Merge remote branch 'origin/stable'

Conflicts:
	debian/control

Makefile.am

@@ -6,7 +6,8 @@ SUBDIRS = . src qa man
 EXTRA_DIST += \
 	src/test/run-cli-tests \
 	src/test/cli \
-	src/test/downloads
+	src/test/downloads \
+	udev/50-rbd.rules
 check-local:
 #	Build gtest before we build our own tests. Doing this instead
 #	of SUBDIRS because with that, gtest's own tests would be run

debian/control

@@ -117,7 +117,7 @@ Description: debugging symbols for libcrush1
  .
  This package contains debugging symbols for libcrush.
 
-Package: libcrush1-dev
+Package: libcrush-dev
 Architecture: i386 amd64
 Section: libdevel
 Depends: ${shlibs:Depends}, ${misc:Depends}, libcrush1 (= ${binary:Version})
@@ -160,7 +160,7 @@ Description: debugging symbols for librados
  .
  This package contains debugging symbols for librados.
 
-Package: librados2-dev
+Package: librados-dev
 Architecture: i386 amd64
 Section: libdevel
 Depends: ${shlibs:Depends}, ${misc:Depends}, librados2 (= ${binary:Version})
@@ -196,7 +196,7 @@ Description: debugging symbols for librbd1
  .
  This package contains debugging symbols for librbd1.
 
-Package: librbd1-dev
+Package: librbd-dev
 Architecture: i386 amd64
 Section: libdevel
 Depends: ${shlibs:Depends}, ${misc:Depends}, librbd1 (= ${binary:Version})
@@ -234,7 +234,7 @@ Description: debugging symbols for libceph1
  .
  This package contains debugging symbols for libceph1.
 
-Package: libceph1-dev
+Package: libceph-dev
 Architecture: i386 amd64
 Section: libdevel
 Depends: ${shlibs:Depends}, ${misc:Depends}, libceph1 (= ${binary:Version})

debian/librbd1.install

@@ -1 +1,3 @@
 usr/lib/librbd.so.*
+usr/bin/crbdnamer
+lib/udev/rules.d/50-rbd.rules

debian/rules

@@ -41,6 +41,7 @@ install: build
 	dh_installdirs
 
 	$(MAKE) DESTDIR=$(DESTDIR) install
+	install -D -m 644 udev/50-rbd.rules $(DESTDIR)/lib/udev/rules.d/50-rbd.rules
 
 # Add here commands to install the package into debian/testpack.
 # Build architecture-independent files here.

src/Makefile.am

@@ -9,7 +9,7 @@ CLEANFILES =
 bin_PROGRAMS =
 sbin_PROGRAMS =
 sbin_SCRIPTS =
-bin_SCRIPTS = crun cclass $(srcdir)/cclsinfo cdebugpack
+bin_SCRIPTS = crun cclass $(srcdir)/cclsinfo cdebugpack crbdnamer
 # C/C++ tests to build will be appended to this
 check_PROGRAMS =
 # tests to actually run on "make check"; if you need extra, non-test,
@@ -467,6 +467,7 @@ EXTRA_DIST = $(srcdir)/verify-mds-journal.sh $(srcdir)/vstart.sh $(srcdir)/stop.
 	$(srcdir)/cclass.in $(srcdir)/cdebugpack.in \
 	$(srcdir)/cclsinfo $(srcdir)/make_version $(srcdir)/check_version \
 	$(srcdir)/.git_version \
+	$(srcdir)/crbdnamer \
 	$(ceph_tool_gui_DATA)
 
 # work around old versions of automake that don't define $docdir

src/crbdnamer

@@ -0,0 +1,5 @@
+#!/bin/sh
+
+echo -n `cat /sys/devices/rbd/$1/pool`
+echo -n " "
+echo -n `cat /sys/devices/rbd/$1/name`

src/include/utime.h

@@ -98,7 +98,7 @@ public:
 
   // cast to double
   operator double() {
-    return (double)sec() + ((double)nsec() / 100000000.0L);
+    return (double)sec() + ((double)nsec() / 1000000000.0L);
   }
   operator ceph_timespec() {
     ceph_timespec ts;

src/os/FileStore.cc

@@ -1990,6 +1990,9 @@ unsigned FileStore::_do_transaction(Transaction& t)
       assert(0);
     }
 
+    if (r == -ENOTEMPTY) {
+      assert(0 == "ENOTEMPTY suggests garbage data in osd data dir");
+    }
     if (r == -ENOSPC) {
       // For now, if we hit _any_ ENOSPC, crash, before we do any damage
       // by partially applying transactions.

src/osd/OSD.cc

@@ -1144,7 +1144,8 @@ void OSD::project_pg_history(pg_t pgid, PG::Info::History& h, epoch_t from,
            << ", start " << h
            << dendl;
 
