mds: properly set COMPLETE flag when merging dirfrags

don't keep the COMPLETE flag when merging dirfrags during journal
replay, because it's inconvenience to check if the all dirfrags
under the 'basefrag' are in the cache and complete. One special case
is that newly created dirfrag get fragmented, then the fragment
operation get rolled back. The COMPLETE flag should be preserved in
this case because the dirfrag still doesn't exist on object store.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>

diff --git a/src/mds/CDir.cc b/src/mds/CDir.cc
index de29c6c312d562101472b7e94c02d5c80e32a0a2..f0dc962f1f9c02ac574e953d80c95b1104526915 100644 (file)
--- a/src/mds/CDir.cc
+++ b/src/mds/CDir.cc
@@ -874,7 +874,7 @@ void CDir::split(int bits, list<CDir*>& subs, list<Context*>& waiters, bool repl
   int n = 0;
   for (list<frag_t>::iterator p = frags.begin(); p != frags.end(); ++p) {
     CDir *f = new CDir(inode, *p, cache, is_auth());
-    f->state_set(state & MASK_STATE_FRAGMENT_KEPT);
+    f->state_set(state & (MASK_STATE_FRAGMENT_KEPT | STATE_COMPLETE));
     f->replica_map = replica_map;
     f->dir_auth = dir_auth;
     f->init_fragment_pins();
@@ -983,6 +983,9 @@ void CDir::merge(list<CDir*>& subs, list<Context*>& waiters, bool replay)
     inode->close_dirfrag(dir->get_frag());
   }
 
+  if (is_auth() && !replay)
+    mark_complete();
+
   // FIXME: merge dirty old rstat
   fnode.rstat.version = rstat_version;
   fnode.accounted_rstat = fnode.rstat;

diff --git a/src/mds/CDir.h b/src/mds/CDir.h
index 3507f10b08d86e6053e67dc79f1034bcc966bcdc..30c01822f6a078e7801f4178783d05fa42fa8bc7 100644 (file)
--- a/src/mds/CDir.h
+++ b/src/mds/CDir.h
@@ -129,8 +129,7 @@ public:
    |STATE_FROZENDIR
    |STATE_STICKY);
   static const unsigned MASK_STATE_FRAGMENT_KEPT = 
-  (STATE_DIRTY |
-   STATE_COMPLETE |
+  (STATE_DIRTY|
    STATE_EXPORTBOUND |
    STATE_IMPORTBOUND);
 

diff --git a/src/mds/MDCache.cc b/src/mds/MDCache.cc
index ee6237cd2f0f7a7b293b5d064efffb1e81742299..3a16ae7d5485e021646254e518996146c9df9c53 100644 (file)
--- a/src/mds/MDCache.cc
+++ b/src/mds/MDCache.cc
@@ -11572,8 +11572,15 @@ void MDCache::add_uncommitted_fragment(dirfrag_t basedirfrag, int bits, list<fra
   uf.bits = bits;
   uf.ls = ls;
   ls->uncommitted_fragments.insert(basedirfrag);
-  if (rollback)
+  if (rollback) {
     uf.rollback.swap(*rollback);
+    // preserve COMPLETE flag for newly created dirfrag
+    if (bits > 0 && basedirfrag.frag == frag_t()) {
+      CDir *dir = get_dirfrag(basedirfrag);
+      if (dir && dir->is_complete())
+       uf.complete = true;
+    }
+  }
 }
 
 void MDCache::finish_uncommitted_fragment(dirfrag_t basedirfrag, int op)
@@ -11658,6 +11665,8 @@ void MDCache::rollback_uncommitted_fragments()
        dir->set_version(rollback.fnode.version);
        dir->fnode = rollback.fnode;
 
+       if (uf.complete)
+         dir->mark_complete();
        dir->_mark_dirty(ls);
 
        if (!(dir->fnode.rstat == dir->fnode.accounted_rstat)) {
@@ -11675,7 +11684,7 @@ void MDCache::rollback_uncommitted_fragments()
 
        le->add_orig_frag(dir->get_frag());
        le->metablob.add_dir_context(dir);
-       le->metablob.add_dir(dir, true);
+       le->metablob.add_dir(dir, true, uf.complete);
       }
     }
 

diff --git a/src/mds/MDCache.h b/src/mds/MDCache.h
index ad0f5ff734bbfa16afe0291f3caca9024b9c603e..6087e984742914a18bf08061e6b350b7b5d4405e 100644 (file)
--- a/src/mds/MDCache.h
+++ b/src/mds/MDCache.h
@@ -942,11 +942,12 @@ private:
   struct ufragment {
     int bits;
     bool committed;
+    bool complete;
     LogSegment *ls;
     list<Context*> waiters;
     list<frag_t> old_frags;
     bufferlist rollback;
-    ufragment() : bits(0), committed(false), ls(NULL) {}
+    ufragment() : bits(0), committed(false), complete(false), ls(NULL) {}
   };
   map<dirfrag_t, ufragment> uncommitted_fragments;