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

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

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

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

fs/ceph/addr.c

@@ -530,14 +530,10 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 	long writeback_stat;
 	u64 truncate_size;
 	u32 truncate_seq;
-	int err = 0, len = PAGE_SIZE;
+	int err, len = PAGE_SIZE;
 
 	dout("writepage %p idx %lu\n", page, page->index);
 
-	if (!page->mapping || !page->mapping->host) {
-		dout("writepage %p - no mapping\n", page);
-		return -EFAULT;
-	}
 	inode = page->mapping->host;
 	ci = ceph_inode(inode);
 	fsc = ceph_inode_to_client(inode);
@@ -547,7 +543,7 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 	snapc = page_snap_context(page);
 	if (snapc == NULL) {
 		dout("writepage %p page %p not dirty?\n", inode, page);
-		goto out;
+		return 0;
 	}
 	oldest = get_oldest_context(inode, &snap_size,
 				    &truncate_size, &truncate_seq);
@@ -555,9 +551,10 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 		dout("writepage %p page %p snapc %p not writeable - noop\n",
 		     inode, page, snapc);
 		/* we should only noop if called by kswapd */
-		WARN_ON((current->flags & PF_MEMALLOC) == 0);
+		WARN_ON(!(current->flags & PF_MEMALLOC));
 		ceph_put_snap_context(oldest);
-		goto out;
+		redirty_page_for_writepage(wbc, page);
+		return 0;
 	}
 	ceph_put_snap_context(oldest);
 
@@ -567,8 +564,9 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 	/* is this a partial page at end of file? */
 	if (page_off >= snap_size) {
 		dout("%p page eof %llu\n", page, snap_size);
-		goto out;
+		return 0;
 	}
+
 	if (snap_size < page_off + len)
 		len = snap_size - page_off;
 
@@ -595,7 +593,7 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 			dout("writepage interrupted page %p\n", page);
 			redirty_page_for_writepage(wbc, page);
 			end_page_writeback(page);
-			goto out;
+			return err;
 		}
 		dout("writepage setting page/mapping error %d %p\n",
 		     err, page);
@@ -611,7 +609,6 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 	end_page_writeback(page);
 	ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 	ceph_put_snap_context(snapc);  /* page's reference */
-out:
 	return err;
 }
 
@@ -1318,7 +1315,7 @@ static int ceph_write_end(struct file *file, struct address_space *mapping,
 			  struct page *page, void *fsdata)
 {
 	struct inode *inode = file_inode(file);
-	int check_cap = 0;
+	bool check_cap = false;
 
 	dout("write_end file %p inode %p page %p %d~%d (%d)\n", file,
 	     inode, page, (int)pos, (int)copied, (int)len);

fs/ceph/cache.c

@@ -35,18 +35,34 @@ struct fscache_netfs ceph_cache_netfs = {
 	.version	= 0,
 };
 
+static DEFINE_MUTEX(ceph_fscache_lock);
+static LIST_HEAD(ceph_fscache_list);
+
+struct ceph_fscache_entry {
+	struct list_head list;
+	struct fscache_cookie *fscache;
+	struct ceph_fsid fsid;
+	size_t uniq_len;
+	char uniquifier[0];
+};
+
 static uint16_t ceph_fscache_session_get_key(const void *cookie_netfs_data,
 					     void *buffer, uint16_t maxbuf)
 {
 	const struct ceph_fs_client* fsc = cookie_netfs_data;
-	uint16_t klen;
+	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
+	uint16_t fsid_len, uniq_len;
 
-	klen = sizeof(fsc->client->fsid);
-	if (klen > maxbuf)
+	fsid_len = sizeof(fsc->client->fsid);
+	uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
+	if (fsid_len + uniq_len > maxbuf)
 		return 0;
 
-	memcpy(buffer, &fsc->client->fsid, klen);
-	return klen;
+	memcpy(buffer, &fsc->client->fsid, fsid_len);
+	if (uniq_len)
+		memcpy(buffer + fsid_len, fscache_uniq, uniq_len);
+
+	return fsid_len + uniq_len;
 }
 
 static const struct fscache_cookie_def ceph_fscache_fsid_object_def = {
@@ -67,13 +83,54 @@ void ceph_fscache_unregister(void)
 
 int ceph_fscache_register_fs(struct ceph_fs_client* fsc)
 {
+	const struct ceph_fsid *fsid = &fsc->client->fsid;
+	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
+	size_t uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
+	struct ceph_fscache_entry *ent;
+	int err = 0;
+
+	mutex_lock(&ceph_fscache_lock);
+	list_for_each_entry(ent, &ceph_fscache_list, list) {
+		if (memcmp(&ent->fsid, fsid, sizeof(*fsid)))
+			continue;
+		if (ent->uniq_len != uniq_len)
+			continue;
+		if (uniq_len && memcmp(ent->uniquifier, fscache_uniq, uniq_len))
+			continue;
+
+		pr_err("fscache cookie already registered for fsid %pU\n", fsid);
+		pr_err("  use fsc=%%s mount option to specify a uniquifier\n");
+		err = -EBUSY;
+		goto out_unlock;
+	}
+
+	ent = kzalloc(sizeof(*ent) + uniq_len, GFP_KERNEL);
+	if (!ent) {
+		err = -ENOMEM;
+		goto out_unlock;
+	}
+
 	fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index,
 					      &ceph_fscache_fsid_object_def,
 					      fsc, true);
-	if (!fsc->fscache)
-		pr_err("Unable to register fsid: %p fscache cookie\n", fsc);
 
-	return 0;
+	if (fsc->fscache) {
+		memcpy(&ent->fsid, fsid, sizeof(*fsid));
+		if (uniq_len > 0) {
+			memcpy(&ent->uniquifier, fscache_uniq, uniq_len);
+			ent->uniq_len = uniq_len;
+		}
+		ent->fscache = fsc->fscache;
+		list_add_tail(&ent->list, &ceph_fscache_list);
+	} else {
+		kfree(ent);
+		pr_err("unable to register fscache cookie for fsid %pU\n",
+		       fsid);
+		/* all other fs ignore this error */
+	}
+out_unlock:
+	mutex_unlock(&ceph_fscache_lock);
+	return err;
 }
 
 static uint16_t ceph_fscache_inode_get_key(const void *cookie_netfs_data,
@@ -349,7 +406,24 @@ void ceph_invalidate_fscache_page(struct inode* inode, struct page *page)
 
 void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc)
 {
-	fscache_relinquish_cookie(fsc->fscache, 0);
+	if (fscache_cookie_valid(fsc->fscache)) {
+		struct ceph_fscache_entry *ent;
+		bool found = false;
+
+		mutex_lock(&ceph_fscache_lock);
+		list_for_each_entry(ent, &ceph_fscache_list, list) {
+			if (ent->fscache == fsc->fscache) {
+				list_del(&ent->list);
+				kfree(ent);
+				found = true;
+				break;
+			}
+		}
+		WARN_ON_ONCE(!found);
+		mutex_unlock(&ceph_fscache_lock);
+
+		__fscache_relinquish_cookie(fsc->fscache, 0);
+	}
 	fsc->fscache = NULL;
 }
 

fs/ceph/caps.c

@@ -1653,6 +1653,21 @@ static int try_nonblocking_invalidate(struct inode *inode)
 	return -1;
 }
 
