io_uring: punt short reads to async context

We can encounter a short read when we're doing buffered reads and the
data is partially cached. Right now we just return the short read, but
that forces the application to read that CQE, then issue another SQE
to finish the read. That read will not be cached, and hence will result
in an async punt.

It's more efficient to do that async punt from within the kernel, as
that will the not need two round trips more to the kernel.

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

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 23e08c10f48606917ea00a16bdb9bbb813a97a74..92debd8be5353eb5feef4dbe308122946f26d7f4 100644 (file)
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1089,7 +1089,7 @@ static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
        struct iov_iter iter;
        struct file *file;
        size_t iov_count;
-       ssize_t ret;
+       ssize_t read_size, ret;
 
        ret = io_prep_rw(req, s, force_nonblock);
        if (ret)
@@ -1105,13 +1105,24 @@ static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
        if (ret < 0)
                return ret;
 
+       read_size = ret;
        iov_count = iov_iter_count(&iter);
        ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
        if (!ret) {
                ssize_t ret2;
 
-               /* Catch -EAGAIN return for forced non-blocking submission */
                ret2 = call_read_iter(file, kiocb, &iter);
+               /*
+                * In case of a short read, punt to async. This can happen
+                * if we have data partially cached. Alternatively we can
+                * return the short read, in which case the application will
+                * need to issue another SQE and wait for it. That SQE will
+                * need async punt anyway, so it's more efficient to do it
+                * here.
+                */
+               if (force_nonblock && ret2 > 0 && ret2 < read_size)
+                       ret2 = -EAGAIN;
+               /* Catch -EAGAIN return for forced non-blocking submission */
                if (!force_nonblock || ret2 != -EAGAIN) {
                        io_rw_done(kiocb, ret2);
                } else {