io_uring/rw: allow pollable non-blocking attempts for !FMODE_NOWAIT

The checking for whether or not io_uring can do a non-blocking read or
write attempt is gated on FMODE_NOWAIT. However, if the file is
pollable, it's feasible to just check if it's currently in a state in
which it can sanely receive or send _some_ data.

This avoids unnecessary io-wq punts, and repeated worthless retries
before doing that punt, by assuming that some data can get delivered
or received if poll tells us that is true. It also allows multishot
reads to properly work with these types of files, enabling a bit of
a cleanup of the logic that:

c9d952b9 ("io_uring/rw: fix cflags posting for single issue multishot read")

had to put in place.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring/rw.c

@@ -31,9 +31,19 @@ struct io_rw {
 	rwf_t				flags;
 };
 
-static inline bool io_file_supports_nowait(struct io_kiocb *req)
+static bool io_file_supports_nowait(struct io_kiocb *req, __poll_t mask)
 {
-	return req->flags & REQ_F_SUPPORT_NOWAIT;
+	/* If FMODE_NOWAIT is set for a file, we're golden */
+	if (req->flags & REQ_F_SUPPORT_NOWAIT)
+		return true;
+	/* No FMODE_NOWAIT, if we can poll, check the status */
+	if (io_file_can_poll(req)) {
+		struct poll_table_struct pt = { ._key = mask };
+
+		return vfs_poll(req->file, &pt) & mask;
+	}
+	/* No FMODE_NOWAIT support, and file isn't pollable. Tough luck. */
+	return false;
 }
 
 #ifdef CONFIG_COMPAT
@@ -796,8 +806,8 @@ static int io_rw_init_file(struct io_kiocb *req, fmode_t mode, int rw_type)
 	 * 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)))
+	if (kiocb->ki_flags & IOCB_NOWAIT ||
+	    ((file->f_flags & O_NONBLOCK && (req->flags & REQ_F_SUPPORT_NOWAIT))))
 		req->flags |= REQ_F_NOWAIT;
 
 	if (ctx->flags & IORING_SETUP_IOPOLL) {
@@ -838,7 +848,7 @@ static int __io_read(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (force_nonblock) {
 		/* If the file doesn't support async, just async punt */
-		if (unlikely(!io_file_supports_nowait(req)))
+		if (unlikely(!io_file_supports_nowait(req, EPOLLIN)))
 			return -EAGAIN;
 		kiocb->ki_flags |= IOCB_NOWAIT;
 	} else {
@@ -951,13 +961,6 @@ int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
 
 	ret = __io_read(req, issue_flags);
 
-	/*
-	 * If the file doesn't support proper NOWAIT, then disable multishot
-	 * and stay in single shot mode.
-	 */
-	if (!io_file_supports_nowait(req))
-		req->flags &= ~REQ_F_APOLL_MULTISHOT;
-
 	/*
 	 * If we get -EAGAIN, recycle our buffer and just let normal poll
 	 * handling arm it.
@@ -984,9 +987,6 @@ int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
 		 * jump to the termination path. This request is then done.
 		 */
 		cflags = io_put_kbuf(req, ret, issue_flags);
-		if (!(req->flags & REQ_F_APOLL_MULTISHOT))
-			goto done;
-
 		rw->len = 0; /* similarly to above, reset len to 0 */
 
 		if (io_req_post_cqe(req, ret, cflags | IORING_CQE_F_MORE)) {
@@ -1007,7 +1007,6 @@ int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
 	 * Either an error, or we've hit overflow posting the CQE. For any
 	 * multishot request, hitting overflow will terminate it.
 	 */
-done:
 	io_req_set_res(req, ret, cflags);
 	io_req_rw_cleanup(req, issue_flags);
 	if (issue_flags & IO_URING_F_MULTISHOT)
@@ -1031,7 +1030,7 @@ int io_write(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (force_nonblock) {
 		/* If the file doesn't support async, just async punt */
-		if (unlikely(!io_file_supports_nowait(req)))
+		if (unlikely(!io_file_supports_nowait(req, EPOLLOUT)))
 			goto ret_eagain;
 
 		/* Check if we can support NOWAIT. */