mds: revert commit 15a5d37a

commit 15a5d37a (mds: fix race between scatter gather and dirfrag export)
is incomplete, it doesn't handles the race that no fragstat/neststat is
gathered. Previous commit prevents scatter gather during exporting dir,
which eliminates races of this type.

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

diff --git a/src/mds/CDir.cc b/src/mds/CDir.cc
index 3697929e86282130101904d7ddb14b2329a536d5..de29c6c312d562101472b7e94c02d5c80e32a0a2 100644 (file)
--- a/src/mds/CDir.cc
+++ b/src/mds/CDir.cc
@@ -154,7 +154,6 @@ ostream& CDir::print_db_line_prefix(ostream& out)
 // CDir
 
 CDir::CDir(CInode *in, frag_t fg, MDCache *mdcache, bool auth) :
-  mseq(0),
   dirty_rstat_inodes(member_offset(CInode, dirty_rstat_item)),
   item_dirty(this), item_new(this),
   pop_me(ceph_clock_now(g_ceph_context)),
@@ -2102,8 +2101,6 @@ void CDir::_committed(version_t v)
 void CDir::encode_export(bufferlist& bl)
 {
   assert(!is_projected());
-  ceph_seq_t seq = mseq + 1;
-  ::encode(seq, bl);
   ::encode(first, bl);
   ::encode(fnode, bl);
   ::encode(dirty_old_rstat, bl);
@@ -2133,7 +2130,6 @@ void CDir::finish_export(utime_t now)
 
 void CDir::decode_import(bufferlist::iterator& blp, utime_t now, LogSegment *ls)
 {
-  ::decode(mseq, blp);
   ::decode(first, blp);
   ::decode(fnode, blp);
   ::decode(dirty_old_rstat, blp);

diff --git a/src/mds/CDir.h b/src/mds/CDir.h
index db11c060623b4a7ac1f97f4b72d4911d89a51bb6..2c0b7e45a67cbd95b9f5bed16a8fcf335efd40cb 100644 (file)
--- a/src/mds/CDir.h
+++ b/src/mds/CDir.h
@@ -170,7 +170,6 @@ public:
 
   fnode_t fnode;
   snapid_t first;
-  ceph_seq_t mseq; // migrate sequence
   map<snapid_t,old_rstat_t> dirty_old_rstat;  // [value.first,key]
 
   // my inodes with dirty rstat data
@@ -555,7 +554,6 @@ public:
   void encode_export(bufferlist& bl);
   void finish_export(utime_t now);
   void abort_export() {
-    mseq += 2;
     put(PIN_TEMPEXPORTING);
   }
   void decode_import(bufferlist::iterator& blp, utime_t now, LogSegment *ls);

diff --git a/src/mds/CInode.cc b/src/mds/CInode.cc
index 71550d155322850d105dc0083fbe3e4a241380af..48ee839da9eb46c9ff9aba0f5cf58d153a5572a6 100644 (file)
--- a/src/mds/CInode.cc
+++ b/src/mds/CInode.cc
@@ -1231,7 +1231,6 @@ void CInode::encode_lock_state(int type, bufferlist& bl)
          dout(20) << fg << "           fragstat " << pf->fragstat << dendl;
          dout(20) << fg << " accounted_fragstat " << pf->accounted_fragstat << dendl;
          ::encode(fg, tmp);
-         ::encode(dir->mseq, tmp);
          ::encode(dir->first, tmp);
          ::encode(pf->fragstat, tmp);
          ::encode(pf->accounted_fragstat, tmp);
@@ -1267,7 +1266,6 @@ void CInode::encode_lock_state(int type, bufferlist& bl)
          dout(10) << fg << " " << pf->rstat << dendl;
          dout(10) << fg << " " << dir->dirty_old_rstat << dendl;
          ::encode(fg, tmp);
-         ::encode(dir->mseq, tmp);
          ::encode(dir->first, tmp);
          ::encode(pf->rstat, tmp);
          ::encode(pf->accounted_rstat, tmp);
@@ -1426,12 +1424,10 @@ void CInode::decode_lock_state(int type, bufferlist& bl)
       dout(10) << " ...got " << n << " fragstats on " << *this << dendl;
       while (n--) {
        frag_t fg;
-       ceph_seq_t mseq;
        snapid_t fgfirst;
        frag_info_t fragstat;
        frag_info_t accounted_fragstat;
        ::decode(fg, p);
-       ::decode(mseq, p);
        ::decode(fgfirst, p);
        ::decode(fragstat, p);
        ::decode(accounted_fragstat, p);
@@ -1444,12 +1440,6 @@ void CInode::decode_lock_state(int type, bufferlist& bl)
          assert(dir);                // i am auth; i had better have this dir open
          dout(10) << fg << " first " << dir->first << " -> " << fgfirst
                   << " on " << *dir << dendl;
-         if (dir->fnode.fragstat.version == get_projected_inode()->dirstat.version &&
-             ceph_seq_cmp(mseq, dir->mseq) < 0) {
-           dout(10) << " mseq " << mseq << " < " << dir->mseq << ", ignoring" << dendl;
-           continue;
-         }
-         dir->mseq = mseq;
          dir->first = fgfirst;
          dir->fnode.fragstat = fragstat;
          dir->fnode.accounted_fragstat = accounted_fragstat;
@@ -1494,13 +1484,11 @@ void CInode::decode_lock_state(int type, bufferlist& bl)
       ::decode(n, p);
       while (n--) {
        frag_t fg;
-       ceph_seq_t mseq;
        snapid_t fgfirst;
        nest_info_t rstat;
        nest_info_t accounted_rstat;
        map<snapid_t,old_rstat_t> dirty_old_rstat;
        ::decode(fg, p);
-       ::decode(mseq, p);
        ::decode(fgfirst, p);
        ::decode(rstat, p);
        ::decode(accounted_rstat, p);
@@ -1515,12 +1503,6 @@ void CInode::decode_lock_state(int type, bufferlist& bl)
          assert(dir);                // i am auth; i had better have this dir open
          dout(10) << fg << " first " << dir->first << " -> " << fgfirst
                   << " on " << *dir << dendl;
-         if (dir->fnode.rstat.version == get_projected_inode()->rstat.version &&
-             ceph_seq_cmp(mseq, dir->mseq) < 0) {
-           dout(10) << " mseq " << mseq << " < " << dir->mseq << ", ignoring" << dendl;
-           continue;
-         }
-         dir->mseq = mseq;
          dir->first = fgfirst;
          dir->fnode.rstat = rstat;
          dir->fnode.accounted_rstat = accounted_rstat;
@@ -1650,36 +1632,6 @@ void CInode::start_scatter(ScatterLock *lock)
   }
 }
 
