nfsd: fix EXDEV checking in rename
authorJ. Bruce Fields <bfields@redhat.com>
Mon, 15 Apr 2013 20:03:46 +0000 (16:03 -0400)
committerJ. Bruce Fields <bfields@redhat.com>
Fri, 26 Apr 2013 20:18:15 +0000 (16:18 -0400)
We again check for the EXDEV a little later on, so the first check is
redundant.  This check is also slightly racier, since a badly timed
eviction from the export cache could leave us with the two fh_export
pointers pointing to two different cache entries which each refer to the
same underlying export.

It's better to compare vfsmounts as the later check does, but that
leaves a minor security hole in the case where the two exports refer to
two different directories especially if (for example) they have
different root-squashing options.

So, compare ex_path.dentry too.

Reported-by: Joe Habermann <joe.habermann@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
fs/nfsd/vfs.c

index 2b2e2396a86913b4d4e69c376852ded82dd9be3a..84ce601d80632ba1dd090cd319766cecbe4cac98 100644 (file)
@@ -1758,10 +1758,6 @@ nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
        tdentry = tfhp->fh_dentry;
        tdir = tdentry->d_inode;
 
-       err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
-       if (ffhp->fh_export != tfhp->fh_export)
-               goto out;
-
        err = nfserr_perm;
        if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
                goto out;
@@ -1802,6 +1798,8 @@ nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
        host_err = -EXDEV;
        if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
                goto out_dput_new;
+       if (ffhp->fh_export->ex_path.dentry != tfhp->fh_export->ex_path.dentry)
+               goto out_dput_new;
 
        host_err = nfsd_break_lease(odentry->d_inode);
        if (host_err)