~hc/RK356X_SDK_RELEASE.git

..	..	@@ -1,3 +1,4 @@
	1	+// SPDX-License-Identifier: GPL-2.0-only
1	2	/*
2	3	* linux/net/sunrpc/svcsock.c
3	4	*
..	..	@@ -43,8 +44,8 @@
43	44	#include <net/tcp.h>
44	45	#include <net/tcp_states.h>
45	46	#include <linux/uaccess.h>
	47	+#include <linux/highmem.h>
46	48	#include <asm/ioctls.h>
47		-#include <trace/events/skb.h>
48	49
49	50	#include <linux/sunrpc/types.h>
50	51	#include <linux/sunrpc/clnt.h>
..	..	@@ -54,6 +55,9 @@
54	55	#include <linux/sunrpc/stats.h>
55	56	#include <linux/sunrpc/xprt.h>
56	57
	58	+#include <trace/events/sunrpc.h>
	59	+
	60	+#include "socklib.h"
57	61	#include "sunrpc.h"
58	62
59	63	#define RPCDBG_FACILITY RPCDBG_SVCXPRT
..	..	@@ -70,13 +74,6 @@
70	74	static struct svc_xprt svc_create_socket(struct svc_serv , int,
71	75	struct net , struct sockaddr ,
72	76	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		-
80	77	#ifdef CONFIG_DEBUG_LOCK_ALLOC
81	78	static struct lock_class_key svc_key[2];
82	79	static struct lock_class_key svc_slock_key[2];
..	..	@@ -113,31 +110,35 @@
113	110	}
114	111	#endif
115	112
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	+ *
118	117	*/
119		-static void svc_release_skb(struct svc_rqst *rqstp)
	118	+static void svc_tcp_release_rqst(struct svc_rqst *rqstp)
120	119	{
121	120	struct sk_buff *skb = rqstp->rq_xprt_ctxt;
122	121
123	122	if (skb) {
124	123	struct svc_sock *svsk =
125	124	container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
126		- rqstp->rq_xprt_ctxt = NULL;
127	125
128		- dprintk("svc: service %p, releasing skb %p\n", rqstp, skb);
	126	+ rqstp->rq_xprt_ctxt = NULL;
129	127	skb_free_datagram_locked(svsk->sk_sk, skb);
130	128	}
131	129	}
132	130
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)
134	137	{
135	138	struct sk_buff *skb = rqstp->rq_xprt_ctxt;
136	139
137	140	if (skb) {
138	141	rqstp->rq_xprt_ctxt = NULL;
139		-
140		- dprintk("svc: service %p, releasing skb %p\n", rqstp, skb);
141	142	consume_skb(skb);
142	143	}
143	144	}
..	..	@@ -180,109 +181,10 @@
180	181	}
181	182	}
182	183
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)
189	186	{
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;
286	188	}
287	189
288	190	/*
..	..	@@ -322,60 +224,69 @@
322	224	return len;
323	225	}
324	226
	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	+
325	246	/*
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.
327	250	*/
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)
330	253	{
331	254	struct svc_sock *svsk =
332	255	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;
337	261
338	262	rqstp->rq_xprt_hlen = 0;
339	263
340	264	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	+
343	283	/* If we read a full record, then assume there may be more
344	284	* data to read (stream based sockets only!)
