On Wed, Aug 22 2001, Jes Sorensen wrote:
> >>>>> "Adam" == Adam Radford <aradford@3WARE.com> writes:
>
> Adam> Thanks for flags type fix and redundant pushf/popf fix.
> Adam> Regarding your question about the error handling routines, the
> Adam> 3ware driver uses the new style scsi error handling, so looking
> Adam> through scsi_error.c, all calls to
> Adam> hostt->eh_abort_handler() and hostt->eh_host_reset_handler() are
> Adam> wrapped with a spin_lock_irqsave(&io_request_lock, flags) and
> Adam> spin_unlock_irqrestore(&io_request_lock, flags) so they should
> Adam> be okay.
>
> Hmm ok. However, since Jens is working on killing the io_request_lock
> so something will need to get done when that happens.
>
> Jens, what is the strategy for that?
Lots of the lower level hooks are done with io_request_lock required,
the only one I've killed so far is ->detect() and that was mainly
because it didn't fit with the new scheme (per-host locking, host not
inited at this time). IMO it was a mistake to ever grab io_request_lock
(or any other lock, for that matter) before calling into the lower
levels -- it's not very clean and it makes progressing to a better
locking structure harder.
Basically we want to move the locking down a notch, so the mid layer is
not responsible for providing reentrance protection for the low level
drivers. It will be painfull...
-- Jens Axboe- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Aug 23 2001 - 21:00:54 EST