On Tue, Dec 18, 2018 at 08:42:10AM -0700, Jens Axboe wrote:
> For the upcoming async polled IO, we can't sleep allocating requests.
> If we do, then we introduce a deadlock where the submitter already
> has async polled IO in-flight, but can't wait for them to complete
> since polled requests must be active found and reaped.
So two comments here. First I think after the next -rc1 we really
need to rename REQ_HIPRI to REQ_POLLED. Unlike the iocb hipri flag
which is an advice to treat this I/O as urgent REQ_POLLED is a hard
charaterisation and having the hipri name is badly misnamed for that.
Second I retract my suggestion to add this _ASYNC variant, as it seems
to cause more confusion than it helps. I think instead we want
a helper like this:
static void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
{
bio->bi_opf |= REQ_POLLED;
if (!is_sync_kiocb(kiocb)
bio->bi_opf |= REQ_NOWAIT;
}
maybe plus comments to better indicate the intentional and consolidate
the is_sync check.
