Re: [PATCH v3] statmount: Fix the null-ptr-deref in do_statmount()

[Andrei Vagin]

On Fri, Feb 13, 2026 at 2:30 AM Qing Wang <wangqing7171@xxxxxxxxx> wrote:
>
> If the mount is internal, it's mnt_ns will be MNT_NS_INTERNAL, which is
> defined as ERR_PTR(-EINVAL). So, in the do_statmount(), need to check ns
> of mount by IS_ERR() and return.
>
> Fixes: 0e5032237ee5 ("statmount: accept fd as a parameter")
> Reported-by: syzbot+9e03a9535ea65f687a44@xxxxxxxxxxxxxxxxxxxxxxxxx
> Closes: https://lore.kernel.org/all/698e287a.a70a0220.2c38d7.009e.GAE@xxxxxxxxxx/
> Signed-off-by: Qing Wang <wangqing7171@xxxxxxxxx>
> ---
>  fs/namespace.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/fs/namespace.c b/fs/namespace.c
> index a67cbe42746d..90700df65f0d 100644
> --- a/fs/namespace.c
> +++ b/fs/namespace.c
> @@ -5678,6 +5678,8 @@ static int do_statmount(struct kstatmount *s, u64 mnt_id, u64 mnt_ns_id,
>
>                 s->mnt = mnt_file->f_path.mnt;
>                 ns = real_mount(s->mnt)->mnt_ns;
> +               if (IS_ERR(ns))
> +                       return PTR_ERR(ns);

nit: EINVAL is overused in the kernel. ENODEV seems like a better fit
for this case.

Otherwise, the patch looks good to me:
Reviewed-by: Andrei Vagin <avagin@xxxxxxxxx>

>                 if (!ns)
>                         /*
>                          * We can't set mount point and mnt_ns_id since we don't have a
> --
> 2.34.1
>