Merge tag 'v4.13-rc1' into fixes

The fixes branch is based off a random pre-rc1 commit, because we had
some fixes that needed to go in before rc1 was released.

However we now need to fix some code that went in after that point, but
before rc1, so merge rc1 to get that code into fixes so we can fix it!

45 files changed, 2868 insertions, 1470 deletions

diff --git a/net/9p/client.c b/net/9p/client.c
index 1218fb3b52da..4674235b0d9b 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -37,6 +37,7 @@
 #include <linux/uio.h>
 #include <net/9p/9p.h>
 #include <linux/parser.h>
+#include <linux/seq_file.h>
 #include <net/9p/client.h>
 #include <net/9p/transport.h>
 #include "protocol.h"
@@ -77,6 +78,30 @@ inline int p9_is_proto_dotu(struct p9_client *clnt)
 }
 EXPORT_SYMBOL(p9_is_proto_dotu);
 
+int p9_show_client_options(struct seq_file *m, struct p9_client *clnt)
+{
+	if (clnt->msize != 8192)
+		seq_printf(m, ",msize=%u", clnt->msize);
+	seq_printf(m, "trans=%s", clnt->trans_mod->name);
+
+	switch (clnt->proto_version) {
+	case p9_proto_legacy:
+		seq_puts(m, ",noextend");
+		break;
+	case p9_proto_2000u:
+		seq_puts(m, ",version=9p2000.u");
+		break;
+	case p9_proto_2000L:
+		/* Default */
+		break;
+	}
+
+	if (clnt->trans_mod->show_options)
+		return clnt->trans_mod->show_options(m, clnt);
+	return 0;
+}
+EXPORT_SYMBOL(p9_show_client_options);
+
 /*
  * Some error codes are taken directly from the server replies,
  * make sure they are valid.
diff --git a/net/9p/trans_fd.c b/net/9p/trans_fd.c
index dca3cdd1a014..ddfa86648f95 100644
--- a/net/9p/trans_fd.c
+++ b/net/9p/trans_fd.c
@@ -41,6 +41,7 @@
 #include <linux/file.h>
 #include <linux/parser.h>
 #include <linux/slab.h>
+#include <linux/seq_file.h>
 #include <net/9p/9p.h>
 #include <net/9p/client.h>
 #include <net/9p/transport.h>
@@ -51,6 +52,9 @@
 #define MAX_SOCK_BUF (64*1024)
 #define MAXPOLLWADDR	2
 
+static struct p9_trans_module p9_tcp_trans;
+static struct p9_trans_module p9_fd_trans;
+
 /**
  * struct p9_fd_opts - per-transport options
  * @rfd: file descriptor for reading (trans=fd)
@@ -63,7 +67,7 @@ struct p9_fd_opts {
 	int rfd;
 	int wfd;
 	u16 port;
-	int privport;
+	bool privport;
 };
 
 /*
@@ -720,6 +724,20 @@ static int p9_fd_cancelled(struct p9_client *client, struct p9_req_t *req)
 	return 0;
 }
 
+static int p9_fd_show_options(struct seq_file *m, struct p9_client *clnt)
+{
+	if (clnt->trans_mod == &p9_tcp_trans) {
+		if (clnt->trans_opts.tcp.port != P9_PORT)
+			seq_printf(m, "port=%u", clnt->trans_opts.tcp.port);
+	} else if (clnt->trans_mod == &p9_fd_trans) {
+		if (clnt->trans_opts.fd.rfd != ~0)
+			seq_printf(m, "rfd=%u", clnt->trans_opts.fd.rfd);
+		if (clnt->trans_opts.fd.wfd != ~0)
+			seq_printf(m, "wfd=%u", clnt->trans_opts.fd.wfd);
+	}
+	return 0;
+}
+
 /**
  * parse_opts - parse mount options into p9_fd_opts structure
  * @params: options string passed from mount
@@ -738,7 +756,7 @@ static int parse_opts(char *params, struct p9_fd_opts *opts)
 	opts->port = P9_PORT;
 	opts->rfd = ~0;
 	opts->wfd = ~0;
-	opts->privport = 0;
+	opts->privport = false;
 
 	if (!params)
 		return 0;
@@ -776,7 +794,7 @@ static int parse_opts(char *params, struct p9_fd_opts *opts)
 			opts->wfd = option;
 			break;
 		case Opt_privport:
-			opts->privport = 1;
+			opts->privport = true;
 			break;
 		default:
 			continue;
@@ -942,6 +960,8 @@ p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
 
 	csocket = NULL;
 
+	client->trans_opts.tcp.port = opts.port;
+	client->trans_opts.tcp.privport = opts.privport;
 	sin_server.sin_family = AF_INET;
 	sin_server.sin_addr.s_addr = in_aton(addr);
 	sin_server.sin_port = htons(opts.port);
@@ -1020,6 +1040,8 @@ p9_fd_create(struct p9_client *client, const char *addr, char *args)
 	struct p9_fd_opts opts;
 
 	parse_opts(args, &opts);
+	client->trans_opts.fd.rfd = opts.rfd;
+	client->trans_opts.fd.wfd = opts.wfd;
 
 	if (opts.rfd == ~0 || opts.wfd == ~0) {
 		pr_err("Insufficient options for proto=fd\n");
@@ -1044,6 +1066,7 @@ static struct p9_trans_module p9_tcp_trans = {
 	.request = p9_fd_request,
 	.cancel = p9_fd_cancel,
 	.cancelled = p9_fd_cancelled,
+	.show_options = p9_fd_show_options,
 	.owner = THIS_MODULE,
 };
 
@@ -1056,6 +1079,7 @@ static struct p9_trans_module p9_unix_trans = {
 	.request = p9_fd_request,
 	.cancel = p9_fd_cancel,
 	.cancelled = p9_fd_cancelled,
+	.show_options = p9_fd_show_options,
 	.owner = THIS_MODULE,
 };
 
@@ -1068,6 +1092,7 @@ static struct p9_trans_module p9_fd_trans = {
 	.request = p9_fd_request,
 	.cancel = p9_fd_cancel,
 	.cancelled = p9_fd_cancelled,
+	.show_options = p9_fd_show_options,
 	.owner = THIS_MODULE,
 };
 
diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c
index 553ed4ecb6a0..6d8e3031978f 100644
--- a/net/9p/trans_rdma.c
+++ b/net/9p/trans_rdma.c
@@ -43,6 +43,7 @@
 #include <linux/parser.h>
 #include <linux/semaphore.h>
 #include <linux/slab.h>
+#include <linux/seq_file.h>
 #include <net/9p/9p.h>
 #include <net/9p/client.h>
 #include <net/9p/transport.h>
@@ -70,6 +71,8 @@
  * @dm_mr: DMA Memory Region pointer
  * @lkey: The local access only memory region key
  * @timeout: Number of uSecs to wait for connection management events
+ * @privport: Whether a privileged port may be used
+ * @port: The port to use
  * @sq_depth: The depth of the Send Queue
  * @sq_sem: Semaphore for the SQ
  * @rq_depth: The depth of the Receive Queue.
@@ -95,6 +98,8 @@ struct p9_trans_rdma {
 	struct ib_qp *qp;
 	struct ib_cq *cq;
 	long timeout;
+	bool privport;
+	u16 port;
 	int sq_depth;
 	struct semaphore sq_sem;
 	int rq_depth;
@@ -133,10 +138,10 @@ struct p9_rdma_context {
  */
 struct p9_rdma_opts {
 	short port;
+	bool privport;
 	int sq_depth;
 	int rq_depth;
 	long timeout;
-	int privport;
 };
 
 /*
@@ -159,6 +164,23 @@ static match_table_t tokens = {
 	{Opt_err, NULL},
 };
 
+static int p9_rdma_show_options(struct seq_file *m, struct p9_client *clnt)
+{
+	struct p9_trans_rdma *rdma = clnt->trans;
+
+	if (rdma->port != P9_PORT)
+		seq_printf(m, ",port=%u", rdma->port);
+	if (rdma->sq_depth != P9_RDMA_SQ_DEPTH)
+		seq_printf(m, ",sq=%u", rdma->sq_depth);
+	if (rdma->rq_depth != P9_RDMA_RQ_DEPTH)
+		seq_printf(m, ",rq=%u", rdma->rq_depth);
+	if (rdma->timeout != P9_RDMA_TIMEOUT)
+		seq_printf(m, ",timeout=%lu", rdma->timeout);
+	if (rdma->privport)
+		seq_puts(m, ",privport");
+	return 0;
+}
+
 /**
  * parse_opts - parse mount options into rdma options structure
  * @params: options string passed from mount
@@ -177,7 +199,7 @@ static int parse_opts(char *params, struct p9_rdma_opts *opts)
 	opts->sq_depth = P9_RDMA_SQ_DEPTH;
 	opts->rq_depth = P9_RDMA_RQ_DEPTH;
 	opts->timeout = P9_RDMA_TIMEOUT;
-	opts->privport = 0;
+	opts->privport = false;
 
 	if (!params)
 		return 0;
@@ -218,7 +240,7 @@ static int parse_opts(char *params, struct p9_rdma_opts *opts)
 			opts->timeout = option;
 			break;
 		case Opt_privport:
-			opts->privport = 1;
+			opts->privport = true;
 			break;
 		default:
 			continue;
@@ -560,6 +582,8 @@ static struct p9_trans_rdma *alloc_rdma(struct p9_rdma_opts *opts)
 	if (!rdma)
 		return NULL;
 
+	rdma->port = opts->port;
+	rdma->privport = opts->privport;
 	rdma->sq_depth = opts->sq_depth;
 	rdma->rq_depth = opts->rq_depth;
 	rdma->timeout = opts->timeout;
@@ -733,6 +757,7 @@ static struct p9_trans_module p9_rdma_trans = {
 	.request = rdma_request,
 	.cancel = rdma_cancel,
 	.cancelled = rdma_cancelled,
+	.show_options = p9_rdma_show_options,
 };
 
 /**
diff --git a/net/bridge/br_mdb.c b/net/bridge/br_mdb.c
index 09dcdb9c0f3c..a0b11e7d67d9 100644
--- a/net/bridge/br_mdb.c
+++ b/net/bridge/br_mdb.c
@@ -323,7 +323,8 @@ static void __br_mdb_notify(struct net_device *dev, struct net_bridge_port *p,
 			__mdb_entry_to_br_ip(entry, &complete_info->ip);
 			mdb.obj.complete_priv = complete_info;
 			mdb.obj.complete = br_mdb_complete;
-			switchdev_port_obj_add(port_dev, &mdb.obj);
+			if (switchdev_port_obj_add(port_dev, &mdb.obj))
+				kfree(complete_info);
 		}
 	} else if (port_dev && type == RTM_DELMDB) {
 		switchdev_port_obj_del(port_dev, &mdb.obj);
diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c
index 47e94b560ba0..5c036d2f401e 100644
--- a/net/ceph/ceph_common.c
+++ b/net/ceph/ceph_common.c
@@ -85,6 +85,7 @@ const char *ceph_msg_type_name(int type)
 	case CEPH_MSG_OSD_OP: return "osd_op";
 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
+	case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
 	default: return "unknown";
 	}
 }
@@ -598,7 +599,11 @@ struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
 {
 	struct ceph_client *client;
 	struct ceph_entity_addr *myaddr = NULL;
-	int err = -ENOMEM;
+	int err;
+
+	err = wait_for_random_bytes();
+	if (err < 0)
+		return ERR_PTR(err);
 
 	client = kzalloc(sizeof(*client), GFP_KERNEL);
 	if (client == NULL)
diff --git a/net/ceph/crush/crush.c b/net/ceph/crush/crush.c
index 5bf94c04f645..4b428f46a8ca 100644
--- a/net/ceph/crush/crush.c
+++ b/net/ceph/crush/crush.c
@@ -1,6 +1,7 @@
 #ifdef __KERNEL__
 # include <linux/slab.h>
 # include <linux/crush/crush.h>
+void clear_choose_args(struct crush_map *c);
 #else
 # include "crush_compat.h"
 # include "crush.h"
@@ -127,6 +128,8 @@ void crush_destroy(struct crush_map *map)
 
 #ifndef __KERNEL__
 	kfree(map->choose_tries);
+#else
+	clear_choose_args(map);
 #endif
 	kfree(map);
 }
diff --git a/net/ceph/crush/mapper.c b/net/ceph/crush/mapper.c
index b5cd8c21bfdf..746b145bfd11 100644
--- a/net/ceph/crush/mapper.c
+++ b/net/ceph/crush/mapper.c
@@ -302,19 +302,42 @@ static __u64 crush_ln(unsigned int xin)
  *
  */
 
+static __u32 *get_choose_arg_weights(const struct crush_bucket_straw2 *bucket,
+				     const struct crush_choose_arg *arg,
+				     int position)
+{
+	if (!arg || !arg->weight_set || arg->weight_set_size == 0)
+		return bucket->item_weights;
+
+	if (position >= arg->weight_set_size)
+		position = arg->weight_set_size - 1;
+	return arg->weight_set[position].weights;
+}
+
+static __s32 *get_choose_arg_ids(const struct crush_bucket_straw2 *bucket,
+				 const struct crush_choose_arg *arg)
+{
+	if (!arg || !arg->ids)
+		return bucket->h.items;
+
+	return arg->ids;
+}
+
 static int bucket_straw2_choose(const struct crush_bucket_straw2 *bucket,
-				int x, int r)
+				int x, int r,
+				const struct crush_choose_arg *arg,
+				int position)
 {
 	unsigned int i, high = 0;
 	unsigned int u;
-	unsigned int w;
 	__s64 ln, draw, high_draw = 0;
+	__u32 *weights = get_choose_arg_weights(bucket, arg, position);
+	__s32 *ids = get_choose_arg_ids(bucket, arg);
 
 	for (i = 0; i < bucket->h.size; i++) {
-		w = bucket->item_weights[i];
-		if (w) {
-			u = crush_hash32_3(bucket->h.hash, x,
-					   bucket->h.items[i], r);
+		dprintk("weight 0x%x item %d\n", weights[i], ids[i]);
+		if (weights[i]) {
+			u = crush_hash32_3(bucket->h.hash, x, ids[i], r);
 			u &= 0xffff;
 
 			/*
@@ -335,7 +358,7 @@ static int bucket_straw2_choose(const struct crush_bucket_straw2 *bucket,
 			 * weight means a larger (less negative) value
 			 * for draw.
 			 */
-			draw = div64_s64(ln, w);
+			draw = div64_s64(ln, weights[i]);
 		} else {
 			draw = S64_MIN;
 		}
@@ -352,7 +375,9 @@ static int bucket_straw2_choose(const struct crush_bucket_straw2 *bucket,
 
 static int crush_bucket_choose(const struct crush_bucket *in,
 			       struct crush_work_bucket *work,
-			       int x, int r)
+			       int x, int r,
+			       const struct crush_choose_arg *arg,
+			       int position)
 {
 	dprintk(" crush_bucket_choose %d x=%d r=%d\n", in->id, x, r);
 	BUG_ON(in->size == 0);
@@ -374,7 +399,7 @@ static int crush_bucket_choose(const struct crush_bucket *in,
 	case CRUSH_BUCKET_STRAW2:
 		return bucket_straw2_choose(
 			(const struct crush_bucket_straw2 *)in,
-			x, r);
+			x, r, arg, position);
 	default:
 		dprintk("unknown bucket %d alg %d\n", in->id, in->alg);
 		return in->items[0];
@@ -436,7 +461,8 @@ static int crush_choose_firstn(const struct crush_map *map,
 			       unsigned int vary_r,
 			       unsigned int stable,
 			       int *out2,
-			       int parent_r)
+			       int parent_r,
+			       const struct crush_choose_arg *choose_args)
 {
 	int rep;
 	unsigned int ftotal, flocal;
@@ -486,7 +512,10 @@ static int crush_choose_firstn(const struct crush_map *map,
 				else
 					item = crush_bucket_choose(
 						in, work->work[-1-in->id],
-						x, r);
+						x, r,
+						(choose_args ?
+						 &choose_args[-1-in->id] : 0),
+						outpos);
 				if (item >= map->max_devices) {
 					dprintk("   bad item %d\n", item);
 					skip_rep = 1;
@@ -543,7 +572,8 @@ static int crush_choose_firstn(const struct crush_map *map,
 							    vary_r,
 							    stable,
 							    NULL,
-							    sub_r) <= outpos)
+							    sub_r,
+							    choose_args) <= outpos)
 							/* didn't get leaf */
 							reject = 1;
 					} else {
@@ -620,7 +650,8 @@ static void crush_choose_indep(const struct crush_map *map,
 			       unsigned int recurse_tries,
 			       int recurse_to_leaf,
 			       int *out2,
-			       int parent_r)
+			       int parent_r,
+			       const struct crush_choose_arg *choose_args)
 {
 	const struct crush_bucket *in = bucket;
 	int endpos = outpos + left;
@@ -692,7 +723,10 @@ static void crush_choose_indep(const struct crush_map *map,
 
 				item = crush_bucket_choose(
 					in, work->work[-1-in->id],
-					x, r);
+					x, r,
+					(choose_args ?
+					 &choose_args[-1-in->id] : 0),
+					outpos);
 				if (item >= map->max_devices) {
 					dprintk("   bad item %d\n", item);
 					out[rep] = CRUSH_ITEM_NONE;
@@ -746,7 +780,8 @@ static void crush_choose_indep(const struct crush_map *map,
 							x, 1, numrep, 0,
 							out2, rep,
 							recurse_tries, 0,
-							0, NULL, r);
+							0, NULL, r,
+							choose_args);
 						if (out2[rep] == CRUSH_ITEM_NONE) {
 							/* placed nothing; no leaf */
 							break;
@@ -823,7 +858,7 @@ void crush_init_workspace(const struct crush_map *map, void *v)
 	 * set the pointer first and then reserve the space for it to
 	 * point to by incrementing the point.
 	 */
-	v += sizeof(struct crush_work *);
+	v += sizeof(struct crush_work);
 	w->work = v;
 	v += map->max_buckets * sizeof(struct crush_work_bucket *);
 	for (b = 0; b < map->max_buckets; ++b) {
@@ -854,11 +889,12 @@ void crush_init_workspace(const struct crush_map *map, void *v)
  * @weight: weight vector (for map leaves)
  * @weight_max: size of weight vector
  * @cwin: pointer to at least crush_work_size() bytes of memory
+ * @choose_args: weights and ids for each known bucket
  */
 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)
+		  void *cwin, const struct crush_choose_arg *choose_args)
 {
 	int result_len;
 	struct crush_work *cw = cwin;
@@ -968,11 +1004,6 @@ int crush_do_rule(const struct crush_map *map,
 
 			for (i = 0; i < wsize; i++) {
 				int bno;
-				/*
-				 * see CRUSH_N, CRUSH_N_MINUS macros.
-				 * basically, numrep <= 0 means relative to
-				 * the provided result_max
-				 */
 				numrep = curstep->arg1;
 				if (numrep <= 0) {
 					numrep += result_max;
@@ -1013,7 +1044,8 @@ int crush_do_rule(const struct crush_map *map,
 						vary_r,
 						stable,
 						c+osize,
-						0);
+						0,
+						choose_args);
 				} else {
 					out_size = ((numrep < (result_max-osize)) ?
 						    numrep : (result_max-osize));
@@ -1030,7 +1062,8 @@ int crush_do_rule(const struct crush_map *map,
 						   choose_leaf_tries : 1,
 						recurse_to_leaf,
 						c+osize,
-						0);
+						0,
+						choose_args);
 					osize += out_size;
 				}
 			}
diff --git a/net/ceph/debugfs.c b/net/ceph/debugfs.c
index 71ba13927b3d..fa5233e0d01c 100644
--- a/net/ceph/debugfs.c
+++ b/net/ceph/debugfs.c
@@ -77,7 +77,7 @@ static int osdmap_show(struct seq_file *s, void *p)
 	}
 	for (i = 0; i < map->max_osd; i++) {
 		struct ceph_entity_addr *addr = &map->osd_addr[i];
-		int state = map->osd_state[i];
+		u32 state = map->osd_state[i];
 		char sb[64];
 
 		seq_printf(s, "osd%d\t%s\t%3d%%\t(%s)\t%3d%%\n",
@@ -104,6 +104,29 @@ static int osdmap_show(struct seq_file *s, void *p)
 		seq_printf(s, "primary_temp %llu.%x %d\n", pg->pgid.pool,
 			   pg->pgid.seed, pg->primary_temp.osd);
 	}
+	for (n = rb_first(&map->pg_upmap); n; n = rb_next(n)) {
+		struct ceph_pg_mapping *pg =
+			rb_entry(n, struct ceph_pg_mapping, node);
+
+		seq_printf(s, "pg_upmap %llu.%x [", pg->pgid.pool,
+			   pg->pgid.seed);
+		for (i = 0; i < pg->pg_upmap.len; i++)
+			seq_printf(s, "%s%d", (i == 0 ? "" : ","),
+				   pg->pg_upmap.osds[i]);
+		seq_printf(s, "]\n");
+	}
+	for (n = rb_first(&map->pg_upmap_items); n; n = rb_next(n)) {
+		struct ceph_pg_mapping *pg =
+			rb_entry(n, struct ceph_pg_mapping, node);
+
+		seq_printf(s, "pg_upmap_items %llu.%x [", pg->pgid.pool,
+			   pg->pgid.seed);
+		for (i = 0; i < pg->pg_upmap_items.len; i++)
+			seq_printf(s, "%s%d->%d", (i == 0 ? "" : ","),
+				   pg->pg_upmap_items.from_to[i][0],
+				   pg->pg_upmap_items.from_to[i][1]);
+		seq_printf(s, "]\n");
+	}
 
 	up_read(&osdc->lock);
 	return 0;
@@ -147,17 +170,26 @@ static int monc_show(struct seq_file *s, void *p)
 	return 0;
 }
 
+static void dump_spgid(struct seq_file *s, const struct ceph_spg *spgid)
+{
+	seq_printf(s, "%llu.%x", spgid->pgid.pool, spgid->pgid.seed);
+	if (spgid->shard != CEPH_SPG_NOSHARD)
+		seq_printf(s, "s%d", spgid->shard);
+}
+
 static void dump_target(struct seq_file *s, struct ceph_osd_request_target *t)
 {
 	int i;
 
-	seq_printf(s, "osd%d\t%llu.%x\t[", t->osd, t->pgid.pool, t->pgid.seed);
+	seq_printf(s, "osd%d\t%llu.%x\t", t->osd, t->pgid.pool, t->pgid.seed);
+	dump_spgid(s, &t->spgid);
+	seq_puts(s, "\t[");
 	for (i = 0; i < t->up.size; i++)
 		seq_printf(s, "%s%d", (!i ? "" : ","), t->up.osds[i]);
 	seq_printf(s, "]/%d\t[", t->up.primary);
 	for (i = 0; i < t->acting.size; i++)
 		seq_printf(s, "%s%d", (!i ? "" : ","), t->acting.osds[i]);
-	seq_printf(s, "]/%d\t", t->acting.primary);
+	seq_printf(s, "]/%d\te%u\t", t->acting.primary, t->epoch);
 	if (t->target_oloc.pool_ns) {
 		seq_printf(s, "%*pE/%*pE\t0x%x",
 			(int)t->target_oloc.pool_ns->len,
@@ -234,6 +266,73 @@ static void dump_linger_requests(struct seq_file *s, struct ceph_osd *osd)
 	mutex_unlock(&osd->lock);
 }
 
+static void dump_snapid(struct seq_file *s, u64 snapid)
+{
+	if (snapid == CEPH_NOSNAP)
+		seq_puts(s, "head");
+	else if (snapid == CEPH_SNAPDIR)
+		seq_puts(s, "snapdir");
+	else
+		seq_printf(s, "%llx", snapid);
+}
+
+static void dump_name_escaped(struct seq_file *s, unsigned char *name,
+			      size_t len)
+{
+	size_t i;
+
+	for (i = 0; i < len; i++) {
+		if (name[i] == '%' || name[i] == ':' || name[i] == '/' ||
+		    name[i] < 32 || name[i] >= 127) {
+			seq_printf(s, "%%%02x", name[i]);
+		} else {
+			seq_putc(s, name[i]);
+		}
+	}
+}
+
+static void dump_hoid(struct seq_file *s, const struct ceph_hobject_id *hoid)
+{
+	if (hoid->snapid == 0 && hoid->hash == 0 && !hoid->is_max &&
+	    hoid->pool == S64_MIN) {
+		seq_puts(s, "MIN");
+		return;
+	}
+	if (hoid->is_max) {
+		seq_puts(s, "MAX");
+		return;
+	}
+	seq_printf(s, "%lld:%08x:", hoid->pool, hoid->hash_reverse_bits);
+	dump_name_escaped(s, hoid->nspace, hoid->nspace_len);
+	seq_putc(s, ':');
+	dump_name_escaped(s, hoid->key, hoid->key_len);
+	seq_putc(s, ':');
+	dump_name_escaped(s, hoid->oid, hoid->oid_len);
+	seq_putc(s, ':');
+	dump_snapid(s, hoid->snapid);
+}
+
+static void dump_backoffs(struct seq_file *s, struct ceph_osd *osd)
+{
+	struct rb_node *n;
+
+	mutex_lock(&osd->lock);
+	for (n = rb_first(&osd->o_backoffs_by_id); n; n = rb_next(n)) {
+		struct ceph_osd_backoff *backoff =
+		    rb_entry(n, struct ceph_osd_backoff, id_node);
+
+		seq_printf(s, "osd%d\t", osd->o_osd);
+		dump_spgid(s, &backoff->spgid);
+		seq_printf(s, "\t%llu\t", backoff->id);
+		dump_hoid(s, backoff->begin);
+		seq_putc(s, '\t');
+		dump_hoid(s, backoff->end);
+		seq_putc(s, '\n');
+	}
+
+	mutex_unlock(&osd->lock);
+}
+
 static int osdc_show(struct seq_file *s, void *pp)
 {
 	struct ceph_client *client = s->private;
@@ -259,6 +358,13 @@ static int osdc_show(struct seq_file *s, void *pp)
 	}
 	dump_linger_requests(s, &osdc->homeless_osd);
 
+	seq_puts(s, "BACKOFFS\n");
+	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
+		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
+
+		dump_backoffs(s, osd);
+	}
+
 	up_read(&osdc->lock);
 	return 0;
 }
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index 588a91930051..0c31035bbfee 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -1288,13 +1288,16 @@ static void prepare_write_message(struct ceph_connection *con)
 		m->hdr.seq = cpu_to_le64(++con->out_seq);
 		m->needs_out_seq = false;
 	}
-	WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len));
+
+	if (con->ops->reencode_message)
+		con->ops->reencode_message(m);
 
 	dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n",
 	     m, con->out_seq, le16_to_cpu(m->hdr.type),
 	     le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len),
 	     m->data_length);
-	BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len);
+	WARN_ON(m->front.iov_len != le32_to_cpu(m->hdr.front_len));
+	WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len));
 
 	/* tag + hdr + front + middle */
 	con_out_kvec_add(con, sizeof (tag_msg), &tag_msg);
@@ -2033,8 +2036,7 @@ static int process_connect(struct ceph_connection *con)
 {
 	u64 sup_feat = from_msgr(con->msgr)->supported_features;
 	u64 req_feat = from_msgr(con->msgr)->required_features;
-	u64 server_feat = ceph_sanitize_features(
-				le64_to_cpu(con->in_reply.features));
+	u64 server_feat = le64_to_cpu(con->in_reply.features);
 	int ret;
 
 	dout("process_connect on %p tag %d\n", con, (int)con->in_tag);
diff --git a/net/ceph/mon_client.c b/net/ceph/mon_client.c
index 250f11f78609..875675765531 100644
--- a/net/ceph/mon_client.c
+++ b/net/ceph/mon_client.c
@@ -6,6 +6,7 @@
 #include <linux/random.h>
 #include <linux/sched.h>
 
+#include <linux/ceph/ceph_features.h>
 #include <linux/ceph/mon_client.h>
 #include <linux/ceph/libceph.h>
 #include <linux/ceph/debugfs.h>
@@ -297,6 +298,10 @@ static void handle_subscribe_ack(struct ceph_mon_client *monc,
 
 	mutex_lock(&monc->mutex);
 	if (monc->sub_renew_sent) {
+		/*
+		 * This is only needed for legacy (infernalis or older)
+		 * MONs -- see delayed_work().
+		 */
 		monc->sub_renew_after = monc->sub_renew_sent +
 					    (seconds >> 1) * HZ - 1;
 		dout("%s sent %lu duration %d renew after %lu\n", __func__,
@@ -955,7 +960,8 @@ static void delayed_work(struct work_struct *work)
 			__validate_auth(monc);
 		}
 
-		if (is_auth) {
+		if (is_auth &&
+		    !(monc->con.peer_features & CEPH_FEATURE_MON_STATEFUL_SUB)) {
 			unsigned long now = jiffies;
 
 			dout("%s renew subs? now %lu renew after %lu\n",
diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index 924f07c36ddb..86a9737d8e3f 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -12,6 +12,7 @@
 #include <linux/bio.h>
 #endif
 
+#include <linux/ceph/ceph_features.h>
 #include <linux/ceph/libceph.h>
 #include <linux/ceph/osd_client.h>
 #include <linux/ceph/messenger.h>
@@ -49,6 +50,7 @@ static void link_linger(struct ceph_osd *osd,
 			struct ceph_osd_linger_request *lreq);
 static void unlink_linger(struct ceph_osd *osd,
 			  struct ceph_osd_linger_request *lreq);
+static void clear_backoffs(struct ceph_osd *osd);
 
 #if 1
 static inline bool rwsem_is_wrlocked(struct rw_semaphore *sem)
@@ -373,6 +375,7 @@ static void target_copy(struct ceph_osd_request_target *dest,
 	ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
 
 	dest->pgid = src->pgid; /* struct */
+	dest->spgid = src->spgid; /* struct */
 	dest->pg_num = src->pg_num;
 	dest->pg_num_mask = src->pg_num_mask;
 	ceph_osds_copy(&dest->acting, &src->acting);
@@ -384,6 +387,9 @@ static void target_copy(struct ceph_osd_request_target *dest,
 	dest->flags = src->flags;
 	dest->paused = src->paused;
 
+	dest->epoch = src->epoch;
+	dest->last_force_resend = src->last_force_resend;
+
 	dest->osd = src->osd;
 }
 
@@ -537,7 +543,7 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
 }
 EXPORT_SYMBOL(ceph_osdc_alloc_request);
 
-static int ceph_oloc_encoding_size(struct ceph_object_locator *oloc)
+static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
 {
 	return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
 }
@@ -552,17 +558,21 @@ int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
 	WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
 
 	/* create request message */
-	msg_size = 4 + 4 + 4; /* client_inc, osdmap_epoch, flags */
-	msg_size += 4 + 4 + 4 + 8; /* mtime, reassert_version */
+	msg_size = CEPH_ENCODING_START_BLK_LEN +
+			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
+	msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
+	msg_size += CEPH_ENCODING_START_BLK_LEN +
+			sizeof(struct ceph_osd_reqid); /* reqid */
+	msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
+	msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
 	msg_size += CEPH_ENCODING_START_BLK_LEN +
 			ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
-	msg_size += 1 + 8 + 4 + 4; /* pgid */
 	msg_size += 4 + req->r_base_oid.name_len; /* oid */
 	msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
 	msg_size += 8; /* snapid */
 	msg_size += 8; /* snap_seq */
 	msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
-	msg_size += 4; /* retry_attempt */
+	msg_size += 4 + 8; /* retry_attempt, features */
 
 	if (req->r_mempool)
 		msg = ceph_msgpool_get(&osdc->msgpool_op, 0);
@@ -1010,6 +1020,8 @@ static void osd_init(struct ceph_osd *osd)
 	RB_CLEAR_NODE(&osd->o_node);
 	osd->o_requests = RB_ROOT;
 	osd->o_linger_requests = RB_ROOT;
+	osd->o_backoff_mappings = RB_ROOT;
+	osd->o_backoffs_by_id = RB_ROOT;
 	INIT_LIST_HEAD(&osd->o_osd_lru);
 	INIT_LIST_HEAD(&osd->o_keepalive_item);
 	osd->o_incarnation = 1;
@@ -1021,6 +1033,8 @@ static void osd_cleanup(struct ceph_osd *osd)
 	WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
+	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
+	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
 	WARN_ON(!list_empty(&osd->o_osd_lru));
 	WARN_ON(!list_empty(&osd->o_keepalive_item));
 