-/*
- * set dirfrag_version to inode_version - 1. so that we can use dirfrag version
- * to check if we have gathered scatter state for a given dirfrag.
- */
-void CInode::start_scatter_gather(ScatterLock *lock, int auth)
-{
-  assert(is_auth());
-  inode_t *pi = get_projected_inode();
-
-  for (map<frag_t,CDir*>::iterator p = dirfrags.begin();
-       p != dirfrags.end();
-       ++p) {
-    CDir *dir = p->second;
-
-    if (dir->is_auth())
-      continue;
-    if (auth >= 0 && dir->authority().first != auth)
-      continue;
-
-    switch (lock->get_type()) {
-      case CEPH_LOCK_IFILE:
-       dir->fnode.fragstat.version = pi->dirstat.version - 1;
-       break;
-      case CEPH_LOCK_INEST:
-       dir->fnode.rstat.version = pi->rstat.version - 1;
-       break;
-    }
-  }
-}
-
 struct C_Inode_FragUpdate : public Context {
   CInode *in;
   CDir *dir;

diff --git a/src/mds/CInode.h b/src/mds/CInode.h
index d10ba24940f3ffc5b3086d724a5bfb16e32d01a7..3b50560fb5491b8f32cfcce12bec44235d6d3f5b 100644 (file)
--- a/src/mds/CInode.h
+++ b/src/mds/CInode.h
@@ -654,7 +654,6 @@ public:
   void clear_scatter_dirty();  // on rejoin ack
 
   void start_scatter(ScatterLock *lock);
-  void start_scatter_gather(ScatterLock *lock, int auth=-1);
   void finish_scatter_update(ScatterLock *lock, CDir *dir,
                             version_t inode_version, version_t dir_accounted_version);
   void finish_scatter_gather_update(int type);

diff --git a/src/mds/Locker.cc b/src/mds/Locker.cc
index 827bb9c26f24d817ec8912c63a6342720bb83832..1906c49bc4c1338124a3ab32b9dd86daf8f0f8dc 100644 (file)
--- a/src/mds/Locker.cc
+++ b/src/mds/Locker.cc
@@ -777,9 +777,8 @@ void Locker::eval_gather(SimpleLock *lock, bool first, bool *pneed_issue, list<C
          lock->get_parent()->is_replicated()) {
        dout(10) << " finished (local) gather for mix->lock, now gathering from replicas" << dendl;
        send_lock_message(lock, LOCK_AC_LOCK);
-       lock->set_state(LOCK_MIX_LOCK2);
        lock->init_gather();
-       in->start_scatter_gather(static_cast<ScatterLock *>(lock));
+       lock->set_state(LOCK_MIX_LOCK2);
        return;
       }
 
@@ -3521,7 +3520,7 @@ bool Locker::simple_sync(SimpleLock *lock, bool *need_issue)
   assert(lock->is_stable());
 
