re: CVS commit: src/sys/arch/amiga/dev

["Michael L. Hitch" <mhitch%NetBSD.org@localhost>]

On Sun, 11 Jan 2009, matthew green wrote:

> the MI esiop driver seems to have the same problem.  before i had
> interrupts working on the sunblade 2000, esiop would hang my system
> when it was probing disks.  other cards gave errors or hung for a
> while, but alawys returned.

  All the MI *siop drives have a timeout routine, although it doesn't
look to me like it would work in oosip (it doesn't reset anything,
and just just calls the done routine).

  I saw the same problem on a card with a dual esiop and tulip interface
behind a bridge on an alpha.  The interrupts for the second esiop and
the tulip weren't being configured properly and would hang trying to
probe the second esiop.  (The tulip would fail, but didn't hang the
system).

  A quick look at esiop seems that there is no timeout checking in polled 
mode, so it would hang if there are no interrupts.  It looks like osiop
does do timeout checking in polled mode, but oosiop doesn't.

  Hmm, it appears that I also had timeout checking during polled io in
the Amiga siop drivers - just not for non-polled io.

--
Michael L. Hitch                        mhitch%NetBSD.org@localhost