@@ -1141,6 +1155,7 @@ static void close_osd(struct ceph_osd *osd)
 		unlink_linger(osd, lreq);
 		link_linger(&osdc->homeless_osd, lreq);
 	}
+	clear_backoffs(osd);
 
 	__remove_osd_from_lru(osd);
 	erase_osd(&osdc->osds, osd);
@@ -1297,7 +1312,7 @@ static bool target_should_be_paused(struct ceph_osd_client *osdc,
 		       ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
 		       __pool_full(pi);
 
-	WARN_ON(pi->id != t->base_oloc.pool);
+	WARN_ON(pi->id != t->target_oloc.pool);
 	return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
 	       ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
 	       (osdc->osdmap->epoch < osdc->epoch_barrier);
@@ -1311,19 +1326,21 @@ enum calc_target_result {
 
 static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 					   struct ceph_osd_request_target *t,
-					   u32 *last_force_resend,
+					   struct ceph_connection *con,
 					   bool any_change)
 {
 	struct ceph_pg_pool_info *pi;
 	struct ceph_pg pgid, last_pgid;
 	struct ceph_osds up, acting;
 	bool force_resend = false;
-	bool need_check_tiering = false;
-	bool need_resend = false;
+	bool unpaused = false;
+	bool legacy_change;
+	bool split = false;
 	bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
 	enum calc_target_result ct_res;
 	int ret;
 
+	t->epoch = osdc->osdmap->epoch;
 	pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
 	if (!pi) {
 		t->osd = CEPH_HOMELESS_OSD;
@@ -1332,33 +1349,33 @@ static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 	}
 
 	if (osdc->osdmap->epoch == pi->last_force_request_resend) {
-		if (last_force_resend &&
-		    *last_force_resend < pi->last_force_request_resend) {
-			*last_force_resend = pi->last_force_request_resend;
+		if (t->last_force_resend < pi->last_force_request_resend) {
+			t->last_force_resend = pi->last_force_request_resend;
 			force_resend = true;
-		} else if (!last_force_resend) {
+		} else if (t->last_force_resend == 0) {
 			force_resend = true;
 		}
 	}
-	if (ceph_oid_empty(&t->target_oid) || force_resend) {
-		ceph_oid_copy(&t->target_oid, &t->base_oid);
-		need_check_tiering = true;
-	}
-	if (ceph_oloc_empty(&t->target_oloc) || force_resend) {
-		ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
-		need_check_tiering = true;
-	}
 
-	if (need_check_tiering &&
-	    (t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
+	/* apply tiering */
+	ceph_oid_copy(&t->target_oid, &t->base_oid);
+	ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
+	if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
 		if (t->flags & CEPH_OSD_FLAG_READ && pi->read_tier >= 0)
 			t->target_oloc.pool = pi->read_tier;
 		if (t->flags & CEPH_OSD_FLAG_WRITE && pi->write_tier >= 0)
 			t->target_oloc.pool = pi->write_tier;
+
+		pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
+		if (!pi) {
+			t->osd = CEPH_HOMELESS_OSD;
+			ct_res = CALC_TARGET_POOL_DNE;
+			goto out;
+		}
 	}
 
-	ret = ceph_object_locator_to_pg(osdc->osdmap, &t->target_oid,
-					&t->target_oloc, &pgid);
+	ret = __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc,
+					  &pgid);
 	if (ret) {
 		WARN_ON(ret != -ENOENT);
 		t->osd = CEPH_HOMELESS_OSD;
@@ -1368,7 +1385,7 @@ static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 	last_pgid.pool = pgid.pool;
 	last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
 
-	ceph_pg_to_up_acting_osds(osdc->osdmap, &pgid, &up, &acting);
+	ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
 	if (any_change &&
 	    ceph_is_new_interval(&t->acting,
 				 &acting,
@@ -1387,13 +1404,16 @@ static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 
 	if (t->paused && !target_should_be_paused(osdc, t, pi)) {
 		t->paused = false;
-		need_resend = true;
+		unpaused = true;
 	}
+	legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
+			ceph_osds_changed(&t->acting, &acting, any_change);
+	if (t->pg_num)
+		split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
 
-	if (ceph_pg_compare(&t->pgid, &pgid) ||
-	    ceph_osds_changed(&t->acting, &acting, any_change) ||
-	    force_resend) {
+	if (legacy_change || force_resend || split) {
 		t->pgid = pgid; /* struct */
+		ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
 		ceph_osds_copy(&t->acting, &acting);
 		ceph_osds_copy(&t->up, &up);
 		t->size = pi->size;
@@ -1403,15 +1423,342 @@ static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
 		t->sort_bitwise = sort_bitwise;
 
 		t->osd = acting.primary;
-		need_resend = true;
 	}
 
-	ct_res = need_resend ? CALC_TARGET_NEED_RESEND : CALC_TARGET_NO_ACTION;
+	if (unpaused || legacy_change || force_resend ||
+	    (split && con && CEPH_HAVE_FEATURE(con->peer_features,
+					       RESEND_ON_SPLIT)))
+		ct_res = CALC_TARGET_NEED_RESEND;
+	else
+		ct_res = CALC_TARGET_NO_ACTION;
+
 out:
 	dout("%s t %p -> ct_res %d osd %d\n", __func__, t, ct_res, t->osd);
 	return ct_res;
 }
 
+static struct ceph_spg_mapping *alloc_spg_mapping(void)
+{
+	struct ceph_spg_mapping *spg;
+
+	spg = kmalloc(sizeof(*spg), GFP_NOIO);
+	if (!spg)
+		return NULL;
+
+	RB_CLEAR_NODE(&spg->node);
+	spg->backoffs = RB_ROOT;
+	return spg;
+}
+
+static void free_spg_mapping(struct ceph_spg_mapping *spg)
+{
+	WARN_ON(!RB_EMPTY_NODE(&spg->node));
+	WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
+
+	kfree(spg);
+}
+
+/*
+ * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
+ * ceph_pg_mapping.  Used to track OSD backoffs -- a backoff [range] is
+ * defined only within a specific spgid; it does not pass anything to
+ * children on split, or to another primary.
+ */
+DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
+		 RB_BYPTR, const struct ceph_spg *, node)
+
+static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
+{
+	return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
+}
+
+static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
+				   void **pkey, size_t *pkey_len)
+{
+	if (hoid->key_len) {
+		*pkey = hoid->key;
+		*pkey_len = hoid->key_len;
+	} else {
+		*pkey = hoid->oid;
+		*pkey_len = hoid->oid_len;
+	}
+}
+
+static int compare_names(const void *name1, size_t name1_len,
+			 const void *name2, size_t name2_len)
+{
+	int ret;
+
+	ret = memcmp(name1, name2, min(name1_len, name2_len));
+	if (!ret) {
+		if (name1_len < name2_len)
+			ret = -1;
+		else if (name1_len > name2_len)
+			ret = 1;
+	}
+	return ret;
+}
+
+static int hoid_compare(const struct ceph_hobject_id *lhs,
+			const struct ceph_hobject_id *rhs)
+{
+	void *effective_key1, *effective_key2;
+	size_t effective_key1_len, effective_key2_len;
+	int ret;
+
+	if (lhs->is_max < rhs->is_max)
+		return -1;
+	if (lhs->is_max > rhs->is_max)
+		return 1;
+
+	if (lhs->pool < rhs->pool)
+		return -1;
+	if (lhs->pool > rhs->pool)
+		return 1;
+
+	if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
+		return -1;
+	if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
+		return 1;
+
+	ret = compare_names(lhs->nspace, lhs->nspace_len,
+			    rhs->nspace, rhs->nspace_len);
+	if (ret)
+		return ret;
+
+	hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
+	hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
+	ret = compare_names(effective_key1, effective_key1_len,
+			    effective_key2, effective_key2_len);
+	if (ret)
+		return ret;
+
+	ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
+	if (ret)
+		return ret;
+
+	if (lhs->snapid < rhs->snapid)
+		return -1;
+	if (lhs->snapid > rhs->snapid)
+		return 1;
+
+	return 0;
+}
+
+/*
+ * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
+ * compat stuff here.
+ *
+ * Assumes @hoid is zero-initialized.
+ */
+static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
+{
+	u8 struct_v;
+	u32 struct_len;
+	int ret;
+
+	ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
+				  &struct_len);
+	if (ret)
+		return ret;
+
+	if (struct_v < 4) {
+		pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
+		goto e_inval;
+	}
+
+	hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
+						GFP_NOIO);
+	if (IS_ERR(hoid->key)) {
+		ret = PTR_ERR(hoid->key);
+		hoid->key = NULL;
+		return ret;
+	}
+
+	hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
+						GFP_NOIO);
+	if (IS_ERR(hoid->oid)) {
+		ret = PTR_ERR(hoid->oid);
+		hoid->oid = NULL;
+		return ret;
+	}
+
+	ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
+	ceph_decode_32_safe(p, end, hoid->hash, e_inval);
+	ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
+
+	hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
+						   GFP_NOIO);
+	if (IS_ERR(hoid->nspace)) {
+		ret = PTR_ERR(hoid->nspace);
+		hoid->nspace = NULL;
+		return ret;
+	}
+
+	ceph_decode_64_safe(p, end, hoid->pool, e_inval);
+
+	ceph_hoid_build_hash_cache(hoid);
+	return 0;
+
+e_inval:
+	return -EINVAL;
+}
+
+static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
+{
+	return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
+	       4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
+}
+
+static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
+{
+	ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
+	ceph_encode_string(p, end, hoid->key, hoid->key_len);
+	ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
+	ceph_encode_64(p, hoid->snapid);
+	ceph_encode_32(p, hoid->hash);
+	ceph_encode_8(p, hoid->is_max);
+	ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
+	ceph_encode_64(p, hoid->pool);
+}
+
+static void free_hoid(struct ceph_hobject_id *hoid)
+{
+	if (hoid) {
+		kfree(hoid->key);
+		kfree(hoid->oid);
+		kfree(hoid->nspace);
+		kfree(hoid);
+	}
+}
+
+static struct ceph_osd_backoff *alloc_backoff(void)
+{
+	struct ceph_osd_backoff *backoff;
+
+	backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
+	if (!backoff)
+		return NULL;
+
+	RB_CLEAR_NODE(&backoff->spg_node);
+	RB_CLEAR_NODE(&backoff->id_node);
+	return backoff;
+}
+
+static void free_backoff(struct ceph_osd_backoff *backoff)
+{
+	WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
+	WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
+
+	free_hoid(backoff->begin);
+	free_hoid(backoff->end);
+	kfree(backoff);
+}
+
+/*
+ * Within a specific spgid, backoffs are managed by ->begin hoid.
+ */
+DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
+			RB_BYVAL, spg_node);
+
+static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
+					    const struct ceph_hobject_id *hoid)
+{
+	struct rb_node *n = root->rb_node;
+
+	while (n) {
+		struct ceph_osd_backoff *cur =
+		    rb_entry(n, struct ceph_osd_backoff, spg_node);
+		int cmp;
+
+		cmp = hoid_compare(hoid, cur->begin);
+		if (cmp < 0) {
+			n = n->rb_left;
+		} else if (cmp > 0) {
+			if (hoid_compare(hoid, cur->end) < 0)
+				return cur;
+
+			n = n->rb_right;
+		} else {
+			return cur;
+		}
+	}
+
+	return NULL;
+}
+
+/*
+ * Each backoff has a unique id within its OSD session.
+ */
+DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
+
+static void clear_backoffs(struct ceph_osd *osd)
+{
+	while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
+		struct ceph_spg_mapping *spg =
+		    rb_entry(rb_first(&osd->o_backoff_mappings),
+			     struct ceph_spg_mapping, node);
+
+		while (!RB_EMPTY_ROOT(&spg->backoffs)) {
+			struct ceph_osd_backoff *backoff =
+			    rb_entry(rb_first(&spg->backoffs),
+				     struct ceph_osd_backoff, spg_node);
+
+			erase_backoff(&spg->backoffs, backoff);
+			erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
+			free_backoff(backoff);
+		}
+		erase_spg_mapping(&osd->o_backoff_mappings, spg);
+		free_spg_mapping(spg);
+	}
+}
+
+/*
+ * Set up a temporary, non-owning view into @t.
+ */
+static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
+				  const struct ceph_osd_request_target *t)
+{
+	hoid->key = NULL;
+	hoid->key_len = 0;
+	hoid->oid = t->target_oid.name;
+	hoid->oid_len = t->target_oid.name_len;
+	hoid->snapid = CEPH_NOSNAP;
+	hoid->hash = t->pgid.seed;
+	hoid->is_max = false;
+	if (t->target_oloc.pool_ns) {
+		hoid->nspace = t->target_oloc.pool_ns->str;
+		hoid->nspace_len = t->target_oloc.pool_ns->len;
+	} else {
+		hoid->nspace = NULL;
+		hoid->nspace_len = 0;
+	}
+	hoid->pool = t->target_oloc.pool;
+	ceph_hoid_build_hash_cache(hoid);
+}
+
+static bool should_plug_request(struct ceph_osd_request *req)
+{
+	struct ceph_osd *osd = req->r_osd;
+	struct ceph_spg_mapping *spg;
+	struct ceph_osd_backoff *backoff;
+	struct ceph_hobject_id hoid;
+
+	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
+	if (!spg)
+		return false;
+
+	hoid_fill_from_target(&hoid, &req->r_t);
+	backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
+	if (!backoff)
+		return false;
+
+	dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
+	     __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
+	     backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
+	return true;
+}
+
 static void setup_request_data(struct ceph_osd_request *req,
 			       struct ceph_msg *msg)
 {
@@ -1483,7 +1830,37 @@ static void setup_request_data(struct ceph_osd_request *req,
 	WARN_ON(data_len != msg->data_length);
 }
 
-static void encode_request(struct ceph_osd_request *req, struct ceph_msg *msg)
+static void encode_pgid(void **p, const struct ceph_pg *pgid)
+{
+	ceph_encode_8(p, 1);
+	ceph_encode_64(p, pgid->pool);
+	ceph_encode_32(p, pgid->seed);
+	ceph_encode_32(p, -1); /* preferred */
+}
+
+static void encode_spgid(void **p, const struct ceph_spg *spgid)
+{
+	ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
+	encode_pgid(p, &spgid->pgid);
+	ceph_encode_8(p, spgid->shard);
+}
+
+static void encode_oloc(void **p, void *end,
+			const struct ceph_object_locator *oloc)
+{
+	ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
+	ceph_encode_64(p, oloc->pool);
+	ceph_encode_32(p, -1); /* preferred */
+	ceph_encode_32(p, 0);  /* key len */
+	if (oloc->pool_ns)
+		ceph_encode_string(p, end, oloc->pool_ns->str,
+				   oloc->pool_ns->len);
+	else
+		ceph_encode_32(p, 0);
+}
+
+static void encode_request_partial(struct ceph_osd_request *req,
+				   struct ceph_msg *msg)
 {
 	void *p = msg->front.iov_base;
 	void *const end = p + msg->front_alloc_len;
@@ -1500,38 +1877,27 @@ static void encode_request(struct ceph_osd_request *req, struct ceph_msg *msg)
 
 	setup_request_data(req, msg);
 
-	ceph_encode_32(&p, 1); /* client_inc, always 1 */
+	encode_spgid(&p, &req->r_t.spgid); /* actual spg */
+	ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
 	ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
 	ceph_encode_32(&p, req->r_flags);
-	ceph_encode_timespec(p, &req->r_mtime);
-	p += sizeof(struct ceph_timespec);
 
-	/* reassert_version */
-	memset(p, 0, sizeof(struct ceph_eversion));
-	p += sizeof(struct ceph_eversion);
-
-	/* oloc */
-	ceph_start_encoding(&p, 5, 4,
-			    ceph_oloc_encoding_size(&req->r_t.target_oloc));
-	ceph_encode_64(&p, req->r_t.target_oloc.pool);
-	ceph_encode_32(&p, -1); /* preferred */
-	ceph_encode_32(&p, 0); /* key len */
-	if (req->r_t.target_oloc.pool_ns)
-		ceph_encode_string(&p, end, req->r_t.target_oloc.pool_ns->str,
-				   req->r_t.target_oloc.pool_ns->len);
-	else
-		ceph_encode_32(&p, 0);
+	/* reqid */
+	ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
+	memset(p, 0, sizeof(struct ceph_osd_reqid));
+	p += sizeof(struct ceph_osd_reqid);
 
-	/* pgid */
-	ceph_encode_8(&p, 1);
-	ceph_encode_64(&p, req->r_t.pgid.pool);
-	ceph_encode_32(&p, req->r_t.pgid.seed);
-	ceph_encode_32(&p, -1); /* preferred */
+	/* trace */
+	memset(p, 0, sizeof(struct ceph_blkin_trace_info));
+	p += sizeof(struct ceph_blkin_trace_info);
+
+	ceph_encode_32(&p, 0); /* client_inc, always 0 */
+	ceph_encode_timespec(p, &req->r_mtime);
+	p += sizeof(struct ceph_timespec);
 
-	/* oid */
-	ceph_encode_32(&p, req->r_t.target_oid.name_len);
-	memcpy(p, req->r_t.target_oid.name, req->r_t.target_oid.name_len);
-	p += req->r_t.target_oid.name_len;
+	encode_oloc(&p, end, &req->r_t.target_oloc);
+	ceph_encode_string(&p, end, req->r_t.target_oid.name,
+			   req->r_t.target_oid.name_len);
 
 	/* ops, can imply data */
 	ceph_encode_16(&p, req->r_num_ops);
@@ -1552,11 +1918,10 @@ static void encode_request(struct ceph_osd_request *req, struct ceph_msg *msg)
 	}
 
 	ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
+	BUG_ON(p != end - 8); /* space for features */
 
-	BUG_ON(p > end);
-	msg->front.iov_len = p - msg->front.iov_base;
-	msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
-	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
+	msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
+	/* front_len is finalized in encode_request_finish() */
 	msg->hdr.data_len = cpu_to_le32(data_len);
 	/*
 	 * The header "data_off" is a hint to the receiver allowing it
@@ -1565,9 +1930,99 @@ static void encode_request(struct ceph_osd_request *req, struct ceph_msg *msg)
 	 */
 	msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
 
-	dout("%s req %p oid %s oid_len %d front %zu data %u\n", __func__,
-	     req, req->r_t.target_oid.name, req->r_t.target_oid.name_len,
-	     msg->front.iov_len, data_len);
+	dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
+	     req->r_t.target_oid.name, req->r_t.target_oid.name_len);
+}
+
+static void encode_request_finish(struct ceph_msg *msg)
+{
+	void *p = msg->front.iov_base;
+	void *const end = p + msg->front_alloc_len;
+
+	if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
+		/* luminous OSD -- encode features and be done */
+		p = end - 8;
+		ceph_encode_64(&p, msg->con->peer_features);
+	} else {
+		struct {
+			char spgid[CEPH_ENCODING_START_BLK_LEN +
+				   CEPH_PGID_ENCODING_LEN + 1];
+			__le32 hash;
+			__le32 epoch;
+			__le32 flags;
+			char reqid[CEPH_ENCODING_START_BLK_LEN +
+				   sizeof(struct ceph_osd_reqid)];
+			char trace[sizeof(struct ceph_blkin_trace_info)];
+			__le32 client_inc;
+			struct ceph_timespec mtime;
+		} __packed head;
+		struct ceph_pg pgid;
+		void *oloc, *oid, *tail;
+		int oloc_len, oid_len, tail_len;
+		int len;
+
+		/*
+		 * Pre-luminous OSD -- reencode v8 into v4 using @head
+		 * as a temporary buffer.  Encode the raw PG; the rest
+		 * is just a matter of moving oloc, oid and tail blobs
+		 * around.
+		 */
+		memcpy(&head, p, sizeof(head));
+		p += sizeof(head);
+
+		oloc = p;
+		p += CEPH_ENCODING_START_BLK_LEN;
+		pgid.pool = ceph_decode_64(&p);
+		p += 4 + 4; /* preferred, key len */
+		len = ceph_decode_32(&p);
+		p += len;   /* nspace */
+		oloc_len = p - oloc;
+
+		oid = p;
+		len = ceph_decode_32(&p);
+		p += len;
+		oid_len = p - oid;
+
+		tail = p;
+		tail_len = (end - p) - 8;
+
+		p = msg->front.iov_base;
+		ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
+		ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
+		ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
+		ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
+
+		/* reassert_version */
+		memset(p, 0, sizeof(struct ceph_eversion));
+		p += sizeof(struct ceph_eversion);
+
+		BUG_ON(p >= oloc);
+		memmove(p, oloc, oloc_len);
+		p += oloc_len;
+
+		pgid.seed = le32_to_cpu(head.hash);
+		encode_pgid(&p, &pgid); /* raw pg */
+
+		BUG_ON(p >= oid);
+		memmove(p, oid, oid_len);
+		p += oid_len;
+
+		/* tail -- ops, snapid, snapc, retry_attempt */
+		BUG_ON(p >= tail);
+		memmove(p, tail, tail_len);
+		p += tail_len;
+
+		msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
+	}
+
+	BUG_ON(p > end);
+	msg->front.iov_len = p - msg->front.iov_base;
+	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
+
+	dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
+	     le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
+	     le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
+	     le16_to_cpu(msg->hdr.version));
 }
 
 /*
@@ -1580,6 +2035,10 @@ static void send_request(struct ceph_osd_request *req)
 	verify_osd_locked(osd);
 	WARN_ON(osd->o_osd != req->r_t.osd);
 
+	/* backoff? */
+	if (should_plug_request(req))
+		return;
+
 	/*
 	 * We may have a previously queued request message hanging
 	 * around.  Cancel it to avoid corrupting the msgr.
@@ -1593,11 +2052,13 @@ static void send_request(struct ceph_osd_request *req)
 	else
 		WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
 
-	encode_request(req, req->r_request);
+	encode_request_partial(req, req->r_request);
 
-	dout("%s req %p tid %llu to pg %llu.%x osd%d flags 0x%x attempt %d\n",
+	dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
 	     __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
-	     req->r_t.osd, req->r_flags, req->r_attempts);
+	     req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
+	     req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
+	     req->r_attempts);
 
 	req->r_t.paused = false;
 	req->r_stamp = jiffies;
@@ -1645,7 +2106,7 @@ static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
 	dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
 
 again:
-	ct_res = calc_target(osdc, &req->r_t, &req->r_last_force_resend, false);
+	ct_res = calc_target(osdc, &req->r_t, NULL, false);
 	if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
 		goto promote;
 
@@ -1737,13 +2198,12 @@ static void submit_request(struct ceph_osd_request *req, bool wrlocked)
 static void finish_request(struct ceph_osd_request *req)
 {
 	struct ceph_osd_client *osdc = req->r_osdc;
-	struct ceph_osd *osd = req->r_osd;
 
-	verify_osd_locked(osd);
+	WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
 
-	WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
-	unlink_request(osd, req);
+	if (req->r_osd)
+		unlink_request(req->r_osd, req);
 	atomic_dec(&osdc->num_requests);
 
 	/*
@@ -2441,7 +2901,7 @@ static void linger_submit(struct ceph_osd_linger_request *lreq)
 	struct ceph_osd_client *osdc = lreq->osdc;
 	struct ceph_osd *osd;
 
-	calc_target(osdc, &lreq->t, &lreq->last_force_resend, false);
+	calc_target(osdc, &lreq->t, NULL, false);
 	osd = lookup_create_osd(osdc, lreq->t.osd, true);
 	link_linger(osd, lreq);
 
@@ -3059,7 +3519,7 @@ recalc_linger_target(struct ceph_osd_linger_request *lreq)
 	struct ceph_osd_client *osdc = lreq->osdc;
 	enum calc_target_result ct_res;
 
-	ct_res = calc_target(osdc, &lreq->t, &lreq->last_force_resend, true);
+	ct_res = calc_target(osdc, &lreq->t, NULL, true);
 	if (ct_res == CALC_TARGET_NEED_RESEND) {
 		struct ceph_osd *osd;
 
@@ -3117,6 +3577,7 @@ static void scan_requests(struct ceph_osd *osd,
 				list_add_tail(&lreq->scan_item, need_resend_linger);
 			break;
 		case CALC_TARGET_POOL_DNE:
+			list_del_init(&lreq->scan_item);
 			check_linger_pool_dne(lreq);
 			break;
 		}
@@ -3130,8 +3591,8 @@ static void scan_requests(struct ceph_osd *osd,
 		n = rb_next(n); /* unlink_request(), check_pool_dne() */
 
 		dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
-		ct_res = calc_target(osdc, &req->r_t,
-				     &req->r_last_force_resend, false);
+		ct_res = calc_target(osdc, &req->r_t, &req->r_osd->o_con,
+				     false);
 		switch (ct_res) {
 		case CALC_TARGET_NO_ACTION:
 			force_resend_writes = cleared_full ||
@@ -3229,8 +3690,25 @@ static void kick_requests(struct ceph_osd_client *osdc,
 			  struct list_head *need_resend_linger)
 {
 	struct ceph_osd_linger_request *lreq, *nlreq;
+	enum calc_target_result ct_res;
 	struct rb_node *n;
 
+	/* make sure need_resend targets reflect latest map */
+	for (n = rb_first(need_resend); n; ) {
+		struct ceph_osd_request *req =
+		    rb_entry(n, struct ceph_osd_request, r_node);
+
+		n = rb_next(n);
+
+		if (req->r_t.epoch < osdc->osdmap->epoch) {
+			ct_res = calc_target(osdc, &req->r_t, NULL, false);
+			if (ct_res == CALC_TARGET_POOL_DNE) {
+				erase_request(need_resend, req);
+				check_pool_dne(req);
+			}
+		}
+	}
+
 	for (n = rb_first(need_resend); n; ) {
 		struct ceph_osd_request *req =
 		    rb_entry(n, struct ceph_osd_request, r_node);
@@ -3239,8 +3717,6 @@ static void kick_requests(struct ceph_osd_client *osdc,
 		n = rb_next(n);
 		erase_request(need_resend, req); /* before link_request() */
 
-		WARN_ON(req->r_osd);
-		calc_target(osdc, &req->r_t, NULL, false);
 		osd = lookup_create_osd(osdc, req->r_t.osd, true);
 		link_request(osd, req);
 		if (!req->r_linger) {
@@ -3383,6 +3859,8 @@ static void kick_osd_requests(struct ceph_osd *osd)
 {
 	struct rb_node *n;
 
+	clear_backoffs(osd);
+
 	for (n = rb_first(&osd->o_requests); n; ) {
 		struct ceph_osd_request *req =
 		    rb_entry(n, struct ceph_osd_request, r_node);
@@ -3428,6 +3906,261 @@ out_unlock:
 	up_write(&osdc->lock);
 }
 
+struct MOSDBackoff {
+	struct ceph_spg spgid;
+	u32 map_epoch;
+	u8 op;
+	u64 id;
+	struct ceph_hobject_id *begin;
+	struct ceph_hobject_id *end;
+};
+
+static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
+{
+	void *p = msg->front.iov_base;
+	void *const end = p + msg->front.iov_len;
+	u8 struct_v;
+	u32 struct_len;
+	int ret;
+
+	ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
+	if (ret)
+		return ret;
+
+	ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
+	if (ret)
+		return ret;
+
+	ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
+	ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
+	ceph_decode_8_safe(&p, end, m->op, e_inval);
+	ceph_decode_64_safe(&p, end, m->id, e_inval);
+
+	m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
+	if (!m->begin)
+		return -ENOMEM;
+
+	ret = decode_hoid(&p, end, m->begin);
+	if (ret) {
+		free_hoid(m->begin);
+		return ret;
+	}
+
+	m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
+	if (!m->end) {
+		free_hoid(m->begin);
+		return -ENOMEM;
+	}
+
+	ret = decode_hoid(&p, end, m->end);
+	if (ret) {
+		free_hoid(m->begin);
+		free_hoid(m->end);
+		return ret;
+	}
+
+	return 0;
+
+e_inval:
+	return -EINVAL;
+}
+
+static struct ceph_msg *create_backoff_message(
+				const struct ceph_osd_backoff *backoff,
+				u32 map_epoch)
+{
+	struct ceph_msg *msg;
+	void *p, *end;
+	int msg_size;
+
+	msg_size = CEPH_ENCODING_START_BLK_LEN +
+			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
+	msg_size += 4 + 1 + 8; /* map_epoch, op, id */
+	msg_size += CEPH_ENCODING_START_BLK_LEN +
+			hoid_encoding_size(backoff->begin);
+	msg_size += CEPH_ENCODING_START_BLK_LEN +
+			hoid_encoding_size(backoff->end);
+
+	msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
+	if (!msg)
+		return NULL;
+
+	p = msg->front.iov_base;
+	end = p + msg->front_alloc_len;
+
+	encode_spgid(&p, &backoff->spgid);
+	ceph_encode_32(&p, map_epoch);
+	ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
+	ceph_encode_64(&p, backoff->id);
+	encode_hoid(&p, end, backoff->begin);
+	encode_hoid(&p, end, backoff->end);
+	BUG_ON(p != end);
+
+	msg->front.iov_len = p - msg->front.iov_base;
+	msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
+	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
+
+	return msg;
+}
+
+static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
+{
+	struct ceph_spg_mapping *spg;
+	struct ceph_osd_backoff *backoff;
+	struct ceph_msg *msg;
+
+	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
+	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
+
+	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
+	if (!spg) {
+		spg = alloc_spg_mapping();
+		if (!spg) {
+			pr_err("%s failed to allocate spg\n", __func__);
+			return;
+		}
+		spg->spgid = m->spgid; /* struct */
+		insert_spg_mapping(&osd->o_backoff_mappings, spg);
+	}
+
+	backoff = alloc_backoff();
+	if (!backoff) {
+		pr_err("%s failed to allocate backoff\n", __func__);
+		return;
+	}
+	backoff->spgid = m->spgid; /* struct */
+	backoff->id = m->id;
+	backoff->begin = m->begin;
+	m->begin = NULL; /* backoff now owns this */
+	backoff->end = m->end;
+	m->end = NULL;   /* ditto */
+
+	insert_backoff(&spg->backoffs, backoff);
+	insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
+
+	/*
+	 * Ack with original backoff's epoch so that the OSD can
+	 * discard this if there was a PG split.
+	 */
+	msg = create_backoff_message(backoff, m->map_epoch);
+	if (!msg) {
+		pr_err("%s failed to allocate msg\n", __func__);
+		return;
+	}
+	ceph_con_send(&osd->o_con, msg);
+}
+
+static bool target_contained_by(const struct ceph_osd_request_target *t,
+				const struct ceph_hobject_id *begin,
+				const struct ceph_hobject_id *end)
+{
+	struct ceph_hobject_id hoid;
+	int cmp;
+
+	hoid_fill_from_target(&hoid, t);
+	cmp = hoid_compare(&hoid, begin);
+	return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
+}
+
+static void handle_backoff_unblock(struct ceph_osd *osd,
+				   const struct MOSDBackoff *m)
+{
+	struct ceph_spg_mapping *spg;
+	struct ceph_osd_backoff *backoff;
+	struct rb_node *n;
+
+	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
+	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
+
+	backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
+	if (!backoff) {
+		pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
+		       __func__, osd->o_osd, m->spgid.pgid.pool,
+		       m->spgid.pgid.seed, m->spgid.shard, m->id);
+		return;
+	}
+
+	if (hoid_compare(backoff->begin, m->begin) &&
+	    hoid_compare(backoff->end, m->end)) {
+		pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
+		       __func__, osd->o_osd, m->spgid.pgid.pool,
+		       m->spgid.pgid.seed, m->spgid.shard, m->id);
+		/* unblock it anyway... */
+	}
+
+	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
+	BUG_ON(!spg);
+
+	erase_backoff(&spg->backoffs, backoff);
+	erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
+	free_backoff(backoff);
+
+	if (RB_EMPTY_ROOT(&spg->backoffs)) {
+		erase_spg_mapping(&osd->o_backoff_mappings, spg);
+		free_spg_mapping(spg);
+	}
+
+	for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
+		struct ceph_osd_request *req =
+		    rb_entry(n, struct ceph_osd_request, r_node);
+
+		if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
+			/*
+			 * Match against @m, not @backoff -- the PG may
+			 * have split on the OSD.
+			 */
+			if (target_contained_by(&req->r_t, m->begin, m->end)) {
+				/*
+				 * If no other installed backoff applies,
+				 * resend.
+				 */
+				send_request(req);
+			}
+		}
+	}
+}
+
+static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
+{
+	struct ceph_osd_client *osdc = osd->o_osdc;
+	struct MOSDBackoff m;
+	int ret;
+
+	down_read(&osdc->lock);
+	if (!osd_registered(osd)) {
+		dout("%s osd%d unknown\n", __func__, osd->o_osd);
+		up_read(&osdc->lock);
+		return;
+	}
+	WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
+
+	mutex_lock(&osd->lock);
+	ret = decode_MOSDBackoff(msg, &m);
+	if (ret) {
+		pr_err("failed to decode MOSDBackoff: %d\n", ret);
+		ceph_msg_dump(msg);
+		goto out_unlock;
+	}
+
+	switch (m.op) {
+	case CEPH_OSD_BACKOFF_OP_BLOCK:
+		handle_backoff_block(osd, &m);
+		break;
+	case CEPH_OSD_BACKOFF_OP_UNBLOCK:
+		handle_backoff_unblock(osd, &m);
+		break;
+	default:
+		pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
+	}
+
+	free_hoid(m.begin);
+	free_hoid(m.end);
+
+out_unlock:
+	mutex_unlock(&osd->lock);
+	up_read(&osdc->lock);
+}
+
 /*
  * Process osd watch notifications
  */
@@ -4365,6 +5098,9 @@ static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
 	case CEPH_MSG_OSD_OPREPLY:
 		handle_reply(osd, msg);
 		break;
+	case CEPH_MSG_OSD_BACKOFF:
+		handle_backoff(osd, msg);
+		break;
 	case CEPH_MSG_WATCH_NOTIFY:
 		handle_watch_notify(osdc, msg);
 		break;
@@ -4487,6 +5223,7 @@ static struct ceph_msg *alloc_msg(struct ceph_connection *con,
 	*skip = 0;
 	switch (type) {
 	case CEPH_MSG_OSD_MAP:
+	case CEPH_MSG_OSD_BACKOFF:
 	case CEPH_MSG_WATCH_NOTIFY:
 		return alloc_msg_with_page_vector(hdr);
 	case CEPH_MSG_OSD_OPREPLY:
@@ -4571,6 +5308,11 @@ static int invalidate_authorizer(struct ceph_connection *con)
 	return ceph_monc_validate_auth(&osdc->client->monc);
 }
 
