In `erofsmount_nbd()`, if the child process forked therein exits erroneously within `erofsmount_startnbd()`, the NBD device will never become operational. Since the parent process does not check the child's status, it will stuck in an infinite loop polling the NBD device state.
This patch ensures the parent process correctly detects this failure and returns an error accordingly. Signed-off-by: Yifan Zhao <[email protected]> --- mount/main.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/mount/main.c b/mount/main.c index b28b8ba..758e8f8 100644 --- a/mount/main.c +++ b/mount/main.c @@ -8,6 +8,7 @@ #include <signal.h> #include <sys/mount.h> #include <sys/types.h> +#include <sys/wait.h> #include <pthread.h> #include <unistd.h> #include "erofs/config.h" @@ -1173,6 +1174,14 @@ static int erofsmount_nbd(struct erofs_nbd_source *source, while (1) { err = erofs_nbd_in_service(num); if (err == -ENOENT || err == -ENOTCONN) { + int status; + + err = waitpid(pid, &status, WNOHANG); + if (err < 0) + return -errno; + else if (err > 0) + return status ? -EIO : 0; + usleep(50000); continue; } -- 2.43.0
