The sense of this test is reversed. There's nothing that prevents
userland from requesting a DIO read that is longer than the available
data. Conversely, we don't expect a buffered read to be short unless it
hits the EOF.
Suggested-by: David Howells <dhowells@xxxxxxxxxx>
Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
---
fs/netfs/io.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
David, feel free to fold this into the patch that adds the condition
so we can avoid the regression.
diff --git a/fs/netfs/io.c b/fs/netfs/io.c
index e5a15a924fc7..8188d43e8044 100644
--- a/fs/netfs/io.c
+++ b/fs/netfs/io.c
@@ -728,7 +728,7 @@ ssize_t netfs_begin_read(struct netfs_io_request *rreq, bool sync)
ret = rreq->error;
if (ret == 0 && rreq->submitted < rreq->len &&
- rreq->origin == NETFS_DIO_READ) {
+ rreq->origin != NETFS_DIO_READ) {
trace_netfs_failure(rreq, NULL, ret, netfs_fail_short_read);
ret = -EIO;
}
--
2.36.1