+static void osd_reencode_message(struct ceph_msg *msg)
+{
+	encode_request_finish(msg);
+}
+
 static int osd_sign_message(struct ceph_msg *msg)
 {
 	struct ceph_osd *o = msg->con->private;
@@ -4595,6 +5337,7 @@ static const struct ceph_connection_operations osd_con_ops = {
 	.verify_authorizer_reply = verify_authorizer_reply,
 	.invalidate_authorizer = invalidate_authorizer,
 	.alloc_msg = alloc_msg,
+	.reencode_message = osd_reencode_message,
 	.sign_message = osd_sign_message,
 	.check_message_signature = osd_check_message_signature,
 	.fault = osd_fault,
diff --git a/net/ceph/osdmap.c b/net/ceph/osdmap.c
index 55e3a477f92d..864789c5974e 100644
--- a/net/ceph/osdmap.c
+++ b/net/ceph/osdmap.c
@@ -11,7 +11,7 @@
 #include <linux/crush/hash.h>
 #include <linux/crush/mapper.h>
 
-char *ceph_osdmap_state_str(char *str, int len, int state)
+char *ceph_osdmap_state_str(char *str, int len, u32 state)
 {
 	if (!len)
 		return str;
@@ -138,19 +138,175 @@ bad:
 	return -EINVAL;
 }
 
-static int skip_name_map(void **p, void *end)
+static struct crush_choose_arg_map *alloc_choose_arg_map(void)
 {
-        int len;
-        ceph_decode_32_safe(p, end, len ,bad);
-        while (len--) {
-                int strlen;
-                *p += sizeof(u32);
-                ceph_decode_32_safe(p, end, strlen, bad);
-                *p += strlen;
+	struct crush_choose_arg_map *arg_map;
+
+	arg_map = kzalloc(sizeof(*arg_map), GFP_NOIO);
+	if (!arg_map)
+		return NULL;
+
+	RB_CLEAR_NODE(&arg_map->node);
+	return arg_map;
 }
-        return 0;
-bad:
-        return -EINVAL;
+
+static void free_choose_arg_map(struct crush_choose_arg_map *arg_map)
+{
+	if (arg_map) {
+		int i, j;
+
+		WARN_ON(!RB_EMPTY_NODE(&arg_map->node));
+
+		for (i = 0; i < arg_map->size; i++) {
+			struct crush_choose_arg *arg = &arg_map->args[i];
+
+			for (j = 0; j < arg->weight_set_size; j++)
+				kfree(arg->weight_set[j].weights);
+			kfree(arg->weight_set);
+			kfree(arg->ids);
+		}
+		kfree(arg_map->args);
+		kfree(arg_map);
+	}
+}
+
+DEFINE_RB_FUNCS(choose_arg_map, struct crush_choose_arg_map, choose_args_index,
+		node);
+
+void clear_choose_args(struct crush_map *c)
+{
+	while (!RB_EMPTY_ROOT(&c->choose_args)) {
+		struct crush_choose_arg_map *arg_map =
+		    rb_entry(rb_first(&c->choose_args),
+			     struct crush_choose_arg_map, node);
+
+		erase_choose_arg_map(&c->choose_args, arg_map);
+		free_choose_arg_map(arg_map);
+	}
+}
+
+static u32 *decode_array_32_alloc(void **p, void *end, u32 *plen)
+{
+	u32 *a = NULL;
+	u32 len;
+	int ret;
+
+	ceph_decode_32_safe(p, end, len, e_inval);
+	if (len) {
+		u32 i;
+
+		a = kmalloc_array(len, sizeof(u32), GFP_NOIO);
+		if (!a) {
+			ret = -ENOMEM;
+			goto fail;
+		}
+
+		ceph_decode_need(p, end, len * sizeof(u32), e_inval);
+		for (i = 0; i < len; i++)
+			a[i] = ceph_decode_32(p);
+	}
+
+	*plen = len;
+	return a;
+
+e_inval:
+	ret = -EINVAL;
+fail:
+	kfree(a);
+	return ERR_PTR(ret);
+}
+
+/*
+ * Assumes @arg is zero-initialized.
+ */
+static int decode_choose_arg(void **p, void *end, struct crush_choose_arg *arg)
+{
+	int ret;
+
+	ceph_decode_32_safe(p, end, arg->weight_set_size, e_inval);
+	if (arg->weight_set_size) {
+		u32 i;
+
+		arg->weight_set = kmalloc_array(arg->weight_set_size,
+						sizeof(*arg->weight_set),
+						GFP_NOIO);
+		if (!arg->weight_set)
+			return -ENOMEM;
+
+		for (i = 0; i < arg->weight_set_size; i++) {
+			struct crush_weight_set *w = &arg->weight_set[i];
+
+			w->weights = decode_array_32_alloc(p, end, &w->size);
+			if (IS_ERR(w->weights)) {
+				ret = PTR_ERR(w->weights);
+				w->weights = NULL;
+				return ret;
+			}
+		}
+	}
+
+	arg->ids = decode_array_32_alloc(p, end, &arg->ids_size);
+	if (IS_ERR(arg->ids)) {
+		ret = PTR_ERR(arg->ids);
+		arg->ids = NULL;
+		return ret;
+	}
+
+	return 0;
+
+e_inval:
+	return -EINVAL;
+}
+
+static int decode_choose_args(void **p, void *end, struct crush_map *c)
+{
+	struct crush_choose_arg_map *arg_map = NULL;
+	u32 num_choose_arg_maps, num_buckets;
+	int ret;
+
+	ceph_decode_32_safe(p, end, num_choose_arg_maps, e_inval);
+	while (num_choose_arg_maps--) {
+		arg_map = alloc_choose_arg_map();
+		if (!arg_map) {
+			ret = -ENOMEM;
+			goto fail;
+		}
+
+		ceph_decode_64_safe(p, end, arg_map->choose_args_index,
+				    e_inval);
+		arg_map->size = c->max_buckets;
+		arg_map->args = kcalloc(arg_map->size, sizeof(*arg_map->args),
+					GFP_NOIO);
+		if (!arg_map->args) {
+			ret = -ENOMEM;
+			goto fail;
+		}
+
+		ceph_decode_32_safe(p, end, num_buckets, e_inval);
+		while (num_buckets--) {
+			struct crush_choose_arg *arg;
+			u32 bucket_index;
+
+			ceph_decode_32_safe(p, end, bucket_index, e_inval);
+			if (bucket_index >= arg_map->size)
+				goto e_inval;
+
+			arg = &arg_map->args[bucket_index];
+			ret = decode_choose_arg(p, end, arg);
+			if (ret)
+				goto fail;
+		}
+
+		insert_choose_arg_map(&c->choose_args, arg_map);
+	}
+
+	return 0;
+
+e_inval:
+	ret = -EINVAL;
+fail:
+	free_choose_arg_map(arg_map);
+	return ret;
 }
 
 static void crush_finalize(struct crush_map *c)
@@ -187,7 +343,6 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 	void **p = &pbyval;
 	void *start = pbyval;
 	u32 magic;
-	u32 num_name_maps;
 
 	dout("crush_decode %p to %p len %d\n", *p, end, (int)(end - *p));
 
@@ -195,6 +350,8 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 	if (c == NULL)
 		return ERR_PTR(-ENOMEM);
 
+	c->choose_args = RB_ROOT;
+
         /* set tunables to default values */
         c->choose_local_tries = 2;
         c->choose_local_fallback_tries = 5;
@@ -353,12 +510,9 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 		}
 	}
 
-	/* ignore trailing name maps. */
-        for (num_name_maps = 0; num_name_maps < 3; num_name_maps++) {
-                err = skip_name_map(p, end);
-                if (err < 0)
-                        goto done;
-        }
+	ceph_decode_skip_map(p, end, 32, string, bad); /* type_map */
+	ceph_decode_skip_map(p, end, 32, string, bad); /* name_map */
+	ceph_decode_skip_map(p, end, 32, string, bad); /* rule_name_map */
 
         /* tunables */
         ceph_decode_need(p, end, 3*sizeof(u32), done);
@@ -391,6 +545,21 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 	dout("crush decode tunable chooseleaf_stable = %d\n",
 	     c->chooseleaf_stable);
 
+	if (*p != end) {
+		/* class_map */
+		ceph_decode_skip_map(p, end, 32, 32, bad);
+		/* class_name */
+		ceph_decode_skip_map(p, end, 32, string, bad);
+		/* class_bucket */
+		ceph_decode_skip_map_of_map(p, end, 32, 32, 32, bad);
+	}
+
+	if (*p != end) {
+		err = decode_choose_args(p, end, c);
+		if (err)
+			goto bad;
+	}
+
 done:
 	crush_finalize(c);
 	dout("crush_decode success\n");
@@ -418,75 +587,49 @@ int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs)
 	return 0;
 }
 
-/*
- * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid
- * to a set of osds) and primary_temp (explicit primary setting)
- */
-static int __insert_pg_mapping(struct ceph_pg_mapping *new,
-			       struct rb_root *root)
+int ceph_spg_compare(const struct ceph_spg *lhs, const struct ceph_spg *rhs)
 {
-	struct rb_node **p = &root->rb_node;
-	struct rb_node *parent = NULL;
-	struct ceph_pg_mapping *pg = NULL;
-	int c;
+	int ret;
 
-	dout("__insert_pg_mapping %llx %p\n", *(u64 *)&new->pgid, new);
-	while (*p) {
-		parent = *p;
-		pg = rb_entry(parent, struct ceph_pg_mapping, node);
-		c = ceph_pg_compare(&new->pgid, &pg->pgid);
-		if (c < 0)
-			p = &(*p)->rb_left;
-		else if (c > 0)
-			p = &(*p)->rb_right;
-		else
-			return -EEXIST;
-	}
+	ret = ceph_pg_compare(&lhs->pgid, &rhs->pgid);
+	if (ret)
+		return ret;
+
+	if (lhs->shard < rhs->shard)
+		return -1;
+	if (lhs->shard > rhs->shard)
+		return 1;
 
-	rb_link_node(&new->node, parent, p);
-	rb_insert_color(&new->node, root);
 	return 0;
 }
 
-static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
-						   struct ceph_pg pgid)
+static struct ceph_pg_mapping *alloc_pg_mapping(size_t payload_len)
 {
-	struct rb_node *n = root->rb_node;
 	struct ceph_pg_mapping *pg;
-	int c;
 
-	while (n) {
-		pg = rb_entry(n, struct ceph_pg_mapping, node);
-		c = ceph_pg_compare(&pgid, &pg->pgid);
-		if (c < 0) {
-			n = n->rb_left;
-		} else if (c > 0) {
-			n = n->rb_right;
-		} else {
-			dout("__lookup_pg_mapping %lld.%x got %p\n",
-			     pgid.pool, pgid.seed, pg);
-			return pg;
-		}
-	}
-	return NULL;
+	pg = kmalloc(sizeof(*pg) + payload_len, GFP_NOIO);
+	if (!pg)
+		return NULL;
+
+	RB_CLEAR_NODE(&pg->node);
+	return pg;
 }
 
-static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg pgid)
+static void free_pg_mapping(struct ceph_pg_mapping *pg)
 {
-	struct ceph_pg_mapping *pg = __lookup_pg_mapping(root, pgid);
+	WARN_ON(!RB_EMPTY_NODE(&pg->node));
 
-	if (pg) {
-		dout("__remove_pg_mapping %lld.%x %p\n", pgid.pool, pgid.seed,
-		     pg);
-		rb_erase(&pg->node, root);
-		kfree(pg);
-		return 0;
-	}
-	dout("__remove_pg_mapping %lld.%x dne\n", pgid.pool, pgid.seed);
-	return -ENOENT;
+	kfree(pg);
 }
 
 /*
+ * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid
+ * to a set of osds) and primary_temp (explicit primary setting)
+ */
+DEFINE_RB_FUNCS2(pg_mapping, struct ceph_pg_mapping, pgid, ceph_pg_compare,
+		 RB_BYPTR, const struct ceph_pg *, node)
+
+/*
  * rbtree of pg pool info
  */
 static int __insert_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *new)
@@ -682,11 +825,48 @@ static int decode_pool(void **p, void *end, struct ceph_pg_pool_info *pi)
 		*p += len;
 	}
 
+	/*
+	 * last_force_op_resend_preluminous, will be overridden if the
+	 * map was encoded with RESEND_ON_SPLIT
+	 */
 	if (ev >= 15)
 		pi->last_force_request_resend = ceph_decode_32(p);
 	else
 		pi->last_force_request_resend = 0;
 
+	if (ev >= 16)
+		*p += 4; /* skip min_read_recency_for_promote */
+
+	if (ev >= 17)
+		*p += 8; /* skip expected_num_objects */
+
+	if (ev >= 19)
+		*p += 4; /* skip cache_target_dirty_high_ratio_micro */
+
+	if (ev >= 20)
+		*p += 4; /* skip min_write_recency_for_promote */
+
+	if (ev >= 21)
+		*p += 1; /* skip use_gmt_hitset */
+
+	if (ev >= 22)
+		*p += 1; /* skip fast_read */
+
+	if (ev >= 23) {
+		*p += 4; /* skip hit_set_grade_decay_rate */
+		*p += 4; /* skip hit_set_search_last_n */
+	}
+
+	if (ev >= 24) {
+		/* skip opts */
+		*p += 1 + 1; /* versions */
+		len = ceph_decode_32(p);
+		*p += len;
+	}
+
+	if (ev >= 25)
+		pi->last_force_request_resend = ceph_decode_32(p);
+
 	/* ignore the rest */
 
 	*p = pool_end;
@@ -743,6 +923,8 @@ struct ceph_osdmap *ceph_osdmap_alloc(void)
 	map->pool_max = -1;
 	map->pg_temp = RB_ROOT;
 	map->primary_temp = RB_ROOT;
+	map->pg_upmap = RB_ROOT;
+	map->pg_upmap_items = RB_ROOT;
 	mutex_init(&map->crush_workspace_mutex);
 
 	return map;
@@ -757,14 +939,28 @@ void ceph_osdmap_destroy(struct ceph_osdmap *map)
 		struct ceph_pg_mapping *pg =
 			rb_entry(rb_first(&map->pg_temp),
 				 struct ceph_pg_mapping, node);
-		rb_erase(&pg->node, &map->pg_temp);
-		kfree(pg);
+		erase_pg_mapping(&map->pg_temp, pg);
+		free_pg_mapping(pg);
 	}
 	while (!RB_EMPTY_ROOT(&map->primary_temp)) {
 		struct ceph_pg_mapping *pg =
 			rb_entry(rb_first(&map->primary_temp),
 				 struct ceph_pg_mapping, node);
-		rb_erase(&pg->node, &map->primary_temp);
+		erase_pg_mapping(&map->primary_temp, pg);
+		free_pg_mapping(pg);
+	}
+	while (!RB_EMPTY_ROOT(&map->pg_upmap)) {
+		struct ceph_pg_mapping *pg =
+			rb_entry(rb_first(&map->pg_upmap),
+				 struct ceph_pg_mapping, node);
+		rb_erase(&pg->node, &map->pg_upmap);
+		kfree(pg);
+	}
+	while (!RB_EMPTY_ROOT(&map->pg_upmap_items)) {
+		struct ceph_pg_mapping *pg =
+			rb_entry(rb_first(&map->pg_upmap_items),
+				 struct ceph_pg_mapping, node);
+		rb_erase(&pg->node, &map->pg_upmap_items);
 		kfree(pg);
 	}
 	while (!RB_EMPTY_ROOT(&map->pg_pools)) {
@@ -788,7 +984,7 @@ void ceph_osdmap_destroy(struct ceph_osdmap *map)
  */
 static int osdmap_set_max_osd(struct ceph_osdmap *map, int max)
 {
-	u8 *state;
+	u32 *state;
 	u32 *weight;
 	struct ceph_entity_addr *addr;
 	int i;
@@ -964,47 +1160,40 @@ static int decode_new_pools(void **p, void *end, struct ceph_osdmap *map)
 	return __decode_pools(p, end, map, true);
 }
 
-static int __decode_pg_temp(void **p, void *end, struct ceph_osdmap *map,
-			    bool incremental)
+typedef struct ceph_pg_mapping *(*decode_mapping_fn_t)(void **, void *, bool);
+
+static int decode_pg_mapping(void **p, void *end, struct rb_root *mapping_root,
+			     decode_mapping_fn_t fn, bool incremental)
 {
 	u32 n;
 
+	WARN_ON(!incremental && !fn);
+
 	ceph_decode_32_safe(p, end, n, e_inval);
 	while (n--) {
+		struct ceph_pg_mapping *pg;
 		struct ceph_pg pgid;
-		u32 len, i;
 		int ret;
 
 		ret = ceph_decode_pgid(p, end, &pgid);
 		if (ret)
 			return ret;
 
-		ceph_decode_32_safe(p, end, len, e_inval);
-
-		ret = __remove_pg_mapping(&map->pg_temp, pgid);
-		BUG_ON(!incremental && ret != -ENOENT);
-
-		if (!incremental || len > 0) {
-			struct ceph_pg_mapping *pg;
-
-			ceph_decode_need(p, end, len*sizeof(u32), e_inval);
-
-			if (len > (UINT_MAX - sizeof(*pg)) / sizeof(u32))
-				return -EINVAL;
-
-			pg = kzalloc(sizeof(*pg) + len*sizeof(u32), GFP_NOFS);
-			if (!pg)
-				return -ENOMEM;
+		pg = lookup_pg_mapping(mapping_root, &pgid);
+		if (pg) {
+			WARN_ON(!incremental);
+			erase_pg_mapping(mapping_root, pg);
+			free_pg_mapping(pg);
+		}
 
-			pg->pgid = pgid;
-			pg->pg_temp.len = len;
-			for (i = 0; i < len; i++)
-				pg->pg_temp.osds[i] = ceph_decode_32(p);
+		if (fn) {
+			pg = fn(p, end, incremental);
+			if (IS_ERR(pg))
+				return PTR_ERR(pg);
 
-			ret = __insert_pg_mapping(pg, &map->pg_temp);
-			if (ret) {
-				kfree(pg);
-				return ret;
+			if (pg) {
+				pg->pgid = pgid; /* struct */
+				insert_pg_mapping(mapping_root, pg);
 			}
 		}
 	}
@@ -1015,69 +1204,77 @@ e_inval:
 	return -EINVAL;
 }
 
+static struct ceph_pg_mapping *__decode_pg_temp(void **p, void *end,
+						bool incremental)
+{
+	struct ceph_pg_mapping *pg;
+	u32 len, i;
+
+	ceph_decode_32_safe(p, end, len, e_inval);
+	if (len == 0 && incremental)
+		return NULL;	/* new_pg_temp: [] to remove */
+	if (len > (SIZE_MAX - sizeof(*pg)) / sizeof(u32))
+		return ERR_PTR(-EINVAL);
+
+	ceph_decode_need(p, end, len * sizeof(u32), e_inval);
+	pg = alloc_pg_mapping(len * sizeof(u32));
+	if (!pg)
+		return ERR_PTR(-ENOMEM);
+
+	pg->pg_temp.len = len;
+	for (i = 0; i < len; i++)
+		pg->pg_temp.osds[i] = ceph_decode_32(p);
+
+	return pg;
+
+e_inval:
+	return ERR_PTR(-EINVAL);
+}
+
 static int decode_pg_temp(void **p, void *end, struct ceph_osdmap *map)
 {
-	return __decode_pg_temp(p, end, map, false);
+	return decode_pg_mapping(p, end, &map->pg_temp, __decode_pg_temp,
+				 false);
 }
 
 static int decode_new_pg_temp(void **p, void *end, struct ceph_osdmap *map)
 {
-	return __decode_pg_temp(p, end, map, true);
+	return decode_pg_mapping(p, end, &map->pg_temp, __decode_pg_temp,
+				 true);
 }
 
-static int __decode_primary_temp(void **p, void *end, struct ceph_osdmap *map,
-				 bool incremental)
+static struct ceph_pg_mapping *__decode_primary_temp(void **p, void *end,
+						     bool incremental)
 {
-	u32 n;
-
-	ceph_decode_32_safe(p, end, n, e_inval);
-	while (n--) {
-		struct ceph_pg pgid;
-		u32 osd;
-		int ret;
-
-		ret = ceph_decode_pgid(p, end, &pgid);
-		if (ret)
-			return ret;
-
-		ceph_decode_32_safe(p, end, osd, e_inval);
-
-		ret = __remove_pg_mapping(&map->primary_temp, pgid);
-		BUG_ON(!incremental && ret != -ENOENT);
-
-		if (!incremental || osd != (u32)-1) {
-			struct ceph_pg_mapping *pg;
-
-			pg = kzalloc(sizeof(*pg), GFP_NOFS);
-			if (!pg)
-				return -ENOMEM;
+	struct ceph_pg_mapping *pg;
+	u32 osd;
 
-			pg->pgid = pgid;
-			pg->primary_temp.osd = osd;
+	ceph_decode_32_safe(p, end, osd, e_inval);
+	if (osd == (u32)-1 && incremental)
+		return NULL;	/* new_primary_temp: -1 to remove */
 
-			ret = __insert_pg_mapping(pg, &map->primary_temp);
-			if (ret) {
-				kfree(pg);
-				return ret;
-			}
-		}
-	}
+	pg = alloc_pg_mapping(0);
+	if (!pg)
+		return ERR_PTR(-ENOMEM);
 
-	return 0;
+	pg->primary_temp.osd = osd;
+	return pg;
 
 e_inval:
-	return -EINVAL;
+	return ERR_PTR(-EINVAL);
 }
 
 static int decode_primary_temp(void **p, void *end, struct ceph_osdmap *map)
 {
-	return __decode_primary_temp(p, end, map, false);
+	return decode_pg_mapping(p, end, &map->primary_temp,
+				 __decode_primary_temp, false);
 }
 
 static int decode_new_primary_temp(void **p, void *end,
 				   struct ceph_osdmap *map)
 {
-	return __decode_primary_temp(p, end, map, true);
+	return decode_pg_mapping(p, end, &map->primary_temp,
+				 __decode_primary_temp, true);
 }
 
 u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd)
@@ -1168,6 +1365,75 @@ e_inval:
 	return -EINVAL;
 }
 
+static struct ceph_pg_mapping *__decode_pg_upmap(void **p, void *end,
+						 bool __unused)
+{
+	return __decode_pg_temp(p, end, false);
+}
+
+static int decode_pg_upmap(void **p, void *end, struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap, __decode_pg_upmap,
+				 false);
+}
+
+static int decode_new_pg_upmap(void **p, void *end, struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap, __decode_pg_upmap,
+				 true);
+}
+
+static int decode_old_pg_upmap(void **p, void *end, struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap, NULL, true);
+}
+
+static struct ceph_pg_mapping *__decode_pg_upmap_items(void **p, void *end,
+						       bool __unused)
+{
+	struct ceph_pg_mapping *pg;
+	u32 len, i;
+
+	ceph_decode_32_safe(p, end, len, e_inval);
+	if (len > (SIZE_MAX - sizeof(*pg)) / (2 * sizeof(u32)))
+		return ERR_PTR(-EINVAL);
+
+	ceph_decode_need(p, end, 2 * len * sizeof(u32), e_inval);
+	pg = kzalloc(sizeof(*pg) + 2 * len * sizeof(u32), GFP_NOIO);
+	if (!pg)
+		return ERR_PTR(-ENOMEM);
+
+	pg->pg_upmap_items.len = len;
+	for (i = 0; i < len; i++) {
+		pg->pg_upmap_items.from_to[i][0] = ceph_decode_32(p);
+		pg->pg_upmap_items.from_to[i][1] = ceph_decode_32(p);
+	}
+
+	return pg;
+
+e_inval:
+	return ERR_PTR(-EINVAL);
+}
+
+static int decode_pg_upmap_items(void **p, void *end, struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap_items,
+				 __decode_pg_upmap_items, false);
+}
+
+static int decode_new_pg_upmap_items(void **p, void *end,
+				     struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap_items,
+				 __decode_pg_upmap_items, true);
+}
+
+static int decode_old_pg_upmap_items(void **p, void *end,
+				     struct ceph_osdmap *map)
+{
+	return decode_pg_mapping(p, end, &map->pg_upmap_items, NULL, true);
+}
+
 /*
  * decode a full map.
  */
@@ -1218,13 +1484,21 @@ static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map)
 
 	/* osd_state, osd_weight, osd_addrs->client_addr */
 	ceph_decode_need(p, end, 3*sizeof(u32) +
-			 map->max_osd*(1 + sizeof(*map->osd_weight) +
+			 map->max_osd*((struct_v >= 5 ? sizeof(u32) :
+							sizeof(u8)) +
+				       sizeof(*map->osd_weight) +
 				       sizeof(*map->osd_addr)), e_inval);
 
 	if (ceph_decode_32(p) != map->max_osd)
 		goto e_inval;
 
-	ceph_decode_copy(p, map->osd_state, map->max_osd);
+	if (struct_v >= 5) {
+		for (i = 0; i < map->max_osd; i++)
+			map->osd_state[i] = ceph_decode_32(p);
+	} else {
+		for (i = 0; i < map->max_osd; i++)
+			map->osd_state[i] = ceph_decode_8(p);
+	}
 
 	if (ceph_decode_32(p) != map->max_osd)
 		goto e_inval;
@@ -1257,9 +1531,7 @@ static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map)
 		if (err)
 			goto bad;
 	} else {
-		/* XXX can this happen? */
-		kfree(map->osd_primary_affinity);
-		map->osd_primary_affinity = NULL;
+		WARN_ON(map->osd_primary_affinity);
 	}
 
 	/* crush */
@@ -1268,6 +1540,26 @@ static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map)
 	if (err)
 		goto bad;
 
+	*p += len;
+	if (struct_v >= 3) {
+		/* erasure_code_profiles */
+		ceph_decode_skip_map_of_map(p, end, string, string, string,
+					    bad);
+	}
+
+	if (struct_v >= 4) {
+		err = decode_pg_upmap(p, end, map);
+		if (err)
+			goto bad;
+
+		err = decode_pg_upmap_items(p, end, map);
+		if (err)
+			goto bad;
+	} else {
+		WARN_ON(!RB_EMPTY_ROOT(&map->pg_upmap));
+		WARN_ON(!RB_EMPTY_ROOT(&map->pg_upmap_items));
+	}
+
 	/* ignore the rest */
 	*p = end;
 
@@ -1314,7 +1606,7 @@ struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end)
  *     new_up_client: { osd=6, addr=... } # set osd_state and addr
  *     new_state: { osd=6, xorstate=EXISTS } # clear osd_state
  */
-static int decode_new_up_state_weight(void **p, void *end,
+static int decode_new_up_state_weight(void **p, void *end, u8 struct_v,
 				      struct ceph_osdmap *map)
 {
 	void *new_up_client;
@@ -1330,7 +1622,7 @@ static int decode_new_up_state_weight(void **p, void *end,
 
 	new_state = *p;
 	ceph_decode_32_safe(p, end, len, e_inval);
-	len *= sizeof(u32) + sizeof(u8);
+	len *= sizeof(u32) + (struct_v >= 5 ? sizeof(u32) : sizeof(u8));
 	ceph_decode_need(p, end, len, e_inval);
 	*p += len;
 
@@ -1366,11 +1658,14 @@ static int decode_new_up_state_weight(void **p, void *end,
 	len = ceph_decode_32(p);
 	while (len--) {
 		s32 osd;
-		u8 xorstate;
+		u32 xorstate;
 		int ret;
 
 		osd = ceph_decode_32(p);
-		xorstate = ceph_decode_8(p);
+		if (struct_v >= 5)
+			xorstate = ceph_decode_32(p);
+		else
+			xorstate = ceph_decode_8(p);
 		if (xorstate == 0)
 			xorstate = CEPH_OSD_UP;
 		BUG_ON(osd >= map->max_osd);
@@ -1504,7 +1799,7 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 	}
 
 	/* new_up_client, new_state, new_weight */
-	err = decode_new_up_state_weight(p, end, map);
+	err = decode_new_up_state_weight(p, end, struct_v, map);
 	if (err)
 		goto bad;
 
@@ -1527,6 +1822,32 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 			goto bad;
 	}
 
+	if (struct_v >= 3) {
+		/* new_erasure_code_profiles */
+		ceph_decode_skip_map_of_map(p, end, string, string, string,
+					    bad);
+		/* old_erasure_code_profiles */
+		ceph_decode_skip_set(p, end, string, bad);
+	}
+
+	if (struct_v >= 4) {
+		err = decode_new_pg_upmap(p, end, map);
+		if (err)
+			goto bad;
+
+		err = decode_old_pg_upmap(p, end, map);
+		if (err)
+			goto bad;
+
+		err = decode_new_pg_upmap_items(p, end, map);
+		if (err)
+			goto bad;
+
+		err = decode_old_pg_upmap_items(p, end, map);
+		if (err)
+			goto bad;
+	}
+
 	/* ignore the rest */
 	*p = end;
 
