Merge remote-tracking branch 'gh/next'

configure.ac

@@ -8,7 +8,7 @@ AC_PREREQ(2.59)
 # VERSION define is not used by the code.  It gets a version string
 # from 'git describe'; see src/ceph_ver.[ch]
 
-AC_INIT([ceph], [0.88], [ceph-devel@vger.kernel.org])
+AC_INIT([ceph], [0.89], [ceph-devel@vger.kernel.org])
 
 # Create release string.  Used with VERSION for RPMs.
 RPM_RELEASE=0

debian/changelog

@@ -1,3 +1,9 @@
+ceph (0.89-1) stable; urgency=low
+
+  * New upstream release
+
+ -- Alfredo Deza <adeza@redhat.com>  Wed, 03 Dec 2014 08:18:33 -0800
+
 ceph (0.88-1) stable; urgency=low
 
   * New upstream release

qa/workunits/cephtool/test.sh

@@ -765,6 +765,21 @@ function test_mon_mds()
   fail_all_mds
   ceph fs rm cephfs --yes-i-really-mean-it
 
+  # Create a FS and check that we can subsequently add a cache tier to it
+  ceph fs new cephfs fs_metadata fs_data
+
+  # Adding overlay to FS pool should be permitted, RADOS clients handle this.
+  ceph osd tier add fs_metadata mds-tier
+  ceph osd tier cache-mode mds-tier writeback
+  ceph osd tier set-overlay fs_metadata mds-tier
+
+  # Clean up FS
+  fail_all_mds
+  ceph fs rm cephfs --yes-i-really-mean-it
+
+  # Clean up overlay/tier relationship
+  ceph osd tier remove-overlay fs_metadata
+  ceph osd tier remove fs_metadata mds-tier
 
   ceph osd pool delete mds-tier mds-tier --yes-i-really-really-mean-it
   ceph osd pool delete mds-ec-pool mds-ec-pool --yes-i-really-really-mean-it
@@ -1299,39 +1314,31 @@ function test_osd_bench()
 #
 
 set +x
-MON_TESTS=(
-  mon_injectargs
-  mon_injectargs_SI
-  tiering
-  auth
-  auth_profiles
-  mon_misc
-  mon_mon
-  mon_osd
-  mon_osd_pool
-  mon_osd_pool_quota
-  mon_pg
-  mon_osd_pool_set
-  mon_osd_tiered_pool_set
-  mon_osd_erasure_code
-  mon_osd_misc
-  mon_heap_profiler
-)
+MON_TESTS+=" mon_injectargs"
+MON_TESTS+=" mon_injectargs_SI"
+MON_TESTS+=" tiering"
+MON_TESTS+=" auth"
+MON_TESTS+=" auth_profiles"
+MON_TESTS+=" mon_misc"
+MON_TESTS+=" mon_mon"
+MON_TESTS+=" mon_osd"
+MON_TESTS+=" mon_osd_pool"
+MON_TESTS+=" mon_osd_pool_quota"
+MON_TESTS+=" mon_pg"
+MON_TESTS+=" mon_osd_pool_set"
+MON_TESTS+=" mon_osd_tiered_pool_set"
+MON_TESTS+=" mon_osd_erasure_code"
+MON_TESTS+=" mon_osd_misc"
+MON_TESTS+=" mon_heap_profiler"
 
-OSD_TESTS=(
-  osd_bench
-)
+OSD_TESTS+=" osd_bench"
 
-MDS_TESTS=(
-  mds_tell
-  mon_mds
-)
+MDS_TESTS+=" mds_tell"
+MDS_TESTS+=" mon_mds"
 
-TESTS=(
-  $MON_TESTS
-  $OSD_TESTS
-  $MDS_TESTS
-)
+TESTS+=$MON_TESTS
+TESTS+=$OSD_TESTS
+TESTS+=$MDS_TESTS
 
 #
 # "main" follows
@@ -1340,7 +1347,7 @@ TESTS=(
 function list_tests()
 {
   echo "AVAILABLE TESTS"
-  for i in ${TESTS[@]}; do
+  for i in $TESTS; do
     echo "  $i"
   done
 }
@@ -1368,13 +1375,13 @@ while [[ $# -gt 0 ]]; do
       sanity_check=false
       ;;
     "--test-mon" )
-      tests_to_run=("${tests_to_run[@]}" $MON_TESTS)
+      tests_to_run+="$MON_TESTS"
       ;;
     "--test-osd" )
-      tests_to_run=("${tests_to_run[@]}" $OSD_TESTS)
+      tests_to_run+="$OSD_TESTS"
       ;;
     "--test-mds" )
-      tests_to_run=("${tests_to_run[@]}" $MDS_TESTS)
+      tests_to_run+="$MDS_TESTS"
       ;;
     "-t" )
       shift
@@ -1383,7 +1390,7 @@ while [[ $# -gt 0 ]]; do
         usage ;
         exit 1
       fi
-      tests_to_run=("${tests_to_run[@]}" "$1")
+      tests_to_run+=" $1"
       ;;
     "-h" )
       usage ;
@@ -1398,14 +1405,14 @@ if [[ $do_list -eq 1 ]]; then
   exit 0
 fi
 
