> i just converted a server over to centos 4 and the rsync daemon is dying
> with a segv. i've checked the redhat bugzilla and googled around some and
> don't see anything obvious.
>
> the client connects and authenticates, and the server starts sending some
> data but then dies after about 8k. strace'ing hasn't shown anything
> obvious other than the segv.
another user was able to confirm similar problems and i finally started
digging around with gdb and friends. what i have determined is that this
is probably a bug in libc, with some help from rsync. the problem seems
to be that rsync is chroot'ing and trying to map uids to names for the -o
option (to preserve ownership). this only occurs if files are owned by
someone other than root and the remote user has requested -o. in my
particular case, i normally use nis and rsync is reading that from
nsswitch.conf before chrooting. when it tries to map uids, it tries
valiantly to find libnss_nis.so in the chroot area and fails and that
seems to cause a segfault. if rsync chroot'ed earlier, this might not
happen, or if i wasn't using nis it might not happen. i haven't been able
to test the latter.
what i was able to do was chown everything to root since then rsync short
circuits the getpwuid and doesn't crash. this is fine for now. i'll be
contacting the rsync maintainer to see if he has any input.
