[PATCH v3 4/4] add listmount(2) syscall

[Ian Kent]

On 5/10/23 12:01, Miklos Szeredi wrote:
> On Wed, 4 Oct 2023 at 21:38, Paul Moore <paul at paul-moore.com> wrote:
>> On Thu, Sep 28, 2023 at 9:04 AM Miklos Szeredi <mszeredi at redhat.com> wrote:
>>> Add way to query the children of a particular mount.  This is a more
>>> flexible way to iterate the mount tree than having to parse the complete
>>> /proc/self/mountinfo.
>>>
>>> Lookup the mount by the new 64bit mount ID.  If a mount needs to be queried
>>> based on path, then statx(2) can be used to first query the mount ID
>>> belonging to the path.
>>>
>>> Return an array of new (64bit) mount ID's.  Without privileges only mounts
>>> are listed which are reachable from the task's root.
>>>
>>> Signed-off-by: Miklos Szeredi <mszeredi at redhat.com>
>>> ---
>>>   arch/x86/entry/syscalls/syscall_32.tbl |  1 +
>>>   arch/x86/entry/syscalls/syscall_64.tbl |  1 +
>>>   fs/namespace.c                         | 69 ++++++++++++++++++++++++++
>>>   include/linux/syscalls.h               |  3 ++
>>>   include/uapi/asm-generic/unistd.h      |  5 +-
>>>   include/uapi/linux/mount.h             |  3 ++
>>>   6 files changed, 81 insertions(+), 1 deletion(-)
>> ...
>>
>>> diff --git a/fs/namespace.c b/fs/namespace.c
>>> index 3326ba2b2810..050e2d2af110 100644
>>> --- a/fs/namespace.c
>>> +++ b/fs/namespace.c
>>> @@ -4970,6 +4970,75 @@ SYSCALL_DEFINE4(statmount, const struct __mount_arg __user *, req,
>>>          return ret;
>>>   }
>>>
>>> +static long do_listmount(struct vfsmount *mnt, u64 __user *buf, size_t bufsize,
>>> +                        const struct path *root, unsigned int flags)
>>> +{
>>> +       struct mount *r, *m = real_mount(mnt);
>>> +       struct path rootmnt = {
>>> +               .mnt = root->mnt,
>>> +               .dentry = root->mnt->mnt_root
>>> +       };
>>> +       long ctr = 0;
>>> +       bool reachable_only = true;
>>> +       int err;
>>> +
>>> +       err = security_sb_statfs(mnt->mnt_root);
>>> +       if (err)
>>> +               return err;
>>> +
>>> +       if (flags & LISTMOUNT_UNREACHABLE) {
>>> +               if (!capable(CAP_SYS_ADMIN))
>>> +                       return -EPERM;
>>> +               reachable_only = false;
>>> +       }
>>> +
>>> +       if (reachable_only && !is_path_reachable(m, mnt->mnt_root, &rootmnt))
>>> +               return capable(CAP_SYS_ADMIN) ? 0 : -EPERM;
>>> +
>>> +       list_for_each_entry(r, &m->mnt_mounts, mnt_child) {
>>> +               if (reachable_only &&
>>> +                   !is_path_reachable(r, r->mnt.mnt_root, root))
>>> +                       continue;
>> I believe we would want to move the security_sb_statfs() call from
>> above to down here; something like this I think ...
>>
>>    err = security_sb_statfs(r->mnt.mnt_root);
>>    if (err)
>>      /* if we can't access the mount, pretend it doesn't exist */
>>      continue;
> Hmm.  Why is this specific to listing mounts (i.e. why doesn't readdir
> have a similar filter)?
>
> Also why hasn't this come up with regards to the proc interfaces that
> list mounts?

The proc interfaces essentially use <mount namespace>->list to provide

the mounts that can be seen so it's filtered by mount namespace of the

task that's doing the open().


See fs/namespace.c:mnt_list_next() and just below the m_start(), m_next(),

etc.


Ian

>
> I just want to understand the big picture here.
>
> Thanks,
> Miklos