forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-10-22 8ac6c7a54ed1b98d142dce24b11c6de6a1e239a5 
 kernel/net/ceph/messenger.c
....@@ -156,7 +156,6 @@
156156 /* Slab caches for frequently-allocated structures */
157157 
158158 static struct kmem_cache	*ceph_msg_cache;
159
-static struct kmem_cache	*ceph_msg_data_cache;
160159 
161160 /* static tag bytes (protocol control messages) */
162161 static char tag_msg = CEPH_MSGR_TAG_MSG;
....@@ -187,30 +186,33 @@
187186 
188187 static struct page *zero_page;		/* used in certain error cases */
189188 
190
-const char *ceph_pr_addr(const struct sockaddr_storage *ss)
189
+const char *ceph_pr_addr(const struct ceph_entity_addr *addr)
191190 {
192191 	int i;
193192 	char *s;
194
-	struct sockaddr_in *in4 = (struct sockaddr_in *) ss;
195
-	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss;
193
+	struct sockaddr_storage ss = addr->in_addr; /* align */
194
+	struct sockaddr_in *in4 = (struct sockaddr_in *)&ss;
195
+	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *)&ss;
196196 
197197 	i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK;
198198 	s = addr_str[i];
199199 
200
-	switch (ss->ss_family) {
200
+	switch (ss.ss_family) {
201201 	case AF_INET:
202
-		snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%hu", &in4->sin_addr,
202
+		snprintf(s, MAX_ADDR_STR_LEN, "(%d)%pI4:%hu",
203
+			 le32_to_cpu(addr->type), &in4->sin_addr,
203204 			 ntohs(in4->sin_port));
204205 		break;
205206 
206207 	case AF_INET6:
207
-		snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%hu", &in6->sin6_addr,
208
+		snprintf(s, MAX_ADDR_STR_LEN, "(%d)[%pI6c]:%hu",
209
+			 le32_to_cpu(addr->type), &in6->sin6_addr,
208210 			 ntohs(in6->sin6_port));
209211 		break;
210212 
211213 	default:
212214 		snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)",
213
-			 ss->ss_family);
215
+			 ss.ss_family);
214216 	}
215217 
216218 	return s;
....@@ -220,7 +222,7 @@
220222 static void encode_my_addr(struct ceph_messenger *msgr)
221223 {
222224 	memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr));
223
-	ceph_encode_addr(&msgr->my_enc_addr);
225
+	ceph_encode_banner_addr(&msgr->my_enc_addr);
224226 }
225227 
226228 /*
....@@ -235,23 +237,11 @@
235237 	if (!ceph_msg_cache)
236238 		return -ENOMEM;
237239 
238
-	BUG_ON(ceph_msg_data_cache);
239
-	ceph_msg_data_cache = KMEM_CACHE(ceph_msg_data, 0);
240
-	if (ceph_msg_data_cache)
241
-		return 0;
242
-
243
-	kmem_cache_destroy(ceph_msg_cache);
244
-	ceph_msg_cache = NULL;
245
-
246
-	return -ENOMEM;
240
+	return 0;
247241 }
248242 
249243 static void ceph_msgr_slab_exit(void)
250244 {
251
-	BUG_ON(!ceph_msg_data_cache);
252
-	kmem_cache_destroy(ceph_msg_data_cache);
253
-	ceph_msg_data_cache = NULL;
254
-
255245 	BUG_ON(!ceph_msg_cache);
256246 	kmem_cache_destroy(ceph_msg_cache);
257247 	ceph_msg_cache = NULL;
....@@ -422,7 +412,7 @@
422412 	switch (sk->sk_state) {
423413 	case TCP_CLOSE:
424414 		dout("%s TCP_CLOSE\n", __func__);
425
-		/* fall through */
415
+		fallthrough;
426416 	case TCP_CLOSE_WAIT:
427417 		dout("%s TCP_CLOSE_WAIT\n", __func__);
428418 		con_sock_state_closing(con);
....@@ -462,7 +452,7 @@
462452  */
463453 static int ceph_tcp_connect(struct ceph_connection *con)
464454 {
465
-	struct sockaddr_storage *paddr = &con->peer_addr.in_addr;
455
+	struct sockaddr_storage ss = con->peer_addr.in_addr; /* align */
466456 	struct socket *sock;
467457 	unsigned int noio_flag;
468458 	int ret;
....@@ -471,7 +461,7 @@
471461 
472462 	/* sock_create_kern() allocates with GFP_KERNEL */
473463 	noio_flag = memalloc_noio_save();
474
-	ret = sock_create_kern(read_pnet(&con->msgr->net), paddr->ss_family,
464
+	ret = sock_create_kern(read_pnet(&con->msgr->net), ss.ss_family,
475465 			       SOCK_STREAM, IPPROTO_TCP, &sock);
476466 	memalloc_noio_restore(noio_flag);
477467 	if (ret)
....@@ -484,31 +474,24 @@
484474 
485475 	set_sock_callbacks(sock, con);
486476 
487
-	dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr));
477
+	dout("connect %s\n", ceph_pr_addr(&con->peer_addr));
488478 
489479 	con_sock_state_connecting(con);
