[Virtio-fs] [PATCH v4 1/2] virtiofsd: Track mounts

[Greg Kurz]

On Wed, 26 Jan 2022 17:47:09 -0500
Vivek Goyal <vgoyal at redhat.com> wrote:

> On Tue, Jan 25, 2022 at 03:12:11PM +0100, Greg Kurz wrote:
> > The upcoming implementation of ->sync_fs() needs to know about all
> > submounts in order to call syncfs() on them when virtiofsd is started
> > without '-o announce_submounts'.
> > 
> > Track every inode that comes up with a new mount id in a GHashTable.
> > If the mount id isn't available, e.g. no statx() on the host, fallback
> > on the device id for the key. This is done during lookup because we
> > only care for the submounts that the client knows about. The inode
> > is removed from the hash table when ultimately unreferenced. This
> > can happen on a per-mount basis when the client posts a FUSE_FORGET
> > request or for all submounts at once with FUSE_DESTROY.
> > 
> > Signed-off-by: Greg Kurz <groug at kaod.org>
> > ---
> >  tools/virtiofsd/passthrough_ll.c | 43 +++++++++++++++++++++++++++++---
> >  1 file changed, 40 insertions(+), 3 deletions(-)
> > 
> > diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
> > index 64b5b4fbb186..7bf31fc129c8 100644
> > --- a/tools/virtiofsd/passthrough_ll.c
> > +++ b/tools/virtiofsd/passthrough_ll.c
> > @@ -117,6 +117,7 @@ struct lo_inode {
> >      GHashTable *posix_locks; /* protected by lo_inode->plock_mutex */
> >  
> >      mode_t filetype;
> > +    bool is_mnt;
> >  };
> >  
> >  struct lo_cred {
> > @@ -164,6 +165,7 @@ struct lo_data {
> >      bool use_statx;
> >      struct lo_inode root;
> >      GHashTable *inodes; /* protected by lo->mutex */
> > +    GHashTable *mnt_inodes; /* protected by lo->mutex */
> >      struct lo_map ino_map; /* protected by lo->mutex */
> >      struct lo_map dirp_map; /* protected by lo->mutex */
> >      struct lo_map fd_map; /* protected by lo->mutex */
> > @@ -1000,6 +1002,31 @@ static int do_statx(struct lo_data *lo, int dirfd, const char *pathname,
> >      return 0;
> >  }
> >  
> 
> Hi Greg,
> 
> Thanks for the patches. Had a quick look. Overall these patches look
> pretty good to me. I will spend more time testing and having a 
> closer look. Some quick thoughts below.
> 
> > +static uint64_t mnt_inode_key(struct lo_inode *inode)
> > +{
> > +    /* Prefer mnt_id, fallback on dev */
> > +    return inode->key.mnt_id ? inode->key.mnt_id : inode->key.dev;
> > +}
> 
> I am not sure if we should use inode->key.dev. This might create problem
> if same file system is bind mounted at two paths in shared dir. So
> say /dev/sdb is mounted at foo1/ and then bind mounted at foo2/ in
> shared dir. A user looks up foo1/ and does some writes. Then we
> lookup foo2/ and release that inode. Release of foo2 will let go
> inode from the hash. And that means if later another write happens
> in foo1/ followed by syncfs(), we will not issue syncfs() on filesystem
> backed by /dev/sdb.
> 

Right, inode->key.dev isn't a safe fallback, which is the whole 
point behind the need for mnt_id :-)

> So what are the options.
> 
> A. Make mnt_id mandatory and do not implement it if mnt_id is not
>    available.
> 

This is still quite recent (kernel 5.7 and glibc 2.33), it would
be unfortunate for users of older versions.

> B. Don't do anything and live with this. It is a corner case and
>    still better than not implement submount syncfs at all.
> 

Having bind mounts isn't such a corner case, is it ? And anyway,
this would be a serious flaw in the implementation. I don't think
this is acceptable.

> C. Instead of adding lo_inode to hash, create another kind of object
>    and reference count that. It could be a mount fd which we open
>    when we add object for the first time. So when foo1/ inode is
>    instantiated, create mountfd object, add it to hash table using
>    device id as the key. When foo2 comes along, we find the object
>    in the hash and just bump up the ref. Now this mountfd object
>    will go away when both foo1 and foo2 inodes have been evicted
>    and will take care of the issue I am referring to.
> 
> I guess B is little extra complexity but probably not too bad.

s/B/C

> WDYT. It sounds litter better than option A and B.
> 

Definitely. I'll give a try.

