On 2025/12/26 14:49, Gao Xiang wrote:
On 2025/12/23 18:04, Yifan Zhao wrote:
If the main process of `erofsmount_nbd()` encounters an error after the
nbd device has been successfully set up, it fails to disconnect it
before exiting, resulting in the subprocess not being cleaned up and
keeping its connection with NBD device.
This patch resolves the issue by disconnecting NBD device before exiting
on error.
Signed-off-by: Yifan Zhao <[email protected]>
---
Note:
- I believe directly killing the child process is unsafe, as it may
leave
in-flight NBD requests from the kernel unhandled, causing soft lockup.
- And I believe using nbdpath here is safe, as the child process
maintains
the NBD device connection throughout, preventing concurrent access by
other
actors.
What if the child process is already exited earlier, and the current NBD
device is reused for others?
How about keeping the previous nbdfd for ioctl interfaces instead to
avoid nbd device reuse.
OK, I will try this way.
Thanks,
Yifan
Thanks,
Gao Xiang
mount/main.c | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/mount/main.c b/mount/main.c
index 2a21979..d2d4815 100644
--- a/mount/main.c
+++ b/mount/main.c
@@ -1287,10 +1287,23 @@ static int erofsmount_nbd(struct
erofs_nbd_source *source,
if (!err) {
err = mount(nbdpath, mountpoint, fstype, flags, options);
- if (err < 0)
+ if (err < 0) {
err = -errno;
+ if (msg.is_netlink) {
+ erofs_nbd_nl_disconnect(msg.nbdnum);
+ } else {
+ int nbdfd;
+
+ nbdfd = open(nbdpath, O_RDWR);
+ if (nbdfd > 0) {
+ erofs_nbd_disconnect(nbdfd);
+ close(nbdfd);
+ }
+ }
+ return err;
+ }
- if (!err && msg.is_netlink) {
+ if (msg.is_netlink) {
id = erofs_nbd_get_identifier(msg.nbdnum);
err = IS_ERR(id) ? PTR_ERR(id) :