-if [[ ${#tests_to_run[@]} -eq 0 ]]; then
-  tests_to_run=("${TESTS[@]}")
+if test -z "$tests_to_run" ; then
+  tests_to_run="$TESTS"
 fi
 
 if $sanity_check ; then
     wait_no_osd_down
 fi
-for i in ${tests_to_run[@]}; do
+for i in $tests_to_run; do
   if $sanity_check ; then
       check_no_osd_down
   fi

src/client/Client.cc

@@ -216,7 +216,7 @@ Client::Client(Messenger *m, MonClient *mc)
 				  cct->_conf->client_oc_max_dirty_age,
 				  true);
   objecter_finisher.start();
-  filer = new Filer(objecter);
+  filer = new Filer(objecter, &objecter_finisher);
 }
 
 
@@ -543,7 +543,7 @@ void Client::trim_cache_for_reconnect(MetaSession *s)
 		 << " trimmed " << trimmed << " dentries" << dendl;
 
   if (s->caps.size() > 0)
-    _invalidate_kernel_dcache();
+    _invalidate_kernel_dcache(s);
 }
 
 void Client::trim_dentry(Dentry *dn)
@@ -3277,16 +3277,19 @@ void Client::remove_session_caps(MetaSession *s)
   sync_cond.Signal();
 }
 
-void Client::_invalidate_kernel_dcache()
+void Client::_invalidate_kernel_dcache(MetaSession *s)
 {
-  // notify kernel to invalidate top level directory entries. As a side effect,
-  // unused inodes underneath these entries get pruned.
-  if (dentry_invalidate_cb && root->dir) {
-    for (ceph::unordered_map<string, Dentry*>::iterator p = root->dir->dentries.begin();
-	 p != root->dir->dentries.end();
-	 ++p) {
-      if (p->second->inode)
-	_schedule_invalidate_dentry_callback(p->second, false);
+  if (!dentry_invalidate_cb)
+    return;
+
+  for (xlist<Cap*>::iterator p = s->caps.begin(); !p.end(); ++p) {
+    Inode *in = (*p)->inode;
+    if (in->dn_set.empty())
+      continue;
+    for (set<Dentry*>::iterator q = in->dn_set.begin();
+	 q != in->dn_set.end();
+	 ++q) {
+      _schedule_invalidate_dentry_callback(*q, false);
     }
   }
 }
@@ -3320,14 +3323,8 @@ void Client::trim_caps(MetaSession *s, int max)
       while (q != in->dn_set.end()) {
 	Dentry *dn = *q++;
 	if (dn->lru_is_expireable()) {
-          if (dn->dir->parent_inode->ino == MDS_INO_ROOT) {
-            // Only issue one of these per DN for inodes in root: handle
-            // others more efficiently by calling for root-child DNs at
-            // the end of this function.
-            _schedule_invalidate_dentry_callback(dn, true);
-          }
+	  _schedule_invalidate_dentry_callback(dn, false);
 	  trim_dentry(dn);
-
         } else {
           ldout(cct, 20) << "  not expirable: " << dn->name << dendl;
 	  all = false;
@@ -3348,9 +3345,6 @@ void Client::trim_caps(MetaSession *s, int max)
     }
   }
   s->s_cap_iterator = NULL;
-
-  if (s->caps.size() > max)
-    _invalidate_kernel_dcache();
 }
 
 void Client::mark_caps_dirty(Inode *in, int caps)

src/client/Client.h

@@ -429,7 +429,7 @@ protected:
   void trim_cache_for_reconnect(MetaSession *s);
   void trim_dentry(Dentry *dn);
   void trim_caps(MetaSession *s, int max);
-  void _invalidate_kernel_dcache();
+  void _invalidate_kernel_dcache(MetaSession *s);
   
   void dump_inode(Formatter *f, Inode *in, set<Inode*>& did, bool disconnected);
   void dump_cache(Formatter *f);  // debug

src/mds/MDS.cc

@@ -129,7 +129,7 @@ MDS::MDS(const std::string &n, Messenger *m, MonClient *mc) :
   objecter = new Objecter(m->cct, messenger, monc, 0, 0);
   objecter->unset_honor_osdmap_full();
 
-  filer = new Filer(objecter);
+  filer = new Filer(objecter, &finisher);
 
   mdcache = new MDCache(this);
   mdlog = new MDLog(this);

src/mon/MDSMonitor.cc

@@ -577,8 +577,8 @@ void MDSMonitor::get_health(list<pair<health_status_t, string> >& summary,
   mdsmap.get_health(summary, detail);
 