@@ -1547,12 +1868,13 @@ bad:
 void ceph_oloc_copy(struct ceph_object_locator *dest,
 		    const struct ceph_object_locator *src)
 {
-	WARN_ON(!ceph_oloc_empty(dest));
-	WARN_ON(dest->pool_ns); /* empty() only covers ->pool */
+	ceph_oloc_destroy(dest);
 
 	dest->pool = src->pool;
 	if (src->pool_ns)
 		dest->pool_ns = ceph_get_string(src->pool_ns);
+	else
+		dest->pool_ns = NULL;
 }
 EXPORT_SYMBOL(ceph_oloc_copy);
 
@@ -1565,14 +1887,15 @@ EXPORT_SYMBOL(ceph_oloc_destroy);
 void ceph_oid_copy(struct ceph_object_id *dest,
 		   const struct ceph_object_id *src)
 {
-	WARN_ON(!ceph_oid_empty(dest));
+	ceph_oid_destroy(dest);
 
 	if (src->name != src->inline_name) {
 		/* very rare, see ceph_object_id definition */
 		dest->name = kmalloc(src->name_len + 1,
 				     GFP_NOIO | __GFP_NOFAIL);
+	} else {
+		dest->name = dest->inline_name;
 	}
-
 	memcpy(dest->name, src->name, src->name_len + 1);
 	dest->name_len = src->name_len;
 }
@@ -1714,9 +2037,8 @@ void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src)
 	dest->primary = src->primary;
 }
 
-static bool is_split(const struct ceph_pg *pgid,
-		     u32 old_pg_num,
-		     u32 new_pg_num)
+bool ceph_pg_is_split(const struct ceph_pg *pgid, u32 old_pg_num,
+		      u32 new_pg_num)
 {
 	int old_bits = calc_bits_of(old_pg_num);
 	int old_mask = (1 << old_bits) - 1;
@@ -1761,7 +2083,7 @@ bool ceph_is_new_interval(const struct ceph_osds *old_acting,
 	       !osds_equal(old_up, new_up) ||
 	       old_size != new_size ||
 	       old_min_size != new_min_size ||
-	       is_split(pgid, old_pg_num, new_pg_num) ||
+	       ceph_pg_is_split(pgid, old_pg_num, new_pg_num) ||
 	       old_sort_bitwise != new_sort_bitwise;
 }
 
@@ -1885,16 +2207,12 @@ EXPORT_SYMBOL(ceph_calc_file_object_mapping);
  * Should only be called with target_oid and target_oloc (as opposed to
  * base_oid and base_oloc), since tiering isn't taken into account.
  */
-int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
-			      struct ceph_object_id *oid,
-			      struct ceph_object_locator *oloc,
-			      struct ceph_pg *raw_pgid)
+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)
 {
-	struct ceph_pg_pool_info *pi;
-
-	pi = ceph_pg_pool_by_id(osdmap, oloc->pool);
-	if (!pi)
-		return -ENOENT;
+	WARN_ON(pi->id != oloc->pool);
 
 	if (!oloc->pool_ns) {
 		raw_pgid->pool = oloc->pool;
@@ -1926,6 +2244,20 @@ int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
 	}
 	return 0;
 }
+
+int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
+			      const struct ceph_object_id *oid,
+			      const struct ceph_object_locator *oloc,
+			      struct ceph_pg *raw_pgid)
+{
+	struct ceph_pg_pool_info *pi;
+
+	pi = ceph_pg_pool_by_id(osdmap, oloc->pool);
+	if (!pi)
+		return -ENOENT;
+
+	return __ceph_object_locator_to_pg(pi, oid, oloc, raw_pgid);
+}
 EXPORT_SYMBOL(ceph_object_locator_to_pg);
 
 /*
@@ -1970,23 +2302,57 @@ static u32 raw_pg_to_pps(struct ceph_pg_pool_info *pi,
 
 static int do_crush(struct ceph_osdmap *map, int ruleno, int x,
 		    int *result, int result_max,
-		    const __u32 *weight, int weight_max)
+		    const __u32 *weight, int weight_max,
+		    u64 choose_args_index)
 {
+	struct crush_choose_arg_map *arg_map;
 	int r;
 
 	BUG_ON(result_max > CEPH_PG_MAX_SIZE);
 
+	arg_map = lookup_choose_arg_map(&map->crush->choose_args,
+					choose_args_index);
+
 	mutex_lock(&map->crush_workspace_mutex);
 	r = crush_do_rule(map->crush, ruleno, x, result, result_max,
-			  weight, weight_max, map->crush_workspace);
+			  weight, weight_max, map->crush_workspace,
+			  arg_map ? arg_map->args : NULL);
 	mutex_unlock(&map->crush_workspace_mutex);
 
 	return r;
 }
 
+static void remove_nonexistent_osds(struct ceph_osdmap *osdmap,
+				    struct ceph_pg_pool_info *pi,
+				    struct ceph_osds *set)
+{
+	int i;
+
+	if (ceph_can_shift_osds(pi)) {
+		int removed = 0;
+
+		/* shift left */
+		for (i = 0; i < set->size; i++) {
+			if (!ceph_osd_exists(osdmap, set->osds[i])) {
+				removed++;
+				continue;
+			}
+			if (removed)
+				set->osds[i - removed] = set->osds[i];
+		}
+		set->size -= removed;
+	} else {
+		/* set dne devices to NONE */
+		for (i = 0; i < set->size; i++) {
+			if (!ceph_osd_exists(osdmap, set->osds[i]))
+				set->osds[i] = CRUSH_ITEM_NONE;
+		}
+	}
+}
+
 /*
- * Calculate raw set (CRUSH output) for given PG.  The result may
- * contain nonexistent OSDs.  ->primary is undefined for a raw set.
+ * Calculate raw set (CRUSH output) for given PG and filter out
+ * nonexistent OSDs.  ->primary is undefined for a raw set.
  *
  * Placement seed (CRUSH input) is returned through @ppps.
  */
@@ -2020,7 +2386,7 @@ static void pg_to_raw_osds(struct ceph_osdmap *osdmap,
 	}
 
 	len = do_crush(osdmap, ruleno, pps, raw->osds, pi->size,
-		       osdmap->osd_weight, osdmap->max_osd);
+		       osdmap->osd_weight, osdmap->max_osd, pi->id);
 	if (len < 0) {
 		pr_err("error %d from crush rule %d: pool %lld ruleset %d type %d size %d\n",
 		       len, ruleno, pi->id, pi->crush_ruleset, pi->type,
@@ -2029,6 +2395,70 @@ static void pg_to_raw_osds(struct ceph_osdmap *osdmap,
 	}
 
 	raw->size = len;
+	remove_nonexistent_osds(osdmap, pi, raw);
+}
+
+/* apply pg_upmap[_items] mappings */
+static void apply_upmap(struct ceph_osdmap *osdmap,
+			const struct ceph_pg *pgid,
+			struct ceph_osds *raw)
+{
+	struct ceph_pg_mapping *pg;
+	int i, j;
+
+	pg = lookup_pg_mapping(&osdmap->pg_upmap, pgid);
+	if (pg) {
+		/* make sure targets aren't marked out */
+		for (i = 0; i < pg->pg_upmap.len; i++) {
+			int osd = pg->pg_upmap.osds[i];
+
+			if (osd != CRUSH_ITEM_NONE &&
+			    osd < osdmap->max_osd &&
+			    osdmap->osd_weight[osd] == 0) {
+				/* reject/ignore explicit mapping */
+				return;
+			}
+		}
+		for (i = 0; i < pg->pg_upmap.len; i++)
+			raw->osds[i] = pg->pg_upmap.osds[i];
+		raw->size = pg->pg_upmap.len;
+		return;
+	}
+
+	pg = lookup_pg_mapping(&osdmap->pg_upmap_items, pgid);
+	if (pg) {
+		/*
+		 * Note: this approach does not allow a bidirectional swap,
+		 * e.g., [[1,2],[2,1]] applied to [0,1,2] -> [0,2,1].
+		 */
+		for (i = 0; i < pg->pg_upmap_items.len; i++) {
+			int from = pg->pg_upmap_items.from_to[i][0];
+			int to = pg->pg_upmap_items.from_to[i][1];
+			int pos = -1;
+			bool exists = false;
+
+			/* make sure replacement doesn't already appear */
+			for (j = 0; j < raw->size; j++) {
+				int osd = raw->osds[j];
+
+				if (osd == to) {
+					exists = true;
+					break;
+				}
+				/* ignore mapping if target is marked out */
+				if (osd == from && pos < 0 &&
+				    !(to != CRUSH_ITEM_NONE &&
+				      to < osdmap->max_osd &&
+				      osdmap->osd_weight[to] == 0)) {
+					pos = j;
+				}
+			}
+			if (!exists && pos >= 0) {
+				raw->osds[pos] = to;
+				return;
+			}
+		}
+	}
 }
 
 /*
@@ -2151,18 +2581,16 @@ static void apply_primary_affinity(struct ceph_osdmap *osdmap,
  */
 static void get_temp_osds(struct ceph_osdmap *osdmap,
 			  struct ceph_pg_pool_info *pi,
-			  const struct ceph_pg *raw_pgid,
+			  const struct ceph_pg *pgid,
 			  struct ceph_osds *temp)
 {
-	struct ceph_pg pgid;
 	struct ceph_pg_mapping *pg;
 	int i;
 
-	raw_pg_to_pg(pi, raw_pgid, &pgid);
 	ceph_osds_init(temp);
 
 	/* pg_temp? */
-	pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid);
+	pg = lookup_pg_mapping(&osdmap->pg_temp, pgid);
 	if (pg) {
 		for (i = 0; i < pg->pg_temp.len; i++) {
 			if (ceph_osd_is_down(osdmap, pg->pg_temp.osds[i])) {
@@ -2185,7 +2613,7 @@ static void get_temp_osds(struct ceph_osdmap *osdmap,
 	}
 
 	/* primary_temp? */
-	pg = __lookup_pg_mapping(&osdmap->primary_temp, pgid);
+	pg = lookup_pg_mapping(&osdmap->primary_temp, pgid);
 	if (pg)
 		temp->primary = pg->primary_temp.osd;
 }
@@ -2198,43 +2626,75 @@ static void get_temp_osds(struct ceph_osdmap *osdmap,
  * resend a request.
  */
 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)
 {
-	struct ceph_pg_pool_info *pi;
+	struct ceph_pg pgid;
 	u32 pps;
 
-	pi = ceph_pg_pool_by_id(osdmap, raw_pgid->pool);
-	if (!pi) {
-		ceph_osds_init(up);
-		ceph_osds_init(acting);
-		goto out;
-	}
+	WARN_ON(pi->id != raw_pgid->pool);
+	raw_pg_to_pg(pi, raw_pgid, &pgid);
 
 	pg_to_raw_osds(osdmap, pi, raw_pgid, up, &pps);
+	apply_upmap(osdmap, &pgid, up);
 	raw_to_up_osds(osdmap, pi, up);
 	apply_primary_affinity(osdmap, pi, pps, up);
-	get_temp_osds(osdmap, pi, raw_pgid, acting);
+	get_temp_osds(osdmap, pi, &pgid, acting);
 	if (!acting->size) {
 		memcpy(acting->osds, up->osds, up->size * sizeof(up->osds[0]));
 		acting->size = up->size;
 		if (acting->primary == -1)
 			acting->primary = up->primary;
 	}
-out:
 	WARN_ON(!osds_valid(up) || !osds_valid(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)
+{
+	struct ceph_pg pgid;
+	struct ceph_osds up, acting;
+	int i;
+
+	WARN_ON(pi->id != raw_pgid->pool);
+	raw_pg_to_pg(pi, raw_pgid, &pgid);
+
+	if (ceph_can_shift_osds(pi)) {
+		spgid->pgid = pgid; /* struct */
+		spgid->shard = CEPH_SPG_NOSHARD;
+		return true;
+	}
+
+	ceph_pg_to_up_acting_osds(osdmap, pi, &pgid, &up, &acting);
+	for (i = 0; i < acting.size; i++) {
+		if (acting.osds[i] == acting.primary) {
+			spgid->pgid = pgid; /* struct */
+			spgid->shard = i;
+			return true;
+		}
+	}
+
+	return false;
+}
+
 /*
  * Return acting primary for given PG, or -1 if none.
  */
 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
 			      const struct ceph_pg *raw_pgid)
 {
+	struct ceph_pg_pool_info *pi;
 	struct ceph_osds up, acting;
 
-	ceph_pg_to_up_acting_osds(osdmap, raw_pgid, &up, &acting);
+	pi = ceph_pg_pool_by_id(osdmap, raw_pgid->pool);
+	if (!pi)
+		return -1;
+
+	ceph_pg_to_up_acting_osds(osdmap, pi, raw_pgid, &up, &acting);
 	return acting.primary;
 }
 EXPORT_SYMBOL(ceph_pg_to_acting_primary);
diff --git a/net/compat.c b/net/compat.c
index aba929e5250f..6ded6c821d7a 100644
--- a/net/compat.c
+++ b/net/compat.c
@@ -37,21 +37,16 @@ int get_compat_msghdr(struct msghdr *kmsg,
 		      struct sockaddr __user **save_addr,
 		      struct iovec **iov)
 {
-	compat_uptr_t uaddr, uiov, tmp3;
-	compat_size_t nr_segs;
+	struct compat_msghdr msg;
 	ssize_t err;
 
-	if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
-	    __get_user(uaddr, &umsg->msg_name) ||
-	    __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
-	    __get_user(uiov, &umsg->msg_iov) ||
-	    __get_user(nr_segs, &umsg->msg_iovlen) ||
-	    __get_user(tmp3, &umsg->msg_control) ||
-	    __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
-	    __get_user(kmsg->msg_flags, &umsg->msg_flags))
+	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
 		return -EFAULT;
 
-	if (!uaddr)
+	kmsg->msg_flags = msg.msg_flags;
+	kmsg->msg_namelen = msg.msg_namelen;
+
+	if (!msg.msg_name)
 		kmsg->msg_namelen = 0;
 
 	if (kmsg->msg_namelen < 0)
@@ -59,14 +54,16 @@ int get_compat_msghdr(struct msghdr *kmsg,
 
 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
-	kmsg->msg_control = compat_ptr(tmp3);
+
+	kmsg->msg_control = compat_ptr(msg.msg_control);
+	kmsg->msg_controllen = msg.msg_controllen;
 
 	if (save_addr)
-		*save_addr = compat_ptr(uaddr);
+		*save_addr = compat_ptr(msg.msg_name);
 
-	if (uaddr && kmsg->msg_namelen) {
+	if (msg.msg_name && kmsg->msg_namelen) {
 		if (!save_addr) {
-			err = move_addr_to_kernel(compat_ptr(uaddr),
+			err = move_addr_to_kernel(compat_ptr(msg.msg_name),
 						  kmsg->msg_namelen,
 						  kmsg->msg_name);
 			if (err < 0)
@@ -77,13 +74,13 @@ int get_compat_msghdr(struct msghdr *kmsg,
 		kmsg->msg_namelen = 0;
 	}
 
-	if (nr_segs > UIO_MAXIOV)
+	if (msg.msg_iovlen > UIO_MAXIOV)
 		return -EMSGSIZE;
 
 	kmsg->msg_iocb = NULL;
 
 	return compat_import_iovec(save_addr ? READ : WRITE,
-				   compat_ptr(uiov), nr_segs,
+				   compat_ptr(msg.msg_iov), msg.msg_iovlen,
 				   UIO_FASTIOV, iov, &kmsg->msg_iter);
 }
 
@@ -316,15 +313,15 @@ struct sock_fprog __user *get_compat_bpf_fprog(char __user *optval)
 {
 	struct compat_sock_fprog __user *fprog32 = (struct compat_sock_fprog __user *)optval;
 	struct sock_fprog __user *kfprog = compat_alloc_user_space(sizeof(struct sock_fprog));
-	compat_uptr_t ptr;
-	u16 len;
-
-	if (!access_ok(VERIFY_READ, fprog32, sizeof(*fprog32)) ||
-	    !access_ok(VERIFY_WRITE, kfprog, sizeof(struct sock_fprog)) ||
-	    __get_user(len, &fprog32->len) ||
-	    __get_user(ptr, &fprog32->filter) ||
-	    __put_user(len, &kfprog->len) ||
-	    __put_user(compat_ptr(ptr), &kfprog->filter))
+	struct compat_sock_fprog f32;
+	struct sock_fprog f;
+
+	if (copy_from_user(&f32, fprog32, sizeof(*fprog32)))
+		return NULL;
+	memset(&f, 0, sizeof(f));
+	f.len = f32.len;
+	f.filter = compat_ptr(f32.filter);
+	if (copy_to_user(kfprog, &f, sizeof(struct sock_fprog)))
 		return NULL;
 
 	return kfprog;
diff --git a/net/core/datagram.c b/net/core/datagram.c
index 6877c43cc92d..ee5647bd91b3 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -203,7 +203,7 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
 /**
  *	__skb_try_recv_datagram - Receive a datagram skbuff
  *	@sk: socket
- *	@flags: MSG_ flags
+ *	@flags: MSG\_ flags
  *	@destructor: invoked under the receive lock on successful dequeue
  *	@peeked: returns non-zero if this packet has been seen before
  *	@off: an offset in bytes to peek skb from. Returns an offset
@@ -375,7 +375,7 @@ EXPORT_SYMBOL(__sk_queue_drop_skb);
  *	skb_kill_datagram - Free a datagram skbuff forcibly
  *	@sk: socket
  *	@skb: datagram skbuff
- *	@flags: MSG_ flags
+ *	@flags: MSG\_ flags
  *
  *	This function frees a datagram skbuff that was received by
  *	skb_recv_datagram.  The flags argument must match the one
@@ -809,7 +809,7 @@ EXPORT_SYMBOL(skb_copy_and_csum_datagram_msg);
  *	sequenced packet sockets providing the socket receive queue
  *	is only ever holding data ready to receive.
  *
- *	Note: when you _don't_ use this routine for this protocol,
+ *	Note: when you *don't* use this routine for this protocol,
  *	and you use a different write policy from sock_writeable()
  *	then please supply your own write_space callback.
  */
diff --git a/net/core/dev.c b/net/core/dev.c
index 02440518dd69..8515f8fe0460 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -7384,7 +7384,7 @@ static int netif_alloc_rx_queues(struct net_device *dev)
 
 	BUG_ON(count < 1);
 
-	rx = kvzalloc(sz, GFP_KERNEL | __GFP_REPEAT);
+	rx = kvzalloc(sz, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
 	if (!rx)
 		return -ENOMEM;
 
@@ -7424,7 +7424,7 @@ static int netif_alloc_netdev_queues(struct net_device *dev)
 	if (count < 1 || count > 0xffff)
 		return -EINVAL;
 
-	tx = kvzalloc(sz, GFP_KERNEL | __GFP_REPEAT);
+	tx = kvzalloc(sz, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
 	if (!tx)
 		return -ENOMEM;
 
@@ -7965,7 +7965,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	/* ensure 32-byte alignment of whole construct */
 	alloc_size += NETDEV_ALIGN - 1;
 
-	p = kvzalloc(alloc_size, GFP_KERNEL | __GFP_REPEAT);
+	p = kvzalloc(alloc_size, GFP_KERNEL | __GFP_RETRY_MAYFAIL);
 	if (!p)
 		return NULL;
 
diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index e31fc11a8000..d0713627deb6 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -347,8 +347,7 @@ out_entries:
 
 static void neigh_get_hash_rnd(u32 *x)
 {
-	get_random_bytes(x, sizeof(*x));
-	*x |= 1;
+	*x = get_random_u32() | 1;
 }
 
 static struct neigh_hash_table *neigh_hash_alloc(unsigned int shift)
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 8b11341ed69a..f990eb8b30a9 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -4747,7 +4747,7 @@ struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
 
 	gfp_head = gfp_mask;
 	if (gfp_head & __GFP_DIRECT_RECLAIM)
-		gfp_head |= __GFP_REPEAT;
+		gfp_head |= __GFP_RETRY_MAYFAIL;
 
 	*errcode = -ENOBUFS;
 	skb = alloc_skb(header_len, gfp_head);
diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index bb909f1d7537..06863ea3fc5b 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -2431,8 +2431,8 @@ static int ipmr_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
 	tableid = tb[RTA_TABLE] ? nla_get_u32(tb[RTA_TABLE]) : 0;
 
 	mrt = ipmr_get_table(net, tableid ? tableid : RT_TABLE_DEFAULT);
-	if (IS_ERR(mrt)) {
-		err = PTR_ERR(mrt);
+	if (!mrt) {
+		err = -ENOENT;
 		goto errout_free;
 	}
 
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index c816cd53f7fc..0383e66f59bc 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -2979,8 +2979,7 @@ static __net_init int rt_genid_init(struct net *net)
 {
 	atomic_set(&net->ipv4.rt_genid, 0);
 	atomic_set(&net->fnhe_genid, 0);
-	get_random_bytes(&net->ipv4.dev_addr_genid,
-			 sizeof(net->ipv4.dev_addr_genid));
+	atomic_set(&net->ipv4.dev_addr_genid, get_random_int());
 	return 0;
 }
 
diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index 1770c1d9b37f..e1648238a9c9 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -1003,14 +1003,10 @@ struct xt_table_info *xt_alloc_table_info(unsigned int size)
 	if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
 		return NULL;
 
-	if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
-		info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
-	if (!info) {
-		info = __vmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY,
-				 PAGE_KERNEL);
-		if (!info)
-			return NULL;
-	}
+	info = kvmalloc(sz, GFP_KERNEL);
+	if (!info)
+		return NULL;
+
 	memset(info, 0, sizeof(*info));
 	info->size = size;
 	return info;
diff --git a/net/sched/sch_fq.c b/net/sched/sch_fq.c
index 147fde73a0f5..263d16e3219e 100644
--- a/net/sched/sch_fq.c
+++ b/net/sched/sch_fq.c
@@ -648,7 +648,7 @@ static int fq_resize(struct Qdisc *sch, u32 log)
 		return 0;
 
 	/* If XPS was setup, we can allocate memory on right NUMA node */
-	array = kvmalloc_node(sizeof(struct rb_root) << log, GFP_KERNEL | __GFP_REPEAT,
+	array = kvmalloc_node(sizeof(struct rb_root) << log, GFP_KERNEL | __GFP_RETRY_MAYFAIL,
 			      netdev_queue_numa_node_read(sch->dev_queue));
 	if (!array)
 		return -ENOMEM;
diff --git a/net/socket.c b/net/socket.c
index 59e902b9df09..bf2122691fba 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -1910,22 +1910,18 @@ static int copy_msghdr_from_user(struct msghdr *kmsg,
 				 struct sockaddr __user **save_addr,
 				 struct iovec **iov)
 {
-	struct sockaddr __user *uaddr;
-	struct iovec __user *uiov;
-	size_t nr_segs;
+	struct user_msghdr msg;
 	ssize_t err;
 
-	if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
-	    __get_user(uaddr, &umsg->msg_name) ||
-	    __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
-	    __get_user(uiov, &umsg->msg_iov) ||
-	    __get_user(nr_segs, &umsg->msg_iovlen) ||
-	    __get_user(kmsg->msg_control, &umsg->msg_control) ||
-	    __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
-	    __get_user(kmsg->msg_flags, &umsg->msg_flags))
+	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
 		return -EFAULT;
 
-	if (!uaddr)
+	kmsg->msg_control = msg.msg_control;
+	kmsg->msg_controllen = msg.msg_controllen;
+	kmsg->msg_flags = msg.msg_flags;
+
+	kmsg->msg_namelen = msg.msg_namelen;
+	if (!msg.msg_name)
 		kmsg->msg_namelen = 0;
 
 	if (kmsg->msg_namelen < 0)
@@ -1935,11 +1931,11 @@ static int copy_msghdr_from_user(struct msghdr *kmsg,
 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
 
 	if (save_addr)
-		*save_addr = uaddr;
+		*save_addr = msg.msg_name;
 
-	if (uaddr && kmsg->msg_namelen) {
+	if (msg.msg_name && kmsg->msg_namelen) {
 		if (!save_addr) {
-			err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
+			err = move_addr_to_kernel(msg.msg_name, kmsg->msg_namelen,
 						  kmsg->msg_name);
 			if (err < 0)
 				return err;
@@ -1949,12 +1945,13 @@ static int copy_msghdr_from_user(struct msghdr *kmsg,
 		kmsg->msg_namelen = 0;
 	}
 
-	if (nr_segs > UIO_MAXIOV)
+	if (msg.msg_iovlen > UIO_MAXIOV)
 		return -EMSGSIZE;
 
 	kmsg->msg_iocb = NULL;
 
-	return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
+	return import_iovec(save_addr ? READ : WRITE,
+			    msg.msg_iov, msg.msg_iovlen,
 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
 }
 
diff --git a/net/sunrpc/auth_gss/gss_krb5_crypto.c b/net/sunrpc/auth_gss/gss_krb5_crypto.c
index fb39284ec174..12649c9fedab 100644
--- a/net/sunrpc/auth_gss/gss_krb5_crypto.c
+++ b/net/sunrpc/auth_gss/gss_krb5_crypto.c
@@ -34,6 +34,7 @@
  * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  */
 
+#include <crypto/algapi.h>
 #include <crypto/hash.h>
 #include <crypto/skcipher.h>
 #include <linux/err.h>
@@ -927,7 +928,7 @@ gss_krb5_aes_decrypt(struct krb5_ctx *kctx, u32 offset, struct xdr_buf *buf,
 	if (ret)
 		goto out_err;
 
-	if (memcmp(pkt_hmac, our_hmac, kctx->gk5e->cksumlength) != 0) {
+	if (crypto_memneq(pkt_hmac, our_hmac, kctx->gk5e->cksumlength) != 0) {
 		ret = GSS_S_BAD_SIG;
 		goto out_err;
 	}
diff --git a/net/sunrpc/auth_gss/gss_rpc_upcall.c b/net/sunrpc/auth_gss/gss_rpc_upcall.c
index f0c6a8c78a56..46b295e4f2b8 100644
--- a/net/sunrpc/auth_gss/gss_rpc_upcall.c
+++ b/net/sunrpc/auth_gss/gss_rpc_upcall.c
@@ -55,15 +55,15 @@ enum {
 #define PROC(proc, name)				\
 [GSSX_##proc] = {					\
 	.p_proc   = GSSX_##proc,			\
-	.p_encode = (kxdreproc_t)gssx_enc_##name,	\
-	.p_decode = (kxdrdproc_t)gssx_dec_##name,	\
+	.p_encode = gssx_enc_##name,	\
+	.p_decode = gssx_dec_##name,	\
 	.p_arglen = GSSX_ARG_##name##_sz,		\
 	.p_replen = GSSX_RES_##name##_sz, 		\
 	.p_statidx = GSSX_##proc,			\
 	.p_name   = #proc,				\
 }
 
-static struct rpc_procinfo gssp_procedures[] = {
+static const struct rpc_procinfo gssp_procedures[] = {
 	PROC(INDICATE_MECHS, indicate_mechs),
         PROC(GET_CALL_CONTEXT, get_call_context),
         PROC(IMPORT_AND_CANON_NAME, import_and_canon_name),
@@ -364,11 +364,12 @@ void gssp_free_upcall_data(struct gssp_upcall_data *data)
 /*
  * Initialization stuff
  */
-
+static unsigned int gssp_version1_counts[ARRAY_SIZE(gssp_procedures)];
 static const struct rpc_version gssp_version1 = {
 	.number		= GSSPROXY_VERS_1,
 	.nrprocs	= ARRAY_SIZE(gssp_procedures),
 	.procs		= gssp_procedures,
+	.counts		= gssp_version1_counts,
 };
 
 static const struct rpc_version *gssp_version[] = {
diff --git a/net/sunrpc/auth_gss/gss_rpc_xdr.c b/net/sunrpc/auth_gss/gss_rpc_xdr.c
index 25d9a9cf7b66..c4778cae58ef 100644
--- a/net/sunrpc/auth_gss/gss_rpc_xdr.c
+++ b/net/sunrpc/auth_gss/gss_rpc_xdr.c
@@ -44,7 +44,7 @@ static int gssx_dec_bool(struct xdr_stream *xdr, u32 *v)
 }
 
 static int gssx_enc_buffer(struct xdr_stream *xdr,
-			   gssx_buffer *buf)
+			   const gssx_buffer *buf)
 {
 	__be32 *p;
 
@@ -56,7 +56,7 @@ static int gssx_enc_buffer(struct xdr_stream *xdr,
 }
 
 static int gssx_enc_in_token(struct xdr_stream *xdr,
-			     struct gssp_in_token *in)
+			     const struct gssp_in_token *in)
 {
 	__be32 *p;
 
@@ -130,7 +130,7 @@ static int gssx_dec_option(struct xdr_stream *xdr,
 }
 
 static int dummy_enc_opt_array(struct xdr_stream *xdr,
-				struct gssx_option_array *oa)
+				const struct gssx_option_array *oa)
 {
 	__be32 *p;
 
@@ -348,7 +348,7 @@ static int gssx_dec_status(struct xdr_stream *xdr,
 }
 
 static int gssx_enc_call_ctx(struct xdr_stream *xdr,
-			     struct gssx_call_ctx *ctx)
+			     const struct gssx_call_ctx *ctx)
 {
 	struct gssx_option opt;
 	__be32 *p;
@@ -733,8 +733,9 @@ static int gssx_enc_cb(struct xdr_stream *xdr, struct gssx_cb *cb)
 
 void gssx_enc_accept_sec_context(struct rpc_rqst *req,
 				 struct xdr_stream *xdr,
-				 struct gssx_arg_accept_sec_context *arg)
+				 const void *data)
 {
+	const struct gssx_arg_accept_sec_context *arg = data;
 	int err;
 
 	err = gssx_enc_call_ctx(xdr, &arg->call_ctx);
@@ -789,8 +790,9 @@ done:
 
 int gssx_dec_accept_sec_context(struct rpc_rqst *rqstp,
 				struct xdr_stream *xdr,
-				struct gssx_res_accept_sec_context *res)
+				void *data)
 {
+	struct gssx_res_accept_sec_context *res = data;
 	u32 value_follows;
 	int err;
 	struct page *scratch;
diff --git a/net/sunrpc/auth_gss/gss_rpc_xdr.h b/net/sunrpc/auth_gss/gss_rpc_xdr.h
index 9d88c6239f01..146c31032917 100644
--- a/net/sunrpc/auth_gss/gss_rpc_xdr.h
+++ b/net/sunrpc/auth_gss/gss_rpc_xdr.h
@@ -179,10 +179,10 @@ struct gssx_res_accept_sec_context {
 #define gssx_dec_init_sec_context NULL
 void gssx_enc_accept_sec_context(struct rpc_rqst *req,
 				 struct xdr_stream *xdr,
-				 struct gssx_arg_accept_sec_context *args);
+				 const void *data);
 int gssx_dec_accept_sec_context(struct rpc_rqst *rqstp,
 				struct xdr_stream *xdr,
-				struct gssx_res_accept_sec_context *res);
+				void *data);
 #define gssx_enc_release_handle NULL
 #define gssx_dec_release_handle NULL
 #define gssx_enc_get_mic NULL
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index a54a7a3d28f5..7b1ee5a0b03c 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -838,6 +838,14 @@ unwrap_integ_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct g
 	struct xdr_netobj mic;
 	struct xdr_buf integ_buf;
 