345	285	*/
346	286	if (len == buflen)
347	287	set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
348	288
349		- dprintk("svc: socket %p recvfrom(%p, %zu) = %d\n",
350		- svsk, iov[0].iov_base, iov[0].iov_len, len);
351	289	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;
379	290	}
380	291
381	292	/*
..	..	@@ -411,13 +322,10 @@
411	322	struct svc_sock svsk = (struct svc_sock )sk->sk_user_data;
412	323
413	324	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		-
418	325	/* Refer to svc_setup_socket() for details. */
419	326	rmb();
420	327	svsk->sk_odata(sk);
	328	+ trace_svcsock_data_ready(&svsk->sk_xprt, 0);
421	329	if (!test_and_set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags))
422	330	svc_xprt_enqueue(&svsk->sk_xprt);
423	331	}
..	..	@@ -431,11 +339,9 @@
431	339	struct svc_sock svsk = (struct svc_sock )(sk->sk_user_data);
432	340
433	341	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		-
437	342	/* Refer to svc_setup_socket() for details. */
438	343	rmb();
	344	+ trace_svcsock_write_space(&svsk->sk_xprt, 0);
439	345	svsk->sk_owspace(sk);
440	346	svc_xprt_enqueue(&svsk->sk_xprt);
441	347	}
..	..	@@ -452,17 +358,9 @@
452	358
453	359	static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
454	360	{
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);
461	362
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);
466	364	}
467	365
468	366	/*
..	..	@@ -520,8 +418,15 @@
520	418	return 0;
521	419	}
522	420
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
525	430	*/
526	431	static int svc_udp_recvfrom(struct svc_rqst *rqstp)
527	432	{
..	..	@@ -555,20 +460,14 @@
555	460	svc_sock_setbufsize(svsk, serv->sv_nrthreads + 3);
556	461
557	462	clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
558		- skb = NULL;
559	463	err = kernel_recvmsg(svsk->sk_sock, &msg, NULL,
560	464	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;
563	470
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		- }
572	471	len = svc_addr_len(svc_addr(rqstp));
573	472	rqstp->rq_addrlen = len;
574	473	if (skb->tstamp == 0) {
..	..	@@ -579,26 +478,21 @@
579	478	sock_write_timestamp(svsk->sk_sk, skb->tstamp);
580	479	set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags); /* there may be more data... */
581	480
582		- len = skb->len;
	481	+ len = skb->len;
583	482	rqstp->rq_arg.len = len;
	483	+ trace_svcsock_udp_recv(&svsk->sk_xprt, len);
584	484
585	485	rqstp->rq_prot = IPPROTO_UDP;
586	486
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;
592	489	rqstp->rq_daddrlen = svc_addr_len(svc_daddr(rqstp));
593	490
594	491	if (skb_is_nonlinear(skb)) {
595	492	/* we have to copy */
596	493	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;
602	496	local_bh_enable();
603	497	consume_skb(skb);
604	498	} else {
..	..	@@ -626,28 +520,79 @@
626	520	serv->sv_stats->netudpcnt++;
627	521
628	522	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();
629	537	out_free:
630	538	kfree_skb(skb);
631	539	return 0;
632	540	}
633	541
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)
636	552	{
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;
638	569
639		- svc_release_udp_skb(rqstp);
	570	+ svc_udp_release_rqst(rqstp);
640	571
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) {
643	582	/* 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);
645	587
646		- return error;
647		-}
	588	+ mutex_unlock(&xprt->xpt_mutex);
	589	+ if (err < 0)
	590	+ return err;
	591	+ return sent;
648	592
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;
651	596	}
652	597
653	598	static int svc_udp_has_wspace(struct svc_xprt *xprt)
..	..	@@ -690,10 +635,10 @@
690	635	.xpo_create = svc_udp_create,
691	636	.xpo_recvfrom = svc_udp_recvfrom,
692	637	.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,
694	640	.xpo_detach = svc_sock_detach,
695	641	.xpo_free = svc_sock_free,
696		- .xpo_prep_reply_hdr = svc_udp_prep_reply_hdr,
697	642	.xpo_has_wspace = svc_udp_has_wspace,
698	643	.xpo_accept = svc_udp_accept,
699	644	.xpo_secure_port = svc_sock_secure_port,
..	..	@@ -710,8 +655,6 @@
710	655
711	656	static void svc_udp_init(struct svc_sock svsk, struct svc_serv serv)
712	657	{
713		- int err, level, optname, one = 1;
714		-
715	658	svc_xprt_init(sock_net(svsk->sk_sock->sk), &svc_udp_class,