   // For each MDS GID...
-  for (std::map<mds_gid_t, MDSMap::mds_info_t>::const_iterator i = pending_mdsmap.mds_info.begin();
-      i != pending_mdsmap.mds_info.end(); ++i) {
+  for (std::map<mds_gid_t, MDSMap::mds_info_t>::const_iterator i = mdsmap.mds_info.begin();
+      i != mdsmap.mds_info.end(); ++i) {
     // Decode MDSHealth
     bufferlist bl;
     mon->store->get(MDS_HEALTH_PREFIX, stringify(i->first), bl);

src/mon/OSDMonitor.cc

@@ -5685,10 +5685,6 @@ done:
       goto reply;
     }
 
-    if (!_check_remove_tier(pool_id, p, &err, &ss)) {
-      goto reply;
-    }
-
     // go
     pg_pool_t *np = pending_inc.get_new_pool(pool_id, p);
     np->read_tier = overlaypool_id;

src/mon/PaxosService.h

@@ -566,7 +566,6 @@ public:
    */
   bool is_readable(version_t ver = 0) {
     if (ver > get_last_committed() ||
-	is_proposing() ||
 	!paxos->is_readable(0) ||
 	get_last_committed() == 0)
       return false;

src/osdc/Filer.cc

@@ -25,6 +25,7 @@
 
 #include "include/Context.h"
 
+#include "common/Finisher.h"
 #include "common/config.h"
 
 #define dout_subsys ceph_subsys_filer
@@ -134,8 +135,9 @@ void Filer::_probe(Probe *probe)
   for (std::vector<ObjectExtent>::iterator i = stat_extents.begin();
       i != stat_extents.end(); ++i) {
     C_Probe *c = new C_Probe(this, probe, i->oid);
-    objecter->stat(i->oid, i->oloc, probe->snapid, &c->size, &c->mtime, 
-                   probe->flags | CEPH_OSD_FLAG_RWORDERED, c);
+    objecter->stat(i->oid, i->oloc, probe->snapid, &c->size, &c->mtime,
+		   probe->flags | CEPH_OSD_FLAG_RWORDERED,
+		   new C_OnFinisher(c, finisher));
   }
 }
 
@@ -339,8 +341,8 @@ void Filer::_do_purge_range(PurgeRange *pr, int fin)
     const OSDMap *osdmap = objecter->get_osdmap_read();
     const object_locator_t oloc = osdmap->file_to_object_locator(pr->layout);
     objecter->put_osdmap_read();
-    objecter->remove(oid, oloc, pr->snapc, pr->mtime, pr->flags,
-		     NULL, new C_PurgeRange(this, pr));
+    objecter->remove(oid, oloc, pr->snapc, pr->mtime, pr->flags, NULL,
+		     new C_OnFinisher(new C_PurgeRange(this, pr), finisher));
   }
 }
 

src/osdc/Filer.h

@@ -35,7 +35,7 @@
 class Context;
 class Messenger;
 class OSDMap;
-
+class Finisher;
 
 
 /**** Filer interface ***/
@@ -43,6 +43,7 @@ class OSDMap;
 class Filer {
   CephContext *cct;
   Objecter   *objecter;
+  Finisher   *finisher;
   
   // probes
   struct Probe {
@@ -88,7 +89,7 @@ class Filer {
   Filer(const Filer& other);
   const Filer operator=(const Filer& other);
 
-  Filer(Objecter *o) : cct(o->cct), objecter(o) {}
+  Filer(Objecter *o, Finisher *f) : cct(o->cct), objecter(o), finisher(f) {}
   ~Filer() {}
 
   bool is_active() {

src/osdc/Journaler.h

@@ -377,7 +377,7 @@ public:
     ino(ino_), pg_pool(pool), readonly(true),
     stream_format(-1), journal_stream(-1),
     magic(mag),
-    objecter(obj), filer(objecter), logger(l), logger_key_lat(lkey),
+    objecter(obj), filer(objecter, f), logger(l), logger_key_lat(lkey),
     timer(tim), delay_flush_event(0),
     state(STATE_UNDEF), error(0),
     prezeroing_pos(0), prezero_pos(0), write_pos(0), flush_pos(0), safe_pos(0),

src/test/vstart_wrapper.sh

@@ -29,7 +29,7 @@ function vstart_setup()
     export LC_ALL=C # some tests are vulnerable to i18n
     ./vstart.sh \
         -o 'paxos propose interval = 0.01' \
-        -n -l $VSTART_ARGS || return 1
+        -n -l $CEPH_START || return 1
     export PATH=.:$PATH
     export CEPH_CONF=$CEPH_DIR/ceph.conf
 

src/tools/cephfs/Dumper.cc

@@ -85,7 +85,7 @@ int Dumper::dump(const char *dump_file)
 
   cout << "journal is " << start << "~" << len << std::endl;
 
-  Filer filer(objecter);
+  Filer filer(objecter, &finisher);
   bufferlist bl;
 
   C_SaferCond cond;
@@ -238,7 +238,7 @@ int Dumper::undump(const char *dump_file)
     return r;
   }
 
-  Filer filer(objecter);
+  Filer filer(objecter, &finisher);
 
   /* Erase any objects at the end of the region to which we shall write
    * the new log data.  This is to avoid leaving trailing junk after