io_uring: fix issue with io_write() not always undoing sb_start_write()
authorJens Axboe <axboe@kernel.dk>
Fri, 24 Jun 2022 16:24:45 +0000 (10:24 -0600)
committerJens Axboe <axboe@kernel.dk>
Mon, 25 Jul 2022 00:39:32 +0000 (18:39 -0600)
commite053aaf4da56cbf0afb33a0fda4a62188e2c0637
tree4dbe314803d0eddfdcea133396452dbb977c03b2
parent4e17aaab54359fa2cdeb0080c822a08f2980f979
io_uring: fix issue with io_write() not always undoing sb_start_write()

This is actually an older issue, but we never used to hit the -EAGAIN
path before having done sb_start_write(). Make sure that we always call
kiocb_end_write() if we need to retry the write, so that we keep the
calls to sb_start_write() etc balanced.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_uring/rw.c