Re: switch block layer polling to a bio based model

[Jens Axboe <axboe@kernel.dk>]

On 4/26/21 9:06 AM, Christoph Hellwig wrote:
> On Mon, Apr 26, 2021 at 08:57:31AM -0600, Jens Axboe wrote:
>> I was separately curious about this as I have a (as of yet unposted)
>> patchset that recycles bio allocations, as we spend quite a bit of time
>> doing that for high rate polled IO. It's good for taking the above 2.97M
>> IOPS to 3.2-3.3M IOPS, and it'd obviously be a bit more problematic with
>> required RCU freeing of bio's. Even without the alloc cache, using RCU
>> will ruin any potential cache locality on back-to-back bio free + bio
>> alloc.
> 
> That sucks indeed.  How do you recycle the bios?  If we make sure the

Here's the series. It's not super clean (yet), but basically allows
users like io_uring to setup a bio cache, and pass that in through
iocb->ki_bi_cache. With that, we can recycle them instead of going
through free+alloc continually. If you look at profiles for high iops,
we're spending more time than desired doing just that.

https://git.kernel.dk/cgit/linux-block/log/?h=io_uring-bio-cache

> bio is only ever recycled as a bio and bi_bdev remaings valid long
> enough we might not need the rcu free.  Even without your recycling
> we could probably do something nasty using SLAB_TYPESAFE_BY_RCU.

It would not be hard to restrict to same bdev for the cache, just one
more check to do for recycling.

Note that the caching series _only_ supports polled IO for now, as
non-polled would require IRQ juggling for free+alloc and that will
definitely take some of the win away and maybe even render it moot.
Have yet to test that part out. Not a huge deal with the RCU free, as
you end up doing that purely for polled IO and hence wouldn't impact the
IRQ side of things negatively.

-- 
Jens Axboe