Re: [Qemu-devel] [PATCH] hw/9pfs: Add new security model mapped-file.

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

On Thu, 22 Dec 2011 21:45:51 +0000, Alex Bradbury <asb@asbradbury.org> wrote:
> On 19 December 2011 13:50, Aneesh Kumar K.V
> <aneesh.kumar@linux.vnet.ibm.com> wrote:
> > From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
> >
> > This enable us to do passthrough equivalent security model on NFS directory.
> > NFS server mostly do root squashing and don't support xattr. Hence we cannot
> > use 'passthrough' or 'mapped' security model
> >
> > Also added "mapped-xattr" security to indicate earlier "mapped" security model
> > Older name is still supported.
> 
> Out of interest, did you do any performance comparison with mapped-xattr?
> 

No, I was mostly looking at making virtFS usable on an NFS setup

> 
> > @@ -404,6 +578,10 @@ static int local_fstat(FsContext *fs_ctx, int fid_type,
> >                       &tmp_dev, sizeof(dev_t)) > 0) {
> >                 stbuf->st_rdev = tmp_dev;
> >         }
> > +#if 0
> > +    } else if (fs_ctx->export_flags & V9FS_SM_MAPPED_FILE) {
> > +        /* FIXME!! how to implement that for MAPPED_FILE */
> > +#endif
> 
> Lacking an implementation of fstat seems like a pretty major omission
> that isn't mentioned in the commit message.
> 

If you look at where local_fstat get used currently, We do it only for
lock call, that too to check whether the file is present. We are not
using the stat value returned. So the patchset should be usable as it
is. What i am contemplating is should i add additional code in server
that fallback to path based stat if fs driver callback returned
EOPNOTSUPP error. May be i should do that in the next version.

-aneesh