716	659	&svsk->sk_xprt, serv);
717	660	clear_bit(XPT_CACHE_AUTH, &svsk->sk_xprt.xpt_flags);
..	..	@@ -731,19 +674,14 @@
731	674	/* make sure we get destination address info */
732	675	switch (svsk->sk_sk->sk_family) {
733	676	case AF_INET:
734		- level = SOL_IP;
735		- optname = IP_PKTINFO;
	677	+ ip_sock_set_pktinfo(svsk->sk_sock->sk);
736	678	break;
737	679	case AF_INET6:
738		- level = SOL_IPV6;
739		- optname = IPV6_RECVPKTINFO;
	680	+ ip6_sock_set_recvpktinfo(svsk->sk_sock->sk);
740	681	break;
741	682	default:
742	683	BUG();
743	684	}
744		- err = kernel_setsockopt(svsk->sk_sock, level, optname,
745		- (char *)&one, sizeof(one));
746		- dprintk("svc: kernel_setsockopt returned %d\n", err);
747	685	}
748	686
749	687	/*
..	..	@@ -753,9 +691,6 @@
753	691	static void svc_tcp_listen_data_ready(struct sock *sk)
754	692	{
755	693	struct svc_sock svsk = (struct svc_sock )sk->sk_user_data;
756		-
757		- dprintk("svc: socket %p TCP (listen) state change %d\n",
758		- sk, sk->sk_state);
759	694
760	695	if (svsk) {
761	696	/* Refer to svc_setup_socket() for details. */
..	..	@@ -777,8 +712,7 @@
777	712	if (svsk) {
778	713	set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
779	714	svc_xprt_enqueue(&svsk->sk_xprt);
780		- } else
781		- printk("svc: socket %p: no user data\n", sk);
	715	+ }
782	716	}
783	717	}
784	718
..	..	@@ -789,15 +723,11 @@
789	723	{
790	724	struct svc_sock svsk = (struct svc_sock )sk->sk_user_data;
791	725
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 {
	726	+ if (svsk) {
798	727	/* Refer to svc_setup_socket() for details. */
799	728	rmb();
800	729	svsk->sk_ostate(sk);
	730	+ trace_svcsock_tcp_state(&svsk->sk_xprt, svsk->sk_sock);
801	731	if (sk->sk_state != TCP_ESTABLISHED) {
802	732	set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
803	733	svc_xprt_enqueue(&svsk->sk_xprt);
..	..	@@ -818,9 +748,7 @@
818	748	struct socket *newsock;
819	749	struct svc_sock *newsvsk;
820	750	int err, slen;
821		- RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
822	751
823		- dprintk("svc: tcp_accept %p sock %p\n", svsk, sock);
824	752	if (!sock)
825	753	return NULL;
826	754
..	..	@@ -833,29 +761,17 @@
833	761	else if (err != -EAGAIN)
834	762	net_warn_ratelimited("%s: accept failed (err %d)!\n",
835	763	serv->sv_name, -err);
	764	+ trace_svcsock_accept_err(xprt, serv->sv_name, err);
836	765	return NULL;
837	766	}
838	767	set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
839	768
840	769	err = kernel_getpeername(newsock, sin);
841	770	if (err < 0) {
842		- net_warn_ratelimited("%s: peername failed (err %d)!\n",
843		- serv->sv_name, -err);
	771	+ trace_svcsock_getpeername_err(xprt, serv->sv_name, err);
844	772	goto failed; /* aborted connection or whatever */
845	773	}
846	774	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)));
859	775
860	776	/* Reset the inherited callbacks before calling svc_setup_socket */
861	777	newsock->sk->sk_state_change = svsk->sk_ostate;
..	..	@@ -874,10 +790,8 @@
874	790	svc_xprt_set_remote(&newsvsk->sk_xprt, sin, slen);
875	791	err = kernel_getsockname(newsock, sin);
876	792	slen = err;
877		- if (unlikely(err < 0)) {
878		- dprintk("svc_tcp_accept: kernel_getsockname error %d\n", -err);
	793	+ if (unlikely(err < 0))
879	794	slen = offsetof(struct sockaddr, sa_data);
880		- }
881	795	svc_xprt_set_local(&newsvsk->sk_xprt, sin, slen);
882	796
883	797	if (sock_is_loopback(newsock->sk))
..	..	@@ -894,13 +808,14 @@
894	808	return NULL;
895	809	}
896	810
897		-static unsigned int svc_tcp_restore_pages(struct svc_sock svsk, struct svc_rqst rqstp)
	811	+static size_t svc_tcp_restore_pages(struct svc_sock *svsk,
	812	+ struct svc_rqst *rqstp)
898	813	{
899		- unsigned int i, len, npages;
	814	+ size_t len = svsk->sk_datalen;
	815	+ unsigned int i, npages;
900	816
901		- if (svsk->sk_datalen == 0)
	817	+ if (!len)
902	818	return 0;
903		- len = svsk->sk_datalen;
904	819	npages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
905	820	for (i = 0; i < npages; i++) {
906	821	if (rqstp->rq_pages[i] != NULL)
..	..	@@ -949,46 +864,45 @@
949	864	}
950	865
951	866	/*
952		- * Receive fragment record header.
953		- * If we haven't gotten the record length yet, get the next four bytes.
	867	+ * Receive fragment record header into sk_marker.
