Re: [PATCH v3] block: move non sync requests complete flow to softirq

From: Christoph Hellwig
Date: Wed Sep 04 2024 - 00:29:43 EST


On Tue, Sep 03, 2024 at 11:49:28AM -0600, Jens Axboe wrote:
> The elephant in the room here is why an 80M completion takes 100 msec?
> That seems... insane.
>
> That aside, doing writes that big isn't great for latencies in general,
> even if they are orders of magnitude smaller (as they should be). Maybe
> this is solvable by just limiting the write size here.
>
> But it really seems out of line for a write that size to take 100 msec
> to process.

pagecache state processing is quite inefficient, we had to limit
the number of them for XFS to avoid latency problems a while ago.
Note that moving to folios means you can process a lot more data
with the same number of completion iterations as well. I'd suggest
the submitter looks into that for whatever file system they are using.