Re: [PATCH 1/3] virtiofsd: Find original inode ID of mount points

[Max Reitz <mreitz@redhat.com>]

On 17.05.21 16:57, Vivek Goyal wrote:
> On Wed, May 12, 2021 at 02:55:42PM +0200, Max Reitz wrote:
>> Mount point directories represent two inodes: On one hand, they are a
>> normal directory on their parent filesystem.  On the other, they are the
>> root node of the filesystem mounted there.  Thus, they have two inode
>> IDs.
>>
>> Right now, we only report the latter inode ID (i.e. the inode ID of the
>> mounted filesystem's root node).  This is fine once the guest has
>> auto-mounted a submount there (so this inode ID goes with a device ID
>> that is distinct from the parent filesystem), but before the auto-mount,
>> they have the device ID of the parent and the inode ID for the submount.
>> This is problematic because this is likely exactly the same
>> st_dev/st_ino combination as the parent filesystem's root node.  This
>> leads to problems for example with `find`, which will thus complain
>> about a filesystem loop if it has visited the parent filesystem's root
>> node before, and then refuse to descend into the submount.
>>
>> There is a way to find the mount directory's original inode ID, and that
>> is to readdir(3) the parent directory, look for the mount directory, and
>> read the dirent.d_ino field.  Using this, we can let lookup and
>> readdirplus return that original inode ID, which the guest will thus
>> show until the submount is auto-mounted.  (Then, it will invoke getattr
>> and that stat(2) call will return the inode ID for the submount.)
>>
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>>   tools/virtiofsd/passthrough_ll.c | 104 +++++++++++++++++++++++++++++--
>>   1 file changed, 99 insertions(+), 5 deletions(-)
>>
>> diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
>> index 1553d2ef45..110b6e7e5b 100644
>> --- a/tools/virtiofsd/passthrough_ll.c
>> +++ b/tools/virtiofsd/passthrough_ll.c
>> @@ -968,14 +968,87 @@ static int do_statx(struct lo_data *lo, int dirfd, const char *pathname,
>>       return 0;
>>   }
>>   
>> +/*
>> + * Use readdir() to find mp_name's inode ID on the parent's filesystem.
>> + * (For mount points, stat() will only return the inode ID on the
>> + * filesystem mounted there, i.e. the root directory's inode ID.  The
>> + * mount point originally was a directory on the parent filesystem,
>> + * though, and so has a different inode ID there.  When passing
>> + * submount information to the guest, we need to pass this other ID,
>> + * so the guest can use it as the inode ID until the submount is
>> + * auto-mounted.  (At which point the guest will invoke getattr and
>> + * find the inode ID on the submount.))
>> + *
>> + * Return 0 on success, and -errno otherwise.  *pino is set only in
>> + * case of success.
>> + */
>> +static int get_mp_ino_on_parent(const struct lo_inode *dir, const char *mp_name,
>> +                                ino_t *pino)
>> +{
>> +    int dirfd = -1;
>> +    int ret;
>> +    DIR *dp = NULL;
>> +
>> +    dirfd = openat(dir->fd, ".", O_RDONLY);
>> +    if (dirfd < 0) {
>> +        ret = -errno;
>> +        goto out;
>> +    }
>> +
>> +    dp = fdopendir(dirfd);
>> +    if (!dp) {
>> +        ret = -errno;
>> +        goto out;
>> +    }
>> +    /* Owned by dp now */
>> +    dirfd = -1;
>> +
>> +    while (true) {
>> +        struct dirent *de;
>> +
>> +        errno = 0;
>> +        de = readdir(dp);
>> +        if (!de) {
>> +            ret = errno ? -errno : -ENOENT;
>> +            goto out;
>> +        }
>> +
>> +        if (!strcmp(de->d_name, mp_name)) {
>> +            *pino = de->d_ino;
>> +            ret = 0;
>> +            goto out;
>> +        }
>> +    }
>> +
>> +out:
>> +    if (dp) {
>> +        closedir(dp);
>> +    }
>> +    if (dirfd >= 0) {
>> +        close(dirfd);
>> +    }
>> +    return ret;
>> +}
>> +
>>   /*
>>    * Increments nlookup on the inode on success. unref_inode_lolocked() must be
>>    * called eventually to decrement nlookup again. If inodep is non-NULL, the
>>    * inode pointer is stored and the caller must call lo_inode_put().
>> + *
>> + * If parent_fs_st_ino is true, the entry is a mount point, and submounts are
>> + * announced to the guest, set e->attr.st_ino to the entry's inode ID on its
>> + * parent filesystem instead of its inode ID on the filesystem mounted on it.
>> + * (For mount points, the entry encompasses two inodes: One on the parent FS,
>> + * and one on the mounted FS (where it is the root node), so it has two inode
>> + * IDs.  When looking up entries, we should show the guest the parent FS's inode
>> + * ID, because as long as the guest has not auto-mounted the submount, it should
>> + * see that original ID.  Once it does perform the auto-mount, it will invoke
>> + * getattr and see the root node's inode ID.)
>>    */
>>   static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
>>                           struct fuse_entry_param *e,
>> -                        struct lo_inode **inodep)
>> +                        struct lo_inode **inodep,
>> +                        bool parent_fs_st_ino)
>>   {
>>       int newfd;
>>       int res;
>> @@ -984,6 +1057,7 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
>>       struct lo_data *lo = lo_data(req);
>>       struct lo_inode *inode = NULL;
>>       struct lo_inode *dir = lo_inode(req, parent);
>> +    ino_t ino_id_for_guest;
>>   
>>       if (inodep) {
>>           *inodep = NULL; /* in case there is an error */
>> @@ -1018,9 +1092,22 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
>>           goto out_err;
>>       }
>>   
>> +    ino_id_for_guest = e->attr.st_ino;
>> +
>>       if (S_ISDIR(e->attr.st_mode) && lo->announce_submounts &&
>>           (e->attr.st_dev != dir->key.dev || mnt_id != dir->key.mnt_id)) {
>>           e->attr_flags |= FUSE_ATTR_SUBMOUNT;
>> +
>> +        if (parent_fs_st_ino) {
>> +            /*
>> +             * Best effort, so ignore errors.
>> +             * Also note that using readdir() means there may be races:
>> +             * The directory entry we find (if any) may be different
>> +             * from newfd.  Again, this is a best effort.  Reporting
>> +             * the wrong inode ID to the guest is not catastrophic.
>> +             */
>> +            get_mp_ino_on_parent(dir, name, &ino_id_for_guest);
> 
> Hi Max,
> 
> [CC virtio-fs list ]
> 
> In general patch looks good to me. A minor nit. get_mp_ino_on_parent()
> is retruning error. It might be better to capture error and print a
> message and continue.

