[PATCH v3v] eventfd: convert to f_op->read_iter()

[Jens Axboe <axboe@kernel.dk>]

eventfd is using ->read() as it's file_operations read handler, but
this prevents passing in information about whether a given IO operation
is blocking or not. We can only use the file flags for that. To support
async (-EAGAIN/poll based) retries for io_uring, we need ->read_iter()
support. Convert eventfd to using ->read_iter().

Signed-off-by: Jens Axboe <axboe@kernel.dk>

---

Actually send out the right patch...

Since v2:

- Cleanup eventfd_read() as per Al's suggestions

Since v1:

- Add FMODE_NOWAIT to the eventfd file

diff --git a/fs/eventfd.c b/fs/eventfd.c
index 78e41c7c3d05..c9fa1e9cf5e3 100644
--- a/fs/eventfd.c
+++ b/fs/eventfd.c
@@ -216,32 +216,32 @@ int eventfd_ctx_remove_wait_queue(struct eventfd_ctx *ctx, wait_queue_entry_t *w
 }
 EXPORT_SYMBOL_GPL(eventfd_ctx_remove_wait_queue);
 
-static ssize_t eventfd_read(struct file *file, char __user *buf, size_t count,
-			    loff_t *ppos)
+static ssize_t eventfd_read(struct kiocb *iocb, struct iov_iter *to)
 {
+	struct file *file = iocb->ki_filp;
 	struct eventfd_ctx *ctx = file->private_data;
-	ssize_t res;
 	__u64 ucnt = 0;
 	DECLARE_WAITQUEUE(wait, current);
 
-	if (count < sizeof(ucnt))
+	if (iov_iter_count(to) < sizeof(ucnt))
 		return -EINVAL;
-
 	spin_lock_irq(&ctx->wqh.lock);
-	res = -EAGAIN;
-	if (ctx->count > 0)
-		res = sizeof(ucnt);
-	else if (!(file->f_flags & O_NONBLOCK)) {
+	if (!ctx->count) {
+		if ((file->f_flags & O_NONBLOCK) ||
+		    (iocb->ki_flags & IOCB_NOWAIT)) {
+			spin_unlock_irq(&ctx->wqh.lock);
+			return -EAGAIN;
+		}
 		__add_wait_queue(&ctx->wqh, &wait);
 		for (;;) {
 			set_current_state(TASK_INTERRUPTIBLE);
-			if (ctx->count > 0) {
-				res = sizeof(ucnt);
+			if (ctx->count)
 				break;
-			}
 			if (signal_pending(current)) {
-				res = -ERESTARTSYS;
-				break;
+				__remove_wait_queue(&ctx->wqh, &wait);
+				__set_current_state(TASK_RUNNING);
+				spin_unlock_irq(&ctx->wqh.lock);
+				return -ERESTARTSYS;
 			}
 			spin_unlock_irq(&ctx->wqh.lock);
 			schedule();
@@ -250,17 +250,14 @@ static ssize_t eventfd_read(struct file *file, char __user *buf, size_t count,
 		__remove_wait_queue(&ctx->wqh, &wait);
 		__set_current_state(TASK_RUNNING);
 	}
-	if (likely(res > 0)) {
-		eventfd_ctx_do_read(ctx, &ucnt);
-		if (waitqueue_active(&ctx->wqh))
-			wake_up_locked_poll(&ctx->wqh, EPOLLOUT);
-	}
+	eventfd_ctx_do_read(ctx, &ucnt);
+	if (waitqueue_active(&ctx->wqh))
+		wake_up_locked_poll(&ctx->wqh, EPOLLOUT);
 	spin_unlock_irq(&ctx->wqh.lock);
-
-	if (res > 0 && put_user(ucnt, (__u64 __user *)buf))
+	if (unlikely(copy_to_iter(&ucnt, sizeof(ucnt), to) != sizeof(ucnt)))
 		return -EFAULT;
 
-	return res;
+	return sizeof(ucnt);
 }
 
 static ssize_t eventfd_write(struct file *file, const char __user *buf, size_t count,
@@ -329,7 +326,7 @@ static const struct file_operations eventfd_fops = {
 #endif
 	.release	= eventfd_release,
 	.poll		= eventfd_poll,
-	.read		= eventfd_read,
+	.read_iter	= eventfd_read,
 	.write		= eventfd_write,
 	.llseek		= noop_llseek,
 };
@@ -427,8 +424,17 @@ static int do_eventfd(unsigned int count, int flags)
 
 	fd = anon_inode_getfd("[eventfd]", &eventfd_fops, ctx,
 			      O_RDWR | (flags & EFD_SHARED_FCNTL_FLAGS));
-	if (fd < 0)
+	if (fd < 0) {
 		eventfd_free_ctx(ctx);
+	} else {
+		struct file *file;
+
+		file = fget(fd);
+		if (file) {
+			file->f_mode |= FMODE_NOWAIT;
+			fput(file);
+		}
+	}
 
 	return fd;
 }

-- 
Jens Axboe