Merge branch 'unstable' into asyncmds

Conflicts:

	src/mon/OSDMonitor.cc

src/TODO

@@ -22,7 +22,6 @@ big items
 - client, user authentication
 - cas
 - osd failure declarations
-- libuuid?
 
 
 repair

src/cosd.cc

@@ -99,7 +99,7 @@ int main(int argc, const char **argv)
 
   if (whoami < 0) {
     nstring magic;
-    ceph_fsid fsid;
+    ceph_fsid_t fsid;
     int r = OSD::peek_super(dev, magic, fsid, whoami);
     if (r < 0) {
       cerr << "unable to determine OSD identity from superblock on " << dev << ": " << strerror(-r) << std::endl;
@@ -109,7 +109,7 @@ int main(int argc, const char **argv)
       cerr << "OSD magic " << magic << " != my " << CEPH_OSD_ONDISK_MAGIC << std::endl;
       exit(1);
     }
-    if (!ceph_fsid_equal(&fsid, &monmap.fsid)) {
+    if (ceph_fsid_compare(&fsid, &monmap.fsid)) {
       cerr << "OSD fsid " << fsid << " != monmap fsid " << monmap.fsid << std::endl;
       exit(1);
     }

src/ebofs/Ebofs.cc

@@ -2493,6 +2493,11 @@ unsigned Ebofs::_apply_transaction(Transaction& t)
     int op = t.get_op();
     switch (op) {
  
+    case Transaction::OP_STARTSYNC:
+      dirty = true;
+      commit_cond.Signal();
+      break;
+
     case Transaction::OP_TOUCH:
       {
 	coll_t cid = t.get_cid();

src/include/ceph_fs.h

@@ -61,19 +61,15 @@ typedef __le64 ceph_version_t;
 typedef __le64 ceph_tid_t;      /* transaction id */
 typedef __le32 ceph_epoch_t;
 
-
 /*
  * fs id
  */
-struct ceph_fsid {
+typedef struct { unsigned char fsid[16]; } ceph_fsid_t;
-	__le64 major;
-	__le64 minor;
-} __attribute__ ((packed));
 
-static inline int ceph_fsid_equal(const struct ceph_fsid *a,
+static inline int ceph_fsid_compare(const ceph_fsid_t *a,
-				  const struct ceph_fsid *b)
+				    const ceph_fsid_t *b)
 {
-	return a->major == b->major && a->minor == b->minor;
+	return memcmp(a, b, sizeof(*a));
 }
 
 
@@ -541,7 +537,7 @@ struct ceph_msg_footer {
 
 
 struct ceph_mon_statfs {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	__le64 tid;
 };
 
@@ -553,18 +549,18 @@ struct ceph_statfs {
 };
 
 struct ceph_mon_statfs_reply {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	__le64 tid;
 	struct ceph_statfs st;
 };
 
 struct ceph_osd_getmap {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	__le32 start;
 } __attribute__ ((packed));
 
 struct ceph_mds_getmap {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	__le32 want;
 } __attribute__ ((packed));
 
@@ -1115,6 +1111,7 @@ enum {
 
 	/* fancy write */
 	CEPH_OSD_OP_APPEND     = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 6,
+	CEPH_OSD_OP_STARTSYNC  = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 7,
 };
 
 static inline int ceph_osd_op_type_lock(int op)

src/include/types.h

@@ -197,7 +197,7 @@ struct ltstr
 
 #include "encoding.h"
 
-WRITE_RAW_ENCODER(ceph_fsid)
+WRITE_RAW_ENCODER(ceph_fsid_t)
 WRITE_RAW_ENCODER(ceph_file_layout)
 WRITE_RAW_ENCODER(ceph_mds_request_head)
 WRITE_RAW_ENCODER(ceph_mds_caps)
@@ -370,8 +370,12 @@ inline ostream& operator<<(ostream& out, const SnapContext& snapc) {
 
 // --
 
-inline ostream& operator<<(ostream& out, const ceph_fsid& f) {
+inline ostream& operator<<(ostream& out, const ceph_fsid_t& f) {
-  return out << hex << f.major << '.' << f.minor << dec;
+  char b[37];
+  sprintf(b, "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
+	  f.fsid[0], f.fsid[1], f.fsid[2], f.fsid[3], f.fsid[4], f.fsid[5], f.fsid[6], f.fsid[7],
+	  f.fsid[8], f.fsid[9], f.fsid[10], f.fsid[11], f.fsid[12], f.fsid[13], f.fsid[14], f.fsid[15]);
+  return out << b;
 }
 
 inline ostream& operator<<(ostream& out, const ceph_osd_op& op) {

src/kernel/mds_client.c

@@ -2199,9 +2199,10 @@ void ceph_mdsc_handle_map(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
 	void *p = msg->front.iov_base;
 	void *end = p + msg->front.iov_len;
 	struct ceph_mdsmap *newmap, *oldmap;
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	int err = -EINVAL;
 	int from;
+	__le64 major, minor;
 
 	if (le32_to_cpu(msg->hdr.src.name.type) == CEPH_ENTITY_TYPE_MDS)
 		from = le32_to_cpu(msg->hdr.src.name.num);
@@ -2209,9 +2210,11 @@ void ceph_mdsc_handle_map(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
 		from = -1;
 
 	ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
-	ceph_decode_64_le(&p, fsid.major);
+	ceph_decode_64_le(&p, major);
-	ceph_decode_64_le(&p, fsid.minor);
+	__ceph_fsid_set_major(&fsid, major);
-	if (!ceph_fsid_equal(&fsid, &mdsc->client->monc.monmap->fsid)) {
+	ceph_decode_64_le(&p, minor);
+	__ceph_fsid_set_minor(&fsid, minor);
+	if (ceph_fsid_compare(&fsid, &mdsc->client->monc.monmap->fsid)) {
 		derr(0, "got mdsmap with wrong fsid\n");
 		return;
 	}

src/kernel/mon_client.c

@@ -19,6 +19,7 @@ struct ceph_monmap *ceph_monmap_decode(void *p, void *end)
 {
 	struct ceph_monmap *m;
 	int i, err = -EINVAL;
+	__le64 major, minor;
 
 	dout(30, "monmap_decode %p %p len %d\n", p, end, (int)(end-p));
 
@@ -28,8 +29,10 @@ struct ceph_monmap *ceph_monmap_decode(void *p, void *end)
 		return ERR_PTR(-ENOMEM);
 
 	ceph_decode_need(&p, end, 2*sizeof(u32) + 2*sizeof(u64), bad);
-	ceph_decode_64_le(&p, m->fsid.major);
+	ceph_decode_64_le(&p, major);
-	ceph_decode_64_le(&p, m->fsid.minor);
+	__ceph_fsid_set_major(&m->fsid, major);
+	ceph_decode_64_le(&p, minor);
+	__ceph_fsid_set_minor(&m->fsid, minor);
 	ceph_decode_32(&p, m->epoch);
 	ceph_decode_32(&p, m->num_mon);
 	ceph_decode_need(&p, end, m->num_mon*sizeof(m->mon_inst[0]), bad);

src/kernel/mon_client.h

@@ -29,7 +29,7 @@ struct ceph_mount_args;
  * ceph_monmap_decode().
  */
 struct ceph_monmap {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	u32 epoch;
 	u32 num_mon;
 	struct ceph_entity_inst mon_inst[0];

src/kernel/osd_client.c

@@ -442,7 +442,8 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
 	u32 epoch;
 	struct ceph_osdmap *newmap = NULL, *oldmap;
 	int err;
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
+	__le64 major, minor;
 
 	dout(2, "handle_map have %u\n", osdc->osdmap ? osdc->osdmap->epoch : 0);
 	p = msg->front.iov_base;
@@ -450,9 +451,11 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
 
 	/* verify fsid */
 	ceph_decode_need(&p, end, sizeof(fsid), bad);
-	ceph_decode_64_le(&p, fsid.major);
+	ceph_decode_64_le(&p, major);
-	ceph_decode_64_le(&p, fsid.minor);
+	__ceph_fsid_set_major(&fsid, major);
-	if (!ceph_fsid_equal(&fsid, &osdc->client->monc.monmap->fsid)) {
+	ceph_decode_64_le(&p, minor);
+	__ceph_fsid_set_minor(&fsid, minor);
+	if (ceph_fsid_compare(&fsid, &osdc->client->monc.monmap->fsid)) {
 		derr(0, "got map with wrong fsid, ignoring\n");
 		return;
 	}

src/kernel/osdmap.c

@@ -341,6 +341,7 @@ struct ceph_osdmap *osdmap_decode(void **p, void *end)
 	u32 len, max, i;
 	int err = -EINVAL;
 	void *start = *p;
+	__le64 major, minor;
 
 	dout(30, "osdmap_decode %p to %p len %d\n", *p, end, (int)(end - *p));
 
@@ -349,8 +350,10 @@ struct ceph_osdmap *osdmap_decode(void **p, void *end)
 		return ERR_PTR(-ENOMEM);
 
 	ceph_decode_need(p, end, 2*sizeof(u64)+11*sizeof(u32), bad);
-	ceph_decode_64_le(p, map->fsid.major);
+	ceph_decode_64_le(p, major);
-	ceph_decode_64_le(p, map->fsid.minor);
+	__ceph_fsid_set_major(&map->fsid, major);
+	ceph_decode_64_le(p, minor);
+	__ceph_fsid_set_minor(&map->fsid, minor);
 	ceph_decode_32(p, map->epoch);
 	ceph_decode_32_le(p, map->ctime.tv_sec);
 	ceph_decode_32_le(p, map->ctime.tv_nsec);
@@ -422,18 +425,21 @@ struct ceph_osdmap *apply_incremental(void **p, void *end,
 {
 	struct ceph_osdmap *newmap = map;
 	struct crush_map *newcrush = NULL;
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	u32 epoch = 0;
 	struct ceph_timespec ctime;
 	u32 len, x;
 	__s32 new_flags, max;
 	void *start = *p;
 	int err = -EINVAL;
+	__le64 major, minor;
 
 	ceph_decode_need(p, end, sizeof(fsid)+sizeof(ctime)+2*sizeof(u32),
 			 bad);
-	ceph_decode_64_le(p, fsid.major);
+	ceph_decode_64_le(p, major);
-	ceph_decode_64_le(p, fsid.minor);
+	__ceph_fsid_set_major(&fsid, major);
+	ceph_decode_64_le(p, minor);
+	__ceph_fsid_set_minor(&fsid, minor);
 	ceph_decode_32(p, epoch);
 	BUG_ON(epoch != map->epoch+1);
 	ceph_decode_32_le(p, ctime.tv_sec);

src/kernel/osdmap.h

@@ -2,6 +2,7 @@
 #define _FS_CEPH_OSDMAP_H
 
 #include "types.h"
+#include "ceph_fs.h"
 #include "crush/crush.h"
 
 /*
@@ -17,7 +18,7 @@
  * the change between two successive epochs, or as a fully encoded map.
  */
 struct ceph_osdmap {
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	u32 epoch;
 	u32 mkfs_epoch;
 	struct ceph_timespec ctime, mtime;

src/kernel/super.c

@@ -105,7 +105,7 @@ static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
 	buf->f_frsize = PAGE_CACHE_SIZE;
 
 	/* leave in little-endian, regardless of host endianness */
-	fsid = monmap->fsid.major ^ monmap->fsid.minor;
+	fsid = __ceph_fsid_major(&monmap->fsid) ^ __ceph_fsid_minor(&monmap->fsid);
 	buf->f_fsid.val[0] = le64_to_cpu(fsid) & 0xffffffff;
 	buf->f_fsid.val[1] = le64_to_cpu(fsid) >> 32;
 
@@ -134,7 +134,7 @@ static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt)
 		seq_printf(m, ",debug=%d", ceph_debug);
 	if (args->flags & CEPH_MOUNT_FSID)
 		seq_printf(m, ",fsidmajor=%llu,fsidminor%llu",
-			   args->fsid.major, args->fsid.minor);
+			   __ceph_fsid_major(&args->fsid), __ceph_fsid_minor(&args->fsid));
 	if (args->flags & CEPH_MOUNT_NOSHARE)
 		seq_puts(m, ",noshare");
 	if (args->flags & CEPH_MOUNT_UNSAFE_WRITEBACK)
@@ -252,8 +252,8 @@ static void handle_monmap(struct ceph_client *client, struct ceph_msg *msg)
 		client->msgr->inst.name = msg->hdr.dst.name;
 		sprintf(name, "client%d", client->whoami);
 		dout(1, "i am %s, fsid is %llx.%llx\n", name,
-		     le64_to_cpu(client->monc.monmap->fsid.major),
+		     le64_to_cpu(__ceph_fsid_major(&client->monc.monmap->fsid)),
-		     le64_to_cpu(client->monc.monmap->fsid.minor));
+		     le64_to_cpu(__ceph_fsid_minor(&client->monc.monmap->fsid)));
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 25)
 		client->client_kobj = kobject_create_and_add(name, ceph_kobj);
@@ -523,10 +523,10 @@ static int parse_mount_args(int flags, char *options, const char *dev_name,
 		}
 		switch (token) {
 		case Opt_fsidmajor:
-			args->fsid.major = cpu_to_le64(intval);
+			__ceph_fsid_set_major(&args->fsid, cpu_to_le64(intval));
 			break;
 		case Opt_fsidminor:
-			args->fsid.minor = cpu_to_le64(intval);
+			__ceph_fsid_set_minor(&args->fsid, cpu_to_le64(intval));
 			break;
 		case Opt_port:
 			args->my_addr.ipaddr.sin_port = htons(intval);
@@ -953,7 +953,7 @@ static int ceph_compare_super(struct super_block *sb, void *data)
 
 	/* either compare fsid, or specified mon_hostname */
 	if (args->flags & CEPH_MOUNT_FSID) {
-		if (!ceph_fsid_equal(&args->fsid, &other->fsid)) {
+		if (ceph_fsid_compare(&args->fsid, &other->fsid)) {
 			dout(30, "fsid doesn't match\n");
 			return 0;
 		}

src/kernel/super.h

@@ -16,6 +16,7 @@
 #include "mon_client.h"
 #include "mds_client.h"
 #include "osd_client.h"
+#include "ceph_fs.h"
 
 /* f_type in struct statfs */
 #define CEPH_SUPER_MAGIC 0x00c36400
@@ -57,7 +58,7 @@ struct ceph_mount_args {
 	int sb_flags;
 	int flags;
 	int mount_timeout;
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 	struct ceph_entity_addr my_addr;
 	int num_mon;
 	struct ceph_entity_addr mon_addr[MAX_MON_MOUNT_ADDR];
@@ -89,7 +90,7 @@ struct ceph_client {
 
 	struct mutex mount_mutex;       /* serialize mount attempts */
 	struct ceph_mount_args mount_args;
-	struct ceph_fsid fsid;
+	ceph_fsid_t fsid;
 
 	struct super_block *sb;
 
@@ -573,6 +574,26 @@ static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 /* super.c */
 extern const char *ceph_msg_type_name(int type);
 
+static inline __le64 __ceph_fsid_minor(ceph_fsid_t *fsid)
+{
+	return *(__le64 *)&fsid->fsid[8];
+}
+
+static inline __le64 __ceph_fsid_major(ceph_fsid_t *fsid)
+{
+	return *(__le64 *)&fsid->fsid[0];
+}
+
+static inline void __ceph_fsid_set_minor(ceph_fsid_t *fsid, __le64 val)
+{
+	*(__le64 *)&fsid->fsid[8] = val;
+}
+
+static inline void __ceph_fsid_set_major(ceph_fsid_t *fsid, __le64 val)
+{
+	*(__le64 *)&fsid->fsid[0] = val;
+}
+
 /* inode.c */
 extern const struct inode_operations ceph_file_iops;
 extern struct kmem_cache *ceph_inode_cachep;

src/messages/MLog.h

@@ -19,14 +19,14 @@
 
 class MLog : public Message {
 public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   deque<LogEntry> entries;
   version_t last;
   
   MLog() : Message(MSG_PGSTATS) {}
-  MLog(ceph_fsid& f, deque<LogEntry>& e) : 
+  MLog(ceph_fsid_t& f, deque<LogEntry>& e) : 
     Message(MSG_LOG), fsid(f), entries(e), last(0) { }
-  MLog(ceph_fsid& f, version_t l) : 
+  MLog(ceph_fsid_t& f, version_t l) : 
     Message(MSG_LOG), fsid(f), last(l) {}
 
   const char *get_type_name() { return "log"; }

src/messages/MMDSBeacon.h

@@ -22,7 +22,7 @@
 #include "mds/MDSMap.h"
 
 class MMDSBeacon : public Message {
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t last_epoch_seen;  // include last mdsmap epoch mds has seen to avoid race with monitor decree
   __u32 state;
   version_t seq;
@@ -30,11 +30,11 @@ class MMDSBeacon : public Message {
 
  public:
   MMDSBeacon() : Message(MSG_MDS_BEACON) {}
-  MMDSBeacon(ceph_fsid &f, epoch_t les, int st, version_t se, int wr) : 
+  MMDSBeacon(ceph_fsid_t &f, epoch_t les, int st, version_t se, int wr) : 
     Message(MSG_MDS_BEACON), 
     fsid(f), last_epoch_seen(les), state(st), seq(se), want_rank(wr) { }
 
-  ceph_fsid& get_fsid() { return fsid; }
+  ceph_fsid_t& get_fsid() { return fsid; }
   epoch_t get_last_epoch_seen() { return last_epoch_seen; }
   int get_state() { return state; }
   version_t get_seq() { return seq; }

src/messages/MMDSGetMap.h

@@ -21,11 +21,11 @@
 
 class MMDSGetMap : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t want;
 
   MMDSGetMap() {}
-  MMDSGetMap(ceph_fsid &f, epoch_t w=0) : 
+  MMDSGetMap(ceph_fsid_t &f, epoch_t w=0) : 
     Message(CEPH_MSG_MDS_GETMAP), 
     fsid(f),
     want(w) { }

src/messages/MMDSMap.h

@@ -45,7 +45,7 @@ class MMDSMap : public Message {
   }
   */
   
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t epoch;
   bufferlist encoded;
 
@@ -54,7 +54,7 @@ class MMDSMap : public Message {
 
   MMDSMap() : 
     Message(CEPH_MSG_MDS_MAP) {}
-  MMDSMap(ceph_fsid &f, MDSMap *mm) :
+  MMDSMap(ceph_fsid_t &f, MDSMap *mm) :
     Message(CEPH_MSG_MDS_MAP),
     fsid(f) {
     epoch = mm->get_epoch();

src/messages/MMonCommand.h

@@ -22,11 +22,11 @@ using std::vector;
 
 class MMonCommand : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   vector<string> cmd;
 
   MMonCommand() : Message(MSG_MON_COMMAND) {}
-  MMonCommand(ceph_fsid &f) : 
+  MMonCommand(ceph_fsid_t &f) : 
     Message(MSG_MON_COMMAND),
     fsid(f) { }
   

src/messages/MMonObserve.h

@@ -22,12 +22,12 @@ using std::vector;
 
 class MMonObserve : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   uint32_t machine_id;
   version_t ver;
 
   MMonObserve() : Message(MSG_MON_OBSERVE) {}
-  MMonObserve(ceph_fsid &f, int mid, version_t v) : 
+  MMonObserve(ceph_fsid_t &f, int mid, version_t v) : 
     Message(MSG_MON_OBSERVE),
     fsid(f), machine_id(mid), ver(v) { }
   

src/messages/MOSDFailure.h

@@ -21,12 +21,12 @@
 
 class MOSDFailure : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   entity_inst_t failed;
   epoch_t       epoch;
 
   MOSDFailure() : Message(MSG_OSD_FAILURE) {}
-  MOSDFailure(ceph_fsid &fs, entity_inst_t f, epoch_t e) : 
+  MOSDFailure(ceph_fsid_t &fs, entity_inst_t f, epoch_t e) : 
     Message(MSG_OSD_FAILURE),
     fsid(fs), failed(f), epoch(e) {}
  

src/messages/MOSDGetMap.h

@@ -21,11 +21,11 @@
 
 class MOSDGetMap : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t start;  // this is the first incremental the sender wants (he has start-1)
 
   MOSDGetMap() : Message(CEPH_MSG_OSD_GETMAP) {}
-  MOSDGetMap(ceph_fsid& f, epoch_t s=0) : 
+  MOSDGetMap(ceph_fsid_t& f, epoch_t s=0) : 
     Message(CEPH_MSG_OSD_GETMAP),
     fsid(f), start(s) { }
 

src/messages/MOSDMap.h

@@ -22,7 +22,7 @@
 
 class MOSDMap : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   map<epoch_t, bufferlist> maps;
   map<epoch_t, bufferlist> incremental_maps;
 
@@ -47,7 +47,7 @@ class MOSDMap : public Message {
 
 
   MOSDMap() : Message(CEPH_MSG_OSD_MAP) { }
-  MOSDMap(ceph_fsid &f, OSDMap *oc=0) : Message(CEPH_MSG_OSD_MAP),
+  MOSDMap(ceph_fsid_t &f, OSDMap *oc=0) : Message(CEPH_MSG_OSD_MAP),
 				      fsid(f) {
     if (oc)
       oc->encode(maps[oc->get_epoch()]);

src/messages/MOSDPing.h

@@ -23,12 +23,12 @@
 
 class MOSDPing : public Message {
  public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t map_epoch, peer_as_of_epoch;
   bool ack;
   osd_peer_stat_t peer_stat;
 
-  MOSDPing(ceph_fsid& f, epoch_t e, epoch_t pe, osd_peer_stat_t& ps, bool a=false) : 
+  MOSDPing(ceph_fsid_t& f, epoch_t e, epoch_t pe, osd_peer_stat_t& ps, bool a=false) : 
     Message(MSG_OSD_PING), fsid(f), map_epoch(e), peer_as_of_epoch(pe), ack(a), peer_stat(ps) { }
   MOSDPing() {}
 

src/messages/MOSDScrub.h

@@ -23,15 +23,15 @@
  */
 
 struct MOSDScrub : public Message {
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   vector<pg_t> scrub_pgs;
   bool repair;
 
   MOSDScrub() {}
-  MOSDScrub(ceph_fsid& f) :
+  MOSDScrub(ceph_fsid_t& f) :
     Message(MSG_OSD_SCRUB),
     fsid(f), repair(false) {}
-  MOSDScrub(ceph_fsid& f, vector<pg_t>& pgs, bool r) :
+  MOSDScrub(ceph_fsid_t& f, vector<pg_t>& pgs, bool r) :
     Message(MSG_OSD_SCRUB),
     fsid(f), scrub_pgs(pgs), repair(r) {}
 

src/messages/MPGStats.h

@@ -19,14 +19,14 @@
 
 class MPGStats : public Message {
 public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   map<pg_t,pg_stat_t> pg_stat;
   osd_stat_t osd_stat;
   epoch_t epoch;
   utime_t had_map_for;
   
   MPGStats() : Message(MSG_PGSTATS) {}
-  MPGStats(ceph_fsid& f, epoch_t e, utime_t had) : 
+  MPGStats(ceph_fsid_t& f, epoch_t e, utime_t had) : 
     Message(MSG_PGSTATS), fsid(f), epoch(e), had_map_for(had) {}
 
   const char *get_type_name() { return "pg_stats"; }

src/messages/MStatfs.h

@@ -20,7 +20,7 @@
 
 class MStatfs : public Message {
 public:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   tid_t tid;
 
   MStatfs() : Message(CEPH_MSG_STATFS) {}

src/messages/MStatfsReply.h

@@ -21,7 +21,7 @@ public:
   struct ceph_mon_statfs_reply h;
 
   MStatfsReply() : Message(CEPH_MSG_STATFS_REPLY) {}
-  MStatfsReply(ceph_fsid &f, tid_t t) : Message(CEPH_MSG_STATFS_REPLY) {
+  MStatfsReply(ceph_fsid_t &f, tid_t t) : Message(CEPH_MSG_STATFS_REPLY) {
     h.fsid = f;
     h.tid = t;
   }

src/mon/LogMonitor.cc

@@ -211,7 +211,7 @@ bool LogMonitor::prepare_log(MLog *m)
 {
   dout(10) << "prepare_log " << *m << " from " << m->get_orig_source() << dendl;
 
-  if (!ceph_fsid_equal(&m->fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &mon->monmap->fsid)) {
     dout(0) << "handle_log on fsid " << m->fsid << " != " << mon->monmap->fsid << dendl;
     delete m;
     return false;

src/mon/MDSMonitor.cc

@@ -159,7 +159,7 @@ bool MDSMonitor::preprocess_beacon(MMDSBeacon *m)
   version_t seq = m->get_seq();
   MDSMap::mds_info_t info;
 
-  if (!ceph_fsid_equal(&m->get_fsid(), &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&m->get_fsid(), &mon->monmap->fsid)) {
     dout(0) << "preprocess_beacon on fsid " << m->get_fsid() << " != " << mon->monmap->fsid << dendl;
     goto out;
   }

src/mon/MonMap.h

@@ -24,7 +24,7 @@
 class MonMap {
  public:
   epoch_t epoch;       // what epoch/version of the monmap
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   vector<entity_inst_t> mon_inst;
 
   int       last_mon;    // last mon i talked to
@@ -33,7 +33,7 @@ class MonMap {
     generate_fsid();
   }
 
-  ceph_fsid& get_fsid() { return fsid; }
+  ceph_fsid_t& get_fsid() { return fsid; }
 
   unsigned size() {
     return mon_inst.size();
@@ -99,8 +99,8 @@ class MonMap {
 
 
   void generate_fsid() {
-    fsid.major = ((uint64_t)rand() << 32) + rand();
+    for (int i=0; i<16; i++)
-    fsid.minor = ((uint64_t)rand() << 32) + rand();
+      fsid.fsid[i] = rand();
   }
 
   // read from/write to a file

src/mon/Monitor.cc

@@ -198,7 +198,7 @@ void Monitor::lose_election(epoch_t epoch, set<int> &q, int l)
 
 void Monitor::handle_command(MMonCommand *m)
 {
-  if (!ceph_fsid_equal(&m->fsid, &monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &monmap->fsid)) {
     dout(0) << "handle_command on fsid " << m->fsid << " != " << monmap->fsid << dendl;
     reply_command(m, -EPERM, "wrong fsid");
     return;

src/mon/OSDMonitor.cc

@@ -340,7 +340,7 @@ void OSDMonitor::handle_osd_getmap(MOSDGetMap *m)
 	  << " start " << m->get_start_epoch()
 	  << dendl;
   
-  if (!ceph_fsid_equal(&m->fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &mon->monmap->fsid)) {
     dout(0) << "handle_osd_getmap on fsid " << m->fsid << " != " << mon->monmap->fsid << dendl;
     goto out;
   }
@@ -369,7 +369,7 @@ bool OSDMonitor::preprocess_failure(MOSDFailure *m)
   // who is failed
   int badboy = m->get_failed().name.num();
 
-  if (!ceph_fsid_equal(&m->fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &mon->monmap->fsid)) {
     dout(0) << "preprocess_failure on fsid " << m->fsid << " != " << mon->monmap->fsid << dendl;
     goto didit;
   }
@@ -457,7 +457,7 @@ void OSDMonitor::_reported_failure(MOSDFailure *m)
 
 bool OSDMonitor::preprocess_boot(MOSDBoot *m)
 {
-  if (!ceph_fsid_equal(&m->sb.fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&m->sb.fsid, &mon->monmap->fsid)) {
     dout(0) << "preprocess_boot on fsid " << m->sb.fsid << " != " << mon->monmap->fsid << dendl;
     delete m;
     return true;
@@ -975,7 +975,7 @@ bool OSDMonitor::prepare_command(MMonCommand *m)
       OSDMap map;
       map.decode(m->get_data());
       epoch_t e = atoi(m->cmd[2].c_str());
-      if (ceph_fsid_equal(&map.fsid, &mon->monmap->fsid)) {
+      if (ceph_fsid_compare(&map.fsid, &mon->monmap->fsid) == 0) {
 	if (pending_inc.epoch == e) {
 	  map.epoch = pending_inc.epoch;  // make sure epoch is correct
 	  map.encode(pending_inc.fullmap);

src/mon/PGMonitor.cc

@@ -203,7 +203,7 @@ void PGMonitor::handle_statfs(MStatfs *statfs)
 
   dout(10) << "handle_statfs " << *statfs << " from " << statfs->get_orig_source() << dendl;
 
-  if (!ceph_fsid_equal(&statfs->fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&statfs->fsid, &mon->monmap->fsid)) {
     dout(0) << "handle_statfs on fsid " << statfs->fsid << " != " << mon->monmap->fsid << dendl;
     goto out;
   }
@@ -261,7 +261,7 @@ bool PGMonitor::prepare_pg_stats(MPGStats *stats)
   dout(10) << "prepare_pg_stats " << *stats << " from " << stats->get_orig_source() << dendl;
   int from = stats->get_orig_source().num();
 
-  if (!ceph_fsid_equal(&stats->fsid, &mon->monmap->fsid)) {
+  if (ceph_fsid_compare(&stats->fsid, &mon->monmap->fsid)) {
     dout(0) << "handle_statfs on fsid " << stats->fsid << " != " << mon->monmap->fsid << dendl;
     delete stats;
     return false;

src/os/FileStore.cc

@@ -694,6 +694,10 @@ unsigned FileStore::_apply_transaction(Transaction& t)
       _collection_rmattr(t.get_cid(), t.get_attrname());
       break;
 
+    case Transaction::OP_STARTSYNC:
+      _start_sync();
+      break;
+
     default:
       cerr << "bad op " << op << std::endl;
       assert(0);
@@ -1431,6 +1435,12 @@ void FileStore::sync_entry()
   lock.Unlock();
 }
 
+void FileStore::_start_sync()
+{
+  dout(10) << "start_sync" << dendl;
+  sync_cond.Signal();
+}
+
 void FileStore::sync()
 {
   Mutex::Locker l(lock);

src/os/FileStore.h

@@ -115,6 +115,8 @@ class FileStore : public JournalingObjectStore {
   int _do_clone_range(int from, int to, __u64 off, __u64 len);
   int _remove(coll_t cid, pobject_t oid);
 
+  void _start_sync();
+
   void sync();
   void sync(Context *onsafe);
 

src/os/ObjectStore.h

@@ -98,6 +98,9 @@ public:
     static const int OP_COLL_RMATTR =  25;  // cid, attrname
     static const int OP_COLL_SETATTRS = 26;  // cid, attrset
 
+    static const int OP_STARTSYNC =    27;  // start a sync 
+
+
   private:
     /*
     int len;
@@ -158,6 +161,10 @@ public:
       return attrsets[attrsetp++];
     }
 
+    void start_sync() {
+      int op = OP_STARTSYNC;
+      ops.push_back(op);
+    }
     void touch(coll_t cid, pobject_t oid) {
       int op = OP_TOUCH;
       ops.push_back(op);

src/osd/OSD.cc

@@ -149,7 +149,7 @@ ObjectStore *OSD::create_object_store(const char *dev)
 }
 
 
-int OSD::mkfs(const char *dev, ceph_fsid fsid, int whoami)
+int OSD::mkfs(const char *dev, ceph_fsid_t fsid, int whoami)
 {
   ObjectStore *store = create_object_store(dev);
   if (!store)
@@ -217,7 +217,7 @@ int OSD::mkfs(const char *dev, ceph_fsid fsid, int whoami)
   return 0;
 }
 
-int OSD::peek_super(const char *dev, nstring& magic, ceph_fsid& fsid, int& whoami)
+int OSD::peek_super(const char *dev, nstring& magic, ceph_fsid_t& fsid, int& whoami)
 {
   ObjectStore *store = create_object_store(dev);
   if (!store)
@@ -561,7 +561,7 @@ int OSD::read_superblock()
 
   dout(10) << "read_superblock " << superblock << dendl;
 
-  if (!ceph_fsid_equal(&superblock.fsid, &monmap->fsid)) {
+  if (ceph_fsid_compare(&superblock.fsid, &monmap->fsid)) {
     derr(0) << "read_superblock fsid " << superblock.fsid << " != monmap " << monmap->fsid << dendl;
     return -1;
   }
@@ -982,7 +982,7 @@ void OSD::handle_osd_ping(MOSDPing *m)
     return;
   }
 
-  if (!ceph_fsid_equal(&superblock.fsid, &m->fsid)) {
+  if (ceph_fsid_compare(&superblock.fsid, &m->fsid)) {
     dout(20) << "handle_osd_ping from " << m->get_source()
 	     << " bad fsid " << m->fsid << " != " << superblock.fsid << dendl;
     delete m;
@@ -1633,7 +1633,7 @@ void OSD::handle_scrub(MOSDScrub *m)
 {
   dout(10) << "handle_scrub " << *m << dendl;
   
-  if (!ceph_fsid_equal(&m->fsid, &monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &monmap->fsid)) {
     dout(0) << "handle_scrub fsid " << m->fsid << " != " << monmap->fsid << dendl;
     delete m;
     return;
@@ -1724,7 +1724,7 @@ void OSD::note_up_osd(int osd)
 void OSD::handle_osd_map(MOSDMap *m)
 {
   assert(osd_lock.is_locked());
-  if (!ceph_fsid_equal(&m->fsid, &monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &monmap->fsid)) {
     dout(0) << "handle_osd_map fsid " << m->fsid << " != " << monmap->fsid << dendl;
     delete m;
     return;
@@ -2098,8 +2098,7 @@ void OSD::advance_map(ObjectStore::Transaction& t, interval_set<snapid_t>& remov
       pg->on_role_change();
 
       // interrupt backlog generation
-      pg->generate_backlog_epoch = 0;
+      cancel_generate_backlog(pg);
-      backlog_wq.dequeue(pg);
 
       // take active waiters
       take_waiters(pg->waiting_for_active);
@@ -3130,6 +3129,13 @@ void OSD::queue_generate_backlog(PG *pg)
   }
 }
 
+void OSD::cancel_generate_backlog(PG *pg)
+{
+  dout(10) << *pg << " cancel_generate_backlog" << dendl;
+  pg->generate_backlog_epoch = 0;
+  backlog_wq.dequeue(pg);
+}
+
 void OSD::generate_backlog(PG *pg)
 {
   pg->lock();

src/osd/OSD.h

@@ -539,6 +539,7 @@ private:
   } backlog_wq;
 
   void queue_generate_backlog(PG *pg);
+  void cancel_generate_backlog(PG *pg);
   void generate_backlog(PG *pg);
 
 
@@ -693,8 +694,8 @@ private:
   // static bits
   static int find_osd_dev(char *result, int whoami);
   static ObjectStore *create_object_store(const char *dev);
-  static int mkfs(const char *dev, ceph_fsid fsid, int whoami);
+  static int mkfs(const char *dev, ceph_fsid_t fsid, int whoami);
-  static int peek_super(const char *dev, nstring& magic, ceph_fsid& fsid, int& whoami);
+  static int peek_super(const char *dev, nstring& magic, ceph_fsid_t& fsid, int& whoami);
 
   // startup/shutdown
   int init();

src/osd/OSDMap.cc

@@ -74,7 +74,7 @@ void OSDMap::print_summary(ostream& out)
 }
 
 
-void OSDMap::build_simple(epoch_t e, ceph_fsid &fsid,
+void OSDMap::build_simple(epoch_t e, ceph_fsid_t &fsid,
 			  int num_osd, int num_dom, int pg_bits, int lpg_bits,
 			  int mds_local_osd)
 {

src/osd/OSDMap.h

@@ -132,7 +132,7 @@ class OSDMap {
 public:
   class Incremental {
   public:
-    ceph_fsid fsid;
+    ceph_fsid_t fsid;
     epoch_t epoch;   // new epoch; we are a diff from epoch-1 to epoch
     utime_t ctime;
     int32_t new_flags;
@@ -227,7 +227,7 @@ public:
 
     Incremental(epoch_t e=0) : epoch(e), new_flags(-1), new_max_osd(-1), 
 			       new_pg_num(0), new_pgp_num(0), new_lpg_num(0), new_lpgp_num(0) {
-      fsid.major = fsid.minor = 0;
+      memset(&fsid, 0, sizeof(fsid));
     }
     Incremental(bufferlist &bl) {
       bufferlist::iterator p = bl.begin();
@@ -239,7 +239,7 @@ public:
   };
   
 private:
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   epoch_t epoch;        // what epoch of the osd cluster descriptor is this
   utime_t ctime, mtime; // epoch start time
 
@@ -294,13 +294,13 @@ private:
 	     last_pg_change(0),
 	     flags(0),
 	     max_osd(0), max_snap(0) { 
-    fsid.major = fsid.minor = 0;
+    memset(&fsid, 0, sizeof(fsid));
     calc_pg_masks();
   }
 
   // map info
-  ceph_fsid& get_fsid() { return fsid; }
+  ceph_fsid_t& get_fsid() { return fsid; }
-  void set_fsid(ceph_fsid& f) { fsid = f; }
+  void set_fsid(ceph_fsid_t& f) { fsid = f; }
 
   epoch_t get_epoch() const { return epoch; }
   void inc_epoch() { epoch++; }
@@ -484,7 +484,7 @@ private:
     if (inc.epoch == 1)
       fsid = inc.fsid;
     else
-      assert(ceph_fsid_equal(&inc.fsid, &fsid));
+      assert(ceph_fsid_compare(&inc.fsid, &fsid) == 0);
     assert(inc.epoch == epoch+1);
     epoch++;
     ctime = inc.ctime;
@@ -899,7 +899,7 @@ private:
   /*
    * handy helpers to build simple maps...
    */
-  void build_simple(epoch_t e, ceph_fsid &fsid,
+  void build_simple(epoch_t e, ceph_fsid_t &fsid,
 		    int num_osd, int num_dom,
 		    int pg_bits, int lpg_bits,
 		    int mds_local_osd);

src/osd/PG.cc

@@ -341,7 +341,7 @@ void PG::merge_log(ObjectStore::Transaction& t,
     list<Log::Entry>::iterator p = log.log.end();
     while (p != log.log.begin()) {
       p--;
-      if (p->version <= log.top) {
+      if (p->version.version <= log.top.version) {
 	dout(10) << "merge_log split point is " << *p << dendl;
 
 	if (p->version < log.top && p->version < oldest_update) {
@@ -371,9 +371,8 @@ void PG::merge_log(ObjectStore::Transaction& t,
 	 p++) {
       Log::Entry &oe = *p;                      // old entry
       if (old_objects.count(oe.oid) &&
-	  old_objects[oe.oid] == &oe) {
+	  old_objects[oe.oid] == &oe)
 	merge_old_entry(t, oe);
-      }
     }
 
     info.last_update = log.top = olog.top;
@@ -1179,7 +1178,6 @@ void PG::peer(ObjectStore::Transaction& t,
     // let's pull info+logs from _everyone_ (strays included, this
     // time) in search of missing objects.
 
-    bool waiting = false;
     for (map<int,Info>::iterator it = peer_info.begin();
          it != peer_info.end();
          it++) {
@@ -1190,19 +1188,17 @@ void PG::peer(ObjectStore::Transaction& t,
 
       if (peer_summary_requested.count(peer)) {
         dout(10) << " already requested summary/backlog from osd" << peer << dendl;
-        waiting = true;
         continue;
       }
 
       dout(10) << " requesting summary/backlog from osd" << peer << dendl;      
       query_map[peer][info.pgid] = Query(Query::BACKLOG, info.history);
       peer_summary_requested.insert(peer);
-      waiting = true;
+      return;
     }
     
-    if (!waiting)
+    dout(10) << (missing.num_missing() - missing_loc.size())
-      dout(10) << (missing.num_missing() - missing_loc.size())
+	     << " objects are still lost, waiting+hoping for a notify from someone else!" << dendl;
-	       << " objects are still lost, waiting+hoping for a notify from someone else!" << dendl;
     return;
   }
 
@@ -1282,6 +1278,9 @@ void PG::activate(ObjectStore::Transaction& t,
   // clear prior set (and dependency info)... we are done peering!
   clear_prior();
 
+  // if we are building a backlog, cancel it!
+  osd->cancel_generate_backlog(this);
+
   // write pg info, log
   write_info(t);
   write_log(t);
@@ -1782,7 +1781,9 @@ void PG::read_log(ObjectStore *store)
       pobject_t poid(info.pgid.pool(), 0, i->oid);
       bufferlist bv;
       int r = osd->store->getattr(info.pgid.to_coll(), poid, "version", bv);
-      eversion_t v(bv);
+      eversion_t v;
+      if (r >= 0)
+	v = eversion_t(bv);
       if (r < 0 || v < i->version) {
 	dout(15) << "read_log  missing " << *i << dendl;
 	missing.add(i->oid, i->version, v);

src/osd/ReplicatedPG.cc

@@ -1131,6 +1131,9 @@ int ReplicatedPG::prepare_simple_op(ObjectStore::Transaction& t, osd_reqid_t req
     }
     break;
 
+  case CEPH_OSD_OP_STARTSYNC:
+    t.start_sync();
+    break;
 
   default:
     return -EINVAL;
@@ -2584,30 +2587,8 @@ void ReplicatedPG::on_change()
     }
   }
   
-  // remove strays from pushing map
+  // clear pushing map
-  {
+  pushing.clear();
-    map<object_t, set<int> >::iterator p = pushing.begin();
-    while (p != pushing.end()) {
-      set<int>::iterator q = p->second.begin();
-      while (q != p->second.end()) {
-	int o = *q++;
-	bool have = false;
-	for (unsigned i=1; i<acting.size(); i++)
-	  if (acting[i] == o) {
-	    have = true;
-	    break;
-	  }
-	if (!have) {
-	  dout(10) << " forgetting push of " << p->first << " to (now stray) osd" << o << dendl;
-	  p->second.erase(o);
-	}
-      }
-      if (p->second.empty())
-	pushing.erase(p++);
-      else
-	p++;	
-    }
-  }
 }
 
 void ReplicatedPG::on_role_change()

src/osd/osd_types.h

@@ -614,7 +614,7 @@ struct ObjectMutation {
 class OSDSuperblock {
 public:
   nstring magic;
-  ceph_fsid fsid;
+  ceph_fsid_t fsid;
   int32_t whoami;    // my role in this fs.
   epoch_t current_epoch;             // most recent epoch
   epoch_t oldest_map, newest_map;    // oldest/newest maps we have.

src/osdc/Objecter.cc

@@ -72,7 +72,7 @@ void Objecter::handle_osd_map(MOSDMap *m)
 {
   assert(osdmap); 
 
-  if (!ceph_fsid_equal(&m->fsid, &monmap->fsid)) {
+  if (ceph_fsid_compare(&m->fsid, &monmap->fsid)) {
     dout(0) << "handle_osd_map fsid " << m->fsid << " != " << monmap->fsid << dendl;
     delete m;
     return;

src/vstart.sh

@@ -91,7 +91,7 @@ fi
 
 
 # lockdep everywhere?
-export CEPH_ARGS="--lockdep 1"
+# export CEPH_ARGS="--lockdep 1"
 
 
 # sudo if btrfs