+	/* NFS READ normally uses splice to send data in-place. However
+	 * the data in cache can change after the reply's MIC is computed
+	 * but before the RPC reply is sent. To prevent the client from
+	 * rejecting the server-computed MIC in this somewhat rare case,
+	 * do not use splice with the GSS integrity service.
+	 */
+	clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
+
 	/* Did we already verify the signature on the original pass through? */
 	if (rqstp->rq_deferred)
 		return 0;
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index b5cb921775a0..2e49d1f892b7 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1517,14 +1517,16 @@ static void
 call_start(struct rpc_task *task)
 {
 	struct rpc_clnt	*clnt = task->tk_client;
+	int idx = task->tk_msg.rpc_proc->p_statidx;
 
 	dprintk("RPC: %5u call_start %s%d proc %s (%s)\n", task->tk_pid,
 			clnt->cl_program->name, clnt->cl_vers,
 			rpc_proc_name(task),
 			(RPC_IS_ASYNC(task) ? "async" : "sync"));
 
-	/* Increment call count */
-	task->tk_msg.rpc_proc->p_count++;
+	/* Increment call count (version might not be valid for ping) */
+	if (clnt->cl_program->version[clnt->cl_vers])
+		clnt->cl_program->version[clnt->cl_vers]->counts[idx]++;
 	clnt->cl_stats->rpccnt++;
 	task->tk_action = call_reserve;
 }
@@ -1672,7 +1674,7 @@ call_allocate(struct rpc_task *task)
 	unsigned int slack = task->tk_rqstp->rq_cred->cr_auth->au_cslack;
 	struct rpc_rqst *req = task->tk_rqstp;
 	struct rpc_xprt *xprt = req->rq_xprt;
-	struct rpc_procinfo *proc = task->tk_msg.rpc_proc;
+	const struct rpc_procinfo *proc = task->tk_msg.rpc_proc;
 	int status;
 
 	dprint_status(task);
@@ -2476,16 +2478,18 @@ out_overflow:
 	goto out_garbage;
 }
 
-static void rpcproc_encode_null(void *rqstp, struct xdr_stream *xdr, void *obj)
+static void rpcproc_encode_null(struct rpc_rqst *rqstp, struct xdr_stream *xdr,
+		const void *obj)
 {
 }
 
-static int rpcproc_decode_null(void *rqstp, struct xdr_stream *xdr, void *obj)
+static int rpcproc_decode_null(struct rpc_rqst *rqstp, struct xdr_stream *xdr,
+		void *obj)
 {
 	return 0;
 }
 
-static struct rpc_procinfo rpcproc_null = {
+static const struct rpc_procinfo rpcproc_null = {
 	.p_encode = rpcproc_encode_null,
 	.p_decode = rpcproc_decode_null,
 };
diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c
index 5b30603596d0..ea0676f199c8 100644
--- a/net/sunrpc/rpcb_clnt.c
+++ b/net/sunrpc/rpcb_clnt.c
@@ -128,13 +128,13 @@ struct rpcbind_args {
 	int			r_status;
 };
 
-static struct rpc_procinfo rpcb_procedures2[];
-static struct rpc_procinfo rpcb_procedures3[];
-static struct rpc_procinfo rpcb_procedures4[];
+static const struct rpc_procinfo rpcb_procedures2[];
+static const struct rpc_procinfo rpcb_procedures3[];
+static const struct rpc_procinfo rpcb_procedures4[];
 
 struct rpcb_info {
 	u32			rpc_vers;
-	struct rpc_procinfo *	rpc_proc;
+	const struct rpc_procinfo *rpc_proc;
 };
 
 static const struct rpcb_info rpcb_next_version[];
@@ -620,7 +620,8 @@ int rpcb_v4_register(struct net *net, const u32 program, const u32 version,
 	return -EAFNOSUPPORT;
 }
 
-static struct rpc_task *rpcb_call_async(struct rpc_clnt *rpcb_clnt, struct rpcbind_args *map, struct rpc_procinfo *proc)
+static struct rpc_task *rpcb_call_async(struct rpc_clnt *rpcb_clnt,
+		struct rpcbind_args *map, const struct rpc_procinfo *proc)
 {
 	struct rpc_message msg = {
 		.rpc_proc = proc,
@@ -671,7 +672,7 @@ static struct rpc_clnt *rpcb_find_transport_owner(struct rpc_clnt *clnt)
 void rpcb_getport_async(struct rpc_task *task)
 {
 	struct rpc_clnt *clnt;
-	struct rpc_procinfo *proc;
+	const struct rpc_procinfo *proc;
 	u32 bind_version;
 	struct rpc_xprt *xprt;
 	struct rpc_clnt	*rpcb_clnt;
@@ -843,8 +844,9 @@ static void rpcb_getport_done(struct rpc_task *child, void *data)
  */
 
 static void rpcb_enc_mapping(struct rpc_rqst *req, struct xdr_stream *xdr,
-			     const struct rpcbind_args *rpcb)
+			     const void *data)
 {
+	const struct rpcbind_args *rpcb = data;
 	__be32 *p;
 
 	dprintk("RPC: %5u encoding PMAP_%s call (%u, %u, %d, %u)\n",
@@ -860,8 +862,9 @@ static void rpcb_enc_mapping(struct rpc_rqst *req, struct xdr_stream *xdr,
 }
 
 static int rpcb_dec_getport(struct rpc_rqst *req, struct xdr_stream *xdr,
-			    struct rpcbind_args *rpcb)
+			    void *data)
 {
+	struct rpcbind_args *rpcb = data;
 	unsigned long port;
 	__be32 *p;
 
@@ -882,8 +885,9 @@ static int rpcb_dec_getport(struct rpc_rqst *req, struct xdr_stream *xdr,
 }
 
 static int rpcb_dec_set(struct rpc_rqst *req, struct xdr_stream *xdr,
-			unsigned int *boolp)
+			void *data)
 {
+	unsigned int *boolp = data;
 	__be32 *p;
 
 	p = xdr_inline_decode(xdr, 4);
@@ -917,8 +921,9 @@ static void encode_rpcb_string(struct xdr_stream *xdr, const char *string,
 }
 
 static void rpcb_enc_getaddr(struct rpc_rqst *req, struct xdr_stream *xdr,
-			     const struct rpcbind_args *rpcb)
+			     const void *data)
 {
+	const struct rpcbind_args *rpcb = data;
 	__be32 *p;
 
 	dprintk("RPC: %5u encoding RPCB_%s call (%u, %u, '%s', '%s')\n",
@@ -937,8 +942,9 @@ static void rpcb_enc_getaddr(struct rpc_rqst *req, struct xdr_stream *xdr,
 }
 
 static int rpcb_dec_getaddr(struct rpc_rqst *req, struct xdr_stream *xdr,
-			    struct rpcbind_args *rpcb)
+			    void *data)
 {
+	struct rpcbind_args *rpcb = data;
 	struct sockaddr_storage address;
 	struct sockaddr *sap = (struct sockaddr *)&address;
 	__be32 *p;
@@ -989,11 +995,11 @@ out_fail:
  * since the Linux kernel RPC code requires only these.
  */
 
-static struct rpc_procinfo rpcb_procedures2[] = {
+static const struct rpc_procinfo rpcb_procedures2[] = {
 	[RPCBPROC_SET] = {
 		.p_proc		= RPCBPROC_SET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_mapping,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_mapping,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_mappingargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_SET,
@@ -1002,8 +1008,8 @@ static struct rpc_procinfo rpcb_procedures2[] = {
 	},
 	[RPCBPROC_UNSET] = {
 		.p_proc		= RPCBPROC_UNSET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_mapping,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_mapping,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_mappingargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_UNSET,
@@ -1012,8 +1018,8 @@ static struct rpc_procinfo rpcb_procedures2[] = {
 	},
 	[RPCBPROC_GETPORT] = {
 		.p_proc		= RPCBPROC_GETPORT,
-		.p_encode	= (kxdreproc_t)rpcb_enc_mapping,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_getport,
+		.p_encode	= rpcb_enc_mapping,
+		.p_decode	= rpcb_dec_getport,
 		.p_arglen	= RPCB_mappingargs_sz,
 		.p_replen	= RPCB_getportres_sz,
 		.p_statidx	= RPCBPROC_GETPORT,
@@ -1022,11 +1028,11 @@ static struct rpc_procinfo rpcb_procedures2[] = {
 	},
 };
 
-static struct rpc_procinfo rpcb_procedures3[] = {
+static const struct rpc_procinfo rpcb_procedures3[] = {
 	[RPCBPROC_SET] = {
 		.p_proc		= RPCBPROC_SET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_SET,
@@ -1035,8 +1041,8 @@ static struct rpc_procinfo rpcb_procedures3[] = {
 	},
 	[RPCBPROC_UNSET] = {
 		.p_proc		= RPCBPROC_UNSET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_UNSET,
@@ -1045,8 +1051,8 @@ static struct rpc_procinfo rpcb_procedures3[] = {
 	},
 	[RPCBPROC_GETADDR] = {
 		.p_proc		= RPCBPROC_GETADDR,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_getaddr,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_getaddr,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_getaddrres_sz,
 		.p_statidx	= RPCBPROC_GETADDR,
@@ -1055,11 +1061,11 @@ static struct rpc_procinfo rpcb_procedures3[] = {
 	},
 };
 
-static struct rpc_procinfo rpcb_procedures4[] = {
+static const struct rpc_procinfo rpcb_procedures4[] = {
 	[RPCBPROC_SET] = {
 		.p_proc		= RPCBPROC_SET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_SET,
@@ -1068,8 +1074,8 @@ static struct rpc_procinfo rpcb_procedures4[] = {
 	},
 	[RPCBPROC_UNSET] = {
 		.p_proc		= RPCBPROC_UNSET,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_set,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_set,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_setres_sz,
 		.p_statidx	= RPCBPROC_UNSET,
@@ -1078,8 +1084,8 @@ static struct rpc_procinfo rpcb_procedures4[] = {
 	},
 	[RPCBPROC_GETADDR] = {
 		.p_proc		= RPCBPROC_GETADDR,
-		.p_encode	= (kxdreproc_t)rpcb_enc_getaddr,
-		.p_decode	= (kxdrdproc_t)rpcb_dec_getaddr,
+		.p_encode	= rpcb_enc_getaddr,
+		.p_decode	= rpcb_dec_getaddr,
 		.p_arglen	= RPCB_getaddrargs_sz,
 		.p_replen	= RPCB_getaddrres_sz,
 		.p_statidx	= RPCBPROC_GETADDR,
@@ -1112,22 +1118,28 @@ static const struct rpcb_info rpcb_next_version6[] = {
 	},
 };
 
+static unsigned int rpcb_version2_counts[ARRAY_SIZE(rpcb_procedures2)];
 static const struct rpc_version rpcb_version2 = {
 	.number		= RPCBVERS_2,
 	.nrprocs	= ARRAY_SIZE(rpcb_procedures2),
-	.procs		= rpcb_procedures2
+	.procs		= rpcb_procedures2,
+	.counts		= rpcb_version2_counts,
 };
 
+static unsigned int rpcb_version3_counts[ARRAY_SIZE(rpcb_procedures3)];
 static const struct rpc_version rpcb_version3 = {
 	.number		= RPCBVERS_3,
 	.nrprocs	= ARRAY_SIZE(rpcb_procedures3),
-	.procs		= rpcb_procedures3
+	.procs		= rpcb_procedures3,
+	.counts		= rpcb_version3_counts,
 };
 
+static unsigned int rpcb_version4_counts[ARRAY_SIZE(rpcb_procedures4)];
 static const struct rpc_version rpcb_version4 = {
 	.number		= RPCBVERS_4,
 	.nrprocs	= ARRAY_SIZE(rpcb_procedures4),
-	.procs		= rpcb_procedures4
+	.procs		= rpcb_procedures4,
+	.counts		= rpcb_version4_counts,
 };
 
 static const struct rpc_version *rpcb_version[] = {
diff --git a/net/sunrpc/stats.c b/net/sunrpc/stats.c
index caeb01ad2b5a..1e671333c3d5 100644
--- a/net/sunrpc/stats.c
+++ b/net/sunrpc/stats.c
@@ -55,8 +55,7 @@ static int rpc_proc_show(struct seq_file *seq, void *v) {
 		seq_printf(seq, "proc%u %u",
 					vers->number, vers->nrprocs);
 		for (j = 0; j < vers->nrprocs; j++)
-			seq_printf(seq, " %u",
-					vers->procs[j].p_count);
+			seq_printf(seq, " %u", vers->counts[j]);
 		seq_putc(seq, '\n');
 	}
 	return 0;
@@ -78,9 +77,9 @@ static const struct file_operations rpc_proc_fops = {
 /*
  * Get RPC server stats
  */
-void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp) {
+void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp)
+{
 	const struct svc_program *prog = statp->program;
-	const struct svc_procedure *proc;
 	const struct svc_version *vers;
 	unsigned int i, j;
 
@@ -99,11 +98,12 @@ void svc_seq_show(struct seq_file *seq, const struct svc_stat *statp) {
 			statp->rpcbadclnt);
 
 	for (i = 0; i < prog->pg_nvers; i++) {
-		if (!(vers = prog->pg_vers[i]) || !(proc = vers->vs_proc))
+		vers = prog->pg_vers[i];
+		if (!vers)
 			continue;
 		seq_printf(seq, "proc%d %u", i, vers->vs_nproc);
-		for (j = 0; j < vers->vs_nproc; j++, proc++)
-			seq_printf(seq, " %u", proc->pc_count);
+		for (j = 0; j < vers->vs_nproc; j++)
+			seq_printf(seq, " %u", vers->vs_count[j]);
 		seq_putc(seq, '\n');
 	}
 }
@@ -192,7 +192,7 @@ void rpc_count_iostats(const struct rpc_task *task, struct rpc_iostats *stats)
 EXPORT_SYMBOL_GPL(rpc_count_iostats);
 
 static void _print_name(struct seq_file *seq, unsigned int op,
-			struct rpc_procinfo *procs)
+			const struct rpc_procinfo *procs)
 {
 	if (procs[op].p_name)
 		seq_printf(seq, "\t%12s: ", procs[op].p_name);
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index bc0f5a0ecbdc..85ce0db5b0a6 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1008,7 +1008,7 @@ int svc_register(const struct svc_serv *serv, struct net *net,
 		 const unsigned short port)
 {
 	struct svc_program	*progp;
-	struct svc_version	*vers;
+	const struct svc_version *vers;
 	unsigned int		i;
 	int			error = 0;
 
@@ -1151,10 +1151,9 @@ static int
 svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 {
 	struct svc_program	*progp;
-	struct svc_version	*versp = NULL;	/* compiler food */
-	struct svc_procedure	*procp = NULL;
+	const struct svc_version *versp = NULL;	/* compiler food */
+	const struct svc_procedure *procp = NULL;
 	struct svc_serv		*serv = rqstp->rq_server;
-	kxdrproc_t		xdr;
 	__be32			*statp;
 	u32			prog, vers, proc;
 	__be32			auth_stat, rpc_stat;
@@ -1166,7 +1165,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 	if (argv->iov_len < 6*4)
 		goto err_short_len;
 
-	/* Will be turned off only in gss privacy case: */
+	/* Will be turned off by GSS integrity and privacy services */
 	set_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 	/* Will be turned off only when NFSv4 Sessions are used */
 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
@@ -1262,7 +1261,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 	svc_putnl(resv, RPC_SUCCESS);
 
 	/* Bump per-procedure stats counter */
-	procp->pc_count++;
+	versp->vs_count[proc]++;
 
 	/* Initialize storage for argp and resp */
 	memset(rqstp->rq_argp, 0, procp->pc_argsize);
@@ -1276,28 +1275,30 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 
 	/* Call the function that processes the request. */
 	if (!versp->vs_dispatch) {
-		/* Decode arguments */
-		xdr = procp->pc_decode;
-		if (xdr && !xdr(rqstp, argv->iov_base, rqstp->rq_argp))
+		/*
+		 * Decode arguments
+		 * XXX: why do we ignore the return value?
+		 */
+		if (procp->pc_decode &&
+		    !procp->pc_decode(rqstp, argv->iov_base))
 			goto err_garbage;
 
-		*statp = procp->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
+		*statp = procp->pc_func(rqstp);
 
 		/* Encode reply */
 		if (*statp == rpc_drop_reply ||
 		    test_bit(RQ_DROPME, &rqstp->rq_flags)) {
 			if (procp->pc_release)
-				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
+				procp->pc_release(rqstp);
 			goto dropit;
 		}
 		if (*statp == rpc_autherr_badcred) {
 			if (procp->pc_release)
-				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
+				procp->pc_release(rqstp);
 			goto err_bad_auth;
 		}
-		if (*statp == rpc_success &&
-		    (xdr = procp->pc_encode) &&
-		    !xdr(rqstp, resv->iov_base+resv->iov_len, rqstp->rq_resp)) {
+		if (*statp == rpc_success && procp->pc_encode &&
+		    !procp->pc_encode(rqstp, resv->iov_base + resv->iov_len)) {
 			dprintk("svc: failed to encode reply\n");
 			/* serv->sv_stats->rpcsystemerr++; */
 			*statp = rpc_system_err;
@@ -1307,7 +1308,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 		if (!versp->vs_dispatch(rqstp, statp)) {
 			/* Release reply info */
 			if (procp->pc_release)
-				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
+				procp->pc_release(rqstp);
 			goto dropit;
 		}
 	}
@@ -1318,7 +1319,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 
 	/* Release reply info */
 	if (procp->pc_release)
-		procp->pc_release(rqstp, NULL, rqstp->rq_resp);
+		procp->pc_release(rqstp);
 
 	if (procp->pc_encode == NULL)
 		goto dropit;
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index 7bfe1fb42add..d16a8b423c20 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -659,11 +659,13 @@ static int svc_alloc_arg(struct svc_rqst *rqstp)
 	int i;
 
 	/* now allocate needed pages.  If we get a failure, sleep briefly */
-	pages = (serv->sv_max_mesg + PAGE_SIZE) / PAGE_SIZE;
-	WARN_ON_ONCE(pages >= RPCSVC_MAXPAGES);
-	if (pages >= RPCSVC_MAXPAGES)
+	pages = (serv->sv_max_mesg + 2 * PAGE_SIZE) >> PAGE_SHIFT;
+	if (pages > RPCSVC_MAXPAGES) {
+		pr_warn_once("svc: warning: pages=%u > RPCSVC_MAXPAGES=%lu\n",
+			     pages, RPCSVC_MAXPAGES);
 		/* use as many pages as possible */
-		pages = RPCSVC_MAXPAGES - 1;
+		pages = RPCSVC_MAXPAGES;
+	}
 	for (i = 0; i < pages ; i++)
 		while (rqstp->rq_pages[i] == NULL) {
 			struct page *p = alloc_page(GFP_KERNEL);
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 3e63c5e97ebe..4654a9934269 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -1047,13 +1047,15 @@ out:
 	return ret;
 }
 
-static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt, gfp_t gfp_flags)
+static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt)
 {
 	struct rpc_rqst *req = ERR_PTR(-EAGAIN);
 
 	if (!atomic_add_unless(&xprt->num_reqs, 1, xprt->max_reqs))
 		goto out;
-	req = kzalloc(sizeof(struct rpc_rqst), gfp_flags);
+	spin_unlock(&xprt->reserve_lock);
+	req = kzalloc(sizeof(struct rpc_rqst), GFP_NOFS);
+	spin_lock(&xprt->reserve_lock);
 	if (req != NULL)
 		goto out;
 	atomic_dec(&xprt->num_reqs);
@@ -1081,7 +1083,7 @@ void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
 		list_del(&req->rq_list);
 		goto out_init_req;
 	}
-	req = xprt_dynamic_alloc_slot(xprt, GFP_NOWAIT|__GFP_NOWARN);
+	req = xprt_dynamic_alloc_slot(xprt);
 	if (!IS_ERR(req))
 		goto out_init_req;
 	switch (PTR_ERR(req)) {
diff --git a/net/sunrpc/xprtrdma/Makefile b/net/sunrpc/xprtrdma/Makefile
index c1ae8142ab73..b8213ddce2f2 100644
--- a/net/sunrpc/xprtrdma/Makefile
+++ b/net/sunrpc/xprtrdma/Makefile
@@ -3,6 +3,6 @@ obj-$(CONFIG_SUNRPC_XPRT_RDMA) += rpcrdma.o
 rpcrdma-y := transport.o rpc_rdma.o verbs.o \
 	fmr_ops.o frwr_ops.o \
 	svc_rdma.o svc_rdma_backchannel.o svc_rdma_transport.o \
-	svc_rdma_marshal.o svc_rdma_sendto.o svc_rdma_recvfrom.o \
-	svc_rdma_rw.o module.o
+	svc_rdma_sendto.o svc_rdma_recvfrom.o svc_rdma_rw.o \
+	module.o
 rpcrdma-$(CONFIG_SUNRPC_BACKCHANNEL) += backchannel.o
diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c
index 59e64025ed96..d3f84bb1d443 100644
--- a/net/sunrpc/xprtrdma/fmr_ops.c
+++ b/net/sunrpc/xprtrdma/fmr_ops.c
@@ -91,7 +91,7 @@ __fmr_unmap(struct rpcrdma_mw *mw)
 
 	list_add(&mw->fmr.fm_mr->list, &l);
 	rc = ib_unmap_fmr(&l);
-	list_del_init(&mw->fmr.fm_mr->list);
+	list_del(&mw->fmr.fm_mr->list);
 	return rc;
 }
 
@@ -213,13 +213,11 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
 			break;
 	}
-	mw->mw_nents = i;
 	mw->mw_dir = rpcrdma_data_dir(writing);
-	if (i == 0)
-		goto out_dmamap_err;
 
-	if (!ib_dma_map_sg(r_xprt->rx_ia.ri_device,
-			   mw->mw_sg, mw->mw_nents, mw->mw_dir))
+	mw->mw_nents = ib_dma_map_sg(r_xprt->rx_ia.ri_device,
+				     mw->mw_sg, i, mw->mw_dir);
+	if (!mw->mw_nents)
 		goto out_dmamap_err;
 
 	for (i = 0, dma_pages = mw->fmr.fm_physaddrs; i < mw->mw_nents; i++)
@@ -237,16 +235,18 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	return mw->mw_nents;
 
 out_dmamap_err:
-	pr_err("rpcrdma: failed to dma map sg %p sg_nents %u\n",
-	       mw->mw_sg, mw->mw_nents);
-	rpcrdma_defer_mr_recovery(mw);
+	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
+	       mw->mw_sg, i);
+	rpcrdma_put_mw(r_xprt, mw);
 	return -EIO;
 
 out_maperr:
 	pr_err("rpcrdma: ib_map_phys_fmr %u@0x%llx+%i (%d) status %i\n",
 	       len, (unsigned long long)dma_pages[0],
 	       pageoff, mw->mw_nents, rc);
-	rpcrdma_defer_mr_recovery(mw);
+	ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
+			mw->mw_sg, mw->mw_nents, mw->mw_dir);
+	rpcrdma_put_mw(r_xprt, mw);
 	return -EIO;
 }
 
@@ -255,24 +255,26 @@ out_maperr:
  * Sleeps until it is safe for the host CPU to access the
  * previously mapped memory regions.
  *
- * Caller ensures that req->rl_registered is not empty.
+ * Caller ensures that @mws is not empty before the call. This
+ * function empties the list.
  */
 static void
-fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 {
-	struct rpcrdma_mw *mw, *tmp;
+	struct rpcrdma_mw *mw;
 	LIST_HEAD(unmap_list);
 	int rc;
 
-	dprintk("RPC:       %s: req %p\n", __func__, req);
-
 	/* ORDER: Invalidate all of the req's MRs first
 	 *
 	 * ib_unmap_fmr() is slow, so use a single call instead
 	 * of one call per mapped FMR.
 	 */
-	list_for_each_entry(mw, &req->rl_registered, mw_list)
+	list_for_each_entry(mw, mws, mw_list) {
+		dprintk("RPC:       %s: unmapping fmr %p\n",
+			__func__, &mw->fmr);
 		list_add_tail(&mw->fmr.fm_mr->list, &unmap_list);
+	}
 	r_xprt->rx_stats.local_inv_needed++;
 	rc = ib_unmap_fmr(&unmap_list);
 	if (rc)
@@ -281,9 +283,11 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 	/* ORDER: Now DMA unmap all of the req's MRs, and return
 	 * them to the free MW list.
 	 */
-	list_for_each_entry_safe(mw, tmp, &req->rl_registered, mw_list) {
-		list_del_init(&mw->mw_list);
-		list_del_init(&mw->fmr.fm_mr->list);
+	while (!list_empty(mws)) {
+		mw = rpcrdma_pop_mw(mws);
+		dprintk("RPC:       %s: DMA unmapping fmr %p\n",
+			__func__, &mw->fmr);
+		list_del(&mw->fmr.fm_mr->list);
 		ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
 				mw->mw_sg, mw->mw_nents, mw->mw_dir);
 		rpcrdma_put_mw(r_xprt, mw);
@@ -294,8 +298,9 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 out_reset:
 	pr_err("rpcrdma: ib_unmap_fmr failed (%i)\n", rc);
 
-	list_for_each_entry_safe(mw, tmp, &req->rl_registered, mw_list) {
-		list_del_init(&mw->fmr.fm_mr->list);
+	while (!list_empty(mws)) {
+		mw = rpcrdma_pop_mw(mws);
+		list_del(&mw->fmr.fm_mr->list);
 		fmr_op_recover_mr(mw);
 	}
 }
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index f81dd93176c0..6aea36a38bfd 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -277,7 +277,7 @@ __frwr_sendcompletion_flush(struct ib_wc *wc, const char *wr)
 }
 
 /**
- * frwr_wc_fastreg - Invoked by RDMA provider for each polled FastReg WC
+ * frwr_wc_fastreg - Invoked by RDMA provider for a flushed FastReg WC
  * @cq:	completion queue (ignored)
  * @wc:	completed WR
  *
@@ -298,7 +298,7 @@ frwr_wc_fastreg(struct ib_cq *cq, struct ib_wc *wc)
 }
 
 /**
- * frwr_wc_localinv - Invoked by RDMA provider for each polled LocalInv WC
+ * frwr_wc_localinv - Invoked by RDMA provider for a flushed LocalInv WC
  * @cq:	completion queue (ignored)
  * @wc:	completed WR
  *
@@ -319,7 +319,7 @@ frwr_wc_localinv(struct ib_cq *cq, struct ib_wc *wc)
 }
 
 /**
- * frwr_wc_localinv - Invoked by RDMA provider for each polled LocalInv WC
+ * frwr_wc_localinv_wake - Invoked by RDMA provider for a signaled LocalInv WC
  * @cq:	completion queue (ignored)
  * @wc:	completed WR
  *
@@ -355,7 +355,7 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	struct ib_mr *mr;
 	struct ib_reg_wr *reg_wr;
 	struct ib_send_wr *bad_wr;
-	int rc, i, n, dma_nents;
+	int rc, i, n;
 	u8 key;
 
 	mw = NULL;
@@ -391,14 +391,10 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
 			break;
 	}
-	mw->mw_nents = i;
 	mw->mw_dir = rpcrdma_data_dir(writing);
-	if (i == 0)
-		goto out_dmamap_err;
 
-	dma_nents = ib_dma_map_sg(ia->ri_device,
-				  mw->mw_sg, mw->mw_nents, mw->mw_dir);
-	if (!dma_nents)
+	mw->mw_nents = ib_dma_map_sg(ia->ri_device, mw->mw_sg, i, mw->mw_dir);
+	if (!mw->mw_nents)
 		goto out_dmamap_err;
 
 	n = ib_map_mr_sg(mr, mw->mw_sg, mw->mw_nents, NULL, PAGE_SIZE);
@@ -436,13 +432,14 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	return mw->mw_nents;
 
 out_dmamap_err:
-	pr_err("rpcrdma: failed to dma map sg %p sg_nents %u\n",
-	       mw->mw_sg, mw->mw_nents);
-	rpcrdma_defer_mr_recovery(mw);
+	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
+	       mw->mw_sg, i);
+	frmr->fr_state = FRMR_IS_INVALID;
+	rpcrdma_put_mw(r_xprt, mw);
 	return -EIO;
 
 out_mapmr_err:
-	pr_err("rpcrdma: failed to map mr %p (%u/%u)\n",
+	pr_err("rpcrdma: failed to map mr %p (%d/%d)\n",
 	       frmr->fr_mr, n, mw->mw_nents);
 	rpcrdma_defer_mr_recovery(mw);
 	return -EIO;
@@ -458,21 +455,19 @@ out_senderr:
  * Sleeps until it is safe for the host CPU to access the
  * previously mapped memory regions.
  *
- * Caller ensures that req->rl_registered is not empty.
+ * Caller ensures that @mws is not empty before the call. This
+ * function empties the list.
  */
 static void
-frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 {
 	struct ib_send_wr *first, **prev, *last, *bad_wr;
-	struct rpcrdma_rep *rep = req->rl_reply;
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rpcrdma_frmr *f;
 	struct rpcrdma_mw *mw;
 	int count, rc;
 
-	dprintk("RPC:       %s: req %p\n", __func__, req);
-
-	/* ORDER: Invalidate all of the req's MRs first
+	/* ORDER: Invalidate all of the MRs first
 	 *
 	 * Chain the LOCAL_INV Work Requests and post them with
 	 * a single ib_post_send() call.
@@ -480,11 +475,10 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 	f = NULL;
 	count = 0;
 	prev = &first;
-	list_for_each_entry(mw, &req->rl_registered, mw_list) {
+	list_for_each_entry(mw, mws, mw_list) {
 		mw->frmr.fr_state = FRMR_IS_INVALID;
 
-		if ((rep->rr_wc_flags & IB_WC_WITH_INVALIDATE) &&
-		    (mw->mw_handle == rep->rr_inv_rkey))
+		if (mw->mw_flags & RPCRDMA_MW_F_RI)
 			continue;
 
 		f = &mw->frmr;
@@ -524,18 +518,19 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 	 * unless ri_id->qp is a valid pointer.
 	 */
 	r_xprt->rx_stats.local_inv_needed++;
+	bad_wr = NULL;
 	rc = ib_post_send(ia->ri_id->qp, first, &bad_wr);
+	if (bad_wr != first)
+		wait_for_completion(&f->fr_linv_done);
 	if (rc)
 		goto reset_mrs;
 
-	wait_for_completion(&f->fr_linv_done);
-
-	/* ORDER: Now DMA unmap all of the req's MRs, and return
+	/* ORDER: Now DMA unmap all of the MRs, and return
 	 * them to the free MW list.
 	 */
 unmap:
-	while (!list_empty(&req->rl_registered)) {
-		mw = rpcrdma_pop_mw(&req->rl_registered);
+	while (!list_empty(mws)) {
+		mw = rpcrdma_pop_mw(mws);
 		dprintk("RPC:       %s: DMA unmapping frmr %p\n",
 			__func__, &mw->frmr);
 		ib_dma_unmap_sg(ia->ri_device,
@@ -546,17 +541,19 @@ unmap:
 
 reset_mrs:
 	pr_err("rpcrdma: FRMR invalidate ib_post_send returned %i\n", rc);
-	rdma_disconnect(ia->ri_id);
 
