Nevermind, problem solved. I was running into ENOSPC on the host file
system while it tried to write to the backing store (a sparse zvol). The
cause wasn't immediately apparent because the zpool had plenty of space.
But there were some reservations on other datasets that starved the zvol of
usabl
Lately whenever I try to buildworld on my main development VM, I get
"vtbd0: hard error" errors, and inevitably UFS panics soon afterwards.
Does anybody else see this? vtbd0 is backed by a zvol that apparently has
no errors, and there's nothing in the host's dmesg. What else