+bool __ceph_should_report_size(struct ceph_inode_info *ci)
+{
+	loff_t size = ci->vfs_inode.i_size;
+	/* mds will adjust max size according to the reported size */
+	if (ci->i_flushing_caps & CEPH_CAP_FILE_WR)
+		return false;
+	if (size >= ci->i_max_size)
+		return true;
+	/* half of previous max_size increment has been used */
+	if (ci->i_max_size > ci->i_reported_size &&
+	    (size << 1) >= ci->i_max_size + ci->i_reported_size)
+		return true;
+	return false;
+}
+
 /*
  * Swiss army knife function to examine currently used and wanted
  * versus held caps.  Release, flush, ack revoked caps to mds as
@@ -1806,8 +1821,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 			}
 
 			/* approaching file_max? */
-			if ((inode->i_size << 1) >= ci->i_max_size &&
-			    (ci->i_reported_size << 1) < ci->i_max_size) {
+			if (__ceph_should_report_size(ci)) {
 				dout("i_size approaching max_size\n");
 				goto ack;
 			}
@@ -3027,8 +3041,10 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 					le32_to_cpu(grant->truncate_seq),
 					le64_to_cpu(grant->truncate_size),
 					size);
-		/* max size increase? */
-		if (ci->i_auth_cap == cap && max_size != ci->i_max_size) {
+	}
+
+	if (ci->i_auth_cap == cap && (newcaps & CEPH_CAP_ANY_FILE_WR)) {
+		if (max_size != ci->i_max_size) {
 			dout("max_size %lld -> %llu\n",
 			     ci->i_max_size, max_size);
 			ci->i_max_size = max_size;
@@ -3037,6 +3053,10 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 				ci->i_requested_max_size = 0;
 			}
 			wake = true;
+		} else if (ci->i_wanted_max_size > ci->i_max_size &&
+			   ci->i_wanted_max_size > ci->i_requested_max_size) {
+			/* CEPH_CAP_OP_IMPORT */
+			wake = true;
 		}
 	}
 
@@ -3554,7 +3574,6 @@ static void handle_cap_import(struct ceph_mds_client *mdsc,
 	}
 
 	/* make sure we re-request max_size, if necessary */
-	ci->i_wanted_max_size = 0;
 	ci->i_requested_max_size = 0;
 
 	*old_issued = issued;
@@ -3790,6 +3809,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
  */
 void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 {
+	struct inode *inode;
 	struct ceph_inode_info *ci;
 	int flags = CHECK_CAPS_NODELAY;
 
@@ -3805,9 +3825,15 @@ void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 		    time_before(jiffies, ci->i_hold_caps_max))
 			break;
 		list_del_init(&ci->i_cap_delay_list);
+
+		inode = igrab(&ci->vfs_inode);
 		spin_unlock(&mdsc->cap_delay_lock);
-		dout("check_delayed_caps on %p\n", &ci->vfs_inode);
-		ceph_check_caps(ci, flags, NULL);
+
+		if (inode) {
+			dout("check_delayed_caps on %p\n", inode);
+			ceph_check_caps(ci, flags, NULL);
+			iput(inode);
+		}
 	}
 	spin_unlock(&mdsc->cap_delay_lock);
 }

fs/ceph/file.c

@@ -1040,8 +1040,8 @@ ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos,
 	int num_pages;
 	int written = 0;
 	int flags;
-	int check_caps = 0;
 	int ret;
+	bool check_caps = false;
 	struct timespec mtime = current_time(inode);
 	size_t count = iov_iter_count(from);
 

fs/ceph/inode.c

@@ -1016,6 +1016,7 @@ static void update_dentry_lease(struct dentry *dentry,
 	long unsigned ttl = from_time + (duration * HZ) / 1000;
 	long unsigned half_ttl = from_time + (duration * HZ / 2) / 1000;
 	struct inode *dir;
+	struct ceph_mds_session *old_lease_session = NULL;
 
 	/*
 	 * Make sure dentry's inode matches tgt_vino. NULL tgt_vino means that
@@ -1051,8 +1052,10 @@ static void update_dentry_lease(struct dentry *dentry,
 	    time_before(ttl, di->time))
 		goto out_unlock;  /* we already have a newer lease. */
 
-	if (di->lease_session && di->lease_session != session)
-		goto out_unlock;
+	if (di->lease_session && di->lease_session != session) {
+		old_lease_session = di->lease_session;
+		di->lease_session = NULL;
+	}
 
 	ceph_dentry_lru_touch(dentry);
 
@@ -1065,6 +1068,8 @@ static void update_dentry_lease(struct dentry *dentry,
 	di->time = ttl;
 out_unlock:
 	spin_unlock(&dentry->d_lock);
+	if (old_lease_session)
+		ceph_put_mds_session(old_lease_session);
 	return;
 }
 
@@ -1653,20 +1658,17 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 	return err;
 }
 
-int ceph_inode_set_size(struct inode *inode, loff_t size)
+bool ceph_inode_set_size(struct inode *inode, loff_t size)
 {
 	struct ceph_inode_info *ci = ceph_inode(inode);
-	int ret = 0;
+	bool ret;
 
 	spin_lock(&ci->i_ceph_lock);
 	dout("set_size %p %llu -> %llu\n", inode, inode->i_size, size);
 	i_size_write(inode, size);
 	inode->i_blocks = calc_inode_blocks(size);
 
-	/* tell the MDS if we are approaching max_size */
-	if ((size << 1) >= ci->i_max_size &&
-	    (ci->i_reported_size << 1) < ci->i_max_size)
-		ret = 1;
+	ret = __ceph_should_report_size(ci);
 
 	spin_unlock(&ci->i_ceph_lock);
 	return ret;

fs/ceph/locks.c

@@ -127,6 +127,29 @@ static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
 	dout("ceph_lock_wait_for_completion: request %llu was interrupted\n",
 	     req->r_tid);
 
+	mutex_lock(&mdsc->mutex);
+	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
+		err = 0;
+	} else {
+		/*
+		 * ensure we aren't running concurrently with
+		 * ceph_fill_trace or ceph_readdir_prepopulate, which
+		 * rely on locks (dir mutex) held by our caller.
+		 */
+		mutex_lock(&req->r_fill_mutex);
+		req->r_err = err;
+		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
+		mutex_unlock(&req->r_fill_mutex);
+
+		if (!req->r_session) {
+			// haven't sent the request
+			err = 0;
+		}
+	}
+	mutex_unlock(&mdsc->mutex);
+	if (!err)
+		return 0;
+
 	intr_req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETFILELOCK,
 					    USE_AUTH_MDS);
 	if (IS_ERR(intr_req))
@@ -146,7 +169,7 @@ static int ceph_lock_wait_for_completion(struct ceph_mds_client *mdsc,
 	if (err && err != -ERESTARTSYS)
 		return err;
 
-	wait_for_completion(&req->r_completion);
+	wait_for_completion_killable(&req->r_safe_completion);
 	return 0;
 }
 

fs/ceph/mds_client.c

@@ -3769,13 +3769,13 @@ static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
 {
 	struct ceph_mds_client *mdsc = fsc->mdsc;
-
 	dout("mdsc_destroy %p\n", mdsc);
-	ceph_mdsc_stop(mdsc);
 
 	/* flush out any connection work with references to us */
 	ceph_msgr_flush();
 
