forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-02-20 ea08eeccae9297f7aabd2ef7f0c2517ac4549acc 
 kernel/net/sunrpc/svcsock.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-only
12 /*
23  * linux/net/sunrpc/svcsock.c
34  *
....@@ -43,8 +44,8 @@
4344 #include <net/tcp.h>
4445 #include <net/tcp_states.h>
4546 #include <linux/uaccess.h>
47
+#include <linux/highmem.h>
4648 #include <asm/ioctls.h>
47
-#include <trace/events/skb.h>
4849 
4950 #include <linux/sunrpc/types.h>
5051 #include <linux/sunrpc/clnt.h>
....@@ -54,6 +55,9 @@
5455 #include <linux/sunrpc/stats.h>
5556 #include <linux/sunrpc/xprt.h>
5657 
58
+#include <trace/events/sunrpc.h>
59
+
60
+#include "socklib.h"
5761 #include "sunrpc.h"
5862 
5963 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
....@@ -70,13 +74,6 @@
7074 static struct svc_xprt *svc_create_socket(struct svc_serv *, int,
7175 					  struct net *, struct sockaddr *,
7276 					  int, int);
73
-#if defined(CONFIG_SUNRPC_BACKCHANNEL)
74
-static struct svc_xprt *svc_bc_create_socket(struct svc_serv *, int,
75
-					     struct net *, struct sockaddr *,
76
-					     int, int);
77
-static void svc_bc_sock_free(struct svc_xprt *xprt);
78
-#endif /* CONFIG_SUNRPC_BACKCHANNEL */
79
-
8077 #ifdef CONFIG_DEBUG_LOCK_ALLOC
8178 static struct lock_class_key svc_key[2];
8279 static struct lock_class_key svc_slock_key[2];
....@@ -113,31 +110,35 @@
113110 }
114111 #endif
115112 
116
-/*
117
- * Release an skbuff after use
113
+/**
114
+ * svc_tcp_release_rqst - Release transport-related resources
115
+ * @rqstp: request structure with resources to be released
116
+ *
118117  */
119
-static void svc_release_skb(struct svc_rqst *rqstp)
118
+static void svc_tcp_release_rqst(struct svc_rqst *rqstp)
120119 {
121120 	struct sk_buff *skb = rqstp->rq_xprt_ctxt;
122121 
123122 	if (skb) {
124123 		struct svc_sock *svsk =
125124 			container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
126
-		rqstp->rq_xprt_ctxt = NULL;
127125 
128
-		dprintk("svc: service %p, releasing skb %p\n", rqstp, skb);
126
+		rqstp->rq_xprt_ctxt = NULL;
129127 		skb_free_datagram_locked(svsk->sk_sk, skb);
130128 	}
131129 }
132130 
133
-static void svc_release_udp_skb(struct svc_rqst *rqstp)
131
+/**
132
+ * svc_udp_release_rqst - Release transport-related resources
133
+ * @rqstp: request structure with resources to be released
134
+ *
135
+ */
136
+static void svc_udp_release_rqst(struct svc_rqst *rqstp)
134137 {
135138 	struct sk_buff *skb = rqstp->rq_xprt_ctxt;
136139 
137140 	if (skb) {
138141 		rqstp->rq_xprt_ctxt = NULL;
139
-
140
-		dprintk("svc: service %p, releasing skb %p\n", rqstp, skb);
141142 		consume_skb(skb);
142143 	}
143144 }
....@@ -180,109 +181,10 @@
180181 	}
181182 }
182183 
183
-/*
184
- * send routine intended to be shared by the fore- and back-channel
185
- */
186
-int svc_send_common(struct socket *sock, struct xdr_buf *xdr,
187
-		    struct page *headpage, unsigned long headoffset,
188
-		    struct page *tailpage, unsigned long tailoffset)
184
+static int svc_sock_read_payload(struct svc_rqst *rqstp, unsigned int offset,
185
+				 unsigned int length)
189186 {
190
-	int		result;
191
-	int		size;
192
-	struct page	**ppage = xdr->pages;
193
-	size_t		base = xdr->page_base;
194
-	unsigned int	pglen = xdr->page_len;
195
-	unsigned int	flags = MSG_MORE | MSG_SENDPAGE_NOTLAST;
196
-	int		slen;
197
-	int		len = 0;
198
-
199
-	slen = xdr->len;
200
-
201
-	/* send head */
202
-	if (slen == xdr->head[0].iov_len)
203
-		flags = 0;
204
-	len = kernel_sendpage(sock, headpage, headoffset,
205
-				  xdr->head[0].iov_len, flags);
206
-	if (len != xdr->head[0].iov_len)
207
-		goto out;
208
-	slen -= xdr->head[0].iov_len;
209
-	if (slen == 0)
210
-		goto out;
211
-
212
-	/* send page data */
213
-	size = PAGE_SIZE - base < pglen ? PAGE_SIZE - base : pglen;
214
-	while (pglen > 0) {
215
-		if (slen == size)
216
-			flags = 0;
217
-		result = kernel_sendpage(sock, *ppage, base, size, flags);
218
-		if (result > 0)
219
-			len += result;
220
-		if (result != size)
221
-			goto out;
222
-		slen -= size;
223
-		pglen -= size;
224
-		size = PAGE_SIZE < pglen ? PAGE_SIZE : pglen;
225
-		base = 0;
226
-		ppage++;
227
-	}
228
-
229
-	/* send tail */
230
-	if (xdr->tail[0].iov_len) {
231
-		result = kernel_sendpage(sock, tailpage, tailoffset,
232
-				   xdr->tail[0].iov_len, 0);
233
-		if (result > 0)
234
-			len += result;
235
-	}
236
-
237
-out:
238
-	return len;
239
-}
240
-
241
-
242
-/*
243
- * Generic sendto routine
244
- */
245
-static int svc_sendto(struct svc_rqst *rqstp, struct xdr_buf *xdr)
246
-{
247
-	struct svc_sock	*svsk =
248
-		container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
249
-	struct socket	*sock = svsk->sk_sock;
250
-	union {
251
-		struct cmsghdr	hdr;
252
-		long		all[SVC_PKTINFO_SPACE / sizeof(long)];
253
-	} buffer;
254
-	struct cmsghdr *cmh = &buffer.hdr;
255
-	int		len = 0;
256
-	unsigned long tailoff;
257
-	unsigned long headoff;
258
-	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
259
-
260
-	if (rqstp->rq_prot == IPPROTO_UDP) {
261
-		struct msghdr msg = {
262
-			.msg_name	= &rqstp->rq_addr,
263
-			.msg_namelen	= rqstp->rq_addrlen,
264
-			.msg_control	= cmh,
265
-			.msg_controllen	= sizeof(buffer),
266
-			.msg_flags	= MSG_MORE,
267
-		};
268
-
269
-		svc_set_cmsg_data(rqstp, cmh);
270
-
271
-		if (sock_sendmsg(sock, &msg) < 0)
272
-			goto out;
273
-	}
274
-
275
-	tailoff = ((unsigned long)xdr->tail[0].iov_base) & (PAGE_SIZE-1);
276
-	headoff = 0;
277
-	len = svc_send_common(sock, xdr, rqstp->rq_respages[0], headoff,
278
-			       rqstp->rq_respages[0], tailoff);
279
-
280
-out:
281
-	dprintk("svc: socket %p sendto([%p %zu... ], %d) = %d (addr %s)\n",
282
-		svsk, xdr->head[0].iov_base, xdr->head[0].iov_len,
283
-		xdr->len, len, svc_print_addr(rqstp, buf, sizeof(buf)));
284
-
285
-	return len;
187
+	return 0;
286188 }
287189 
288190 /*
....@@ -322,60 +224,69 @@
322224 	return len;
323225 }
324226 
227
+#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
228
+static void svc_flush_bvec(const struct bio_vec *bvec, size_t size, size_t seek)
229
+{
230
+	struct bvec_iter bi = {
231
+		.bi_size	= size + seek,
232
+	};
233
+	struct bio_vec bv;
234
+
235
+	bvec_iter_advance(bvec, &bi, seek & PAGE_MASK);
236
+	for_each_bvec(bv, bvec, bi, bi)
237
+		flush_dcache_page(bv.bv_page);
238
+}
239
+#else
240
+static inline void svc_flush_bvec(const struct bio_vec *bvec, size_t size,
241
+				  size_t seek)
242
+{
243
+}
244
+#endif
245
+
325246 /*
326
- * Generic recvfrom routine.
