With the current implementation, blocking flock can lead to
deadlock. Thus, it's better to return EOPNOTSUPP if a user attempts
to perform a blocking flock request.

Signed-off-by: Sebastian Hasler <sebastian.hasler@stuvus.uni-stuttgart.de>
---
 tools/virtiofsd/passthrough_ll.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
index 64b5b4fbb1..f3cc307f6d 100644
--- a/tools/virtiofsd/passthrough_ll.c
+++ b/tools/virtiofsd/passthrough_ll.c
@@ -2442,6 +2442,12 @@ static void lo_flock(fuse_req_t req, fuse_ino_t ino, struct fuse_file_info *fi,
     int res;
     (void)ino;
 
+    if (!(op & LOCK_NB)) {
+        /* Blocking flock is not supported */
+        fuse_reply_err(req, EOPNOTSUPP);
+        return;
+    }
+
     res = flock(lo_fi_fd(req, fi), op);
 
     fuse_reply_err(req, res == -1 ? errno : 0);
-- 
2.33.1

On Tue, 11 Jan 2022 19:10:43 +0100
Sebastian Hasler <sebastian.hasler@stuvus.uni-stuttgart.de> wrote:

> With the current implementation, blocking flock can lead to
> deadlock. Thus, it's better to return EOPNOTSUPP if a user attempts
> to perform a blocking flock request.
> 
> Signed-off-by: Sebastian Hasler <sebastian.hasler@stuvus.uni-stuttgart.de>
> ---
>  tools/virtiofsd/passthrough_ll.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
> index 64b5b4fbb1..f3cc307f6d 100644
> --- a/tools/virtiofsd/passthrough_ll.c
> +++ b/tools/virtiofsd/passthrough_ll.c
> @@ -2442,6 +2442,12 @@ static void lo_flock(fuse_req_t req, fuse_ino_t ino, struct fuse_file_info *fi,
>      int res;
>      (void)ino;
>  
> +    if (!(op & LOCK_NB)) {
> +        /* Blocking flock is not supported */

This paraphrases the code. It would be more informative to provide
an explanation, something like /* Blocking flock can deadlock */ .

No big deal.

Reviewed-by: Greg Kurz <groug@kaod.org>

> +        fuse_reply_err(req, EOPNOTSUPP);
> +        return;
> +    }
> +
>      res = flock(lo_fi_fd(req, fi), op);
>  
>      fuse_reply_err(req, res == -1 ? errno : 0);

On 11/01/2022 19:10, Sebastian Hasler wrote:
> With the current implementation, blocking flock can lead to
> deadlock. Thus, it's better to return EOPNOTSUPP if a user attempts
> to perform a blocking flock request.
>
> Signed-off-by: Sebastian Hasler <sebastian.hasler@stuvus.uni-stuttgart.de>
> ---
>   tools/virtiofsd/passthrough_ll.c | 6 ++++++
>   1 file changed, 6 insertions(+)
>
> diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
> index 64b5b4fbb1..f3cc307f6d 100644
> --- a/tools/virtiofsd/passthrough_ll.c
> +++ b/tools/virtiofsd/passthrough_ll.c
> @@ -2442,6 +2442,12 @@ static void lo_flock(fuse_req_t req, fuse_ino_t ino, struct fuse_file_info *fi,
>       int res;
>       (void)ino;
>   
> +    if (!(op & LOCK_NB)) {
> +        /* Blocking flock is not supported */
> +        fuse_reply_err(req, EOPNOTSUPP);
> +        return;
> +    }
> +
>       res = flock(lo_fi_fd(req, fi), op);
>   
>       fuse_reply_err(req, res == -1 ? errno : 0);

I tested this patch by cherry-picking it on v6.1.0 and using it with 
Kata Containers 2.3.0. The bash code

     exec 42>/lock/flock
     flock -w 120 42

outputs

     flock: 42: Operation not supported

while the bash code

     exec 42>/lock/flock
     flock --nonblock 42

still works. So it works as intended.

-- 
Sebastian Hasler

stuvus – Studierendenvertretung Universität Stuttgart