[This has been sitting in my drafts folder for a while; sorry for the late response.]
On Thu, 2012-03-22 at 13:12 +0100, Rik Theys wrote: > Hi, > > On 03/21/2012 05:27 PM, Jonathan Nieder wrote: > >> When booting with the quiet option off, I see that the lvm command in the > >> initrd > >> segfaults. > > > > Can you get a backtrace, for example by saving a core file to some > > other (virtual) disk and analyzing it afterward with gdb? > > > > See http://wiki.debian.org/InitramfsDebug for some hints. > > I captured the boot via the serial console and provided the "debug" > command line parameter to get into the initramfs prompt. > > It seems it's not only lvm that segfaults, but also "ls" etc in the > initramfs. > > In attach the log of my session. Which shows that the segfault is always at the same code address: [ 56.663596] lvm[540]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007fff25461ec8 error 5 [ 76.174282] exe[541]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007fffa69b3388 error 5 [ 78.307062] exe[542]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007fff33270d08 error 5 [ 87.775183] exe[543]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007ffffb125068 error 5 [ 97.937356] exe[545]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007fffb53be498 error 5 [ 108.789157] lvm[547]: segfault at ffffffffff600400 ip ffffffffff600400 sp 00007fff0e012348 error 5 This address is not accessible in user-mode, and probably isn't used by the kernel either. > Would unpacking the initramfs on a sid box with the 3.3 kernel and > chrooting into it be able to provide any additional info? I believe the > initramfs uses another C library than glibc, or is that no longer the case? You later wrote: > I updated the initramfses for all kernels on the system (2.6.32, 3.2, > 3.3). The new 3.3 image still has the segfaults, the others boot OK. > > I ran a diff between the update-initramfs -v output from a working and > non-working kernel but I can't spot a difference. So this certainly seems to be a bug in the kernel or possibly the hypervisor. Ben. -- Ben Hutchings This sentence contradicts itself - no actually it doesn't.
signature.asc
Description: This is a digitally signed message part