247
+ * Read from @rqstp's transport socket. The incoming message fills whole
248
+ * pages in @rqstp's rq_pages array until the last page of the message
249
+ * has been received into a partial page.
327250  */
328
-static int svc_recvfrom(struct svc_rqst *rqstp, struct kvec *iov, int nr,
329
-			int buflen)
251
+static ssize_t svc_tcp_read_msg(struct svc_rqst *rqstp, size_t buflen,
252
+				size_t seek)
330253 {
331254 	struct svc_sock *svsk =
332255 		container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
333
-	struct msghdr msg = {
334
-		.msg_flags	= MSG_DONTWAIT,
335
-	};
336
-	int len;
256
+	struct bio_vec *bvec = rqstp->rq_bvec;
257
+	struct msghdr msg = { NULL };
258
+	unsigned int i;
259
+	ssize_t len;
260
+	size_t t;
337261 
338262 	rqstp->rq_xprt_hlen = 0;
339263 
340264 	clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
341
-	iov_iter_kvec(&msg.msg_iter, READ | ITER_KVEC, iov, nr, buflen);
342
-	len = sock_recvmsg(svsk->sk_sock, &msg, msg.msg_flags);
265
+
266
+	for (i = 0, t = 0; t < buflen; i++, t += PAGE_SIZE) {
267
+		bvec[i].bv_page = rqstp->rq_pages[i];
268
+		bvec[i].bv_len = PAGE_SIZE;
269
+		bvec[i].bv_offset = 0;
270
+	}
271
+	rqstp->rq_respages = &rqstp->rq_pages[i];
272
+	rqstp->rq_next_page = rqstp->rq_respages + 1;
273
+
274
+	iov_iter_bvec(&msg.msg_iter, READ, bvec, i, buflen);
275
+	if (seek) {
276
+		iov_iter_advance(&msg.msg_iter, seek);
277
+		buflen -= seek;
278
+	}
279
+	len = sock_recvmsg(svsk->sk_sock, &msg, MSG_DONTWAIT);
280
+	if (len > 0)
281
+		svc_flush_bvec(bvec, len, seek);
282
+
343283 	/* If we read a full record, then assume there may be more
344284 	 * data to read (stream based sockets only!)
345285 	 */
346286 	if (len == buflen)
347287 		set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
348288 
349
-	dprintk("svc: socket %p recvfrom(%p, %zu) = %d\n",
350
-		svsk, iov[0].iov_base, iov[0].iov_len, len);
351289 	return len;
352
-}
353
-
354
-static int svc_partial_recvfrom(struct svc_rqst *rqstp,
355
-				struct kvec *iov, int nr,
356
-				int buflen, unsigned int base)
357
-{
358
-	size_t save_iovlen;
359
-	void *save_iovbase;
360
-	unsigned int i;
361
-	int ret;
362
-
363
-	if (base == 0)
364
-		return svc_recvfrom(rqstp, iov, nr, buflen);
365
-
366
-	for (i = 0; i < nr; i++) {
367
-		if (iov[i].iov_len > base)
368
-			break;
369
-		base -= iov[i].iov_len;
370
-	}
371
-	save_iovlen = iov[i].iov_len;
372
-	save_iovbase = iov[i].iov_base;
373
-	iov[i].iov_len -= base;
374
-	iov[i].iov_base += base;
375
-	ret = svc_recvfrom(rqstp, &iov[i], nr - i, buflen);
376
-	iov[i].iov_len = save_iovlen;
377
-	iov[i].iov_base = save_iovbase;
378
-	return ret;
379290 }
380291 
381292 /*
....@@ -411,13 +322,10 @@
411322 	struct svc_sock	*svsk = (struct svc_sock *)sk->sk_user_data;
412323 
413324 	if (svsk) {
414
-		dprintk("svc: socket %p(inet %p), busy=%d\n",
415
-			svsk, sk,
416
-			test_bit(XPT_BUSY, &svsk->sk_xprt.xpt_flags));
417
-
418325 		/* Refer to svc_setup_socket() for details. */
419326 		rmb();
420327 		svsk->sk_odata(sk);
328
+		trace_svcsock_data_ready(&svsk->sk_xprt, 0);
421329 		if (!test_and_set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags))
422330 			svc_xprt_enqueue(&svsk->sk_xprt);
423331 	}
....@@ -431,11 +339,9 @@
431339 	struct svc_sock	*svsk = (struct svc_sock *)(sk->sk_user_data);
432340 
433341 	if (svsk) {
434
-		dprintk("svc: socket %p(inet %p), write_space busy=%d\n",
435
-			svsk, sk, test_bit(XPT_BUSY, &svsk->sk_xprt.xpt_flags));
436
-
437342 		/* Refer to svc_setup_socket() for details. */
438343 		rmb();
344
+		trace_svcsock_write_space(&svsk->sk_xprt, 0);
439345 		svsk->sk_owspace(sk);
440346 		svc_xprt_enqueue(&svsk->sk_xprt);
441347 	}
....@@ -452,17 +358,9 @@
452358 
453359 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
454360 {
455
-	struct svc_sock *svsk;
456
-	struct socket *sock;
457
-	struct linger no_linger = {
458
-		.l_onoff = 1,
459
-		.l_linger = 0,
460
-	};
361
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
461362 
462
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
463
-	sock = svsk->sk_sock;
464
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
465
-			  (char *)&no_linger, sizeof(no_linger));
363
+	sock_no_linger(svsk->sk_sock->sk);
466364 }
467365 
468366 /*
....@@ -520,8 +418,15 @@
520418 	return 0;
521419 }
522420 
523
-/*
524
- * Receive a datagram from a UDP socket.
421
+/**
422
+ * svc_udp_recvfrom - Receive a datagram from a UDP socket.
423
+ * @rqstp: request structure into which to receive an RPC Call
424
+ *
425
+ * Called in a loop when XPT_DATA has been set.
