Here is what happens when one I boot a RH9 IBM xseries box.
EXT3-fs: mounted filesystem with ordered data mode. Freeing unused kernel mamory: 132k freed INIT: version 2.84 booting Settomg defai;t fpmt (latarcyrheb-sun16): [ ok ] Welcome to /etc/rc.d/rc.synsinit: line 70: 35 Segmentation fault LC_ALL=C grep -q "Red Hat" /etc/redhat-release Red Hat Linux Press 'I' to enter interactive startup. Mounting proc filesystem: [failed] /etc/rc.d/rc.sysinit: line 90: 42 Segmantation fault LC_ALL=C grep -q .initrd /proc/mounts Confuguring kernel parameters: Error: /proc must ve mounted To mount /proc at boot you need an /etcfstab line like: /proc /proc proc defaults In the meantime, mount /proc /proc -t proc [failed] Setting clock (localtime): Fi Jan 23 09:06:35 MST 2004 [ ok ] Loading default keymap (us): [ ok ] Setting hosname mgserver.net: [ ok ] /etc/rc,d/rc.sysinit: line 182: 65 Segmentation fault LC-ALL=C grep -iq"nousb"
I know RH9 is old, but lets not debate on that for now. Checked all hardware, raid, disks, even replaced RAM. So not a hardware issue. Gives same error while booting to single user mode too. Able to boot to rescue mode. Examined fstab. Has proper proc line. Grub is set up fine too, Did nothing special to trigger the error. Was getting segfaults with grep, rebooted, got stuck.
Consulted lots of experts, google, RH kbase. No go. Perhaps you mortals can save us a reinstall? Why on earth won't proc mount?
Regards, NMK.
On Sunday 17 Jun 2007 22:23:38 Nadeem M. Khan wrote: <snip>
Is your glibc fine?
On 6/17/07, Mrugesh Karnik mrugeshkarnik@gmail.com wrote:
On Sunday 17 Jun 2007 22:23:38 Nadeem M. Khan wrote:
<snip>
Is your glibc fine?
How would I check that?
Regards, NMK.
On Sunday 17 Jun 2007 23:59:04 Nadeem M. Khan wrote:
On 6/17/07, Mrugesh Karnik mrugeshkarnik@gmail.com wrote:
On Sunday 17 Jun 2007 22:23:38 Nadeem M. Khan wrote:
<snip>
Is your glibc fine?
How would I check that?
Refresh the package or something?
On Monday 18 June 2007 01:02, Mrugesh Karnik wrote:
On Sunday 17 Jun 2007 23:59:04 Nadeem M. Khan wrote:
On 6/17/07, Mrugesh Karnik mrugeshkarnik@gmail.com wrote:
On Sunday 17 Jun 2007 22:23:38 Nadeem M. Khan wrote:
<snip>
Is your glibc fine?
How would I check that?
Refresh the package or something?
I woul suspect that /lib is corrupt. I had a similiar problem on a debian install with reiserfs while testing for power outages. Also afaik ext3 is less resilient than jfs.
On 6/18/07, jtd jtd@mtnl.net.in wrote:
I woul suspect that /lib is corrupt. I had a similiar problem on a debian install with reiserfs while testing for power outages. Also afaik ext3 is less resilient than jfs.
In rescue mode, running grep gives a segfault. Lately, even running mount gives a segfault. Mount was ok till yesterday.
From what I know, proc is mounted by the kernel on /proc. All that is
required is the existence of a proc directory in /. I suspect something wrong with the kernel mechanism that mounts proc. I wanted to build a new kernel, but I don't see how that can be done from the rescue mode. The rpm command would complain of missing modules and compiling a tarball would be even more difficult.
Regards, NMK.
On Monday 18 Jun 2007 11:58:52 Nadeem M. Khan wrote:
On 6/18/07, jtd jtd@mtnl.net.in wrote:
I woul suspect that /lib is corrupt. I had a similiar problem on a debian install with reiserfs while testing for power outages. Also afaik ext3 is less resilient than jfs.
In rescue mode, running grep gives a segfault. Lately, even running mount gives a segfault. Mount was ok till yesterday.
From what I know, proc is mounted by the kernel on /proc. All that is
required is the existence of a proc directory in /. I suspect something wrong with the kernel mechanism that mounts proc. I wanted to build a new kernel, but I don't see how that can be done from the rescue mode. The rpm command would complain of missing modules and compiling a tarball would be even more difficult.
Your userspace seems broken to me. I doubt it has anything to do with the kernel.
If you notice, the reason /proc isn't being mounted by the init script is because grep segfaults. I wonder if there is something like Gentoo's revdep-rebuild for Red Hat.
I'd run an fsck to start with. If necessary, restore from the superblock backup. Then I'd reinstall glibc and the lib packages.
On 6/18/07, Mrugesh Karnik mrugeshkarnik@gmail.com wrote:
If you notice, the reason /proc isn't being mounted by the init script is because grep segfaults. I wonder if there is something like Gentoo's revdep-rebuild for Red Hat.
I replaced the grep binary from a similar system. Grep works now, but proc still doesnt mount.
I'd run an fsck to start with.
Done that yesterday. Clean.
If necessary, restore from the superblock
backup. Then I'd reinstall glibc and the lib packages.
To be on the safe side, I suggested upgrading to atleast Fedora core 5. Today mount is segfaulting, tomorrow more commands would segfault. The suggestion has been accepted. but I'm real curious. BTW. I can't use the rpm command in rescue mode, so reinstalling glibc would be difficult.
Regards, NMK.
On Monday 18 June 2007 12:39, Nadeem M. Khan wrote:
On 6/18/07, Mrugesh Karnik mrugeshkarnik@gmail.com wrote:
If you notice, the reason /proc isn't being mounted by the init script is because grep segfaults. I wonder if there is something like Gentoo's revdep-rebuild for Red Hat.
I replaced the grep binary from a similar system. Grep works now, but proc still doesnt mount.
I'd run an fsck to start with.
Done that yesterday. Clean.
If necessary, restore from the superblock
backup. Then I'd reinstall glibc and the lib packages.
To be on the safe side, I suggested upgrading to atleast Fedora core 5. Today mount is segfaulting, tomorrow more commands would segfault.
Are you sure that the moboard / PSU is ok?. Noisy PSUs including the VRs on the mobo can cause arbitary segfaults.
On 6/18/07, jtd jtd@mtnl.net.in wrote:
Are you sure that the moboard / PSU is ok?. Noisy PSUs including the VRs on the mobo can cause arbitary segfaults.
Well its an IBM xseries with light path diagnostics. No amber attention LEDs have lit up.
Regards, NMK.
On Monday 18 June 2007 15:07, Nadeem M. Khan wrote:
On 6/18/07, jtd jtd@mtnl.net.in wrote:
Are you sure that the moboard / PSU is ok?. Noisy PSUs including the VRs on the mobo can cause arbitary segfaults.
Well its an IBM xseries with light path diagnostics. No amber attention LEDs have lit up.
Those on board monitors are not designed to monitor noisy psus. and according to the linuxbios devs never trust any IPMI.
-
jtd -
Mrugesh Karnik -
Nadeem M. Khan