954	868	*/
955		-static int svc_tcp_recv_record(struct svc_sock svsk, struct svc_rqst rqstp)
	869	+static ssize_t svc_tcp_read_marker(struct svc_sock *svsk,
	870	+ struct svc_rqst *rqstp)
956	871	{
957		- struct svc_serv *serv = svsk->sk_xprt.xpt_server;
958		- unsigned int want;
959		- int len;
	872	+ ssize_t want, len;
960	873
	874	+ /* If we haven't gotten the record length yet,
	875	+ * get the next four bytes.
	876	+ */
961	877	if (svsk->sk_tcplen < sizeof(rpc_fraghdr)) {
	878	+ struct msghdr msg = { NULL };
962	879	struct kvec iov;
963	880
964	881	want = sizeof(rpc_fraghdr) - svsk->sk_tcplen;
965		- iov.iov_base = ((char *) &svsk->sk_reclen) + svsk->sk_tcplen;
	882	+ iov.iov_base = ((char *)&svsk->sk_marker) + svsk->sk_tcplen;
966	883	iov.iov_len = want;
967		- if ((len = svc_recvfrom(rqstp, &iov, 1, want)) < 0)
968		- goto error;
	884	+ iov_iter_kvec(&msg.msg_iter, READ, &iov, 1, want);
	885	+ len = sock_recvmsg(svsk->sk_sock, &msg, MSG_DONTWAIT);
	886	+ if (len < 0)
	887	+ return len;
969	888	svsk->sk_tcplen += len;
970		-
971	889	if (len < want) {
972		- dprintk("svc: short recvfrom while reading record "
973		- "length (%d of %d)\n", len, want);
974		- return -EAGAIN;
	890	+ /* call again to read the remaining bytes */
	891	+ goto err_short;
975	892	}
976		-
977		- dprintk("svc: TCP record, %d bytes\n", svc_sock_reclen(svsk));
	893	+ trace_svcsock_marker(&svsk->sk_xprt, svsk->sk_marker);
978	894	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		- }
	895	+ svsk->sk_xprt.xpt_server->sv_max_mesg)
	896	+ goto err_too_large;
984	897	}
985		-
986	898	return svc_sock_reclen(svsk);
987		-error:
988		- dprintk("RPC: TCP recv_record got %d\n", len);
989		- return len;
990		-err_delete:
	899	+
	900	+err_too_large:
	901	+ net_notice_ratelimited("svc: %s %s RPC fragment too large: %d\n",
	902	+ __func__, svsk->sk_xprt.xpt_server->sv_name,
	903	+ svc_sock_reclen(svsk));
991	904	set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
	905	+err_short:
992	906	return -EAGAIN;
993	907	}
994	908
..	..	@@ -1006,7 +920,7 @@
1006	920
1007	921	if (!bc_xprt)
1008	922	return -EAGAIN;
1009		- spin_lock(&bc_xprt->recv_lock);
	923	+ spin_lock(&bc_xprt->queue_lock);
1010	924	req = xprt_lookup_rqst(bc_xprt, xid);
1011	925	if (!req)
1012	926	goto unlock_notfound;
..	..	@@ -1024,7 +938,7 @@
1024	938	memcpy(dst->iov_base, src->iov_base, src->iov_len);
1025	939	xprt_complete_rqst(req->rq_task, rqstp->rq_arg.len);
1026	940	rqstp->rq_arg.len = 0;
1027		- spin_unlock(&bc_xprt->recv_lock);
	941	+ spin_unlock(&bc_xprt->queue_lock);
1028	942	return 0;
1029	943	unlock_notfound:
1030	944	printk(KERN_NOTICE
..	..	@@ -1033,92 +947,62 @@
1033	947	__func__, ntohl(calldir),
1034	948	bc_xprt, ntohl(xid));
1035	949	unlock_eagain:
1036		- spin_unlock(&bc_xprt->recv_lock);
	950	+ spin_unlock(&bc_xprt->queue_lock);
1037	951	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;
1052	952	}
1053	953
1054	954	static void svc_tcp_fragment_received(struct svc_sock *svsk)
1055	955	{
1056	956	/* 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));
1060	957	svsk->sk_tcplen = 0;
1061		- svsk->sk_reclen = 0;
	958	+ svsk->sk_marker = xdr_zero;
1062	959	}
1063	960
1064		-/*
1065		- * Receive data from a TCP socket.
	961	+/**
	962	+ * svc_tcp_recvfrom - Receive data from a TCP socket
	963	+ * @rqstp: request structure into which to receive an RPC Call
	964	+ *
	965	+ * Called in a loop when XPT_DATA has been set.
	966	+ *
	967	+ * Read the 4-byte stream record marker, then use the record length
	968	+ * in that marker to set up exactly the resources needed to receive
	969	+ * the next RPC message into @rqstp.
