Re: dm-crypt parallelization patches

From: Tejun Heo
Date: Tue Apr 09 2013 - 17:02:13 EST


Hey,

On Tue, Apr 09, 2013 at 04:32:28PM -0400, Mikulas Patocka wrote:
> > and we probably need to change that to bio_associate_task().
>
> Generally, we shouldn't associate bios with "current" task in device
> mapper targets. For example suppose that we have two stacked dm-crypt
> targets:

It only follows the first association so it doesn't matter how many
layers it goes through. That said, yeah, there could be situations
where @task is availalbe but the bio's already in the hand of a
different task. If that's the case, change it to
associate_task(@task).

> It doesn't work because device mapper on the underlying layers ignores
> bi_ioc and bi_css.
>
> If I make device mapper forward bi_ioc and bi_css to outgoing bios, it
> improves performance (from 2:30 to 1:30), but it is still far from
> perfect.

For testing, copying bi_ioc and bi_css directly is fine but please add
another interface to copy those for the actual code. Say,
bio_copy_association(@to_bio, @from_bio) or whatever.

As for the performance loss, I'm somewhat confident in saying the
remaining difference would be from ignoring plugging boundaries.

Thanks.

--
tejun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/