io_uring: allow retry for O_NONBLOCK if async is supported

A common complaint is that using O_NONBLOCK files with io_uring can be a
bit of a pain. Be a bit nicer and allow normal retry IFF the file does
support async behavior. This makes it possible to use io_uring more
reliably with O_NONBLOCK files, for use cases where it either isn't
possible or feasible to modify the file flags.

Cc: stable@vger.kernel.org
Reported-and-tested-by: Dan Melnic <dmm@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

fs/io_uring.c

@@ -2843,7 +2843,8 @@ static bool io_file_supports_nowait(struct io_kiocb *req, int rw)
 	return __io_file_supports_nowait(req->file, rw);
 }
 
-static int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+static int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe,
+		      int rw)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 	struct kiocb *kiocb = &req->rw.kiocb;
@@ -2865,8 +2866,13 @@ static int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	if (unlikely(ret))
 		return ret;
 
-	/* don't allow async punt for O_NONBLOCK or RWF_NOWAIT */
-	if ((kiocb->ki_flags & IOCB_NOWAIT) || (file->f_flags & O_NONBLOCK))
+	/*
+	 * If the file is marked O_NONBLOCK, still allow retry for it if it
+	 * supports async. Otherwise it's impossible to use O_NONBLOCK files
+	 * reliably. If not, or it IOCB_NOWAIT is set, don't retry.
+	 */
+	if ((kiocb->ki_flags & IOCB_NOWAIT) ||
+	    ((file->f_flags & O_NONBLOCK) && !io_file_supports_nowait(req, rw)))
 		req->flags |= REQ_F_NOWAIT;
 
 	ioprio = READ_ONCE(sqe->ioprio);
@@ -3349,7 +3355,7 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	if (unlikely(!(req->file->f_mode & FMODE_READ)))
 		return -EBADF;
-	return io_prep_rw(req, sqe);
+	return io_prep_rw(req, sqe, READ);
 }
 
 /*
@@ -3542,7 +3548,7 @@ static int io_write_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	if (unlikely(!(req->file->f_mode & FMODE_WRITE)))
 		return -EBADF;
-	return io_prep_rw(req, sqe);
+	return io_prep_rw(req, sqe, WRITE);
 }
 
 static int io_write(struct io_kiocb *req, unsigned int issue_flags)