	970	+ *
	971	+ * Returns:
	972	+ * On success, the number of bytes in a received RPC Call, or
	973	+ * %0 if a complete RPC Call message was not ready to return
	974	+ *
	975	+ * The zero return case handles partial receives and callback Replies.
	976	+ * The state of a partial receive is preserved in the svc_sock for
	977	+ * the next call to svc_tcp_recvfrom.
1066	978	*/
1067	979	static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
1068	980	{
1069	981	struct svc_sock *svsk =
1070	982	container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
1071	983	struct svc_serv *serv = svsk->sk_xprt.xpt_server;
1072		- int len;
1073		- struct kvec *vec;
1074		- unsigned int want, base;
	984	+ size_t want, base;
	985	+ ssize_t len;
1075	986	__be32 *p;
1076	987	__be32 calldir;
1077		- int pnum;
1078	988
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);
	989	+ clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
	990	+ len = svc_tcp_read_marker(svsk, rqstp);
1085	991	if (len < 0)
1086	992	goto error;
1087	993
1088	994	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);
	995	+ want = len - (svsk->sk_tcplen - sizeof(rpc_fraghdr));
	996	+ len = svc_tcp_read_msg(rqstp, base + want, base);
1101	997	if (len >= 0) {
	998	+ trace_svcsock_tcp_recv(&svsk->sk_xprt, len);
1102	999	svsk->sk_tcplen += len;
1103	1000	svsk->sk_datalen += len;
1104	1001	}
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		- }
	1002	+ if (len != want \|\| !svc_sock_final_rec(svsk))
	1003	+ goto err_incomplete;
	1004	+ if (svsk->sk_datalen < 8)
	1005	+ goto err_nuts;
1122	1006
1123	1007	rqstp->rq_arg.len = svsk->sk_datalen;
1124	1008	rqstp->rq_arg.page_base = 0;
..	..	@@ -1153,61 +1037,165 @@
1153	1037
1154	1038	return rqstp->rq_arg.len;
1155	1039
	1040	+err_incomplete:
	1041	+ svc_tcp_save_pages(svsk, rqstp);
	1042	+ if (len < 0 && len != -EAGAIN)
	1043	+ goto err_delete;
	1044	+ if (len == want)
	1045	+ svc_tcp_fragment_received(svsk);
	1046	+ else
	1047	+ trace_svcsock_tcp_recv_short(&svsk->sk_xprt,
	1048	+ svc_sock_reclen(svsk),
	1049	+ svsk->sk_tcplen - sizeof(rpc_fraghdr));
	1050	+ goto err_noclose;
1156	1051	error:
1157	1052	if (len != -EAGAIN)
1158	1053	goto err_delete;
1159		- dprintk("RPC: TCP recvfrom got EAGAIN\n");
	1054	+ trace_svcsock_tcp_recv_eagain(&svsk->sk_xprt, 0);
1160	1055	return 0;
	1056	+err_nuts:
	1057	+ svsk->sk_datalen = 0;
1161	1058	err_delete:
1162		- printk(KERN_NOTICE "%s: recvfrom returned errno %d\n",
1163		- svsk->sk_xprt.xpt_server->sv_name, -len);
	1059	+ trace_svcsock_tcp_recv_err(&svsk->sk_xprt, len);
1164	1060	set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
1165	1061	err_noclose:
1166	1062	return 0; /* record not complete */
1167	1063	}
1168	1064
1169		-/*
1170		- * Send out data on TCP socket.
1171		- */
1172		-static int svc_tcp_sendto(struct svc_rqst *rqstp)
	1065	+static int svc_tcp_send_kvec(struct socket sock, const struct kvec vec,
	1066	+ int flags)
1173	1067	{
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;
	1068	+ return kernel_sendpage(sock, virt_to_page(vec->iov_base),
	1069	+ offset_in_page(vec->iov_base),
	1070	+ vec->iov_len, flags);
1200	1071	}
1201	1072
1202	1073	/*
1203		- * Setup response header. TCP has a 4B record length field.
	1074	+ * kernel_sendpage() is used exclusively to reduce the number of
	1075	+ * copy operations in this path. Therefore the caller must ensure
	1076	+ * that the pages backing @xdr are unchanging.
	1077	+ *
	1078	+ * In addition, the logic assumes that * .bv_len is never larger
	1079	+ * than PAGE_SIZE.
