Red Hat Bugzilla – Bug 202173
Last modified: 2007-11-30 17:11:40 EST
# pvcreate /dev/sdb
(gdb) run /dev/sdb
Starting program: /usr/sbin/pvcreate /dev/sdb
Reading symbols from shared object read from target memory...done.
Loaded system supplied DSO at 0x94a000
File descriptor 3 left open
File descriptor 4 left open
File descriptor 5 left open
Program received signal SIGSEGV, Segmentation fault.
0x00000000 in ?? ()
#0 0x00000000 in ?? ()
#1 0x0807304e in dev_create_file (filename=0x8b4c7d0 "/etc/lvm/lvm.conf",
dev=0x0, alias=0x8b471a0, use_malloc=1)
#2 0x080728ce in read_config_file (cft=0x8b4c7b8) at config/config.c:237
#3 0x0806f7a3 in _load_config_file (cmd=0x8b47690, tag=0x80b55d3 "") at
#4 0x0806fa3f in _init_lvm_conf (cmd=0x8b47690) at commands/toolcontext.c:377
#5 0x08070474 in create_toolcontext (the_args=0x80c5760) at
#6 0x0805734c in lvm2_main (argc=2, argv=0xbfd83034, is_static=0) at
#7 0x0806b3ea in main (argc=1735287144, argv=0x11) at lvm.c:20
It segfaults without the /dev/sdb too.
Please confirm exact versions, which kernel & glibc packages, hardware platform
and whether there's anything unusual like xen involved; also run it under 'strace'.
also attach a copy of /etc/lvm/lvm.conf to this bugzilla
Created attachment 134035 [details]
I haven't edited lvm.conf, it's stock (8ca8f0cdab06cf78cf57bbc997357f60)
Which device-mapper package - 1.02.02-3.2 ?
Try running under 'ltrace' too and attach the output.
Is selinux enabled? Enforcing?
Does 'lvm version' work?
Does 'lvm' (with no args) give you an lvm shell prompt?
Does pvcreate work if you prefix the command with 'lvm.static'? ('lvm.static
what does 'ldd /usr/sbin/pvcreate' say?
which architecture (i.e. the 'Hardware:' field on this bugzilla - what cpu?)
[I have same packages & lvm.conf as you have installed here - though different
kernel - and it works for me. So still need to work out the essential
difference in your system.]
Here you go:
# rpm -q device-mapper
# ltrace lvm
ltrace: Couldn't find .dynsym or .dynstr in "/sbin/lvm"
# selinuxenabled ; echo $?
# lvm version
# ldd $(which lvm)
not a dynamic executable
# lvm.static pvcreate
Please enter a physical volume path
# ldd /usr/sbin/pvcreate
linux-gate.so.1 => (0x0094a000)
libdevmapper.so.1.02 => /lib/libdevmapper.so.1.02 (0x00aa2000)
libreadline.so.5 => /usr/lib/libreadline.so.5 (0x04200000)
libselinux.so.1 => /lib/libselinux.so.1 (0x03f88000)
libdl.so.2 => /lib/libdl.so.2 (0x00ac4000)
libncurses.so.5 => /usr/lib/libncurses.so.5 (0x03fc8000)
libc.so.6 => /lib/libc.so.6 (0x00968000)
libsepol.so.1 => /lib/libsepol.so.1 (0x03f43000)
# head -5 /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 15
model : 3
model name : Intel(R) Pentium(R) 4 CPU 3.00GHz
Ah. I have vmware and nvidia both installed on this box.. want to leave this, or?
Try those commands with explicit path: /usr/sbin/lvm
(only interested in ltrace from a command that segfaults, by the way)
If it's related to vmware or nvidia, it would still be good to know that was the
So far, you seem to have shown that 'lvm.static' is working OK.
(that's the statically-linked version of the executable used during boot)
Created attachment 134038 [details]
(In reply to comment #11)
> Try those commands with explicit path: /usr/sbin/lvm
dm_strdup(0x9dda7d0, 0x80b4736, 65, 0xaaa63b, 0x80b4736 <unfinished ...>
--- SIGSEGV (Segmentation fault) ---
Is that related to
possibly - you have a different device-mapper rpm which must not be compatible
So missing a requires in the rpm then.
The FC-5 one is still:
If you want to use the newer one, you need to use the corresponding new lvm2
package as well.
The rpm says:
Install Date: Tue Jul 4 09:59:26 2006
# zgrep device-mapper /var/log/yum.log*
/var/log/yum.log.1:Jul 04 09:59:39 Updated: device-mapper.i386 1.02.07-2.0
but doesn't show the repo. Where on earth did it come from, if it's not in FC5?
Ah from FC4. This is an upgrade box. I will sort it out.
Thanks very much.
It works now. I did:
# rpm -e device-mapper lvm2 --nodeps
# yum install device-mapper lvm2
the new dm package should probably have a conflicts line or similar
Ah yes - we did release an update for fc4 because that package was too old to
handle the latest kernels. We could release it for fc5 too if people wanted it
- it's not necessary, but it might avoid problems like this hitting other people.
and I suppose releasing an update to the dm package with a conflict would cause
a mess. This bug will probably show up by the lvm gui tool failing to work.