io_uring: fix IOPOLL -EAGAIN retries
authorJens Axboe <axboe@kernel.dk>
Thu, 27 Aug 2020 22:40:19 +0000 (16:40 -0600)
committerJens Axboe <axboe@kernel.dk>
Thu, 27 Aug 2020 22:40:29 +0000 (16:40 -0600)
commiteefdf30f3dcb5c1d47bee2b3afdb9d4d05343ff3
treee5811c224ae3d95cc27d3a0d03142f07dc5fdca0
parent56450c20fe10d4d93f58019109aa4e06fc0b9206
io_uring: fix IOPOLL -EAGAIN retries

This normally isn't hit, as polling is mostly done on NVMe with deep
queue depths. But if we do run into request starvation, we need to
ensure that retries are properly serialized.

Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
fs/io_uring.c