From: Jens Axboe axboe@kernel.dk
[ Upstream commit 1bacd264d3c3a05de4afdd1712c9dd6ccebb9490 ]
If we mark for reissue, we assume that the buffer will remain stable. Hence if are using a provided buffer, we need to ensure that we stick with it for the duration of that request.
This only affects block devices that use provided buffers, as those are the only ones that get marked with REQ_F_REISSUE.
Signed-off-by: Jens Axboe axboe@kernel.dk Signed-off-by: Sasha Levin sashal@kernel.org --- fs/io_uring.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c index 68aab48838e4..725c59c734f1 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -3009,7 +3009,7 @@ static bool __io_complete_rw_common(struct io_kiocb *req, long res) if (unlikely(res != req->result)) { if ((res == -EAGAIN || res == -EOPNOTSUPP) && io_rw_should_reissue(req)) { - req->flags |= REQ_F_REISSUE; + req->flags |= REQ_F_REISSUE | REQ_F_PARTIAL_IO; return true; } req_set_fail(req); @@ -3059,7 +3059,7 @@ static void io_complete_rw_iopoll(struct kiocb *kiocb, long res) kiocb_end_write(req); if (unlikely(res != req->result)) { if (res == -EAGAIN && io_rw_should_reissue(req)) { - req->flags |= REQ_F_REISSUE; + req->flags |= REQ_F_REISSUE | REQ_F_PARTIAL_IO; return; } req->result = res;