 	/* Find and reset the MRs in the LOCAL_INV WRs that did not
-	 * get posted. This is synchronous, and slow.
+	 * get posted.
 	 */
-	list_for_each_entry(mw, &req->rl_registered, mw_list) {
-		f = &mw->frmr;
-		if (mw->mw_handle == bad_wr->ex.invalidate_rkey) {
-			__frwr_reset_mr(ia, mw);
-			bad_wr = bad_wr->next;
-		}
+	rpcrdma_init_cqcount(&r_xprt->rx_ep, -count);
+	while (bad_wr) {
+		f = container_of(bad_wr, struct rpcrdma_frmr,
+				 fr_invwr);
+		mw = container_of(f, struct rpcrdma_mw, frmr);
+
+		__frwr_reset_mr(ia, mw);
+
+		bad_wr = bad_wr->next;
 	}
 	goto unmap;
 }
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 694e9b13ecf0..ca4d6e4528f3 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -141,7 +141,7 @@ static bool rpcrdma_args_inline(struct rpcrdma_xprt *r_xprt,
 
 	if (xdr->page_len) {
 		remaining = xdr->page_len;
-		offset = xdr->page_base & ~PAGE_MASK;
+		offset = offset_in_page(xdr->page_base);
 		count = 0;
 		while (remaining) {
 			remaining -= min_t(unsigned int,
@@ -222,7 +222,7 @@ rpcrdma_convert_iovs(struct rpcrdma_xprt *r_xprt, struct xdr_buf *xdrbuf,
 
 	len = xdrbuf->page_len;
 	ppages = xdrbuf->pages + (xdrbuf->page_base >> PAGE_SHIFT);
-	page_base = xdrbuf->page_base & ~PAGE_MASK;
+	page_base = offset_in_page(xdrbuf->page_base);
 	p = 0;
 	while (len && n < RPCRDMA_MAX_SEGS) {
 		if (!ppages[p]) {
@@ -540,7 +540,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 			goto out;
 
 		page = virt_to_page(xdr->tail[0].iov_base);
-		page_base = (unsigned long)xdr->tail[0].iov_base & ~PAGE_MASK;
+		page_base = offset_in_page(xdr->tail[0].iov_base);
 
 		/* If the content in the page list is an odd length,
 		 * xdr_write_pages() has added a pad at the beginning
@@ -557,7 +557,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 	 */
 	if (xdr->page_len) {
 		ppages = xdr->pages + (xdr->page_base >> PAGE_SHIFT);
-		page_base = xdr->page_base & ~PAGE_MASK;
+		page_base = offset_in_page(xdr->page_base);
 		remaining = xdr->page_len;
 		while (remaining) {
 			sge_no++;
@@ -587,7 +587,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 	 */
 	if (xdr->tail[0].iov_len) {
 		page = virt_to_page(xdr->tail[0].iov_base);
-		page_base = (unsigned long)xdr->tail[0].iov_base & ~PAGE_MASK;
+		page_base = offset_in_page(xdr->tail[0].iov_base);
 		len = xdr->tail[0].iov_len;
 
 map_tail:
@@ -734,6 +734,9 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst)
 		rpclen = 0;
 	}
 
+	req->rl_xid = rqst->rq_xid;
+	rpcrdma_insert_req(&r_xprt->rx_buf, req);
+
 	/* This implementation supports the following combinations
 	 * of chunk lists in one RPC-over-RDMA Call message:
 	 *
@@ -875,9 +878,9 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
 	srcp += curlen;
 	copy_len -= curlen;
 
-	page_base = rqst->rq_rcv_buf.page_base;
-	ppages = rqst->rq_rcv_buf.pages + (page_base >> PAGE_SHIFT);
-	page_base &= ~PAGE_MASK;
+	ppages = rqst->rq_rcv_buf.pages +
+		(rqst->rq_rcv_buf.page_base >> PAGE_SHIFT);
+	page_base = offset_in_page(rqst->rq_rcv_buf.page_base);
 	fixup_copy_count = 0;
 	if (copy_len && rqst->rq_rcv_buf.page_len) {
 		int pagelist_len;
@@ -928,6 +931,24 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
 	return fixup_copy_count;
 }
 
+/* Caller must guarantee @rep remains stable during this call.
+ */
+static void
+rpcrdma_mark_remote_invalidation(struct list_head *mws,
+				 struct rpcrdma_rep *rep)
+{
+	struct rpcrdma_mw *mw;
+
+	if (!(rep->rr_wc_flags & IB_WC_WITH_INVALIDATE))
+		return;
+
+	list_for_each_entry(mw, mws, mw_list)
+		if (mw->mw_handle == rep->rr_inv_rkey) {
+			mw->mw_flags = RPCRDMA_MW_F_RI;
+			break; /* only one invalidated MR per RPC */
+		}
+}
+
 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
 /* By convention, backchannel calls arrive via rdma_msg type
  * messages, and never populate the chunk lists. This makes
@@ -969,14 +990,16 @@ rpcrdma_reply_handler(struct work_struct *work)
 {
 	struct rpcrdma_rep *rep =
 			container_of(work, struct rpcrdma_rep, rr_work);
+	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
+	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
+	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
 	struct rpcrdma_msg *headerp;
 	struct rpcrdma_req *req;
 	struct rpc_rqst *rqst;
-	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
-	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
 	__be32 *iptr;
 	int rdmalen, status, rmerr;
 	unsigned long cwnd;
+	struct list_head mws;
 
 	dprintk("RPC:       %s: incoming rep %p\n", __func__, rep);
 
@@ -994,27 +1017,45 @@ rpcrdma_reply_handler(struct work_struct *work)
 	/* Match incoming rpcrdma_rep to an rpcrdma_req to
 	 * get context for handling any incoming chunks.
 	 */
-	spin_lock_bh(&xprt->transport_lock);
-	rqst = xprt_lookup_rqst(xprt, headerp->rm_xid);
-	if (!rqst)
+	spin_lock(&buf->rb_lock);
+	req = rpcrdma_lookup_req_locked(&r_xprt->rx_buf,
+					headerp->rm_xid);
+	if (!req)
 		goto out_nomatch;
-
-	req = rpcr_to_rdmar(rqst);
 	if (req->rl_reply)
 		goto out_duplicate;
 
-	/* Sanity checking has passed. We are now committed
-	 * to complete this transaction.
+	list_replace_init(&req->rl_registered, &mws);
+	rpcrdma_mark_remote_invalidation(&mws, rep);
+
+	/* Avoid races with signals and duplicate replies
+	 * by marking this req as matched.
 	 */
-	list_del_init(&rqst->rq_list);
-	spin_unlock_bh(&xprt->transport_lock);
+	req->rl_reply = rep;
+	spin_unlock(&buf->rb_lock);
+
 	dprintk("RPC:       %s: reply %p completes request %p (xid 0x%08x)\n",
 		__func__, rep, req, be32_to_cpu(headerp->rm_xid));
 
-	/* from here on, the reply is no longer an orphan */
-	req->rl_reply = rep;
-	xprt->reestablish_timeout = 0;
+	/* Invalidate and unmap the data payloads before waking the
+	 * waiting application. This guarantees the memory regions
+	 * are properly fenced from the server before the application
+	 * accesses the data. It also ensures proper send flow control:
+	 * waking the next RPC waits until this RPC has relinquished
+	 * all its Send Queue entries.
+	 */
+	if (!list_empty(&mws))
+		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt, &mws);
 
+	/* Perform XID lookup, reconstruction of the RPC reply, and
+	 * RPC completion while holding the transport lock to ensure
+	 * the rep, rqst, and rq_task pointers remain stable.
+	 */
+	spin_lock_bh(&xprt->transport_lock);
+	rqst = xprt_lookup_rqst(xprt, headerp->rm_xid);
+	if (!rqst)
+		goto out_norqst;
+	xprt->reestablish_timeout = 0;
 	if (headerp->rm_vers != rpcrdma_version)
 		goto out_badversion;
 
@@ -1024,12 +1065,9 @@ rpcrdma_reply_handler(struct work_struct *work)
 	case rdma_msg:
 		/* never expect read chunks */
 		/* never expect reply chunks (two ways to check) */
-		/* never expect write chunks without having offered RDMA */
 		if (headerp->rm_body.rm_chunks[0] != xdr_zero ||
 		    (headerp->rm_body.rm_chunks[1] == xdr_zero &&
-		     headerp->rm_body.rm_chunks[2] != xdr_zero) ||
-		    (headerp->rm_body.rm_chunks[1] != xdr_zero &&
-		     list_empty(&req->rl_registered)))
+		     headerp->rm_body.rm_chunks[2] != xdr_zero))
 			goto badheader;
 		if (headerp->rm_body.rm_chunks[1] != xdr_zero) {
 			/* count any expected write chunks in read reply */
@@ -1066,8 +1104,7 @@ rpcrdma_reply_handler(struct work_struct *work)
 		/* never expect read or write chunks, always reply chunks */
 		if (headerp->rm_body.rm_chunks[0] != xdr_zero ||
 		    headerp->rm_body.rm_chunks[1] != xdr_zero ||
-		    headerp->rm_body.rm_chunks[2] != xdr_one ||
-		    list_empty(&req->rl_registered))
+		    headerp->rm_body.rm_chunks[2] != xdr_one)
 			goto badheader;
 		iptr = (__be32 *)((unsigned char *)headerp +
 							RPCRDMA_HDRLEN_MIN);
@@ -1093,17 +1130,6 @@ badheader:
 	}
 
 out:
-	/* Invalidate and flush the data payloads before waking the
-	 * waiting application. This guarantees the memory region is
-	 * properly fenced from the server before the application
-	 * accesses the data. It also ensures proper send flow
-	 * control: waking the next RPC waits until this RPC has
-	 * relinquished all its Send Queue entries.
-	 */
-	if (!list_empty(&req->rl_registered))
-		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt, req);
-
-	spin_lock_bh(&xprt->transport_lock);
 	cwnd = xprt->cwnd;
 	xprt->cwnd = atomic_read(&r_xprt->rx_buf.rb_credits) << RPC_CWNDSHIFT;
 	if (xprt->cwnd > cwnd)
@@ -1112,7 +1138,7 @@ out:
 	xprt_complete_rqst(rqst->rq_task, status);
 	spin_unlock_bh(&xprt->transport_lock);
 	dprintk("RPC:       %s: xprt_complete_rqst(0x%p, 0x%p, %d)\n",
-			__func__, xprt, rqst, status);
+		__func__, xprt, rqst, status);
 	return;
 
 out_badstatus:
@@ -1161,26 +1187,37 @@ out_rdmaerr:
 	r_xprt->rx_stats.bad_reply_count++;
 	goto out;
 
-/* If no pending RPC transaction was matched, post a replacement
- * receive buffer before returning.
+/* The req was still available, but by the time the transport_lock
+ * was acquired, the rqst and task had been released. Thus the RPC
+ * has already been terminated.
  */
+out_norqst:
+	spin_unlock_bh(&xprt->transport_lock);
+	rpcrdma_buffer_put(req);
+	dprintk("RPC:       %s: race, no rqst left for req %p\n",
+		__func__, req);
+	return;
+
 out_shortreply:
 	dprintk("RPC:       %s: short/invalid reply\n", __func__);
 	goto repost;
 
 out_nomatch:
-	spin_unlock_bh(&xprt->transport_lock);
+	spin_unlock(&buf->rb_lock);
 	dprintk("RPC:       %s: no match for incoming xid 0x%08x len %d\n",
 		__func__, be32_to_cpu(headerp->rm_xid),
 		rep->rr_len);
 	goto repost;
 
 out_duplicate:
-	spin_unlock_bh(&xprt->transport_lock);
+	spin_unlock(&buf->rb_lock);
 	dprintk("RPC:       %s: "
 		"duplicate reply %p to RPC request %p: xid 0x%08x\n",
 		__func__, rep, req, be32_to_cpu(headerp->rm_xid));
 
+/* If no pending RPC transaction was matched, post a replacement
+ * receive buffer before returning.
+ */
 repost:
 	r_xprt->rx_stats.bad_reply_count++;
 	if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, rep))
diff --git a/net/sunrpc/xprtrdma/svc_rdma_marshal.c b/net/sunrpc/xprtrdma/svc_rdma_marshal.c
deleted file mode 100644
index bdcf7d85a3dc..000000000000
--- a/net/sunrpc/xprtrdma/svc_rdma_marshal.c
+++ /dev/null
@@ -1,168 +0,0 @@
-/*
- * Copyright (c) 2016 Oracle. All rights reserved.
- * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the BSD-type
- * license below:
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- *      Redistributions of source code must retain the above copyright
- *      notice, this list of conditions and the following disclaimer.
- *
- *      Redistributions in binary form must reproduce the above
- *      copyright notice, this list of conditions and the following
- *      disclaimer in the documentation and/or other materials provided
- *      with the distribution.
- *
- *      Neither the name of the Network Appliance, Inc. nor the names of
- *      its contributors may be used to endorse or promote products
- *      derived from this software without specific prior written
- *      permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- * Author: Tom Tucker <tom@opengridcomputing.com>
- */
-
-#include <linux/sunrpc/xdr.h>
-#include <linux/sunrpc/debug.h>
-#include <asm/unaligned.h>
-#include <linux/sunrpc/rpc_rdma.h>
-#include <linux/sunrpc/svc_rdma.h>
-
-#define RPCDBG_FACILITY	RPCDBG_SVCXPRT
-
-static __be32 *xdr_check_read_list(__be32 *p, __be32 *end)
-{
-	__be32 *next;
-
-	while (*p++ != xdr_zero) {
-		next = p + rpcrdma_readchunk_maxsz - 1;
-		if (next > end)
-			return NULL;
-		p = next;
-	}
-	return p;
-}
-
-static __be32 *xdr_check_write_list(__be32 *p, __be32 *end)
-{
-	__be32 *next;
-
-	while (*p++ != xdr_zero) {
-		next = p + 1 + be32_to_cpup(p) * rpcrdma_segment_maxsz;
-		if (next > end)
-			return NULL;
-		p = next;
-	}
-	return p;
-}
-
-static __be32 *xdr_check_reply_chunk(__be32 *p, __be32 *end)
-{
-	__be32 *next;
-
-	if (*p++ != xdr_zero) {
-		next = p + 1 + be32_to_cpup(p) * rpcrdma_segment_maxsz;
-		if (next > end)
-			return NULL;
-		p = next;
-	}
-	return p;
-}
-
-/**
- * svc_rdma_xdr_decode_req - Parse incoming RPC-over-RDMA header
- * @rq_arg: Receive buffer
- *
- * On entry, xdr->head[0].iov_base points to first byte in the
- * RPC-over-RDMA header.
- *
- * On successful exit, head[0] points to first byte past the
- * RPC-over-RDMA header. For RDMA_MSG, this is the RPC message.
- * The length of the RPC-over-RDMA header is returned.
- */
-int svc_rdma_xdr_decode_req(struct xdr_buf *rq_arg)
-{
-	__be32 *p, *end, *rdma_argp;
-	unsigned int hdr_len;
-
-	/* Verify that there's enough bytes for header + something */
-	if (rq_arg->len <= RPCRDMA_HDRLEN_ERR)
-		goto out_short;
-
-	rdma_argp = rq_arg->head[0].iov_base;
-	if (*(rdma_argp + 1) != rpcrdma_version)
-		goto out_version;
-
-	switch (*(rdma_argp + 3)) {
-	case rdma_msg:
-	case rdma_nomsg:
-		break;
-
-	case rdma_done:
-		goto out_drop;
-
-	case rdma_error:
-		goto out_drop;
-
-	default:
-		goto out_proc;
-	}
-
-	end = (__be32 *)((unsigned long)rdma_argp + rq_arg->len);
-	p = xdr_check_read_list(rdma_argp + 4, end);
-	if (!p)
-		goto out_inval;
-	p = xdr_check_write_list(p, end);
-	if (!p)
-		goto out_inval;
-	p = xdr_check_reply_chunk(p, end);
-	if (!p)
-		goto out_inval;
-	if (p > end)
-		goto out_inval;
-
-	rq_arg->head[0].iov_base = p;
-	hdr_len = (unsigned long)p - (unsigned long)rdma_argp;
-	rq_arg->head[0].iov_len -= hdr_len;
-	return hdr_len;
-
-out_short:
-	dprintk("svcrdma: header too short = %d\n", rq_arg->len);
-	return -EINVAL;
-
-out_version:
-	dprintk("svcrdma: bad xprt version: %u\n",
-		be32_to_cpup(rdma_argp + 1));
-	return -EPROTONOSUPPORT;
-
-out_drop:
-	dprintk("svcrdma: dropping RDMA_DONE/ERROR message\n");
-	return 0;
-
-out_proc:
-	dprintk("svcrdma: bad rdma procedure (%u)\n",
-		be32_to_cpup(rdma_argp + 3));
-	return -EINVAL;
-
-out_inval:
-	dprintk("svcrdma: failed to parse transport header\n");
-	return -EINVAL;
-}
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 27a99bf5b1a6..ad4bd62eebf1 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2016, 2017 Oracle. All rights reserved.
  * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved.
  * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
  *
@@ -40,12 +41,66 @@
  * Author: Tom Tucker <tom@opengridcomputing.com>
  */
 
-#include <linux/sunrpc/debug.h>
-#include <linux/sunrpc/rpc_rdma.h>
-#include <linux/spinlock.h>
+/* Operation
+ *
+ * The main entry point is svc_rdma_recvfrom. This is called from
+ * svc_recv when the transport indicates there is incoming data to
+ * be read. "Data Ready" is signaled when an RDMA Receive completes,
+ * or when a set of RDMA Reads complete.
+ *
+ * An svc_rqst is passed in. This structure contains an array of
+ * free pages (rq_pages) that will contain the incoming RPC message.
+ *
+ * Short messages are moved directly into svc_rqst::rq_arg, and
+ * the RPC Call is ready to be processed by the Upper Layer.
+ * svc_rdma_recvfrom returns the length of the RPC Call message,
+ * completing the reception of the RPC Call.
+ *
+ * However, when an incoming message has Read chunks,
+ * svc_rdma_recvfrom must post RDMA Reads to pull the RPC Call's
+ * data payload from the client. svc_rdma_recvfrom sets up the
+ * RDMA Reads using pages in svc_rqst::rq_pages, which are
+ * transferred to an svc_rdma_op_ctxt for the duration of the
+ * I/O. svc_rdma_recvfrom then returns zero, since the RPC message
+ * is still not yet ready.
+ *
+ * When the Read chunk payloads have become available on the
+ * server, "Data Ready" is raised again, and svc_recv calls
+ * svc_rdma_recvfrom again. This second call may use a different
+ * svc_rqst than the first one, thus any information that needs
+ * to be preserved across these two calls is kept in an
+ * svc_rdma_op_ctxt.
+ *
+ * The second call to svc_rdma_recvfrom performs final assembly
+ * of the RPC Call message, using the RDMA Read sink pages kept in
+ * the svc_rdma_op_ctxt. The xdr_buf is copied from the
+ * svc_rdma_op_ctxt to the second svc_rqst. The second call returns
+ * the length of the completed RPC Call message.
+ *
+ * Page Management
+ *
+ * Pages under I/O must be transferred from the first svc_rqst to an
+ * svc_rdma_op_ctxt before the first svc_rdma_recvfrom call returns.
+ *
+ * The first svc_rqst supplies pages for RDMA Reads. These are moved
+ * from rqstp::rq_pages into ctxt::pages. The consumed elements of
+ * the rq_pages array are set to NULL and refilled with the first
+ * svc_rdma_recvfrom call returns.
+ *
+ * During the second svc_rdma_recvfrom call, RDMA Read sink pages
+ * are transferred from the svc_rdma_op_ctxt to the second svc_rqst
+ * (see rdma_read_complete() below).
+ */
+
 #include <asm/unaligned.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/rdma_cm.h>
+
+#include <linux/spinlock.h>
+
+#include <linux/sunrpc/xdr.h>
+#include <linux/sunrpc/debug.h>
+#include <linux/sunrpc/rpc_rdma.h>
 #include <linux/sunrpc/svc_rdma.h>
 
 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
@@ -59,7 +114,6 @@ static void rdma_build_arg_xdr(struct svc_rqst *rqstp,
 			       struct svc_rdma_op_ctxt *ctxt,
 			       u32 byte_count)
 {
-	struct rpcrdma_msg *rmsgp;
 	struct page *page;
 	u32 bc;
 	int sge_no;
@@ -83,20 +137,12 @@ static void rdma_build_arg_xdr(struct svc_rqst *rqstp,
 	rqstp->rq_arg.page_len = bc;
 	rqstp->rq_arg.page_base = 0;
 
-	/* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */
-	rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
-	if (rmsgp->rm_type == rdma_nomsg)
-		rqstp->rq_arg.pages = &rqstp->rq_pages[0];
-	else
-		rqstp->rq_arg.pages = &rqstp->rq_pages[1];
-
 	sge_no = 1;
 	while (bc && sge_no < ctxt->count) {
 		page = ctxt->pages[sge_no];
 		put_page(rqstp->rq_pages[sge_no]);
 		rqstp->rq_pages[sge_no] = page;
 		bc -= min_t(u32, bc, ctxt->sge[sge_no].length);
-		rqstp->rq_arg.buflen += ctxt->sge[sge_no].length;
 		sge_no++;
 	}
 	rqstp->rq_respages = &rqstp->rq_pages[sge_no];
@@ -115,406 +161,208 @@ static void rdma_build_arg_xdr(struct svc_rqst *rqstp,
 	rqstp->rq_arg.tail[0].iov_len = 0;
 }
 
-/* Issue an RDMA_READ using the local lkey to map the data sink */
-int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt,
-			struct svc_rqst *rqstp,
-			struct svc_rdma_op_ctxt *head,
-			int *page_no,
-			u32 *page_offset,
-			u32 rs_handle,
-			u32 rs_length,
-			u64 rs_offset,
-			bool last)
-{
-	struct ib_rdma_wr read_wr;
-	int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT;
-	struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt);
-	int ret, read, pno;
-	u32 pg_off = *page_offset;
-	u32 pg_no = *page_no;
-
-	ctxt->direction = DMA_FROM_DEVICE;
-	ctxt->read_hdr = head;
-	pages_needed = min_t(int, pages_needed, xprt->sc_max_sge_rd);
-	read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset,
-		     rs_length);
-
-	for (pno = 0; pno < pages_needed; pno++) {
-		int len = min_t(int, rs_length, PAGE_SIZE - pg_off);
-
-		head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no];
-		head->arg.page_len += len;
-
-		head->arg.len += len;
-		if (!pg_off)
-			head->count++;
-		rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1];
-		rqstp->rq_next_page = rqstp->rq_respages + 1;
-		ctxt->sge[pno].addr =
-			ib_dma_map_page(xprt->sc_cm_id->device,
-					head->arg.pages[pg_no], pg_off,
-					PAGE_SIZE - pg_off,
-					DMA_FROM_DEVICE);
-		ret = ib_dma_mapping_error(xprt->sc_cm_id->device,
-					   ctxt->sge[pno].addr);
-		if (ret)
-			goto err;
-		svc_rdma_count_mappings(xprt, ctxt);
-
-		ctxt->sge[pno].lkey = xprt->sc_pd->local_dma_lkey;
-		ctxt->sge[pno].length = len;
-		ctxt->count++;
-
-		/* adjust offset and wrap to next page if needed */
-		pg_off += len;
-		if (pg_off == PAGE_SIZE) {
-			pg_off = 0;
-			pg_no++;
-		}
-		rs_length -= len;
-	}
-
-	if (last && rs_length == 0)
-		set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
-	else
-		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
-
-	memset(&read_wr, 0, sizeof(read_wr));
-	ctxt->cqe.done = svc_rdma_wc_read;
-	read_wr.wr.wr_cqe = &ctxt->cqe;
-	read_wr.wr.opcode = IB_WR_RDMA_READ;
-	read_wr.wr.send_flags = IB_SEND_SIGNALED;
-	read_wr.rkey = rs_handle;
-	read_wr.remote_addr = rs_offset;
-	read_wr.wr.sg_list = ctxt->sge;
-	read_wr.wr.num_sge = pages_needed;
-
-	ret = svc_rdma_send(xprt, &read_wr.wr);
-	if (ret) {
-		pr_err("svcrdma: Error %d posting RDMA_READ\n", ret);
-		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
-		goto err;
-	}
+/* This accommodates the largest possible Write chunk,
+ * in one segment.
+ */
+#define MAX_BYTES_WRITE_SEG	((u32)(RPCSVC_MAXPAGES << PAGE_SHIFT))
 
-	/* return current location in page array */
-	*page_no = pg_no;
-	*page_offset = pg_off;
-	ret = read;
-	atomic_inc(&rdma_stat_read);
-	return ret;
- err:
-	svc_rdma_unmap_dma(ctxt);
-	svc_rdma_put_context(ctxt, 0);
-	return ret;
-}
+/* This accommodates the largest possible Position-Zero
+ * Read chunk or Reply chunk, in one segment.
+ */
+#define MAX_BYTES_SPECIAL_SEG	((u32)((RPCSVC_MAXPAGES + 2) << PAGE_SHIFT))
 
-/* Issue an RDMA_READ using an FRMR to map the data sink */
-int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
-			 struct svc_rqst *rqstp,
-			 struct svc_rdma_op_ctxt *head,
-			 int *page_no,
-			 u32 *page_offset,
-			 u32 rs_handle,
-			 u32 rs_length,
-			 u64 rs_offset,
-			 bool last)
+/* Sanity check the Read list.
+ *
+ * Implementation limits:
+ * - This implementation supports only one Read chunk.
+ *
+ * Sanity checks:
+ * - Read list does not overflow buffer.
+ * - Segment size limited by largest NFS data payload.
+ *
+ * The segment count is limited to how many segments can
+ * fit in the transport header without overflowing the
+ * buffer. That's about 40 Read segments for a 1KB inline
+ * threshold.
+ *
+ * Returns pointer to the following Write list.
+ */
+static __be32 *xdr_check_read_list(__be32 *p, const __be32 *end)
 {
-	struct ib_rdma_wr read_wr;
-	struct ib_send_wr inv_wr;
-	struct ib_reg_wr reg_wr;
-	u8 key;
-	int nents = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT;
-	struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt);
-	struct svc_rdma_fastreg_mr *frmr = svc_rdma_get_frmr(xprt);
-	int ret, read, pno, dma_nents, n;
-	u32 pg_off = *page_offset;
-	u32 pg_no = *page_no;
-
-	if (IS_ERR(frmr))
-		return -ENOMEM;
-
-	ctxt->direction = DMA_FROM_DEVICE;
-	ctxt->frmr = frmr;
-	nents = min_t(unsigned int, nents, xprt->sc_frmr_pg_list_len);
-	read = min_t(int, (nents << PAGE_SHIFT) - *page_offset, rs_length);
-
-	frmr->direction = DMA_FROM_DEVICE;
-	frmr->access_flags = (IB_ACCESS_LOCAL_WRITE|IB_ACCESS_REMOTE_WRITE);
-	frmr->sg_nents = nents;
-
-	for (pno = 0; pno < nents; pno++) {
-		int len = min_t(int, rs_length, PAGE_SIZE - pg_off);
-
-		head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no];
-		head->arg.page_len += len;
-		head->arg.len += len;
-		if (!pg_off)
-			head->count++;
-
-		sg_set_page(&frmr->sg[pno], rqstp->rq_arg.pages[pg_no],
-			    len, pg_off);
-
-		rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1];
-		rqstp->rq_next_page = rqstp->rq_respages + 1;
-
-		/* adjust offset and wrap to next page if needed */
-		pg_off += len;
-		if (pg_off == PAGE_SIZE) {
-			pg_off = 0;
-			pg_no++;
+	u32 position;
+	bool first;
+
+	first = true;
+	while (*p++ != xdr_zero) {
+		if (first) {
+			position = be32_to_cpup(p++);
+			first = false;
+		} else if (be32_to_cpup(p++) != position) {
+			return NULL;
 		}
-		rs_length -= len;
-	}
+		p++;	/* handle */
+		if (be32_to_cpup(p++) > MAX_BYTES_SPECIAL_SEG)
+			return NULL;
+		p += 2;	/* offset */
 
-	if (last && rs_length == 0)
-		set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
-	else
-		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
-
-	dma_nents = ib_dma_map_sg(xprt->sc_cm_id->device,
-				  frmr->sg, frmr->sg_nents,
-				  frmr->direction);
-	if (!dma_nents) {
-		pr_err("svcrdma: failed to dma map sg %p\n",
-		       frmr->sg);
-		return -ENOMEM;
+		if (p > end)
+			return NULL;
 	}
+	return p;
+}
 
-	n = ib_map_mr_sg(frmr->mr, frmr->sg, frmr->sg_nents, NULL, PAGE_SIZE);
-	if (unlikely(n != frmr->sg_nents)) {
-		pr_err("svcrdma: failed to map mr %p (%d/%d elements)\n",
-		       frmr->mr, n, frmr->sg_nents);
-		return n < 0 ? n : -EINVAL;
-	}
+/* The segment count is limited to how many segments can
+ * fit in the transport header without overflowing the
+ * buffer. That's about 60 Write segments for a 1KB inline
+ * threshold.
+ */
+static __be32 *xdr_check_write_chunk(__be32 *p, const __be32 *end,
+				     u32 maxlen)
+{
+	u32 i, segcount;
 
-	/* Bump the key */
-	key = (u8)(frmr->mr->lkey & 0x000000FF);
-	ib_update_fast_reg_key(frmr->mr, ++key);
-
-	ctxt->sge[0].addr = frmr->mr->iova;
-	ctxt->sge[0].lkey = frmr->mr->lkey;
-	ctxt->sge[0].length = frmr->mr->length;
-	ctxt->count = 1;
-	ctxt->read_hdr = head;
-
-	/* Prepare REG WR */
-	ctxt->reg_cqe.done = svc_rdma_wc_reg;
-	reg_wr.wr.wr_cqe = &ctxt->reg_cqe;
-	reg_wr.wr.opcode = IB_WR_REG_MR;
-	reg_wr.wr.send_flags = IB_SEND_SIGNALED;
-	reg_wr.wr.num_sge = 0;
-	reg_wr.mr = frmr->mr;
-	reg_wr.key = frmr->mr->lkey;
-	reg_wr.access = frmr->access_flags;
-	reg_wr.wr.next = &read_wr.wr;
-
-	/* Prepare RDMA_READ */
-	memset(&read_wr, 0, sizeof(read_wr));
-	ctxt->cqe.done = svc_rdma_wc_read;
-	read_wr.wr.wr_cqe = &ctxt->cqe;
-	read_wr.wr.send_flags = IB_SEND_SIGNALED;
-	read_wr.rkey = rs_handle;
-	read_wr.remote_addr = rs_offset;
-	read_wr.wr.sg_list = ctxt->sge;
-	read_wr.wr.num_sge = 1;
-	if (xprt->sc_dev_caps & SVCRDMA_DEVCAP_READ_W_INV) {
-		read_wr.wr.opcode = IB_WR_RDMA_READ_WITH_INV;
-		read_wr.wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey;
-	} else {
-		read_wr.wr.opcode = IB_WR_RDMA_READ;
-		read_wr.wr.next = &inv_wr;
-		/* Prepare invalidate */
-		memset(&inv_wr, 0, sizeof(inv_wr));
-		ctxt->inv_cqe.done = svc_rdma_wc_inv;
-		inv_wr.wr_cqe = &ctxt->inv_cqe;
-		inv_wr.opcode = IB_WR_LOCAL_INV;
-		inv_wr.send_flags = IB_SEND_SIGNALED | IB_SEND_FENCE;
-		inv_wr.ex.invalidate_rkey = frmr->mr->lkey;
-	}
+	segcount = be32_to_cpup(p++);
+	for (i = 0; i < segcount; i++) {
+		p++;	/* handle */
+		if (be32_to_cpup(p++) > maxlen)
+			return NULL;
+		p += 2;	/* offset */
 
-	/* Post the chain */
-	ret = svc_rdma_send(xprt, &reg_wr.wr);
-	if (ret) {
-		pr_err("svcrdma: Error %d posting RDMA_READ\n", ret);
-		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
-		goto err;
+		if (p > end)
+			return NULL;
 	}
 
-	/* return current location in page array */
-	*page_no = pg_no;
-	*page_offset = pg_off;
-	ret = read;
-	atomic_inc(&rdma_stat_read);
-	return ret;
- err:
-	svc_rdma_put_context(ctxt, 0);
-	svc_rdma_put_frmr(xprt, frmr);
-	return ret;
-}
-
-static unsigned int
-rdma_rcl_chunk_count(struct rpcrdma_read_chunk *ch)
-{
-	unsigned int count;
-
-	for (count = 0; ch->rc_discrim != xdr_zero; ch++)
-		count++;
-	return count;
+	return p;
 }
 