1204	1080	*/
1205		-void svc_tcp_prep_reply_hdr(struct svc_rqst *rqstp)
	1081	+static int svc_tcp_sendmsg(struct socket sock, struct msghdr msg,
	1082	+ struct xdr_buf *xdr, rpc_fraghdr marker,
	1083	+ unsigned int *sentp)
1206	1084	{
1207		- struct kvec *resv = &rqstp->rq_res.head[0];
	1085	+ const struct kvec *head = xdr->head;
	1086	+ const struct kvec *tail = xdr->tail;
	1087	+ struct kvec rm = {
	1088	+ .iov_base = &marker,
	1089	+ .iov_len = sizeof(marker),
	1090	+ };
	1091	+ int flags, ret;
1208	1092
1209		- /* tcp needs a space for the record length... */
1210		- svc_putnl(resv, 0);
	1093	+ *sentp = 0;
	1094	+ ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
	1095	+ if (ret < 0)
	1096	+ return ret;
	1097	+
	1098	+ msg->msg_flags = MSG_MORE;
	1099	+ ret = kernel_sendmsg(sock, msg, &rm, 1, rm.iov_len);
	1100	+ if (ret < 0)
	1101	+ return ret;
	1102	+ *sentp += ret;
	1103	+ if (ret != rm.iov_len)
	1104	+ return -EAGAIN;
	1105	+
	1106	+ flags = head->iov_len < xdr->len ? MSG_MORE \| MSG_SENDPAGE_NOTLAST : 0;
	1107	+ ret = svc_tcp_send_kvec(sock, head, flags);
	1108	+ if (ret < 0)
	1109	+ return ret;
	1110	+ *sentp += ret;
	1111	+ if (ret != head->iov_len)
	1112	+ goto out;
	1113	+
	1114	+ if (xdr->page_len) {
	1115	+ unsigned int offset, len, remaining;
	1116	+ struct bio_vec *bvec;
	1117	+
	1118	+ bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
	1119	+ offset = offset_in_page(xdr->page_base);
	1120	+ remaining = xdr->page_len;
	1121	+ flags = MSG_MORE \| MSG_SENDPAGE_NOTLAST;
	1122	+ while (remaining > 0) {
	1123	+ if (remaining <= PAGE_SIZE && tail->iov_len == 0)
	1124	+ flags = 0;
	1125	+
	1126	+ len = min(remaining, bvec->bv_len - offset);
	1127	+ ret = kernel_sendpage(sock, bvec->bv_page,
	1128	+ bvec->bv_offset + offset,
	1129	+ len, flags);
	1130	+ if (ret < 0)
	1131	+ return ret;
	1132	+ *sentp += ret;
	1133	+ if (ret != len)
	1134	+ goto out;
	1135	+ remaining -= len;
	1136	+ offset = 0;
	1137	+ bvec++;
	1138	+ }
	1139	+ }
	1140	+
	1141	+ if (tail->iov_len) {
	1142	+ ret = svc_tcp_send_kvec(sock, tail, 0);
	1143	+ if (ret < 0)
	1144	+ return ret;
	1145	+ *sentp += ret;
	1146	+ }
	1147	+
	1148	+out:
	1149	+ return 0;
	1150	+}
	1151	+
	1152	+/**
	1153	+ * svc_tcp_sendto - Send out a reply on a TCP socket
	1154	+ * @rqstp: completed svc_rqst
	1155	+ *
	1156	+ * xpt_mutex ensures @rqstp's whole message is written to the socket
	1157	+ * without interruption.
	1158	+ *
	1159	+ * Returns the number of bytes sent, or a negative errno.
