looping symlink = reboot ?
im not sure who to mail on this, so i figure best to
start with a general mailing list for some pointers.
Basic problem, never happened to me before, on 2
of my systems, when attempting to access a symlink
that points back to itself crashes the system. either
a reboot(4/5 cases) or a lockup(1/5). I first
encountered it today when trying to scp a directory
tree to another system. caused a reboot on the
system doing the scp'n. Then i rsync'd the same
directory tree to another system and it worked fine.
then i attempted another scp -r and it rebooted
the system. so i tried it again, and it rebooted
the system at the exact same point. i went to
the console and did it again and it locked up
hard. i scp'd it to another system and it rebooted
again.
eventually i traced it down to these particular
2 systems(i don't have many systems to crash
to test). It hit me when i copied the directory
tree to an ext2 partition(from reiserfs) and
tried a scp -r and it worked fine. i noticed
at the EXACT point where it crashed on reiserfs,
the ext2 filesystem spit out a buncha
"too many levels of symbolic links" or something.
so i looked at the file and it was a simple
symlink back to itself:
blah -> blah
i tried to do a ls -l bla<TAB FILENAME COMPLETION>
system rebooted again.
so i thought it was reiserfs. so i ssh'd to a machine
at home and tried it, no crash, no errors. behaved
as expected(same as ext2). so i thought maybe it
was the 3ware raid driver(!), so i went to another
system that has a raid1 array(opposed to raid10
on the affected systems) and it also uses reiserfs,
and no problem.
all of them are running virtually identical kernels,
all are 2.2.19 with the same reiserfs patch
3.something. all machines are running potato.
no other problems otherwise(been using the same
kernels since june). the only differences are
minor configuration changes depending on
the hardware in the machine and it's purpose.
im at a loss as to who i should attempt to
contact to attempt to find a resolution to
this problem.
so i suppose my next test will be to
re allocate the swap partition on the raid
array as an ext2 filesystem and see if it
locks up.
has anyone ever heard of a controller driver
possibly affecting filesystem access like this?
oh, and there is no logs, the console showed
nothing when i was at it(locked hard), and
i have it logging to a remote syslog
server, and it shows nothing.
oh and as for why there is a looping symlink
i don't know. i didn't create the directory
tree, i was just trying to migrate it to another
machine.
nate
Reply to: