panic: initiate_write_inodeblock_ufs2: already started

I upgraded a few days ago from a year-old RELENG_7 to a fresh RELENG_7
(right before
the 7.1 release, basically), and then started an upgrade to CURRENT but noticed
that my gmirror was degraded when booting to single-user.
It had been degraded for a few days, corresponding roughly
to when I physically moved the machine to a different room.

I hesitated for a while before sending this, because there are so many
correlated things going on, and it's hard to rule out hardware problems.
However, I have now seen this panic a couple times; once, the
panic string was interleaved with "ad2: FAILURE - load data"
and another where the string was interleaved with "ad0: FAILURE - load data",
so I am using the working hypothesis that this failure is not
tied to a particular one of the disks.

Such sparse data as I have collected is up here:
http://stuff.mit.edu/afs/sipb.mit.edu/user/kaduk/freebsd/periphrasis/20090107/panic.txt
I have seen three four panics so far; I was in X for the first, and thus
have no data for it; the last three are in the above file.
I was only able to get to KDB for the middle of the three, and I sadly
did not print the panicstr from it, so I'm not sure if it's the same
panic as the subject line or not.  The machine had just errored
about failing to set up DMA for both disks, so I was unable to
get a dump :(

I'll try to see if I can see if there's a difference between
kernel.old and kernel.

-Ben Kaduk