panic: lockmgr: locking against myself, 5.X

[Jonathan Fosburgh]

Yesterday (Thursday) I updated my sources so I could upgrade from 5.0- 
RELEASE to 5.1-BETA. The buildworld and buildkernel went fine, and while 
the kernel build was finishing I went home for the day.  I came back this 
morning to find that, after the build had finished and I was sitting at a 
prompt, I had a panic in lockmgr.  Unfortunately, most of the panic message 
was truncated and I did not have a debug kernel built at that time.  This 
was a fresh install of 5.0-R I performed on a clean, brand new HD (Seagate) 
maybe two weeks ago and I had had no panics on it before that time.  I 
rebooted the system and after it passed through the point where it mounts 
filesystems and schedules background fsck's I received the same panic, and 
additionally I received a second panic when it was trying to sync the 
disks.  The second panic is: syncing disks, buffers remaining... panic: 
bdwrite: buffer is not busy.

I rebooted again into single user mode and manually fsck'd the filesystems. 
 My filesystems are UFS2 with softupdates.  After completing the fsck and 
mounting the filesystems, I typed exit from the prompt and let the system 
continue to boot, only to receive the exact same panic.  When I manually 
fsck the filesystems the next steps I see are a Fastboot message and the 
system attempting to gather entropy.

I have tried this without softupdates as well, and also disabling 
background fscks, all to no avail.  Finally, I decided to proceed with the 
upgrade and install my kernel.  This went fine, and I rebooted into single 
user mode fine.  I ran mergemaster -p and installworld, all without 
trouble. However, when running mergemaster after the install I get the same 
panic, this time while it is setting up the temporary root environment.  I 
proceeded to build a debug kernel and just allowed the system to boot to 
multi-user mode.  As expected, I get the same set of panics in the same 
place during boot.  Dropping into the debugger I run trace and get a 
pageful of output.  I will provide the full output of trace if it is 
necessary, but as I am transcribing it by hand I copied down the first two 
lines and the last one, anticipating those are probably the most 
interesting:

ffs_balloc_ufs2 (c351bb68, 24000, 0, 4000, clldfe80) at 
ffs_balloc_ufs2+0xff8
ffs_copyonwrite (c34af920, cae5f3do, cae5f3d0, d1dea740, c0271b87) at 
ffs_copyonwrite+0x443
...
--- syscall (5, FreeBSD ELF32, open), eip= -x807b2b3, esp = 0xbfbff5ec, ebp 
= 0xbfbff678 ---

Interestingly, at this point my second panic has changed to:

syncing disks, buffers remaining ... panic: ffs_coponwrite : recursive call

It has been some time since I have tried to debug a kernel panic so I am 
rusty on this.  I've searched through the mailing lists so I can find a way 
to generate a dump and try to use gdb to debug the kernel.

My FreeBSD installation is currently unbootable except to single-user mode. 
 This email is being sent from Windows 2000 installed on a different disk.  
Can anyone help me with this problem? I haven't seen anyone, at least 
recently, with a problem quite like this, though I have seen several people 
receive the "locking against myself" panic before.  I am going to try to 
get a kernel dump now and hopefully I will be able to provide more info 
soon.


-- 
Jonathan Fosburgh
Software Systems Specialist IV
AIX/SAN Administrator
University of Texas MD Anderson Cancer Center
Houston, TX