From 295219ab7d6250e56cba480cb8b46302401d9cce Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 25 Aug 2022 10:19:08 -0600 Subject: [PATCH] io_uring: fix issue with io_write() not always undoing sb_start_write() commit e053aaf4da56cbf0afb33a0fda4a62188e2c0637 upstream. 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 Signed-off-by: Greg Kroah-Hartman --- fs/io_uring.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/io_uring.c b/fs/io_uring.c index 89f24b54fe5e8..2680e9756b1d4 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -3720,7 +3720,12 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags) copy_iov: iov_iter_restore(iter, state); ret = io_setup_async_rw(req, iovec, inline_vecs, iter, false); - return ret ?: -EAGAIN; + if (!ret) { + if (kiocb->ki_flags & IOCB_WRITE) + kiocb_end_write(req); + return -EAGAIN; + } + return ret; } out_free: /* it's reportedly faster than delegating the null check to kfree() */