+	ceph_mdsc_stop(mdsc);
+
 	fsc->mdsc = NULL;
 	kfree(mdsc);
 	dout("mdsc_destroy %p done\n", mdsc);

fs/ceph/super.c

@@ -121,6 +121,7 @@ enum {
 	/* int args above */
 	Opt_snapdirname,
 	Opt_mds_namespace,
+	Opt_fscache_uniq,
 	Opt_last_string,
 	/* string args above */
 	Opt_dirstat,
@@ -158,6 +159,7 @@ static match_table_t fsopt_tokens = {
 	/* int args above */
 	{Opt_snapdirname, "snapdirname=%s"},
 	{Opt_mds_namespace, "mds_namespace=%s"},
+	{Opt_fscache_uniq, "fsc=%s"},
 	/* string args above */
 	{Opt_dirstat, "dirstat"},
 	{Opt_nodirstat, "nodirstat"},
@@ -223,6 +225,14 @@ static int parse_fsopt_token(char *c, void *private)
 		if (!fsopt->mds_namespace)
 			return -ENOMEM;
 		break;
+	case Opt_fscache_uniq:
+		fsopt->fscache_uniq = kstrndup(argstr[0].from,
+					       argstr[0].to-argstr[0].from,
+					       GFP_KERNEL);
+		if (!fsopt->fscache_uniq)
+			return -ENOMEM;
+		fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
+		break;
 		/* misc */
 	case Opt_wsize:
 		fsopt->wsize = intval;
@@ -317,6 +327,7 @@ static void destroy_mount_options(struct ceph_mount_options *args)
 	kfree(args->snapdir_name);
 	kfree(args->mds_namespace);
 	kfree(args->server_path);
+	kfree(args->fscache_uniq);
 	kfree(args);
 }
 
@@ -350,8 +361,10 @@ static int compare_mount_options(struct ceph_mount_options *new_fsopt,
 	ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
 	if (ret)
 		return ret;
-
 	ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
+	if (ret)
+		return ret;
+	ret = strcmp_null(fsopt1->fscache_uniq, fsopt2->fscache_uniq);
 	if (ret)
 		return ret;
 
@@ -475,8 +488,12 @@ static int ceph_show_options(struct seq_file *m, struct dentry *root)
 		seq_puts(m, ",noasyncreaddir");
 	if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
 		seq_puts(m, ",nodcache");
-	if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE)
-		seq_puts(m, ",fsc");
+	if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) {
+		if (fsopt->fscache_uniq)
+			seq_printf(m, ",fsc=%s", fsopt->fscache_uniq);
+		else
+			seq_puts(m, ",fsc");
+	}
 	if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
 		seq_puts(m, ",nopoolperm");
 
@@ -597,18 +614,11 @@ static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
 	if (!fsc->wb_pagevec_pool)
 		goto fail_trunc_wq;
 
-	/* setup fscache */
-	if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) &&
-	    (ceph_fscache_register_fs(fsc) != 0))
-		goto fail_fscache;
-
 	/* caps */
 	fsc->min_caps = fsopt->max_readdir;
 
 	return fsc;
 
-fail_fscache:
-	ceph_fscache_unregister_fs(fsc);
 fail_trunc_wq:
 	destroy_workqueue(fsc->trunc_wq);
 fail_pg_inv_wq:
@@ -626,8 +636,6 @@ static void destroy_fs_client(struct ceph_fs_client *fsc)
 {
 	dout("destroy_fs_client %p\n", fsc);
 
-	ceph_fscache_unregister_fs(fsc);
-
 	destroy_workqueue(fsc->wb_wq);
 	destroy_workqueue(fsc->pg_inv_wq);
 	destroy_workqueue(fsc->trunc_wq);
@@ -636,8 +644,6 @@ static void destroy_fs_client(struct ceph_fs_client *fsc)
 
 	destroy_mount_options(fsc->mount_options);
 
-	ceph_fs_debugfs_cleanup(fsc);
-
 	ceph_destroy_client(fsc->client);
 
 	kfree(fsc);
@@ -822,6 +828,13 @@ static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc)
 		if (err < 0)
 			goto out;
 
+		/* setup fscache */
+		if (fsc->mount_options->flags & CEPH_MOUNT_OPT_FSCACHE) {
+			err = ceph_fscache_register_fs(fsc);
+			if (err < 0)
+				goto out;
+		}
+
 		if (!fsc->mount_options->server_path) {
 			path = "";
 			dout("mount opening path \\t\n");
@@ -1040,6 +1053,12 @@ static void ceph_kill_sb(struct super_block *s)
 
 	ceph_mdsc_pre_umount(fsc->mdsc);
 	generic_shutdown_super(s);
+
+	fsc->client->extra_mon_dispatch = NULL;
+	ceph_fs_debugfs_cleanup(fsc);
+
+	ceph_fscache_unregister_fs(fsc);
+
 	ceph_mdsc_destroy(fsc);
 
 	destroy_fs_client(fsc);

fs/ceph/super.h

@@ -73,6 +73,7 @@ struct ceph_mount_options {
 	char *snapdir_name;   /* default ".snap" */
 	char *mds_namespace;  /* default NULL */
 	char *server_path;    /* default  "/" */
+	char *fscache_uniq;   /* default NULL */
 };
 
 struct ceph_fs_client {
@@ -793,7 +794,7 @@ extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 
 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 
-extern int ceph_inode_set_size(struct inode *inode, loff_t size);
+extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 extern void ceph_queue_vmtruncate(struct inode *inode);
 
@@ -918,6 +919,7 @@ extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 				       struct ceph_snap_context *snapc);
 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
 			     struct ceph_mds_session **psession);
+extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 			    struct ceph_mds_session *session);
 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);

fs/ceph/xattr.c

@@ -756,6 +756,9 @@ ssize_t __ceph_getxattr(struct inode *inode, const char *name, void *value,
 	/* let's see if a virtual xattr was requested */
 	vxattr = ceph_match_vxattr(inode, name);
 	if (vxattr) {
+		err = ceph_do_getattr(inode, 0, true);
+		if (err)
+			return err;
 		err = -ENODATA;
 		if (!(vxattr->exists_cb && !vxattr->exists_cb(ci)))
 			err = vxattr->getxattr_cb(ci, value, size);

