qemu_open() in blkio_virtio_blk_common_open() is used to open the
character device (e.g. /dev/vhost-vdpa-0 or /dev/vfio/vfio) or in
the future eventually the unix socket.
In all these cases we cannot open the path in read-only mode,
when the `read-only` option of blockdev is on, because the exchange
of IOCTL commands for example will fail.
In order to open the device read-only, we have to use the `read-only`
property of the libblkio driver as we already do in blkio_file_open().
Fixes: cad2ccc395 ("block/blkio: use qemu_open() to support fd passing for
virtio-blk")
Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=2225439
Reported-by: Qing Wang <qinw...@redhat.com>
Signed-off-by: Stefano Garzarella <sgarz...@redhat.com>
---
Notes:
v2:
- added comment on top of qemu_open() [Daniel]
v1:
https://lore.kernel.org/qemu-devel/20230725111155.85426-1-sgarz...@redhat.com/
block/blkio.c | 21 ++++++++++++---------
1 file changed, 12 insertions(+), 9 deletions(-)
diff --git a/block/blkio.c b/block/blkio.c
index 1798648134..cd6d2e55e7 100644
--- a/block/blkio.c
+++ b/block/blkio.c
@@ -686,15 +686,18 @@ static int blkio_virtio_blk_common_open(BlockDriverState
*bs,
* layer through the "/dev/fdset/N" special path.
*/
if (fd_supported) {
- int open_flags;
-
- if (flags & BDRV_O_RDWR) {
- open_flags = O_RDWR;
- } else {
- open_flags = O_RDONLY;
- }
-
- fd = qemu_open(path, open_flags, errp);
+ /*
+ * `path` can contain the path of a character device
+ * (e.g. /dev/vhost-vdpa-0 or /dev/vfio/vfio) or a unix socket.
+ *
+ * So, we should always open it with O_RDWR flag, also if BDRV_O_RDWR
+ * is not set in the open flags, because the exchange of IOCTL commands
+ * for example will fail.
+ *
+ * In order to open the device read-only, we are using the `read-only`
+ * property of the libblkio driver in blkio_file_open().
+ */
+ fd = qemu_open(path, O_RDWR, errp);
if (fd < 0) {
return -EINVAL;
}
--
2.41.0