   CInode *in = 0;
-  if (lock->get_type() != CEPH_LOCK_DN)
+  if (lock->get_cap_shift())
     in = static_cast<CInode *>(lock->get_parent());
 
   int old_state = lock->get_state();
@@ -3543,11 +3542,10 @@ bool Locker::simple_sync(SimpleLock *lock, bool *need_issue)
     if (lock->get_parent()->is_replicated() && old_state == LOCK_MIX) {
       send_lock_message(lock, LOCK_AC_SYNC);
       lock->init_gather();
-      in->start_scatter_gather(static_cast<ScatterLock *>(lock));
       gather++;
     }
     
-    if (lock->get_cap_shift() && in->is_head()) {
+    if (in && in->is_head()) {
       if (in->issued_caps_need_gather(lock)) {
        if (need_issue)
          *need_issue = true;
@@ -3658,7 +3656,7 @@ void Locker::simple_lock(SimpleLock *lock, bool *need_issue)
   assert(lock->get_state() != LOCK_LOCK);
   
   CInode *in = 0;
-  if (lock->get_type() != CEPH_LOCK_DN)
+  if (lock->get_cap_shift())
     in = static_cast<CInode *>(lock->get_parent());
 
   int old_state = lock->get_state();
@@ -3685,7 +3683,7 @@ void Locker::simple_lock(SimpleLock *lock, bool *need_issue)
   }
   if (lock->is_rdlocked())
     gather++;
-  if (lock->get_cap_shift() && in->is_head()) {
+  if (in && in->is_head()) {
     if (in->issued_caps_need_gather(lock)) {
       if (need_issue)
        *need_issue = true;
@@ -3718,8 +3716,6 @@ void Locker::simple_lock(SimpleLock *lock, bool *need_issue)
       gather++;
       send_lock_message(lock, LOCK_AC_LOCK);
       lock->init_gather();
-      if (lock->get_state() == LOCK_MIX_LOCK2)
-       in->start_scatter_gather(static_cast<ScatterLock *>(lock));
     }
   }
 
@@ -4125,9 +4121,8 @@ void Locker::scatter_tempsync(ScatterLock *lock, bool *need_issue)
 
   if (lock->get_state() == LOCK_MIX_TSYN &&
       in->is_replicated()) {
-    send_lock_message(lock, LOCK_AC_LOCK);
     lock->init_gather();
-    in->start_scatter_gather(static_cast<ScatterLock *>(lock));
+    send_lock_message(lock, LOCK_AC_LOCK);
     gather++;
   }
 
@@ -4451,8 +4446,6 @@ void Locker::file_excl(ScatterLock *lock, bool *need_issue)
       lock->get_state() != LOCK_XSYN_EXCL) {  // if we were lock, replicas are already lock.
     send_lock_message(lock, LOCK_AC_LOCK);
     lock->init_gather();
-    if (lock->get_state() == LOCK_MIX_EXCL)
-      in->start_scatter_gather(static_cast<ScatterLock *>(lock));
     gather++;
   }
   if (lock->is_leased()) {

diff --git a/src/mds/MDCache.cc b/src/mds/MDCache.cc
index 2e0f1c734238709ee49b22e0758b6d10f03358ec..0e2073223ab50e9e40baff30b79b053098dba757 100644 (file)
--- a/src/mds/MDCache.cc
+++ b/src/mds/MDCache.cc
@@ -3812,7 +3812,6 @@ void MDCache::rejoin_walk(CDir *dir, MMDSCacheRejoin *rejoin)
     dout(15) << " add_strong_dirfrag " << *dir << dendl;
     rejoin->add_strong_dirfrag(dir->dirfrag(), dir->get_replica_nonce(), dir->get_dir_rep());
     dir->state_set(CDir::STATE_REJOINING);
-    dir->mseq = 0;
 
     for (CDir::map_t::iterator p = dir->items.begin();
         p != dir->items.end();
@@ -3974,15 +3973,11 @@ void MDCache::handle_cache_rejoin_weak(MMDSCacheRejoin *weak)
        ++p) {
     CInode *in = get_inode(p->first);
     assert(in);
-    if (survivor) {
-      in->start_scatter_gather(&in->filelock, from);
-      in->start_scatter_gather(&in->nestlock, from);
-    } else {
-      rejoin_potential_updated_scatterlocks.insert(in);
-    }
     in->decode_lock_state(CEPH_LOCK_IFILE, p->second.file);
     in->decode_lock_state(CEPH_LOCK_INEST, p->second.nest);
     in->decode_lock_state(CEPH_LOCK_IDFT, p->second.dft);
+    if (!survivor)
+      rejoin_potential_updated_scatterlocks.insert(in);
   }
 
   // recovering peer may send incorrect dirfrags here.  we need to