	1160	+ */
	1161	+static int svc_tcp_sendto(struct svc_rqst *rqstp)
	1162	+{
	1163	+ struct svc_xprt *xprt = rqstp->rq_xprt;
	1164	+ struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
	1165	+ struct xdr_buf *xdr = &rqstp->rq_res;
	1166	+ rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT \|
	1167	+ (u32)xdr->len);
	1168	+ struct msghdr msg = {
	1169	+ .msg_flags = 0,
	1170	+ };
	1171	+ unsigned int sent;
	1172	+ int err;
	1173	+
	1174	+ svc_tcp_release_rqst(rqstp);
	1175	+
	1176	+ mutex_lock(&xprt->xpt_mutex);
	1177	+ if (svc_xprt_is_dead(xprt))
	1178	+ goto out_notconn;
	1179	+ err = svc_tcp_sendmsg(svsk->sk_sock, &msg, xdr, marker, &sent);
	1180	+ xdr_free_bvec(xdr);
	1181	+ trace_svcsock_tcp_send(xprt, err < 0 ? (long)err : sent);
	1182	+ if (err < 0 \|\| sent != (xdr->len + sizeof(marker)))
	1183	+ goto out_close;
	1184	+ mutex_unlock(&xprt->xpt_mutex);
	1185	+ return sent;
	1186	+
	1187	+out_notconn:
	1188	+ mutex_unlock(&xprt->xpt_mutex);
	1189	+ return -ENOTCONN;
	1190	+out_close:
	1191	+ pr_notice("rpc-srv/tcp: %s: %s %d when sending %d bytes - shutting down socket\n",
	1192	+ xprt->xpt_server->sv_name,
	1193	+ (err < 0) ? "got error" : "sent",
	1194	+ (err < 0) ? err : sent, xdr->len);
	1195	+ set_bit(XPT_CLOSE, &xprt->xpt_flags);
	1196	+ svc_xprt_enqueue(xprt);
	1197	+ mutex_unlock(&xprt->xpt_mutex);
	1198	+ return -EAGAIN;
1211	1199	}
1212	1200
1213	1201	static struct svc_xprt svc_tcp_create(struct svc_serv serv,
..	..	@@ -1218,66 +1206,14 @@
1218	1206	return svc_create_socket(serv, IPPROTO_TCP, net, sa, salen, flags);
1219	1207	}
1220	1208
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		-
1273	1209	static const struct svc_xprt_ops svc_tcp_ops = {
1274	1210	.xpo_create = svc_tcp_create,
1275	1211	.xpo_recvfrom = svc_tcp_recvfrom,
1276	1212	.xpo_sendto = svc_tcp_sendto,
1277		- .xpo_release_rqst = svc_release_skb,
	1213	+ .xpo_read_payload = svc_sock_read_payload,
	1214	+ .xpo_release_rqst = svc_tcp_release_rqst,
1278	1215	.xpo_detach = svc_tcp_sock_detach,
1279	1216	.xpo_free = svc_sock_free,
1280		- .xpo_prep_reply_hdr = svc_tcp_prep_reply_hdr,
1281	1217	.xpo_has_wspace = svc_tcp_has_wspace,
1282	1218	.xpo_accept = svc_tcp_accept,
1283	1219	.xpo_secure_port = svc_sock_secure_port,
..	..	@@ -1296,14 +1232,12 @@
1296	1232	{
1297	1233	svc_reg_xprt_class(&svc_tcp_class);
1298	1234	svc_reg_xprt_class(&svc_udp_class);
1299		- svc_init_bc_xprt_sock();
1300	1235	}
1301	1236
1302	1237	void svc_cleanup_xprt_sock(void)
1303	1238	{
1304	1239	svc_unreg_xprt_class(&svc_tcp_class);
1305	1240	svc_unreg_xprt_class(&svc_udp_class);
1306		- svc_cleanup_bc_xprt_sock();
1307	1241	}
1308	1242
1309	1243	static void svc_tcp_init(struct svc_sock svsk, struct svc_serv serv)
..	..	@@ -1315,18 +1249,16 @@
1315	1249	set_bit(XPT_CACHE_AUTH, &svsk->sk_xprt.xpt_flags);
1316	1250	set_bit(XPT_CONG_CTRL, &svsk->sk_xprt.xpt_flags);
1317	1251	if (sk->sk_state == TCP_LISTEN) {
1318		- dprintk("setting up TCP socket for listening\n");
1319	1252	strcpy(svsk->sk_xprt.xpt_remotebuf, "listener");
1320	1253	set_bit(XPT_LISTENER, &svsk->sk_xprt.xpt_flags);
1321	1254	sk->sk_data_ready = svc_tcp_listen_data_ready;
1322	1255	set_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags);
1323	1256	} else {
1324		- dprintk("setting up TCP socket for reading\n");
1325	1257	sk->sk_state_change = svc_tcp_state_change;
1326	1258	sk->sk_data_ready = svc_data_ready;
1327	1259	sk->sk_write_space = svc_write_space;
1328	1260
1329		- svsk->sk_reclen = 0;
	1261	+ svsk->sk_marker = xdr_zero;
1330	1262	svsk->sk_tcplen = 0;
1331	1263	svsk->sk_datalen = 0;
1332	1264	memset(&svsk->sk_pages[0], 0, sizeof(svsk->sk_pages));
..	..	@@ -1371,7 +1303,6 @@
1371	1303	int pmap_register = !(flags & SVC_SOCK_ANONYMOUS);
1372	1304	int err = 0;
1373	1305
1374		- dprintk("svc: svc_setup_socket %p\n", sock);
1375	1306	svsk = kzalloc(sizeof(*svsk), GFP_KERNEL);
1376	1307	if (!svsk)
1377	1308	return ERR_PTR(-ENOMEM);
..	..	@@ -1408,12 +1339,7 @@
1408	1339	else
1409	1340	svc_tcp_init(svsk, serv);
1410	1341
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		-
	1342	+ trace_svcsock_new_socket(sock);
1417	1343	return svsk;
1418	1344	}
1419	1345
..	..	@@ -1439,13 +1365,14 @@
1439	1365	* @fd: file descriptor of the new listener
1440	1366	* @name_return: pointer to buffer to fill in with name of listener
1441	1367	* @len: size of the buffer
	1368	+ * @cred: credential
1442	1369	*
1443	1370	* Fills in socket name and returns positive length of name if successful.