include/linux/ceph/ceph_features.h

@@ -2,103 +2,174 @@
 #define __CEPH_FEATURES
 
 /*
- * feature bits
+ * Each time we reclaim bits for reuse we need to specify another bit
+ * that, if present, indicates we have the new incarnation of that
+ * feature.  Base case is 1 (first use).
  */
-#define CEPH_FEATURE_UID            (1ULL<<0)
-#define CEPH_FEATURE_NOSRCADDR      (1ULL<<1)
-#define CEPH_FEATURE_MONCLOCKCHECK  (1ULL<<2)
-#define CEPH_FEATURE_FLOCK          (1ULL<<3)
-#define CEPH_FEATURE_SUBSCRIBE2     (1ULL<<4)
-#define CEPH_FEATURE_MONNAMES       (1ULL<<5)
-#define CEPH_FEATURE_RECONNECT_SEQ  (1ULL<<6)
-#define CEPH_FEATURE_DIRLAYOUTHASH  (1ULL<<7)
-#define CEPH_FEATURE_OBJECTLOCATOR  (1ULL<<8)
-#define CEPH_FEATURE_PGID64         (1ULL<<9)
-#define CEPH_FEATURE_INCSUBOSDMAP   (1ULL<<10)
-#define CEPH_FEATURE_PGPOOL3        (1ULL<<11)
-#define CEPH_FEATURE_OSDREPLYMUX    (1ULL<<12)
-#define CEPH_FEATURE_OSDENC         (1ULL<<13)
-#define CEPH_FEATURE_OMAP           (1ULL<<14)
-#define CEPH_FEATURE_MONENC         (1ULL<<15)
-#define CEPH_FEATURE_QUERY_T        (1ULL<<16)
-#define CEPH_FEATURE_INDEP_PG_MAP   (1ULL<<17)
-#define CEPH_FEATURE_CRUSH_TUNABLES (1ULL<<18)
-#define CEPH_FEATURE_CHUNKY_SCRUB   (1ULL<<19)
-#define CEPH_FEATURE_MON_NULLROUTE  (1ULL<<20)
-#define CEPH_FEATURE_MON_GV         (1ULL<<21)
-#define CEPH_FEATURE_BACKFILL_RESERVATION (1ULL<<22)
-#define CEPH_FEATURE_MSG_AUTH	    (1ULL<<23)
-#define CEPH_FEATURE_RECOVERY_RESERVATION (1ULL<<24)
-#define CEPH_FEATURE_CRUSH_TUNABLES2 (1ULL<<25)
-#define CEPH_FEATURE_CREATEPOOLID   (1ULL<<26)
-#define CEPH_FEATURE_REPLY_CREATE_INODE   (1ULL<<27)
-#define CEPH_FEATURE_OSD_HBMSGS     (1ULL<<28)
-#define CEPH_FEATURE_MDSENC         (1ULL<<29)
-#define CEPH_FEATURE_OSDHASHPSPOOL  (1ULL<<30)
-#define CEPH_FEATURE_MON_SINGLE_PAXOS (1ULL<<31)
-#define CEPH_FEATURE_OSD_SNAPMAPPER (1ULL<<32)
-#define CEPH_FEATURE_MON_SCRUB      (1ULL<<33)
-#define CEPH_FEATURE_OSD_PACKED_RECOVERY (1ULL<<34)
-#define CEPH_FEATURE_OSD_CACHEPOOL (1ULL<<35)
-#define CEPH_FEATURE_CRUSH_V2      (1ULL<<36)  /* new indep; SET_* steps */
-#define CEPH_FEATURE_EXPORT_PEER   (1ULL<<37)
-#define CEPH_FEATURE_OSD_ERASURE_CODES (1ULL<<38)
-#define CEPH_FEATURE_OSD_TMAP2OMAP (1ULL<<38)   /* overlap with EC */
-/* The process supports new-style OSDMap encoding. Monitors also use
-   this bit to determine if peers support NAK messages. */
-#define CEPH_FEATURE_OSDMAP_ENC    (1ULL<<39)
-#define CEPH_FEATURE_MDS_INLINE_DATA     (1ULL<<40)
-#define CEPH_FEATURE_CRUSH_TUNABLES3     (1ULL<<41)
-#define CEPH_FEATURE_OSD_PRIMARY_AFFINITY (1ULL<<41)  /* overlap w/ tunables3 */
-#define CEPH_FEATURE_MSGR_KEEPALIVE2   (1ULL<<42)
-#define CEPH_FEATURE_OSD_POOLRESEND    (1ULL<<43)
-#define CEPH_FEATURE_ERASURE_CODE_PLUGINS_V2 (1ULL<<44)
-#define CEPH_FEATURE_OSD_SET_ALLOC_HINT (1ULL<<45)
-#define CEPH_FEATURE_OSD_FADVISE_FLAGS (1ULL<<46)
-#define CEPH_FEATURE_OSD_REPOP         (1ULL<<46)   /* overlap with fadvise */
-#define CEPH_FEATURE_OSD_OBJECT_DIGEST  (1ULL<<46)  /* overlap with fadvise */
-#define CEPH_FEATURE_OSD_TRANSACTION_MAY_LAYOUT (1ULL<<46) /* overlap w/ fadvise */
-#define CEPH_FEATURE_MDS_QUOTA      (1ULL<<47)
-#define CEPH_FEATURE_CRUSH_V4      (1ULL<<48)  /* straw2 buckets */
-#define CEPH_FEATURE_OSD_MIN_SIZE_RECOVERY (1ULL<<49)
-// duplicated since it was introduced at the same time as MIN_SIZE_RECOVERY
-#define CEPH_FEATURE_OSD_PROXY_FEATURES (1ULL<<49)  /* overlap w/ above */
-#define CEPH_FEATURE_MON_METADATA (1ULL<<50)
-#define CEPH_FEATURE_OSD_BITWISE_HOBJ_SORT (1ULL<<51) /* can sort objs bitwise */
-#define CEPH_FEATURE_OSD_PROXY_WRITE_FEATURES (1ULL<<52)
-#define CEPH_FEATURE_ERASURE_CODE_PLUGINS_V3 (1ULL<<53)
-#define CEPH_FEATURE_OSD_HITSET_GMT (1ULL<<54)
-#define CEPH_FEATURE_HAMMER_0_94_4 (1ULL<<55)
-#define CEPH_FEATURE_NEW_OSDOP_ENCODING   (1ULL<<56) /* New, v7 encoding */
-#define CEPH_FEATURE_MON_STATEFUL_SUB (1ULL<<57) /* stateful mon subscription */
-#define CEPH_FEATURE_MON_ROUTE_OSDMAP (1ULL<<57) /* peon sends osdmaps */
-#define CEPH_FEATURE_CRUSH_TUNABLES5	(1ULL<<58) /* chooseleaf stable mode */
-// duplicated since it was introduced at the same time as CEPH_FEATURE_CRUSH_TUNABLES5
-#define CEPH_FEATURE_NEW_OSDOPREPLY_ENCODING   (1ULL<<58) /* New, v7 encoding */
-#define CEPH_FEATURE_FS_FILE_LAYOUT_V2       (1ULL<<58) /* file_layout_t */
+#define CEPH_FEATURE_INCARNATION_1 (0ull)
+#define CEPH_FEATURE_INCARNATION_2 (1ull<<57) // CEPH_FEATURE_SERVER_JEWEL
+
+#define DEFINE_CEPH_FEATURE(bit, incarnation, name)			\
+	const static uint64_t CEPH_FEATURE_##name = (1ULL<<bit);		\
+	const static uint64_t CEPH_FEATUREMASK_##name =			\
+		(1ULL<<bit | CEPH_FEATURE_INCARNATION_##incarnation);
+
+/* this bit is ignored but still advertised by release *when* */
+#define DEFINE_CEPH_FEATURE_DEPRECATED(bit, incarnation, name, when) \
+	const static uint64_t DEPRECATED_CEPH_FEATURE_##name = (1ULL<<bit); \
+	const static uint64_t DEPRECATED_CEPH_FEATUREMASK_##name =		\
+		(1ULL<<bit | CEPH_FEATURE_INCARNATION_##incarnation);
 
 /*
- * The introduction of CEPH_FEATURE_OSD_SNAPMAPPER caused the feature
- * vector to evaluate to 64 bit ~0.  To cope, we designate 1ULL << 63
- * to mean 33 bit ~0, and introduce a helper below to do the
- * translation.
+ * this bit is ignored by release *unused* and not advertised by
+ * release *unadvertised*
+ */
+#define DEFINE_CEPH_FEATURE_RETIRED(bit, inc, name, unused, unadvertised)
+
+
+/*
+ * test for a feature.  this test is safer than a typical mask against
+ * the bit because it ensures that we have the bit AND the marker for the
+ * bit's incarnation.  this must be used in any case where the features
+ * bits may include an old meaning of the bit.
+ */
+#define CEPH_HAVE_FEATURE(x, name)			\
+	(((x) & (CEPH_FEATUREMASK_##name)) == (CEPH_FEATUREMASK_##name))
+
+
+/*
+ * Notes on deprecation:
+ *
+ * A *major* release is a release through which all upgrades must pass
+ * (e.g., jewel).  For example, no pre-jewel server will ever talk to
+ * a post-jewel server (mon, osd, etc).
+ *
+ * For feature bits used *only* on the server-side:
+ *
+ *  - In the first phase we indicate that a feature is DEPRECATED as of
+ *    a particular release.  This is the first major release X (say,
+ *    jewel) that does not depend on its peers advertising the feature.
+ *    That is, it safely assumes its peers all have the feature.  We
+ *    indicate this with the DEPRECATED macro.  For example,
+ *
+ *      DEFINE_CEPH_FEATURE_DEPRECATED( 2, 1, MONCLOCKCHECK, JEWEL)
+ *
+ *    because 10.2.z (jewel) did not care if its peers advertised this
+ *    feature bit.
+ *
+ *  - In the second phase we stop advertising the the bit and call it
+ *    RETIRED.  This can normally be done in the *next* major release
+ *    following the one in which we marked the feature DEPRECATED.  In
+ *    the above example, for 12.0.z (luminous) we can say:
+ *
+ *      DEFINE_CEPH_FEATURE_RETIRED( 2, 1, MONCLOCKCHECK, JEWEL, LUMINOUS)
  *
- * This was introduced by ceph.git commit
- *   9ea02b84104045c2ffd7e7f4e7af512953855ecd v0.58-657-g9ea02b8
- * and fixed by ceph.git commit
- *   4255b5c2fb54ae40c53284b3ab700fdfc7e61748 v0.65-263-g4255b5c
+ *  - The bit can be reused in the first post-luminous release, 13.0.z
+ *    (m).
+ *
+ * This ensures that no two versions who have different meanings for
+ * the bit ever speak to each other.
  */
-#define CEPH_FEATURE_RESERVED (1ULL<<63)
-
-static inline u64 ceph_sanitize_features(u64 features)
-{
-	if (features & CEPH_FEATURE_RESERVED) {
-		/* everything through OSD_SNAPMAPPER */
-		return 0x1ffffffffull;
-	} else {
-		return features;
-	}
-}
+
+DEFINE_CEPH_FEATURE( 0, 1, UID)
+DEFINE_CEPH_FEATURE( 1, 1, NOSRCADDR)
+DEFINE_CEPH_FEATURE_RETIRED( 2, 1, MONCLOCKCHECK, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE( 3, 1, FLOCK)
+DEFINE_CEPH_FEATURE( 4, 1, SUBSCRIBE2)
+DEFINE_CEPH_FEATURE( 5, 1, MONNAMES)
+DEFINE_CEPH_FEATURE( 6, 1, RECONNECT_SEQ)
+DEFINE_CEPH_FEATURE( 7, 1, DIRLAYOUTHASH)
+DEFINE_CEPH_FEATURE( 8, 1, OBJECTLOCATOR)
+DEFINE_CEPH_FEATURE( 9, 1, PGID64)
+DEFINE_CEPH_FEATURE(10, 1, INCSUBOSDMAP)
+DEFINE_CEPH_FEATURE(11, 1, PGPOOL3)
+DEFINE_CEPH_FEATURE(12, 1, OSDREPLYMUX)
+DEFINE_CEPH_FEATURE(13, 1, OSDENC)
+DEFINE_CEPH_FEATURE_RETIRED(14, 1, OMAP, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(14, 2, SERVER_KRAKEN)
+DEFINE_CEPH_FEATURE(15, 1, MONENC)
+DEFINE_CEPH_FEATURE_RETIRED(16, 1, QUERY_T, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(17, 1, INDEP_PG_MAP, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(18, 1, CRUSH_TUNABLES)
+DEFINE_CEPH_FEATURE_RETIRED(19, 1, CHUNKY_SCRUB, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(20, 1, MON_NULLROUTE, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(21, 1, MON_GV, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(21, 2, SERVER_LUMINOUS)
+DEFINE_CEPH_FEATURE(21, 2, RESEND_ON_SPLIT)  // overlap
+DEFINE_CEPH_FEATURE(21, 2, RADOS_BACKOFF)    // overlap
+DEFINE_CEPH_FEATURE(21, 2, OSDMAP_PG_UPMAP)  // overlap
+DEFINE_CEPH_FEATURE(21, 2, CRUSH_CHOOSE_ARGS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(22, 1, BACKFILL_RESERVATION, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(23, 1, MSG_AUTH)
+DEFINE_CEPH_FEATURE_RETIRED(24, 1, RECOVERY_RESERVATION, JEWEL, LUNINOUS)
+
+DEFINE_CEPH_FEATURE(25, 1, CRUSH_TUNABLES2)
+DEFINE_CEPH_FEATURE(26, 1, CREATEPOOLID)
+DEFINE_CEPH_FEATURE(27, 1, REPLY_CREATE_INODE)
+DEFINE_CEPH_FEATURE_RETIRED(28, 1, OSD_HBMSGS, HAMMER, JEWEL)
+DEFINE_CEPH_FEATURE(28, 2, SERVER_M)
+DEFINE_CEPH_FEATURE(29, 1, MDSENC)
+DEFINE_CEPH_FEATURE(30, 1, OSDHASHPSPOOL)
+DEFINE_CEPH_FEATURE(31, 1, MON_SINGLE_PAXOS)  // deprecate me
+DEFINE_CEPH_FEATURE_RETIRED(32, 1, OSD_SNAPMAPPER, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(33, 1, MON_SCRUB, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE_RETIRED(34, 1, OSD_PACKED_RECOVERY, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(35, 1, OSD_CACHEPOOL)
+DEFINE_CEPH_FEATURE(36, 1, CRUSH_V2)
+DEFINE_CEPH_FEATURE(37, 1, EXPORT_PEER)
+DEFINE_CEPH_FEATURE(38, 1, OSD_ERASURE_CODES)
+DEFINE_CEPH_FEATURE(38, 1, OSD_OSD_TMAP2OMAP) // overlap
+DEFINE_CEPH_FEATURE(39, 1, OSDMAP_ENC)
+DEFINE_CEPH_FEATURE(40, 1, MDS_INLINE_DATA)
+DEFINE_CEPH_FEATURE(41, 1, CRUSH_TUNABLES3)
+DEFINE_CEPH_FEATURE(41, 1, OSD_PRIMARY_AFFINITY) // overlap
+DEFINE_CEPH_FEATURE(42, 1, MSGR_KEEPALIVE2)
+DEFINE_CEPH_FEATURE(43, 1, OSD_POOLRESEND)
+DEFINE_CEPH_FEATURE(44, 1, ERASURE_CODE_PLUGINS_V2)
+DEFINE_CEPH_FEATURE_RETIRED(45, 1, OSD_SET_ALLOC_HINT, JEWEL, LUMINOUS)
+
+DEFINE_CEPH_FEATURE(46, 1, OSD_FADVISE_FLAGS)
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_REPOP, JEWEL, LUMINOUS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_OBJECT_DIGEST, JEWEL, LUMINOUS) // overlap
+DEFINE_CEPH_FEATURE_RETIRED(46, 1, OSD_TRANSACTION_MAY_LAYOUT, JEWEL, LUMINOUS) // overlap
+
+DEFINE_CEPH_FEATURE(47, 1, MDS_QUOTA)
+DEFINE_CEPH_FEATURE(48, 1, CRUSH_V4)
+DEFINE_CEPH_FEATURE_RETIRED(49, 1, OSD_MIN_SIZE_RECOVERY, JEWEL, LUMINOUS)
+DEFINE_CEPH_FEATURE_RETIRED(49, 1, OSD_PROXY_FEATURES, JEWEL, LUMINOUS) // overlap
+
+DEFINE_CEPH_FEATURE(50, 1, MON_METADATA)
+DEFINE_CEPH_FEATURE(51, 1, OSD_BITWISE_HOBJ_SORT)
+DEFINE_CEPH_FEATURE(52, 1, OSD_PROXY_WRITE_FEATURES)
+DEFINE_CEPH_FEATURE(53, 1, ERASURE_CODE_PLUGINS_V3)
+DEFINE_CEPH_FEATURE(54, 1, OSD_HITSET_GMT)
+DEFINE_CEPH_FEATURE(55, 1, HAMMER_0_94_4)
+DEFINE_CEPH_FEATURE(56, 1, NEW_OSDOP_ENCODING)
+DEFINE_CEPH_FEATURE(57, 1, MON_STATEFUL_SUB)
+DEFINE_CEPH_FEATURE(57, 1, MON_ROUTE_OSDMAP) // overlap
+DEFINE_CEPH_FEATURE(57, 1, OSDSUBOP_NO_SNAPCONTEXT) // overlap
+DEFINE_CEPH_FEATURE(57, 1, SERVER_JEWEL) // overlap
+DEFINE_CEPH_FEATURE(58, 1, CRUSH_TUNABLES5)
+DEFINE_CEPH_FEATURE(58, 1, NEW_OSDOPREPLY_ENCODING) // overlap
+DEFINE_CEPH_FEATURE(58, 1, FS_FILE_LAYOUT_V2) // overlap
+DEFINE_CEPH_FEATURE(59, 1, FS_BTIME)
+DEFINE_CEPH_FEATURE(59, 1, FS_CHANGE_ATTR) // overlap
+DEFINE_CEPH_FEATURE(59, 1, MSG_ADDR2) // overlap
+DEFINE_CEPH_FEATURE(60, 1, BLKIN_TRACING)  // *do not share this bit*
+
+DEFINE_CEPH_FEATURE(61, 1, RESERVED2)          // unused, but slow down!
+DEFINE_CEPH_FEATURE(62, 1, RESERVED)           // do not use; used as a sentinal
+DEFINE_CEPH_FEATURE_DEPRECATED(63, 1, RESERVED_BROKEN, LUMINOUS) // client-facing
+
 
 /*
  * Features supported.
@@ -113,6 +184,11 @@ static inline u64 ceph_sanitize_features(u64 features)
 	 CEPH_FEATURE_PGPOOL3 |			\
 	 CEPH_FEATURE_OSDENC |			\
 	 CEPH_FEATURE_CRUSH_TUNABLES |		\
+	 CEPH_FEATURE_SERVER_LUMINOUS |		\
+	 CEPH_FEATURE_RESEND_ON_SPLIT |		\
+	 CEPH_FEATURE_RADOS_BACKOFF |		\
+	 CEPH_FEATURE_OSDMAP_PG_UPMAP |		\
+	 CEPH_FEATURE_CRUSH_CHOOSE_ARGS |	\
 	 CEPH_FEATURE_MSG_AUTH |		\
 	 CEPH_FEATURE_CRUSH_TUNABLES2 |		\
 	 CEPH_FEATURE_REPLY_CREATE_INODE |	\
@@ -126,7 +202,11 @@ static inline u64 ceph_sanitize_features(u64 features)
 	 CEPH_FEATURE_CRUSH_TUNABLES3 |		\
 	 CEPH_FEATURE_OSD_PRIMARY_AFFINITY |	\
 	 CEPH_FEATURE_MSGR_KEEPALIVE2 |		\
+	 CEPH_FEATURE_OSD_POOLRESEND |		\
 	 CEPH_FEATURE_CRUSH_V4 |		\
+	 CEPH_FEATURE_NEW_OSDOP_ENCODING |	\
+	 CEPH_FEATURE_SERVER_JEWEL |		\
+	 CEPH_FEATURE_MON_STATEFUL_SUB |	\
 	 CEPH_FEATURE_CRUSH_TUNABLES5 |		\
 	 CEPH_FEATURE_NEW_OSDOPREPLY_ENCODING)
 

include/linux/ceph/ceph_fs.h

@@ -147,6 +147,7 @@ struct ceph_dir_layout {
 #define CEPH_MSG_OSD_OP                 42
 #define CEPH_MSG_OSD_OPREPLY            43
 #define CEPH_MSG_WATCH_NOTIFY           44
+#define CEPH_MSG_OSD_BACKOFF            61
 
 
 /* watch-notify operations */

include/linux/ceph/decode.h

@@ -132,6 +132,66 @@ static inline char *ceph_extract_encoded_string(void **p, void *end,
 	return ERR_PTR(-ERANGE);
 }
 
+/*
+ * skip helpers
+ */
+#define ceph_decode_skip_n(p, end, n, bad)			\
+	do {							\
+		ceph_decode_need(p, end, n, bad);		\
+                *p += n;					\
+	} while (0)
+
+#define ceph_decode_skip_64(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u64), bad)
+
+#define ceph_decode_skip_32(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u32), bad)
+
+#define ceph_decode_skip_16(p, end, bad)			\
+ceph_decode_skip_n(p, end, sizeof(u16), bad)
+
+#define ceph_decode_skip_8(p, end, bad)				\
+ceph_decode_skip_n(p, end, sizeof(u8), bad)
+
+#define ceph_decode_skip_string(p, end, bad)			\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		ceph_decode_skip_n(p, end, len, bad);		\
+	} while (0)
+
+#define ceph_decode_skip_set(p, end, type, bad)			\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--)					\
+			ceph_decode_skip_##type(p, end, bad);	\
+	} while (0)
+
+#define ceph_decode_skip_map(p, end, ktype, vtype, bad)		\
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--) {					\
+			ceph_decode_skip_##ktype(p, end, bad);	\
+			ceph_decode_skip_##vtype(p, end, bad);	\
+		}						\
+	} while (0)
+
+#define ceph_decode_skip_map_of_map(p, end, ktype1, ktype2, vtype2, bad) \
+	do {							\
+		u32 len;					\
+								\
+		ceph_decode_32_safe(p, end, len, bad);		\
+		while (len--) {					\
+			ceph_decode_skip_##ktype1(p, end, bad);	\
+			ceph_decode_skip_map(p, end, ktype2, vtype2, bad); \
+		}						\
+	} while (0)
+
 /*
  * struct ceph_timespec <-> struct timespec
  */

include/linux/ceph/libceph.h

@@ -184,10 +184,11 @@ static inline int calc_pages_for(u64 off, u64 len)
 		(off >> PAGE_SHIFT);
 }
 
-/*
- * These are not meant to be generic - an integer key is assumed.
- */
-#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
+#define RB_BYVAL(a)      (a)
+#define RB_BYPTR(a)      (&(a))
+#define RB_CMP3WAY(a, b) ((a) < (b) ? -1 : (a) > (b))
+
+#define DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld) \
 static void insert_##name(struct rb_root *root, type *t)		\
 {									\
 	struct rb_node **n = &root->rb_node;				\
@@ -197,11 +198,13 @@ static void insert_##name(struct rb_root *root, type *t)		\
 									\
 	while (*n) {							\
 		type *cur = rb_entry(*n, type, nodefld);		\
+		int cmp;						\
 									\
 		parent = *n;						\
-		if (t->keyfld < cur->keyfld)				\
+		cmp = cmpexp(keyexp(t->keyfld), keyexp(cur->keyfld));	\
+		if (cmp < 0)						\
 			n = &(*n)->rb_left;				\
-		else if (t->keyfld > cur->keyfld)			\
+		else if (cmp > 0)					\
 			n = &(*n)->rb_right;				\
 		else							\
 			BUG();						\
@@ -217,19 +220,24 @@ static void erase_##name(struct rb_root *root, type *t)			\
 	RB_CLEAR_NODE(&t->nodefld);					\
 }
 
-#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
-extern type __lookup_##name##_key;					\
-static type *lookup_##name(struct rb_root *root,			\
-			   typeof(__lookup_##name##_key.keyfld) key)	\
+/*
+ * @lookup_param_type is a parameter and not constructed from (@type,
+ * @keyfld) with typeof() because adding const is too unwieldy.
+ */
+#define DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,	\
+			       lookup_param_type, nodefld)		\
+static type *lookup_##name(struct rb_root *root, lookup_param_type key)	\
 {									\
 	struct rb_node *n = root->rb_node;				\
 									\
 	while (n) {							\
 		type *cur = rb_entry(n, type, nodefld);			\
+		int cmp;						\
 									\
-		if (key < cur->keyfld)					\
+		cmp = cmpexp(key, keyexp(cur->keyfld));			\
+		if (cmp < 0)						\
 			n = n->rb_left;					\
-		else if (key > cur->keyfld)				\
+		else if (cmp > 0)					\
 			n = n->rb_right;				\
 		else							\
 			return cur;					\
@@ -238,6 +246,23 @@ static type *lookup_##name(struct rb_root *root,			\
 	return NULL;							\
 }
 
+#define DEFINE_RB_FUNCS2(name, type, keyfld, cmpexp, keyexp,		\
+			 lookup_param_type, nodefld)			\
+DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, cmpexp, keyexp, nodefld)	\
+DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, cmpexp, keyexp,		\
+		       lookup_param_type, nodefld)
+
+/*
+ * Shorthands for integer keys.
+ */
+#define DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)		\
+DEFINE_RB_INSDEL_FUNCS2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL, nodefld)
+
+#define DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)		\
+extern type __lookup_##name##_key;					\
+DEFINE_RB_LOOKUP_FUNC2(name, type, keyfld, RB_CMP3WAY, RB_BYVAL,	\
+		       typeof(__lookup_##name##_key.keyfld), nodefld)
+
 #define DEFINE_RB_FUNCS(name, type, keyfld, nodefld)			\
 DEFINE_RB_INSDEL_FUNCS(name, type, keyfld, nodefld)			\
 DEFINE_RB_LOOKUP_FUNC(name, type, keyfld, nodefld)

include/linux/ceph/messenger.h

@@ -44,6 +44,8 @@ struct ceph_connection_operations {
 					struct ceph_msg_header *hdr,
 					int *skip);
 
+	void (*reencode_message) (struct ceph_msg *msg);
+
 	int (*sign_message) (struct ceph_msg *msg);
 	int (*check_message_signature) (struct ceph_msg *msg);
 };

include/linux/ceph/osd_client.h

 #ifndef _FS_CEPH_OSD_CLIENT_H
 #define _FS_CEPH_OSD_CLIENT_H
 
+#include <linux/bitrev.h>
 #include <linux/completion.h>
 #include <linux/kref.h>
 #include <linux/mempool.h>
@@ -36,6 +37,8 @@ struct ceph_osd {
 	struct ceph_connection o_con;
 	struct rb_root o_requests;
 	struct rb_root o_linger_requests;
+	struct rb_root o_backoff_mappings;
+	struct rb_root o_backoffs_by_id;
 	struct list_head o_osd_lru;
 	struct ceph_auth_handshake o_auth;
 	unsigned long lru_ttl;
@@ -136,7 +139,8 @@ struct ceph_osd_request_target {
 	struct ceph_object_id target_oid;
 	struct ceph_object_locator target_oloc;
 
-	struct ceph_pg pgid;
+	struct ceph_pg pgid;               /* last raw pg we mapped to */
+	struct ceph_spg spgid;             /* last actual spg we mapped to */
 	u32 pg_num;
 	u32 pg_num_mask;
 	struct ceph_osds acting;
@@ -148,6 +152,9 @@ struct ceph_osd_request_target {
 	unsigned int flags;                /* CEPH_OSD_FLAG_* */
 	bool paused;
 
+	u32 epoch;
+	u32 last_force_resend;
+
 	int osd;
 };
 
@@ -193,7 +200,6 @@ struct ceph_osd_request {
 	unsigned long r_stamp;                /* jiffies, send or check time */
 	unsigned long r_start_stamp;          /* jiffies */
 	int r_attempts;
-	u32 r_last_force_resend;
 	u32 r_map_dne_bound;
 
 	struct ceph_osd_req_op r_ops[];
@@ -203,6 +209,23 @@ struct ceph_request_redirect {
 	struct ceph_object_locator oloc;
 };
 
+/*
+ * osd request identifier
+ *
+ * caller name + incarnation# + tid to unique identify this request
+ */
+struct ceph_osd_reqid {
+	struct ceph_entity_name name;
+	__le64 tid;
+	__le32 inc;
+} __packed;
+
+struct ceph_blkin_trace_info {
+	__le64 trace_id;
+	__le64 span_id;
+	__le64 parent_span_id;
+} __packed;
+
 typedef void (*rados_watchcb2_t)(void *arg, u64 notify_id, u64 cookie,
 				 u64 notifier_id, void *data, size_t data_len);
 typedef void (*rados_watcherrcb_t)(void *arg, u64 cookie, int err);
@@ -221,7 +244,6 @@ struct ceph_osd_linger_request {
 	struct list_head pending_lworks;
 
 	struct ceph_osd_request_target t;
-	u32 last_force_resend;
 	u32 map_dne_bound;
 
 	struct timespec mtime;
@@ -256,6 +278,48 @@ struct ceph_watch_item {
 	struct ceph_entity_addr addr;
 };
 
+struct ceph_spg_mapping {
+	struct rb_node node;
+	struct ceph_spg spgid;
+
+	struct rb_root backoffs;
+};
+
+struct ceph_hobject_id {
+	void *key;
+	size_t key_len;
+	void *oid;
+	size_t oid_len;
+	u64 snapid;
+	u32 hash;
+	u8 is_max;
+	void *nspace;
+	size_t nspace_len;
+	s64 pool;
+
+	/* cache */
+	u32 hash_reverse_bits;
+};
+
+static inline void ceph_hoid_build_hash_cache(struct ceph_hobject_id *hoid)
+{
+	hoid->hash_reverse_bits = bitrev32(hoid->hash);
+}
+
+/*
+ * PG-wide backoff: [begin, end)
+ * per-object backoff: begin == end
+ */
+struct ceph_osd_backoff {
+	struct rb_node spg_node;
+	struct rb_node id_node;
+
+	struct ceph_spg spgid;
+	u64 id;
+	struct ceph_hobject_id *begin;
+	struct ceph_hobject_id *end;
+};
+
 #define CEPH_LINGER_ID_START	0xffff000000000000ULL
 
 struct ceph_osd_client {

include/linux/ceph/osdmap.h

@@ -24,7 +24,15 @@ struct ceph_pg {
 	uint32_t seed;
 };
 
+#define CEPH_SPG_NOSHARD	-1
+
+struct ceph_spg {
+	struct ceph_pg pgid;
+	s8 shard;
+};
+
 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs);
+int ceph_spg_compare(const struct ceph_spg *lhs, const struct ceph_spg *rhs);
 
 #define CEPH_POOL_FLAG_HASHPSPOOL	(1ULL << 0) /* hash pg seed and pool id
 						       together */
@@ -135,10 +143,14 @@ struct ceph_pg_mapping {
 		struct {
 			int len;
 			int osds[];
-		} pg_temp;
+		} pg_temp, pg_upmap;
 		struct {
 			int osd;
 		} primary_temp;
+		struct {
+			int len;
+			int from_to[][2];
+		} pg_upmap_items;
 	};
 };
 
