an /initrd directory in your root filesystem ?
I did install/upgrade to filesystem-2.1.0-2.1 from the noarch directory and yes I do have an /initrd directory in root. Question... my /boot directory is /dev/sda2 and my / is /dev/sda3 (Dell utils are /dev/sda1) could that cause a problem?
So, what is our status now?
Is this issue resolved for the recently released kernel 2.4.9-12 that fixes the syn-cookie issue?
I'm not sure whose problem this is. It appears to be a RHN problem because the customer simply ran up2date and was left with an unusable system. If packages were upgraded manually (and all of them were updated), I believe it worked fine.
I think this bug report has way too many possibly not related threads going on. hunter: you updated with up2date correct? and then got a kernel panic on reboot? Did up2date report any errors? Can you provide any more information about the kernel panic? (did it complain about missing modules, "unable to mount /" etc?) Matt_Domsch: can the bug report from john.emerson be ignored as part of this report? edg.edu: You did manual installs of all the different kernel packages, including a mkinitrd and filesystem update, and manually running "mkinitrd" correct? And still saw all of them fail?
I ran up2date and there were no errors. Kernel panics like this at boot: /lib/aacraid.o: unresolved symbol ERROR: /bin/insmod exited abnormally ... Kernel panic: No init found.
>can the bug report from john.emerson > be ignored as part of this report? Yes. Upgrading RPMs by hand and doing the kernel separately worked fine for him. They seemed related, but apparently are not.
I did everything by hand, including deleting the initrd's created by installing the new kernel and making new ones. I have subsequently tried this on 2 other systems with the same results. (PowerEdge 2400 and another 2550) Everything installs fine and then when I reboot the kernel panics and the machine hangs.