-/* If there was additional inline content, append it to the end of arg.pages.
- * Tail copy has to be done after the reader function has determined how many
- * pages are needed for RDMA READ.
+/* Sanity check the Write list.
+ *
+ * Implementation limits:
+ * - This implementation supports only one Write chunk.
+ *
+ * Sanity checks:
+ * - Write list does not overflow buffer.
+ * - Segment size limited by largest NFS data payload.
+ *
+ * Returns pointer to the following Reply chunk.
  */
-static int
-rdma_copy_tail(struct svc_rqst *rqstp, struct svc_rdma_op_ctxt *head,
-	       u32 position, u32 byte_count, u32 page_offset, int page_no)
+static __be32 *xdr_check_write_list(__be32 *p, const __be32 *end)
 {
-	char *srcp, *destp;
-
-	srcp = head->arg.head[0].iov_base + position;
-	byte_count = head->arg.head[0].iov_len - position;
-	if (byte_count > PAGE_SIZE) {
-		dprintk("svcrdma: large tail unsupported\n");
-		return 0;
-	}
-
-	/* Fit as much of the tail on the current page as possible */
-	if (page_offset != PAGE_SIZE) {
-		destp = page_address(rqstp->rq_arg.pages[page_no]);
-		destp += page_offset;
-		while (byte_count--) {
-			*destp++ = *srcp++;
-			page_offset++;
-			if (page_offset == PAGE_SIZE && byte_count)
-				goto more;
-		}
-		goto done;
+	u32 chcount;
+
+	chcount = 0;
+	while (*p++ != xdr_zero) {
+		p = xdr_check_write_chunk(p, end, MAX_BYTES_WRITE_SEG);
+		if (!p)
+			return NULL;
+		if (chcount++ > 1)
+			return NULL;
 	}
-
-more:
-	/* Fit the rest on the next page */
-	page_no++;
-	destp = page_address(rqstp->rq_arg.pages[page_no]);
-	while (byte_count--)
-		*destp++ = *srcp++;
-
-	rqstp->rq_respages = &rqstp->rq_arg.pages[page_no+1];
-	rqstp->rq_next_page = rqstp->rq_respages + 1;
-
-done:
-	byte_count = head->arg.head[0].iov_len - position;
-	head->arg.page_len += byte_count;
-	head->arg.len += byte_count;
-	head->arg.buflen += byte_count;
-	return 1;
+	return p;
 }
 
-/* Returns the address of the first read chunk or <nul> if no read chunk
- * is present
+/* Sanity check the Reply chunk.
+ *
+ * Sanity checks:
+ * - Reply chunk does not overflow buffer.
+ * - Segment size limited by largest NFS data payload.
+ *
+ * Returns pointer to the following RPC header.
  */
-static struct rpcrdma_read_chunk *
-svc_rdma_get_read_chunk(struct rpcrdma_msg *rmsgp)
+static __be32 *xdr_check_reply_chunk(__be32 *p, const __be32 *end)
 {
-	struct rpcrdma_read_chunk *ch =
-		(struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
-
-	if (ch->rc_discrim == xdr_zero)
-		return NULL;
-	return ch;
+	if (*p++ != xdr_zero) {
+		p = xdr_check_write_chunk(p, end, MAX_BYTES_SPECIAL_SEG);
+		if (!p)
+			return NULL;
+	}
+	return p;
 }
 
-static int rdma_read_chunks(struct svcxprt_rdma *xprt,
-			    struct rpcrdma_msg *rmsgp,
-			    struct svc_rqst *rqstp,
-			    struct svc_rdma_op_ctxt *head)
+/* On entry, xdr->head[0].iov_base points to first byte in the
+ * RPC-over-RDMA header.
+ *
+ * On successful exit, head[0] points to first byte past the
+ * RPC-over-RDMA header. For RDMA_MSG, this is the RPC message.
+ * The length of the RPC-over-RDMA header is returned.
+ *
+ * Assumptions:
+ * - The transport header is entirely contained in the head iovec.
+ */
+static int svc_rdma_xdr_decode_req(struct xdr_buf *rq_arg)
 {
-	int page_no, ret;
-	struct rpcrdma_read_chunk *ch;
-	u32 handle, page_offset, byte_count;
-	u32 position;
-	u64 rs_offset;
-	bool last;
-
-	/* If no read list is present, return 0 */
-	ch = svc_rdma_get_read_chunk(rmsgp);
-	if (!ch)
-		return 0;
+	__be32 *p, *end, *rdma_argp;
+	unsigned int hdr_len;
+	char *proc;
+
+	/* Verify that there's enough bytes for header + something */
+	if (rq_arg->len <= RPCRDMA_HDRLEN_ERR)
+		goto out_short;
+
+	rdma_argp = rq_arg->head[0].iov_base;
+	if (*(rdma_argp + 1) != rpcrdma_version)
+		goto out_version;
+
+	switch (*(rdma_argp + 3)) {
+	case rdma_msg:
+		proc = "RDMA_MSG";
+		break;
+	case rdma_nomsg:
+		proc = "RDMA_NOMSG";
+		break;
+
+	case rdma_done:
+		goto out_drop;
 
-	if (rdma_rcl_chunk_count(ch) > RPCSVC_MAXPAGES)
-		return -EINVAL;
-
-	/* The request is completed when the RDMA_READs complete. The
-	 * head context keeps all the pages that comprise the
-	 * request.
-	 */
-	head->arg.head[0] = rqstp->rq_arg.head[0];
-	head->arg.tail[0] = rqstp->rq_arg.tail[0];
-	head->hdr_count = head->count;
-	head->arg.page_base = 0;
-	head->arg.page_len = 0;
-	head->arg.len = rqstp->rq_arg.len;
-	head->arg.buflen = rqstp->rq_arg.buflen;
-
-	/* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */
-	position = be32_to_cpu(ch->rc_position);
-	if (position == 0) {
-		head->arg.pages = &head->pages[0];
-		page_offset = head->byte_len;
-	} else {
-		head->arg.pages = &head->pages[head->count];
-		page_offset = 0;
-	}
+	case rdma_error:
+		goto out_drop;
 
-	ret = 0;
-	page_no = 0;
-	for (; ch->rc_discrim != xdr_zero; ch++) {
-		if (be32_to_cpu(ch->rc_position) != position)
-			goto err;
-
-		handle = be32_to_cpu(ch->rc_target.rs_handle),
-		byte_count = be32_to_cpu(ch->rc_target.rs_length);
-		xdr_decode_hyper((__be32 *)&ch->rc_target.rs_offset,
-				 &rs_offset);
-
-		while (byte_count > 0) {
-			last = (ch + 1)->rc_discrim == xdr_zero;
-			ret = xprt->sc_reader(xprt, rqstp, head,
-					      &page_no, &page_offset,
-					      handle, byte_count,
-					      rs_offset, last);
-			if (ret < 0)
-				goto err;
-			byte_count -= ret;
-			rs_offset += ret;
-			head->arg.buflen += ret;
-		}
+	default:
+		goto out_proc;
 	}
 
-	/* Read list may need XDR round-up (see RFC 5666, s. 3.7) */
-	if (page_offset & 3) {
-		u32 pad = 4 - (page_offset & 3);
-
-		head->arg.tail[0].iov_len += pad;
-		head->arg.len += pad;
-		head->arg.buflen += pad;
-		page_offset += pad;
-	}
+	end = (__be32 *)((unsigned long)rdma_argp + rq_arg->len);
+	p = xdr_check_read_list(rdma_argp + 4, end);
+	if (!p)
+		goto out_inval;
+	p = xdr_check_write_list(p, end);
+	if (!p)
+		goto out_inval;
+	p = xdr_check_reply_chunk(p, end);
+	if (!p)
+		goto out_inval;
+	if (p > end)
+		goto out_inval;
+
+	rq_arg->head[0].iov_base = p;
+	hdr_len = (unsigned long)p - (unsigned long)rdma_argp;
+	rq_arg->head[0].iov_len -= hdr_len;
+	rq_arg->len -= hdr_len;
+	dprintk("svcrdma: received %s request for XID 0x%08x, hdr_len=%u\n",
+		proc, be32_to_cpup(rdma_argp), hdr_len);
+	return hdr_len;
+
+out_short:
+	dprintk("svcrdma: header too short = %d\n", rq_arg->len);
+	return -EINVAL;
+
+out_version:
+	dprintk("svcrdma: bad xprt version: %u\n",
+		be32_to_cpup(rdma_argp + 1));
+	return -EPROTONOSUPPORT;
 
-	ret = 1;
-	if (position && position < head->arg.head[0].iov_len)
-		ret = rdma_copy_tail(rqstp, head, position,
-				     byte_count, page_offset, page_no);
-	head->arg.head[0].iov_len = position;
-	head->position = position;
+out_drop:
+	dprintk("svcrdma: dropping RDMA_DONE/ERROR message\n");
+	return 0;
 
- err:
-	/* Detach arg pages. svc_recv will replenish them */
-	for (page_no = 0;
-	     &rqstp->rq_pages[page_no] < rqstp->rq_respages; page_no++)
-		rqstp->rq_pages[page_no] = NULL;
+out_proc:
+	dprintk("svcrdma: bad rdma procedure (%u)\n",
+		be32_to_cpup(rdma_argp + 3));
+	return -EINVAL;
 
-	return ret;
+out_inval:
+	dprintk("svcrdma: failed to parse transport header\n");
+	return -EINVAL;
 }
 
 static void rdma_read_complete(struct svc_rqst *rqstp,
@@ -528,24 +376,9 @@ static void rdma_read_complete(struct svc_rqst *rqstp,
 		rqstp->rq_pages[page_no] = head->pages[page_no];
 	}
 
-	/* Adjustments made for RDMA_NOMSG type requests */
-	if (head->position == 0) {
-		if (head->arg.len <= head->sge[0].length) {
-			head->arg.head[0].iov_len = head->arg.len -
-							head->byte_len;
-			head->arg.page_len = 0;
-		} else {
-			head->arg.head[0].iov_len = head->sge[0].length -
-								head->byte_len;
-			head->arg.page_len = head->arg.len -
-						head->sge[0].length;
-		}
-	}
-
 	/* Point rq_arg.pages past header */
 	rqstp->rq_arg.pages = &rqstp->rq_pages[head->hdr_count];
 	rqstp->rq_arg.page_len = head->arg.page_len;
-	rqstp->rq_arg.page_base = head->arg.page_base;
 
 	/* rq_respages starts after the last arg page */
 	rqstp->rq_respages = &rqstp->rq_pages[page_no];
@@ -642,21 +475,44 @@ static bool svc_rdma_is_backchannel_reply(struct svc_xprt *xprt,
 	return true;
 }
 
-/*
- * Set up the rqstp thread context to point to the RQ buffer. If
- * necessary, pull additional data from the client with an RDMA_READ
- * request.
+/**
+ * svc_rdma_recvfrom - Receive an RPC call
+ * @rqstp: request structure into which to receive an RPC Call
+ *
+ * Returns:
+ *	The positive number of bytes in the RPC Call message,
+ *	%0 if there were no Calls ready to return,
+ *	%-EINVAL if the Read chunk data is too large,
+ *	%-ENOMEM if rdma_rw context pool was exhausted,
+ *	%-ENOTCONN if posting failed (connection is lost),
+ *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
+ *
+ * Called in a loop when XPT_DATA is set. XPT_DATA is cleared only
+ * when there are no remaining ctxt's to process.
+ *
+ * The next ctxt is removed from the "receive" lists.
+ *
+ * - If the ctxt completes a Read, then finish assembling the Call
+ *   message and return the number of bytes in the message.
+ *
+ * - If the ctxt completes a Receive, then construct the Call
+ *   message from the contents of the Receive buffer.
+ *
+ *   - If there are no Read chunks in this message, then finish
+ *     assembling the Call message and return the number of bytes
+ *     in the message.
+ *
+ *   - If there are Read chunks in this message, post Read WRs to
+ *     pull that payload and return 0.
  */
 int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 {
 	struct svc_xprt *xprt = rqstp->rq_xprt;
 	struct svcxprt_rdma *rdma_xprt =
 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
-	struct svc_rdma_op_ctxt *ctxt = NULL;
-	struct rpcrdma_msg *rmsgp;
-	int ret = 0;
-
-	dprintk("svcrdma: rqstp=%p\n", rqstp);
+	struct svc_rdma_op_ctxt *ctxt;
+	__be32 *p;
+	int ret;
 
 	spin_lock(&rdma_xprt->sc_rq_dto_lock);
 	if (!list_empty(&rdma_xprt->sc_read_complete_q)) {
@@ -671,22 +527,14 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 					struct svc_rdma_op_ctxt, list);
 		list_del(&ctxt->list);
 	} else {
-		atomic_inc(&rdma_stat_rq_starve);
+		/* No new incoming requests, terminate the loop */
 		clear_bit(XPT_DATA, &xprt->xpt_flags);
-		ctxt = NULL;
+		spin_unlock(&rdma_xprt->sc_rq_dto_lock);
+		return 0;
 	}
 	spin_unlock(&rdma_xprt->sc_rq_dto_lock);
-	if (!ctxt) {
-		/* This is the EAGAIN path. The svc_recv routine will
-		 * return -EAGAIN, the nfsd thread will go to call into
-		 * svc_recv again and we shouldn't be on the active
-		 * transport list
-		 */
-		if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
-			goto defer;
-		goto out;
-	}
-	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p\n",
+
+	dprintk("svcrdma: recvfrom: ctxt=%p on xprt=%p, rqstp=%p\n",
 		ctxt, rdma_xprt, rqstp);
 	atomic_inc(&rdma_stat_recv);
 
@@ -694,7 +542,7 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 	rdma_build_arg_xdr(rqstp, ctxt, ctxt->byte_len);
 
 	/* Decode the RDMA header. */
-	rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
+	p = (__be32 *)rqstp->rq_arg.head[0].iov_base;
 	ret = svc_rdma_xdr_decode_req(&rqstp->rq_arg);
 	if (ret < 0)
 		goto out_err;
@@ -702,9 +550,8 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 		goto out_drop;
 	rqstp->rq_xprt_hlen = ret;
 
-	if (svc_rdma_is_backchannel_reply(xprt, &rmsgp->rm_xid)) {
-		ret = svc_rdma_handle_bc_reply(xprt->xpt_bc_xprt,
-					       &rmsgp->rm_xid,
+	if (svc_rdma_is_backchannel_reply(xprt, p)) {
+		ret = svc_rdma_handle_bc_reply(xprt->xpt_bc_xprt, p,
 					       &rqstp->rq_arg);
 		svc_rdma_put_context(ctxt, 0);
 		if (ret)
@@ -712,39 +559,34 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 		return ret;
 	}
 
-	/* Read read-list data. */
-	ret = rdma_read_chunks(rdma_xprt, rmsgp, rqstp, ctxt);
-	if (ret > 0) {
-		/* read-list posted, defer until data received from client. */
-		goto defer;
-	} else if (ret < 0) {
-		/* Post of read-list failed, free context. */
-		svc_rdma_put_context(ctxt, 1);
-		return 0;
-	}
+	p += rpcrdma_fixed_maxsz;
+	if (*p != xdr_zero)
+		goto out_readchunk;
 
 complete:
-	ret = rqstp->rq_arg.head[0].iov_len
-		+ rqstp->rq_arg.page_len
-		+ rqstp->rq_arg.tail[0].iov_len;
 	svc_rdma_put_context(ctxt, 0);
- out:
-	dprintk("svcrdma: ret=%d, rq_arg.len=%u, "
-		"rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len=%zd\n",
-		ret, rqstp->rq_arg.len,
-		rqstp->rq_arg.head[0].iov_base,
-		rqstp->rq_arg.head[0].iov_len);
+	dprintk("svcrdma: recvfrom: xprt=%p, rqstp=%p, rq_arg.len=%u\n",
+		rdma_xprt, rqstp, rqstp->rq_arg.len);
 	rqstp->rq_prot = IPPROTO_MAX;
 	svc_xprt_copy_addrs(rqstp, xprt);
-	return ret;
+	return rqstp->rq_arg.len;
+
+out_readchunk:
+	ret = svc_rdma_recv_read_chunk(rdma_xprt, rqstp, ctxt, p);
+	if (ret < 0)
+		goto out_postfail;
+	return 0;
 
 out_err:
-	svc_rdma_send_error(rdma_xprt, &rmsgp->rm_xid, ret);
+	svc_rdma_send_error(rdma_xprt, p, ret);
 	svc_rdma_put_context(ctxt, 0);
 	return 0;
 
-defer:
-	return 0;
+out_postfail:
+	if (ret == -EINVAL)
+		svc_rdma_send_error(rdma_xprt, p, ret);
+	svc_rdma_put_context(ctxt, 1);
+	return ret;
 
 out_drop:
 	svc_rdma_put_context(ctxt, 1);
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index 0cf620277693..933f79bed270 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -12,6 +12,9 @@
 
 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
 
+static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc);
+static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc);
+
 /* Each R/W context contains state for one chain of RDMA Read or
  * Write Work Requests.
  *
@@ -113,22 +116,20 @@ struct svc_rdma_chunk_ctxt {
 	struct svcxprt_rdma	*cc_rdma;
 	struct list_head	cc_rwctxts;
 	int			cc_sqecount;
-	enum dma_data_direction cc_dir;
 };
 
 static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
-			     struct svc_rdma_chunk_ctxt *cc,
-			     enum dma_data_direction dir)
+			     struct svc_rdma_chunk_ctxt *cc)
 {
 	cc->cc_rdma = rdma;
 	svc_xprt_get(&rdma->sc_xprt);
 
 	INIT_LIST_HEAD(&cc->cc_rwctxts);
 	cc->cc_sqecount = 0;
-	cc->cc_dir = dir;
 }
 
-static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc)
+static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc,
+				enum dma_data_direction dir)
 {
 	struct svcxprt_rdma *rdma = cc->cc_rdma;
 	struct svc_rdma_rw_ctxt *ctxt;
@@ -138,7 +139,7 @@ static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc)
 
 		rdma_rw_ctx_destroy(&ctxt->rw_ctx, rdma->sc_qp,
 				    rdma->sc_port_num, ctxt->rw_sg_table.sgl,
-				    ctxt->rw_nents, cc->cc_dir);
+				    ctxt->rw_nents, dir);
 		svc_rdma_put_rw_ctxt(rdma, ctxt);
 	}
 	svc_xprt_put(&rdma->sc_xprt);
@@ -176,13 +177,14 @@ svc_rdma_write_info_alloc(struct svcxprt_rdma *rdma, __be32 *chunk)
 	info->wi_seg_no = 0;
 	info->wi_nsegs = be32_to_cpup(++chunk);
 	info->wi_segs = ++chunk;
-	svc_rdma_cc_init(rdma, &info->wi_cc, DMA_TO_DEVICE);
+	svc_rdma_cc_init(rdma, &info->wi_cc);
+	info->wi_cc.cc_cqe.done = svc_rdma_write_done;
 	return info;
 }
 
 static void svc_rdma_write_info_free(struct svc_rdma_write_info *info)
 {
-	svc_rdma_cc_release(&info->wi_cc);
+	svc_rdma_cc_release(&info->wi_cc, DMA_TO_DEVICE);
 	kfree(info);
 }
 
@@ -216,6 +218,76 @@ static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
 	svc_rdma_write_info_free(info);
 }
 
+/* State for pulling a Read chunk.
+ */
+struct svc_rdma_read_info {
+	struct svc_rdma_op_ctxt		*ri_readctxt;
+	unsigned int			ri_position;
+	unsigned int			ri_pageno;
+	unsigned int			ri_pageoff;
+	unsigned int			ri_chunklen;
+
+	struct svc_rdma_chunk_ctxt	ri_cc;
+};
+
+static struct svc_rdma_read_info *
+svc_rdma_read_info_alloc(struct svcxprt_rdma *rdma)
+{
+	struct svc_rdma_read_info *info;
+
+	info = kmalloc(sizeof(*info), GFP_KERNEL);
+	if (!info)
+		return info;
+
+	svc_rdma_cc_init(rdma, &info->ri_cc);
+	info->ri_cc.cc_cqe.done = svc_rdma_wc_read_done;
+	return info;
+}
+
+static void svc_rdma_read_info_free(struct svc_rdma_read_info *info)
+{
+	svc_rdma_cc_release(&info->ri_cc, DMA_FROM_DEVICE);
+	kfree(info);
+}
+
+/**
+ * svc_rdma_wc_read_done - Handle completion of an RDMA Read ctx
+ * @cq: controlling Completion Queue
+ * @wc: Work Completion
+ *
+ */
+static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct ib_cqe *cqe = wc->wr_cqe;
+	struct svc_rdma_chunk_ctxt *cc =
+			container_of(cqe, struct svc_rdma_chunk_ctxt, cc_cqe);
+	struct svcxprt_rdma *rdma = cc->cc_rdma;
+	struct svc_rdma_read_info *info =
+			container_of(cc, struct svc_rdma_read_info, ri_cc);
+
+	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
+	wake_up(&rdma->sc_send_wait);
+
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
+		if (wc->status != IB_WC_WR_FLUSH_ERR)
+			pr_err("svcrdma: read ctx: %s (%u/0x%x)\n",
+			       ib_wc_status_msg(wc->status),
+			       wc->status, wc->vendor_err);
+		svc_rdma_put_context(info->ri_readctxt, 1);
+	} else {
+		spin_lock(&rdma->sc_rq_dto_lock);
+		list_add_tail(&info->ri_readctxt->list,
+			      &rdma->sc_read_complete_q);
+		spin_unlock(&rdma->sc_rq_dto_lock);
+
+		set_bit(XPT_DATA, &rdma->sc_xprt.xpt_flags);
+		svc_xprt_enqueue(&rdma->sc_xprt);
+	}
+
+	svc_rdma_read_info_free(info);
+}
+
 /* This function sleeps when the transport's Send Queue is congested.
  *
  * Assumptions:
@@ -232,6 +304,9 @@ static int svc_rdma_post_chunk_ctxt(struct svc_rdma_chunk_ctxt *cc)
 	struct ib_cqe *cqe;
 	int ret;
 
+	if (cc->cc_sqecount > rdma->sc_sq_depth)
+		return -EINVAL;
+
 	first_wr = NULL;
 	cqe = &cc->cc_cqe;
 	list_for_each(tmp, &cc->cc_rwctxts) {
@@ -295,8 +370,9 @@ static void svc_rdma_pagelist_to_sg(struct svc_rdma_write_info *info,
 	struct scatterlist *sg;
 	struct page **page;
 
-	page_off = (info->wi_next_off + xdr->page_base) & ~PAGE_MASK;
-	page_no = (info->wi_next_off + xdr->page_base) >> PAGE_SHIFT;
+	page_off = info->wi_next_off + xdr->page_base;
+	page_no = page_off >> PAGE_SHIFT;
+	page_off = offset_in_page(page_off);
 	page = xdr->pages + page_no;
 	info->wi_next_off += remaining;
 	sg = ctxt->rw_sg_table.sgl;
@@ -332,7 +408,6 @@ svc_rdma_build_writes(struct svc_rdma_write_info *info,
 	__be32 *seg;
 	int ret;
 
-	cc->cc_cqe.done = svc_rdma_write_done;
 	seg = info->wi_segs + info->wi_seg_no * rpcrdma_segment_maxsz;
 	do {
 		unsigned int write_len;
@@ -425,6 +500,7 @@ static int svc_rdma_send_xdr_pagelist(struct svc_rdma_write_info *info,
  *
  * Returns a non-negative number of bytes the chunk consumed, or
  *	%-E2BIG if the payload was larger than the Write chunk,
+ *	%-EINVAL if client provided too many segments,
  *	%-ENOMEM if rdma_rw context pool was exhausted,
  *	%-ENOTCONN if posting failed (connection is lost),
  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
@@ -465,6 +541,7 @@ out_err:
  *
  * Returns a non-negative number of bytes the chunk consumed, or
  *	%-E2BIG if the payload was larger than the Reply chunk,
+ *	%-EINVAL if client provided too many segments,
  *	%-ENOMEM if rdma_rw context pool was exhausted,
  *	%-ENOTCONN if posting failed (connection is lost),
  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
@@ -510,3 +587,353 @@ out_err:
 	svc_rdma_write_info_free(info);
 	return ret;
 }
+
+static int svc_rdma_build_read_segment(struct svc_rdma_read_info *info,
+				       struct svc_rqst *rqstp,
+				       u32 rkey, u32 len, u64 offset)
+{
+	struct svc_rdma_op_ctxt *head = info->ri_readctxt;
+	struct svc_rdma_chunk_ctxt *cc = &info->ri_cc;
+	struct svc_rdma_rw_ctxt *ctxt;
+	unsigned int sge_no, seg_len;
+	struct scatterlist *sg;
+	int ret;
+
+	sge_no = PAGE_ALIGN(info->ri_pageoff + len) >> PAGE_SHIFT;
+	ctxt = svc_rdma_get_rw_ctxt(cc->cc_rdma, sge_no);
+	if (!ctxt)
+		goto out_noctx;
+	ctxt->rw_nents = sge_no;
+
+	dprintk("svcrdma: reading segment %u@0x%016llx:0x%08x (%u sges)\n",
+		len, offset, rkey, sge_no);
+
+	sg = ctxt->rw_sg_table.sgl;
+	for (sge_no = 0; sge_no < ctxt->rw_nents; sge_no++) {
+		seg_len = min_t(unsigned int, len,
+				PAGE_SIZE - info->ri_pageoff);
+
+		head->arg.pages[info->ri_pageno] =
+			rqstp->rq_pages[info->ri_pageno];
+		if (!info->ri_pageoff)
+			head->count++;
+
+		sg_set_page(sg, rqstp->rq_pages[info->ri_pageno],
+			    seg_len, info->ri_pageoff);
+		sg = sg_next(sg);
+
+		info->ri_pageoff += seg_len;
+		if (info->ri_pageoff == PAGE_SIZE) {
+			info->ri_pageno++;
+			info->ri_pageoff = 0;
+		}
+		len -= seg_len;
+
+		/* Safety check */
+		if (len &&
+		    &rqstp->rq_pages[info->ri_pageno + 1] > rqstp->rq_page_end)
+			goto out_overrun;
+	}
+
+	ret = rdma_rw_ctx_init(&ctxt->rw_ctx, cc->cc_rdma->sc_qp,
+			       cc->cc_rdma->sc_port_num,
+			       ctxt->rw_sg_table.sgl, ctxt->rw_nents,
+			       0, offset, rkey, DMA_FROM_DEVICE);
+	if (ret < 0)
+		goto out_initerr;
+
+	list_add(&ctxt->rw_list, &cc->cc_rwctxts);
+	cc->cc_sqecount += ret;
+	return 0;
+
+out_noctx:
+	dprintk("svcrdma: no R/W ctxs available\n");
+	return -ENOMEM;
+
+out_overrun:
+	dprintk("svcrdma: request overruns rq_pages\n");
+	return -EINVAL;
+
+out_initerr:
+	svc_rdma_put_rw_ctxt(cc->cc_rdma, ctxt);
+	pr_err("svcrdma: failed to map pagelist (%d)\n", ret);
+	return -EIO;
+}
+
+static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
+				     struct svc_rdma_read_info *info,
+				     __be32 *p)
+{
+	int ret;
+
+	info->ri_chunklen = 0;
+	while (*p++ != xdr_zero) {
+		u32 rs_handle, rs_length;
+		u64 rs_offset;
+
+		if (be32_to_cpup(p++) != info->ri_position)
+			break;
+		rs_handle = be32_to_cpup(p++);
+		rs_length = be32_to_cpup(p++);
+		p = xdr_decode_hyper(p, &rs_offset);
+
+		ret = svc_rdma_build_read_segment(info, rqstp,
+						  rs_handle, rs_length,
+						  rs_offset);
+		if (ret < 0)
+			break;
+
+		info->ri_chunklen += rs_length;
+	}
+
+	return ret;
+}
+
+/* If there is inline content following the Read chunk, append it to
+ * the page list immediately following the data payload. This has to
+ * be done after the reader function has determined how many pages
+ * were consumed for RDMA Read.
+ *
+ * On entry, ri_pageno and ri_pageoff point directly to the end of the
+ * page list. On exit, both have been updated to the new "next byte".
+ *
+ * Assumptions:
+ *	- Inline content fits entirely in rq_pages[0]
+ *	- Trailing content is only a handful of bytes
+ */
+static int svc_rdma_copy_tail(struct svc_rqst *rqstp,
+			      struct svc_rdma_read_info *info)
+{
+	struct svc_rdma_op_ctxt *head = info->ri_readctxt;
+	unsigned int tail_length, remaining;
+	u8 *srcp, *destp;
+
+	/* Assert that all inline content fits in page 0. This is an
+	 * implementation limit, not a protocol limit.
+	 */
+	if (head->arg.head[0].iov_len > PAGE_SIZE) {
+		pr_warn_once("svcrdma: too much trailing inline content\n");
+		return -EINVAL;
+	}
+
+	srcp = head->arg.head[0].iov_base;
+	srcp += info->ri_position;
+	tail_length = head->arg.head[0].iov_len - info->ri_position;
+	remaining = tail_length;
+
+	/* If there is room on the last page in the page list, try to
+	 * fit the trailing content there.
+	 */
+	if (info->ri_pageoff > 0) {
+		unsigned int len;
+
+		len = min_t(unsigned int, remaining,
+			    PAGE_SIZE - info->ri_pageoff);
+		destp = page_address(rqstp->rq_pages[info->ri_pageno]);
+		destp += info->ri_pageoff;
+
+		memcpy(destp, srcp, len);
+		srcp += len;
+		destp += len;
+		info->ri_pageoff += len;
+		remaining -= len;
+
+		if (info->ri_pageoff == PAGE_SIZE) {
+			info->ri_pageno++;
+			info->ri_pageoff = 0;
+		}
+	}
+
+	/* Otherwise, a fresh page is needed. */
+	if (remaining) {
+		head->arg.pages[info->ri_pageno] =
+				rqstp->rq_pages[info->ri_pageno];
+		head->count++;
+
+		destp = page_address(rqstp->rq_pages[info->ri_pageno]);
+		memcpy(destp, srcp, remaining);
+		info->ri_pageoff += remaining;
+	}
+
+	head->arg.page_len += tail_length;
+	head->arg.len += tail_length;
+	head->arg.buflen += tail_length;
+	return 0;
+}
+
+/* Construct RDMA Reads to pull over a normal Read chunk. The chunk
+ * data lands in the page list of head->arg.pages.
+ *
+ * Currently NFSD does not look at the head->arg.tail[0] iovec.
+ * Therefore, XDR round-up of the Read chunk and trailing
+ * inline content must both be added at the end of the pagelist.
+ */
+static int svc_rdma_build_normal_read_chunk(struct svc_rqst *rqstp,
+					    struct svc_rdma_read_info *info,
+					    __be32 *p)
+{
+	struct svc_rdma_op_ctxt *head = info->ri_readctxt;
+	int ret;
+
+	dprintk("svcrdma: Reading Read chunk at position %u\n",
+		info->ri_position);
+
+	info->ri_pageno = head->hdr_count;
+	info->ri_pageoff = 0;
+
+	ret = svc_rdma_build_read_chunk(rqstp, info, p);
+	if (ret < 0)
+		goto out;
+
+	/* Read chunk may need XDR round-up (see RFC 5666, s. 3.7).
+	 */
+	if (info->ri_chunklen & 3) {
+		u32 padlen = 4 - (info->ri_chunklen & 3);
+
+		info->ri_chunklen += padlen;
+
+		/* NB: data payload always starts on XDR alignment,
+		 * thus the pad can never contain a page boundary.
+		 */
+		info->ri_pageoff += padlen;
+		if (info->ri_pageoff == PAGE_SIZE) {
+			info->ri_pageno++;
+			info->ri_pageoff = 0;
+		}
+	}
+
+	head->arg.page_len = info->ri_chunklen;
+	head->arg.len += info->ri_chunklen;
+	head->arg.buflen += info->ri_chunklen;
+
+	if (info->ri_position < head->arg.head[0].iov_len) {
+		ret = svc_rdma_copy_tail(rqstp, info);
+		if (ret < 0)
+			goto out;
+	}
+	head->arg.head[0].iov_len = info->ri_position;
+
+out:
+	return ret;
+}
+
+/* Construct RDMA Reads to pull over a Position Zero Read chunk.
+ * The start of the data lands in the first page just after
+ * the Transport header, and the rest lands in the page list of
+ * head->arg.pages.
+ *
+ * Assumptions:
+ *	- A PZRC has an XDR-aligned length (no implicit round-up).
+ *	- There can be no trailing inline content (IOW, we assume
+ *	  a PZRC is never sent in an RDMA_MSG message, though it's
+ *	  allowed by spec).
+ */
+static int svc_rdma_build_pz_read_chunk(struct svc_rqst *rqstp,
+					struct svc_rdma_read_info *info,
+					__be32 *p)
+{
+	struct svc_rdma_op_ctxt *head = info->ri_readctxt;
+	int ret;
+
+	dprintk("svcrdma: Reading Position Zero Read chunk\n");
+
+	info->ri_pageno = head->hdr_count - 1;
+	info->ri_pageoff = offset_in_page(head->byte_len);
+
+	ret = svc_rdma_build_read_chunk(rqstp, info, p);
+	if (ret < 0)
+		goto out;
+
+	head->arg.len += info->ri_chunklen;
+	head->arg.buflen += info->ri_chunklen;
+
+	if (head->arg.buflen <= head->sge[0].length) {
+		/* Transport header and RPC message fit entirely
+		 * in page where head iovec resides.
+		 */
+		head->arg.head[0].iov_len = info->ri_chunklen;
+	} else {
+		/* Transport header and part of RPC message reside
+		 * in the head iovec's page.
+		 */
+		head->arg.head[0].iov_len =
+				head->sge[0].length - head->byte_len;
+		head->arg.page_len =
+				info->ri_chunklen - head->arg.head[0].iov_len;
+	}
+
+out:
+	return ret;
+}
+
+/**
+ * svc_rdma_recv_read_chunk - Pull a Read chunk from the client
+ * @rdma: controlling RDMA transport
+ * @rqstp: set of pages to use as Read sink buffers
+ * @head: pages under I/O collect here
+ * @p: pointer to start of Read chunk
+ *
+ * Returns:
+ *	%0 if all needed RDMA Reads were posted successfully,
+ *	%-EINVAL if client provided too many segments,
+ *	%-ENOMEM if rdma_rw context pool was exhausted,
+ *	%-ENOTCONN if posting failed (connection is lost),
+ *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
+ *
+ * Assumptions:
+ * - All Read segments in @p have the same Position value.
+ */
+int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma, struct svc_rqst *rqstp,
+			     struct svc_rdma_op_ctxt *head, __be32 *p)
+{
+	struct svc_rdma_read_info *info;
+	struct page **page;
+	int ret;
+
+	/* The request (with page list) is constructed in
+	 * head->arg. Pages involved with RDMA Read I/O are
+	 * transferred there.
+	 */
+	head->hdr_count = head->count;
+	head->arg.head[0] = rqstp->rq_arg.head[0];
+	head->arg.tail[0] = rqstp->rq_arg.tail[0];
+	head->arg.pages = head->pages;
+	head->arg.page_base = 0;
+	head->arg.page_len = 0;
+	head->arg.len = rqstp->rq_arg.len;
+	head->arg.buflen = rqstp->rq_arg.buflen;
+
+	info = svc_rdma_read_info_alloc(rdma);
+	if (!info)
+		return -ENOMEM;
+	info->ri_readctxt = head;
+
+	info->ri_position = be32_to_cpup(p + 1);
+	if (info->ri_position)
+		ret = svc_rdma_build_normal_read_chunk(rqstp, info, p);
+	else
+		ret = svc_rdma_build_pz_read_chunk(rqstp, info, p);
+
+	/* Mark the start of the pages that can be used for the reply */
+	if (info->ri_pageoff > 0)
+		info->ri_pageno++;
+	rqstp->rq_respages = &rqstp->rq_pages[info->ri_pageno];
+	rqstp->rq_next_page = rqstp->rq_respages + 1;
+
+	if (ret < 0)
+		goto out;
+
+	ret = svc_rdma_post_chunk_ctxt(&info->ri_cc);
+
+out:
+	/* Read sink pages have been moved from rqstp->rq_pages to
+	 * head->arg.pages. Force svc_recv to refill those slots
+	 * in rq_pages.
+	 */
+	for (page = rqstp->rq_pages; page < rqstp->rq_respages; page++)
+		*page = NULL;
+
+	if (ret < 0)
+		svc_rdma_read_info_free(info);
+	return ret;
+}
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 1736337f3a55..7c3a211e0e9a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -313,13 +313,17 @@ static int svc_rdma_dma_map_buf(struct svcxprt_rdma *rdma,
 	dma_addr = ib_dma_map_page(dev, virt_to_page(base),
 				   offset, len, DMA_TO_DEVICE);
 	if (ib_dma_mapping_error(dev, dma_addr))
-		return -EIO;
+		goto out_maperr;
 
