client: nfs-ganesha with cephfs client, removing dir reports not empty

the problem was due to the cephfs client side wrongly filled the
directory offset from its readdir_cache to nfs-ganesha as its mdcache
dirent cookie

Fixes: http://tracker.ceph.com/issues/40746
Signed-off-by: Peng Xie <peng.hse@xtaotech.com>
(cherry picked from commit 2cce2d4c7d5f3a274c206b4090057760f357ecb0)

diff --git a/src/client/Client.cc b/src/client/Client.cc
index 35ea7100eae8388b52552a57a328c379e3ea4980..49c15f942818a31cbdaeb3f1e3d3d703ebb414b2 100644 (file)
--- a/src/client/Client.cc
+++ b/src/client/Client.cc
@@ -7851,12 +7851,12 @@ int Client::_readdir_cache_cb(dir_result_t *dirp, add_dirent_cb_t cb, void *p,
     fill_statx(dn->inode, caps, &stx);
 
     uint64_t next_off = dn->offset + 1;
+    fill_dirent(&de, dn->name.c_str(), stx.stx_mode, stx.stx_ino, next_off);
     ++pd;
     if (pd == dir->readdir_cache.end())
       next_off = dir_result_t::END;
 
     Inode *in = NULL;
-    fill_dirent(&de, dn->name.c_str(), stx.stx_mode, stx.stx_ino, next_off);
     if (getref) {
       in = dn->inode.get();
       _ll_get(in);