Red Hat Bugzilla – Full Text Bug Listing
|Summary:||init segfaults at boot on system upgraded from F8|
|Product:||[Fedora] Fedora||Reporter:||Will Woods <wwoods>|
|Component:||mkinitrd||Assignee:||Peter Jones <pjones>|
|Status:||CLOSED RAWHIDE||QA Contact:||Fedora Extras Quality Assurance <extras-qa>|
|Version:||rawhide||CC:||cookiecaper, dcantrell, katzj, lsof, notting, wtogami|
|Fixed In Version:||Doc Type:||Bug Fix|
|Doc Text:||Story Points:||---|
|Last Closed:||2008-04-17 02:29:00 EDT||Type:||---|
|oVirt Team:||---||RHEL 7.3 requirements from Atomic Host:|
|Bug Depends On:|
Description Will Woods 2008-04-01 14:16:01 EDT
I upgraded my system from F8 to rawhide using preupgrade. At boot I get the following message from init: init: segfault at d ip 004a266b sp bfa52b60 error 4 in libc-2.7.90.so[48c000+16d000] Bootup hangs at this point.
Comment 1 Bill Nottingham 2008-04-01 14:17:27 EDT
What happens if you boot with enforcing=0/selinux=0? If you boot with init=/bin/bash and do: rpm -V upstart event-compat-sysv what do you see? (In reply to comment #0) > I upgraded my system from F8 to rawhide using preupgrade. > > At boot I get the following message from init: > > init: segfault at d ip 004a266b sp bfa52b60 error 4 in > libc-2.7.90.so[48c000+16d000] > > Bootup hangs at this point.
Comment 2 Will Woods 2008-04-01 14:27:39 EDT
All combinations of those boot args give slightly different SPs but that's it. e.g. selinux=0 init=/bin/bash: init: segfault at d ip 004a266b sp bfacb960 error 4 in libc-2.7.90.so[48c000+16d000] Trying to get a working rescue CD so I can poke this machine a bit further.
Comment 3 Bill Nottingham 2008-04-01 15:02:37 EDT
Well, that eliminates upstart as the culprit.
Comment 4 Jeremy Katz 2008-04-01 15:58:46 EDT
Is that init or is it /init from the initrd (in which case, nash)?
Comment 5 Bill Nottingham 2008-04-01 16:28:44 EDT
Poked at the box. It's the return of the 'mkinitrd chooses mismatching ld.so and libc' problem.
Comment 6 Jeremy Katz 2008-04-01 19:37:08 EDT
Blargh. This is because ld-2.7.90.so sorts before ld-2.7.so in shell sort. But we don't want to change the sort to be reversed as that would break things for ld-2.7 -> ld-2.8. Not really sure if there's any better heuristic to use off-hand :-/
Comment 7 Bill Nottingham 2008-04-02 11:35:24 EDT
Can't we just readlink ld-linux.so.2?
Comment 8 Jeremy Katz 2008-04-02 11:53:24 EDT
The name isn't consistent across arches... it's ld-linux-x86_64.so.2 on x86_64, ld-linux-ia64.so.2 on itanium, ld.so.1 on ppc and ld64.so.1 on ppc64. I didn't feel motivated enough to find an s390 or a sparc at that point. And keeping a hardcoded list in mkinitrd per-arch feels like loss
Comment 9 Bill Nottingham 2008-04-02 14:21:56 EDT
You don't need a list, you can just ldd something...
Comment 10 Adam Jackson 2008-04-03 14:51:35 EDT
(In reply to comment #9) > You don't need a list, you can just ldd something... Like say, the libc you're about to install.
Comment 11 Dave Jones 2008-04-03 21:06:20 EDT
*** Bug 440514 has been marked as a duplicate of this bug. ***
Comment 12 Dave Jones 2008-04-04 13:07:51 EDT
*** Bug 440583 has been marked as a duplicate of this bug. ***
Comment 13 Need Real Name 2008-04-04 13:17:29 EDT
Comment 14 Will Woods 2008-04-04 13:25:47 EDT
Yes, anything that recreates the initrd (e.g. installing a new kernel) post-upgrade will work fine. But the problem happens when you run mkinitrd *during* the upgrade.
Comment 15 Peter Jones 2008-04-07 10:46:22 EDT
Can you test this with 6.0.42 ? (to be built very soon...)
Comment 16 Jeremy Katz 2008-04-17 02:29:00 EDT
Just did an upgrade from F8 and things are fine now