1444	1371	* Name is terminated with '\n'. On error, returns a negative errno
1445	1372	* value.
1446	1373	*/
1447	1374	int svc_addsock(struct svc_serv serv, const int fd, char name_return,
1448		- const size_t len)
	1375	+ const size_t len, const struct cred *cred)
1449	1376	{
1450	1377	int err = 0;
1451	1378	struct socket *so = sockfd_lookup(fd, &err);
..	..	@@ -1478,6 +1405,7 @@
1478	1405	salen = kernel_getsockname(svsk->sk_sock, sin);
1479	1406	if (salen >= 0)
1480	1407	svc_xprt_set_local(&svsk->sk_xprt, sin, salen);
	1408	+ svsk->sk_xprt.xpt_cred = get_cred(cred);
1481	1409	svc_add_new_perm_xprt(serv, &svsk->sk_xprt);
1482	1410	return svc_one_sock_name(svsk, name_return, len);
1483	1411	out:
..	..	@@ -1503,12 +1431,6 @@
1503	1431	struct sockaddr newsin = (struct sockaddr )&addr;
1504	1432	int newlen;
1505	1433	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)));
1512	1434
1513	1435	if (protocol != IPPROTO_UDP && protocol != IPPROTO_TCP) {
1514	1436	printk(KERN_WARNING "svc: only UDP and TCP "
..	..	@@ -1539,11 +1461,8 @@
1539	1461	* getting requests from IPv4 remotes. Those should
1540	1462	* be shunted to a PF_INET listener via rpcbind.
1541	1463	*/
1542		- val = 1;
1543	1464	if (family == PF_INET6)
1544		- kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
1545		- (char *)&val, sizeof(val));
1546		-
	1465	+ ip6_sock_set_v6only(sock->sk);
1547	1466	if (type == SOCK_STREAM)
1548	1467	sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
1549	1468	error = kernel_bind(sock, sin, len);
..	..	@@ -1568,7 +1487,6 @@
1568	1487	svc_xprt_set_local(&svsk->sk_xprt, newsin, newlen);
1569	1488	return (struct svc_xprt *)svsk;
1570	1489	bummer:
1571		- dprintk("svc: svc_create_socket error = %d\n", -error);
1572	1490	sock_release(sock);
1573	1491	return ERR_PTR(error);
1574	1492	}
..	..	@@ -1581,8 +1499,6 @@
1581	1499	{
1582	1500	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
1583	1501	struct sock *sk = svsk->sk_sk;
1584		-
1585		- dprintk("svc: svc_sock_detach(%p)\n", svsk);
1586	1502
1587	1503	/* put back the old socket callbacks */
1588	1504	lock_sock(sk);
..	..	@@ -1600,8 +1516,6 @@
1600	1516	{
1601	1517	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
1602	1518
1603		- dprintk("svc: svc_tcp_sock_detach(%p)\n", svsk);
1604		-
1605	1519	svc_sock_detach(xprt);
1606	1520
1607	1521	if (!test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
..	..	@@ -1616,7 +1530,6 @@
1616	1530	static void svc_sock_free(struct svc_xprt *xprt)
1617	1531	{
1618	1532	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
1619		- dprintk("svc: svc_sock_free(%p)\n", svsk);
1620	1533
1621	1534	if (svsk->sk_sock->file)
1622	1535	sockfd_put(svsk->sk_sock);
..	..	@@ -1624,45 +1537,3 @@
1624	1537	sock_release(svsk->sk_sock);
1625	1538	kfree(svsk);
1626	1539	}
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 */