amd64/129426: FreeBSD 7.0 crash after subdiskXX: detached

[Mark Kamichoff]

>Number:         129426
>Category:       amd64
>Synopsis:       FreeBSD 7.0 crash after subdiskXX: detached
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    freebsd-amd64
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Thu Dec 04 23:20:01 UTC 2008
>Closed-Date:
>Last-Modified:
>Originator:     Mark Kamichoff
>Release:        FreeBSD 7.0-RELEASE-p1
>Organization:
>Environment:
FreeBSD dax.prolixium.com 7.0-RELEASE-p1 FreeBSD 7.0-RELEASE-p1 #0: Thu May 22 22:20:18 EDT 2008     prox at dax1.prolixium.com:/usr/obj/usr/src/sys/DAX  amd64
>Description:
Hi - 

Every 30-60 days my FreeBSD-7 server will crash with the following output:

subdisk12: detached
ad12: detached
g_vfs_done():ad12s1d[WRITE(offset=160650805248, length=16384)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=67040542720, length=16384)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=80158326784, length=16384)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=160691107840, length=2048)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=155648786432, length=2048)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=6144000, length=16384)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=6160384, length=6144)]error = 6
g_vfs_done():ad12s1d[WRITE(offset=65536, length=2048)]error = 6
g_vfs_done():ad12s1d[READ(offset=160702988288, length=16384)]error = 6
g_vfs_done():ad12s1d[READ(offset=160710213632, length=16384)]error = 6
panic: vinvalbuf: dirty bufs
cpuid = 0
Uptime: 30d8h8m29s
Physical memory: 998 MB
Dumping 316 MB: 301 285 269 253 237 221 205 ... <serial console output hangs>

This has been happening since the box was built about two years ago, initially running FreeBSD 6.1.  I've replaced all the hardware in this box recently (including the chassis) and upgraded to FreeBSD 7.0.  The crashes are all the same, varying only in the disk that becomes detached (subdisk12 or subdisk8 - either 250GB SATA disk).

I haven't been able to obtain a dump of physical memory, since the machine always hard locks halfway through the dump operation.  Since this is a dedicated server at a colocation facility, I don't have physical access to the server - only serial console.

Please let me know what information to provide.  Since I have replaced all the hardware in the machine at this point, I believe this can be eliminated as a possible culprit.

Thanks!

- Mark
>How-To-Repeat:
Unknown.  Box has never attained uptime for more than 100-120 days, though.
>Fix:
Unknown.

>Release-Note:
>Audit-Trail:
>Unformatted: