Merge tag 'ceph-for-4.13-rc1' of git://github.com/ceph/ceph-client

Pull ceph updates from Ilya Dryomov:
 "The main item here is support for v12.y.z ("Luminous") clusters:
  RESEND_ON_SPLIT, RADOS_BACKOFF, OSDMAP_PG_UPMAP and CRUSH_CHOOSE_ARGS
  feature bits, and various other changes in the RADOS client protocol.

  On top of that we have a new fsc mount option to allow supplying
  fscache uniquifier (similar to NFS) and the usual pile of filesystem
  fixes from Zheng"

* tag 'ceph-for-4.13-rc1' of git://github.com/ceph/ceph-client: (44 commits)
  libceph: advertise support for NEW_OSDOP_ENCODING and SERVER_LUMINOUS
  libceph: osd_state is 32 bits wide in luminous
  crush: remove an obsolete comment
  crush: crush_init_workspace starts with struct crush_work
  libceph, crush: per-pool crush_choose_arg_map for crush_do_rule()
  crush: implement weight and id overrides for straw2
  libceph: apply_upmap()
  libceph: compute actual pgid in ceph_pg_to_up_acting_osds()
  libceph: pg_upmap[_items] infrastructure
  libceph: ceph_decode_skip_* helpers
  libceph: kill __{insert,lookup,remove}_pg_mapping()
  libceph: introduce and switch to decode_pg_mapping()
  libceph: don't pass pgid by value
  libceph: respect RADOS_BACKOFF backoffs
  libceph: make DEFINE_RB_* helpers more general
  libceph: avoid unnecessary pi lookups in calc_target()
  libceph: use target pi for calc_target() calculations
  libceph: always populate t->target_{oid,oloc} in calc_target()
  libceph: make sure need_resend targets reflect latest map
  libceph: delete from need_resend_linger before check_linger_pool_dne()
  ...

10 files changed, 450 insertions, 118 deletions

diff --git a/include/linux/ceph/ceph_features.h b/include/linux/ceph/ceph_features.h
index fd8b2953c78f..f0f6c537b64c 100644
--- a/include/linux/ceph/ceph_features.h
+++ b/include/linux/ceph/ceph_features.h
@@ -2,103 +2,174 @@
 #define __CEPH_FEATURES
 
 /*
- * feature bits
+ * Each time we reclaim bits for reuse we need to specify another bit
+ * that, if present, indicates we have the new incarnation of that
+ * feature.  Base case is 1 (first use).
  */
-#define CEPH_FEATURE_UID            (1ULL<<0)
-#define CEPH_FEATURE_NOSRCADDR      (1ULL<<1)
-#define CEPH_FEATURE_MONCLOCKCHECK  (1ULL<<2)
-#define CEPH_FEATURE_FLOCK          (1ULL<<3)
-#define CEPH_FEATURE_SUBSCRIBE2     (1ULL<<4)
-#define CEPH_FEATURE_MONNAMES       (1ULL<<5)
-#define CEPH_FEATURE_RECONNECT_SEQ  (1ULL<<6)
-#define CEPH_FEATURE_DIRLAYOUTHASH  (1ULL<<7)
-#define CEPH_FEATURE_OBJECTLOCATOR  (1ULL<<8)
-#define CEPH_FEATURE_PGID64         (1ULL<<9)
-#define CEPH_FEATURE_INCSUBOSDMAP   (1ULL<<10)
-#define CEPH_FEATURE_PGPOOL3        (1ULL<<11)
-#define CEPH_FEATURE_OSDREPLYMUX    (1ULL<<12)
-#define CEPH_FEATURE_OSDENC         (1ULL<<13)
-#define CEPH_FEATURE_OMAP           (1ULL<<14)
-#define CEPH_FEATURE_MONENC         (1ULL<<15)
-#define CEPH_FEATURE_QUERY_T        (1ULL<<16)
-#define CEPH_FEATURE_INDEP_PG_MAP   (1ULL<<17)
-#define CEPH_FEATURE_CRUSH_TUNABLES (1ULL<<18)
-#define CEPH_FEATURE_CHUNKY_SCRUB   (1ULL<<19)
-#define CEPH_FEATURE_MON_NULLROUTE  (1ULL<<20)
-#define CEPH_FEATURE_MON_GV         (1ULL<<21)
-#define CEPH_FEATURE_BACKFILL_RESERVATION (1ULL<<22)
-#define CEPH_FEATURE_MSG_AUTH	    (1ULL<<23)
-#define CEPH_FEATURE_RECOVERY_RESERVATION (1ULL<<24)
-#define CEPH_FEATURE_CRUSH_TUNABLES2 (1ULL<<25)
-#define CEPH_FEATURE_CREATEPOOLID   (1ULL<<26)
-#define CEPH_FEATURE_REPLY_CREATE_INODE   (1ULL<<27)
-#define CEPH_FEATURE_OSD_HBMSGS     (1ULL<<28)
-#define CEPH_FEATURE_MDSENC         (1ULL<<29)
-#define CEPH_FEATURE_OSDHASHPSPOOL  (1ULL<<30)
-#define CEPH_FEATURE_MON_SINGLE_PAXOS (1ULL<<31)
-#define CEPH_FEATURE_OSD_SNAPMAPPER (1ULL<<32)
-#define CEPH_FEATURE_MON_SCRUB      (1ULL<<33)
-#define CEPH_FEATURE_OSD_PACKED_RECOVERY (1ULL<<34)
-#define CEPH_FEATURE_OSD_CACHEPOOL (1ULL<<35)
-#define CEPH_FEATURE_CRUSH_V2      (1ULL<<36)  /* new indep; SET_* steps */
-#define CEPH_FEATURE_EXPORT_PEER   (1ULL<<37)
-#define CEPH_FEATURE_OSD_ERASURE_CODES (1ULL<<38)
-#define CEPH_FEATURE_OSD_TMAP2OMAP (1ULL<<38)   /* overlap with EC */
-/* The process supports new-style OSDMap encoding. Monitors also use
-   this bit to determine if peers support NAK messages. */
-#define CEPH_FEATURE_OSDMAP_ENC    (1ULL<<39)
-#define CEPH_FEATURE_MDS_INLINE_DATA     (1ULL<<40)
-#define CEPH_FEATURE_CRUSH_TUNABLES3     (1ULL<<41)
-#define CEPH_FEATURE_OSD_PRIMARY_AFFINITY (1ULL<<41)  /* overlap w/ tunables3 */
-#define CEPH_FEATURE_MSGR_KEEPALIVE2   (1ULL<<42)
-#define CEPH_FEATURE_OSD_POOLRESEND    (1ULL<<43)
-#define CEPH_FEATURE_ERASURE_CODE_PLUGINS_V2 (1ULL<<44)
-#define CEPH_FEATURE_OSD_SET_ALLOC_HINT (1ULL<<45)
-#define CEPH_FEATURE_OSD_FADVISE_FLAGS (1ULL<<46)
-#define CEPH_FEATURE_OSD_REPOP         (1ULL<<46)   /* overlap with fadvise */
-#define CEPH_FEATURE_OSD_OBJECT_DIGEST  (1ULL<<46)  /* overlap with fadvise */
-#define CEPH_FEATURE_OSD_TRANSACTION_MAY_LAYOUT (1ULL<<46) /* overlap w/ fadvise */
-#define CEPH_FEATURE_MDS_QUOTA      (1ULL<<47)
-#define CEPH_FEATURE_CRUSH_V4      (1ULL<<48)  /* straw2 buckets */
-#define CEPH_FEATURE_OSD_MIN_SIZE_RECOVERY (1ULL<<49)
-// duplicated since it was introduced at the same time as MIN_SIZE_RECOVERY
-#define CEPH_FEATURE_OSD_PROXY_FEATURES (1ULL<<49)  /* overlap w/ above */
-#define CEPH_FEATURE_MON_METADATA (1ULL<<50)
-#define CEPH_FEATURE_OSD_BITWISE_HOBJ_SORT (1ULL<<51) /* can sort objs bitwise */
-#define CEPH_FEATURE_OSD_PROXY_WRITE_FEATURES (1ULL<<52)
-#define CEPH_FEATURE_ERASURE_CODE_PLUGINS_V3 (1ULL<<53)
-#define CEPH_FEATURE_OSD_HITSET_GMT (1ULL<<54)
-#define CEPH_FEATURE_HAMMER_0_94_4 (1ULL<<55)
-#define CEPH_FEATURE_NEW_OSDOP_ENCODING   (1ULL<<56) /* New, v7 encoding */
-#define CEPH_FEATURE_MON_STATEFUL_SUB (1ULL<<57) /* stateful mon subscription */
-#define CEPH_FEATURE_MON_ROUTE_OSDMAP (1ULL<<57) /* peon sends osdmaps */
-#define CEPH_FEATURE_CRUSH_TUNABLES5	(1ULL<<58) /* chooseleaf stable mode */
-// duplicated since it was introduced at the same time as CEPH_FEATURE_CRUSH_TUNABLES5
-#define CEPH_FEATURE_NEW_OSDOPREPLY_ENCODING   (1ULL<<58) /* New, v7 encoding */
-#define CEPH_FEATURE_FS_FILE_LAYOUT_V2       (1ULL<<58) /* file_layout_t */
+#define CEPH_FEATURE_INCARNATION_1 (0ull)
+#define CEPH_FEATURE_INCARNATION_2 (1ull<<57) // CEPH_FEATURE_SERVER_JEWEL
+
+#define DEFINE_CEPH_FEATURE(bit, incarnation, name)			\
+	const static uint64_t CEPH_FEATURE_##name = (1ULL<<bit);		\
+	const static uint64_t CEPH_FEATUREMASK_##name =			\
+		(1ULL<<bit | CEPH_FEATURE_INCARNATION_##incarnation);
+
+/* this bit is ignored but still advertised by release *when* */
+#define DEFINE_CEPH_FEATURE_DEPRECATED(bit, incarnation, name, when) \
+	const static uint64_t DEPRECATED_CEPH_FEATURE_##name = (1ULL<<bit); \
+	const static uint64_t DEPRECATED_CEPH_FEATUREMASK_##name =		\
+		(1ULL<<bit | CEPH_FEATURE_INCARNATION_##incarnation);
 
 /*
- * The introduction of CEPH_FEATURE_OSD_SNAPMAPPER caused the feature
- * vector to evaluate to 64 bit ~0.  To cope, we designate 1ULL << 63
- * to mean 33 bit ~0, and introduce a helper below to do the
- * translation.
+ * this bit is ignored by release *unused* and not advertised by
+ * release *unadvertised*
+ */
+#define DEFINE_CEPH_FEATURE_RETIRED(bit, inc, name, unused, unadvertised)
+
+
+/*
+ * test for a feature.  this test is safer than a typical mask against
+ * the bit because it ensures that we have the bit AND the marker for the
+ * bit's incarnation.  this must be used in any case where the features
+ * bits may include an old meaning of the bit.
+ */
+#define CEPH_HAVE_FEATURE(x, name)			\
+	(((x) & (CEPH_FEATUREMASK_##name)) == (CEPH_FEATUREMASK_##name))
+
+
+/*
+ * Notes on deprecation:
+ *
+ * A *major* release is a release through which all upgrades must pass
+ * (e.g., jewel).  For example, no pre-jewel server will ever talk to
+ * a post-jewel server (mon, osd, etc).
+ *
+ * For feature bits used *only* on the server-side:
+ *
+ *  - In the first phase we indicate that a feature is DEPRECATED as of
+ *    a particular release.  This is the first major release X (say,
+ *    jewel) that does not depend on its peers advertising the feature.
+ *    That is, it safely assumes its peers all have the feature.  We
+ *    indicate this with the DEPRECATED macro.  For example,
+ *
+ *      DEFINE_CEPH_FEATURE_DEPRECATED( 2, 1, MONCLOCKCHECK, JEWEL)
+ *
+ *    because 10.2.z (jewel) did not care if its peers advertised this
+ *    feature bit.
+ *
+ *  - In the second phase we stop advertising the the bit and call it
+ *    RETIRED.  This can normally be done in the *next* major release
+ *    following the one in which we marked the feature DEPRECATED.  In
+ *    the above example, for 12.0.z (luminous) we can say:
+ *
+ *      DEFINE_CEPH_FEATURE_RETIRED( 2, 1, MONCLOCKCHECK, JEWEL, LUMINOUS)
  *
- * This was introduced by ceph.git commit
- *   9ea02b84104045c2ffd7e7f4e7af512953855ecd v0.58-657-g9ea02b8
- * and fixed by ceph.git commit
- *   4255b5c2fb54ae40c53284b3ab700fdfc7e61748 v0.65-263-g4255b5c
+ *  - The bit can be reused in the first post-luminous release, 13.0.z
+ *    (m).
+ *
+ * This ensures that no two versions who have different meanings for
+ * the bit ever speak to each other.
  */
-#define CEPH_FEATURE_RESERVED (1ULL<<63)
-
-static inline u64 ceph_sanitize_features(u64 features)
-{
-	if (features & CEPH_FEATURE_RESERVED) {
-		/* everything through OSD_SNAPMAPPER */
-		return 0x1ffffffffull;
-	} else {
-		return features;
-	}
-}
+
+DEFINE_CEPH_FEATURE( 0, 1, UID)
+DEFINE_CEPH_FEATURE( 1, 1, NOSRCADDR)
+DEFINE_CEPH_FEATURE_RETIRED( 2, 1, MONCLOCKCHECK, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE( 3, 1, FLOCK)
+DEFINE_CEPH_FEATURE( 4, 1, SUBSCRIBE2)
+DEFINE_CEPH_FEATURE( 5, 1, MONNAMES)
+DEFINE_CEPH_FEATURE( 6, 1, RECONNECT_SEQ)
+DEFINE_CEPH_FEATURE( 7, 1, DIRLAYOUTHASH)
+DEFINE_CEPH_FEATURE( 8, 1, OBJECTLOCATOR)
+DEFINE_CEPH_FEATURE( 9, 1, PGID64)
+DEFINE_CEPH_FEATURE(10, 1, INCSUBOSDMAP)
+DEFINE_CEPH_FEATURE(11, 1, PGPOOL3)
+DEFINE_CEPH_FEATURE(12, 1, OSDREPLYMUX)
+DEFINE_CEPH_FEATURE(13, 1, OSDENC)
+DEFINE_CEPH_FEATURE_RETIRED(14, 1, OMAP, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(14, 2, SERVER_KRAKEN)
+DEFINE_CEPH_FEATURE(15, 1, MONENC)
+DEFINE_CEPH_FEATURE_RETIRED(16, 1, QUERY_T, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(17, 1, INDEP_PG_MAP, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(18, 1, CRUSH_TUNABLES)
+DEFINE_CEPH_FEATURE_RETIRED(19, 1, CHUNKY_SCRUB, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(20, 1, MON_NULLROUTE, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(21, 1, MON_GV, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(21, 2, SERVER_LUMINOUS)
+DEFINE_CEPH_FEATURE(21, 2, RESEND_ON_SPLIT)  // overlap
+DEFINE_CEPH_FEATURE(21, 2, RADOS_BACKOFF)    // overlap
+DEFINE_CEPH_FEATURE(21, 2, OSDMAP_PG_UPMAP)  // overlap
+DEFINE_CEPH_FEATURE(21, 2, CRUSH_CHOOSE_ARGS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(22, 1, BACKFILL_RESERVATION, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(23, 1, MSG_AUTH)
+DEFINE_CEPH_FEATURE_RETIRED(24, 1, RECOVERY_RESERVATION, JEWEL, LUNINOUS)
+
+DEFINE_CEPH_FEATURE(25, 1, CRUSH_TUNABLES2)
+DEFINE_CEPH_FEATURE(26, 1, CREATEPOOLID)
+DEFINE_CEPH_FEATURE(27, 1, REPLY_CREATE_INODE)
+DEFINE_CEPH_FEATURE_RETIRED(28, 1, OSD_HBMSGS, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(28, 2, SERVER_M)
+DEFINE_CEPH_FEATURE(29, 1, MDSENC)
+DEFINE_CEPH_FEATURE(30, 1, OSDHASHPSPOOL)
+DEFINE_CEPH_FEATURE(31, 1, MON_SINGLE_PAXOS)  // deprecate me
+DEFINE_CEPH_FEATURE_RETIRED(32, 1, OSD_SNAPMAPPER, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(33, 1, MON_SCRUB, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(34, 1, OSD_PACKED_RECOVERY, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(35, 1, OSD_CACHEPOOL)
+DEFINE_CEPH_FEATURE(36, 1, CRUSH_V2)
+DEFINE_CEPH_FEATURE(37, 1, EXPORT_PEER)
+DEFINE_CEPH_FEATURE(38, 1, OSD_ERASURE_CODES)
+DEFINE_CEPH_FEATURE(38, 1, OSD_OSD_TMAP2OMAP) // overlap
+DEFINE_CEPH_FEATURE(39, 1, OSDMAP_ENC)
+DEFINE_CEPH_FEATURE(40, 1, MDS_INLINE_DATA)
+DEFINE_CEPH_FEATURE(41, 1, CRUSH_TUNABLES3)
+DEFINE_CEPH_FEATURE(41, 1, OSD_PRIMARY_AFFINITY) // overlap
+DEFINE_CEPH_FEATURE(42, 1, MSGR_KEEPALIVE2)
+DEFINE_CEPH_FEATURE(43, 1, OSD_POOLRESEND)
+DEFINE_CEPH_FEATURE(44, 1, ERASURE_CODE_PLUGINS_V2)
+DEFINE_CEPH_FEATURE_RETIRED(45, 1, OSD_SET_ALLOC_HINT, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(46, 1, OSD_FADVISE_FLAGS)
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_REPOP, JEWEL, LUMINOUS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_OBJECT_DIGEST, JEWEL, LUMINOUS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_TRANSACTION_MAY_LAYOUT, JEWEL, LUMINOUS) // overlap
+
+DEFINE_CEPH_FEATURE(47, 1, MDS_QUOTA)
+DEFINE_CEPH_FEATURE(48, 1, CRUSH_V4)
+DEFINE_CEPH_FEATURE_RETIRED(49, 1, OSD_MIN_SIZE_RECOVERY, JEWEL, LUMINOUS)
+DEFINE_CEPH_FEATURE_RETIRED(49, 1, OSD_PROXY_FEATURES, JEWEL, LUMINOUS) // overlap
+
+DEFINE_CEPH_FEATURE(50, 1, MON_METADATA)
+DEFINE_CEPH_FEATURE(51, 1, OSD_BITWISE_HOBJ_SORT)
+DEFINE_CEPH_FEATURE(52, 1, OSD_PROXY_WRITE_FEATURES)
+DEFINE_CEPH_FEATURE(53, 1, ERASURE_CODE_PLUGINS_V3)
+DEFINE_CEPH_FEATURE(54, 1, OSD_HITSET_GMT)
+DEFINE_CEPH_FEATURE(55, 1, HAMMER_0_94_4)
+DEFINE_CEPH_FEATURE(56, 1, NEW_OSDOP_ENCODING)
+DEFINE_CEPH_FEATURE(57, 1, MON_STATEFUL_SUB)
+DEFINE_CEPH_FEATURE(57, 1, MON_ROUTE_OSDMAP) // overlap
+DEFINE_CEPH_FEATURE(57, 1, OSDSUBOP_NO_SNAPCONTEXT) // overlap
+DEFINE_CEPH_FEATURE(57, 1, SERVER_JEWEL) // overlap
+DEFINE_CEPH_FEATURE(58, 1, CRUSH_TUNABLES5)
+DEFINE_CEPH_FEATURE(58, 1, NEW_OSDOPREPLY_ENCODING) // overlap
+DEFINE_CEPH_FEATURE(58, 1, FS_FILE_LAYOUT_V2) // overlap
+DEFINE_CEPH_FEATURE(59, 1, FS_BTIME)
+DEFINE_CEPH_FEATURE(59, 1, FS_CHANGE_ATTR) // overlap
+DEFINE_CEPH_FEATURE(59, 1, MSG_ADDR2) // overlap
+DEFINE_CEPH_FEATURE(60, 1, BLKIN_TRACING)  // *do not share this bit*
+
+DEFINE_CEPH_FEATURE(61, 1, RESERVED2)          // unused, but slow down!
+DEFINE_CEPH_FEATURE(62, 1, RESERVED)           // do not use; used as a sentinal
+DEFINE_CEPH_FEATURE_DEPRECATED(63, 1, RESERVED_BROKEN, LUMINOUS) // client-facing
+
 
 /*
  * Features supported.
@@ -113,6 +184,11 @@ static inline u64 ceph_sanitize_features(u64 features)
 	 CEPH_FEATURE_PGPOOL3 |			\
 	 CEPH_FEATURE_OSDENC |			\
 	 CEPH_FEATURE_CRUSH_TUNABLES |		\
+	 CEPH_FEATURE_SERVER_LUMINOUS |		\
+	 CEPH_FEATURE_RESEND_ON_SPLIT |		\
+	 CEPH_FEATURE_RADOS_BACKOFF |		\
+	 CEPH_FEATURE_OSDMAP_PG_UPMAP |		\
+	 CEPH_FEATURE_CRUSH_CHOOSE_ARGS |	\
 	 CEPH_FEATURE_MSG_AUTH |		\
 	 CEPH_FEATURE_CRUSH_TUNABLES2 |		\
 	 CEPH_FEATURE_REPLY_CREATE_INODE |	\
@@ -126,7 +202,11 @@ static inline u64 ceph_sanitize_features(u64 features)
 	 CEPH_FEATURE_CRUSH_TUNABLES3 |		\
 	 CEPH_FEATURE_OSD_PRIMARY_AFFINITY |	\
 	 CEPH_FEATURE_MSGR_KEEPALIVE2 |		\
+	 CEPH_FEATURE_OSD_POOLRESEND |		\
 	 CEPH_FEATURE_CRUSH_V4 |		\
+	 CEPH_FEATURE_NEW_OSDOP_ENCODING |	\
+	 CEPH_FEATURE_SERVER_JEWEL |		\
+	 CEPH_FEATURE_MON_STATEFUL_SUB |	\
 	 CEPH_FEATURE_CRUSH_TUNABLES5 |		\
 	 CEPH_FEATURE_NEW_OSDOPREPLY_ENCODING)
 
diff --git a/include/linux/ceph/ceph_fs.h b/include/linux/ceph/ceph_fs.h
index ad078ebe25d6..edf5b04b918a 100644
--- a/include/linux/ceph/ceph_fs.h
+++ b/include/linux/ceph/ceph_fs.h
@@ -147,6 +147,7 @@ struct ceph_dir_layout {
 #define CEPH_MSG_OSD_OP                 42
 #define CEPH_MSG_OSD_OPREPLY            43
 #define CEPH_MSG_WATCH_NOTIFY           44
+#define CEPH_MSG_OSD_BACKOFF            61
 
 
 /* watch-notify operations */
diff --git a/include/linux/ceph/decode.h b/include/linux/ceph/decode.h
index f990f2cc907a..14af9b70d301 100644
--- a/include/linux/ceph/decode.h
+++ b/include/linux/ceph/decode.h
@@ -133,6 +133,66 @@ bad:
 }
 
 /*
+ * skip helpers
+ */
+#define ceph_decode_skip_n(p, end, n, bad)			\
+	do {							\
+		ceph_decode_need(p, end, n, bad);		\
+                *p += n;					\
+	} while (0)
+
+#define ceph_decode_skip_64(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u64), bad)
+
+#define ceph_decode_skip_32(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u32), bad)
+
+#define ceph_decode_skip_16(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u16), bad)
+
+#define ceph_decode_skip_8(p, end, bad)				\
+ceph_decode_skip_n(p, end, sizeof(u8), bad)
+
+#define ceph_decode_skip_string(p, end, bad)			\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		ceph_decode_skip_n(p, end, len, bad);		\
+	} while (0)
+
+#define ceph_decode_skip_set(p, end, type, bad)			\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--)					\
+			ceph_decode_skip_##type(p, end, bad);	\
+	} while (0)
+
+#define ceph_decode_skip_map(p, end, ktype, vtype, bad)		\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--) {					\
+			ceph_decode_skip_##ktype(p, end, bad);	\
+			ceph_decode_skip_##vtype(p, end, bad);	\
+		}						\
+	} while (0)
+
+#define ceph_decode_skip_map_of_map(p, end, ktype1, ktype2, vtype2, bad) \
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--) {					\
+			ceph_decode_skip_##ktype1(p, end, bad);	\
+			ceph_decode_skip_map(p, end, ktype2, vtype2, bad); \
+		}						\
+	} while (0)
+
+/*
  * struct ceph_timespec <-> struct timespec
  */
 static inline void ceph_decode_timespec(struct timespec *ts,
diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h
index 3229ae6c7846..8a79587e1317 100644
--- a/include/linux/ceph/libceph.h
+++ b/include/linux/ceph/libceph.h
@@ -184,10 +184,11 @@ static inline int calc_pages_for(u64 off, u64 len)
 		(off >> PAGE_SHIFT);
 }
 
-/*
- * These are not meant to be generic - an integer key is assumed.
- */
-#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
+#define RB_BYVAL(a)      (a)
+#define RB_BYPTR(a)      (&(a))
+#define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
+
+#define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
 static void insert_##name(struct rb_root *root, type *t)		\
 {									\
 	struct rb_node **n = &root->rb_node;				\
@@ -197,11 +198,13 @@ static void insert_##name(struct rb_root *root, type *t)		\
 									\
 	while (*n) {							\
 		type *cur = rb_entry(*n, type, nodefld);		\
+		int cmp;						\
 									\
 		parent = *n;						\
-		if (t->keyfld < cur->keyfld)				\
+		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
+		if (cmp < 0)						\
 			n = &(*n)->rb_left;				\
-		else if (t->keyfld > cur->keyfld)			\
+		else if (cmp > 0)					\
 			n = &(*n)->rb_right;				\
 		else							\
 			BUG();						\
@@ -217,19 +220,24 @@ static void erase_##name(struct rb_root *root, type *t)			\
 	RB_CLEAR_NODE(&t->nodefld);					\
 }
 
-#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
-extern type __lookup_##name##_key;					\
-static type *lookup_##name(struct rb_root *root,			\
-			   typeof(__lookup_##name##_key.keyfld) key)	\
+/*
+ * @lookup_param_type is a parameter and not constructed from (@type,
+ * @keyfld) with typeof() because adding const is too unwieldy.
+ */
+#define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
+			       lookup_param_type, nodefld)		\
+static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
 {									\
 	struct rb_node *n = root->rb_node;				\
 									\
 	while (n) {							\
 		type *cur = rb_entry(n, type, nodefld);			\
+		int cmp;						\
 									\
-		if (key < cur->keyfld)					\
+		cmp = cmpexp(key, keyexp(cur->keyfld));			\
+		if (cmp < 0)						\
 			n = n->rb_left;					\
-		else if (key > cur->keyfld)				\
+		else if (cmp > 0)					\
 			n = n->rb_right;				\
 		else							\
 			return cur;					\
@@ -238,6 +246,23 @@ static type *lookup_##name(struct rb_root *root,			\
 	return NULL;							\
 }
 
+#define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
+			 lookup_param_type, nodefld)			\
+DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
+DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
+		       lookup_param_type, nodefld)
+
+/*
+ * Shorthands for integer keys.
+ */
+#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
+DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
+
+#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
+extern type __lookup_##name##_key;					\
+DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
+		       typeof(__lookup_##name##_key.keyfld), nodefld)
+
 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)
diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h
index c5c4c713e00f..fbd94d9fa5dd 100644
--- a/include/linux/ceph/messenger.h
+++ b/include/linux/ceph/messenger.h
@@ -44,6 +44,8 @@ struct ceph_connection_operations {
 					struct ceph_msg_header *hdr,
 					int *skip);
 
+	void (*reencode_message) (struct ceph_msg *msg);
+
 	int (*sign_message) (struct ceph_msg *msg);
 	int (*check_message_signature) (struct ceph_msg *msg);
 };
diff --git a/include/linux/ceph/osd_client.h b/include/linux/ceph/osd_client.h
index 85650b415e73..c6d96a5f46fd 100644
--- a/include/linux/ceph/osd_client.h
+++ b/include/linux/ceph/osd_client.h
@@ -1,6 +1,7 @@
 #ifndef _FS_CEPH_OSD_CLIENT_H
 #define _FS_CEPH_OSD_CLIENT_H
 
+#include <linux/bitrev.h>
 #include <linux/completion.h>
 #include <linux/kref.h>
 #include <linux/mempool.h>
@@ -36,6 +37,8 @@ struct ceph_osd {
 	struct ceph_connection o_con;
 	struct rb_root o_requests;
 	struct rb_root o_linger_requests;
+	struct rb_root o_backoff_mappings;
+	struct rb_root o_backoffs_by_id;
 	struct list_head o_osd_lru;
 	struct ceph_auth_handshake o_auth;
 	unsigned long lru_ttl;
@@ -136,7 +139,8 @@ struct ceph_osd_request_target {
 	struct ceph_object_id target_oid;
 	struct ceph_object_locator target_oloc;
 
-	struct ceph_pg pgid;
+	struct ceph_pg pgid;               /* last raw pg we mapped to */
+	struct ceph_spg spgid;             /* last actual spg we mapped to */
 	u32 pg_num;
 	u32 pg_num_mask;
 	struct ceph_osds acting;
@@ -148,6 +152,9 @@ struct ceph_osd_request_target {
 	unsigned int flags;                /* CEPH_OSD_FLAG_* */
 	bool paused;
 
+	u32 epoch;
+	u32 last_force_resend;
+
 	int osd;
 };
 
@@ -193,7 +200,6 @@ struct ceph_osd_request {
 	unsigned long r_stamp;                /* jiffies, send or check time */
 	unsigned long r_start_stamp;          /* jiffies */
 	int r_attempts;
-	u32 r_last_force_resend;
 	u32 r_map_dne_bound;
 
 	struct ceph_osd_req_op r_ops[];
@@ -203,6 +209,23 @@ struct ceph_request_redirect {
 	struct ceph_object_locator oloc;
 };
 
+/*
+ * osd request identifier
+ *
+ * caller name + incarnation# + tid to unique identify this request
+ */
+struct ceph_osd_reqid {
+	struct ceph_entity_name name;
+	__le64 tid;
+	__le32 inc;
+} __packed;
+
+struct ceph_blkin_trace_info {
+	__le64 trace_id;
+	__le64 span_id;
+	__le64 parent_span_id;
+} __packed;
+
 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie,
 				 u64 notifier_id, void *data, size_t data_len);
 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err);
@@ -221,7 +244,6 @@ struct ceph_osd_linger_request {
 	struct list_head pending_lworks;
 
 	struct ceph_osd_request_target t;
-	u32 last_force_resend;
 	u32 map_dne_bound;
 
 	struct timespec mtime;
@@ -256,6 +278,48 @@ struct ceph_watch_item {
 	struct ceph_entity_addr addr;
 };
 
+struct ceph_spg_mapping {
+	struct rb_node node;
+	struct ceph_spg spgid;
+
+	struct rb_root backoffs;
+};
+
+struct ceph_hobject_id {
+	void *key;
+	size_t key_len;
+	void *oid;
+	size_t oid_len;
+	u64 snapid;
+	u32 hash;
+	u8 is_max;
+	void *nspace;
+	size_t nspace_len;
+	s64 pool;
+
+	/* cache */
+	u32 hash_reverse_bits;
+};
+
+static inline void ceph_hoid_build_hash_cache(struct ceph_hobject_id *hoid)
+{
+	hoid->hash_reverse_bits = bitrev32(hoid->hash);
+}
+
+/*
+ * PG-wide backoff: [begin, end)
+ * per-object backoff: begin == end
+ */
+struct ceph_osd_backoff {
+	struct rb_node spg_node;
+	struct rb_node id_node;
+
+	struct ceph_spg spgid;
+	u64 id;
+	struct ceph_hobject_id *begin;
+	struct ceph_hobject_id *end;
+};
+
 #define CEPH_LINGER_ID_START	0xffff000000000000ULL
 
 struct ceph_osd_client {
diff --git a/include/linux/ceph/osdmap.h b/include/linux/ceph/osdmap.h
index 938656f70807..a0996cb9faed 100644
--- a/include/linux/ceph/osdmap.h
+++ b/include/linux/ceph/osdmap.h
@@ -24,7 +24,15 @@ struct ceph_pg {
 	uint32_t seed;
 };
 
+#define CEPH_SPG_NOSHARD	-1
+
+struct ceph_spg {
+	struct ceph_pg pgid;
+	s8 shard;
+};
+
 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs);
+int ceph_spg_compare(const struct ceph_spg *lhs, const struct ceph_spg *rhs);
 
 #define CEPH_POOL_FLAG_HASHPSPOOL	(1ULL << 0) /* hash pg seed and pool id
 						       together */
@@ -135,10 +143,14 @@ struct ceph_pg_mapping {
 		struct {
 			int len;
 			int osds[];
-		} pg_temp;
+		} pg_temp, pg_upmap;
 		struct {
 			int osd;
 		} primary_temp;
+		struct {
+			int len;
+			int from_to[][2];
+		} pg_upmap_items;
 	};
 };
 
@@ -150,13 +162,17 @@ struct ceph_osdmap {
 	u32 flags;         /* CEPH_OSDMAP_* */
 
 	u32 max_osd;       /* size of osd_state, _offload, _addr arrays */
-	u8 *osd_state;     /* CEPH_OSD_* */
+	u32 *osd_state;    /* CEPH_OSD_* */
 	u32 *osd_weight;   /* 0 = failed, 0x10000 = 100% normal */
 	struct ceph_entity_addr *osd_addr;
 
 	struct rb_root pg_temp;
 	struct rb_root primary_temp;
 
+	/* remap (post-CRUSH, pre-up) */
+	struct rb_root pg_upmap;	/* PG := raw set */
+	struct rb_root pg_upmap_items;	/* from -> to within raw set */
+
 	u32 *osd_primary_affinity;
 
 	struct rb_root pg_pools;
@@ -187,7 +203,7 @@ static inline bool ceph_osd_is_down(struct ceph_osdmap *map, int osd)
 	return !ceph_osd_is_up(map, osd);
 }
 
-extern char *ceph_osdmap_state_str(char *str, int len, int state);
+char *ceph_osdmap_state_str(char *str, int len, u32 state);
 extern u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd);
 
 static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
@@ -198,11 +214,13 @@ static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
 	return &map->osd_addr[osd];
 }
 
+#define CEPH_PGID_ENCODING_LEN		(1 + 8 + 4 + 4)
+
 static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
 {
 	__u8 version;
 
-	if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
+	if (!ceph_has_room(p, end, CEPH_PGID_ENCODING_LEN)) {
 		pr_warn("incomplete pg encoding\n");
 		return -EINVAL;
 	}
@@ -240,6 +258,8 @@ static inline void ceph_osds_init(struct ceph_osds *set)
 
 void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src);
 
+bool ceph_pg_is_split(const struct ceph_pg *pgid, u32 old_pg_num,
+		      u32 new_pg_num);
 bool ceph_is_new_interval(const struct ceph_osds *old_acting,
 			  const struct ceph_osds *new_acting,
 			  const struct ceph_osds *old_up,
@@ -262,15 +282,24 @@ extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
 					 u64 off, u64 len,
 					 u64 *bno, u64 *oxoff, u64 *oxlen);
 
+int __ceph_object_locator_to_pg(struct ceph_pg_pool_info *pi,
+				const struct ceph_object_id *oid,
+				const struct ceph_object_locator *oloc,
+				struct ceph_pg *raw_pgid);
 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
-			      struct ceph_object_id *oid,
-			      struct ceph_object_locator *oloc,
+			      const struct ceph_object_id *oid,
+			      const struct ceph_object_locator *oloc,
 			      struct ceph_pg *raw_pgid);
 
 void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap,
+			       struct ceph_pg_pool_info *pi,
 			       const struct ceph_pg *raw_pgid,
 			       struct ceph_osds *up,
 			       struct ceph_osds *acting);
+bool ceph_pg_to_primary_shard(struct ceph_osdmap *osdmap,
+			      struct ceph_pg_pool_info *pi,
+			      const struct ceph_pg *raw_pgid,
+			      struct ceph_spg *spgid);
 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
 			      const struct ceph_pg *raw_pgid);
 
diff --git a/include/linux/ceph/rados.h b/include/linux/ceph/rados.h
index 5d0018782d50..385db08bb8b2 100644
--- a/include/linux/ceph/rados.h
+++ b/include/linux/ceph/rados.h
@@ -439,6 +439,12 @@ enum {
 
 const char *ceph_osd_watch_op_name(int o);
 
+enum {
+	CEPH_OSD_BACKOFF_OP_BLOCK = 1,
+	CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
+	CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
+};
+
 /*
  * an individual object operation.  each may be accompanied by some data
  * payload
diff --git a/include/linux/crush/crush.h b/include/linux/crush/crush.h
index fbecbd089d75..92e165d417a6 100644
--- a/include/linux/crush/crush.h
+++ b/include/linux/crush/crush.h
@@ -2,6 +2,7 @@
 #define CEPH_CRUSH_CRUSH_H
 
 #ifdef __KERNEL__
+# include <linux/rbtree.h>
 # include <linux/types.h>
 #else
 # include "crush_compat.h"
@@ -137,6 +138,68 @@ struct crush_bucket {
 
 };
 
+/** @ingroup API
+ *
+ * Replacement weights for each item in a bucket. The size of the
+ * array must be exactly the size of the straw2 bucket, just as the
+ * item_weights array.
+ *
+ */
+struct crush_weight_set {
+	__u32 *weights; /*!< 16.16 fixed point weights
+                             in the same order as items */
+	__u32 size;     /*!< size of the __weights__ array */
+};
+
+/** @ingroup API
+ *
+ * Replacement weights and ids for a given straw2 bucket, for
+ * placement purposes.
+ *
+ * When crush_do_rule() chooses the Nth item from a straw2 bucket, the
+ * replacement weights found at __weight_set[N]__ are used instead of
+ * the weights from __item_weights__. If __N__ is greater than
+ * __weight_set_size__, the weights found at __weight_set_size-1__ are
+ * used instead. For instance if __weight_set__ is:
+ *
+ *    [ [ 0x10000, 0x20000 ],   // position 0
+ *      [ 0x20000, 0x40000 ] ]  // position 1
+ *
+ * choosing the 0th item will use position 0 weights [ 0x10000, 0x20000 ]
+ * choosing the 1th item will use position 1 weights [ 0x20000, 0x40000 ]
+ * choosing the 2th item will use position 1 weights [ 0x20000, 0x40000 ]
+ * etc.
+ *
+ */
+struct crush_choose_arg {
+	__s32 *ids;            /*!< values to use instead of items */
+	__u32 ids_size;        /*!< size of the __ids__ array */
+	struct crush_weight_set *weight_set; /*!< weight replacements for
+                                                  a given position */
+	__u32 weight_set_size; /*!< size of the __weight_set__ array */
+};
+
+/** @ingroup API
+ *
+ * Replacement weights and ids for each bucket in the crushmap. The
+ * __size__ of the __args__ array must be exactly the same as the
+ * __map->max_buckets__.
+ *
+ * The __crush_choose_arg__ at index N will be used when choosing
+ * an item from the bucket __map->buckets[N]__ bucket, provided it
+ * is a straw2 bucket.
+ *
+ */
+struct crush_choose_arg_map {
+#ifdef __KERNEL__
+	struct rb_node node;
+	u64 choose_args_index;
+#endif
+	struct crush_choose_arg *args; /*!< replacement for each bucket
+                                            in the crushmap */
+	__u32 size;                    /*!< size of the __args__ array */
+};
+
 struct crush_bucket_uniform {
 	struct crush_bucket h;
 	__u32 item_weight;  /* 16-bit fixed point; all items equally weighted */
@@ -236,6 +299,9 @@ struct crush_map {
 	__u32 allowed_bucket_algs;
 
 	__u32 *choose_tries;
+#else
+	/* CrushWrapper::choose_args */
+	struct rb_root choose_args;
 #endif
 };
 
diff --git a/include/linux/crush/mapper.h b/include/linux/crush/mapper.h
index c95e19e1ff11..141edabb947e 100644
--- a/include/linux/crush/mapper.h
+++ b/include/linux/crush/mapper.h
@@ -11,11 +11,10 @@
 #include "crush.h"
 
 extern int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size);
-extern int crush_do_rule(const struct crush_map *map,
-			 int ruleno,
-			 int x, int *result, int result_max,
-			 const __u32 *weights, int weight_max,
-			 void *cwin);
+int crush_do_rule(const struct crush_map *map,
+		  int ruleno, int x, int *result, int result_max,
+		  const __u32 *weight, int weight_max,
+		  void *cwin, const struct crush_choose_arg *choose_args);
 
 /*
  * Returns the exact amount of workspace that will need to be used