490
-	ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr),
480
+	ret = sock->ops->connect(sock, (struct sockaddr *)&ss, sizeof(ss),
491481 				 O_NONBLOCK);
492482 	if (ret == -EINPROGRESS) {
493483 		dout("connect %s EINPROGRESS sk_state = %u\n",
494
-		     ceph_pr_addr(&con->peer_addr.in_addr),
484
+		     ceph_pr_addr(&con->peer_addr),
495485 		     sock->sk->sk_state);
496486 	} else if (ret < 0) {
497487 		pr_err("connect %s error %d\n",
498
-		       ceph_pr_addr(&con->peer_addr.in_addr), ret);
488
+		       ceph_pr_addr(&con->peer_addr), ret);
499489 		sock_release(sock);
500490 		return ret;
501491 	}
502492 
503
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
504
-		int optval = 1;
505
-
506
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
507
-					(char *)&optval, sizeof(optval));
508
-		if (ret)
509
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
510
-			       ret);
511
-	}
493
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
494
+		tcp_sock_set_nodelay(sock->sk);
512495 
513496 	con->sock = sock;
514497 	return 0;
....@@ -526,7 +509,7 @@
526509 	if (!buf)
527510 		msg.msg_flags |= MSG_TRUNC;
528511 
529
-	iov_iter_kvec(&msg.msg_iter, READ | ITER_KVEC, &iov, 1, len);
512
+	iov_iter_kvec(&msg.msg_iter, READ, &iov, 1, len);
530513 	r = sock_recvmsg(sock, &msg, msg.msg_flags);
531514 	if (r == -EAGAIN)
532515 		r = 0;
....@@ -545,7 +528,7 @@
545528 	int r;
546529 
547530 	BUG_ON(page_offset + length > PAGE_SIZE);
548
-	iov_iter_bvec(&msg.msg_iter, READ | ITER_BVEC, &bvec, 1, length);
531
+	iov_iter_bvec(&msg.msg_iter, READ, &bvec, 1, length);
549532 	r = sock_recvmsg(sock, &msg, msg.msg_flags);
550533 	if (r == -EAGAIN)
551534 		r = 0;
....@@ -557,7 +540,7 @@
557540  * shortly.
558541  */
559542 static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
560
-		     size_t kvlen, size_t len, int more)
543
+			    size_t kvlen, size_t len, bool more)
561544 {
562545 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
563546 	int r;
....@@ -573,24 +556,15 @@
573556 	return r;
574557 }
575558 
576
-static int __ceph_tcp_sendpage(struct socket *sock, struct page *page,
577
-		     int offset, size_t size, bool more)
578
-{
579
-	int flags = MSG_DONTWAIT | MSG_NOSIGNAL | (more ? MSG_MORE : MSG_EOR);
580
-	int ret;
581
-
582
-	ret = kernel_sendpage(sock, page, offset, size, flags);
583
-	if (ret == -EAGAIN)
584
-		ret = 0;
585
-
586
-	return ret;
587
-}
588
-
559
+/*
560
+ * @more: either or both of MSG_MORE and MSG_SENDPAGE_NOTLAST
561
+ */
589562 static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
590
-		     int offset, size_t size, bool more)
563
+			     int offset, size_t size, int more)
591564 {
592
-	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
593
-	struct bio_vec bvec;
565
+	ssize_t (*sendpage)(struct socket *sock, struct page *page,
566
+			    int offset, size_t size, int flags);
567
+	int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
594568 	int ret;
595569 
596570 	/*
....@@ -601,20 +575,12 @@
601575 	 * coalescing neighboring slab objects into a single frag which
602576 	 * triggers one of hardened usercopy checks.
603577 	 */
604
-	if (page_count(page) >= 1 && !PageSlab(page))
605
-		return __ceph_tcp_sendpage(sock, page, offset, size, more);
606
-
607
-	bvec.bv_page = page;
608
-	bvec.bv_offset = offset;
609
-	bvec.bv_len = size;
610
-
611
-	if (more)
612
-		msg.msg_flags |= MSG_MORE;
578
+	if (sendpage_ok(page))
579
+		sendpage = sock->ops->sendpage;
613580 	else
614
-		msg.msg_flags |= MSG_EOR;  /* superfluous, but what the hell */
581
+		sendpage = sock_no_sendpage;
615582 
616
-	iov_iter_bvec(&msg.msg_iter, WRITE | ITER_BVEC, &bvec, 1, size);
617
-	ret = sock_sendmsg(sock, &msg);
583
+	ret = sendpage(sock, page, offset, size, flags);
618584 	if (ret == -EAGAIN)
619585 		ret = 0;
620586 
....@@ -699,8 +665,7 @@
699665 void ceph_con_close(struct ceph_connection *con)
700666 {
701667 	mutex_lock(&con->mutex);
702
-	dout("con_close %p peer %s\n", con,
703
-	     ceph_pr_addr(&con->peer_addr.in_addr));
668
