nfs3: Dont ref dir fd_t used in hard fh resolution

..because the extra ref was under the mistaken assumption that
directory fd_t will be cached even during hard fh resolution and
that is not the case.

Signed-off-by: Shehjar Tikoo <shehjart@gluster.com>
Signed-off-by: Vijay Bellur <vijay@dev.gluster.com>

BUG: 1397 (Cached dir fd_ts are a leakin')
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=1397

1 files changed, 9 insertions, 7 deletions

diff --git a/xlators/nfs/server/src/nfs3-helpers.c b/xlators/nfs/server/src/nfs3-helpers.c
index 116d145033a..dae2d6a61fc 100644
--- a/xlators/nfs/server/src/nfs3-helpers.c
+++ b/xlators/nfs/server/src/nfs3-helpers.c
@@ -2704,14 +2704,16 @@ nfs3_fh_resolve_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         cs->resolvedloc.path);
 
         nfs_user_root_create (&nfu);
-        /* This function can be called in a recursive code path, so if another
-         * directory was opened in an earlier call, we must unref through this
-         * reference before opening another fd_t.
+        /* Keep this directory fd_t around till we have either:
+         * a. found the entry,
+         * b. exhausted all the entries,
+         * c. decide to step into a child directory.
+         *
+         * This decision is made in nfs3_fh_resolve_check_response.
          */
-        if (cs->resolve_dir_fd)
-                fd_unref (cs->resolve_dir_fd);
-
-        cs->resolve_dir_fd = fd_ref (fd);
+        cs->resolve_dir_fd = fd;
+        gf_log (GF_NFS3, GF_LOG_TRACE, "resolve new fd refed: 0x%lx, ref: %d",
+                (long)cs->resolve_dir_fd, cs->resolve_dir_fd->refcount);
         ret = nfs_readdirp (cs->nfsx, cs->vol, &nfu, fd, GF_NFS3_DTPREF, 0,
                             nfs3_fh_resolve_readdir_cbk, cs);