Re: block device direct I/O fast path

From: Jens Axboe
Date: Tue Nov 01 2016 - 14:22:49 EST


On 11/01/2016 11:54 AM, Jens Axboe wrote:
On Tue, Nov 01 2016, Christoph Hellwig wrote:
On Tue, Nov 01, 2016 at 11:00:19AM -0600, Jens Axboe wrote:
#2 is a bit more problematic, I'm pondering how we can implement that on
top of the bio approach. The nice thing about the request based approach
is that we have a 1:1 mapping with the unit on the driver side. And we
have a place to store the timer. I don't particularly love the embedded
timer, however, it'd be great to implement that differently. Trying to
think of options there, haven't found any yet.

I have a couple ideas for that. Give me a few weeks and I'll send
patches.

I'm not that patient :-)

For the SYNC part, it should be easy enough to do by just using an
on-stack hrtimer. I guess that will do for now, since we don't have
poll support for async O_DIRECT right now anyway.

http://git.kernel.dk/cgit/linux-block/commit/?h=for-4.10/dio&id=e96d9afd56791a61d463cb88f8f3b48393b71020

Untested, but should get the point across. I'll fire up a test box.

http://git.kernel.dk/cgit/linux-block/commit/?h=for-4.10/dio&id=b879a97d749c5c1755d1e2f20c721eb6fde0c291

Now tested, inadvertently used the absolute timer instead of the
relative. Works for me.

--
Jens Axboe