Sure, why not.

> I have couple of general questions about submounts.
> 
> - What happens in case of single file mounted on top of another file.
> 
>    mount --bind foo.txt bar.txt
> 
> Do submounts work when mount point is not a directory.

No, as you can see in the condition quoted above, we only set the 
FUSE_ATTR_SUBMOUNT flag for directories.  That seemed the most common 
case for me, and I didn’t want to have to worry about weirdness that 
might ensue for file mounts.

> - Say a directory is not a mount point yet and lookup instantiates an
>    inode. Later user mounts something on that directory. When does
>    client/server notice this change. I am assuming this is probably
>    part of revalidation path.

I guess at least before this patch this is no different from any other 
filesystem change.  Because st_dev+st_ino changed, it should basically 
look like the old directory was removed and a different one was put in 
its place.

Now, with this patch, we will return the old st_ino to the guest, but 
internally virtiofsd will still use the submount’s st_dev/st_ino, so a 
new lo_inode should be created, and so fuse_dentry_revalidate()’s lookup 
should return a different node ID, resulting it to consider the entry 
expired.

Besides, fuse_dentry_revalidate() has a condition on IS_AUTOMOUNT(inode) 
!= flags & FUSE_ATTR_SUBMOUNT.  Considering the previous paragraph, this 
doesn’t seem necessary to me, but it can’t hurt.

Max