I can't reproduce the problem. The sequence of comands that failed
repeatedly are are now working as I expect. The cheap'o thumdrive is
indeed slow. mkfs took just under 15min and didn't generate the journal
entry. The device busy message from cryptsetup is gone.
Between the bugreport and re-testing I updated my system, rebooted and
copied /dev/zero to the thumbdrive. So I don't know what may have caused
the 'blocked for more than 120 seconds' journal entry. I see that the
system was in fsync(). I monitored the meminfo.writeback with /proc and
could see it going down while the mkfs was in the "writing superblocks
and filesystem ..." step.
- Greg
