HEADS UP [Re: thread+preemption stability improvement]

[Daniel Eriksson]

Pawel Jakub Dawidek wrote:

> Could you give me a procedure which I can use to reproduce it?

Not really. The use case is so simple that I really don't think it is a
general problem. Otherwise everyone would be complaining about it. I'll see
if I can schedule some more downtime for the server to run some more tests
of this. (I only did it once, and because the machine was live at the moment
the lockup of the ataraid arrays made it quite a mess, so I'm hesitant to
try it again unless I'm in single user mode.)

This was the command I issued that resulted in the crash. Nothing strange
about it as far as I can tell (taken directly from the gstripe man page).
gstripe label -v -s 4096 data /dev/da15 /dev/da16

/dev/da15 and /dev/da16 had been labeled and newfs'ed, but were not mounted
at the time.

As I said, if I get a chance I'll look at this next time the machine is down
in single user mode. I forgot to do that earlier this morning when I did a
lot of testing to try to get around the SATA vs. preemtion problem I'm
seeing.

/Daniel Eriksson