 	ctxt->sge[sge_no].addr = dma_addr;
 	ctxt->sge[sge_no].length = len;
 	ctxt->sge[sge_no].lkey = rdma->sc_pd->local_dma_lkey;
 	svc_rdma_count_mappings(rdma, ctxt);
 	return 0;
+
+out_maperr:
+	pr_err("svcrdma: failed to map buffer\n");
+	return -EIO;
 }
 
 static int svc_rdma_dma_map_page(struct svcxprt_rdma *rdma,
@@ -334,13 +338,17 @@ static int svc_rdma_dma_map_page(struct svcxprt_rdma *rdma,
 
 	dma_addr = ib_dma_map_page(dev, page, offset, len, DMA_TO_DEVICE);
 	if (ib_dma_mapping_error(dev, dma_addr))
-		return -EIO;
+		goto out_maperr;
 
 	ctxt->sge[sge_no].addr = dma_addr;
 	ctxt->sge[sge_no].length = len;
 	ctxt->sge[sge_no].lkey = rdma->sc_pd->local_dma_lkey;
 	svc_rdma_count_mappings(rdma, ctxt);
 	return 0;
+
+out_maperr:
+	pr_err("svcrdma: failed to map page\n");
+	return -EIO;
 }
 
 /**
@@ -547,7 +555,6 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
 	return 0;
 
 err:
-	pr_err("svcrdma: failed to post Send WR (%d)\n", ret);
 	svc_rdma_unmap_dma(ctxt);
 	svc_rdma_put_context(ctxt, 1);
 	return ret;
@@ -677,7 +684,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	return 0;
 
  err2:
-	if (ret != -E2BIG)
+	if (ret != -E2BIG && ret != -EINVAL)
 		goto err1;
 
 	ret = svc_rdma_post_recv(rdma, GFP_KERNEL);
diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index a9d9cb1ba4c6..e660d4965b18 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -202,7 +202,6 @@ struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *xprt)
 out:
 	ctxt->count = 0;
 	ctxt->mapped_sges = 0;
-	ctxt->frmr = NULL;
 	return ctxt;
 
 out_empty:
@@ -226,22 +225,13 @@ void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt)
 {
 	struct svcxprt_rdma *xprt = ctxt->xprt;
 	struct ib_device *device = xprt->sc_cm_id->device;
-	u32 lkey = xprt->sc_pd->local_dma_lkey;
 	unsigned int i;
 
-	for (i = 0; i < ctxt->mapped_sges; i++) {
-		/*
-		 * Unmap the DMA addr in the SGE if the lkey matches
-		 * the local_dma_lkey, otherwise, ignore it since it is
-		 * an FRMR lkey and will be unmapped later when the
-		 * last WR that uses it completes.
-		 */
-		if (ctxt->sge[i].lkey == lkey)
-			ib_dma_unmap_page(device,
-					    ctxt->sge[i].addr,
-					    ctxt->sge[i].length,
-					    ctxt->direction);
-	}
+	for (i = 0; i < ctxt->mapped_sges; i++)
+		ib_dma_unmap_page(device,
+				  ctxt->sge[i].addr,
+				  ctxt->sge[i].length,
+				  ctxt->direction);
 	ctxt->mapped_sges = 0;
 }
 
@@ -346,36 +336,6 @@ out:
 	svc_xprt_put(&xprt->sc_xprt);
 }
 
-static void svc_rdma_send_wc_common(struct svcxprt_rdma *xprt,
-				    struct ib_wc *wc,
-				    const char *opname)
-{
-	if (wc->status != IB_WC_SUCCESS)
-		goto err;
-
-out:
-	atomic_inc(&xprt->sc_sq_avail);
-	wake_up(&xprt->sc_send_wait);
-	return;
-
-err:
-	set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
-	if (wc->status != IB_WC_WR_FLUSH_ERR)
-		pr_err("svcrdma: %s: %s (%u/0x%x)\n",
-		       opname, ib_wc_status_msg(wc->status),
-		       wc->status, wc->vendor_err);
-	goto out;
-}
-
-static void svc_rdma_send_wc_common_put(struct ib_cq *cq, struct ib_wc *wc,
-					const char *opname)
-{
-	struct svcxprt_rdma *xprt = cq->cq_context;
-
-	svc_rdma_send_wc_common(xprt, wc, opname);
-	svc_xprt_put(&xprt->sc_xprt);
-}
-
 /**
  * svc_rdma_wc_send - Invoked by RDMA provider for each polled Send WC
  * @cq:        completion queue
@@ -384,73 +344,28 @@ static void svc_rdma_send_wc_common_put(struct ib_cq *cq, struct ib_wc *wc,
  */
 void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct ib_cqe *cqe = wc->wr_cqe;
-	struct svc_rdma_op_ctxt *ctxt;
-
-	svc_rdma_send_wc_common_put(cq, wc, "send");
-
-	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
-	svc_rdma_unmap_dma(ctxt);
-	svc_rdma_put_context(ctxt, 1);
-}
-
-/**
- * svc_rdma_wc_reg - Invoked by RDMA provider for each polled FASTREG WC
- * @cq:        completion queue
- * @wc:        completed WR
- *
- */
-void svc_rdma_wc_reg(struct ib_cq *cq, struct ib_wc *wc)
-{
-	svc_rdma_send_wc_common_put(cq, wc, "fastreg");
-}
-
-/**
- * svc_rdma_wc_read - Invoked by RDMA provider for each polled Read WC
- * @cq:        completion queue
- * @wc:        completed WR
- *
- */
-void svc_rdma_wc_read(struct ib_cq *cq, struct ib_wc *wc)
-{
 	struct svcxprt_rdma *xprt = cq->cq_context;
 	struct ib_cqe *cqe = wc->wr_cqe;
 	struct svc_rdma_op_ctxt *ctxt;
 
-	svc_rdma_send_wc_common(xprt, wc, "read");
+	atomic_inc(&xprt->sc_sq_avail);
+	wake_up(&xprt->sc_send_wait);
 
 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
 	svc_rdma_unmap_dma(ctxt);
-	svc_rdma_put_frmr(xprt, ctxt->frmr);
-
-	if (test_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags)) {
-		struct svc_rdma_op_ctxt *read_hdr;
-
-		read_hdr = ctxt->read_hdr;
-		spin_lock(&xprt->sc_rq_dto_lock);
-		list_add_tail(&read_hdr->list,
-			      &xprt->sc_read_complete_q);
-		spin_unlock(&xprt->sc_rq_dto_lock);
+	svc_rdma_put_context(ctxt, 1);
 
-		set_bit(XPT_DATA, &xprt->sc_xprt.xpt_flags);
-		svc_xprt_enqueue(&xprt->sc_xprt);
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
+		if (wc->status != IB_WC_WR_FLUSH_ERR)
+			pr_err("svcrdma: Send: %s (%u/0x%x)\n",
+			       ib_wc_status_msg(wc->status),
+			       wc->status, wc->vendor_err);
 	}
 
-	svc_rdma_put_context(ctxt, 0);
 	svc_xprt_put(&xprt->sc_xprt);
 }
 
-/**
- * svc_rdma_wc_inv - Invoked by RDMA provider for each polled LOCAL_INV WC
- * @cq:        completion queue
- * @wc:        completed WR
- *
- */
-void svc_rdma_wc_inv(struct ib_cq *cq, struct ib_wc *wc)
-{
-	svc_rdma_send_wc_common_put(cq, wc, "localInv");
-}
-
 static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *serv,
 					     int listener)
 {
@@ -462,14 +377,12 @@ static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *serv,
 	INIT_LIST_HEAD(&cma_xprt->sc_accept_q);
 	INIT_LIST_HEAD(&cma_xprt->sc_rq_dto_q);
 	INIT_LIST_HEAD(&cma_xprt->sc_read_complete_q);
-	INIT_LIST_HEAD(&cma_xprt->sc_frmr_q);
 	INIT_LIST_HEAD(&cma_xprt->sc_ctxts);
 	INIT_LIST_HEAD(&cma_xprt->sc_rw_ctxts);
 	init_waitqueue_head(&cma_xprt->sc_send_wait);
 
 	spin_lock_init(&cma_xprt->sc_lock);
 	spin_lock_init(&cma_xprt->sc_rq_dto_lock);
-	spin_lock_init(&cma_xprt->sc_frmr_q_lock);
 	spin_lock_init(&cma_xprt->sc_ctxt_lock);
 	spin_lock_init(&cma_xprt->sc_rw_ctxt_lock);
 
@@ -780,86 +693,6 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
 	return ERR_PTR(ret);
 }
 
-static struct svc_rdma_fastreg_mr *rdma_alloc_frmr(struct svcxprt_rdma *xprt)
-{
-	struct ib_mr *mr;
-	struct scatterlist *sg;
-	struct svc_rdma_fastreg_mr *frmr;
-	u32 num_sg;
-
-	frmr = kmalloc(sizeof(*frmr), GFP_KERNEL);
-	if (!frmr)
-		goto err;
-
-	num_sg = min_t(u32, RPCSVC_MAXPAGES, xprt->sc_frmr_pg_list_len);
-	mr = ib_alloc_mr(xprt->sc_pd, IB_MR_TYPE_MEM_REG, num_sg);
-	if (IS_ERR(mr))
-		goto err_free_frmr;
-
-	sg = kcalloc(RPCSVC_MAXPAGES, sizeof(*sg), GFP_KERNEL);
-	if (!sg)
-		goto err_free_mr;
-
-	sg_init_table(sg, RPCSVC_MAXPAGES);
-
-	frmr->mr = mr;
-	frmr->sg = sg;
-	INIT_LIST_HEAD(&frmr->frmr_list);
-	return frmr;
-
- err_free_mr:
-	ib_dereg_mr(mr);
- err_free_frmr:
-	kfree(frmr);
- err:
-	return ERR_PTR(-ENOMEM);
-}
-
-static void rdma_dealloc_frmr_q(struct svcxprt_rdma *xprt)
-{
-	struct svc_rdma_fastreg_mr *frmr;
-
-	while (!list_empty(&xprt->sc_frmr_q)) {
-		frmr = list_entry(xprt->sc_frmr_q.next,
-				  struct svc_rdma_fastreg_mr, frmr_list);
-		list_del_init(&frmr->frmr_list);
-		kfree(frmr->sg);
-		ib_dereg_mr(frmr->mr);
-		kfree(frmr);
-	}
-}
-
-struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *rdma)
-{
-	struct svc_rdma_fastreg_mr *frmr = NULL;
-
-	spin_lock(&rdma->sc_frmr_q_lock);
-	if (!list_empty(&rdma->sc_frmr_q)) {
-		frmr = list_entry(rdma->sc_frmr_q.next,
-				  struct svc_rdma_fastreg_mr, frmr_list);
-		list_del_init(&frmr->frmr_list);
-		frmr->sg_nents = 0;
-	}
-	spin_unlock(&rdma->sc_frmr_q_lock);
-	if (frmr)
-		return frmr;
-
-	return rdma_alloc_frmr(rdma);
-}
-
-void svc_rdma_put_frmr(struct svcxprt_rdma *rdma,
-		       struct svc_rdma_fastreg_mr *frmr)
-{
-	if (frmr) {
-		ib_dma_unmap_sg(rdma->sc_cm_id->device,
-				frmr->sg, frmr->sg_nents, frmr->direction);
-		spin_lock(&rdma->sc_frmr_q_lock);
-		WARN_ON_ONCE(!list_empty(&frmr->frmr_list));
-		list_add(&frmr->frmr_list, &rdma->sc_frmr_q);
-		spin_unlock(&rdma->sc_frmr_q_lock);
-	}
-}
-
 /*
  * This is the xpo_recvfrom function for listening endpoints. Its
  * purpose is to accept incoming connections. The CMA callback handler
@@ -908,8 +741,6 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	 * capabilities of this particular device */
 	newxprt->sc_max_sge = min((size_t)dev->attrs.max_sge,
 				  (size_t)RPCSVC_MAXPAGES);
-	newxprt->sc_max_sge_rd = min_t(size_t, dev->attrs.max_sge_rd,
-				       RPCSVC_MAXPAGES);
 	newxprt->sc_max_req_size = svcrdma_max_req_size;
 	newxprt->sc_max_requests = min_t(u32, dev->attrs.max_qp_wr,
 					 svcrdma_max_requests);
@@ -952,7 +783,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	memset(&qp_attr, 0, sizeof qp_attr);
 	qp_attr.event_handler = qp_event_handler;
 	qp_attr.qp_context = &newxprt->sc_xprt;
-	qp_attr.port_num = newxprt->sc_cm_id->port_num;
+	qp_attr.port_num = newxprt->sc_port_num;
 	qp_attr.cap.max_rdma_ctxs = newxprt->sc_max_requests;
 	qp_attr.cap.max_send_wr = newxprt->sc_sq_depth;
 	qp_attr.cap.max_recv_wr = newxprt->sc_rq_depth;
@@ -976,47 +807,12 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	}
 	newxprt->sc_qp = newxprt->sc_cm_id->qp;
 
-	/*
-	 * Use the most secure set of MR resources based on the
-	 * transport type and available memory management features in
-	 * the device. Here's the table implemented below:
-	 *
-	 *		Fast	Global	DMA	Remote WR
-	 *		Reg	LKEY	MR	Access
-	 *		Sup'd	Sup'd	Needed	Needed
-	 *
-	 * IWARP	N	N	Y	Y
-	 *		N	Y	Y	Y
-	 *		Y	N	Y	N
-	 *		Y	Y	N	-
-	 *
-	 * IB		N	N	Y	N
-	 *		N	Y	N	-
-	 *		Y	N	Y	N
-	 *		Y	Y	N	-
-	 *
-	 * NB:	iWARP requires remote write access for the data sink
-	 *	of an RDMA_READ. IB does not.
-	 */
-	newxprt->sc_reader = rdma_read_chunk_lcl;
-	if (dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) {
-		newxprt->sc_frmr_pg_list_len =
-			dev->attrs.max_fast_reg_page_list_len;
-		newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_FAST_REG;
-		newxprt->sc_reader = rdma_read_chunk_frmr;
-	} else
+	if (!(dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS))
 		newxprt->sc_snd_w_inv = false;
-
-	/*
-	 * Determine if a DMA MR is required and if so, what privs are required
-	 */
-	if (!rdma_protocol_iwarp(dev, newxprt->sc_cm_id->port_num) &&
-	    !rdma_ib_or_roce(dev, newxprt->sc_cm_id->port_num))
+	if (!rdma_protocol_iwarp(dev, newxprt->sc_port_num) &&
+	    !rdma_ib_or_roce(dev, newxprt->sc_port_num))
 		goto errout;
 
-	if (rdma_protocol_iwarp(dev, newxprt->sc_cm_id->port_num))
-		newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_READ_W_INV;
-
 	/* Post receive buffers */
 	for (i = 0; i < newxprt->sc_max_requests; i++) {
 		ret = svc_rdma_post_recv(newxprt, GFP_KERNEL);
@@ -1056,7 +852,6 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr;
 	dprintk("    remote address  : %pIS:%u\n", sap, rpc_get_port(sap));
 	dprintk("    max_sge         : %d\n", newxprt->sc_max_sge);
-	dprintk("    max_sge_rd      : %d\n", newxprt->sc_max_sge_rd);
 	dprintk("    sq_depth        : %d\n", newxprt->sc_sq_depth);
 	dprintk("    max_requests    : %d\n", newxprt->sc_max_requests);
 	dprintk("    ord             : %d\n", newxprt->sc_ord);
@@ -1117,12 +912,6 @@ static void __svc_rdma_free(struct work_struct *work)
 		pr_err("svcrdma: sc_xprt still in use? (%d)\n",
 		       kref_read(&xprt->xpt_ref));
 
-	/*
-	 * Destroy queued, but not processed read completions. Note
-	 * that this cleanup has to be done before destroying the
-	 * cm_id because the device ptr is needed to unmap the dma in
-	 * svc_rdma_put_context.
-	 */
 	while (!list_empty(&rdma->sc_read_complete_q)) {
 		struct svc_rdma_op_ctxt *ctxt;
 		ctxt = list_first_entry(&rdma->sc_read_complete_q,
@@ -1130,8 +919,6 @@ static void __svc_rdma_free(struct work_struct *work)
 		list_del(&ctxt->list);
 		svc_rdma_put_context(ctxt, 1);
 	}
-
-	/* Destroy queued, but not processed recv completions */
 	while (!list_empty(&rdma->sc_rq_dto_q)) {
 		struct svc_rdma_op_ctxt *ctxt;
 		ctxt = list_first_entry(&rdma->sc_rq_dto_q,
@@ -1151,7 +938,6 @@ static void __svc_rdma_free(struct work_struct *work)
 		xprt->xpt_bc_xprt = NULL;
 	}
 
-	rdma_dealloc_frmr_q(rdma);
 	svc_rdma_destroy_rw_ctxts(rdma);
 	svc_rdma_destroy_ctxts(rdma);
 
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 62ecbccd9748..d1c458e5ec4d 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -684,7 +684,8 @@ xprt_rdma_free(struct rpc_task *task)
 
 	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);
 
-	if (unlikely(!list_empty(&req->rl_registered)))
+	rpcrdma_remove_req(&r_xprt->rx_buf, req);
+	if (!list_empty(&req->rl_registered))
 		ia->ri_ops->ro_unmap_safe(r_xprt, req, !RPC_IS_ASYNC(task));
 	rpcrdma_unmap_sges(ia, req);
 	rpcrdma_buffer_put(req);
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 3dbce9ac4327..e4171f2abe37 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -243,8 +243,6 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 	struct sockaddr *sap = (struct sockaddr *)&ep->rep_remote_addr;
 #endif
-	struct ib_qp_attr *attr = &ia->ri_qp_attr;
-	struct ib_qp_init_attr *iattr = &ia->ri_qp_init_attr;
 	int connstate = 0;
 
 	switch (event->event) {
@@ -267,7 +265,8 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 		break;
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
-		pr_info("rpcrdma: removing device for %pIS:%u\n",
+		pr_info("rpcrdma: removing device %s for %pIS:%u\n",
+			ia->ri_device->name,
 			sap, rpc_get_port(sap));
 #endif
 		set_bit(RPCRDMA_IAF_REMOVING, &ia->ri_flags);
@@ -282,13 +281,6 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 		return 1;
 	case RDMA_CM_EVENT_ESTABLISHED:
 		connstate = 1;
-		ib_query_qp(ia->ri_id->qp, attr,
-			    IB_QP_MAX_QP_RD_ATOMIC | IB_QP_MAX_DEST_RD_ATOMIC,
-			    iattr);
-		dprintk("RPC:       %s: %d responder resources"
-			" (%d initiator)\n",
-			__func__, attr->max_dest_rd_atomic,
-			attr->max_rd_atomic);
 		rpcrdma_update_connect_private(xprt, &event->param.conn);
 		goto connected;
 	case RDMA_CM_EVENT_CONNECT_ERROR:
@@ -298,11 +290,9 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 		connstate = -ENETDOWN;
 		goto connected;
 	case RDMA_CM_EVENT_REJECTED:
-#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
-		pr_info("rpcrdma: connection to %pIS:%u on %s rejected: %s\n",
-			sap, rpc_get_port(sap), ia->ri_device->name,
+		dprintk("rpcrdma: connection to %pIS:%u rejected: %s\n",
+			sap, rpc_get_port(sap),
 			rdma_reject_msg(id, event->status));
-#endif
 		connstate = -ECONNREFUSED;
 		if (event->status == IB_CM_REJ_STALE_CONN)
 			connstate = -EAGAIN;
@@ -310,37 +300,19 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 	case RDMA_CM_EVENT_DISCONNECTED:
 		connstate = -ECONNABORTED;
 connected:
-		dprintk("RPC:       %s: %sconnected\n",
-					__func__, connstate > 0 ? "" : "dis");
 		atomic_set(&xprt->rx_buf.rb_credits, 1);
 		ep->rep_connected = connstate;
 		rpcrdma_conn_func(ep);
 		wake_up_all(&ep->rep_connect_wait);
 		/*FALLTHROUGH*/
 	default:
-		dprintk("RPC:       %s: %pIS:%u (ep 0x%p): %s\n",
-			__func__, sap, rpc_get_port(sap), ep,
-			rdma_event_msg(event->event));
+		dprintk("RPC:       %s: %pIS:%u on %s/%s (ep 0x%p): %s\n",
+			__func__, sap, rpc_get_port(sap),
+			ia->ri_device->name, ia->ri_ops->ro_displayname,
+			ep, rdma_event_msg(event->event));
 		break;
 	}
 
-#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
-	if (connstate == 1) {
-		int ird = attr->max_dest_rd_atomic;
-		int tird = ep->rep_remote_cma.responder_resources;
-
-		pr_info("rpcrdma: connection to %pIS:%u on %s, memreg '%s', %d credits, %d responders%s\n",
-			sap, rpc_get_port(sap),
-			ia->ri_device->name,
-			ia->ri_ops->ro_displayname,
-			xprt->rx_buf.rb_max_requests,
-			ird, ird < 4 && ird < tird / 2 ? " (low!)" : "");
-	} else if (connstate < 0) {
-		pr_info("rpcrdma: connection to %pIS:%u closed (%d)\n",
-			sap, rpc_get_port(sap), connstate);
-	}
-#endif
-
 	return 0;
 }
 
@@ -971,7 +943,6 @@ rpcrdma_create_req(struct rpcrdma_xprt *r_xprt)
 	if (req == NULL)
 		return ERR_PTR(-ENOMEM);
 
-	INIT_LIST_HEAD(&req->rl_free);
 	spin_lock(&buffer->rb_reqslock);
 	list_add(&req->rl_all, &buffer->rb_allreqs);
 	spin_unlock(&buffer->rb_reqslock);
@@ -1033,6 +1004,7 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 	spin_lock_init(&buf->rb_recovery_lock);
 	INIT_LIST_HEAD(&buf->rb_mws);
 	INIT_LIST_HEAD(&buf->rb_all);
+	INIT_LIST_HEAD(&buf->rb_pending);
 	INIT_LIST_HEAD(&buf->rb_stale_mrs);
 	INIT_DELAYED_WORK(&buf->rb_refresh_worker,
 			  rpcrdma_mr_refresh_worker);
@@ -1055,7 +1027,7 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 			goto out;
 		}
 		req->rl_backchannel = false;
-		list_add(&req->rl_free, &buf->rb_send_bufs);
+		list_add(&req->rl_list, &buf->rb_send_bufs);
 	}
 
 	INIT_LIST_HEAD(&buf->rb_recv_bufs);
@@ -1084,8 +1056,8 @@ rpcrdma_buffer_get_req_locked(struct rpcrdma_buffer *buf)
 	struct rpcrdma_req *req;
 
 	req = list_first_entry(&buf->rb_send_bufs,
-			       struct rpcrdma_req, rl_free);
-	list_del(&req->rl_free);
+			       struct rpcrdma_req, rl_list);
+	list_del_init(&req->rl_list);
 	return req;
 }
 
@@ -1187,6 +1159,7 @@ rpcrdma_get_mw(struct rpcrdma_xprt *r_xprt)
 
 	if (!mw)
 		goto out_nomws;
+	mw->mw_flags = 0;
 	return mw;
 
 out_nomws:
@@ -1267,7 +1240,7 @@ rpcrdma_buffer_put(struct rpcrdma_req *req)
 
 	spin_lock(&buffers->rb_lock);
 	buffers->rb_send_count--;
-	list_add_tail(&req->rl_free, &buffers->rb_send_bufs);
+	list_add_tail(&req->rl_list, &buffers->rb_send_bufs);
 	if (rep) {
 		buffers->rb_recv_count--;
 		list_add_tail(&rep->rr_list, &buffers->rb_recv_bufs);
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 1d66acf1a723..b282d3f8cdd8 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -271,6 +271,7 @@ struct rpcrdma_mw {
 	struct scatterlist	*mw_sg;
 	int			mw_nents;
 	enum dma_data_direction	mw_dir;
+	unsigned long		mw_flags;
 	union {
 		struct rpcrdma_fmr	fmr;
 		struct rpcrdma_frmr	frmr;
@@ -282,6 +283,11 @@ struct rpcrdma_mw {
 	struct list_head	mw_all;
 };
 
+/* mw_flags */
+enum {
+	RPCRDMA_MW_F_RI		= 1,
+};
+
 /*
  * struct rpcrdma_req -- structure central to the request/reply sequence.
  *
@@ -334,7 +340,8 @@ enum {
 
 struct rpcrdma_buffer;
 struct rpcrdma_req {
-	struct list_head	rl_free;
+	struct list_head	rl_list;
+	__be32			rl_xid;
 	unsigned int		rl_mapped_sges;
 	unsigned int		rl_connect_cookie;
 	struct rpcrdma_buffer	*rl_buffer;
@@ -396,6 +403,7 @@ struct rpcrdma_buffer {
 	int			rb_send_count, rb_recv_count;
 	struct list_head	rb_send_bufs;
 	struct list_head	rb_recv_bufs;
+	struct list_head	rb_pending;
 	u32			rb_max_requests;
 	atomic_t		rb_credits;	/* most recent credit grant */
 
@@ -461,7 +469,7 @@ struct rpcrdma_memreg_ops {
 				  struct rpcrdma_mr_seg *, int, bool,
 				  struct rpcrdma_mw **);
 	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
-					 struct rpcrdma_req *);
+					 struct list_head *);
 	void		(*ro_unmap_safe)(struct rpcrdma_xprt *,
 					 struct rpcrdma_req *, bool);
 	void		(*ro_recover_mr)(struct rpcrdma_mw *);
@@ -544,6 +552,34 @@ void rpcrdma_destroy_req(struct rpcrdma_req *);
 int rpcrdma_buffer_create(struct rpcrdma_xprt *);
 void rpcrdma_buffer_destroy(struct rpcrdma_buffer *);
 
+static inline void
+rpcrdma_insert_req(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
+{
+	spin_lock(&buffers->rb_lock);
+	if (list_empty(&req->rl_list))
+		list_add_tail(&req->rl_list, &buffers->rb_pending);
+	spin_unlock(&buffers->rb_lock);
+}
+
+static inline struct rpcrdma_req *
+rpcrdma_lookup_req_locked(struct rpcrdma_buffer *buffers, __be32 xid)
+{
+	struct rpcrdma_req *pos;
+
+	list_for_each_entry(pos, &buffers->rb_pending, rl_list)
+		if (pos->rl_xid == xid)
+			return pos;
+	return NULL;
+}
+
+static inline void
+rpcrdma_remove_req(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
+{
+	spin_lock(&buffers->rb_lock);
+	list_del(&req->rl_list);
+	spin_unlock(&buffers->rb_lock);
+}
+
 struct rpcrdma_mw *rpcrdma_get_mw(struct rpcrdma_xprt *);
 void rpcrdma_put_mw(struct rpcrdma_xprt *, struct rpcrdma_mw *);
 struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *);