projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ALSA: hda - Update descriptions about new position_fix values
[linux-2.6-block.git] / net / sunrpc / xprtsock.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
a246b010
CL		 2/*
3 * linux/net/sunrpc/xprtsock.c
4 *
5 * Client-side transport implementation for sockets.
6 *
113aa838
AC		 7 * TCP callback races fixes (C) 1998 Red Hat
8 * TCP send fixes (C) 1998 Red Hat
a246b010
CL		 9 * TCP NFS related read + write fixes
10 * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
11 *
12 * Rewrite of larges part of the code in order to stabilize TCP stuff.
13 * Fix behaviour when socket buffer is full.
14 * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
55aa4f58
CL		 15 *
16 * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com>
8f9d5b1a
CL		 17 *
18 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
19 * <gilles.quillard@bull.net>
a246b010
CL		 20 */
21
22#include <linux/types.h>
176e21ee 23#include <linux/string.h>
a246b010 24#include <linux/slab.h>
bc25571e 25#include <linux/module.h>
a246b010 26#include <linux/capability.h>
a246b010
CL		 27#include <linux/pagemap.h>
28#include <linux/errno.h>
29#include <linux/socket.h>
30#include <linux/in.h>
31#include <linux/net.h>
32#include <linux/mm.h>
176e21ee 33#include <linux/un.h>
a246b010
CL		 34#include <linux/udp.h>
35#include <linux/tcp.h>
36#include <linux/sunrpc/clnt.h>
5976687a 37#include <linux/sunrpc/addr.h>
02107148 38#include <linux/sunrpc/sched.h>
4cfc7e60 39#include <linux/sunrpc/svcsock.h>
49c36fcc 40#include <linux/sunrpc/xprtsock.h>
a246b010 41#include <linux/file.h>
9e00abc3 42#ifdef CONFIG_SUNRPC_BACKCHANNEL
44b98efd
RL		 43#include <linux/sunrpc/bc_xprt.h>
44#endif
a246b010
CL		 45
46#include <net/sock.h>
47#include <net/checksum.h>
48#include <net/udp.h>
49#include <net/tcp.h>
277e4ab7 50#include <linux/bvec.h>
6a829eb8 51#include <linux/highmem.h>
277e4ab7 52#include <linux/uio.h>
a1231fda 53#include <linux/sched/mm.h>
a246b010 54
40b5ea0c
TM		 55#include <trace/events/sunrpc.h>
56
4cfc7e60 57#include "sunrpc.h"
176e21ee
CL		 58
59static void xs_close(struct rpc_xprt *xprt);
7196dbb0
TM		 60static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
61 struct socket *sock);
176e21ee 62
c556b754
CL		 63/*
64 * xprtsock tunables
65 */
09acfea5
TM		 66static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
67static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
68static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
c556b754 69
09acfea5
TM		 70static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
71static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
c556b754 72
7d1e8255 73#define XS_TCP_LINGER_TO (15U * HZ)
25fe6142 74static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
7d1e8255 75
fbf76683
CL		 76/*
77 * We can register our own files under /proc/sys/sunrpc by
78 * calling register_sysctl_table() again. The files in that
79 * directory become the union of all files registered there.
80 *
81 * We simply need to make sure that we don't collide with
82 * someone else's file names!
83 */
84
fbf76683
CL		 85static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
86static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
d9ba131d 87static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
fbf76683
CL		 88static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
89static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
90
91static struct ctl_table_header *sunrpc_table_header;
92
93/*
94 * FIXME: changing the UDP slot table size should also resize the UDP
95 * socket buffers for existing UDP transports
96 */
fe2c6338 97static struct ctl_table xs_tunables_table[] = {
fbf76683 98 {
fbf76683
CL		 99 .procname = "udp_slot_table_entries",
100 .data = &xprt_udp_slot_table_entries,
101 .maxlen = sizeof(unsigned int),
102 .mode = 0644,
6d456111 103 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 104 .extra1 = &min_slot_table_size,
105 .extra2 = &max_slot_table_size
106 },
107 {
fbf76683
CL		 108 .procname = "tcp_slot_table_entries",
109 .data = &xprt_tcp_slot_table_entries,
110 .maxlen = sizeof(unsigned int),
111 .mode = 0644,
6d456111 112 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 113 .extra1 = &min_slot_table_size,
114 .extra2 = &max_slot_table_size
115 },
d9ba131d
TM		 116 {
117 .procname = "tcp_max_slot_table_entries",
118 .data = &xprt_max_tcp_slot_table_entries,
119 .maxlen = sizeof(unsigned int),
120 .mode = 0644,
121 .proc_handler = proc_dointvec_minmax,
122 .extra1 = &min_slot_table_size,
123 .extra2 = &max_tcp_slot_table_limit
124 },
fbf76683 125 {
fbf76683
CL		 126 .procname = "min_resvport",
127 .data = &xprt_min_resvport,
128 .maxlen = sizeof(unsigned int),
129 .mode = 0644,
6d456111 130 .proc_handler = proc_dointvec_minmax,
fbf76683 131 .extra1 = &xprt_min_resvport_limit,
826799e6 132 .extra2 = &xprt_max_resvport_limit
fbf76683
CL		 133 },
134 {
fbf76683
CL		 135 .procname = "max_resvport",
136 .data = &xprt_max_resvport,
137 .maxlen = sizeof(unsigned int),
138 .mode = 0644,
6d456111 139 .proc_handler = proc_dointvec_minmax,
826799e6 140 .extra1 = &xprt_min_resvport_limit,
fbf76683
CL		 141 .extra2 = &xprt_max_resvport_limit
142 },
25fe6142
TM		 143 {
144 .procname = "tcp_fin_timeout",
145 .data = &xs_tcp_fin_timeout,
146 .maxlen = sizeof(xs_tcp_fin_timeout),
147 .mode = 0644,
6d456111 148 .proc_handler = proc_dointvec_jiffies,
fbf76683 149 },
f8572d8f 150 { },
fbf76683
CL		 151};
152
fe2c6338 153static struct ctl_table sunrpc_table[] = {
fbf76683 154 {
fbf76683
CL		 155 .procname = "sunrpc",
156 .mode = 0555,
157 .child = xs_tunables_table
158 },
f8572d8f 159 { },
fbf76683
CL		 160};
161
03bf4b70
CL		 162/*
163 * Wait duration for a reply from the RPC portmapper.
164 */
165#define XS_BIND_TO (60U * HZ)
166
167/*
168 * Delay if a UDP socket connect error occurs. This is most likely some
169 * kind of resource problem on the local host.
170 */
171#define XS_UDP_REEST_TO (2U * HZ)
172
173/*
174 * The reestablish timeout allows clients to delay for a bit before attempting
175 * to reconnect to a server that just dropped our connection.
176 *
177 * We implement an exponential backoff when trying to reestablish a TCP
178 * transport connection with the server. Some servers like to drop a TCP
179 * connection when they are overworked, so we start with a short timeout and
180 * increase over time if the server is down or not responding.
181 */
182#define XS_TCP_INIT_REEST_TO (3U * HZ)
03bf4b70
CL		 183
184/*
185 * TCP idle timeout; client drops the transport socket if it is idle
186 * for this long. Note that we also timeout UDP sockets to prevent
187 * holding port numbers when there is no RPC traffic.
188 */
189#define XS_IDLE_DISC_TO (5U * 60 * HZ)
190
f895b252 191#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
a246b010 192# undef RPC_DEBUG_DATA
9903cd1c 193# define RPCDBG_FACILITY RPCDBG_TRANS
a246b010
CL		 194#endif
195
a246b010 196#ifdef RPC_DEBUG_DATA
9903cd1c 197static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010 198{
9903cd1c
CL		 199 u8 *buf = (u8 *) packet;
200 int j;
a246b010 201
46121cf7 202 dprintk("RPC: %s\n", msg);
a246b010
CL		 203 for (j = 0; j < count && j < 128; j += 4) {
204 if (!(j & 31)) {
205 if (j)
206 dprintk("\n");
207 dprintk("0x%04x ", j);
208 }
209 dprintk("%02x%02x%02x%02x ",
210 buf[j], buf[j+1], buf[j+2], buf[j+3]);
211 }
212 dprintk("\n");
213}
214#else
9903cd1c 215static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010
CL		 216{
217 /* NOP */
218}
219#endif
220
2118071d
TM		 221static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
222{
223 return (struct rpc_xprt *) sk->sk_user_data;
224}
225
95392c59
CL		 226static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
227{
228 return (struct sockaddr *) &xprt->addr;
229}
230
176e21ee
CL		 231static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
232{
233 return (struct sockaddr_un *) &xprt->addr;
234}
235
95392c59 236static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
edb267a6 237{
95392c59
CL		 238 return (struct sockaddr_in *) &xprt->addr;
239}
240
241static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
242{
243 return (struct sockaddr_in6 *) &xprt->addr;
244}
245
c877b849 246static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
edb267a6 247{
c877b849 248 struct sockaddr *sap = xs_addr(xprt);
9dc3b095
CL		 249 struct sockaddr_in6 *sin6;
250 struct sockaddr_in *sin;
176e21ee 251 struct sockaddr_un *sun;
c877b849 252 char buf[128];
edb267a6 253
9dc3b095 254 switch (sap->sa_family) {
176e21ee
CL		 255 case AF_LOCAL:
256 sun = xs_addr_un(xprt);
257 strlcpy(buf, sun->sun_path, sizeof(buf));
258 xprt->address_strings[RPC_DISPLAY_ADDR] =
259 kstrdup(buf, GFP_KERNEL);
260 break;
9dc3b095 261 case AF_INET:
176e21ee
CL		 262 (void)rpc_ntop(sap, buf, sizeof(buf));
263 xprt->address_strings[RPC_DISPLAY_ADDR] =
264 kstrdup(buf, GFP_KERNEL);
9dc3b095 265 sin = xs_addr_in(xprt);
fc0b5791 266 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
9dc3b095
CL		 267 break;
268 case AF_INET6:
176e21ee
CL		 269 (void)rpc_ntop(sap, buf, sizeof(buf));
270 xprt->address_strings[RPC_DISPLAY_ADDR] =
271 kstrdup(buf, GFP_KERNEL);
9dc3b095 272 sin6 = xs_addr_in6(xprt);
fc0b5791 273 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
9dc3b095
CL		 274 break;
275 default:
276 BUG();
756805e7 277 }
176e21ee 278
9dc3b095 279 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
edb267a6
CL		 280}
281
9dc3b095 282static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
4b6473fb 283{
9dc3b095
CL		 284 struct sockaddr *sap = xs_addr(xprt);
285 char buf[128];
4b6473fb 286
81160e66 287 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
c877b849 288 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
4b6473fb 289
81160e66 290 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
c877b849
CL		 291 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
292}
4b6473fb 293
9dc3b095
CL		 294static void xs_format_peer_addresses(struct rpc_xprt *xprt,
295 const char *protocol,
296 const char *netid)
c877b849 297{
b454ae90 298 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
b454ae90 299 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
c877b849 300 xs_format_common_peer_addresses(xprt);
9dc3b095 301 xs_format_common_peer_ports(xprt);
edb267a6 302}
4b6473fb 303
9dc3b095 304static void xs_update_peer_port(struct rpc_xprt *xprt)
4b6473fb 305{
9dc3b095
CL		 306 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
307 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
4417c8c4 308
9dc3b095 309 xs_format_common_peer_ports(xprt);
edb267a6
CL		 310}
311
312static void xs_free_peer_addresses(struct rpc_xprt *xprt)
313{
33e01dc7
CL		 314 unsigned int i;
315
316 for (i = 0; i < RPC_DISPLAY_MAX; i++)
317 switch (i) {
318 case RPC_DISPLAY_PROTO:
319 case RPC_DISPLAY_NETID:
320 continue;
321 default:
322 kfree(xprt->address_strings[i]);
323 }
edb267a6
CL		 324}
325
277e4ab7
TM		 326static size_t
327xs_alloc_sparse_pages(struct xdr_buf *buf, size_t want, gfp_t gfp)
328{
329 size_t i,n;
330
16e5e90f 331 if (!want || !(buf->flags & XDRBUF_SPARSE_PAGES))
277e4ab7 332 return want;
277e4ab7
TM		 333 n = (buf->page_base + want + PAGE_SIZE - 1) >> PAGE_SHIFT;
334 for (i = 0; i < n; i++) {
335 if (buf->pages[i])
336 continue;
337 buf->bvec[i].bv_page = buf->pages[i] = alloc_page(gfp);
338 if (!buf->pages[i]) {
16e5e90f
TM		 339 i *= PAGE_SIZE;
340 return i > buf->page_base ? i - buf->page_base : 0;
277e4ab7
TM		 341 }
342 }
343 return want;
344}
345
346static ssize_t
347xs_sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags, size_t seek)
348{
349 ssize_t ret;
350 if (seek != 0)
351 iov_iter_advance(&msg->msg_iter, seek);
352 ret = sock_recvmsg(sock, msg, flags);
353 return ret > 0 ? ret + seek : ret;
354}
355
356static ssize_t
357xs_read_kvec(struct socket *sock, struct msghdr *msg, int flags,
358 struct kvec *kvec, size_t count, size_t seek)
359{
0e9b4a82 360 iov_iter_kvec(&msg->msg_iter, READ, kvec, 1, count);
277e4ab7
TM		 361 return xs_sock_recvmsg(sock, msg, flags, seek);
362}
363
364static ssize_t
365xs_read_bvec(struct socket *sock, struct msghdr *msg, int flags,
366 struct bio_vec *bvec, unsigned long nr, size_t count,
367 size_t seek)
368{
0e9b4a82 369 iov_iter_bvec(&msg->msg_iter, READ, bvec, nr, count);
277e4ab7
TM		 370 return xs_sock_recvmsg(sock, msg, flags, seek);
371}
372
373static ssize_t
374xs_read_discard(struct socket *sock, struct msghdr *msg, int flags,
375 size_t count)
376{
b76a5afd
TM		 377 iov_iter_discard(&msg->msg_iter, READ, count);
378 return sock_recvmsg(sock, msg, flags);
277e4ab7
TM		 379}
380
6a829eb8
TM		 381#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
382static void
383xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
384{
385 struct bvec_iter bi = {
386 .bi_size = count,
387 };
388 struct bio_vec bv;
389
390 bvec_iter_advance(bvec, &bi, seek & PAGE_MASK);
391 for_each_bvec(bv, bvec, bi, bi)
392 flush_dcache_page(bv.bv_page);
393}
394#else
395static inline void
396xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
397{
398}
399#endif
400
277e4ab7
TM		 401static ssize_t
402xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags,
403 struct xdr_buf *buf, size_t count, size_t seek, size_t *read)
404{
405 size_t want, seek_init = seek, offset = 0;
406 ssize_t ret;
407
e92053a5
TM		 408 want = min_t(size_t, count, buf->head[0].iov_len);
409 if (seek < want) {
277e4ab7
TM		 410 ret = xs_read_kvec(sock, msg, flags, &buf->head[0], want, seek);
411 if (ret <= 0)
412 goto sock_err;
413 offset += ret;
414 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
415 goto out;
416 if (ret != want)
c4433055 417 goto out;
277e4ab7
TM		 418 seek = 0;
419 } else {
e92053a5
TM		 420 seek -= want;
421 offset += want;
277e4ab7 422 }
16e5e90f
TM		 423
424 want = xs_alloc_sparse_pages(buf,
425 min_t(size_t, count - offset, buf->page_len),
12a3ad61 426 GFP_KERNEL);
16e5e90f 427 if (seek < want) {
277e4ab7
TM		 428 ret = xs_read_bvec(sock, msg, flags, buf->bvec,
429 xdr_buf_pagecount(buf),
430 want + buf->page_base,
431 seek + buf->page_base);
432 if (ret <= 0)
433 goto sock_err;
6a829eb8 434 xs_flush_bvec(buf->bvec, ret, seek + buf->page_base);
277e4ab7
TM		 435 offset += ret - buf->page_base;
436 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
437 goto out;
438 if (ret != want)
c4433055 439 goto out;
277e4ab7
TM		 440 seek = 0;
441 } else {
16e5e90f
TM		 442 seek -= want;
443 offset += want;
277e4ab7 444 }
16e5e90f 445
e92053a5
TM		 446 want = min_t(size_t, count - offset, buf->tail[0].iov_len);
447 if (seek < want) {
277e4ab7
TM		 448 ret = xs_read_kvec(sock, msg, flags, &buf->tail[0], want, seek);
449 if (ret <= 0)
450 goto sock_err;
451 offset += ret;
452 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
453 goto out;
454 if (ret != want)
c4433055 455 goto out;
9734ad57 456 } else if (offset < seek_init)
e92053a5 457 offset = seek_init;
277e4ab7 458 ret = -EMSGSIZE;
277e4ab7
TM		 459out:
460 *read = offset - seek_init;
461 return ret;
277e4ab7
TM		 462sock_err:
463 offset += seek;
464 goto out;
465}
466
467static void
468xs_read_header(struct sock_xprt *transport, struct xdr_buf *buf)
469{
470 if (!transport->recv.copied) {
471 if (buf->head[0].iov_len >= transport->recv.offset)
472 memcpy(buf->head[0].iov_base,
473 &transport->recv.xid,
474 transport->recv.offset);
475 transport->recv.copied = transport->recv.offset;
476 }
477}
478
479static bool
480xs_read_stream_request_done(struct sock_xprt *transport)
481{
482 return transport->recv.fraghdr & cpu_to_be32(RPC_LAST_STREAM_FRAGMENT);
483}
484
e92053a5
TM		 485static void
486xs_read_stream_check_eor(struct sock_xprt *transport,
487 struct msghdr *msg)
488{
489 if (xs_read_stream_request_done(transport))
490 msg->msg_flags |= MSG_EOR;
491}
492
277e4ab7
TM		 493static ssize_t
494xs_read_stream_request(struct sock_xprt *transport, struct msghdr *msg,
495 int flags, struct rpc_rqst *req)
496{
497 struct xdr_buf *buf = &req->rq_private_buf;
01f2f5b8
AH		 498 size_t want, uninitialized_var(read);
499 ssize_t uninitialized_var(ret);
277e4ab7
TM		 500
501 xs_read_header(transport, buf);
502
503 want = transport->recv.len - transport->recv.offset;
e92053a5
TM		 504 if (want != 0) {
505 ret = xs_read_xdr_buf(transport->sock, msg, flags, buf,
506 transport->recv.copied + want,
507 transport->recv.copied,
508 &read);
509 transport->recv.offset += read;
510 transport->recv.copied += read;
727fcc64 511 }
e92053a5 512
727fcc64 513 if (transport->recv.offset == transport->recv.len)
e92053a5 514 xs_read_stream_check_eor(transport, msg);
277e4ab7 515
e92053a5
TM		 516 if (want == 0)
517 return 0;
518
277e4ab7 519 switch (ret) {
c4433055
TM		 520 default:
521 break;
26781eab 522 case -EFAULT:
277e4ab7 523 case -EMSGSIZE:
26781eab 524 msg->msg_flags |= MSG_TRUNC;
c4433055 525 return read;
277e4ab7
TM		 526 case 0:
527 return -ESHUTDOWN;
277e4ab7 528 }
c4433055 529 return ret < 0 ? ret : read;
277e4ab7
TM		 530}
531
532static size_t
533xs_read_stream_headersize(bool isfrag)
534{
535 if (isfrag)
536 return sizeof(__be32);
537 return 3 * sizeof(__be32);
538}
539
540static ssize_t
541xs_read_stream_header(struct sock_xprt *transport, struct msghdr *msg,
542 int flags, size_t want, size_t seek)
543{
544 struct kvec kvec = {
545 .iov_base = &transport->recv.fraghdr,
546 .iov_len = want,
547 };
548 return xs_read_kvec(transport->sock, msg, flags, &kvec, want, seek);
549}
550
551#if defined(CONFIG_SUNRPC_BACKCHANNEL)
552static ssize_t
553xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
554{
555 struct rpc_xprt *xprt = &transport->xprt;
556 struct rpc_rqst *req;
557 ssize_t ret;
558
559 /* Look up and lock the request corresponding to the given XID */
560 req = xprt_lookup_bc_request(xprt, transport->recv.xid);
561 if (!req) {
562 printk(KERN_WARNING "Callback slot table overflowed\n");
563 return -ESHUTDOWN;
564 }
565
566 ret = xs_read_stream_request(transport, msg, flags, req);
567 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 568 xprt_complete_bc_request(req, transport->recv.copied);
277e4ab7
TM		 569
570 return ret;
571}
572#else /* CONFIG_SUNRPC_BACKCHANNEL */
573static ssize_t
574xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
575{
576 return -ESHUTDOWN;
577}
578#endif /* CONFIG_SUNRPC_BACKCHANNEL */
579
580static ssize_t
581xs_read_stream_reply(struct sock_xprt *transport, struct msghdr *msg, int flags)
582{
583 struct rpc_xprt *xprt = &transport->xprt;
584 struct rpc_rqst *req;
585 ssize_t ret = 0;
586
587 /* Look up and lock the request corresponding to the given XID */
588 spin_lock(&xprt->queue_lock);
589 req = xprt_lookup_rqst(xprt, transport->recv.xid);
590 if (!req) {
591 msg->msg_flags |= MSG_TRUNC;
592 goto out;
593 }
594 xprt_pin_rqst(req);
595 spin_unlock(&xprt->queue_lock);
596
597 ret = xs_read_stream_request(transport, msg, flags, req);
598
599 spin_lock(&xprt->queue_lock);
600 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 601 xprt_complete_rqst(req->rq_task, transport->recv.copied);
277e4ab7
TM		 602 xprt_unpin_rqst(req);
603out:
604 spin_unlock(&xprt->queue_lock);
605 return ret;
606}
607
608static ssize_t
609xs_read_stream(struct sock_xprt *transport, int flags)
610{
611 struct msghdr msg = { 0 };
612 size_t want, read = 0;
613 ssize_t ret = 0;
614
615 if (transport->recv.len == 0) {
616 want = xs_read_stream_headersize(transport->recv.copied != 0);
617 ret = xs_read_stream_header(transport, &msg, flags, want,
618 transport->recv.offset);
619 if (ret <= 0)
620 goto out_err;
621 transport->recv.offset = ret;
c4433055
TM		 622 if (transport->recv.offset != want)
623 return transport->recv.offset;
277e4ab7
TM		 624 transport->recv.len = be32_to_cpu(transport->recv.fraghdr) &
625 RPC_FRAGMENT_SIZE_MASK;
626 transport->recv.offset -= sizeof(transport->recv.fraghdr);
627 read = ret;
628 }
629
630 switch (be32_to_cpu(transport->recv.calldir)) {
c4433055
TM		 631 default:
632 msg.msg_flags |= MSG_TRUNC;
633 break;
277e4ab7
TM		 634 case RPC_CALL:
635 ret = xs_read_stream_call(transport, &msg, flags);
636 break;
637 case RPC_REPLY:
638 ret = xs_read_stream_reply(transport, &msg, flags);
639 }
640 if (msg.msg_flags & MSG_TRUNC) {
641 transport->recv.calldir = cpu_to_be32(-1);
642 transport->recv.copied = -1;
643 }
644 if (ret < 0)
645 goto out_err;
646 read += ret;
647 if (transport->recv.offset < transport->recv.len) {
c4433055
TM		 648 if (!(msg.msg_flags & MSG_TRUNC))
649 return read;
b76a5afd 650 msg.msg_flags = 0;
277e4ab7
TM		 651 ret = xs_read_discard(transport->sock, &msg, flags,
652 transport->recv.len - transport->recv.offset);
653 if (ret <= 0)
654 goto out_err;
655 transport->recv.offset += ret;
656 read += ret;
657 if (transport->recv.offset != transport->recv.len)
c4433055 658 return read;
277e4ab7
TM		 659 }
660 if (xs_read_stream_request_done(transport)) {
c50b8ee0 661 trace_xs_stream_read_request(transport);
277e4ab7
TM		 662 transport->recv.copied = 0;
663 }
664 transport->recv.offset = 0;
665 transport->recv.len = 0;
666 return read;
667out_err:
79462857 668 return ret != 0 ? ret : -ESHUTDOWN;
277e4ab7
TM		 669}
670
0ffe86f4
TM		 671static __poll_t xs_poll_socket(struct sock_xprt *transport)
672{
a73881c9
TM		 673 return transport->sock->ops->poll(transport->file, transport->sock,
674 NULL);
0ffe86f4
TM		 675}
676
677static bool xs_poll_socket_readable(struct sock_xprt *transport)
678{
679 __poll_t events = xs_poll_socket(transport);
680
681 return (events & (EPOLLIN | EPOLLRDNORM)) && !(events & EPOLLRDHUP);
682}
683
684static void xs_poll_check_readable(struct sock_xprt *transport)
685{
686
687 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
688 if (!xs_poll_socket_readable(transport))
689 return;
690 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
691 queue_work(xprtiod_workqueue, &transport->recv_worker);
692}
693
c50b8ee0
TM		 694static void xs_stream_data_receive(struct sock_xprt *transport)
695{
696 size_t read = 0;
697 ssize_t ret = 0;
698
699 mutex_lock(&transport->recv_mutex);
700 if (transport->sock == NULL)
701 goto out;
c50b8ee0
TM		 702 for (;;) {
703 ret = xs_read_stream(transport, MSG_DONTWAIT);
c4433055 704 if (ret < 0)
c50b8ee0
TM		 705 break;
706 read += ret;
707 cond_resched();
708 }
5f52a9d4
TM		 709 if (ret == -ESHUTDOWN)
710 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
711 else
712 xs_poll_check_readable(transport);
c50b8ee0
TM		 713out:
714 mutex_unlock(&transport->recv_mutex);
715 trace_xs_stream_read_data(&transport->xprt, ret, read);
716}
717
718static void xs_stream_data_receive_workfn(struct work_struct *work)
719{
720 struct sock_xprt *transport =
721 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 722 unsigned int pflags = memalloc_nofs_save();
723
c50b8ee0 724 xs_stream_data_receive(transport);
a1231fda 725 memalloc_nofs_restore(pflags);
c50b8ee0
TM		 726}
727
550aebfe
TM		 728static void
729xs_stream_reset_connect(struct sock_xprt *transport)
730{
731 transport->recv.offset = 0;
732 transport->recv.len = 0;
733 transport->recv.copied = 0;
734 transport->xmit.offset = 0;
ae053551
TM		 735}
736
737static void
738xs_stream_start_connect(struct sock_xprt *transport)
739{
550aebfe
TM		 740 transport->xprt.stat.connect_count++;
741 transport->xprt.stat.connect_start = jiffies;
742}
743
b4b5cc85
CL		 744#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
745
e791f8e9 746static int xs_sendmsg(struct socket *sock, struct msghdr *msg, size_t seek)
b4b5cc85 747{
e791f8e9
TM		 748 if (seek)
749 iov_iter_advance(&msg->msg_iter, seek);
750 return sock_sendmsg(sock, msg);
751}
b4b5cc85 752
e791f8e9
TM		 753static int xs_send_kvec(struct socket *sock, struct msghdr *msg, struct kvec *vec, size_t seek)
754{
e791f8e9
TM		 755 iov_iter_kvec(&msg->msg_iter, WRITE, vec, 1, vec->iov_len);
756 return xs_sendmsg(sock, msg, seek);
b4b5cc85
CL		 757}
758
0472e476 759static int xs_send_pagedata(struct socket *sock, struct msghdr *msg, struct xdr_buf *xdr, size_t base)
b4b5cc85 760{
f279cd00 761 int err;
24c5684b 762
0472e476
TM		 763 err = xdr_alloc_bvec(xdr, GFP_KERNEL);
764 if (err < 0)
765 return err;
b4b5cc85 766
0472e476
TM		 767 iov_iter_bvec(&msg->msg_iter, WRITE, xdr->bvec,
768 xdr_buf_pagecount(xdr),
769 xdr->page_len + xdr->page_base);
770 return xs_sendmsg(sock, msg, base + xdr->page_base);
b4b5cc85
CL		 771}
772
06b5fc3a
TM		 773#define xs_record_marker_len() sizeof(rpc_fraghdr)
774
067fb11b
CL		 775/* Common case:
776 * - stream transport
777 * - sending from byte 0 of the message
778 * - the message is wholly contained in @xdr's head iovec
779 */
06b5fc3a
TM		 780static int xs_send_rm_and_kvec(struct socket *sock, struct msghdr *msg,
781 rpc_fraghdr marker, struct kvec *vec, size_t base)
067fb11b 782{
067fb11b 783 struct kvec iov[2] = {
06b5fc3a 784 [0] = {
067fb11b
CL		 785 .iov_base = &marker,
786 .iov_len = sizeof(marker)
787 },
06b5fc3a 788 [1] = *vec,
b4b5cc85 789 };
06b5fc3a 790 size_t len = iov[0].iov_len + iov[1].iov_len;
b4b5cc85 791
06b5fc3a
TM		 792 iov_iter_kvec(&msg->msg_iter, WRITE, iov, 2, len);
793 return xs_sendmsg(sock, msg, base);
b4b5cc85
CL		 794}
795
9903cd1c
CL		 796/**
797 * xs_sendpages - write pages directly to a socket
798 * @sock: socket to send on
799 * @addr: UDP only -- address of destination
800 * @addrlen: UDP only -- length of destination address
801 * @xdr: buffer containing this request
802 * @base: starting position in the buffer
06b5fc3a 803 * @rm: stream record marker field
f279cd00 804 * @sent_p: return the total number of bytes successfully queued for sending
9903cd1c 805 *
a246b010 806 */
06b5fc3a 807static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, rpc_fraghdr rm, int *sent_p)
a246b010 808{
e791f8e9
TM		 809 struct msghdr msg = {
810 .msg_name = addr,
811 .msg_namelen = addrlen,
812 .msg_flags = XS_SENDMSG_FLAGS | MSG_MORE,
813 };
06b5fc3a
TM		 814 unsigned int rmsize = rm ? sizeof(rm) : 0;
815 unsigned int remainder = rmsize + xdr->len - base;
816 unsigned int want;
f279cd00 817 int err = 0;
a246b010 818
262965f5 819 if (unlikely(!sock))
fba91afb 820 return -ENOTSOCK;
262965f5 821
06b5fc3a
TM		 822 want = xdr->head[0].iov_len + rmsize;
823 if (base < want) {
824 unsigned int len = want - base;
24c5684b 825 remainder -= len;
e791f8e9
TM		 826 if (remainder == 0)
827 msg.msg_flags &= ~MSG_MORE;
06b5fc3a
TM		 828 if (rmsize)
829 err = xs_send_rm_and_kvec(sock, &msg, rm,
830 &xdr->head[0], base);
067fb11b 831 else
06b5fc3a 832 err = xs_send_kvec(sock, &msg, &xdr->head[0], base);
24c5684b 833 if (remainder == 0 || err != len)
a246b010 834 goto out;
f279cd00 835 *sent_p += err;
a246b010
CL		 836 base = 0;
837 } else
06b5fc3a 838 base -= want;
a246b010 839
24c5684b
TM		 840 if (base < xdr->page_len) {
841 unsigned int len = xdr->page_len - base;
842 remainder -= len;
0472e476
TM		 843 if (remainder == 0)
844 msg.msg_flags &= ~MSG_MORE;
845 err = xs_send_pagedata(sock, &msg, xdr, base);
846 if (remainder == 0 || err != len)
a246b010 847 goto out;
0472e476 848 *sent_p += err;
a246b010 849 base = 0;
24c5684b
TM		 850 } else
851 base -= xdr->page_len;
852
853 if (base >= xdr->tail[0].iov_len)
f279cd00 854 return 0;
e791f8e9
TM		 855 msg.msg_flags &= ~MSG_MORE;
856 err = xs_send_kvec(sock, &msg, &xdr->tail[0], base);
a246b010 857out:
f279cd00
JB		 858 if (err > 0) {
859 *sent_p += err;
860 err = 0;
861 }
862 return err;
a246b010
CL		 863}
864
9903cd1c 865/**
c544577d 866 * xs_nospace - handle transmit was incomplete
50f484e2 867 * @req: pointer to RPC request
9903cd1c 868 *
a246b010 869 */
c544577d 870static int xs_nospace(struct rpc_rqst *req)
a246b010 871{
262965f5 872 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 873 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
06ea0bfe 874 struct sock *sk = transport->inet;
24ca9a84 875 int ret = -EAGAIN;
a246b010 876
46121cf7 877 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
c544577d
TM		 878 req->rq_task->tk_pid,
879 req->rq_slen - transport->xmit.offset,
262965f5
CL		 880 req->rq_slen);
881
b6ddf64f 882 /* Protect against races with write_space */
b5e92419 883 spin_lock(&xprt->transport_lock);
b6ddf64f
TM		 884
885 /* Don't race with disconnect */
886 if (xprt_connected(xprt)) {
13331a55
TM		 887 /* wait for more buffer space */
888 sk->sk_write_pending++;
c544577d 889 xprt_wait_for_buffer_space(xprt);
13331a55 890 } else
5e3771ce 891 ret = -ENOTCONN;
262965f5 892
b5e92419 893 spin_unlock(&xprt->transport_lock);
06ea0bfe
TM		 894
895 /* Race breaker in case memory is freed before above code is called */
d48f9ce7
DV		 896 if (ret == -EAGAIN) {
897 struct socket_wq *wq;
898
899 rcu_read_lock();
900 wq = rcu_dereference(sk->sk_wq);
901 set_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags);
902 rcu_read_unlock();
903
904 sk->sk_write_space(sk);
905 }
5e3771ce 906 return ret;
262965f5
CL		 907}
908
277e4ab7
TM		 909static void
910xs_stream_prepare_request(struct rpc_rqst *req)
911{
75369089 912 xdr_free_bvec(&req->rq_rcv_buf);
12a3ad61 913 req->rq_task->tk_status = xdr_alloc_bvec(&req->rq_rcv_buf, GFP_KERNEL);
277e4ab7
TM		 914}
915
4cd34e7c
TM		 916/*
917 * Determine if the previous message in the stream was aborted before it
918 * could complete transmission.
919 */
920static bool
921xs_send_request_was_aborted(struct sock_xprt *transport, struct rpc_rqst *req)
922{
923 return transport->xmit.offset != 0 && req->rq_bytes_sent == 0;
924}
925
61677eee 926/*
06b5fc3a 927 * Return the stream record marker field for a record of length < 2^31-1
61677eee 928 */
06b5fc3a
TM		 929static rpc_fraghdr
930xs_stream_record_marker(struct xdr_buf *xdr)
61677eee 931{
06b5fc3a
TM		 932 if (!xdr->len)
933 return 0;
934 return cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | (u32)xdr->len);
61677eee
CL		 935}
936
176e21ee
CL		 937/**
938 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
50f484e2 939 * @req: pointer to RPC request
176e21ee
CL		 940 *
941 * Return values:
942 * 0: The request has been sent
943 * EAGAIN: The socket was blocked, please call again later to
944 * complete the request
945 * ENOTCONN: Caller needs to invoke connect logic then call again
946 * other: Some other error occured, the request was not sent
947 */
adfa7144 948static int xs_local_send_request(struct rpc_rqst *req)
176e21ee 949{
176e21ee
CL		 950 struct rpc_xprt *xprt = req->rq_xprt;
951 struct sock_xprt *transport =
952 container_of(xprt, struct sock_xprt, xprt);
953 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 954 rpc_fraghdr rm = xs_stream_record_marker(xdr);
955 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
176e21ee 956 int status;
f279cd00 957 int sent = 0;
176e21ee 958
4cd34e7c
TM		 959 /* Close the stream if the previous transmission was incomplete */
960 if (xs_send_request_was_aborted(transport, req)) {
961 xs_close(xprt);
962 return -ENOTCONN;
963 }
964
176e21ee
CL		 965 xs_pktdump("packet data:",
966 req->rq_svec->iov_base, req->rq_svec->iov_len);
967
78215759 968 req->rq_xtime = ktime_get();
6c7a64e5 969 status = xs_sendpages(transport->sock, NULL, 0, xdr,
7e3d3620 970 transport->xmit.offset, rm, &sent);
176e21ee 971 dprintk("RPC: %s(%u) = %d\n",
6c7a64e5 972 __func__, xdr->len - transport->xmit.offset, status);
743c69e7
N		 973
974 if (status == -EAGAIN && sock_writeable(transport->inet))
975 status = -ENOBUFS;
976
f279cd00 977 if (likely(sent > 0) || status == 0) {
6c7a64e5
TM		 978 transport->xmit.offset += sent;
979 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 980 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 981 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 982 transport->xmit.offset = 0;
176e21ee
CL		 983 return 0;
984 }
985 status = -EAGAIN;
986 }
987
988 switch (status) {
3601c4a9 989 case -ENOBUFS:
b5872f0c 990 break;
176e21ee 991 case -EAGAIN:
c544577d 992 status = xs_nospace(req);
176e21ee
CL		 993 break;
994 default:
995 dprintk("RPC: sendmsg returned unrecognized error %d\n",
996 -status);
e9d47639 997 /* fall through */
176e21ee
CL		 998 case -EPIPE:
999 xs_close(xprt);
1000 status = -ENOTCONN;
1001 }
1002
1003 return status;
1004}
1005
262965f5
CL		 1006/**
1007 * xs_udp_send_request - write an RPC request to a UDP socket
50f484e2 1008 * @req: pointer to RPC request
262965f5
CL		 1009 *
1010 * Return values:
1011 * 0: The request has been sent
1012 * EAGAIN: The socket was blocked, please call again later to
1013 * complete the request
1014 * ENOTCONN: Caller needs to invoke connect logic then call again
25985edc 1015 * other: Some other error occurred, the request was not sent
262965f5 1016 */
adfa7144 1017static int xs_udp_send_request(struct rpc_rqst *req)
262965f5 1018{
262965f5 1019 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 1020 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 1021 struct xdr_buf *xdr = &req->rq_snd_buf;
f279cd00 1022 int sent = 0;
262965f5 1023 int status;
a246b010 1024
9903cd1c 1025 xs_pktdump("packet data:",
a246b010
CL		 1026 req->rq_svec->iov_base,
1027 req->rq_svec->iov_len);
1028
01d37c42
TM		 1029 if (!xprt_bound(xprt))
1030 return -ENOTCONN;
75891f50
TM		 1031
1032 if (!xprt_request_get_cong(xprt, req))
1033 return -EBADSLT;
1034
78215759 1035 req->rq_xtime = ktime_get();
f279cd00 1036 status = xs_sendpages(transport->sock, xs_addr(xprt), xprt->addrlen,
06b5fc3a 1037 xdr, 0, 0, &sent);
a246b010 1038
46121cf7 1039 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
6c7a64e5 1040 xdr->len, status);
a246b010 1041
3dedbb5c
JB		 1042 /* firewall is blocking us, don't return -EAGAIN or we end up looping */
1043 if (status == -EPERM)
1044 goto process_status;
1045
743c69e7
N		 1046 if (status == -EAGAIN && sock_writeable(transport->inet))
1047 status = -ENOBUFS;
1048
f279cd00
JB		 1049 if (sent > 0 || status == 0) {
1050 req->rq_xmit_bytes_sent += sent;
1051 if (sent >= req->rq_slen)
2199700f
TM		 1052 return 0;
1053 /* Still some bytes left; set up for a retry later. */
262965f5 1054 status = -EAGAIN;
2199700f 1055 }
a246b010 1056
3dedbb5c 1057process_status:
262965f5 1058 switch (status) {
fba91afb
TM		 1059 case -ENOTSOCK:
1060 status = -ENOTCONN;
1061 /* Should we call xs_close() here? */
1062 break;
b6ddf64f 1063 case -EAGAIN:
c544577d 1064 status = xs_nospace(req);
b6ddf64f 1065 break;
262965f5 1066 case -ENETUNREACH:
3601c4a9 1067 case -ENOBUFS:
262965f5 1068 case -EPIPE:
a246b010 1069 case -ECONNREFUSED:
3dedbb5c 1070 case -EPERM:
a246b010 1071 /* When the server has died, an ICMP port unreachable message
9903cd1c 1072 * prompts ECONNREFUSED. */
13331a55
TM		 1073 break;
1074 default:
1075 dprintk("RPC: sendmsg returned unrecognized error %d\n",
1076 -status);
a246b010 1077 }
5fe46e9d 1078
262965f5 1079 return status;
a246b010
CL		 1080}
1081
9903cd1c 1082/**
262965f5 1083 * xs_tcp_send_request - write an RPC request to a TCP socket
50f484e2 1084 * @req: pointer to RPC request
9903cd1c
CL		 1085 *
1086 * Return values:
262965f5
CL		 1087 * 0: The request has been sent
1088 * EAGAIN: The socket was blocked, please call again later to
1089 * complete the request
1090 * ENOTCONN: Caller needs to invoke connect logic then call again
25985edc 1091 * other: Some other error occurred, the request was not sent
9903cd1c
CL		 1092 *
1093 * XXX: In the case of soft timeouts, should we eventually give up
262965f5 1094 * if sendmsg is not able to make progress?
9903cd1c 1095 */
adfa7144 1096static int xs_tcp_send_request(struct rpc_rqst *req)
a246b010 1097{
a246b010 1098 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 1099 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 1100 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 1101 rpc_fraghdr rm = xs_stream_record_marker(xdr);
1102 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
9ffadfbc 1103 bool vm_wait = false;
b595bb15 1104 int status;
f279cd00 1105 int sent;
a246b010 1106
4cd34e7c
TM		 1107 /* Close the stream if the previous transmission was incomplete */
1108 if (xs_send_request_was_aborted(transport, req)) {
1109 if (transport->sock != NULL)
1110 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
1111 return -ENOTCONN;
1112 }
1113
262965f5
CL		 1114 xs_pktdump("packet data:",
1115 req->rq_svec->iov_base,
1116 req->rq_svec->iov_len);
a246b010 1117
7196dbb0
TM		 1118 if (test_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state))
1119 xs_tcp_set_socket_timeouts(xprt, transport->sock);
1120
a246b010
CL		 1121 /* Continue transmitting the packet/record. We must be careful
1122 * to cope with writespace callbacks arriving _after_ we have
262965f5 1123 * called sendmsg(). */
78215759 1124 req->rq_xtime = ktime_get();
a246b010 1125 while (1) {
f279cd00
JB		 1126 sent = 0;
1127 status = xs_sendpages(transport->sock, NULL, 0, xdr,
7e3d3620 1128 transport->xmit.offset, rm, &sent);
a246b010 1129
46121cf7 1130 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
6c7a64e5 1131 xdr->len - transport->xmit.offset, status);
a246b010 1132
262965f5
CL		 1133 /* If we've sent the entire packet, immediately
1134 * reset the count of bytes sent. */
6c7a64e5
TM		 1135 transport->xmit.offset += sent;
1136 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 1137 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 1138 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 1139 transport->xmit.offset = 0;
262965f5
CL		 1140 return 0;
1141 }
a246b010 1142
9ffadfbc
TM		 1143 WARN_ON_ONCE(sent == 0 && status == 0);
1144
1145 if (status == -EAGAIN ) {
1146 /*
1147 * Return EAGAIN if we're sure we're hitting the
1148 * socket send buffer limits.
1149 */
1150 if (test_bit(SOCK_NOSPACE, &transport->sock->flags))
1151 break;
1152 /*
1153 * Did we hit a memory allocation failure?
1154 */
1155 if (sent == 0) {
1156 status = -ENOBUFS;
1157 if (vm_wait)
1158 break;
1159 /* Retry, knowing now that we're below the
1160 * socket send buffer limit
1161 */
1162 vm_wait = true;
1163 }
1164 continue;
1165 }
f580dd04
TM		 1166 if (status < 0)
1167 break;
9ffadfbc 1168 vm_wait = false;
a246b010
CL		 1169 }
1170
262965f5 1171 switch (status) {
fba91afb
TM		 1172 case -ENOTSOCK:
1173 status = -ENOTCONN;
1174 /* Should we call xs_close() here? */
1175 break;
262965f5 1176 case -EAGAIN:
c544577d 1177 status = xs_nospace(req);
262965f5 1178 break;
262965f5 1179 case -ECONNRESET:
2a9e1cfa 1180 case -ECONNREFUSED:
262965f5 1181 case -ENOTCONN:
3913c78c 1182 case -EADDRINUSE:
b5872f0c 1183 case -ENOBUFS:
b9d2bb2e 1184 case -EPIPE:
13331a55
TM		 1185 break;
1186 default:
1187 dprintk("RPC: sendmsg returned unrecognized error %d\n",
1188 -status);
a246b010 1189 }
5fe46e9d 1190
a246b010
CL		 1191 return status;
1192}
1193
2a9e1cfa
TM		 1194static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1195{
1196 transport->old_data_ready = sk->sk_data_ready;
1197 transport->old_state_change = sk->sk_state_change;
1198 transport->old_write_space = sk->sk_write_space;
2118071d 1199 transport->old_error_report = sk->sk_error_report;
2a9e1cfa
TM		 1200}
1201
1202static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1203{
1204 sk->sk_data_ready = transport->old_data_ready;
1205 sk->sk_state_change = transport->old_state_change;
1206 sk->sk_write_space = transport->old_write_space;
2118071d
TM		 1207 sk->sk_error_report = transport->old_error_report;
1208}
1209
42d42a5b
TM		 1210static void xs_sock_reset_state_flags(struct rpc_xprt *xprt)
1211{
1212 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1213
1214 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
4f8943f8
TM		 1215 clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state);
1216 clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state);
1217 clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state);
1218}
1219
1220static void xs_run_error_worker(struct sock_xprt *transport, unsigned int nr)
1221{
1222 set_bit(nr, &transport->sock_state);
1223 queue_work(xprtiod_workqueue, &transport->error_worker);
42d42a5b
TM		 1224}
1225
b70ae915
TM		 1226static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
1227{
1228 smp_mb__before_atomic();
1229 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1230 clear_bit(XPRT_CLOSING, &xprt->state);
42d42a5b 1231 xs_sock_reset_state_flags(xprt);
b70ae915
TM		 1232 smp_mb__after_atomic();
1233}
1234
2118071d
TM		 1235/**
1236 * xs_error_report - callback to handle TCP socket state errors
1237 * @sk: socket
1238 *
1239 * Note: we don't call sock_error() since there may be a rpc_task
1240 * using the socket, and so we don't want to clear sk->sk_err.
1241 */
1242static void xs_error_report(struct sock *sk)
1243{
4f8943f8 1244 struct sock_xprt *transport;
2118071d
TM		 1245 struct rpc_xprt *xprt;
1246 int err;
1247
1248 read_lock_bh(&sk->sk_callback_lock);
1249 if (!(xprt = xprt_from_sock(sk)))
1250 goto out;
1251
4f8943f8 1252 transport = container_of(xprt, struct sock_xprt, xprt);
2118071d
TM		 1253 err = -sk->sk_err;
1254 if (err == 0)
1255 goto out;
1256 dprintk("RPC: xs_error_report client %p, error=%d...\n",
1257 xprt, -err);
e8353c76 1258 trace_rpc_socket_error(xprt, sk->sk_socket, err);
4f8943f8 1259 xs_run_error_worker(transport, XPRT_SOCK_WAKE_ERROR);
2118071d
TM		 1260 out:
1261 read_unlock_bh(&sk->sk_callback_lock);
2a9e1cfa
TM		 1262}
1263
fe315e76 1264static void xs_reset_transport(struct sock_xprt *transport)
a246b010 1265{
ee0ac0c2
CL		 1266 struct socket *sock = transport->sock;
1267 struct sock *sk = transport->inet;
6cc7e908 1268 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1269 struct file *filp = transport->file;
a246b010 1270
fe315e76
CL		 1271 if (sk == NULL)
1272 return;
9903cd1c 1273
264d1df3
JL		 1274 if (atomic_read(&transport->xprt.swapper))
1275 sk_clear_memalloc(sk);
1276
09939204
TM		 1277 kernel_sock_shutdown(sock, SHUT_RDWR);
1278
edc1b01c 1279 mutex_lock(&transport->recv_mutex);
a246b010 1280 write_lock_bh(&sk->sk_callback_lock);
ee0ac0c2
CL		 1281 transport->inet = NULL;
1282 transport->sock = NULL;
a73881c9 1283 transport->file = NULL;
a246b010 1284
9903cd1c 1285 sk->sk_user_data = NULL;
2a9e1cfa
TM		 1286
1287 xs_restore_old_callbacks(transport, sk);
0c78789e 1288 xprt_clear_connected(xprt);
a246b010 1289 write_unlock_bh(&sk->sk_callback_lock);
6cc7e908 1290 xs_sock_reset_connection_flags(xprt);
ae053551
TM		 1291 /* Reset stream record info */
1292 xs_stream_reset_connect(transport);
edc1b01c 1293 mutex_unlock(&transport->recv_mutex);
a246b010 1294
6cc7e908 1295 trace_rpc_socket_close(xprt, sock);
a73881c9 1296 fput(filp);
0445f92c
TM		 1297
1298 xprt_disconnect_done(xprt);
fe315e76
CL		 1299}
1300
1301/**
1302 * xs_close - close a socket
1303 * @xprt: transport
1304 *
1305 * This is used when all requests are complete; ie, no DRC state remains
1306 * on the server we want to save.
f75e6745
TM		 1307 *
1308 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
1309 * xs_reset_transport() zeroing the socket from underneath a writer.
fe315e76
CL		 1310 */
1311static void xs_close(struct rpc_xprt *xprt)
1312{
1313 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1314
1315 dprintk("RPC: xs_close xprt %p\n", xprt);
1316
1317 xs_reset_transport(transport);
61d0a8e6 1318 xprt->reestablish_timeout = 0;
a246b010
CL		 1319}
1320
4a068258
CL		 1321static void xs_inject_disconnect(struct rpc_xprt *xprt)
1322{
1323 dprintk("RPC: injecting transport disconnect on xprt=%p\n",
1324 xprt);
1325 xprt_disconnect_done(xprt);
1326}
1327
315f3812
KM		 1328static void xs_xprt_free(struct rpc_xprt *xprt)
1329{
1330 xs_free_peer_addresses(xprt);
1331 xprt_free(xprt);
1332}
1333
9903cd1c
CL		 1334/**
1335 * xs_destroy - prepare to shutdown a transport
1336 * @xprt: doomed transport
1337 *
1338 */
1339static void xs_destroy(struct rpc_xprt *xprt)
a246b010 1340{
03c78827
TM		 1341 struct sock_xprt *transport = container_of(xprt,
1342 struct sock_xprt, xprt);
46121cf7 1343 dprintk("RPC: xs_destroy xprt %p\n", xprt);
9903cd1c 1344
03c78827 1345 cancel_delayed_work_sync(&transport->connect_worker);
a1311d87 1346 xs_close(xprt);
edc1b01c 1347 cancel_work_sync(&transport->recv_worker);
b5e92419 1348 cancel_work_sync(&transport->error_worker);
315f3812 1349 xs_xprt_free(xprt);
a1311d87 1350 module_put(THIS_MODULE);
a246b010
CL		 1351}
1352
9903cd1c 1353/**
f9b2ee71
TM		 1354 * xs_udp_data_read_skb - receive callback for UDP sockets
1355 * @xprt: transport
1356 * @sk: socket
1357 * @skb: skbuff
9903cd1c 1358 *
a246b010 1359 */
f9b2ee71
TM		 1360static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
1361 struct sock *sk,
1362 struct sk_buff *skb)
a246b010 1363{
9903cd1c 1364 struct rpc_task *task;
a246b010 1365 struct rpc_rqst *rovr;
f9b2ee71 1366 int repsize, copied;
d8ed029d
AD		 1367 u32 _xid;
1368 __be32 *xp;
a246b010 1369
1da8c681 1370 repsize = skb->len;
a246b010 1371 if (repsize < 4) {
46121cf7 1372 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
f9b2ee71 1373 return;
a246b010
CL		 1374 }
1375
1376 /* Copy the XID from the skb... */
1da8c681 1377 xp = skb_header_pointer(skb, 0, sizeof(_xid), &_xid);
a246b010 1378 if (xp == NULL)
f9b2ee71 1379 return;
a246b010
CL		 1380
1381 /* Look up and lock the request corresponding to the given XID */
75c84151 1382 spin_lock(&xprt->queue_lock);
a246b010
CL		 1383 rovr = xprt_lookup_rqst(xprt, *xp);
1384 if (!rovr)
1385 goto out_unlock;
729749bb 1386 xprt_pin_rqst(rovr);
ecd465ee 1387 xprt_update_rtt(rovr->rq_task);
75c84151 1388 spin_unlock(&xprt->queue_lock);
a246b010
CL		 1389 task = rovr->rq_task;
1390
a246b010
CL		 1391 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1392 copied = repsize;
1393
1394 /* Suck it into the iovec, verify checksum if not done by hw. */
1781f7f5 1395 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
75c84151 1396 spin_lock(&xprt->queue_lock);
0afa6b44 1397 __UDPX_INC_STATS(sk, UDP_MIB_INERRORS);
729749bb 1398 goto out_unpin;
1781f7f5
HX		 1399 }
1400
a246b010 1401
b5e92419 1402 spin_lock(&xprt->transport_lock);
6a24dfb6 1403 xprt_adjust_cwnd(xprt, task, copied);
b5e92419 1404 spin_unlock(&xprt->transport_lock);
75c84151 1405 spin_lock(&xprt->queue_lock);
1570c1e4 1406 xprt_complete_rqst(task, copied);
0afa6b44 1407 __UDPX_INC_STATS(sk, UDP_MIB_INDATAGRAMS);
729749bb
TM		 1408out_unpin:
1409 xprt_unpin_rqst(rovr);
a246b010 1410 out_unlock:
75c84151 1411 spin_unlock(&xprt->queue_lock);
f9b2ee71
TM		 1412}
1413
1414static void xs_udp_data_receive(struct sock_xprt *transport)
1415{
1416 struct sk_buff *skb;
1417 struct sock *sk;
1418 int err;
1419
1420 mutex_lock(&transport->recv_mutex);
1421 sk = transport->inet;
1422 if (sk == NULL)
1423 goto out;
1424 for (;;) {
7c13f97f 1425 skb = skb_recv_udp(sk, 0, 1, &err);
4f546149 1426 if (skb == NULL)
f9b2ee71 1427 break;
4f546149
TM		 1428 xs_udp_data_read_skb(&transport->xprt, sk, skb);
1429 consume_skb(skb);
1430 cond_resched();
f9b2ee71 1431 }
0ffe86f4 1432 xs_poll_check_readable(transport);
f9b2ee71
TM		 1433out:
1434 mutex_unlock(&transport->recv_mutex);
1435}
1436
1437static void xs_udp_data_receive_workfn(struct work_struct *work)
1438{
1439 struct sock_xprt *transport =
1440 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 1441 unsigned int pflags = memalloc_nofs_save();
1442
f9b2ee71 1443 xs_udp_data_receive(transport);
a1231fda 1444 memalloc_nofs_restore(pflags);
f9b2ee71
TM		 1445}
1446
1447/**
1448 * xs_data_ready - "data ready" callback for UDP sockets
1449 * @sk: socket with data to read
1450 *
1451 */
1452static void xs_data_ready(struct sock *sk)
1453{
1454 struct rpc_xprt *xprt;
1455
1456 read_lock_bh(&sk->sk_callback_lock);
1457 dprintk("RPC: xs_data_ready...\n");
1458 xprt = xprt_from_sock(sk);
1459 if (xprt != NULL) {
1460 struct sock_xprt *transport = container_of(xprt,
1461 struct sock_xprt, xprt);
5157b956
TM		 1462 transport->old_data_ready(sk);
1463 /* Any data means we had a useful conversation, so
1464 * then we don't need to delay the next reconnect
1465 */
1466 if (xprt->reestablish_timeout)
1467 xprt->reestablish_timeout = 0;
42d42a5b 1468 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
40a5f1b1 1469 queue_work(xprtiod_workqueue, &transport->recv_worker);
f9b2ee71 1470 }
f064af1e 1471 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1472}
1473
a519fc7a
TM		 1474/*
1475 * Helper function to force a TCP close if the server is sending
1476 * junk and/or it has put us in CLOSE_WAIT
1477 */
1478static void xs_tcp_force_close(struct rpc_xprt *xprt)
1479{
a519fc7a
TM		 1480 xprt_force_disconnect(xprt);
1481}
1482
9e00abc3 1483#if defined(CONFIG_SUNRPC_BACKCHANNEL)
6b26cc8c
CL		 1484static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt)
1485{
1486 return PAGE_SIZE;
1487}
9e00abc3 1488#endif /* CONFIG_SUNRPC_BACKCHANNEL */
44b98efd 1489
9903cd1c
CL		 1490/**
1491 * xs_tcp_state_change - callback to handle TCP socket state changes
1492 * @sk: socket whose state has changed
1493 *
1494 */
1495static void xs_tcp_state_change(struct sock *sk)
a246b010 1496{
9903cd1c 1497 struct rpc_xprt *xprt;
0fdea1e8 1498 struct sock_xprt *transport;
a246b010 1499
f064af1e 1500 read_lock_bh(&sk->sk_callback_lock);
a246b010
CL		 1501 if (!(xprt = xprt_from_sock(sk)))
1502 goto out;
46121cf7 1503 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
669502ff 1504 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
46121cf7
CL		 1505 sk->sk_state, xprt_connected(xprt),
1506 sock_flag(sk, SOCK_DEAD),
669502ff
AC		 1507 sock_flag(sk, SOCK_ZAPPED),
1508 sk->sk_shutdown);
a246b010 1509
0fdea1e8 1510 transport = container_of(xprt, struct sock_xprt, xprt);
40b5ea0c 1511 trace_rpc_socket_state_change(xprt, sk->sk_socket);
a246b010
CL		 1512 switch (sk->sk_state) {
1513 case TCP_ESTABLISHED:
a246b010 1514 if (!xprt_test_and_set_connected(xprt)) {
8b71798c 1515 xprt->connect_cookie++;
0fdea1e8
TM		 1516 clear_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
1517 xprt_clear_connecting(xprt);
51971139 1518
3968a8a5
CL		 1519 xprt->stat.connect_count++;
1520 xprt->stat.connect_time += (long)jiffies -
1521 xprt->stat.connect_start;
4f8943f8 1522 xs_run_error_worker(transport, XPRT_SOCK_WAKE_PENDING);
a246b010 1523 }
a246b010 1524 break;
3b948ae5
TM		 1525 case TCP_FIN_WAIT1:
1526 /* The client initiated a shutdown of the socket */
7c1d71cf 1527 xprt->connect_cookie++;
663b8858 1528 xprt->reestablish_timeout = 0;
3b948ae5 1529 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1530 smp_mb__before_atomic();
3b948ae5 1531 clear_bit(XPRT_CONNECTED, &xprt->state);
ef803670 1532 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
4e857c58 1533 smp_mb__after_atomic();
a246b010 1534 break;
632e3bdc 1535 case TCP_CLOSE_WAIT:
3b948ae5 1536 /* The server initiated a shutdown of the socket */
7c1d71cf 1537 xprt->connect_cookie++;
d0bea455 1538 clear_bit(XPRT_CONNECTED, &xprt->state);
4f8943f8 1539 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
e9d47639 1540 /* fall through */
663b8858
TM		 1541 case TCP_CLOSING:
1542 /*
1543 * If the server closed down the connection, make sure that
1544 * we back off before reconnecting
1545 */
1546 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1547 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3b948ae5
TM		 1548 break;
1549 case TCP_LAST_ACK:
670f9457 1550 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1551 smp_mb__before_atomic();
3b948ae5 1552 clear_bit(XPRT_CONNECTED, &xprt->state);
4e857c58 1553 smp_mb__after_atomic();
3b948ae5
TM		 1554 break;
1555 case TCP_CLOSE:
0fdea1e8
TM		 1556 if (test_and_clear_bit(XPRT_SOCK_CONNECTING,
1557 &transport->sock_state))
1558 xprt_clear_connecting(xprt);
9b30889c 1559 clear_bit(XPRT_CLOSING, &xprt->state);
9b30889c 1560 /* Trigger the socket release */
4f8943f8 1561 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
a246b010
CL		 1562 }
1563 out:
f064af1e 1564 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1565}
1566
1f0fa154
IJ		 1567static void xs_write_space(struct sock *sk)
1568{
13331a55 1569 struct socket_wq *wq;
4f8943f8 1570 struct sock_xprt *transport;
1f0fa154
IJ		 1571 struct rpc_xprt *xprt;
1572
13331a55 1573 if (!sk->sk_socket)
1f0fa154 1574 return;
13331a55 1575 clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1f0fa154
IJ		 1576
1577 if (unlikely(!(xprt = xprt_from_sock(sk))))
1578 return;
4f8943f8 1579 transport = container_of(xprt, struct sock_xprt, xprt);
13331a55
TM		 1580 rcu_read_lock();
1581 wq = rcu_dereference(sk->sk_wq);
1582 if (!wq || test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags) == 0)
1583 goto out;
1f0fa154 1584
4f8943f8
TM		 1585 xs_run_error_worker(transport, XPRT_SOCK_WAKE_WRITE);
1586 sk->sk_write_pending--;
13331a55
TM		 1587out:
1588 rcu_read_unlock();
1f0fa154
IJ		 1589}
1590
9903cd1c 1591/**
c7b2cae8
CL		 1592 * xs_udp_write_space - callback invoked when socket buffer space
1593 * becomes available
9903cd1c
CL		 1594 * @sk: socket whose state has changed
1595 *
a246b010
CL		 1596 * Called when more output buffer space is available for this socket.
1597 * We try not to wake our writers until they can make "significant"
c7b2cae8 1598 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
a246b010
CL		 1599 * with a bunch of small requests.
1600 */
c7b2cae8 1601static void xs_udp_write_space(struct sock *sk)
a246b010 1602{
f064af1e 1603 read_lock_bh(&sk->sk_callback_lock);
a246b010 1604
c7b2cae8 1605 /* from net/core/sock.c:sock_def_write_space */
1f0fa154
IJ		 1606 if (sock_writeable(sk))
1607 xs_write_space(sk);
a246b010 1608
f064af1e 1609 read_unlock_bh(&sk->sk_callback_lock);
c7b2cae8 1610}
a246b010 1611
c7b2cae8
CL		 1612/**
1613 * xs_tcp_write_space - callback invoked when socket buffer space
1614 * becomes available
1615 * @sk: socket whose state has changed
1616 *
1617 * Called when more output buffer space is available for this socket.
1618 * We try not to wake our writers until they can make "significant"
1619 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1620 * with a bunch of small requests.
1621 */
1622static void xs_tcp_write_space(struct sock *sk)
1623{
f064af1e 1624 read_lock_bh(&sk->sk_callback_lock);
c7b2cae8
CL		 1625
1626 /* from net/core/stream.c:sk_stream_write_space */
64dc6130 1627 if (sk_stream_is_writeable(sk))
1f0fa154 1628 xs_write_space(sk);
b6ddf64f 1629
f064af1e 1630 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1631}
1632
470056c2 1633static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
a246b010 1634{
ee0ac0c2
CL		 1635 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1636 struct sock *sk = transport->inet;
a246b010 1637
7c6e066e 1638 if (transport->rcvsize) {
a246b010 1639 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
7c6e066e 1640 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
a246b010 1641 }
7c6e066e 1642 if (transport->sndsize) {
a246b010 1643 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
7c6e066e 1644 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
a246b010
CL		 1645 sk->sk_write_space(sk);
1646 }
1647}
1648
43118c29 1649/**
470056c2 1650 * xs_udp_set_buffer_size - set send and receive limits
43118c29 1651 * @xprt: generic transport
470056c2
CL		 1652 * @sndsize: requested size of send buffer, in bytes
1653 * @rcvsize: requested size of receive buffer, in bytes
43118c29 1654 *
470056c2 1655 * Set socket send and receive buffer size limits.
43118c29 1656 */
470056c2 1657static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
43118c29 1658{
7c6e066e
CL		 1659 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1660
1661 transport->sndsize = 0;
470056c2 1662 if (sndsize)
7c6e066e
CL		 1663 transport->sndsize = sndsize + 1024;
1664 transport->rcvsize = 0;
470056c2 1665 if (rcvsize)
7c6e066e 1666 transport->rcvsize = rcvsize + 1024;
470056c2
CL		 1667
1668 xs_udp_do_set_buffer_size(xprt);
43118c29
CL		 1669}
1670
46c0ee8b
CL		 1671/**
1672 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
acf0a39f 1673 * @xprt: controlling transport
46c0ee8b
CL		 1674 * @task: task that timed out
1675 *
1676 * Adjust the congestion window after a retransmit timeout has occurred.
1677 */
6a24dfb6 1678static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
46c0ee8b 1679{
b5e92419 1680 spin_lock(&xprt->transport_lock);
6a24dfb6 1681 xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
b5e92419 1682 spin_unlock(&xprt->transport_lock);
46c0ee8b
CL		 1683}
1684
826799e6 1685static int xs_get_random_port(void)
b85d8806 1686{
826799e6
BF		 1687 unsigned short min = xprt_min_resvport, max = xprt_max_resvport;
1688 unsigned short range;
1689 unsigned short rand;
1690
1691 if (max < min)
1692 return -EADDRINUSE;
1693 range = max - min + 1;
1694 rand = (unsigned short) prandom_u32() % range;
1695 return rand + min;
b85d8806
CL		 1696}
1697
4dda9c8a
TM		 1698/**
1699 * xs_set_reuseaddr_port - set the socket's port and address reuse options
1700 * @sock: socket
1701 *
1702 * Note that this function has to be called on all sockets that share the
1703 * same port, and it must be called before binding.
1704 */
1705static void xs_sock_set_reuseport(struct socket *sock)
1706{
402e23b4 1707 int opt = 1;
4dda9c8a 1708
402e23b4
TM		 1709 kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
1710 (char *)&opt, sizeof(opt));
4dda9c8a
TM		 1711}
1712
1713static unsigned short xs_sock_getport(struct socket *sock)
1714{
1715 struct sockaddr_storage buf;
4dda9c8a
TM		 1716 unsigned short port = 0;
1717
9b2c45d4 1718 if (kernel_getsockname(sock, (struct sockaddr *)&buf) < 0)
4dda9c8a
TM		 1719 goto out;
1720 switch (buf.ss_family) {
1721 case AF_INET6:
1722 port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port);
1723 break;
1724 case AF_INET:
1725 port = ntohs(((struct sockaddr_in *)&buf)->sin_port);
1726 }
1727out:
1728 return port;
1729}
1730
92200412
CL		 1731/**
1732 * xs_set_port - reset the port number in the remote endpoint address
1733 * @xprt: generic transport
1734 * @port: new port number
1735 *
1736 */
1737static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1738{
46121cf7 1739 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
c4efcb1d 1740
9dc3b095
CL		 1741 rpc_set_port(xs_addr(xprt), port);
1742 xs_update_peer_port(xprt);
92200412
CL		 1743}
1744
4dda9c8a
TM		 1745static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
1746{
1747 if (transport->srcport == 0)
1748 transport->srcport = xs_sock_getport(sock);
1749}
1750
826799e6 1751static int xs_get_srcport(struct sock_xprt *transport)
67a391d7 1752{
826799e6 1753 int port = transport->srcport;
67a391d7
TM		 1754
1755 if (port == 0 && transport->xprt.resvport)
1756 port = xs_get_random_port();
1757 return port;
1758}
1759
baaf4e48 1760static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
67a391d7 1761{
fbfffbd5
CL		 1762 if (transport->srcport != 0)
1763 transport->srcport = 0;
67a391d7
TM		 1764 if (!transport->xprt.resvport)
1765 return 0;
1766 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1767 return xprt_max_resvport;
1768 return --port;
1769}
beb59b68 1770static int xs_bind(struct sock_xprt *transport, struct socket *sock)
a246b010 1771{
beb59b68 1772 struct sockaddr_storage myaddr;
67a391d7 1773 int err, nloop = 0;
826799e6 1774 int port = xs_get_srcport(transport);
67a391d7 1775 unsigned short last;
a246b010 1776
0f7a622c
CP		 1777 /*
1778 * If we are asking for any ephemeral port (i.e. port == 0 &&
1779 * transport->xprt.resvport == 0), don't bind. Let the local
1780 * port selection happen implicitly when the socket is used
1781 * (for example at connect time).
1782 *
1783 * This ensures that we can continue to establish TCP
1784 * connections even when all local ephemeral ports are already
1785 * a part of some TCP connection. This makes no difference
1786 * for UDP sockets, but also doens't harm them.
1787 *
1788 * If we're asking for any reserved port (i.e. port == 0 &&
1789 * transport->xprt.resvport == 1) xs_get_srcport above will
1790 * ensure that port is non-zero and we will bind as needed.
1791 */
826799e6
BF		 1792 if (port <= 0)
1793 return port;
0f7a622c 1794
beb59b68 1795 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
a246b010 1796 do {
beb59b68
PE		 1797 rpc_set_port((struct sockaddr *)&myaddr, port);
1798 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1799 transport->xprt.addrlen);
a246b010 1800 if (err == 0) {
fbfffbd5 1801 transport->srcport = port;
d3bc9a1d 1802 break;
a246b010 1803 }
67a391d7 1804 last = port;
baaf4e48 1805 port = xs_next_srcport(transport, port);
67a391d7
TM		 1806 if (port > last)
1807 nloop++;
1808 } while (err == -EADDRINUSE && nloop != 2);
90058d37 1809
4232e863 1810 if (myaddr.ss_family == AF_INET)
beb59b68
PE		 1811 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1812 &((struct sockaddr_in *)&myaddr)->sin_addr,
1813 port, err ? "failed" : "ok", err);
1814 else
1815 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1816 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1817 port, err ? "failed" : "ok", err);
a246b010
CL		 1818 return err;
1819}
1820
176e21ee
CL		 1821/*
1822 * We don't support autobind on AF_LOCAL sockets
1823 */
1824static void xs_local_rpcbind(struct rpc_task *task)
1825{
fb43d172 1826 xprt_set_bound(task->tk_xprt);
176e21ee
CL		 1827}
1828
1829static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1830{
1831}
beb59b68 1832
ed07536e
PZ		 1833#ifdef CONFIG_DEBUG_LOCK_ALLOC
1834static struct lock_class_key xs_key[2];
1835static struct lock_class_key xs_slock_key[2];
1836
176e21ee
CL		 1837static inline void xs_reclassify_socketu(struct socket *sock)
1838{
1839 struct sock *sk = sock->sk;
1840
176e21ee
CL		 1841 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1842 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1843}
1844
8945ee5e 1845static inline void xs_reclassify_socket4(struct socket *sock)
ed07536e
PZ		 1846{
1847 struct sock *sk = sock->sk;
8945ee5e 1848
8945ee5e
CL		 1849 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1850 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1851}
ed07536e 1852
8945ee5e
CL		 1853static inline void xs_reclassify_socket6(struct socket *sock)
1854{
1855 struct sock *sk = sock->sk;
ed07536e 1856
8945ee5e
CL		 1857 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1858 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
ed07536e 1859}
6bc9638a
PE		 1860
1861static inline void xs_reclassify_socket(int family, struct socket *sock)
1862{
fafc4e1e 1863 if (WARN_ON_ONCE(!sock_allow_reclassification(sock->sk)))
1b7a1819
WAA		 1864 return;
1865
4232e863 1866 switch (family) {
176e21ee
CL		 1867 case AF_LOCAL:
1868 xs_reclassify_socketu(sock);
1869 break;
4232e863 1870 case AF_INET:
6bc9638a 1871 xs_reclassify_socket4(sock);
4232e863
CL		 1872 break;
1873 case AF_INET6:
6bc9638a 1874 xs_reclassify_socket6(sock);
4232e863
CL		 1875 break;
1876 }
6bc9638a 1877}
ed07536e 1878#else
6bc9638a
PE		 1879static inline void xs_reclassify_socket(int family, struct socket *sock)
1880{
1881}
ed07536e
PZ		 1882#endif
1883
93dc41bd
N		 1884static void xs_dummy_setup_socket(struct work_struct *work)
1885{
1886}
1887
6bc9638a 1888static struct socket *xs_create_sock(struct rpc_xprt *xprt,
4dda9c8a
TM		 1889 struct sock_xprt *transport, int family, int type,
1890 int protocol, bool reuseport)
22f79326 1891{
a73881c9 1892 struct file *filp;
22f79326
PE		 1893 struct socket *sock;
1894 int err;
1895
6bc9638a 1896 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
22f79326
PE		 1897 if (err < 0) {
1898 dprintk("RPC: can't create %d transport socket (%d).\n",
1899 protocol, -err);
1900 goto out;
1901 }
6bc9638a 1902 xs_reclassify_socket(family, sock);
22f79326 1903
4dda9c8a
TM		 1904 if (reuseport)
1905 xs_sock_set_reuseport(sock);
1906
4cea288a
BH		 1907 err = xs_bind(transport, sock);
1908 if (err) {
22f79326
PE		 1909 sock_release(sock);
1910 goto out;
1911 }
1912
a73881c9
TM		 1913 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1914 if (IS_ERR(filp))
1915 return ERR_CAST(filp);
1916 transport->file = filp;
1917
22f79326
PE		 1918 return sock;
1919out:
1920 return ERR_PTR(err);
1921}
1922
176e21ee
CL		 1923static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1924 struct socket *sock)
1925{
1926 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1927 xprt);
1928
1929 if (!transport->inet) {
1930 struct sock *sk = sock->sk;
1931
1932 write_lock_bh(&sk->sk_callback_lock);
1933
1934 xs_save_old_callbacks(transport, sk);
1935
1936 sk->sk_user_data = xprt;
a2648094 1937 sk->sk_data_ready = xs_data_ready;
176e21ee 1938 sk->sk_write_space = xs_udp_write_space;
b4411457 1939 sock_set_flag(sk, SOCK_FASYNC);
2118071d 1940 sk->sk_error_report = xs_error_report;
176e21ee
CL		 1941
1942 xprt_clear_connected(xprt);
1943
1944 /* Reset to new socket */
1945 transport->sock = sock;
1946 transport->inet = sk;
1947
1948 write_unlock_bh(&sk->sk_callback_lock);
1949 }
1950
ae053551 1951 xs_stream_start_connect(transport);
6c7a64e5 1952
176e21ee
CL		 1953 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1954}
1955
1956/**
1957 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
176e21ee 1958 * @transport: socket transport to connect
176e21ee 1959 */
dc107402 1960static int xs_local_setup_socket(struct sock_xprt *transport)
176e21ee 1961{
176e21ee 1962 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1963 struct file *filp;
176e21ee
CL		 1964 struct socket *sock;
1965 int status = -EIO;
1966
176e21ee
CL		 1967 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1968 SOCK_STREAM, 0, &sock, 1);
1969 if (status < 0) {
1970 dprintk("RPC: can't create AF_LOCAL "
1971 "transport socket (%d).\n", -status);
1972 goto out;
1973 }
d1358917 1974 xs_reclassify_socket(AF_LOCAL, sock);
176e21ee 1975
a73881c9
TM		 1976 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1977 if (IS_ERR(filp)) {
1978 status = PTR_ERR(filp);
1979 goto out;
1980 }
1981 transport->file = filp;
1982
176e21ee
CL		 1983 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1984 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1985
1986 status = xs_local_finish_connecting(xprt, sock);
40b5ea0c 1987 trace_rpc_socket_connect(xprt, sock, status);
176e21ee
CL		 1988 switch (status) {
1989 case 0:
1990 dprintk("RPC: xprt %p connected to %s\n",
1991 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
3968a8a5
CL		 1992 xprt->stat.connect_count++;
1993 xprt->stat.connect_time += (long)jiffies -
1994 xprt->stat.connect_start;
176e21ee 1995 xprt_set_connected(xprt);
3601c4a9 1996 case -ENOBUFS:
176e21ee
CL		 1997 break;
1998 case -ENOENT:
1999 dprintk("RPC: xprt %p: socket %s does not exist\n",
2000 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2001 break;
4a20a988
TM		 2002 case -ECONNREFUSED:
2003 dprintk("RPC: xprt %p: connection refused for %s\n",
2004 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2005 break;
176e21ee
CL		 2006 default:
2007 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
2008 __func__, -status,
2009 xprt->address_strings[RPC_DISPLAY_ADDR]);
2010 }
2011
2012out:
2013 xprt_clear_connecting(xprt);
2014 xprt_wake_pending_tasks(xprt, status);
dc107402
BF		 2015 return status;
2016}
2017
b6669737 2018static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
dc107402 2019{
dc107402
BF		 2020 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2021 int ret;
2022
2023 if (RPC_IS_ASYNC(task)) {
2024 /*
2025 * We want the AF_LOCAL connect to be resolved in the
2026 * filesystem namespace of the process making the rpc
2027 * call. Thus we connect synchronously.
2028 *
2029 * If we want to support asynchronous AF_LOCAL calls,
2030 * we'll need to figure out how to pass a namespace to
2031 * connect.
2032 */
5ad64b36 2033 task->tk_rpc_status = -ENOTCONN;
dc107402
BF		 2034 rpc_exit(task, -ENOTCONN);
2035 return;
2036 }
2037 ret = xs_local_setup_socket(transport);
2038 if (ret && !RPC_IS_SOFTCONN(task))
2039 msleep_interruptible(15000);
176e21ee
CL		 2040}
2041
3c87ef6e 2042#if IS_ENABLED(CONFIG_SUNRPC_SWAP)
d6e971d8
JL		 2043/*
2044 * Note that this should be called with XPRT_LOCKED held (or when we otherwise
2045 * know that we have exclusive access to the socket), to guard against
2046 * races with xs_reset_transport.
2047 */
a564b8f0
MG		 2048static void xs_set_memalloc(struct rpc_xprt *xprt)
2049{
2050 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
2051 xprt);
2052
d6e971d8
JL		 2053 /*
2054 * If there's no sock, then we have nothing to set. The
2055 * reconnecting process will get it for us.
2056 */
2057 if (!transport->inet)
2058 return;
8e228133 2059 if (atomic_read(&xprt->swapper))
a564b8f0
MG		 2060 sk_set_memalloc(transport->inet);
2061}
2062
2063/**
d67fa4d8 2064 * xs_enable_swap - Tag this transport as being used for swap.
a564b8f0 2065 * @xprt: transport to tag
a564b8f0 2066 *
8e228133
JL		 2067 * Take a reference to this transport on behalf of the rpc_clnt, and
2068 * optionally mark it for swapping if it wasn't already.
a564b8f0 2069 */
d67fa4d8
JL		 2070static int
2071xs_enable_swap(struct rpc_xprt *xprt)
a564b8f0 2072{
d6e971d8 2073 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
a564b8f0 2074
d6e971d8
JL		 2075 if (atomic_inc_return(&xprt->swapper) != 1)
2076 return 0;
2077 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2078 return -ERESTARTSYS;
2079 if (xs->inet)
2080 sk_set_memalloc(xs->inet);
2081 xprt_release_xprt(xprt, NULL);
8e228133
JL		 2082 return 0;
2083}
a564b8f0 2084
8e228133 2085/**
d67fa4d8 2086 * xs_disable_swap - Untag this transport as being used for swap.
8e228133
JL		 2087 * @xprt: transport to tag
2088 *
2089 * Drop a "swapper" reference to this xprt on behalf of the rpc_clnt. If the
2090 * swapper refcount goes to 0, untag the socket as a memalloc socket.
2091 */
d67fa4d8
JL		 2092static void
2093xs_disable_swap(struct rpc_xprt *xprt)
8e228133 2094{
d6e971d8 2095 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
8e228133 2096
d6e971d8
JL		 2097 if (!atomic_dec_and_test(&xprt->swapper))
2098 return;
2099 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2100 return;
2101 if (xs->inet)
2102 sk_clear_memalloc(xs->inet);
2103 xprt_release_xprt(xprt, NULL);
a564b8f0 2104}
a564b8f0
MG		 2105#else
2106static void xs_set_memalloc(struct rpc_xprt *xprt)
2107{
2108}
d67fa4d8
JL		 2109
2110static int
2111xs_enable_swap(struct rpc_xprt *xprt)
2112{
2113 return -EINVAL;
2114}
2115
2116static void
2117xs_disable_swap(struct rpc_xprt *xprt)
2118{
2119}
a564b8f0
MG		 2120#endif
2121
16be2d20
CL		 2122static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
2123{
2124 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2125
2126 if (!transport->inet) {
2127 struct sock *sk = sock->sk;
2128
2129 write_lock_bh(&sk->sk_callback_lock);
2130
2a9e1cfa
TM		 2131 xs_save_old_callbacks(transport, sk);
2132
16be2d20 2133 sk->sk_user_data = xprt;
f9b2ee71 2134 sk->sk_data_ready = xs_data_ready;
16be2d20 2135 sk->sk_write_space = xs_udp_write_space;
b4411457 2136 sock_set_flag(sk, SOCK_FASYNC);
16be2d20
CL		 2137
2138 xprt_set_connected(xprt);
2139
2140 /* Reset to new socket */
2141 transport->sock = sock;
2142 transport->inet = sk;
2143
a564b8f0
MG		 2144 xs_set_memalloc(xprt);
2145
16be2d20
CL		 2146 write_unlock_bh(&sk->sk_callback_lock);
2147 }
2148 xs_udp_do_set_buffer_size(xprt);
02910177
TM		 2149
2150 xprt->stat.connect_start = jiffies;
16be2d20
CL		 2151}
2152
8c14ff2a 2153static void xs_udp_setup_socket(struct work_struct *work)
a246b010 2154{
34161db6
TM		 2155 struct sock_xprt *transport =
2156 container_of(work, struct sock_xprt, connect_worker.work);
c8475461 2157 struct rpc_xprt *xprt = &transport->xprt;
d099b8af 2158 struct socket *sock;
b65c0310 2159 int status = -EIO;
9903cd1c 2160
8c14ff2a 2161 sock = xs_create_sock(xprt, transport,
4dda9c8a
TM		 2162 xs_addr(xprt)->sa_family, SOCK_DGRAM,
2163 IPPROTO_UDP, false);
b65c0310 2164 if (IS_ERR(sock))
b0d93ad5 2165 goto out;
68e220bd 2166
c740eff8
CL		 2167 dprintk("RPC: worker connecting xprt %p via %s to "
2168 "%s (port %s)\n", xprt,
2169 xprt->address_strings[RPC_DISPLAY_PROTO],
2170 xprt->address_strings[RPC_DISPLAY_ADDR],
2171 xprt->address_strings[RPC_DISPLAY_PORT]);
68e220bd
CL		 2172
2173 xs_udp_finish_connecting(xprt, sock);
40b5ea0c 2174 trace_rpc_socket_connect(xprt, sock, 0);
b0d93ad5
CL		 2175 status = 0;
2176out:
b0d93ad5 2177 xprt_clear_connecting(xprt);
cf76785d 2178 xprt_unlock_connect(xprt, transport);
7d1e8255 2179 xprt_wake_pending_tasks(xprt, status);
a246b010
CL		 2180}
2181
4876cc77
TM		 2182/**
2183 * xs_tcp_shutdown - gracefully shut down a TCP socket
2184 * @xprt: transport
2185 *
2186 * Initiates a graceful shutdown of the TCP socket by calling the
2187 * equivalent of shutdown(SHUT_RDWR);
2188 */
2189static void xs_tcp_shutdown(struct rpc_xprt *xprt)
2190{
2191 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2192 struct socket *sock = transport->sock;
9b30889c 2193 int skst = transport->inet ? transport->inet->sk_state : TCP_CLOSE;
4876cc77
TM		 2194
2195 if (sock == NULL)
2196 return;
9b30889c
TM		 2197 switch (skst) {
2198 default:
4876cc77
TM		 2199 kernel_sock_shutdown(sock, SHUT_RDWR);
2200 trace_rpc_socket_shutdown(xprt, sock);
9b30889c
TM		 2201 break;
2202 case TCP_CLOSE:
2203 case TCP_TIME_WAIT:
4876cc77 2204 xs_reset_transport(transport);
9b30889c 2205 }
4876cc77
TM		 2206}
2207
8d1b8c62
TM		 2208static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
2209 struct socket *sock)
2210{
7196dbb0
TM		 2211 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2212 unsigned int keepidle;
2213 unsigned int keepcnt;
8d1b8c62
TM		 2214 unsigned int opt_on = 1;
2215 unsigned int timeo;
2216
b5e92419 2217 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2218 keepidle = DIV_ROUND_UP(xprt->timeout->to_initval, HZ);
2219 keepcnt = xprt->timeout->to_retries + 1;
2220 timeo = jiffies_to_msecs(xprt->timeout->to_initval) *
2221 (xprt->timeout->to_retries + 1);
2222 clear_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2223 spin_unlock(&xprt->transport_lock);
7196dbb0 2224
8d1b8c62
TM		 2225 /* TCP Keepalive options */
2226 kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
2227 (char *)&opt_on, sizeof(opt_on));
2228 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
2229 (char *)&keepidle, sizeof(keepidle));
2230 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
2231 (char *)&keepidle, sizeof(keepidle));
2232 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
2233 (char *)&keepcnt, sizeof(keepcnt));
2234
2235 /* TCP user timeout (see RFC5482) */
8d1b8c62
TM		 2236 kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
2237 (char *)&timeo, sizeof(timeo));
2238}
2239
7196dbb0
TM		 2240static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
2241 unsigned long connect_timeout,
2242 unsigned long reconnect_timeout)
2243{
2244 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2245 struct rpc_timeout to;
2246 unsigned long initval;
2247
b5e92419 2248 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2249 if (reconnect_timeout < xprt->max_reconnect_timeout)
2250 xprt->max_reconnect_timeout = reconnect_timeout;
2251 if (connect_timeout < xprt->connect_timeout) {
2252 memcpy(&to, xprt->timeout, sizeof(to));
2253 initval = DIV_ROUND_UP(connect_timeout, to.to_retries + 1);
2254 /* Arbitrary lower limit */
2255 if (initval < XS_TCP_INIT_REEST_TO << 1)
2256 initval = XS_TCP_INIT_REEST_TO << 1;
2257 to.to_initval = initval;
2258 to.to_maxval = initval;
2259 memcpy(&transport->tcp_timeout, &to,
2260 sizeof(transport->tcp_timeout));
2261 xprt->timeout = &transport->tcp_timeout;
2262 xprt->connect_timeout = connect_timeout;
2263 }
2264 set_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2265 spin_unlock(&xprt->transport_lock);
7196dbb0
TM		 2266}
2267
16be2d20 2268static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
a246b010 2269{
16be2d20 2270 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
fe19a96b 2271 int ret = -ENOTCONN;
edb267a6 2272
ee0ac0c2 2273 if (!transport->inet) {
b0d93ad5 2274 struct sock *sk = sock->sk;
d88e4d82 2275 unsigned int addr_pref = IPV6_PREFER_SRC_PUBLIC;
7f260e85 2276
d88e4d82
N		 2277 /* Avoid temporary address, they are bad for long-lived
2278 * connections such as NFS mounts.
2279 * RFC4941, section 3.6 suggests that:
2280 * Individual applications, which have specific
2281 * knowledge about the normal duration of connections,
2282 * MAY override this as appropriate.
2283 */
2284 kernel_setsockopt(sock, SOL_IPV6, IPV6_ADDR_PREFERENCES,
2285 (char *)&addr_pref, sizeof(addr_pref));
2286
8d1b8c62 2287 xs_tcp_set_socket_timeouts(xprt, sock);
775f06ab 2288
b0d93ad5
CL		 2289 write_lock_bh(&sk->sk_callback_lock);
2290
2a9e1cfa
TM		 2291 xs_save_old_callbacks(transport, sk);
2292
b0d93ad5 2293 sk->sk_user_data = xprt;
5157b956 2294 sk->sk_data_ready = xs_data_ready;
b0d93ad5
CL		 2295 sk->sk_state_change = xs_tcp_state_change;
2296 sk->sk_write_space = xs_tcp_write_space;
b4411457 2297 sock_set_flag(sk, SOCK_FASYNC);
2118071d 2298 sk->sk_error_report = xs_error_report;
3167e12c
CL		 2299
2300 /* socket options */
3167e12c 2301 sock_reset_flag(sk, SOCK_LINGER);
3167e12c 2302 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
b0d93ad5
CL		 2303
2304 xprt_clear_connected(xprt);
2305
2306 /* Reset to new socket */
ee0ac0c2
CL		 2307 transport->sock = sock;
2308 transport->inet = sk;
b0d93ad5
CL		 2309
2310 write_unlock_bh(&sk->sk_callback_lock);
2311 }
2312
01d37c42 2313 if (!xprt_bound(xprt))
fe19a96b 2314 goto out;
01d37c42 2315
a564b8f0
MG		 2316 xs_set_memalloc(xprt);
2317
ae053551 2318 xs_stream_start_connect(transport);
e1806c7b 2319
b0d93ad5 2320 /* Tell the socket layer to start connecting... */
0fdea1e8 2321 set_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
fe19a96b
TM		 2322 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2323 switch (ret) {
2324 case 0:
4dda9c8a 2325 xs_set_srcport(transport, sock);
e9d47639 2326 /* fall through */
fe19a96b
TM		 2327 case -EINPROGRESS:
2328 /* SYN_SENT! */
fe19a96b
TM		 2329 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2330 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
1f4c17a0
TM		 2331 break;
2332 case -EADDRNOTAVAIL:
2333 /* Source port number is unavailable. Try a new one! */
2334 transport->srcport = 0;
fe19a96b
TM		 2335 }
2336out:
2337 return ret;
16be2d20
CL		 2338}
2339
9903cd1c 2340/**
b61d59ff 2341 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
acf0a39f 2342 * @work: queued work item
9903cd1c
CL		 2343 *
2344 * Invoked by a work queue tasklet.
a246b010 2345 */
cdd518d5 2346static void xs_tcp_setup_socket(struct work_struct *work)
a246b010 2347{
cdd518d5
PE		 2348 struct sock_xprt *transport =
2349 container_of(work, struct sock_xprt, connect_worker.work);
ee0ac0c2 2350 struct socket *sock = transport->sock;
a9f5f0f7 2351 struct rpc_xprt *xprt = &transport->xprt;
b61d59ff 2352 int status = -EIO;
a246b010 2353
ee0ac0c2 2354 if (!sock) {
cdd518d5 2355 sock = xs_create_sock(xprt, transport,
4dda9c8a
TM		 2356 xs_addr(xprt)->sa_family, SOCK_STREAM,
2357 IPPROTO_TCP, true);
b61d59ff
TM		 2358 if (IS_ERR(sock)) {
2359 status = PTR_ERR(sock);
3167e12c
CL		 2360 goto out;
2361 }
7d1e8255 2362 }
a246b010 2363
c740eff8
CL		 2364 dprintk("RPC: worker connecting xprt %p via %s to "
2365 "%s (port %s)\n", xprt,
2366 xprt->address_strings[RPC_DISPLAY_PROTO],
2367 xprt->address_strings[RPC_DISPLAY_ADDR],
2368 xprt->address_strings[RPC_DISPLAY_PORT]);
edb267a6 2369
16be2d20 2370 status = xs_tcp_finish_connecting(xprt, sock);
40b5ea0c 2371 trace_rpc_socket_connect(xprt, sock, status);
46121cf7
CL		 2372 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2373 xprt, -status, xprt_connected(xprt),
2374 sock->sk->sk_state);
2a491991 2375 switch (status) {
f75e6745
TM		 2376 default:
2377 printk("%s: connect returned unhandled error %d\n",
2378 __func__, status);
e9d47639 2379 /* fall through */
f75e6745
TM		 2380 case -EADDRNOTAVAIL:
2381 /* We're probably in TIME_WAIT. Get rid of existing socket,
2382 * and retry
2383 */
a519fc7a 2384 xs_tcp_force_close(xprt);
88b5ed73 2385 break;
2a491991
TM		 2386 case 0:
2387 case -EINPROGRESS:
2388 case -EALREADY:
718ba5b8 2389 xprt_unlock_connect(xprt, transport);
7d1e8255 2390 return;
9fcfe0c8
TM		 2391 case -EINVAL:
2392 /* Happens, for instance, if the user specified a link
2393 * local IPv6 address without a scope-id.
2394 */
3ed5e2a2
TM		 2395 case -ECONNREFUSED:
2396 case -ECONNRESET:
eb5b46fa 2397 case -ENETDOWN:
3ed5e2a2 2398 case -ENETUNREACH:
4ba161a7 2399 case -EHOSTUNREACH:
3913c78c 2400 case -EADDRINUSE:
3601c4a9 2401 case -ENOBUFS:
6ea44adc
N		 2402 /*
2403 * xs_tcp_force_close() wakes tasks with -EIO.
2404 * We need to wake them first to ensure the
2405 * correct error code.
2406 */
2407 xprt_wake_pending_tasks(xprt, status);
4efdd92c 2408 xs_tcp_force_close(xprt);
9fcfe0c8 2409 goto out;
a246b010 2410 }
2a491991 2411 status = -EAGAIN;
a246b010 2412out:
68e220bd 2413 xprt_clear_connecting(xprt);
cf76785d 2414 xprt_unlock_connect(xprt, transport);
7d1e8255 2415 xprt_wake_pending_tasks(xprt, status);
68e220bd 2416}
b0d93ad5 2417
9903cd1c
CL		 2418/**
2419 * xs_connect - connect a socket to a remote endpoint
1b092092 2420 * @xprt: pointer to transport structure
9903cd1c
CL		 2421 * @task: address of RPC task that manages state of connect request
2422 *
2423 * TCP: If the remote end dropped the connection, delay reconnecting.
03bf4b70
CL		 2424 *
2425 * UDP socket connects are synchronous, but we use a work queue anyway
2426 * to guarantee that even unprivileged user processes can set up a
2427 * socket on a privileged port.
2428 *
2429 * If a UDP socket connect fails, the delay behavior here prevents
2430 * retry floods (hard mounts).
9903cd1c 2431 */
1b092092 2432static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
a246b010 2433{
ee0ac0c2 2434 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
02910177 2435 unsigned long delay = 0;
a246b010 2436
718ba5b8
TM		 2437 WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport));
2438
99b1a4c3 2439 if (transport->sock != NULL) {
46121cf7
CL		 2440 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2441 "seconds\n",
03bf4b70 2442 xprt, xprt->reestablish_timeout / HZ);
99b1a4c3
TM		 2443
2444 /* Start by resetting any existing state */
2445 xs_reset_transport(transport);
2446
675dd90a
CL		 2447 delay = xprt_reconnect_delay(xprt);
2448 xprt_reconnect_backoff(xprt, XS_TCP_INIT_REEST_TO);
02910177 2449
02910177 2450 } else
46121cf7 2451 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
02910177
TM		 2452
2453 queue_delayed_work(xprtiod_workqueue,
2454 &transport->connect_worker,
2455 delay);
a246b010
CL		 2456}
2457
4f8943f8
TM		 2458static void xs_wake_disconnect(struct sock_xprt *transport)
2459{
2460 if (test_and_clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state))
2461 xs_tcp_force_close(&transport->xprt);
2462}
2463
2464static void xs_wake_write(struct sock_xprt *transport)
2465{
2466 if (test_and_clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state))
2467 xprt_write_space(&transport->xprt);
2468}
2469
2470static void xs_wake_error(struct sock_xprt *transport)
2471{
2472 int sockerr;
2473 int sockerr_len = sizeof(sockerr);
2474
2475 if (!test_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2476 return;
2477 mutex_lock(&transport->recv_mutex);
2478 if (transport->sock == NULL)
2479 goto out;
2480 if (!test_and_clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2481 goto out;
2482 if (kernel_getsockopt(transport->sock, SOL_SOCKET, SO_ERROR,
2483 (char *)&sockerr, &sockerr_len) != 0)
2484 goto out;
2485 if (sockerr < 0)
2486 xprt_wake_pending_tasks(&transport->xprt, sockerr);
2487out:
2488 mutex_unlock(&transport->recv_mutex);
2489}
2490
2491static void xs_wake_pending(struct sock_xprt *transport)
2492{
2493 if (test_and_clear_bit(XPRT_SOCK_WAKE_PENDING, &transport->sock_state))
2494 xprt_wake_pending_tasks(&transport->xprt, -EAGAIN);
2495}
2496
2497static void xs_error_handle(struct work_struct *work)
2498{
2499 struct sock_xprt *transport = container_of(work,
2500 struct sock_xprt, error_worker);
2501
2502 xs_wake_disconnect(transport);
2503 xs_wake_write(transport);
2504 xs_wake_error(transport);
2505 xs_wake_pending(transport);
2506}
2507
176e21ee
CL		 2508/**
2509 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2510 * @xprt: rpc_xprt struct containing statistics
2511 * @seq: output file
2512 *
2513 */
2514static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2515{
2516 long idle_time = 0;
2517
2518 if (xprt_connected(xprt))
2519 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2520
2521 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
15a45206 2522 "%llu %llu %lu %llu %llu\n",
176e21ee
CL		 2523 xprt->stat.bind_count,
2524 xprt->stat.connect_count,
8440a886 2525 xprt->stat.connect_time / HZ,
176e21ee
CL		 2526 idle_time,
2527 xprt->stat.sends,
2528 xprt->stat.recvs,
2529 xprt->stat.bad_xids,
2530 xprt->stat.req_u,
15a45206
AA		 2531 xprt->stat.bklog_u,
2532 xprt->stat.max_slots,
2533 xprt->stat.sending_u,
2534 xprt->stat.pending_u);
176e21ee
CL		 2535}
2536
262ca07d
CL		 2537/**
2538 * xs_udp_print_stats - display UDP socket-specifc stats
2539 * @xprt: rpc_xprt struct containing statistics
2540 * @seq: output file
2541 *
2542 */
2543static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2544{
c8475461
CL		 2545 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2546
15a45206
AA		 2547 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2548 "%lu %llu %llu\n",
fbfffbd5 2549 transport->srcport,
262ca07d
CL		 2550 xprt->stat.bind_count,
2551 xprt->stat.sends,
2552 xprt->stat.recvs,
2553 xprt->stat.bad_xids,
2554 xprt->stat.req_u,
15a45206
AA		 2555 xprt->stat.bklog_u,
2556 xprt->stat.max_slots,
2557 xprt->stat.sending_u,
2558 xprt->stat.pending_u);
262ca07d
CL		 2559}
2560
2561/**
2562 * xs_tcp_print_stats - display TCP socket-specifc stats
2563 * @xprt: rpc_xprt struct containing statistics
2564 * @seq: output file
2565 *
2566 */
2567static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2568{
c8475461 2569 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262ca07d
CL		 2570 long idle_time = 0;
2571
2572 if (xprt_connected(xprt))
2573 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2574
15a45206
AA		 2575 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2576 "%llu %llu %lu %llu %llu\n",
fbfffbd5 2577 transport->srcport,
262ca07d
CL		 2578 xprt->stat.bind_count,
2579 xprt->stat.connect_count,
8440a886 2580 xprt->stat.connect_time / HZ,
262ca07d
CL		 2581 idle_time,
2582 xprt->stat.sends,
2583 xprt->stat.recvs,
2584 xprt->stat.bad_xids,
2585 xprt->stat.req_u,
15a45206
AA		 2586 xprt->stat.bklog_u,
2587 xprt->stat.max_slots,
2588 xprt->stat.sending_u,
2589 xprt->stat.pending_u);
262ca07d
CL		 2590}
2591
4cfc7e60
RI		 2592/*
2593 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2594 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2595 * to use the server side send routines.
2596 */
5fe6eaa1 2597static int bc_malloc(struct rpc_task *task)
4cfc7e60 2598{
5fe6eaa1
CL		 2599 struct rpc_rqst *rqst = task->tk_rqstp;
2600 size_t size = rqst->rq_callsize;
4cfc7e60
RI		 2601 struct page *page;
2602 struct rpc_buffer *buf;
2603
5fe6eaa1
CL		 2604 if (size > PAGE_SIZE - sizeof(struct rpc_buffer)) {
2605 WARN_ONCE(1, "xprtsock: large bc buffer request (size %zu)\n",
2606 size);
2607 return -EINVAL;
2608 }
4cfc7e60 2609
b8a13d03 2610 page = alloc_page(GFP_KERNEL);
4cfc7e60 2611 if (!page)
5fe6eaa1 2612 return -ENOMEM;
4cfc7e60
RI		 2613
2614 buf = page_address(page);
2615 buf->len = PAGE_SIZE;
2616
5fe6eaa1 2617 rqst->rq_buffer = buf->data;
18e601d6 2618 rqst->rq_rbuffer = (char *)rqst->rq_buffer + rqst->rq_callsize;
5fe6eaa1 2619 return 0;
4cfc7e60
RI		 2620}
2621
2622/*
2623 * Free the space allocated in the bc_alloc routine
2624 */
3435c74a 2625static void bc_free(struct rpc_task *task)
4cfc7e60 2626{
3435c74a 2627 void *buffer = task->tk_rqstp->rq_buffer;
4cfc7e60
RI		 2628 struct rpc_buffer *buf;
2629
4cfc7e60
RI		 2630 buf = container_of(buffer, struct rpc_buffer, data);
2631 free_page((unsigned long)buf);
2632}
2633
2634/*
2635 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2636 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2637 */
2638static int bc_sendto(struct rpc_rqst *req)
2639{
2640 int len;
2641 struct xdr_buf *xbufp = &req->rq_snd_buf;
4cfc7e60 2642 struct sock_xprt *transport =
067fb11b 2643 container_of(req->rq_xprt, struct sock_xprt, xprt);
4cfc7e60
RI		 2644 unsigned long headoff;
2645 unsigned long tailoff;
067fb11b
CL		 2646 struct page *tailpage;
2647 struct msghdr msg = {
2648 .msg_flags = MSG_MORE
2649 };
2650 rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT |
2651 (u32)xbufp->len);
2652 struct kvec iov = {
2653 .iov_base = &marker,
2654 .iov_len = sizeof(marker),
2655 };
4cfc7e60 2656
067fb11b
CL		 2657 len = kernel_sendmsg(transport->sock, &msg, &iov, 1, iov.iov_len);
2658 if (len != iov.iov_len)
2659 return -EAGAIN;
4cfc7e60 2660
067fb11b
CL		 2661 tailpage = NULL;
2662 if (xbufp->tail[0].iov_len)
2663 tailpage = virt_to_page(xbufp->tail[0].iov_base);
4cfc7e60
RI		 2664 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2665 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
067fb11b 2666 len = svc_send_common(transport->sock, xbufp,
4cfc7e60 2667 virt_to_page(xbufp->head[0].iov_base), headoff,
067fb11b
CL		 2668 tailpage, tailoff);
2669 if (len != xbufp->len)
2670 return -EAGAIN;
4cfc7e60
RI		 2671 return len;
2672}
2673
2674/*
2675 * The send routine. Borrows from svc_send
2676 */
adfa7144 2677static int bc_send_request(struct rpc_rqst *req)
4cfc7e60 2678{
4cfc7e60 2679 struct svc_xprt *xprt;
7fc56136 2680 int len;
4cfc7e60
RI		 2681
2682 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2683 /*
2684 * Get the server socket associated with this callback xprt
2685 */
2686 xprt = req->rq_xprt->bc_xprt;
4cfc7e60
RI		 2687
2688 /*
2689 * Grab the mutex to serialize data as the connection is shared
2690 * with the fore channel
2691 */
c544577d 2692 mutex_lock(&xprt->xpt_mutex);
4cfc7e60
RI		 2693 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2694 len = -ENOTCONN;
2695 else
2696 len = bc_sendto(req);
2697 mutex_unlock(&xprt->xpt_mutex);
2698
2699 if (len > 0)
2700 len = 0;
2701
2702 return len;
2703}
2704
2705/*
2706 * The close routine. Since this is client initiated, we do nothing
2707 */
2708
2709static void bc_close(struct rpc_xprt *xprt)
2710{
4cfc7e60
RI		 2711}
2712
2713/*
2714 * The xprt destroy routine. Again, because this connection is client
2715 * initiated, we do nothing
2716 */
2717
2718static void bc_destroy(struct rpc_xprt *xprt)
2719{
47f72efa
KM		 2720 dprintk("RPC: bc_destroy xprt %p\n", xprt);
2721
2722 xs_xprt_free(xprt);
2723 module_put(THIS_MODULE);
4cfc7e60
RI		 2724}
2725
d31ae254 2726static const struct rpc_xprt_ops xs_local_ops = {
176e21ee 2727 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2728 .release_xprt = xprt_release_xprt,
f39c1bfb 2729 .alloc_slot = xprt_alloc_slot,
a9cde23a 2730 .free_slot = xprt_free_slot,
176e21ee
CL		 2731 .rpcbind = xs_local_rpcbind,
2732 .set_port = xs_local_set_port,
dc107402 2733 .connect = xs_local_connect,
176e21ee
CL		 2734 .buf_alloc = rpc_malloc,
2735 .buf_free = rpc_free,
550aebfe 2736 .prepare_request = xs_stream_prepare_request,
176e21ee 2737 .send_request = xs_local_send_request,
8ba6a92d 2738 .wait_for_reply_request = xprt_wait_for_reply_request_def,
176e21ee 2739 .close = xs_close,
a1311d87 2740 .destroy = xs_destroy,
176e21ee 2741 .print_stats = xs_local_print_stats,
d67fa4d8
JL		 2742 .enable_swap = xs_enable_swap,
2743 .disable_swap = xs_disable_swap,
176e21ee
CL		 2744};
2745
d31ae254 2746static const struct rpc_xprt_ops xs_udp_ops = {
43118c29 2747 .set_buffer_size = xs_udp_set_buffer_size,
12a80469 2748 .reserve_xprt = xprt_reserve_xprt_cong,
49e9a890 2749 .release_xprt = xprt_release_xprt_cong,
f39c1bfb 2750 .alloc_slot = xprt_alloc_slot,
a9cde23a 2751 .free_slot = xprt_free_slot,
45160d62 2752 .rpcbind = rpcb_getport_async,
92200412 2753 .set_port = xs_set_port,
262965f5 2754 .connect = xs_connect,
02107148
CL		 2755 .buf_alloc = rpc_malloc,
2756 .buf_free = rpc_free,
262965f5 2757 .send_request = xs_udp_send_request,
8ba6a92d 2758 .wait_for_reply_request = xprt_wait_for_reply_request_rtt,
46c0ee8b 2759 .timer = xs_udp_timer,
a58dd398 2760 .release_request = xprt_release_rqst_cong,
262965f5
CL		 2761 .close = xs_close,
2762 .destroy = xs_destroy,
262ca07d 2763 .print_stats = xs_udp_print_stats,
d67fa4d8
JL		 2764 .enable_swap = xs_enable_swap,
2765 .disable_swap = xs_disable_swap,
4a068258 2766 .inject_disconnect = xs_inject_disconnect,
262965f5
CL		 2767};
2768
d31ae254 2769static const struct rpc_xprt_ops xs_tcp_ops = {
12a80469 2770 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2771 .release_xprt = xprt_release_xprt,
36bd7de9 2772 .alloc_slot = xprt_alloc_slot,
a9cde23a 2773 .free_slot = xprt_free_slot,
45160d62 2774 .rpcbind = rpcb_getport_async,
92200412 2775 .set_port = xs_set_port,
0b9e7943 2776 .connect = xs_connect,
02107148
CL		 2777 .buf_alloc = rpc_malloc,
2778 .buf_free = rpc_free,
277e4ab7 2779 .prepare_request = xs_stream_prepare_request,
262965f5 2780 .send_request = xs_tcp_send_request,
8ba6a92d 2781 .wait_for_reply_request = xprt_wait_for_reply_request_def,
c627d31b 2782 .close = xs_tcp_shutdown,
9903cd1c 2783 .destroy = xs_destroy,
7196dbb0 2784 .set_connect_timeout = xs_tcp_set_connect_timeout,
262ca07d 2785 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2786 .enable_swap = xs_enable_swap,
2787 .disable_swap = xs_disable_swap,
4a068258 2788 .inject_disconnect = xs_inject_disconnect,
42e5c3e2
CL		 2789#ifdef CONFIG_SUNRPC_BACKCHANNEL
2790 .bc_setup = xprt_setup_bc,
6b26cc8c 2791 .bc_maxpayload = xs_tcp_bc_maxpayload,
7402a4fe 2792 .bc_num_slots = xprt_bc_max_slots,
42e5c3e2
CL		 2793 .bc_free_rqst = xprt_free_bc_rqst,
2794 .bc_destroy = xprt_destroy_bc,
2795#endif
a246b010
CL		 2796};
2797
4cfc7e60
RI		 2798/*
2799 * The rpc_xprt_ops for the server backchannel
2800 */
2801
d31ae254 2802static const struct rpc_xprt_ops bc_tcp_ops = {
4cfc7e60
RI		 2803 .reserve_xprt = xprt_reserve_xprt,
2804 .release_xprt = xprt_release_xprt,
84e28a30 2805 .alloc_slot = xprt_alloc_slot,
a9cde23a 2806 .free_slot = xprt_free_slot,
4cfc7e60
RI		 2807 .buf_alloc = bc_malloc,
2808 .buf_free = bc_free,
2809 .send_request = bc_send_request,
8ba6a92d 2810 .wait_for_reply_request = xprt_wait_for_reply_request_def,
4cfc7e60
RI		 2811 .close = bc_close,
2812 .destroy = bc_destroy,
2813 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2814 .enable_swap = xs_enable_swap,
2815 .disable_swap = xs_disable_swap,
4a068258 2816 .inject_disconnect = xs_inject_disconnect,
4cfc7e60
RI		 2817};
2818
92476850
CL		 2819static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2820{
2821 static const struct sockaddr_in sin = {
2822 .sin_family = AF_INET,
2823 .sin_addr.s_addr = htonl(INADDR_ANY),
2824 };
2825 static const struct sockaddr_in6 sin6 = {
2826 .sin6_family = AF_INET6,
2827 .sin6_addr = IN6ADDR_ANY_INIT,
2828 };
2829
2830 switch (family) {
176e21ee
CL		 2831 case AF_LOCAL:
2832 break;
92476850
CL		 2833 case AF_INET:
2834 memcpy(sap, &sin, sizeof(sin));
2835 break;
2836 case AF_INET6:
2837 memcpy(sap, &sin6, sizeof(sin6));
2838 break;
2839 default:
2840 dprintk("RPC: %s: Bad address family\n", __func__);
2841 return -EAFNOSUPPORT;
2842 }
2843 return 0;
2844}
2845
3c341b0b 2846static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
d9ba131d
TM		 2847 unsigned int slot_table_size,
2848 unsigned int max_slot_table_size)
c8541ecd
CL		 2849{
2850 struct rpc_xprt *xprt;
ffc2e518 2851 struct sock_xprt *new;
c8541ecd 2852
96802a09 2853 if (args->addrlen > sizeof(xprt->addr)) {
46121cf7 2854 dprintk("RPC: xs_setup_xprt: address too large\n");
c8541ecd
CL		 2855 return ERR_PTR(-EBADF);
2856 }
2857
d9ba131d
TM		 2858 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2859 max_slot_table_size);
bd1722d4 2860 if (xprt == NULL) {
46121cf7
CL		 2861 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2862 "rpc_xprt\n");
c8541ecd
CL		 2863 return ERR_PTR(-ENOMEM);
2864 }
c8541ecd 2865
bd1722d4 2866 new = container_of(xprt, struct sock_xprt, xprt);
edc1b01c 2867 mutex_init(&new->recv_mutex);
96802a09
FM		 2868 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2869 xprt->addrlen = args->addrlen;
d3bc9a1d 2870 if (args->srcaddr)
fbfffbd5 2871 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
92476850
CL		 2872 else {
2873 int err;
2874 err = xs_init_anyaddr(args->dstaddr->sa_family,
2875 (struct sockaddr *)&new->srcaddr);
2aa13531
SK		 2876 if (err != 0) {
2877 xprt_free(xprt);
92476850 2878 return ERR_PTR(err);
2aa13531 2879 }
92476850 2880 }
c8541ecd
CL		 2881
2882 return xprt;
2883}
2884
176e21ee
CL		 2885static const struct rpc_timeout xs_local_default_timeout = {
2886 .to_initval = 10 * HZ,
2887 .to_maxval = 10 * HZ,
2888 .to_retries = 2,
2889};
2890
2891/**
2892 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2893 * @args: rpc transport creation arguments
2894 *
2895 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2896 */
2897static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2898{
2899 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2900 struct sock_xprt *transport;
2901 struct rpc_xprt *xprt;
2902 struct rpc_xprt *ret;
2903
d9ba131d
TM		 2904 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2905 xprt_max_tcp_slot_table_entries);
176e21ee
CL		 2906 if (IS_ERR(xprt))
2907 return xprt;
2908 transport = container_of(xprt, struct sock_xprt, xprt);
2909
2910 xprt->prot = 0;
176e21ee
CL		 2911 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2912
2913 xprt->bind_timeout = XS_BIND_TO;
2914 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2915 xprt->idle_timeout = XS_IDLE_DISC_TO;
2916
2917 xprt->ops = &xs_local_ops;
2918 xprt->timeout = &xs_local_default_timeout;
2919
550aebfe 2920 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 2921 INIT_WORK(&transport->error_worker, xs_error_handle);
550aebfe 2922 INIT_DELAYED_WORK(&transport->connect_worker, xs_dummy_setup_socket);
93dc41bd 2923
176e21ee
CL		 2924 switch (sun->sun_family) {
2925 case AF_LOCAL:
2926 if (sun->sun_path[0] != '/') {
2927 dprintk("RPC: bad AF_LOCAL address: %s\n",
2928 sun->sun_path);
2929 ret = ERR_PTR(-EINVAL);
2930 goto out_err;
2931 }
2932 xprt_set_bound(xprt);
176e21ee 2933 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
7073ea87
BF		 2934 ret = ERR_PTR(xs_local_setup_socket(transport));
2935 if (ret)
2936 goto out_err;
176e21ee
CL		 2937 break;
2938 default:
2939 ret = ERR_PTR(-EAFNOSUPPORT);
2940 goto out_err;
2941 }
2942
2943 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2944 xprt->address_strings[RPC_DISPLAY_ADDR]);
2945
2946 if (try_module_get(THIS_MODULE))
2947 return xprt;
2948 ret = ERR_PTR(-EINVAL);
2949out_err:
315f3812 2950 xs_xprt_free(xprt);
176e21ee
CL		 2951 return ret;
2952}
2953
2881ae74
TM		 2954static const struct rpc_timeout xs_udp_default_timeout = {
2955 .to_initval = 5 * HZ,
2956 .to_maxval = 30 * HZ,
2957 .to_increment = 5 * HZ,
2958 .to_retries = 5,
2959};
2960
9903cd1c
CL		 2961/**
2962 * xs_setup_udp - Set up transport to use a UDP socket
96802a09 2963 * @args: rpc transport creation arguments
9903cd1c
CL		 2964 *
2965 */
483066d6 2966static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
a246b010 2967{
8f9d5b1a 2968 struct sockaddr *addr = args->dstaddr;
c8541ecd 2969 struct rpc_xprt *xprt;
c8475461 2970 struct sock_xprt *transport;
0a68b0be 2971 struct rpc_xprt *ret;
a246b010 2972
d9ba131d
TM		 2973 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2974 xprt_udp_slot_table_entries);
c8541ecd
CL		 2975 if (IS_ERR(xprt))
2976 return xprt;
c8475461 2977 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 2978
ec739ef0 2979 xprt->prot = IPPROTO_UDP;
a246b010
CL		 2980 /* XXX: header size can vary due to auth type, IPv6, etc. */
2981 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2982
03bf4b70 2983 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 2984 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2985 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 2986
262965f5 2987 xprt->ops = &xs_udp_ops;
a246b010 2988
ba7392bb 2989 xprt->timeout = &xs_udp_default_timeout;
a246b010 2990
f9b2ee71 2991 INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn);
4f8943f8 2992 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 2993 INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket);
2994
8f9d5b1a
CL		 2995 switch (addr->sa_family) {
2996 case AF_INET:
2997 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2998 xprt_set_bound(xprt);
2999
9dc3b095 3000 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
8f9d5b1a
CL		 3001 break;
3002 case AF_INET6:
3003 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
3004 xprt_set_bound(xprt);
3005
9dc3b095 3006 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
8f9d5b1a
CL		 3007 break;
3008 default:
0a68b0be
BF		 3009 ret = ERR_PTR(-EAFNOSUPPORT);
3010 goto out_err;
8f9d5b1a
CL		 3011 }
3012
c740eff8
CL		 3013 if (xprt_bound(xprt))
3014 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3015 xprt->address_strings[RPC_DISPLAY_ADDR],
3016 xprt->address_strings[RPC_DISPLAY_PORT],
3017 xprt->address_strings[RPC_DISPLAY_PROTO]);
3018 else
3019 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
3020 xprt->address_strings[RPC_DISPLAY_ADDR],
3021 xprt->address_strings[RPC_DISPLAY_PROTO]);
edb267a6 3022
bc25571e
TT		 3023 if (try_module_get(THIS_MODULE))
3024 return xprt;
0a68b0be
BF		 3025 ret = ERR_PTR(-EINVAL);
3026out_err:
315f3812 3027 xs_xprt_free(xprt);
0a68b0be 3028 return ret;
a246b010
CL		 3029}
3030
2881ae74
TM		 3031static const struct rpc_timeout xs_tcp_default_timeout = {
3032 .to_initval = 60 * HZ,
3033 .to_maxval = 60 * HZ,
3034 .to_retries = 2,
3035};
3036
9903cd1c
CL		 3037/**
3038 * xs_setup_tcp - Set up transport to use a TCP socket
96802a09 3039 * @args: rpc transport creation arguments
9903cd1c
CL		 3040 *
3041 */
483066d6 3042static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
a246b010 3043{
8f9d5b1a 3044 struct sockaddr *addr = args->dstaddr;
c8541ecd 3045 struct rpc_xprt *xprt;
c8475461 3046 struct sock_xprt *transport;
0a68b0be 3047 struct rpc_xprt *ret;
b7993ceb
TM		 3048 unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
3049
3050 if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
3051 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
a246b010 3052
d9ba131d 3053 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
b7993ceb 3054 max_slot_table_size);
c8541ecd
CL		 3055 if (IS_ERR(xprt))
3056 return xprt;
c8475461 3057 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 3058
ec739ef0 3059 xprt->prot = IPPROTO_TCP;
808012fb 3060 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
a246b010 3061
03bf4b70 3062 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 3063 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3064 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 3065
262965f5 3066 xprt->ops = &xs_tcp_ops;
ba7392bb 3067 xprt->timeout = &xs_tcp_default_timeout;
a246b010 3068
3851f1cd 3069 xprt->max_reconnect_timeout = xprt->timeout->to_maxval;
7196dbb0
TM		 3070 xprt->connect_timeout = xprt->timeout->to_initval *
3071 (xprt->timeout->to_retries + 1);
3851f1cd 3072
c50b8ee0 3073 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 3074 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 3075 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket);
3076
8f9d5b1a
CL		 3077 switch (addr->sa_family) {
3078 case AF_INET:
3079 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
3080 xprt_set_bound(xprt);
3081
9dc3b095 3082 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
8f9d5b1a
CL		 3083 break;
3084 case AF_INET6:
3085 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
3086 xprt_set_bound(xprt);
3087
9dc3b095 3088 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
8f9d5b1a
CL		 3089 break;
3090 default:
0a68b0be
BF		 3091 ret = ERR_PTR(-EAFNOSUPPORT);
3092 goto out_err;
8f9d5b1a
CL		 3093 }
3094
c740eff8
CL		 3095 if (xprt_bound(xprt))
3096 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3097 xprt->address_strings[RPC_DISPLAY_ADDR],
3098 xprt->address_strings[RPC_DISPLAY_PORT],
3099 xprt->address_strings[RPC_DISPLAY_PROTO]);
3100 else
3101 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
3102 xprt->address_strings[RPC_DISPLAY_ADDR],
3103 xprt->address_strings[RPC_DISPLAY_PROTO]);
3104
bc25571e
TT		 3105 if (try_module_get(THIS_MODULE))
3106 return xprt;
0a68b0be
BF		 3107 ret = ERR_PTR(-EINVAL);
3108out_err:
315f3812 3109 xs_xprt_free(xprt);
0a68b0be 3110 return ret;
a246b010 3111}
282b32e1 3112
f300baba
AB		 3113/**
3114 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
3115 * @args: rpc transport creation arguments
3116 *
3117 */
3118static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
3119{
3120 struct sockaddr *addr = args->dstaddr;
3121 struct rpc_xprt *xprt;
3122 struct sock_xprt *transport;
3123 struct svc_sock *bc_sock;
0a68b0be 3124 struct rpc_xprt *ret;
f300baba 3125
d9ba131d
TM		 3126 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
3127 xprt_tcp_slot_table_entries);
f300baba
AB		 3128 if (IS_ERR(xprt))
3129 return xprt;
3130 transport = container_of(xprt, struct sock_xprt, xprt);
3131
3132 xprt->prot = IPPROTO_TCP;
f300baba
AB		 3133 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
3134 xprt->timeout = &xs_tcp_default_timeout;
3135
3136 /* backchannel */
3137 xprt_set_bound(xprt);
3138 xprt->bind_timeout = 0;
f300baba
AB		 3139 xprt->reestablish_timeout = 0;
3140 xprt->idle_timeout = 0;
3141
f300baba
AB		 3142 xprt->ops = &bc_tcp_ops;
3143
3144 switch (addr->sa_family) {
3145 case AF_INET:
3146 xs_format_peer_addresses(xprt, "tcp",
3147 RPCBIND_NETID_TCP);
3148 break;
3149 case AF_INET6:
3150 xs_format_peer_addresses(xprt, "tcp",
3151 RPCBIND_NETID_TCP6);
3152 break;
3153 default:
0a68b0be
BF		 3154 ret = ERR_PTR(-EAFNOSUPPORT);
3155 goto out_err;
f300baba
AB		 3156 }
3157
50fa0d40
PE		 3158 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3159 xprt->address_strings[RPC_DISPLAY_ADDR],
3160 xprt->address_strings[RPC_DISPLAY_PORT],
3161 xprt->address_strings[RPC_DISPLAY_PROTO]);
f300baba 3162
99de8ea9
BF		 3163 /*
3164 * Once we've associated a backchannel xprt with a connection,
28303ca3
WM		 3165 * we want to keep it around as long as the connection lasts,
3166 * in case we need to start using it for a backchannel again;
3167 * this reference won't be dropped until bc_xprt is destroyed.
99de8ea9
BF		 3168 */
3169 xprt_get(xprt);
3170 args->bc_xprt->xpt_bc_xprt = xprt;
3171 xprt->bc_xprt = args->bc_xprt;
3172 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
3173 transport->sock = bc_sock->sk_sock;
3174 transport->inet = bc_sock->sk_sk;
3175
f300baba
AB		 3176 /*
3177 * Since we don't want connections for the backchannel, we set
3178 * the xprt status to connected
3179 */
3180 xprt_set_connected(xprt);
3181
f300baba
AB		 3182 if (try_module_get(THIS_MODULE))
3183 return xprt;
642aab58
KM		 3184
3185 args->bc_xprt->xpt_bc_xprt = NULL;
39a9beab 3186 args->bc_xprt->xpt_bc_xps = NULL;
99de8ea9 3187 xprt_put(xprt);
0a68b0be
BF		 3188 ret = ERR_PTR(-EINVAL);
3189out_err:
315f3812 3190 xs_xprt_free(xprt);
0a68b0be 3191 return ret;
f300baba
AB		 3192}
3193
176e21ee
CL		 3194static struct xprt_class xs_local_transport = {
3195 .list = LIST_HEAD_INIT(xs_local_transport.list),
3196 .name = "named UNIX socket",
3197 .owner = THIS_MODULE,
3198 .ident = XPRT_TRANSPORT_LOCAL,
3199 .setup = xs_setup_local,
3200};
3201
bc25571e
TT		 3202static struct xprt_class xs_udp_transport = {
3203 .list = LIST_HEAD_INIT(xs_udp_transport.list),
3204 .name = "udp",
3205 .owner = THIS_MODULE,
f300baba 3206 .ident = XPRT_TRANSPORT_UDP,
bc25571e
TT		 3207 .setup = xs_setup_udp,
3208};
3209
3210static struct xprt_class xs_tcp_transport = {
3211 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
3212 .name = "tcp",
3213 .owner = THIS_MODULE,
f300baba 3214 .ident = XPRT_TRANSPORT_TCP,
bc25571e
TT		 3215 .setup = xs_setup_tcp,
3216};
3217
f300baba
AB		 3218static struct xprt_class xs_bc_tcp_transport = {
3219 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
3220 .name = "tcp NFSv4.1 backchannel",
3221 .owner = THIS_MODULE,
3222 .ident = XPRT_TRANSPORT_BC_TCP,
3223 .setup = xs_setup_bc_tcp,
3224};
3225
282b32e1 3226/**
bc25571e 3227 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
282b32e1
CL		 3228 *
3229 */
3230int init_socket_xprt(void)
3231{
2b1bec5f 3232 if (!sunrpc_table_header)
0b4d4147 3233 sunrpc_table_header = register_sysctl_table(sunrpc_table);
fbf76683 3234
176e21ee 3235 xprt_register_transport(&xs_local_transport);
bc25571e
TT		 3236 xprt_register_transport(&xs_udp_transport);
3237 xprt_register_transport(&xs_tcp_transport);
f300baba 3238 xprt_register_transport(&xs_bc_tcp_transport);
bc25571e 3239
282b32e1
CL		 3240 return 0;
3241}
3242
3243/**
bc25571e 3244 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
282b32e1
CL		 3245 *
3246 */
3247void cleanup_socket_xprt(void)
3248{
fbf76683
CL		 3249 if (sunrpc_table_header) {
3250 unregister_sysctl_table(sunrpc_table_header);
3251 sunrpc_table_header = NULL;
3252 }
bc25571e 3253
176e21ee 3254 xprt_unregister_transport(&xs_local_transport);
bc25571e
TT		 3255 xprt_unregister_transport(&xs_udp_transport);
3256 xprt_unregister_transport(&xs_tcp_transport);
f300baba 3257 xprt_unregister_transport(&xs_bc_tcp_transport);
282b32e1 3258}
cbf11071 3259
9bbb9e5a
RR		 3260static int param_set_uint_minmax(const char *val,
3261 const struct kernel_param *kp,
cbf11071
TM		 3262 unsigned int min, unsigned int max)
3263{
00cfaa94 3264 unsigned int num;
cbf11071
TM		 3265 int ret;
3266
3267 if (!val)
3268 return -EINVAL;
00cfaa94 3269 ret = kstrtouint(val, 0, &num);
9761a246
DC		 3270 if (ret)
3271 return ret;
3272 if (num < min || num > max)
cbf11071
TM		 3273 return -EINVAL;
3274 *((unsigned int *)kp->arg) = num;
3275 return 0;
3276}
3277
9bbb9e5a 3278static int param_set_portnr(const char *val, const struct kernel_param *kp)
cbf11071 3279{
ffb6ca33 3280 return param_set_uint_minmax(val, kp,
826799e6 3281 RPC_MIN_RESVPORT,
cbf11071
TM		 3282 RPC_MAX_RESVPORT);
3283}
3284
9c27847d 3285static const struct kernel_param_ops param_ops_portnr = {
9bbb9e5a
RR		 3286 .set = param_set_portnr,
3287 .get = param_get_uint,
3288};
3289
cbf11071
TM		 3290#define param_check_portnr(name, p) \
3291 __param_check(name, p, unsigned int);
3292
3293module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3294module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3295
9bbb9e5a
RR		 3296static int param_set_slot_table_size(const char *val,
3297 const struct kernel_param *kp)
cbf11071
TM		 3298{
3299 return param_set_uint_minmax(val, kp,
3300 RPC_MIN_SLOT_TABLE,
3301 RPC_MAX_SLOT_TABLE);
3302}
3303
9c27847d 3304static const struct kernel_param_ops param_ops_slot_table_size = {
9bbb9e5a
RR		 3305 .set = param_set_slot_table_size,
3306 .get = param_get_uint,
3307};
3308
cbf11071
TM		 3309#define param_check_slot_table_size(name, p) \
3310 __param_check(name, p, unsigned int);
3311
d9ba131d
TM		 3312static int param_set_max_slot_table_size(const char *val,
3313 const struct kernel_param *kp)
3314{
3315 return param_set_uint_minmax(val, kp,
3316 RPC_MIN_SLOT_TABLE,
3317 RPC_MAX_SLOT_TABLE_LIMIT);
3318}
3319
9c27847d 3320static const struct kernel_param_ops param_ops_max_slot_table_size = {
d9ba131d
TM		 3321 .set = param_set_max_slot_table_size,
3322 .get = param_get_uint,
3323};
3324
3325#define param_check_max_slot_table_size(name, p) \
3326 __param_check(name, p, unsigned int);
3327
cbf11071
TM		 3328module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3329 slot_table_size, 0644);
d9ba131d
TM		 3330module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3331 max_slot_table_size, 0644);
cbf11071
TM		 3332module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3333 slot_table_size, 0644);
Linux 6.x block layer and io_uring tree(s)