io_uring: fix off-by one bvec index

commit d6fef34ee4d102be448146f24caf96d7b4a05401 upstream.

If the offset equals the bv_len of the first registered bvec, then the
request does not include any of that first bvec. Skip it so that drivers
don't have to deal with a zero length bvec, which was observed to break
NVMe's PRP list creation.

Cc: stable@vger.kernel.org
Fixes: bd11b3a391e3 ("io_uring: don't use iov_iter_advance() for fixed buffers")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Link: https://lore.kernel.org/r/20231120221831.2646460-1-kbusch@meta.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index d9c853d..dde501a 100644 (file)
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -1261,7 +1261,7 @@ int io_import_fixed(int ddir, struct iov_iter *iter,
                 */
                const struct bio_vec *bvec = imu->bvec;
 
-               if (offset <= bvec->bv_len) {
+               if (offset < bvec->bv_len) {
                        /*
                         * Note, huge pages buffers consists of one large
                         * bvec entry and should always go this way. The other