new lockmgr leak

From: Kris Kennaway <kris_at_obsecurity.org>
Date: Sat, 1 Oct 2005 13:36:25 -0400
This machine was running HEAD from a week or two ago.  It does not use
nullfs, which was a previous source of lockmgr leaks (now might be
fixed, although I've seen indications they're not completely fixed).
It only mounts ufs filesystems and nfs, and none of these are
unmounted at runtime..so the source of the leak must be elsewhere.
Unfortunately, the ddb backtrace appears to be corrupted:

panic(c07125c0,2,0,0,0) at panic+0x60
userret(c61acaf0,f0530d38,9f,424,3) at userret+0xf3
syscall(3b,69a003b,bfbf003b,0,66efe00) at syscall+0x3c6
Xint0x80_syscall() at Xint0x80_syscall+0x1f
--- syscall (1024), eip = 0x28149a97, esp = 0xbfbfbd6c, ebp = 0xbfbfc288 ---

syscall 1024?  No such animal.

0xc07125c0:     userret: Returning with %d locks held.

The thread with the leak was:

41187 c621ea3c 4203 41186 41184 0004002 [CPU 0] md5

This was running md5 over some package files..about all I can think
might have happened is something removed a package at the time it was
being checksummed, but this shouldn't of course cause any problems.

Unfortunately I can't even take a dump on this machine, since twe is
broken.  So this probably isn't a very useful bug report, but it's
here in case it happens again to someone else.

Kris

Received on Sat Oct 01 2005 - 16:07:45 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:44 UTC