add mrsas_shutdown method

[avg]

Summary

It should be safer to flush controller and disk caches on the shutdown.
And to gracefully shut down the controller as well.

[avg]

• update

[imp]

sys/dev/mrsas/mrsas.c

1197 ↗

(On Diff #55797)

is there an upper bound here? And why only for panicstr != NULL? I get it's only while we're panicing, but why do we have to do this when we panic and not otherwise.

[avg]

sys/dev/mrsas/mrsas.c

1197 ↗

(On Diff #55797)

I copied this code verbatim from mrsas_detach, so there is no upper limit in both places.

And it's actually the other way around, we do this only when we are not panic-ing. This code wakes another thread and waits for it to clear a flag. But the other thread, of course, cannot run if we are in panic.

[avg]

remove a wrong panicstr check that was left by accident

[imp]

This looks good to me, however I'd wait a few days to see if Scott can spot something that I can't.

sys/dev/mrsas/mrsas.c

1197 ↗

(On Diff #55797)

I'm glad to see you got rid of the panicstr here, that's what I was really commenting about.

[avg]

In D19817#426388, @imp wrote:

This looks good to me, however I'd wait a few days to see if Scott can spot something that I can't.

I also sent a link to this review to Kashyap Desai.
Waiting for his / Broadcom's reply as well.

[avg]

sys/dev/mrsas/mrsas.c

1197 ↗

(On Diff #55797)

I was so sure that I removed it that I just didn't see it.

[scottl]

sys/dev/mrsas/mrsas.c

1199 ↗

(On Diff #55992)

This can be done with pps_ratecheck() instead.

1209 ↗

(On Diff #55992)

This isn't incorrect, but I'm not comfortable with mesas_shutdown_ctlr() having an unbounded timeout.