-  for (epoch_t e = osdmap->get_epoch();
+  epoch_t e;
+  for (e = osdmap->get_epoch();
        e > from;
        e--) {
     // verify during intermediate epoch (e-1)
@@ -1177,6 +1178,17 @@ void OSD::project_pg_history(pg_t pgid, PG::Info::History& h, epoch_t from,
       break;
   }
 
+  // base case: these floors should be the creation epoch if we didn't
+  // find any changes.
+  if (e == h.epoch_created) {
+    if (!h.same_acting_since)
+      h.same_acting_since = e;
+    if (!h.same_up_since)
+      h.same_up_since = e;
+    if (!h.same_primary_since)
+      h.same_primary_since = e;
+  }
+
   dout(15) << "project_pg_history end " << h << dendl;
 }
 
@@ -3894,6 +3906,8 @@ void OSD::handle_pg_create(MOSDPGCreate *m)
 
     // figure history
     PG::Info::History history;
+    history.epoch_created = created;
+    history.last_epoch_clean = created;
     project_pg_history(pgid, history, created, up, acting);
     
     // register.

src/osd/PG.cc

@@ -351,7 +351,9 @@ void PG::merge_log(ObjectStore::Transaction& t,
       if (p->version <= log.head) {
 	dout(10) << "merge_log split point is " << *p << dendl;
 
-	if (p->version == log.head)
+	hash_map<sobject_t,Log::Entry*>::const_iterator oldobj = old_objects.find(p->soid);
+	if (oldobj != old_objects.end() &&
+	    oldobj->second->version == p->version)
 	  p++;       // move past the split point, if it also exists in our old log...
 	break;
       }
@@ -843,7 +845,7 @@ void PG::generate_past_intervals()
   }
 
   epoch_t first_epoch = 0;
-  epoch_t stop = MAX(1, info.history.last_epoch_clean);
+  epoch_t stop = MAX(info.history.epoch_created, info.history.last_epoch_clean);
   epoch_t last_epoch = info.history.same_acting_since - 1;
   dout(10) << __func__ << " over epochs " << stop << "-" << last_epoch << dendl;
 
@@ -1632,8 +1634,11 @@ void PG::do_peer(ObjectStore::Transaction& t, list<Context*>& tfin,
     if (pi.is_empty())
       continue;
     if (peer_missing.find(peer) == peer_missing.end()) {
-      if (pi.last_update == pi.last_complete) {
-	dout(10) << " infering no missing (last_update==last_complete) for osd" << peer << dendl;
+      if (pi.last_update == pi.last_complete &&  // peer has no missing
+	  pi.last_update == info.last_update) {  // peer is up to date
+	// replica has no missing and identical log as us.  no need to
+	// pull anything.
+	dout(10) << " infering up to date and no missing (last_update==last_complete) for osd" << peer << dendl;
 	peer_missing[peer].num_missing();  // just create the entry.
 	search_for_missing(peer_info[peer], &peer_missing[peer], peer);
 	continue;
@@ -3723,9 +3728,9 @@ void PG::Missing::add(const sobject_t& oid, eversion_t need, eversion_t have)
   rmissing[need] = oid;
 }
 
-void PG::Missing::rm(const sobject_t& oid, eversion_t when)
+void PG::Missing::rm(const sobject_t& oid, eversion_t v)
 {
-  if (missing.count(oid) && missing[oid].need < when) {
+  if (missing.count(oid) && missing[oid].need <= v) {
     rmissing.erase(missing[oid].need);
     missing.erase(oid);
   }

src/osd/PG.h

@@ -586,7 +586,7 @@ public:
     void add_event(Log::Entry& e);
     void revise_need(sobject_t oid, eversion_t need);
     void add(const sobject_t& oid, eversion_t need, eversion_t have);
-    void rm(const sobject_t& oid, eversion_t when);
+    void rm(const sobject_t& oid, eversion_t v);
     void got(const sobject_t& oid, eversion_t v);
     void got(const std::map<sobject_t, Missing::item>::iterator &m);
 

src/osdc/Filer.cc

@@ -107,7 +107,8 @@ void Filer::_probe(Probe *probe)
        p++) {
     dout(10) << "_probe  probing " << p->oid << dendl;
     C_Probe *c = new C_Probe(this, probe, p->oid);
-    probe->ops[p->oid] = objecter->stat(p->oid, p->oloc, probe->snapid, &c->size, &c->mtime, probe->flags, c);
+    probe->ops[p->oid] = objecter->stat(p->oid, p->oloc, probe->snapid, &c->size, &c->mtime, 
+					probe->flags | CEPH_OSD_FLAG_WRITE, c);
   }
 }
 

udev/50-rbd.rules

@@ -0,0 +1 @@
+KERNEL=="rbd[0-9]*", PROGRAM="/usr/bin/crbdnamer %n", SYMLINK+="rbd/%c{1}/%c{2}:%n"