@@ -150,13 +162,17 @@ struct ceph_osdmap {
 	u32 flags;         /* CEPH_OSDMAP_* */
 
 	u32 max_osd;       /* size of osd_state, _offload, _addr arrays */
-	u8 *osd_state;     /* CEPH_OSD_* */
+	u32 *osd_state;    /* CEPH_OSD_* */
 	u32 *osd_weight;   /* 0 = failed, 0x10000 = 100% normal */
 	struct ceph_entity_addr *osd_addr;
 
 	struct rb_root pg_temp;
 	struct rb_root primary_temp;
 
+	/* remap (post-CRUSH, pre-up) */
+	struct rb_root pg_upmap;	/* PG := raw set */
+	struct rb_root pg_upmap_items;	/* from -> to within raw set */
+
 	u32 *osd_primary_affinity;
 
 	struct rb_root pg_pools;
@@ -187,7 +203,7 @@ static inline bool ceph_osd_is_down(struct ceph_osdmap *map, int osd)
 	return !ceph_osd_is_up(map, osd);
 }
 
-extern char *ceph_osdmap_state_str(char *str, int len, int state);
+char *ceph_osdmap_state_str(char *str, int len, u32 state);
 extern u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd);
 
 static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
@@ -198,11 +214,13 @@ static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
 	return &map->osd_addr[osd];
 }
 
+#define CEPH_PGID_ENCODING_LEN		(1 + 8 + 4 + 4)
+
 static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
 {
 	__u8 version;
 
-	if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
+	if (!ceph_has_room(p, end, CEPH_PGID_ENCODING_LEN)) {
 		pr_warn("incomplete pg encoding\n");
 		return -EINVAL;
 	}
@@ -240,6 +258,8 @@ static inline void ceph_osds_init(struct ceph_osds *set)
 
 void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src);
 
+bool ceph_pg_is_split(const struct ceph_pg *pgid, u32 old_pg_num,
+		      u32 new_pg_num);
 bool ceph_is_new_interval(const struct ceph_osds *old_acting,
 			  const struct ceph_osds *new_acting,
 			  const struct ceph_osds *old_up,
@@ -262,15 +282,24 @@ extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
 					 u64 off, u64 len,
 					 u64 *bno, u64 *oxoff, u64 *oxlen);
 
+int __ceph_object_locator_to_pg(struct ceph_pg_pool_info *pi,
+				const struct ceph_object_id *oid,
+				const struct ceph_object_locator *oloc,
+				struct ceph_pg *raw_pgid);
 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
-			      struct ceph_object_id *oid,
-			      struct ceph_object_locator *oloc,
+			      const struct ceph_object_id *oid,
+			      const struct ceph_object_locator *oloc,
 			      struct ceph_pg *raw_pgid);
 
 void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap,
+			       struct ceph_pg_pool_info *pi,
 			       const struct ceph_pg *raw_pgid,
 			       struct ceph_osds *up,
 			       struct ceph_osds *acting);
+bool ceph_pg_to_primary_shard(struct ceph_osdmap *osdmap,
+			      struct ceph_pg_pool_info *pi,
+			      const struct ceph_pg *raw_pgid,
+			      struct ceph_spg *spgid);
 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
 			      const struct ceph_pg *raw_pgid);
 

include/linux/ceph/rados.h

@@ -439,6 +439,12 @@ enum {
 
 const char *ceph_osd_watch_op_name(int o);
 
+enum {
+	CEPH_OSD_BACKOFF_OP_BLOCK = 1,
+	CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
+	CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
+};
+
 /*
  * an individual object operation.  each may be accompanied by some data
  * payload

include/linux/crush/crush.h

@@ -2,6 +2,7 @@
 #define CEPH_CRUSH_CRUSH_H
 
 #ifdef __KERNEL__
+# include <linux/rbtree.h>
 # include <linux/types.h>
 #else
 # include "crush_compat.h"
@@ -137,6 +138,68 @@ struct crush_bucket {
 
 };
 
+/** @ingroup API
+ *
+ * Replacement weights for each item in a bucket. The size of the
+ * array must be exactly the size of the straw2 bucket, just as the
+ * item_weights array.
+ *
+ */
+struct crush_weight_set {
+	__u32 *weights; /*!< 16.16 fixed point weights
+                             in the same order as items */
+	__u32 size;     /*!< size of the __weights__ array */
+};
+
+/** @ingroup API
+ *
+ * Replacement weights and ids for a given straw2 bucket, for
+ * placement purposes.
+ *
+ * When crush_do_rule() chooses the Nth item from a straw2 bucket, the
+ * replacement weights found at __weight_set[N]__ are used instead of
+ * the weights from __item_weights__. If __N__ is greater than
+ * __weight_set_size__, the weights found at __weight_set_size-1__ are
+ * used instead. For instance if __weight_set__ is:
+ *
+ *    [ [ 0x10000, 0x20000 ],   // position 0
+ *      [ 0x20000, 0x40000 ] ]  // position 1
+ *
+ * choosing the 0th item will use position 0 weights [ 0x10000, 0x20000 ]
+ * choosing the 1th item will use position 1 weights [ 0x20000, 0x40000 ]
+ * choosing the 2th item will use position 1 weights [ 0x20000, 0x40000 ]
+ * etc.
+ *
+ */
+struct crush_choose_arg {
+	__s32 *ids;            /*!< values to use instead of items */
+	__u32 ids_size;        /*!< size of the __ids__ array */
+	struct crush_weight_set *weight_set; /*!< weight replacements for
+                                                  a given position */
+	__u32 weight_set_size; /*!< size of the __weight_set__ array */
+};
+
+/** @ingroup API
+ *
+ * Replacement weights and ids for each bucket in the crushmap. The
+ * __size__ of the __args__ array must be exactly the same as the
+ * __map->max_buckets__.
+ *
+ * The __crush_choose_arg__ at index N will be used when choosing
+ * an item from the bucket __map->buckets[N]__ bucket, provided it
+ * is a straw2 bucket.
+ *
+ */
+struct crush_choose_arg_map {
+#ifdef __KERNEL__
+	struct rb_node node;
+	u64 choose_args_index;
+#endif
+	struct crush_choose_arg *args; /*!< replacement for each bucket
+                                            in the crushmap */
+	__u32 size;                    /*!< size of the __args__ array */
+};
+
 struct crush_bucket_uniform {
 	struct crush_bucket h;
 	__u32 item_weight;  /* 16-bit fixed point; all items equally weighted */
@@ -236,6 +299,9 @@ struct crush_map {
 	__u32 allowed_bucket_algs;
 
 	__u32 *choose_tries;
+#else
+	/* CrushWrapper::choose_args */
+	struct rb_root choose_args;
 #endif
 };
 

include/linux/crush/mapper.h

@@ -11,11 +11,10 @@
 #include "crush.h"
 
 extern int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size);
-extern int crush_do_rule(const struct crush_map *map,
-			 int ruleno,
-			 int x, int *result, int result_max,
-			 const __u32 *weights, int weight_max,
-			 void *cwin);
+int crush_do_rule(const struct crush_map *map,
+		  int ruleno, int x, int *result, int result_max,
+		  const __u32 *weight, int weight_max,
+		  void *cwin, const struct crush_choose_arg *choose_args);
 
 /*
  * Returns the exact amount of workspace that will need to be used

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";
 	}
 }

net/ceph/crush/crush.c

 #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);
 }

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;
 				}
 			}

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;
 }

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);

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",

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 @@ static void osd_fault(struct ceph_connection *con)
 	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,

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 @@ static int crush_decode_straw2_bucket(void **p, void *end,
 	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");
@@ -418,74 +587,48 @@ 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
  */
@@ -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 @@ static int __decode_pg_temp(void **p, void *end, struct ceph_osdmap *map,
 	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 @@ static int decode_new_primary_affinity(void **p, void *end,
 	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 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 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);