426
+ *
427
+ * Returns:
428
+ *   On success, the number of bytes in a received RPC Call, or
429
+ *   %0 if a complete RPC Call message was not ready to return
525430  */
526431 static int svc_udp_recvfrom(struct svc_rqst *rqstp)
527432 {
....@@ -555,20 +460,14 @@
555460 	    svc_sock_setbufsize(svsk, serv->sv_nrthreads + 3);
556461 
557462 	clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
558
-	skb = NULL;
559463 	err = kernel_recvmsg(svsk->sk_sock, &msg, NULL,
560464 			     0, 0, MSG_PEEK | MSG_DONTWAIT);
561
-	if (err >= 0)
562
-		skb = skb_recv_udp(svsk->sk_sk, 0, 1, &err);
465
+	if (err < 0)
466
+		goto out_recv_err;
467
+	skb = skb_recv_udp(svsk->sk_sk, 0, 1, &err);
468
+	if (!skb)
469
+		goto out_recv_err;
563470 
564
-	if (skb == NULL) {
565
-		if (err != -EAGAIN) {
566
-			/* possibly an icmp error */
567
-			dprintk("svc: recvfrom returned error %d\n", -err);
568
-			set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
569
-		}
570
-		return 0;
571
-	}
572471 	len = svc_addr_len(svc_addr(rqstp));
573472 	rqstp->rq_addrlen = len;
574473 	if (skb->tstamp == 0) {
....@@ -579,26 +478,21 @@
579478 	sock_write_timestamp(svsk->sk_sk, skb->tstamp);
580479 	set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags); /* there may be more data... */
581480 
582
-	len  = skb->len;
481
+	len = skb->len;
583482 	rqstp->rq_arg.len = len;
483
+	trace_svcsock_udp_recv(&svsk->sk_xprt, len);
584484 
585485 	rqstp->rq_prot = IPPROTO_UDP;
586486 
587
-	if (!svc_udp_get_dest_address(rqstp, cmh)) {
588
-		net_warn_ratelimited("svc: received unknown control message %d/%d; dropping RPC reply datagram\n",
589
-				     cmh->cmsg_level, cmh->cmsg_type);
590
-		goto out_free;
591
-	}
487
+	if (!svc_udp_get_dest_address(rqstp, cmh))
488
+		goto out_cmsg_err;
592489 	rqstp->rq_daddrlen = svc_addr_len(svc_daddr(rqstp));
593490 
594491 	if (skb_is_nonlinear(skb)) {
595492 		/* we have to copy */
596493 		local_bh_disable();
597
-		if (csum_partial_copy_to_xdr(&rqstp->rq_arg, skb)) {
598
-			local_bh_enable();
599
-			/* checksum error */
600
-			goto out_free;
601
-		}
494
+		if (csum_partial_copy_to_xdr(&rqstp->rq_arg, skb))
495
+			goto out_bh_enable;
602496 		local_bh_enable();
603497 		consume_skb(skb);
604498 	} else {
....@@ -626,28 +520,79 @@
626520 		serv->sv_stats->netudpcnt++;
627521 
628522 	return len;
523
+
524
+out_recv_err:
525
+	if (err != -EAGAIN) {
526
+		/* possibly an icmp error */
527
+		set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
528
+	}
529
+	trace_svcsock_udp_recv_err(&svsk->sk_xprt, err);
530
+	return 0;
531
+out_cmsg_err:
532
+	net_warn_ratelimited("svc: received unknown control message %d/%d; dropping RPC reply datagram\n",
533
+			     cmh->cmsg_level, cmh->cmsg_type);
534
+	goto out_free;
535
+out_bh_enable:
536
+	local_bh_enable();
629537 out_free:
630538 	kfree_skb(skb);
631539 	return 0;
632540 }
633541 
634
-static int
635
-svc_udp_sendto(struct svc_rqst *rqstp)
542
+/**
543
+ * svc_udp_sendto - Send out a reply on a UDP socket
544
+ * @rqstp: completed svc_rqst
545
+ *
546
+ * xpt_mutex ensures @rqstp's whole message is written to the socket
547
+ * without interruption.
548
+ *
549
+ * Returns the number of bytes sent, or a negative errno.
550
+ */
551
+static int svc_udp_sendto(struct svc_rqst *rqstp)
636552 {
637
-	int		error;
553
+	struct svc_xprt *xprt = rqstp->rq_xprt;
554
+	struct svc_sock	*svsk = container_of(xprt, struct svc_sock, sk_xprt);
555
+	struct xdr_buf *xdr = &rqstp->rq_res;
556
+	union {
557
+		struct cmsghdr	hdr;
558
+		long		all[SVC_PKTINFO_SPACE / sizeof(long)];
559
+	} buffer;
560
+	struct cmsghdr *cmh = &buffer.hdr;
561
+	struct msghdr msg = {
562
+		.msg_name	= &rqstp->rq_addr,
563
+		.msg_namelen	= rqstp->rq_addrlen,
564
+		.msg_control	= cmh,
565
+		.msg_controllen	= sizeof(buffer),
566
+	};
567
+	unsigned int sent;
568
+	int err;
638569 
639
-	svc_release_udp_skb(rqstp);
570
+	svc_udp_release_rqst(rqstp);
640571 
641
-	error = svc_sendto(rqstp, &rqstp->rq_res);
642
-	if (error == -ECONNREFUSED)
572
+	svc_set_cmsg_data(rqstp, cmh);
573
+
574
+	mutex_lock(&xprt->xpt_mutex);
575
+
576
+	if (svc_xprt_is_dead(xprt))
577
+		goto out_notconn;
578
+
579
+	err = xprt_sock_sendmsg(svsk->sk_sock, &msg, xdr, 0, 0, &sent);
580
+	xdr_free_bvec(xdr);
581
+	if (err == -ECONNREFUSED) {
643582 		/* ICMP error on earlier request. */
644
-		error = svc_sendto(rqstp, &rqstp->rq_res);
583
+		err = xprt_sock_sendmsg(svsk->sk_sock, &msg, xdr, 0, 0, &sent);
584
+		xdr_free_bvec(xdr);
585
+	}
586
+	trace_svcsock_udp_send(xprt, err);
645587 
646
-	return error;
647
-}
588
+	mutex_unlock(&xprt->xpt_mutex);
589
+	if (err < 0)
590
+		return err;
591
+	return sent;
648592 
649
-static void svc_udp_prep_reply_hdr(struct svc_rqst *rqstp)
650
-{
593
+out_notconn:
594
+	mutex_unlock(&xprt->xpt_mutex);
595
+	return -ENOTCONN;
651596 }
652597 
653598 static int svc_udp_has_wspace(struct svc_xprt *xprt)
....@@ -690,10 +635,10 @@
690635 	.xpo_create = svc_udp_create,
691636 	.xpo_recvfrom = svc_udp_recvfrom,
692637 	.xpo_sendto = svc_udp_sendto,
693
-	.xpo_release_rqst = svc_release_udp_skb,
638
+	.xpo_read_payload = svc_sock_read_payload,
639
+	.xpo_release_rqst = svc_udp_release_rqst,
694640 	.xpo_detach = svc_sock_detach,
695641 	.xpo_free = svc_sock_free,
696
-	.xpo_prep_reply_hdr = svc_udp_prep_reply_hdr,
697642 	.xpo_has_wspace = svc_udp_has_wspace,
698643 	.xpo_accept = svc_udp_accept,
699644 	.xpo_secure_port = svc_sock_secure_port,
....@@ -710,8 +655,6 @@
710655 
711656 static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
712657 {
713
-	int err, level, optname, one = 1;
714
-
715658 	svc_xprt_init(sock_net(svsk->sk_sock->sk), &svc_udp_class,
716659 		      &svsk->sk_xprt, serv);
717660 	clear_bit(XPT_CACHE_AUTH, &svsk->sk_xprt.xpt_flags);
....@@ -731,19 +674,14 @@
731674 	/* make sure we get destination address info */
732675 	switch (svsk->sk_sk->sk_family) {
733676 	case AF_INET:
734
-		level = SOL_IP;
735
-		optname = IP_PKTINFO;
677
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
736678 		break;
737679 	case AF_INET6:
738
-		level = SOL_IPV6;
739
-		optname = IPV6_RECVPKTINFO;
680
+		ip6_sock_set_recvpktinfo(svsk->sk_sock->sk);
740681 		break;
741682 	default:
742683 		BUG();
743684 	}
744
-	err = kernel_setsockopt(svsk->sk_sock, level, optname,
745
-					(char *)&one, sizeof(one));
746
-	dprintk("svc: kernel_setsockopt returned %d\n", err);
747685 }
748686 
749687 /*
....@@ -754,15 +692,6 @@
754692 {
755693 	struct svc_sock	*svsk = (struct svc_sock *)sk->sk_user_data;
756694 
757
-	dprintk("svc: socket %p TCP (listen) state change %d\n",
758
-		sk, sk->sk_state);
759
-
760
-	if (svsk) {
761
-		/* Refer to svc_setup_socket() for details. */
762
-		rmb();
763
-		svsk->sk_odata(sk);
764
-	}
765
-
766695 	/*
767696 	 * This callback may called twice when a new connection
768697 	 * is established as a child socket inherits everything
....@@ -771,14 +700,18 @@
771700 	 *    when one of child sockets become ESTABLISHED.
772701 	 * 2) data_ready method of the child socket may be called
773702 	 *    when it receives data before the socket is accepted.
774
-	 * In case of 2, we should ignore it silently.
703
+	 * In case of 2, we should ignore it silently and DO NOT
704
+	 * dereference svsk.
775705 	 */
776
-	if (sk->sk_state == TCP_LISTEN) {
777
-		if (svsk) {
778
-			set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
779
-			svc_xprt_enqueue(&svsk->sk_xprt);
780
-		} else
781
-			printk("svc: socket %p: no user data\n", sk);
706
+	if (sk->sk_state != TCP_LISTEN)
707
+		return;
708
+
709
+	if (svsk) {
710
+		/* Refer to svc_setup_socket() for details. */
711
+		rmb();
712
+		svsk->sk_odata(sk);
713
+		set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
714
+		svc_xprt_enqueue(&svsk->sk_xprt);
782715 	}
783716 }
784717 
....@@ -789,15 +722,11 @@
789722 {
790723 	struct svc_sock	*svsk = (struct svc_sock *)sk->sk_user_data;
791724 
792
-	dprintk("svc: socket %p TCP (connected) state change %d (svsk %p)\n",
793
-		sk, sk->sk_state, sk->sk_user_data);
794
-
795
-	if (!svsk)
796
-		printk("svc: socket %p: no user data\n", sk);
797
-	else {
725
+	if (svsk) {
798726 		/* Refer to svc_setup_socket() for details. */
799727 		rmb();
800728 		svsk->sk_ostate(sk);
729
+		trace_svcsock_tcp_state(&svsk->sk_xprt, svsk->sk_sock);
801730 		if (sk->sk_state != TCP_ESTABLISHED) {
802731 			set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
803732 			svc_xprt_enqueue(&svsk->sk_xprt);
....@@ -818,9 +747,7 @@
818747 	struct socket	*newsock;
819748 	struct svc_sock	*newsvsk;
820749 	int		err, slen;
821
-	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
822750 
823
-	dprintk("svc: tcp_accept %p sock %p\n", svsk, sock);
824751 	if (!sock)
825752 		return NULL;
826753 
....@@ -833,29 +760,17 @@
833760 		else if (err != -EAGAIN)
834761 			net_warn_ratelimited("%s: accept failed (err %d)!\n",
835762 					     serv->sv_name, -err);
763
+		trace_svcsock_accept_err(xprt, serv->sv_name, err);
836764 		return NULL;
837765 	}
838766 	set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
839767 
840768 	err = kernel_getpeername(newsock, sin);
841769 	if (err < 0) {
842
-		net_warn_ratelimited("%s: peername failed (err %d)!\n",
843
-				     serv->sv_name, -err);
770
+		trace_svcsock_getpeername_err(xprt, serv->sv_name, err);
844771 		goto failed;		/* aborted connection or whatever */
845772 	}
846773 	slen = err;
847
-
848
-	/* Ideally, we would want to reject connections from unauthorized
849
-	 * hosts here, but when we get encryption, the IP of the host won't
850
-	 * tell us anything.  For now just warn about unpriv connections.
851
-	 */
852
-	if (!svc_port_is_privileged(sin)) {
853
-		dprintk("%s: connect from unprivileged port: %s\n",
854
-			serv->sv_name,
855
-			__svc_print_addr(sin, buf, sizeof(buf)));
856
-	}
857
-	dprintk("%s: connect from %s\n", serv->sv_name,
858
-		__svc_print_addr(sin, buf, sizeof(buf)));
859774 
860775 	/* Reset the inherited callbacks before calling svc_setup_socket */
861776 	newsock->sk->sk_state_change = svsk->sk_ostate;
....@@ -874,10 +789,8 @@
874789 	svc_xprt_set_remote(&newsvsk->sk_xprt, sin, slen);
875790 	err = kernel_getsockname(newsock, sin);
876791 	slen = err;
877
-	if (unlikely(err < 0)) {
878
-		dprintk("svc_tcp_accept: kernel_getsockname error %d\n", -err);
792
+	if (unlikely(err < 0))
879793 		slen = offsetof(struct sockaddr, sa_data);
880
-	}
881794 	svc_xprt_set_local(&newsvsk->sk_xprt, sin, slen);
882795 
883796 	if (sock_is_loopback(newsock->sk))
....@@ -894,13 +807,14 @@
894807 	return NULL;
895808 }
896809 
897
-static unsigned int svc_tcp_restore_pages(struct svc_sock *svsk, struct svc_rqst *rqstp)
810
+static size_t svc_tcp_restore_pages(struct svc_sock *svsk,
811
+				    struct svc_rqst *rqstp)
898812 {
899
-	unsigned int i, len, npages;
813
+	size_t len = svsk->sk_datalen;
814
+	unsigned int i, npages;
900815 
901
-	if (svsk->sk_datalen == 0)
816
+	if (!len)
902817 		return 0;
903
-	len = svsk->sk_datalen;
904818 	npages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
905819 	for (i = 0; i < npages; i++) {
906820 		if (rqstp->rq_pages[i] != NULL)
....@@ -949,46 +863,45 @@
949863 }
950864 
951865 /*
952
- * Receive fragment record header.
953
- * If we haven't gotten the record length yet, get the next four bytes.
866
+ * Receive fragment record header into sk_marker.
954867  */
955
-static int svc_tcp_recv_record(struct svc_sock *svsk, struct svc_rqst *rqstp)
868
+static ssize_t svc_tcp_read_marker(struct svc_sock *svsk,
869
+				   struct svc_rqst *rqstp)
956870 {
957
-	struct svc_serv	*serv = svsk->sk_xprt.xpt_server;
958
-	unsigned int want;
959
-	int len;
871
+	ssize_t want, len;
960872 
873
+	/* If we haven't gotten the record length yet,
874
+	 * get the next four bytes.
875
+	 */
961876 	if (svsk->sk_tcplen < sizeof(rpc_fraghdr)) {
877
+		struct msghdr	msg = { NULL };
962878 		struct kvec	iov;
963879 
964880 		want = sizeof(rpc_fraghdr) - svsk->sk_tcplen;
965
-		iov.iov_base = ((char *) &svsk->sk_reclen) + svsk->sk_tcplen;
881
+		iov.iov_base = ((char *)&svsk->sk_marker) + svsk->sk_tcplen;
966882 		iov.iov_len  = want;
967
-		if ((len = svc_recvfrom(rqstp, &iov, 1, want)) < 0)
968
-			goto error;
883
+		iov_iter_kvec(&msg.msg_iter, READ, &iov, 1, want);
884
+		len = sock_recvmsg(svsk->sk_sock, &msg, MSG_DONTWAIT);
885
+		if (len < 0)
886
+			return len;
969887 		svsk->sk_tcplen += len;
970
-
971888 		if (len < want) {
972
-			dprintk("svc: short recvfrom while reading record "
973
-				"length (%d of %d)\n", len, want);
974
-			return -EAGAIN;
889
+			/* call again to read the remaining bytes */
890
+			goto err_short;
975891 		}
976
-
977
-		dprintk("svc: TCP record, %d bytes\n", svc_sock_reclen(svsk));
892
+		trace_svcsock_marker(&svsk->sk_xprt, svsk->sk_marker);
978893 		if (svc_sock_reclen(svsk) + svsk->sk_datalen >
979
-							serv->sv_max_mesg) {
980
-			net_notice_ratelimited("RPC: fragment too large: %d\n",
981
-					svc_sock_reclen(svsk));
982
-			goto err_delete;
983
-		}
894
+		    svsk->sk_xprt.xpt_server->sv_max_mesg)
895
+			goto err_too_large;
984896 	}
985
-
986897 	return svc_sock_reclen(svsk);
987
-error:
988
-	dprintk("RPC: TCP recv_record got %d\n", len);
989
-	return len;
990
-err_delete:
898
+
899
+err_too_large:
900
+	net_notice_ratelimited("svc: %s %s RPC fragment too large: %d\n",
901
+			       __func__, svsk->sk_xprt.xpt_server->sv_name,
902
+			       svc_sock_reclen(svsk));
991903 	set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
904
+err_short:
992905 	return -EAGAIN;
993906 }
994907 
....@@ -1006,7 +919,7 @@
1006919 
1007920 	if (!bc_xprt)
1008921 		return -EAGAIN;
1009
-	spin_lock(&bc_xprt->recv_lock);
922
+	spin_lock(&bc_xprt->queue_lock);
1010923 	req = xprt_lookup_rqst(bc_xprt, xid);
1011924 	if (!req)
1012925 		goto unlock_notfound;
....@@ -1024,7 +937,7 @@
1024937 	memcpy(dst->iov_base, src->iov_base, src->iov_len);
1025938 	xprt_complete_rqst(req->rq_task, rqstp->rq_arg.len);
1026939 	rqstp->rq_arg.len = 0;
1027
-	spin_unlock(&bc_xprt->recv_lock);
940
+	spin_unlock(&bc_xprt->queue_lock);
1028941 	return 0;
1029942 unlock_notfound:
1030943 	printk(KERN_NOTICE
....@@ -1033,92 +946,62 @@
1033946 		__func__, ntohl(calldir),
1034947 		bc_xprt, ntohl(xid));
1035948 unlock_eagain:
1036
-	spin_unlock(&bc_xprt->recv_lock);
949
+	spin_unlock(&bc_xprt->queue_lock);
1037950 	return -EAGAIN;
1038
-}
1039
-
1040
-static int copy_pages_to_kvecs(struct kvec *vec, struct page **pages, int len)
1041
-{
1042
-	int i = 0;
1043
-	int t = 0;
1044
-
1045
-	while (t < len) {
1046
-		vec[i].iov_base = page_address(pages[i]);
1047
-		vec[i].iov_len = PAGE_SIZE;
1048
-		i++;
1049
-		t += PAGE_SIZE;
1050
-	}
1051
-	return i;
1052951 }
1053952 
1054953 static void svc_tcp_fragment_received(struct svc_sock *svsk)
1055954 {
1056955 	/* If we have more data, signal svc_xprt_enqueue() to try again */
1057
-	dprintk("svc: TCP %s record (%d bytes)\n",
1058
-		svc_sock_final_rec(svsk) ? "final" : "nonfinal",
1059
-		svc_sock_reclen(svsk));
1060956 	svsk->sk_tcplen = 0;
1061
-	svsk->sk_reclen = 0;
957
+	svsk->sk_marker = xdr_zero;
1062958 }
1063959 
1064
-/*
1065
- * Receive data from a TCP socket.
960
+/**
961
+ * svc_tcp_recvfrom - Receive data from a TCP socket
962
+ * @rqstp: request structure into which to receive an RPC Call
963
+ *
964
+ * Called in a loop when XPT_DATA has been set.
965
+ *
966
+ * Read the 4-byte stream record marker, then use the record length
967
+ * in that marker to set up exactly the resources needed to receive
968
+ * the next RPC message into @rqstp.
969
+ *
970
+ * Returns:
971
+ *   On success, the number of bytes in a received RPC Call, or
972
+ *   %0 if a complete RPC Call message was not ready to return
973
+ *
974
+ * The zero return case handles partial receives and callback Replies.
975
+ * The state of a partial receive is preserved in the svc_sock for
976
+ * the next call to svc_tcp_recvfrom.
1066977  */
1067978 static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
1068979 {
1069980 	struct svc_sock	*svsk =
1070981 		container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
1071982 	struct svc_serv	*serv = svsk->sk_xprt.xpt_server;
1072
-	int		len;
1073
-	struct kvec *vec;
1074
-	unsigned int want, base;
983
+	size_t want, base;
984
+	ssize_t len;
1075985 	__be32 *p;
1076986 	__be32 calldir;
1077
-	int pnum;
1078987 
1079
-	dprintk("svc: tcp_recv %p data %d conn %d close %d\n",
1080
-		svsk, test_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags),
1081
-		test_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags),
1082
-		test_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags));
1083
-
1084
-	len = svc_tcp_recv_record(svsk, rqstp);
988
+	clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
989
+	len = svc_tcp_read_marker(svsk, rqstp);
1085990 	if (len < 0)
1086991 		goto error;
1087992 
1088993 	base = svc_tcp_restore_pages(svsk, rqstp);
1089
-	want = svc_sock_reclen(svsk) - (svsk->sk_tcplen - sizeof(rpc_fraghdr));
1090
-
1091
-	vec = rqstp->rq_vec;
1092
-
1093
-	pnum = copy_pages_to_kvecs(&vec[0], &rqstp->rq_pages[0],
1094
-						svsk->sk_datalen + want);
1095
-
1096
-	rqstp->rq_respages = &rqstp->rq_pages[pnum];
1097
-	rqstp->rq_next_page = rqstp->rq_respages + 1;
1098
-
1099
-	/* Now receive data */
1100
-	len = svc_partial_recvfrom(rqstp, vec, pnum, want, base);
994
+	want = len - (svsk->sk_tcplen - sizeof(rpc_fraghdr));
995
+	len = svc_tcp_read_msg(rqstp, base + want, base);
1101996 	if (len >= 0) {
997
+		trace_svcsock_tcp_recv(&svsk->sk_xprt, len);
1102998 		svsk->sk_tcplen += len;
1103999 		svsk->sk_datalen += len;
11041000 	}
1105
-	if (len != want || !svc_sock_final_rec(svsk)) {
1106
-		svc_tcp_save_pages(svsk, rqstp);
1107
-		if (len < 0 && len != -EAGAIN)
1108
-			goto err_delete;
1109
-		if (len == want)
1110
-			svc_tcp_fragment_received(svsk);
1111
-		else
1112
-			dprintk("svc: incomplete TCP record (%d of %d)\n",
1113
-				(int)(svsk->sk_tcplen - sizeof(rpc_fraghdr)),
1114
-				svc_sock_reclen(svsk));
1115
-		goto err_noclose;
1116
-	}
1117
-
1118
-	if (svsk->sk_datalen < 8) {
1119
-		svsk->sk_datalen = 0;
1120
-		goto err_delete; /* client is nuts. */
1121
-	}
1001
+	if (len != want || !svc_sock_final_rec(svsk))
1002
+		goto err_incomplete;
1003
+	if (svsk->sk_datalen < 8)
1004
+		goto err_nuts;
11221005 
11231006 	rqstp->rq_arg.len = svsk->sk_datalen;
11241007 	rqstp->rq_arg.page_base = 0;
....@@ -1153,61 +1036,165 @@
11531036 
11541037 	return rqstp->rq_arg.len;
11551038 
1039
+err_incomplete:
1040
+	svc_tcp_save_pages(svsk, rqstp);
1041
+	if (len < 0 && len != -EAGAIN)
1042
+		goto err_delete;
1043
+	if (len == want)
1044
+		svc_tcp_fragment_received(svsk);
1045
+	else
1046
+		trace_svcsock_tcp_recv_short(&svsk->sk_xprt,
1047
+				svc_sock_reclen(svsk),
1048
+				svsk->sk_tcplen - sizeof(rpc_fraghdr));
1049
+	goto err_noclose;
11561050 error:
11571051 	if (len != -EAGAIN)
11581052 		goto err_delete;
1159
-	dprintk("RPC: TCP recvfrom got EAGAIN\n");
1053
+	trace_svcsock_tcp_recv_eagain(&svsk->sk_xprt, 0);
11601054 	return 0;
1055
+err_nuts:
1056
+	svsk->sk_datalen = 0;
11611057 err_delete:
1162
-	printk(KERN_NOTICE "%s: recvfrom returned errno %d\n",
1163
-	       svsk->sk_xprt.xpt_server->sv_name, -len);
1058
+	trace_svcsock_tcp_recv_err(&svsk->sk_xprt, len);
11641059 	set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
11651060 err_noclose:
11661061 	return 0;	/* record not complete */
11671062 }
11681063 
1169
-/*
1170
- * Send out data on TCP socket.
1171
- */
1172
-static int svc_tcp_sendto(struct svc_rqst *rqstp)
1064
+static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
1065
+			      int flags)
11731066 {
1174
-	struct xdr_buf	*xbufp = &rqstp->rq_res;
1175
-	int sent;
1176
-	__be32 reclen;
1177
-
1178
-	svc_release_skb(rqstp);
1179
-
1180
-	/* Set up the first element of the reply kvec.
1181
-	 * Any other kvecs that may be in use have been taken
1182
-	 * care of by the server implementation itself.
1183
-	 */
1184
-	reclen = htonl(0x80000000|((xbufp->len ) - 4));
1185
-	memcpy(xbufp->head[0].iov_base, &reclen, 4);
1186
-
1187
-	sent = svc_sendto(rqstp, &rqstp->rq_res);
1188
-	if (sent != xbufp->len) {
1189
-		printk(KERN_NOTICE
1190
-		       "rpc-srv/tcp: %s: %s %d when sending %d bytes "
1191
-		       "- shutting down socket\n",
1192
-		       rqstp->rq_xprt->xpt_server->sv_name,
1193
-		       (sent<0)?"got error":"sent only",
1194
-		       sent, xbufp->len);
1195
-		set_bit(XPT_CLOSE, &rqstp->rq_xprt->xpt_flags);
1196
-		svc_xprt_enqueue(rqstp->rq_xprt);
1197
-		sent = -EAGAIN;
1198
-	}
1199
-	return sent;
1067
+	return kernel_sendpage(sock, virt_to_page(vec->iov_base),
1068
+			       offset_in_page(vec->iov_base),
1069
+			       vec->iov_len, flags);
12001070 }
12011071 
12021072 /*
1203
- * Setup response header. TCP has a 4B record length field.
1073
+ * kernel_sendpage() is used exclusively to reduce the number of
1074
+ * copy operations in this path. Therefore the caller must ensure
1075
+ * that the pages backing @xdr are unchanging.
1076
+ *
1077
+ * In addition, the logic assumes that * .bv_len is never larger
1078
+ * than PAGE_SIZE.
12041079  */
1205
-void svc_tcp_prep_reply_hdr(struct svc_rqst *rqstp)
1080
+static int svc_tcp_sendmsg(struct socket *sock, struct msghdr *msg,
1081
+			   struct xdr_buf *xdr, rpc_fraghdr marker,
1082
+			   unsigned int *sentp)
12061083 {
1207
-	struct kvec *resv = &rqstp->rq_res.head[0];
1084
+	const struct kvec *head = xdr->head;
1085
+	const struct kvec *tail = xdr->tail;
1086
+	struct kvec rm = {
1087
+		.iov_base	= &marker,
1088
+		.iov_len	= sizeof(marker),
1089
+	};
1090
+	int flags, ret;
12081091 
1209
-	/* tcp needs a space for the record length... */
1210
-	svc_putnl(resv, 0);
1092
+	*sentp = 0;
1093
+	ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
1094
+	if (ret < 0)
1095
+		return ret;
1096
+
1097
+	msg->msg_flags = MSG_MORE;
1098
+	ret = kernel_sendmsg(sock, msg, &rm, 1, rm.iov_len);
1099
+	if (ret < 0)
1100
+		return ret;
1101
+	*sentp += ret;
1102
+	if (ret != rm.iov_len)
1103
+		return -EAGAIN;
1104
+
1105
+	flags = head->iov_len < xdr->len ? MSG_MORE | MSG_SENDPAGE_NOTLAST : 0;
1106
+	ret = svc_tcp_send_kvec(sock, head, flags);
1107
+	if (ret < 0)
1108
+		return ret;
1109
+	*sentp += ret;
1110
+	if (ret != head->iov_len)
1111
+		goto out;
1112
+
1113
+	if (xdr->page_len) {
1114
+		unsigned int offset, len, remaining;
1115
+		struct bio_vec *bvec;
1116
+
1117
+		bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
1118
+		offset = offset_in_page(xdr->page_base);
1119
+		remaining = xdr->page_len;
1120
+		flags = MSG_MORE | MSG_SENDPAGE_NOTLAST;
1121
+		while (remaining > 0) {
1122
+			if (remaining <= PAGE_SIZE && tail->iov_len == 0)
1123
+				flags = 0;
1124
+
1125
+			len = min(remaining, bvec->bv_len - offset);
1126
+			ret = kernel_sendpage(sock, bvec->bv_page,
1127
+					      bvec->bv_offset + offset,
1128
+					      len, flags);
1129
+			if (ret < 0)
1130
+				return ret;
1131
+			*sentp += ret;
1132
+			if (ret != len)
1133
+				goto out;
1134
+			remaining -= len;
1135
+			offset = 0;
1136
+			bvec++;
1137
+		}
1138
+	}
1139
+
1140
+	if (tail->iov_len) {
1141
+		ret = svc_tcp_send_kvec(sock, tail, 0);
1142
+		if (ret < 0)
1143
+			return ret;
1144
+		*sentp += ret;
1145
+	}
1146
+
1147
+out:
1148
+	return 0;
1149
+}
1150
+
1151
+/**
1152
+ * svc_tcp_sendto - Send out a reply on a TCP socket
1153
+ * @rqstp: completed svc_rqst
1154
+ *
1155
+ * xpt_mutex ensures @rqstp's whole message is written to the socket
1156
+ * without interruption.
1157
+ *
1158
+ * Returns the number of bytes sent, or a negative errno.
1159
+ */
1160
+static int svc_tcp_sendto(struct svc_rqst *rqstp)
1161
+{
1162
+	struct svc_xprt *xprt = rqstp->rq_xprt;
1163
+	struct svc_sock	*svsk = container_of(xprt, struct svc_sock, sk_xprt);
1164
+	struct xdr_buf *xdr = &rqstp->rq_res;
1165
+	rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT |
1166
+					 (u32)xdr->len);
1167
+	struct msghdr msg = {
1168
+		.msg_flags	= 0,
1169
+	};
1170
+	unsigned int sent;
1171
+	int err;
1172
+
1173
+	svc_tcp_release_rqst(rqstp);
1174
+
1175
+	mutex_lock(&xprt->xpt_mutex);
1176
+	if (svc_xprt_is_dead(xprt))
1177
+		goto out_notconn;
1178
+	err = svc_tcp_sendmsg(svsk->sk_sock, &msg, xdr, marker, &sent);
1179
+	xdr_free_bvec(xdr);
1180
+	trace_svcsock_tcp_send(xprt, err < 0 ? (long)err : sent);
1181
+	if (err < 0 || sent != (xdr->len + sizeof(marker)))
1182
+		goto out_close;
1183
+	mutex_unlock(&xprt->xpt_mutex);
1184
+	return sent;
1185
+
1186
+out_notconn:
1187
+	mutex_unlock(&xprt->xpt_mutex);
1188
+	return -ENOTCONN;
1189
+out_close:
1190
+	pr_notice("rpc-srv/tcp: %s: %s %d when sending %d bytes - shutting down socket\n",
1191
+		  xprt->xpt_server->sv_name,
1192
+		  (err < 0) ? "got error" : "sent",
1193
+		  (err < 0) ? err : sent, xdr->len);
1194
+	set_bit(XPT_CLOSE, &xprt->xpt_flags);
1195
+	svc_xprt_enqueue(xprt);
1196
+	mutex_unlock(&xprt->xpt_mutex);
1197
+	return -EAGAIN;
12111198 }
12121199 
12131200 static struct svc_xprt *svc_tcp_create(struct svc_serv *serv,
....@@ -1218,66 +1205,14 @@
12181205 	return svc_create_socket(serv, IPPROTO_TCP, net, sa, salen, flags);
12191206 }
12201207 
1221
-#if defined(CONFIG_SUNRPC_BACKCHANNEL)
1222
-static struct svc_xprt *svc_bc_create_socket(struct svc_serv *, int,
1223
-					     struct net *, struct sockaddr *,
1224
-					     int, int);
1225
-static void svc_bc_sock_free(struct svc_xprt *xprt);
1226
-
1227
-static struct svc_xprt *svc_bc_tcp_create(struct svc_serv *serv,
1228
-				       struct net *net,
1229
-				       struct sockaddr *sa, int salen,
1230
-				       int flags)
1231
-{
1232
-	return svc_bc_create_socket(serv, IPPROTO_TCP, net, sa, salen, flags);
1233
-}
1234
-
1235
-static void svc_bc_tcp_sock_detach(struct svc_xprt *xprt)
1236
-{
1237
-}
1238
-
1239
-static const struct svc_xprt_ops svc_tcp_bc_ops = {
1240
-	.xpo_create = svc_bc_tcp_create,
1241
-	.xpo_detach = svc_bc_tcp_sock_detach,
1242
-	.xpo_free = svc_bc_sock_free,
1243
-	.xpo_prep_reply_hdr = svc_tcp_prep_reply_hdr,
1244
-	.xpo_secure_port = svc_sock_secure_port,
1245
-};
1246
-
1247
-static struct svc_xprt_class svc_tcp_bc_class = {
1248
-	.xcl_name = "tcp-bc",
1249
-	.xcl_owner = THIS_MODULE,
1250
-	.xcl_ops = &svc_tcp_bc_ops,
1251
-	.xcl_max_payload = RPCSVC_MAXPAYLOAD_TCP,
1252
-};
1253
-
1254
-static void svc_init_bc_xprt_sock(void)
1255
-{
1256
-	svc_reg_xprt_class(&svc_tcp_bc_class);
1257
-}
1258
-
1259
-static void svc_cleanup_bc_xprt_sock(void)
1260
-{
1261
-	svc_unreg_xprt_class(&svc_tcp_bc_class);
1262
-}
1263
-#else /* CONFIG_SUNRPC_BACKCHANNEL */
1264
-static void svc_init_bc_xprt_sock(void)
1265
-{
1266
-}
1267
-
1268
-static void svc_cleanup_bc_xprt_sock(void)
1269
-{
1270
-}
1271
-#endif /* CONFIG_SUNRPC_BACKCHANNEL */
1272
-
12731208 static const struct svc_xprt_ops svc_tcp_ops = {
12741209 	.xpo_create = svc_tcp_create,
12751210 	.xpo_recvfrom = svc_tcp_recvfrom,
12761211 	.xpo_sendto = svc_tcp_sendto,
1277
-	.xpo_release_rqst = svc_release_skb,
1212
+	.xpo_read_payload = svc_sock_read_payload,
1213
+	.xpo_release_rqst = svc_tcp_release_rqst,
12781214 	.xpo_detach = svc_tcp_sock_detach,
12791215 	.xpo_free = svc_sock_free,
1280
-	.xpo_prep_reply_hdr = svc_tcp_prep_reply_hdr,
12811216 	.xpo_has_wspace = svc_tcp_has_wspace,
12821217 	.xpo_accept = svc_tcp_accept,
12831218 	.xpo_secure_port = svc_sock_secure_port,
....@@ -1296,14 +1231,12 @@
12961231 {
12971232 	svc_reg_xprt_class(&svc_tcp_class);
12981233 	svc_reg_xprt_class(&svc_udp_class);
1299
-	svc_init_bc_xprt_sock();
13001234 }
13011235 
13021236 void svc_cleanup_xprt_sock(void)
13031237 {
13041238 	svc_unreg_xprt_class(&svc_tcp_class);
13051239 	svc_unreg_xprt_class(&svc_udp_class);
1306
-	svc_cleanup_bc_xprt_sock();
13071240 }
13081241 
13091242 static void svc_tcp_init(struct svc_sock *svsk, struct svc_serv *serv)
....@@ -1315,18 +1248,16 @@
13151248 	set_bit(XPT_CACHE_AUTH, &svsk->sk_xprt.xpt_flags);
13161249 	set_bit(XPT_CONG_CTRL, &svsk->sk_xprt.xpt_flags);
13171250 	if (sk->sk_state == TCP_LISTEN) {
1318
-		dprintk("setting up TCP socket for listening\n");
13191251 		strcpy(svsk->sk_xprt.xpt_remotebuf, "listener");
13201252 		set_bit(XPT_LISTENER, &svsk->sk_xprt.xpt_flags);
13211253 		sk->sk_data_ready = svc_tcp_listen_data_ready;
13221254 		set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
13231255 	} else {
1324
-		dprintk("setting up TCP socket for reading\n");
13251256 		sk->sk_state_change = svc_tcp_state_change;
13261257 		sk->sk_data_ready = svc_data_ready;
13271258 		sk->sk_write_space = svc_write_space;
13281259 
1329
-		svsk->sk_reclen = 0;
1260
+		svsk->sk_marker = xdr_zero;
13301261 		svsk->sk_tcplen = 0;
13311262 		svsk->sk_datalen = 0;
13321263 		memset(&svsk->sk_pages[0], 0, sizeof(svsk->sk_pages));
....@@ -1371,7 +1302,6 @@
13711302 	int		pmap_register = !(flags & SVC_SOCK_ANONYMOUS);
13721303 	int		err = 0;
13731304 
1374
-	dprintk("svc: svc_setup_socket %p\n", sock);
13751305 	svsk = kzalloc(sizeof(*svsk), GFP_KERNEL);
13761306 	if (!svsk)
13771307 		return ERR_PTR(-ENOMEM);
....@@ -1408,12 +1338,7 @@
14081338 	else
14091339 		svc_tcp_init(svsk, serv);
14101340 
1411
-	dprintk("svc: svc_setup_socket created %p (inet %p), "
1412
-			"listen %d close %d\n",
1413
-			svsk, svsk->sk_sk,
1414
-			test_bit(XPT_LISTENER, &svsk->sk_xprt.xpt_flags),
1415
-			test_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags));
1416
-
1341
+	trace_svcsock_new_socket(sock);
14171342 	return svsk;
14181343 }
14191344 
....@@ -1439,13 +1364,14 @@
14391364  * @fd: file descriptor of the new listener
14401365  * @name_return: pointer to buffer to fill in with name of listener
14411366  * @len: size of the buffer
1367
+ * @cred: credential
14421368  *
14431369  * Fills in socket name and returns positive length of name if successful.
14441370  * Name is terminated with '\n'.  On error, returns a negative errno
14451371  * value.
14461372  */
14471373 int svc_addsock(struct svc_serv *serv, const int fd, char *name_return,
1448
-		const size_t len)
1374
+		const size_t len, const struct cred *cred)
14491375 {
14501376 	int err = 0;
14511377 	struct socket *so = sockfd_lookup(fd, &err);
....@@ -1478,6 +1404,7 @@
14781404 	salen = kernel_getsockname(svsk->sk_sock, sin);
14791405 	if (salen >= 0)
14801406 		svc_xprt_set_local(&svsk->sk_xprt, sin, salen);
1407
+	svsk->sk_xprt.xpt_cred = get_cred(cred);
14811408 	svc_add_new_perm_xprt(serv, &svsk->sk_xprt);
14821409 	return svc_one_sock_name(svsk, name_return, len);
14831410 out:
....@@ -1503,12 +1430,6 @@
15031430 	struct sockaddr *newsin = (struct sockaddr *)&addr;
15041431 	int		newlen;
15051432 	int		family;
1506
-	int		val;
1507
-	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
1508
-
1509
-	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
1510
-			serv->sv_program->pg_name, protocol,
1511
-			__svc_print_addr(sin, buf, sizeof(buf)));
15121433 
15131434 	if (protocol != IPPROTO_UDP && protocol != IPPROTO_TCP) {
15141435 		printk(KERN_WARNING "svc: only UDP and TCP "
....@@ -1539,11 +1460,8 @@
15391460 	 * getting requests from IPv4 remotes.  Those should
15401461 	 * be shunted to a PF_INET listener via rpcbind.
15411462 	 */
1542
-	val = 1;
15431463 	if (family == PF_INET6)
1544
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
1545
-					(char *)&val, sizeof(val));
1546
-
1464
+		ip6_sock_set_v6only(sock->sk);
15471465 	if (type == SOCK_STREAM)
15481466 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
15491467 	error = kernel_bind(sock, sin, len);
....@@ -1568,7 +1486,6 @@
15681486 	svc_xprt_set_local(&svsk->sk_xprt, newsin, newlen);
15691487 	return (struct svc_xprt *)svsk;
15701488 bummer:
1571
-	dprintk("svc: svc_create_socket error = %d\n", -error);
15721489 	sock_release(sock);
15731490 	return ERR_PTR(error);
15741491 }
....@@ -1581,8 +1498,6 @@
15811498 {
15821499 	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
15831500 	struct sock *sk = svsk->sk_sk;
1584
-
1585
-	dprintk("svc: svc_sock_detach(%p)\n", svsk);
15861501 
15871502 	/* put back the old socket callbacks */
15881503 	lock_sock(sk);
....@@ -1600,8 +1515,6 @@
16001515 {
16011516 	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
16021517 
1603
-	dprintk("svc: svc_tcp_sock_detach(%p)\n", svsk);
1604
-
16051518 	svc_sock_detach(xprt);
16061519 
16071520 	if (!test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
....@@ -1616,7 +1529,6 @@
16161529 static void svc_sock_free(struct svc_xprt *xprt)
16171530 {
16181531 	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
1619
-	dprintk("svc: svc_sock_free(%p)\n", svsk);
16201532 
16211533 	if (svsk->sk_sock->file)
16221534 		sockfd_put(svsk->sk_sock);
....@@ -1624,45 +1536,3 @@
16241536 		sock_release(svsk->sk_sock);
16251537 	kfree(svsk);
16261538 }
1627
-
1628
-#if defined(CONFIG_SUNRPC_BACKCHANNEL)
1629
-/*
1630
- * Create a back channel svc_xprt which shares the fore channel socket.
1631
- */
1632
-static struct svc_xprt *svc_bc_create_socket(struct svc_serv *serv,
1633
-					     int protocol,
1634
-					     struct net *net,
1635
-					     struct sockaddr *sin, int len,
1636
-					     int flags)
1637
-{
1638
-	struct svc_sock *svsk;
1639
-	struct svc_xprt *xprt;
1640
-
1641
-	if (protocol != IPPROTO_TCP) {
1642
-		printk(KERN_WARNING "svc: only TCP sockets"
1643
-			" supported on shared back channel\n");
1644
-		return ERR_PTR(-EINVAL);
1645
-	}
1646
-
1647
-	svsk = kzalloc(sizeof(*svsk), GFP_KERNEL);
1648
-	if (!svsk)
1649
-		return ERR_PTR(-ENOMEM);
1650
-
1651
-	xprt = &svsk->sk_xprt;
1652
-	svc_xprt_init(net, &svc_tcp_bc_class, xprt, serv);
1653
-	set_bit(XPT_CONG_CTRL, &svsk->sk_xprt.xpt_flags);
1654
-
1655
-	serv->sv_bc_xprt = xprt;
1656
-
1657
-	return xprt;
1658
-}
1659
-
1660
-/*
1661
- * Free a back channel svc_sock.
1662
- */
1663
-static void svc_bc_sock_free(struct svc_xprt *xprt)
1664
-{
1665
-	if (xprt)
1666
-		kfree(container_of(xprt, struct svc_sock, sk_xprt));
1667
-}
1668
-#endif /* CONFIG_SUNRPC_BACKCHANNEL */