[Virtio-fs] [PATCH] virtiofsd: Fix xattr operations

[Vivek Goyal]

On Tue, Jan 14, 2020 at 01:17:37AM +0000, misono.tomohiro at fujitsu.com wrote:
> > On Mon, Jan 13, 2020 at 7:42 PM Vivek Goyal <vgoyal at redhat.com> wrote:
> > >
> > > On Fri, Jan 10, 2020 at 10:07:22AM +0900, Misono Tomohiro wrote:
> > >
> > > [..]
> > > > --- a/tools/virtiofsd/passthrough_ll.c
> > > > +++ b/tools/virtiofsd/passthrough_ll.c
> > > > @@ -133,6 +133,7 @@ struct lo_inode {
> > > >      GHashTable *posix_locks; /* protected by lo_inode->plock_mutex
> > > > */
> > > >
> > > >      bool is_symlink;
> > > > +    bool is_regular;
> > > >  };
> > > >
> > > >  struct lo_cred {
> > > > @@ -1038,6 +1039,7 @@ static int lo_do_lookup(fuse_req_t req, fuse_ino_t parent, const char *name,
> > > >          }
> > > >
> > > >          inode->is_symlink = S_ISLNK(e->attr.st_mode);
> > > > +        inode->is_regular = S_ISREG(e->attr.st_mode) +
> > > > + S_ISDIR(e->attr.st_mode);
> > >
> > > How about having two variables. One for regular files and one for
> > > directories. Say ->is_regular and ->is_dir.  That way if there are
> > > other users later, these can cleary differentiate between regular
> > > files and directories.
> 
> Thanks for the comments
> Maybe we should just hold st_mode in lo_inode? I'm fine with either approach and will update.

Caching st_mode (only file type and not mode) sounds good to me. That's
even more generic and more usable for future use cases.

Thanks
Vivek