> 
> > +
> > +static void add_mnt_inode(struct lo_data *lo, struct lo_inode *inode)
> > +{
> > +    uint64_t mnt_key = mnt_inode_key(inode);
> > +
> > +    if (!g_hash_table_contains(lo->mnt_inodes, &mnt_key)) {
> > +        inode->is_mnt = true;
> > +        g_hash_table_insert(lo->mnt_inodes, &mnt_key, inode);
> > +    }
> > +}
> > +
> > +static void remove_mnt_inode(struct lo_data *lo, struct lo_inode *inode)
> > +{
> > +    uint64_t mnt_key = mnt_inode_key(inode);
> > +
> > +    if (inode->is_mnt) {
> > +        g_hash_table_remove(lo->mnt_inodes, &mnt_key);
> > +    }
> > +}
> 
> Should we issue syncfs() on this inode when we are removing it? It
> is possible guest did some writes, let go inode and later issued
> a syncfs(). By that time inode is gone and we will not issue any
> syncfs() on this filesystem. Hence leaving data in host page cache.
> 

Right, I guess we can do that.


Cheers,

--
Greg

> Thanks
> Vivek
> 
> > +
> >  /*
> >   * Increments nlookup on the inode on success. unref_inode_lolocked() must be
> >   * called eventually to decrement nlookup again. If inodep is non-NULL, the
> > @@ -1086,10 +1113,15 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
> >          pthread_mutex_lock(&lo->mutex);
> >          inode->fuse_ino = lo_add_inode_mapping(req, inode);
> >          g_hash_table_insert(lo->inodes, &inode->key, inode);
> > +        add_mnt_inode(lo, inode);
> >          pthread_mutex_unlock(&lo->mutex);
> >      }
> >      e->ino = inode->fuse_ino;
> >  
> > +    fuse_log(FUSE_LOG_DEBUG, "  %lli/%s -> %lli%s\n",
> > +             (unsigned long long) parent, name, (unsigned long long) e->ino,
> > +             inode->is_mnt ? " (submount)" : "");
> > +
> >      /* Transfer ownership of inode pointer to caller or drop it */
> >      if (inodep) {
> >          *inodep = inode;
> > @@ -1099,9 +1131,6 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
> >  
> >      lo_inode_put(lo, &dir);
> >  
> > -    fuse_log(FUSE_LOG_DEBUG, "  %lli/%s -> %lli\n", (unsigned long long)parent,
> > -             name, (unsigned long long)e->ino);
> > -
> >      return 0;
> >  
> >  out_err:
> > @@ -1563,6 +1592,7 @@ static void unref_inode(struct lo_data *lo, struct lo_inode *inode, uint64_t n)
> >              g_hash_table_destroy(inode->posix_locks);
> >              pthread_mutex_destroy(&inode->plock_mutex);
> >          }
> > +        remove_mnt_inode(lo, inode);
> >          /* Drop our refcount from lo_do_lookup() */
> >          lo_inode_put(lo, &inode);
> >      }
> > @@ -3337,6 +3367,7 @@ static void lo_destroy(void *userdata)
> >      struct lo_data *lo = (struct lo_data *)userdata;
> >  
> >      pthread_mutex_lock(&lo->mutex);
> > +    g_hash_table_remove_all(lo->mnt_inodes);
> >      while (true) {
> >          GHashTableIter iter;
> >          gpointer key, value;
> > @@ -3850,6 +3881,7 @@ static void setup_root(struct lo_data *lo, struct lo_inode *root)
> >          root->posix_locks = g_hash_table_new_full(
> >              g_direct_hash, g_direct_equal, NULL, posix_locks_value_destroy);
> >      }
> > +    add_mnt_inode(lo, root);
> >  }
> >  
> >  static guint lo_key_hash(gconstpointer key)
> > @@ -3869,6 +3901,10 @@ static gboolean lo_key_equal(gconstpointer a, gconstpointer b)
> >  
> >  static void fuse_lo_data_cleanup(struct lo_data *lo)
> >  {
> > +    if (lo->mnt_inodes) {
> > +        g_hash_table_destroy(lo->mnt_inodes);
> > +    }
> > +
> >      if (lo->inodes) {
> >          g_hash_table_destroy(lo->inodes);
> >      }
> > @@ -3931,6 +3967,7 @@ int main(int argc, char *argv[])
> >      lo.root.fd = -1;
> >      lo.root.fuse_ino = FUSE_ROOT_ID;
> >      lo.cache = CACHE_AUTO;
> > +    lo.mnt_inodes = g_hash_table_new(g_int64_hash, g_int64_equal);
> >  
> >      /*
> >       * Set up the ino map like this:
> > -- 
> > 2.34.1
> > 
>