+	dout("con_close %p peer %s\n", con, ceph_pr_addr(&con->peer_addr));
704669 	con->state = CON_STATE_CLOSED;
705670 
706671 	con_flag_clear(con, CON_FLAG_LOSSYTX);	/* so we retry next connect */
....@@ -724,7 +689,7 @@
724689 		   struct ceph_entity_addr *addr)
725690 {
726691 	mutex_lock(&con->mutex);
727
-	dout("con_open %p %s\n", con, ceph_pr_addr(&addr->in_addr));
692
+	dout("con_open %p %s\n", con, ceph_pr_addr(addr));
728693 
729694 	WARN_ON(con->state != CON_STATE_CLOSED);
730695 	con->state = CON_STATE_PREOPEN;
....@@ -870,6 +835,7 @@
870835 					size_t bytes)
871836 {
872837 	struct ceph_bio_iter *it = &cursor->bio_iter;
838
+	struct page *page = bio_iter_page(it->bio, it->iter);
873839 
874840 	BUG_ON(bytes > cursor->resid);
875841 	BUG_ON(bytes > bio_iter_len(it->bio, it->iter));
....@@ -881,7 +847,8 @@
881847 		return false;   /* no more data */
882848 	}
883849 
884
-	if (!bytes || (it->iter.bi_size && it->iter.bi_bvec_done))
850
+	if (!bytes || (it->iter.bi_size && it->iter.bi_bvec_done &&
851
+		       page == bio_iter_page(it->bio, it->iter)))
885852 		return false;	/* more bytes to process in this segment */
886853 
887854 	if (!it->iter.bi_size) {
....@@ -929,6 +896,7 @@
929896 					size_t bytes)
930897 {
931898 	struct bio_vec *bvecs = cursor->data->bvec_pos.bvecs;
899
+	struct page *page = bvec_iter_page(bvecs, cursor->bvec_iter);
932900 
933901 	BUG_ON(bytes > cursor->resid);
934902 	BUG_ON(bytes > bvec_iter_len(bvecs, cursor->bvec_iter));
....@@ -940,7 +908,8 @@
940908 		return false;   /* no more data */
941909 	}
942910 
943
-	if (!bytes || cursor->bvec_iter.bi_bvec_done)
911
+	if (!bytes || (cursor->bvec_iter.bi_bvec_done &&
912
+		       page == bvec_iter_page(bvecs, cursor->bvec_iter)))
944913 		return false;	/* more bytes to process in this segment */
945914 
946915 	BUG_ON(cursor->last_piece);
....@@ -1147,16 +1116,13 @@
11471116 static void ceph_msg_data_cursor_init(struct ceph_msg *msg, size_t length)
11481117 {
11491118 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
1150
-	struct ceph_msg_data *data;
11511119 
11521120 	BUG_ON(!length);
11531121 	BUG_ON(length > msg->data_length);
1154
-	BUG_ON(list_empty(&msg->data));
1122
+	BUG_ON(!msg->num_data_items);
11551123 
1156
-	cursor->data_head = &msg->data;
11571124 	cursor->total_resid = length;
1158
-	data = list_first_entry(&msg->data, struct ceph_msg_data, links);
1159
-	cursor->data = data;
1125
+	cursor->data = msg->data;
11601126 
11611127 	__ceph_msg_data_cursor_init(cursor);
11621128 }
....@@ -1237,8 +1203,7 @@
12371203 
12381204 	if (!cursor->resid && cursor->total_resid) {
12391205 		WARN_ON(!cursor->last_piece);
1240
-		BUG_ON(list_is_last(&cursor->data->links, cursor->data_head));
1241
-		cursor->data = list_next_entry(cursor->data, links);
1206
+		cursor->data++;
12421207 		__ceph_msg_data_cursor_init(cursor);
12431208 		new_piece = true;
12441209 	}
....@@ -1254,9 +1219,6 @@
12541219 
12551220 static void prepare_message_data(struct ceph_msg *msg, u32 data_len)
12561221 {
1257
-	BUG_ON(!msg);
1258
-	BUG_ON(!data_len);
1259
-
12601222 	/* Initialize data cursor */
12611223 
12621224 	ceph_msg_data_cursor_init(msg, (size_t)data_len);
....@@ -1592,11 +1554,12 @@
15921554 	struct ceph_msg *msg = con->out_msg;
15931555 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
15941556 	bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
1557
+	int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
15951558 	u32 crc;
15961559 
15971560 	dout("%s %p msg %p\n", __func__, con, msg);
15981561 
1599
-	if (list_empty(&msg->data))
1562
+	if (!msg->num_data_items)
16001563 		return -EINVAL;
16011564 
16021565 	/*
....@@ -1612,7 +1575,6 @@
16121575 		struct page *page;
16131576 		size_t page_offset;
16141577 		size_t length;
1615
-		bool last_piece;
16161578 		int ret;
16171579 
16181580 		if (!cursor->resid) {
....@@ -1620,10 +1582,11 @@
16201582 			continue;
16211583 		}
16221584 
1623
-		page = ceph_msg_data_next(cursor, &page_offset, &length,
1624
-					  &last_piece);
1625
-		ret = ceph_tcp_sendpage(con->sock, page, page_offset,
1626
-					length, !last_piece);
1585
+		page = ceph_msg_data_next(cursor, &page_offset, &length, NULL);
1586
+		if (length == cursor->total_resid)
1587
+			more = MSG_MORE;
1588
+		ret = ceph_tcp_sendpage(con->sock, page, page_offset, length,
1589
+					more);
16271590 		if (ret <= 0) {
16281591 			if (do_datacrc)
16291592 				msg->footer.data_crc = cpu_to_le32(crc);
....@@ -1653,13 +1616,16 @@
16531616  */
16541617 static int write_partial_skip(struct ceph_connection *con)
16551618 {
1619
+	int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
16561620 	int ret;
16571621 
16581622 	dout("%s %p %d left\n", __func__, con, con->out_skip);
16591623 	while (con->out_skip > 0) {
16601624 		size_t size = min(con->out_skip, (int) PAGE_SIZE);
16611625 
1662
-		ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, true);
1626
+		if (size == con->out_skip)
1627
+			more = MSG_MORE;
1628
+		ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, more);
16631629 		if (ret <= 0)
16641630 			goto out;
16651631 		con->out_skip -= ret;
....@@ -1761,12 +1727,14 @@
17611727 	ret = read_partial(con, end, size, &con->actual_peer_addr);
17621728 	if (ret <= 0)
17631729 		goto out;
1730
+	ceph_decode_banner_addr(&con->actual_peer_addr);
17641731 
17651732 	size = sizeof (con->peer_addr_for_me);
17661733 	end += size;
17671734 	ret = read_partial(con, end, size, &con->peer_addr_for_me);
17681735 	if (ret <= 0)
17691736 		goto out;
1737
+	ceph_decode_banner_addr(&con->peer_addr_for_me);
17701738 
17711739 out:
17721740 	return ret;
....@@ -1817,21 +1785,22 @@
18171785 {
18181786 	if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) {
18191787 		pr_err("connect to %s got bad banner\n",
1820
-		       ceph_pr_addr(&con->peer_addr.in_addr));
1788
+		       ceph_pr_addr(&con->peer_addr));
18211789 		con->error_msg = "protocol error, bad banner";
18221790 		return -1;
18231791 	}
18241792 	return 0;
18251793 }
18261794 
1827
-static bool addr_is_blank(struct sockaddr_storage *ss)
1795
+static bool addr_is_blank(struct ceph_entity_addr *addr)
18281796 {
1829
-	struct in_addr *addr = &((struct sockaddr_in *)ss)->sin_addr;
1830
-	struct in6_addr *addr6 = &((struct sockaddr_in6 *)ss)->sin6_addr;
1797
+	struct sockaddr_storage ss = addr->in_addr; /* align */
1798
+	struct in_addr *addr4 = &((struct sockaddr_in *)&ss)->sin_addr;
1799
+	struct in6_addr *addr6 = &((struct sockaddr_in6 *)&ss)->sin6_addr;
18311800 
1832
-	switch (ss->ss_family) {
1801
+	switch (ss.ss_family) {
18331802 	case AF_INET:
1834
-		return addr->s_addr == htonl(INADDR_ANY);
1803
+		return addr4->s_addr == htonl(INADDR_ANY);
18351804 	case AF_INET6:
18361805 		return ipv6_addr_any(addr6);
18371806 	default:
....@@ -1839,25 +1808,25 @@
18391808 	}
18401809 }
18411810 
1842
-static int addr_port(struct sockaddr_storage *ss)
1811
+static int addr_port(struct ceph_entity_addr *addr)
18431812 {
1844
-	switch (ss->ss_family) {
1813
+	switch (get_unaligned(&addr->in_addr.ss_family)) {
18451814 	case AF_INET:
1846
-		return ntohs(((struct sockaddr_in *)ss)->sin_port);
1815
+		return ntohs(get_unaligned(&((struct sockaddr_in *)&addr->in_addr)->sin_port));
18471816 	case AF_INET6:
1848
-		return ntohs(((struct sockaddr_in6 *)ss)->sin6_port);
1817
+		return ntohs(get_unaligned(&((struct sockaddr_in6 *)&addr->in_addr)->sin6_port));
18491818 	}
18501819 	return 0;
18511820 }
18521821 
1853
-static void addr_set_port(struct sockaddr_storage *ss, int p)
1822
+static void addr_set_port(struct ceph_entity_addr *addr, int p)
18541823 {
1855
-	switch (ss->ss_family) {
1824
+	switch (get_unaligned(&addr->in_addr.ss_family)) {
18561825 	case AF_INET:
1857
-		((struct sockaddr_in *)ss)->sin_port = htons(p);
1826
+		put_unaligned(htons(p), &((struct sockaddr_in *)&addr->in_addr)->sin_port);
18581827 		break;
18591828 	case AF_INET6:
1860
-		((struct sockaddr_in6 *)ss)->sin6_port = htons(p);
1829
+		put_unaligned(htons(p), &((struct sockaddr_in6 *)&addr->in_addr)->sin6_port);
18611830 		break;
18621831 	}
18631832 }
....@@ -1865,21 +1834,18 @@
18651834 /*
18661835  * Unlike other *_pton function semantics, zero indicates success.
18671836  */
1868
-static int ceph_pton(const char *str, size_t len, struct sockaddr_storage *ss,
1837
+static int ceph_pton(const char *str, size_t len, struct ceph_entity_addr *addr,
18691838 		char delim, const char **ipend)
18701839 {
1871
-	struct sockaddr_in *in4 = (struct sockaddr_in *) ss;
1872
-	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss;
1840
+	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
18731841 
1874
-	memset(ss, 0, sizeof(*ss));
1875
-
1876
-	if (in4_pton(str, len, (u8 *)&in4->sin_addr.s_addr, delim, ipend)) {
1877
-		ss->ss_family = AF_INET;
1842
+	if (in4_pton(str, len, (u8 *)&((struct sockaddr_in *)&addr->in_addr)->sin_addr.s_addr, delim, ipend)) {
1843
+		put_unaligned(AF_INET, &addr->in_addr.ss_family);
18781844 		return 0;
18791845 	}
18801846 
1881
-	if (in6_pton(str, len, (u8 *)&in6->sin6_addr.s6_addr, delim, ipend)) {
1882
-		ss->ss_family = AF_INET6;
1847
+	if (in6_pton(str, len, (u8 *)&((struct sockaddr_in6 *)&addr->in_addr)->sin6_addr.s6_addr, delim, ipend)) {
1848
+		put_unaligned(AF_INET6, &addr->in_addr.ss_family);
18831849 		return 0;
18841850 	}
18851851 
....@@ -1891,7 +1857,7 @@
18911857  */
18921858 #ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER
18931859 static int ceph_dns_resolve_name(const char *name, size_t namelen,
1894
-		struct sockaddr_storage *ss, char delim, const char **ipend)
1860
+		struct ceph_entity_addr *addr, char delim, const char **ipend)
18951861 {
18961862 	const char *end, *delim_p;
18971863 	char *colon_p, *ip_addr = NULL;
....@@ -1918,9 +1884,10 @@
19181884 		return -EINVAL;
19191885 
19201886 	/* do dns_resolve upcall */
1921
-	ip_len = dns_query(NULL, name, end - name, NULL, &ip_addr, NULL);
1887
+	ip_len = dns_query(current->nsproxy->net_ns,
1888
+			   NULL, name, end - name, NULL, &ip_addr, NULL, false);
19221889 	if (ip_len > 0)
1923
-		ret = ceph_pton(ip_addr, ip_len, ss, -1, NULL);
1890
+		ret = ceph_pton(ip_addr, ip_len, addr, -1, NULL);
19241891 	else
19251892 		ret = -ESRCH;
19261893 
....@@ -1929,13 +1896,13 @@
19291896 	*ipend = end;
19301897 
19311898 	pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name,
1932
-			ret, ret ? "failed" : ceph_pr_addr(ss));
1899
+			ret, ret ? "failed" : ceph_pr_addr(addr));
19331900 
19341901 	return ret;
19351902 }
19361903 #else
19371904 static inline int ceph_dns_resolve_name(const char *name, size_t namelen,
1938
-		struct sockaddr_storage *ss, char delim, const char **ipend)
1905
+		struct ceph_entity_addr *addr, char delim, const char **ipend)
19391906 {
19401907 	return -EINVAL;
19411908 }
....@@ -1946,13 +1913,13 @@
19461913  * then try to extract a hostname to resolve using userspace DNS upcall.
19471914  */
19481915 static int ceph_parse_server_name(const char *name, size_t namelen,
1949
-			struct sockaddr_storage *ss, char delim, const char **ipend)
1916
+		struct ceph_entity_addr *addr, char delim, const char **ipend)
19501917 {
19511918 	int ret;
19521919 
1953
-	ret = ceph_pton(name, namelen, ss, delim, ipend);
1920
+	ret = ceph_pton(name, namelen, addr, delim, ipend);
19541921 	if (ret)
1955
-		ret = ceph_dns_resolve_name(name, namelen, ss, delim, ipend);
1922
+		ret = ceph_dns_resolve_name(name, namelen, addr, delim, ipend);
19561923 
19571924 	return ret;
19581925 }
....@@ -1971,7 +1938,6 @@
19711938 	dout("parse_ips on '%.*s'\n", (int)(end-c), c);
19721939 	for (i = 0; i < max_count; i++) {
19731940 		const char *ipend;
1974
-		struct sockaddr_storage *ss = &addr[i].in_addr;
19751941 		int port;
19761942 		char delim = ',';
19771943 
....@@ -1980,7 +1946,7 @@
19801946 			p++;
19811947 		}
19821948 
1983
-		ret = ceph_parse_server_name(p, end - p, ss, delim, &ipend);
1949
+		ret = ceph_parse_server_name(p, end - p, &addr[i], delim, &ipend);
19841950 		if (ret)
19851951 			goto bad;
19861952 		ret = -EINVAL;
....@@ -2011,9 +1977,10 @@
20111977 			port = CEPH_MON_PORT;
20121978 		}
20131979 
2014
-		addr_set_port(ss, port);
1980
+		addr_set_port(&addr[i], port);
1981
+		addr[i].type = CEPH_ENTITY_ADDR_TYPE_LEGACY;
20151982 
2016
-		dout("parse_ips got %s\n", ceph_pr_addr(ss));
1983
+		dout("parse_ips got %s\n", ceph_pr_addr(&addr[i]));
20171984 
20181985 		if (p == end)
20191986 			break;
....@@ -2030,10 +1997,8 @@
20301997 	return 0;
20311998 
20321999 bad:
2033
-	pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c);
20342000 	return ret;
20352001 }
2036
-EXPORT_SYMBOL(ceph_parse_ips);
20372002 
20382003 static int process_banner(struct ceph_connection *con)
20392004 {
....@@ -2042,9 +2007,6 @@
20422007 	if (verify_hello(con) < 0)
20432008 		return -1;
20442009 
2045
-	ceph_decode_addr(&con->actual_peer_addr);
2046
-	ceph_decode_addr(&con->peer_addr_for_me);
2047
-
20482010 	/*
20492011 	 * Make sure the other end is who we wanted.  note that the other
20502012 	 * end may not yet know their ip address, so if it's 0.0.0.0, give
....@@ -2052,13 +2014,13 @@
20522014 	 */
20532015 	if (memcmp(&con->peer_addr, &con->actual_peer_addr,
20542016 		   sizeof(con->peer_addr)) != 0 &&
2055
-	    !(addr_is_blank(&con->actual_peer_addr.in_addr) &&
2017
+	    !(addr_is_blank(&con->actual_peer_addr) &&
20562018 	      con->actual_peer_addr.nonce == con->peer_addr.nonce)) {
2057
-		pr_warn("wrong peer, want %s/%d, got %s/%d\n",
2058
-			ceph_pr_addr(&con->peer_addr.in_addr),
2059
-			(int)le32_to_cpu(con->peer_addr.nonce),
2060
-			ceph_pr_addr(&con->actual_peer_addr.in_addr),
2061
-			(int)le32_to_cpu(con->actual_peer_addr.nonce));
2019
+		pr_warn("wrong peer, want %s/%u, got %s/%u\n",
2020
+			ceph_pr_addr(&con->peer_addr),
2021
+			le32_to_cpu(con->peer_addr.nonce),
2022
+			ceph_pr_addr(&con->actual_peer_addr),
2023
+			le32_to_cpu(con->actual_peer_addr.nonce));
20622024 		con->error_msg = "wrong peer at address";
20632025 		return -1;
20642026 	}
....@@ -2066,16 +2028,16 @@
20662028 	/*
20672029 	 * did we learn our address?
20682030 	 */
2069
-	if (addr_is_blank(&con->msgr->inst.addr.in_addr)) {
2070
-		int port = addr_port(&con->msgr->inst.addr.in_addr);
2031
+	if (addr_is_blank(&con->msgr->inst.addr)) {
2032
+		int port = addr_port(&con->msgr->inst.addr);
20712033 
20722034 		memcpy(&con->msgr->inst.addr.in_addr,
20732035 		       &con->peer_addr_for_me.in_addr,
20742036 		       sizeof(con->peer_addr_for_me.in_addr));
2075
-		addr_set_port(&con->msgr->inst.addr.in_addr, port);
2037
+		addr_set_port(&con->msgr->inst.addr, port);
20762038 		encode_my_addr(con->msgr);
20772039 		dout("process_banner learned my addr is %s\n",
2078
-		     ceph_pr_addr(&con->msgr->inst.addr.in_addr));
2040
+		     ceph_pr_addr(&con->msgr->inst.addr));
20792041 	}
20802042 
20812043 	return 0;
....@@ -2126,7 +2088,7 @@
21262088 		pr_err("%s%lld %s feature set mismatch,"
21272089 		       " my %llx < server's %llx, missing %llx\n",
21282090 		       ENTITY_NAME(con->peer_name),
2129
-		       ceph_pr_addr(&con->peer_addr.in_addr),
2091
+		       ceph_pr_addr(&con->peer_addr),
21302092 		       sup_feat, server_feat, server_feat & ~sup_feat);
21312093 		con->error_msg = "missing required protocol features";
21322094 		reset_connection(con);
....@@ -2136,7 +2098,7 @@
21362098 		pr_err("%s%lld %s protocol version mismatch,"
21372099 		       " my %d != server's %d\n",
21382100 		       ENTITY_NAME(con->peer_name),
2139
-		       ceph_pr_addr(&con->peer_addr.in_addr),
2101
+		       ceph_pr_addr(&con->peer_addr),
21402102 		       le32_to_cpu(con->out_connect.protocol_version),
21412103 		       le32_to_cpu(con->in_reply.protocol_version));
21422104 		con->error_msg = "protocol version mismatch";
....@@ -2170,7 +2132,7 @@
21702132 		     le32_to_cpu(con->in_reply.connect_seq));
21712133 		pr_err("%s%lld %s connection reset\n",
21722134 		       ENTITY_NAME(con->peer_name),
2173
-		       ceph_pr_addr(&con->peer_addr.in_addr));
2135
+		       ceph_pr_addr(&con->peer_addr));
21742136 		reset_connection(con);
21752137 		con_out_kvec_reset(con);
21762138 		ret = prepare_write_connect(con);
....@@ -2227,7 +2189,7 @@
22272189 			pr_err("%s%lld %s protocol feature mismatch,"
22282190 			       " my required %llx > server's %llx, need %llx\n",
22292191 			       ENTITY_NAME(con->peer_name),
2230
-			       ceph_pr_addr(&con->peer_addr.in_addr),
2192
+			       ceph_pr_addr(&con->peer_addr),
22312193 			       req_feat, server_feat, req_feat & ~server_feat);
22322194 			con->error_msg = "missing required protocol features";
22332195 			reset_connection(con);
....@@ -2356,8 +2318,7 @@
23562318 	u32 crc = 0;
23572319 	int ret;
23582320 
2359
-	BUG_ON(!msg);
2360
-	if (list_empty(&msg->data))
2321
+	if (!msg->num_data_items)
23612322 		return -EIO;
23622323 
23632324 	if (do_datacrc)
....@@ -2435,7 +2396,7 @@
24352396 	if ((s64)seq - (s64)con->in_seq < 1) {
24362397 		pr_info("skipping %s%lld %s seq %lld expected %lld\n",
24372398 			ENTITY_NAME(con->peer_name),
2438
-			ceph_pr_addr(&con->peer_addr.in_addr),
2399
+			ceph_pr_addr(&con->peer_addr),
24392400 			seq, con->in_seq + 1);
24402401 		con->in_base_pos = -front_len - middle_len - data_len -
24412402 			sizeof_footer(con);
....@@ -2790,7 +2751,7 @@
27902751 			switch (ret) {
27912752 			case -EBADMSG:
27922753 				con->error_msg = "bad crc/signature";
2793
-				/* fall through */
2754
+				fallthrough;
27942755 			case -EBADE:
27952756 				ret = -EIO;
27962757 				break;
....@@ -2850,13 +2811,13 @@
28502811 		return -ENOENT;
28512812 	}
28522813 
2814
+	dout("%s %p %lu\n", __func__, con, delay);
28532815 	if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) {
28542816 		dout("%s %p - already queued\n", __func__, con);
28552817 		con->ops->put(con);
28562818 		return -EBUSY;
28572819 	}
28582820 
2859
-	dout("%s %p %lu\n", __func__, con, delay);
28602821 	return 0;
28612822 }
28622823 
....@@ -3014,10 +2975,10 @@
30142975 static void con_fault(struct ceph_connection *con)
30152976 {
30162977 	dout("fault %p state %lu to peer %s\n",
3017
-	     con, con->state, ceph_pr_addr(&con->peer_addr.in_addr));
2978
+	     con, con->state, ceph_pr_addr(&con->peer_addr));
30182979 
30192980 	pr_warn("%s%lld %s %s\n", ENTITY_NAME(con->peer_name),
3020
-		ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg);
2981
+		ceph_pr_addr(&con->peer_addr), con->error_msg);
30212982 	con->error_msg = NULL;
30222983 
30232984 	WARN_ON(con->state != CON_STATE_CONNECTING &&
....@@ -3066,6 +3027,12 @@
30663027 }
30673028 
30683029 
3030
+void ceph_messenger_reset_nonce(struct ceph_messenger *msgr)
3031
+{
3032
+	u32 nonce = le32_to_cpu(msgr->inst.addr.nonce) + 1000000;
3033
+	msgr->inst.addr.nonce = cpu_to_le32(nonce);
3034
+	encode_my_addr(msgr);
3035
+}
30693036 
30703037 /*
30713038  * initialize a new messenger instance
....@@ -3271,49 +3238,37 @@
32713238 	return false;
32723239 }
32733240 
3274
-static struct ceph_msg_data *ceph_msg_data_create(enum ceph_msg_data_type type)
3241
+static struct ceph_msg_data *ceph_msg_data_add(struct ceph_msg *msg)
32753242 {
3276
-	struct ceph_msg_data *data;
3277
-
3278
-	if (WARN_ON(!ceph_msg_data_type_valid(type)))
3279
-		return NULL;
3280
-
3281
-	data = kmem_cache_zalloc(ceph_msg_data_cache, GFP_NOFS);
3282
-	if (!data)
3283
-		return NULL;
3284
-
3285
-	data->type = type;
3286
-	INIT_LIST_HEAD(&data->links);
3287
-
3288
-	return data;
3243
+	BUG_ON(msg->num_data_items >= msg->max_data_items);
3244
+	return &msg->data[msg->num_data_items++];
32893245 }
32903246 
32913247 static void ceph_msg_data_destroy(struct ceph_msg_data *data)
32923248 {
3293
-	if (!data)
3294
-		return;
3295
-
3296
-	WARN_ON(!list_empty(&data->links));
3297
-	if (data->type == CEPH_MSG_DATA_PAGELIST)
3249
+	if (data->type == CEPH_MSG_DATA_PAGES && data->own_pages) {
3250
+		int num_pages = calc_pages_for(data->alignment, data->length);
3251
+		ceph_release_page_vector(data->pages, num_pages);
3252
+	} else if (data->type == CEPH_MSG_DATA_PAGELIST) {
32983253 		ceph_pagelist_release(data->pagelist);
3299
-	kmem_cache_free(ceph_msg_data_cache, data);
3254
+	}
33003255 }
33013256 
33023257 void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
3303
-		size_t length, size_t alignment)
3258
+			     size_t length, size_t alignment, bool own_pages)
33043259 {
33053260 	struct ceph_msg_data *data;
33063261 
33073262 	BUG_ON(!pages);
33083263 	BUG_ON(!length);
33093264 
3310
-	data = ceph_msg_data_create(CEPH_MSG_DATA_PAGES);
3311
-	BUG_ON(!data);
3265
+	data = ceph_msg_data_add(msg);
3266
+	data->type = CEPH_MSG_DATA_PAGES;
33123267 	data->pages = pages;
33133268 	data->length = length;
33143269 	data->alignment = alignment & ~PAGE_MASK;
3270
+	data->own_pages = own_pages;
33153271 
3316
-	list_add_tail(&data->links, &msg->data);
33173272 	msg->data_length += length;
33183273 }
33193274 EXPORT_SYMBOL(ceph_msg_data_add_pages);
....@@ -3326,11 +3281,11 @@
33263281 	BUG_ON(!pagelist);
33273282 	BUG_ON(!pagelist->length);
33283283 
3329
-	data = ceph_msg_data_create(CEPH_MSG_DATA_PAGELIST);
3330
-	BUG_ON(!data);
3284
+	data = ceph_msg_data_add(msg);
3285
+	data->type = CEPH_MSG_DATA_PAGELIST;
3286
+	refcount_inc(&pagelist->refcnt);
33313287 	data->pagelist = pagelist;
33323288 
3333
-	list_add_tail(&data->links, &msg->data);
33343289 	msg->data_length += pagelist->length;
33353290 }
33363291 EXPORT_SYMBOL(ceph_msg_data_add_pagelist);
....@@ -3341,12 +3296,11 @@
33413296 {
33423297 	struct ceph_msg_data *data;
33433298 
3344
-	data = ceph_msg_data_create(CEPH_MSG_DATA_BIO);
3345
-	BUG_ON(!data);
3299
+	data = ceph_msg_data_add(msg);
3300
+	data->type = CEPH_MSG_DATA_BIO;
33463301 	data->bio_pos = *bio_pos;
33473302 	data->bio_length = length;
33483303 
3349
-	list_add_tail(&data->links, &msg->data);
33503304 	msg->data_length += length;
33513305 }
33523306 EXPORT_SYMBOL(ceph_msg_data_add_bio);
....@@ -3357,11 +3311,10 @@
33573311 {
33583312 	struct ceph_msg_data *data;
33593313 
3360
-	data = ceph_msg_data_create(CEPH_MSG_DATA_BVECS);
3361
-	BUG_ON(!data);
3314
+	data = ceph_msg_data_add(msg);
3315
+	data->type = CEPH_MSG_DATA_BVECS;
33623316 	data->bvec_pos = *bvec_pos;
33633317 
3364
-	list_add_tail(&data->links, &msg->data);
33653318 	msg->data_length += bvec_pos->iter.bi_size;
33663319 }
33673320 EXPORT_SYMBOL(ceph_msg_data_add_bvecs);
....@@ -3370,8 +3323,8 @@
33703323  * construct a new message with given type, size
33713324  * the new msg has a ref count of 1.
33723325  */
3373
-struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
3374
-			      bool can_fail)
3326
+struct ceph_msg *ceph_msg_new2(int type, int front_len, int max_data_items,
3327
+			       gfp_t flags, bool can_fail)
33753328 {
33763329 	struct ceph_msg *m;
33773330 
....@@ -3385,7 +3338,6 @@
33853338 
33863339 	INIT_LIST_HEAD(&m->list_head);
33873340 	kref_init(&m->kref);
3388
-	INIT_LIST_HEAD(&m->data);
33893341 
33903342 	/* front */
33913343 	if (front_len) {
....@@ -3399,6 +3351,15 @@
33993351 		m->front.iov_base = NULL;
34003352 	}
34013353 	m->front_alloc_len = m->front.iov_len = front_len;
3354
+
3355
+	if (max_data_items) {
3356
+		m->data = kmalloc_array(max_data_items, sizeof(*m->data),
3357
+					flags);
3358
+		if (!m->data)
3359
+			goto out2;
3360
+
3361
+		m->max_data_items = max_data_items;
3362
+	}
34023363 
34033364 	dout("ceph_msg_new %p front %d\n", m, front_len);
34043365 	return m;
....@@ -3415,6 +3376,13 @@
34153376 		     front_len);
34163377 	}
34173378 	return NULL;
3379
+}
3380
+EXPORT_SYMBOL(ceph_msg_new2);
3381
+
3382
+struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
3383
+			      bool can_fail)
3384
+{
3385
+	return ceph_msg_new2(type, front_len, 0, flags, can_fail);
34183386 }
34193387 EXPORT_SYMBOL(ceph_msg_new);
34203388 
....@@ -3511,13 +3479,14 @@
35113479 {
35123480 	dout("%s %p\n", __func__, m);
35133481 	kvfree(m->front.iov_base);
3482
+	kfree(m->data);
35143483 	kmem_cache_free(ceph_msg_cache, m);
35153484 }
35163485 
35173486 static void ceph_msg_release(struct kref *kref)
35183487 {
35193488 	struct ceph_msg *m = container_of(kref, struct ceph_msg, kref);
3520
-	struct ceph_msg_data *data, *next;
3489
+	int i;
35213490 
35223491 	dout("%s %p\n", __func__, m);
35233492 	WARN_ON(!list_empty(&m->list_head));
....@@ -3530,11 +3499,8 @@
35303499 		m->middle = NULL;
35313500 	}
35323501 
3533
-	list_for_each_entry_safe(data, next, &m->data, links) {
3534
-		list_del_init(&data->links);
3535
-		ceph_msg_data_destroy(data);
3536
-	}
3537
-	m->data_length = 0;
3502
+	for (i = 0; i < m->num_data_items; i++)
3503
+		ceph_msg_data_destroy(&m->data[i]);
35383504 
35393505 	if (m->pool)
35403506 		ceph_msgpool_put(m->pool, m);