projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
treewide: use prandom_u32_max() when possible, part 1
[linux-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
9e55eef4 57#include "socklib.h"
4cfc7e60 58#include "sunrpc.h"
176e21ee
CL		 59
60static void xs_close(struct rpc_xprt *xprt);
3b21f757 61static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock);
7196dbb0
TM		 62static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
63 struct socket *sock);
176e21ee 64
c556b754
CL		 65/*
66 * xprtsock tunables
67 */
09acfea5
TM		 68static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
69static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
70static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
c556b754 71
09acfea5
TM		 72static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
73static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
c556b754 74
7d1e8255 75#define XS_TCP_LINGER_TO (15U * HZ)
25fe6142 76static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
7d1e8255 77
fbf76683
CL		 78/*
79 * We can register our own files under /proc/sys/sunrpc by
80 * calling register_sysctl_table() again. The files in that
81 * directory become the union of all files registered there.
82 *
83 * We simply need to make sure that we don't collide with
84 * someone else's file names!
85 */
86
fbf76683
CL		 87static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
88static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
d9ba131d 89static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
fbf76683
CL		 90static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
91static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
92
93static struct ctl_table_header *sunrpc_table_header;
94
d3abc739
OK		 95static struct xprt_class xs_local_transport;
96static struct xprt_class xs_udp_transport;
97static struct xprt_class xs_tcp_transport;
98static struct xprt_class xs_bc_tcp_transport;
99
fbf76683
CL		 100/*
101 * FIXME: changing the UDP slot table size should also resize the UDP
102 * socket buffers for existing UDP transports
103 */
fe2c6338 104static struct ctl_table xs_tunables_table[] = {
fbf76683 105 {
fbf76683
CL		 106 .procname = "udp_slot_table_entries",
107 .data = &xprt_udp_slot_table_entries,
108 .maxlen = sizeof(unsigned int),
109 .mode = 0644,
6d456111 110 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 111 .extra1 = &min_slot_table_size,
112 .extra2 = &max_slot_table_size
113 },
114 {
fbf76683
CL		 115 .procname = "tcp_slot_table_entries",
116 .data = &xprt_tcp_slot_table_entries,
117 .maxlen = sizeof(unsigned int),
118 .mode = 0644,
6d456111 119 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 120 .extra1 = &min_slot_table_size,
121 .extra2 = &max_slot_table_size
122 },
d9ba131d
TM		 123 {
124 .procname = "tcp_max_slot_table_entries",
125 .data = &xprt_max_tcp_slot_table_entries,
126 .maxlen = sizeof(unsigned int),
127 .mode = 0644,
128 .proc_handler = proc_dointvec_minmax,
129 .extra1 = &min_slot_table_size,
130 .extra2 = &max_tcp_slot_table_limit
131 },
fbf76683 132 {
fbf76683
CL		 133 .procname = "min_resvport",
134 .data = &xprt_min_resvport,
135 .maxlen = sizeof(unsigned int),
136 .mode = 0644,
6d456111 137 .proc_handler = proc_dointvec_minmax,
fbf76683 138 .extra1 = &xprt_min_resvport_limit,
826799e6 139 .extra2 = &xprt_max_resvport_limit
fbf76683
CL		 140 },
141 {
fbf76683
CL		 142 .procname = "max_resvport",
143 .data = &xprt_max_resvport,
144 .maxlen = sizeof(unsigned int),
145 .mode = 0644,
6d456111 146 .proc_handler = proc_dointvec_minmax,
826799e6 147 .extra1 = &xprt_min_resvport_limit,
fbf76683
CL		 148 .extra2 = &xprt_max_resvport_limit
149 },
25fe6142
TM		 150 {
151 .procname = "tcp_fin_timeout",
152 .data = &xs_tcp_fin_timeout,
153 .maxlen = sizeof(xs_tcp_fin_timeout),
154 .mode = 0644,
6d456111 155 .proc_handler = proc_dointvec_jiffies,
fbf76683 156 },
f8572d8f 157 { },
fbf76683
CL		 158};
159
fe2c6338 160static struct ctl_table sunrpc_table[] = {
fbf76683 161 {
fbf76683
CL		 162 .procname = "sunrpc",
163 .mode = 0555,
164 .child = xs_tunables_table
165 },
f8572d8f 166 { },
fbf76683
CL		 167};
168
03bf4b70
CL		 169/*
170 * Wait duration for a reply from the RPC portmapper.
171 */
172#define XS_BIND_TO (60U * HZ)
173
174/*
175 * Delay if a UDP socket connect error occurs. This is most likely some
176 * kind of resource problem on the local host.
177 */
178#define XS_UDP_REEST_TO (2U * HZ)
179
180/*
181 * The reestablish timeout allows clients to delay for a bit before attempting
182 * to reconnect to a server that just dropped our connection.
183 *
184 * We implement an exponential backoff when trying to reestablish a TCP
185 * transport connection with the server. Some servers like to drop a TCP
186 * connection when they are overworked, so we start with a short timeout and
187 * increase over time if the server is down or not responding.
188 */
189#define XS_TCP_INIT_REEST_TO (3U * HZ)
03bf4b70
CL		 190
191/*
192 * TCP idle timeout; client drops the transport socket if it is idle
193 * for this long. Note that we also timeout UDP sockets to prevent
194 * holding port numbers when there is no RPC traffic.
195 */
196#define XS_IDLE_DISC_TO (5U * 60 * HZ)
197
f895b252 198#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
a246b010 199# undef RPC_DEBUG_DATA
9903cd1c 200# define RPCDBG_FACILITY RPCDBG_TRANS
a246b010
CL		 201#endif
202
a246b010 203#ifdef RPC_DEBUG_DATA
9903cd1c 204static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010 205{
9903cd1c
CL		 206 u8 *buf = (u8 *) packet;
207 int j;
a246b010 208
46121cf7 209 dprintk("RPC: %s\n", msg);
a246b010
CL		 210 for (j = 0; j < count && j < 128; j += 4) {
211 if (!(j & 31)) {
212 if (j)
213 dprintk("\n");
214 dprintk("0x%04x ", j);
215 }
216 dprintk("%02x%02x%02x%02x ",
217 buf[j], buf[j+1], buf[j+2], buf[j+3]);
218 }
219 dprintk("\n");
220}
221#else
9903cd1c 222static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010
CL		 223{
224 /* NOP */
225}
226#endif
227
2118071d
TM		 228static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
229{
230 return (struct rpc_xprt *) sk->sk_user_data;
231}
232
95392c59
CL		 233static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
234{
235 return (struct sockaddr *) &xprt->addr;
236}
237
176e21ee
CL		 238static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
239{
240 return (struct sockaddr_un *) &xprt->addr;
241}
242
95392c59 243static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
edb267a6 244{
95392c59
CL		 245 return (struct sockaddr_in *) &xprt->addr;
246}
247
248static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
249{
250 return (struct sockaddr_in6 *) &xprt->addr;
251}
252
c877b849 253static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
edb267a6 254{
c877b849 255 struct sockaddr *sap = xs_addr(xprt);
9dc3b095
CL		 256 struct sockaddr_in6 *sin6;
257 struct sockaddr_in *sin;
176e21ee 258 struct sockaddr_un *sun;
c877b849 259 char buf[128];
edb267a6 260
9dc3b095 261 switch (sap->sa_family) {
176e21ee
CL		 262 case AF_LOCAL:
263 sun = xs_addr_un(xprt);
264 strlcpy(buf, sun->sun_path, sizeof(buf));
265 xprt->address_strings[RPC_DISPLAY_ADDR] =
266 kstrdup(buf, GFP_KERNEL);
267 break;
9dc3b095 268 case AF_INET:
176e21ee
CL		 269 (void)rpc_ntop(sap, buf, sizeof(buf));
270 xprt->address_strings[RPC_DISPLAY_ADDR] =
271 kstrdup(buf, GFP_KERNEL);
9dc3b095 272 sin = xs_addr_in(xprt);
fc0b5791 273 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
9dc3b095
CL		 274 break;
275 case AF_INET6:
176e21ee
CL		 276 (void)rpc_ntop(sap, buf, sizeof(buf));
277 xprt->address_strings[RPC_DISPLAY_ADDR] =
278 kstrdup(buf, GFP_KERNEL);
9dc3b095 279 sin6 = xs_addr_in6(xprt);
fc0b5791 280 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
9dc3b095
CL		 281 break;
282 default:
283 BUG();
756805e7 284 }
176e21ee 285
9dc3b095 286 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
edb267a6
CL		 287}
288
9dc3b095 289static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
4b6473fb 290{
9dc3b095
CL		 291 struct sockaddr *sap = xs_addr(xprt);
292 char buf[128];
4b6473fb 293
81160e66 294 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
c877b849 295 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
4b6473fb 296
81160e66 297 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
c877b849
CL		 298 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
299}
4b6473fb 300
9dc3b095
CL		 301static void xs_format_peer_addresses(struct rpc_xprt *xprt,
302 const char *protocol,
303 const char *netid)
c877b849 304{
b454ae90 305 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
b454ae90 306 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
c877b849 307 xs_format_common_peer_addresses(xprt);
9dc3b095 308 xs_format_common_peer_ports(xprt);
edb267a6 309}
4b6473fb 310
9dc3b095 311static void xs_update_peer_port(struct rpc_xprt *xprt)
4b6473fb 312{
9dc3b095
CL		 313 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
314 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
4417c8c4 315
9dc3b095 316 xs_format_common_peer_ports(xprt);
edb267a6
CL		 317}
318
319static void xs_free_peer_addresses(struct rpc_xprt *xprt)
320{
33e01dc7
CL		 321 unsigned int i;
322
323 for (i = 0; i < RPC_DISPLAY_MAX; i++)
324 switch (i) {
325 case RPC_DISPLAY_PROTO:
326 case RPC_DISPLAY_NETID:
327 continue;
328 default:
329 kfree(xprt->address_strings[i]);
330 }
edb267a6
CL		 331}
332
277e4ab7
TM		 333static size_t
334xs_alloc_sparse_pages(struct xdr_buf *buf, size_t want, gfp_t gfp)
335{
336 size_t i,n;
337
16e5e90f 338 if (!want || !(buf->flags & XDRBUF_SPARSE_PAGES))
277e4ab7 339 return want;
277e4ab7
TM		 340 n = (buf->page_base + want + PAGE_SIZE - 1) >> PAGE_SHIFT;
341 for (i = 0; i < n; i++) {
342 if (buf->pages[i])
343 continue;
344 buf->bvec[i].bv_page = buf->pages[i] = alloc_page(gfp);
345 if (!buf->pages[i]) {
16e5e90f
TM		 346 i *= PAGE_SIZE;
347 return i > buf->page_base ? i - buf->page_base : 0;
277e4ab7
TM		 348 }
349 }
350 return want;
351}
352
353static ssize_t
354xs_sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags, size_t seek)
355{
356 ssize_t ret;
357 if (seek != 0)
358 iov_iter_advance(&msg->msg_iter, seek);
359 ret = sock_recvmsg(sock, msg, flags);
360 return ret > 0 ? ret + seek : ret;
361}
362
363static ssize_t
364xs_read_kvec(struct socket *sock, struct msghdr *msg, int flags,
365 struct kvec *kvec, size_t count, size_t seek)
366{
0e9b4a82 367 iov_iter_kvec(&msg->msg_iter, READ, kvec, 1, count);
277e4ab7
TM		 368 return xs_sock_recvmsg(sock, msg, flags, seek);
369}
370
371static ssize_t
372xs_read_bvec(struct socket *sock, struct msghdr *msg, int flags,
373 struct bio_vec *bvec, unsigned long nr, size_t count,
374 size_t seek)
375{
0e9b4a82 376 iov_iter_bvec(&msg->msg_iter, READ, bvec, nr, count);
277e4ab7
TM		 377 return xs_sock_recvmsg(sock, msg, flags, seek);
378}
379
380static ssize_t
381xs_read_discard(struct socket *sock, struct msghdr *msg, int flags,
382 size_t count)
383{
b76a5afd
TM		 384 iov_iter_discard(&msg->msg_iter, READ, count);
385 return sock_recvmsg(sock, msg, flags);
277e4ab7
TM		 386}
387
6a829eb8
TM		 388#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
389static void
390xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
391{
392 struct bvec_iter bi = {
393 .bi_size = count,
394 };
395 struct bio_vec bv;
396
397 bvec_iter_advance(bvec, &bi, seek & PAGE_MASK);
398 for_each_bvec(bv, bvec, bi, bi)
399 flush_dcache_page(bv.bv_page);
400}
401#else
402static inline void
403xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
404{
405}
406#endif
407
277e4ab7
TM		 408static ssize_t
409xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags,
410 struct xdr_buf *buf, size_t count, size_t seek, size_t *read)
411{
412 size_t want, seek_init = seek, offset = 0;
413 ssize_t ret;
414
e92053a5
TM		 415 want = min_t(size_t, count, buf->head[0].iov_len);
416 if (seek < want) {
277e4ab7
TM		 417 ret = xs_read_kvec(sock, msg, flags, &buf->head[0], want, seek);
418 if (ret <= 0)
419 goto sock_err;
420 offset += ret;
421 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
422 goto out;
423 if (ret != want)
c4433055 424 goto out;
277e4ab7
TM		 425 seek = 0;
426 } else {
e92053a5
TM		 427 seek -= want;
428 offset += want;
277e4ab7 429 }
16e5e90f 430
b2648015
TM		 431 want = xs_alloc_sparse_pages(
432 buf, min_t(size_t, count - offset, buf->page_len),
433 GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN);
16e5e90f 434 if (seek < want) {
277e4ab7
TM		 435 ret = xs_read_bvec(sock, msg, flags, buf->bvec,
436 xdr_buf_pagecount(buf),
437 want + buf->page_base,
438 seek + buf->page_base);
439 if (ret <= 0)
440 goto sock_err;
6a829eb8 441 xs_flush_bvec(buf->bvec, ret, seek + buf->page_base);
ac9645c8
DA		 442 ret -= buf->page_base;
443 offset += ret;
277e4ab7
TM		 444 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
445 goto out;
446 if (ret != want)
c4433055 447 goto out;
277e4ab7
TM		 448 seek = 0;
449 } else {
16e5e90f
TM		 450 seek -= want;
451 offset += want;
277e4ab7 452 }
16e5e90f 453
e92053a5
TM		 454 want = min_t(size_t, count - offset, buf->tail[0].iov_len);
455 if (seek < want) {
277e4ab7
TM		 456 ret = xs_read_kvec(sock, msg, flags, &buf->tail[0], want, seek);
457 if (ret <= 0)
458 goto sock_err;
459 offset += ret;
460 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
461 goto out;
462 if (ret != want)
c4433055 463 goto out;
9734ad57 464 } else if (offset < seek_init)
e92053a5 465 offset = seek_init;
277e4ab7 466 ret = -EMSGSIZE;
277e4ab7
TM		 467out:
468 *read = offset - seek_init;
469 return ret;
277e4ab7
TM		 470sock_err:
471 offset += seek;
472 goto out;
473}
474
475static void
476xs_read_header(struct sock_xprt *transport, struct xdr_buf *buf)
477{
478 if (!transport->recv.copied) {
479 if (buf->head[0].iov_len >= transport->recv.offset)
480 memcpy(buf->head[0].iov_base,
481 &transport->recv.xid,
482 transport->recv.offset);
483 transport->recv.copied = transport->recv.offset;
484 }
485}
486
487static bool
488xs_read_stream_request_done(struct sock_xprt *transport)
489{
490 return transport->recv.fraghdr & cpu_to_be32(RPC_LAST_STREAM_FRAGMENT);
491}
492
e92053a5
TM		 493static void
494xs_read_stream_check_eor(struct sock_xprt *transport,
495 struct msghdr *msg)
496{
497 if (xs_read_stream_request_done(transport))
498 msg->msg_flags |= MSG_EOR;
499}
500
277e4ab7
TM		 501static ssize_t
502xs_read_stream_request(struct sock_xprt *transport, struct msghdr *msg,
503 int flags, struct rpc_rqst *req)
504{
505 struct xdr_buf *buf = &req->rq_private_buf;
3f649ab7
KC		 506 size_t want, read;
507 ssize_t ret;
277e4ab7
TM		 508
509 xs_read_header(transport, buf);
510
511 want = transport->recv.len - transport->recv.offset;
e92053a5
TM		 512 if (want != 0) {
513 ret = xs_read_xdr_buf(transport->sock, msg, flags, buf,
514 transport->recv.copied + want,
515 transport->recv.copied,
516 &read);
517 transport->recv.offset += read;
518 transport->recv.copied += read;
727fcc64 519 }
e92053a5 520
727fcc64 521 if (transport->recv.offset == transport->recv.len)
e92053a5 522 xs_read_stream_check_eor(transport, msg);
277e4ab7 523
e92053a5
TM		 524 if (want == 0)
525 return 0;
526
277e4ab7 527 switch (ret) {
c4433055
TM		 528 default:
529 break;
26781eab 530 case -EFAULT:
277e4ab7 531 case -EMSGSIZE:
26781eab 532 msg->msg_flags |= MSG_TRUNC;
c4433055 533 return read;
277e4ab7
TM		 534 case 0:
535 return -ESHUTDOWN;
277e4ab7 536 }
c4433055 537 return ret < 0 ? ret : read;
277e4ab7
TM		 538}
539
540static size_t
541xs_read_stream_headersize(bool isfrag)
542{
543 if (isfrag)
544 return sizeof(__be32);
545 return 3 * sizeof(__be32);
546}
547
548static ssize_t
549xs_read_stream_header(struct sock_xprt *transport, struct msghdr *msg,
550 int flags, size_t want, size_t seek)
551{
552 struct kvec kvec = {
553 .iov_base = &transport->recv.fraghdr,
554 .iov_len = want,
555 };
556 return xs_read_kvec(transport->sock, msg, flags, &kvec, want, seek);
557}
558
559#if defined(CONFIG_SUNRPC_BACKCHANNEL)
560static ssize_t
561xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
562{
563 struct rpc_xprt *xprt = &transport->xprt;
564 struct rpc_rqst *req;
565 ssize_t ret;
566
98b5cee3
BC		 567 /* Is this transport associated with the backchannel? */
568 if (!xprt->bc_serv)
569 return -ESHUTDOWN;
570
277e4ab7
TM		 571 /* Look up and lock the request corresponding to the given XID */
572 req = xprt_lookup_bc_request(xprt, transport->recv.xid);
573 if (!req) {
574 printk(KERN_WARNING "Callback slot table overflowed\n");
575 return -ESHUTDOWN;
576 }
45835a63
TM		 577 if (transport->recv.copied && !req->rq_private_buf.len)
578 return -ESHUTDOWN;
277e4ab7
TM		 579
580 ret = xs_read_stream_request(transport, msg, flags, req);
581 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 582 xprt_complete_bc_request(req, transport->recv.copied);
45835a63
TM		 583 else
584 req->rq_private_buf.len = transport->recv.copied;
277e4ab7
TM		 585
586 return ret;
587}
588#else /* CONFIG_SUNRPC_BACKCHANNEL */
589static ssize_t
590xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
591{
592 return -ESHUTDOWN;
593}
594#endif /* CONFIG_SUNRPC_BACKCHANNEL */
595
596static ssize_t
597xs_read_stream_reply(struct sock_xprt *transport, struct msghdr *msg, int flags)
598{
599 struct rpc_xprt *xprt = &transport->xprt;
600 struct rpc_rqst *req;
601 ssize_t ret = 0;
602
603 /* Look up and lock the request corresponding to the given XID */
604 spin_lock(&xprt->queue_lock);
605 req = xprt_lookup_rqst(xprt, transport->recv.xid);
45835a63 606 if (!req || (transport->recv.copied && !req->rq_private_buf.len)) {
277e4ab7
TM		 607 msg->msg_flags |= MSG_TRUNC;
608 goto out;
609 }
610 xprt_pin_rqst(req);
611 spin_unlock(&xprt->queue_lock);
612
613 ret = xs_read_stream_request(transport, msg, flags, req);
614
615 spin_lock(&xprt->queue_lock);
616 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 617 xprt_complete_rqst(req->rq_task, transport->recv.copied);
45835a63
TM		 618 else
619 req->rq_private_buf.len = transport->recv.copied;
277e4ab7
TM		 620 xprt_unpin_rqst(req);
621out:
622 spin_unlock(&xprt->queue_lock);
623 return ret;
624}
625
626static ssize_t
627xs_read_stream(struct sock_xprt *transport, int flags)
628{
629 struct msghdr msg = { 0 };
630 size_t want, read = 0;
631 ssize_t ret = 0;
632
633 if (transport->recv.len == 0) {
634 want = xs_read_stream_headersize(transport->recv.copied != 0);
635 ret = xs_read_stream_header(transport, &msg, flags, want,
636 transport->recv.offset);
637 if (ret <= 0)
638 goto out_err;
639 transport->recv.offset = ret;
c4433055
TM		 640 if (transport->recv.offset != want)
641 return transport->recv.offset;
277e4ab7
TM		 642 transport->recv.len = be32_to_cpu(transport->recv.fraghdr) &
643 RPC_FRAGMENT_SIZE_MASK;
644 transport->recv.offset -= sizeof(transport->recv.fraghdr);
645 read = ret;
646 }
647
648 switch (be32_to_cpu(transport->recv.calldir)) {
c4433055
TM		 649 default:
650 msg.msg_flags |= MSG_TRUNC;
651 break;
277e4ab7
TM		 652 case RPC_CALL:
653 ret = xs_read_stream_call(transport, &msg, flags);
654 break;
655 case RPC_REPLY:
656 ret = xs_read_stream_reply(transport, &msg, flags);
657 }
658 if (msg.msg_flags & MSG_TRUNC) {
659 transport->recv.calldir = cpu_to_be32(-1);
660 transport->recv.copied = -1;
661 }
662 if (ret < 0)
663 goto out_err;
664 read += ret;
665 if (transport->recv.offset < transport->recv.len) {
c4433055
TM		 666 if (!(msg.msg_flags & MSG_TRUNC))
667 return read;
b76a5afd 668 msg.msg_flags = 0;
277e4ab7
TM		 669 ret = xs_read_discard(transport->sock, &msg, flags,
670 transport->recv.len - transport->recv.offset);
671 if (ret <= 0)
672 goto out_err;
673 transport->recv.offset += ret;
674 read += ret;
675 if (transport->recv.offset != transport->recv.len)
c4433055 676 return read;
277e4ab7
TM		 677 }
678 if (xs_read_stream_request_done(transport)) {
c50b8ee0 679 trace_xs_stream_read_request(transport);
277e4ab7
TM		 680 transport->recv.copied = 0;
681 }
682 transport->recv.offset = 0;
683 transport->recv.len = 0;
684 return read;
685out_err:
79462857 686 return ret != 0 ? ret : -ESHUTDOWN;
277e4ab7
TM		 687}
688
0ffe86f4
TM		 689static __poll_t xs_poll_socket(struct sock_xprt *transport)
690{
a73881c9
TM		 691 return transport->sock->ops->poll(transport->file, transport->sock,
692 NULL);
0ffe86f4
TM		 693}
694
695static bool xs_poll_socket_readable(struct sock_xprt *transport)
696{
697 __poll_t events = xs_poll_socket(transport);
698
699 return (events & (EPOLLIN | EPOLLRDNORM)) && !(events & EPOLLRDHUP);
700}
701
702static void xs_poll_check_readable(struct sock_xprt *transport)
703{
704
705 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
706 if (!xs_poll_socket_readable(transport))
707 return;
708 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
709 queue_work(xprtiod_workqueue, &transport->recv_worker);
710}
711
c50b8ee0
TM		 712static void xs_stream_data_receive(struct sock_xprt *transport)
713{
714 size_t read = 0;
715 ssize_t ret = 0;
716
717 mutex_lock(&transport->recv_mutex);
718 if (transport->sock == NULL)
719 goto out;
c50b8ee0
TM		 720 for (;;) {
721 ret = xs_read_stream(transport, MSG_DONTWAIT);
c4433055 722 if (ret < 0)
c50b8ee0
TM		 723 break;
724 read += ret;
725 cond_resched();
726 }
5f52a9d4
TM		 727 if (ret == -ESHUTDOWN)
728 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
729 else
730 xs_poll_check_readable(transport);
c50b8ee0
TM		 731out:
732 mutex_unlock(&transport->recv_mutex);
733 trace_xs_stream_read_data(&transport->xprt, ret, read);
734}
735
736static void xs_stream_data_receive_workfn(struct work_struct *work)
737{
738 struct sock_xprt *transport =
739 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 740 unsigned int pflags = memalloc_nofs_save();
741
c50b8ee0 742 xs_stream_data_receive(transport);
a1231fda 743 memalloc_nofs_restore(pflags);
c50b8ee0
TM		 744}
745
550aebfe
TM		 746static void
747xs_stream_reset_connect(struct sock_xprt *transport)
748{
749 transport->recv.offset = 0;
750 transport->recv.len = 0;
751 transport->recv.copied = 0;
752 transport->xmit.offset = 0;
ae053551
TM		 753}
754
755static void
756xs_stream_start_connect(struct sock_xprt *transport)
757{
550aebfe
TM		 758 transport->xprt.stat.connect_count++;
759 transport->xprt.stat.connect_start = jiffies;
760}
761
b4b5cc85
CL		 762#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
763
9903cd1c 764/**
c544577d 765 * xs_nospace - handle transmit was incomplete
50f484e2 766 * @req: pointer to RPC request
7496b59f 767 * @transport: pointer to struct sock_xprt
9903cd1c 768 *
a246b010 769 */
7496b59f 770static int xs_nospace(struct rpc_rqst *req, struct sock_xprt *transport)
a246b010 771{
7496b59f 772 struct rpc_xprt *xprt = &transport->xprt;
06ea0bfe 773 struct sock *sk = transport->inet;
24ca9a84 774 int ret = -EAGAIN;
a246b010 775
015747d2 776 trace_rpc_socket_nospace(req, transport);
262965f5 777
b6ddf64f 778 /* Protect against races with write_space */
b5e92419 779 spin_lock(&xprt->transport_lock);
b6ddf64f
TM		 780
781 /* Don't race with disconnect */
782 if (xprt_connected(xprt)) {
13331a55 783 /* wait for more buffer space */
2790a624 784 set_bit(XPRT_SOCK_NOSPACE, &transport->sock_state);
7496b59f 785 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
13331a55 786 sk->sk_write_pending++;
c544577d 787 xprt_wait_for_buffer_space(xprt);
13331a55 788 } else
5e3771ce 789 ret = -ENOTCONN;
262965f5 790
b5e92419 791 spin_unlock(&xprt->transport_lock);
7496b59f
TM		 792 return ret;
793}
06ea0bfe 794
7496b59f
TM		 795static int xs_sock_nospace(struct rpc_rqst *req)
796{
797 struct sock_xprt *transport =
798 container_of(req->rq_xprt, struct sock_xprt, xprt);
799 struct sock *sk = transport->inet;
800 int ret = -EAGAIN;
d48f9ce7 801
7496b59f
TM		 802 lock_sock(sk);
803 if (!sock_writeable(sk))
804 ret = xs_nospace(req, transport);
805 release_sock(sk);
806 return ret;
807}
d48f9ce7 808
d0afde5f 809static int xs_stream_nospace(struct rpc_rqst *req, bool vm_wait)
7496b59f
TM		 810{
811 struct sock_xprt *transport =
812 container_of(req->rq_xprt, struct sock_xprt, xprt);
813 struct sock *sk = transport->inet;
814 int ret = -EAGAIN;
815
d0afde5f
TM		 816 if (vm_wait)
817 return -ENOBUFS;
7496b59f
TM		 818 lock_sock(sk);
819 if (!sk_stream_memory_free(sk))
820 ret = xs_nospace(req, transport);
821 release_sock(sk);
5e3771ce 822 return ret;
262965f5
CL		 823}
824
72691a26 825static int xs_stream_prepare_request(struct rpc_rqst *req, struct xdr_buf *buf)
277e4ab7 826{
72691a26 827 return xdr_alloc_bvec(buf, rpc_task_gfp_mask());
277e4ab7
TM		 828}
829
4cd34e7c
TM		 830/*
831 * Determine if the previous message in the stream was aborted before it
832 * could complete transmission.
833 */
834static bool
835xs_send_request_was_aborted(struct sock_xprt *transport, struct rpc_rqst *req)
836{
837 return transport->xmit.offset != 0 && req->rq_bytes_sent == 0;
838}
839
61677eee 840/*
06b5fc3a 841 * Return the stream record marker field for a record of length < 2^31-1
61677eee 842 */
06b5fc3a
TM		 843static rpc_fraghdr
844xs_stream_record_marker(struct xdr_buf *xdr)
61677eee 845{
06b5fc3a
TM		 846 if (!xdr->len)
847 return 0;
848 return cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | (u32)xdr->len);
61677eee
CL		 849}
850
176e21ee
CL		 851/**
852 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
50f484e2 853 * @req: pointer to RPC request
176e21ee
CL		 854 *
855 * Return values:
856 * 0: The request has been sent
857 * EAGAIN: The socket was blocked, please call again later to
858 * complete the request
859 * ENOTCONN: Caller needs to invoke connect logic then call again
12b20ce3 860 * other: Some other error occurred, the request was not sent
176e21ee 861 */
adfa7144 862static int xs_local_send_request(struct rpc_rqst *req)
176e21ee 863{
176e21ee
CL		 864 struct rpc_xprt *xprt = req->rq_xprt;
865 struct sock_xprt *transport =
866 container_of(xprt, struct sock_xprt, xprt);
867 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 868 rpc_fraghdr rm = xs_stream_record_marker(xdr);
869 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
9e55eef4
CL		 870 struct msghdr msg = {
871 .msg_flags = XS_SENDMSG_FLAGS,
872 };
d0afde5f 873 bool vm_wait;
3f649ab7 874 unsigned int sent;
176e21ee
CL		 875 int status;
876
4cd34e7c
TM		 877 /* Close the stream if the previous transmission was incomplete */
878 if (xs_send_request_was_aborted(transport, req)) {
f0043206 879 xprt_force_disconnect(xprt);
4cd34e7c
TM		 880 return -ENOTCONN;
881 }
882
176e21ee
CL		 883 xs_pktdump("packet data:",
884 req->rq_svec->iov_base, req->rq_svec->iov_len);
885
d0afde5f
TM		 886 vm_wait = sk_stream_is_writeable(transport->inet) ? true : false;
887
78215759 888 req->rq_xtime = ktime_get();
9e55eef4
CL		 889 status = xprt_sock_sendmsg(transport->sock, &msg, xdr,
890 transport->xmit.offset, rm, &sent);
176e21ee 891 dprintk("RPC: %s(%u) = %d\n",
6c7a64e5 892 __func__, xdr->len - transport->xmit.offset, status);
743c69e7 893
f279cd00 894 if (likely(sent > 0) || status == 0) {
6c7a64e5
TM		 895 transport->xmit.offset += sent;
896 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 897 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 898 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 899 transport->xmit.offset = 0;
176e21ee
CL		 900 return 0;
901 }
902 status = -EAGAIN;
d0afde5f 903 vm_wait = false;
176e21ee
CL		 904 }
905
906 switch (status) {
907 case -EAGAIN:
d0afde5f 908 status = xs_stream_nospace(req, vm_wait);
176e21ee
CL		 909 break;
910 default:
911 dprintk("RPC: sendmsg returned unrecognized error %d\n",
912 -status);
df561f66 913 fallthrough;
176e21ee 914 case -EPIPE:
f0043206 915 xprt_force_disconnect(xprt);
176e21ee
CL		 916 status = -ENOTCONN;
917 }
918
919 return status;
920}
921
262965f5
CL		 922/**
923 * xs_udp_send_request - write an RPC request to a UDP socket
50f484e2 924 * @req: pointer to RPC request
262965f5
CL		 925 *
926 * Return values:
927 * 0: The request has been sent
928 * EAGAIN: The socket was blocked, please call again later to
929 * complete the request
930 * ENOTCONN: Caller needs to invoke connect logic then call again
25985edc 931 * other: Some other error occurred, the request was not sent
262965f5 932 */
adfa7144 933static int xs_udp_send_request(struct rpc_rqst *req)
262965f5 934{
262965f5 935 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 936 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 937 struct xdr_buf *xdr = &req->rq_snd_buf;
9e55eef4
CL		 938 struct msghdr msg = {
939 .msg_name = xs_addr(xprt),
940 .msg_namelen = xprt->addrlen,
941 .msg_flags = XS_SENDMSG_FLAGS,
942 };
3f649ab7 943 unsigned int sent;
262965f5 944 int status;
a246b010 945
9903cd1c 946 xs_pktdump("packet data:",
a246b010
CL		 947 req->rq_svec->iov_base,
948 req->rq_svec->iov_len);
949
01d37c42
TM		 950 if (!xprt_bound(xprt))
951 return -ENOTCONN;
75891f50
TM		 952
953 if (!xprt_request_get_cong(xprt, req))
954 return -EBADSLT;
955
ff053dbb
TM		 956 status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
957 if (status < 0)
958 return status;
78215759 959 req->rq_xtime = ktime_get();
9e55eef4 960 status = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, 0, &sent);
a246b010 961
46121cf7 962 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
6c7a64e5 963 xdr->len, status);
a246b010 964
3dedbb5c
JB		 965 /* firewall is blocking us, don't return -EAGAIN or we end up looping */
966 if (status == -EPERM)
967 goto process_status;
968
743c69e7
N		 969 if (status == -EAGAIN && sock_writeable(transport->inet))
970 status = -ENOBUFS;
971
f279cd00
JB		 972 if (sent > 0 || status == 0) {
973 req->rq_xmit_bytes_sent += sent;
974 if (sent >= req->rq_slen)
2199700f
TM		 975 return 0;
976 /* Still some bytes left; set up for a retry later. */
262965f5 977 status = -EAGAIN;
2199700f 978 }
a246b010 979
3dedbb5c 980process_status:
262965f5 981 switch (status) {
fba91afb
TM		 982 case -ENOTSOCK:
983 status = -ENOTCONN;
984 /* Should we call xs_close() here? */
985 break;
b6ddf64f 986 case -EAGAIN:
7496b59f 987 status = xs_sock_nospace(req);
b6ddf64f 988 break;
262965f5 989 case -ENETUNREACH:
3601c4a9 990 case -ENOBUFS:
262965f5 991 case -EPIPE:
a246b010 992 case -ECONNREFUSED:
3dedbb5c 993 case -EPERM:
a246b010 994 /* When the server has died, an ICMP port unreachable message
9903cd1c 995 * prompts ECONNREFUSED. */
13331a55
TM		 996 break;
997 default:
998 dprintk("RPC: sendmsg returned unrecognized error %d\n",
999 -status);
a246b010 1000 }
5fe46e9d 1001
262965f5 1002 return status;
a246b010
CL		 1003}
1004
9903cd1c 1005/**
262965f5 1006 * xs_tcp_send_request - write an RPC request to a TCP socket
50f484e2 1007 * @req: pointer to RPC request
9903cd1c
CL		 1008 *
1009 * Return values:
262965f5
CL		 1010 * 0: The request has been sent
1011 * EAGAIN: The socket was blocked, please call again later to
1012 * complete the request
1013 * ENOTCONN: Caller needs to invoke connect logic then call again
25985edc 1014 * other: Some other error occurred, the request was not sent
9903cd1c
CL		 1015 *
1016 * XXX: In the case of soft timeouts, should we eventually give up
262965f5 1017 * if sendmsg is not able to make progress?
9903cd1c 1018 */
adfa7144 1019static int xs_tcp_send_request(struct rpc_rqst *req)
a246b010 1020{
a246b010 1021 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 1022 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 1023 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 1024 rpc_fraghdr rm = xs_stream_record_marker(xdr);
1025 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
9e55eef4
CL		 1026 struct msghdr msg = {
1027 .msg_flags = XS_SENDMSG_FLAGS,
1028 };
d0afde5f 1029 bool vm_wait;
3f649ab7 1030 unsigned int sent;
b595bb15 1031 int status;
a246b010 1032
4cd34e7c
TM		 1033 /* Close the stream if the previous transmission was incomplete */
1034 if (xs_send_request_was_aborted(transport, req)) {
1035 if (transport->sock != NULL)
1036 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
1037 return -ENOTCONN;
1038 }
d275880a
TM		 1039 if (!transport->inet)
1040 return -ENOTCONN;
4cd34e7c 1041
262965f5
CL		 1042 xs_pktdump("packet data:",
1043 req->rq_svec->iov_base,
1044 req->rq_svec->iov_len);
a246b010 1045
7196dbb0
TM		 1046 if (test_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state))
1047 xs_tcp_set_socket_timeouts(xprt, transport->sock);
1048
3b21f757
TM		 1049 xs_set_srcport(transport, transport->sock);
1050
a246b010
CL		 1051 /* Continue transmitting the packet/record. We must be careful
1052 * to cope with writespace callbacks arriving _after_ we have
262965f5 1053 * called sendmsg(). */
78215759 1054 req->rq_xtime = ktime_get();
d737e5d4 1055 tcp_sock_set_cork(transport->inet, true);
d0afde5f
TM		 1056
1057 vm_wait = sk_stream_is_writeable(transport->inet) ? true : false;
1058
1059 do {
9e55eef4
CL		 1060 status = xprt_sock_sendmsg(transport->sock, &msg, xdr,
1061 transport->xmit.offset, rm, &sent);
a246b010 1062
46121cf7 1063 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
6c7a64e5 1064 xdr->len - transport->xmit.offset, status);
a246b010 1065
262965f5
CL		 1066 /* If we've sent the entire packet, immediately
1067 * reset the count of bytes sent. */
6c7a64e5
TM		 1068 transport->xmit.offset += sent;
1069 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 1070 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 1071 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 1072 transport->xmit.offset = 0;
d737e5d4
TM		 1073 if (atomic_long_read(&xprt->xmit_queuelen) == 1)
1074 tcp_sock_set_cork(transport->inet, false);
262965f5
CL		 1075 return 0;
1076 }
a246b010 1077
9ffadfbc
TM		 1078 WARN_ON_ONCE(sent == 0 && status == 0);
1079
d0afde5f
TM		 1080 if (sent > 0)
1081 vm_wait = false;
1082
1083 } while (status == 0);
a246b010 1084
262965f5 1085 switch (status) {
fba91afb
TM		 1086 case -ENOTSOCK:
1087 status = -ENOTCONN;
1088 /* Should we call xs_close() here? */
1089 break;
262965f5 1090 case -EAGAIN:
d0afde5f 1091 status = xs_stream_nospace(req, vm_wait);
262965f5 1092 break;
262965f5 1093 case -ECONNRESET:
2a9e1cfa 1094 case -ECONNREFUSED:
262965f5 1095 case -ENOTCONN:
3913c78c 1096 case -EADDRINUSE:
b5872f0c 1097 case -ENOBUFS:
b9d2bb2e 1098 case -EPIPE:
13331a55
TM		 1099 break;
1100 default:
1101 dprintk("RPC: sendmsg returned unrecognized error %d\n",
1102 -status);
a246b010 1103 }
5fe46e9d 1104
a246b010
CL		 1105 return status;
1106}
1107
2a9e1cfa
TM		 1108static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1109{
1110 transport->old_data_ready = sk->sk_data_ready;
1111 transport->old_state_change = sk->sk_state_change;
1112 transport->old_write_space = sk->sk_write_space;
2118071d 1113 transport->old_error_report = sk->sk_error_report;
2a9e1cfa
TM		 1114}
1115
1116static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1117{
1118 sk->sk_data_ready = transport->old_data_ready;
1119 sk->sk_state_change = transport->old_state_change;
1120 sk->sk_write_space = transport->old_write_space;
2118071d
TM		 1121 sk->sk_error_report = transport->old_error_report;
1122}
1123
42d42a5b
TM		 1124static void xs_sock_reset_state_flags(struct rpc_xprt *xprt)
1125{
1126 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1127
1128 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
4f8943f8
TM		 1129 clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state);
1130 clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state);
1131 clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state);
2790a624 1132 clear_bit(XPRT_SOCK_NOSPACE, &transport->sock_state);
4f8943f8
TM		 1133}
1134
1135static void xs_run_error_worker(struct sock_xprt *transport, unsigned int nr)
1136{
1137 set_bit(nr, &transport->sock_state);
1138 queue_work(xprtiod_workqueue, &transport->error_worker);
42d42a5b
TM		 1139}
1140
b70ae915
TM		 1141static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
1142{
d896ba83 1143 xprt->connect_cookie++;
b70ae915
TM		 1144 smp_mb__before_atomic();
1145 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1146 clear_bit(XPRT_CLOSING, &xprt->state);
42d42a5b 1147 xs_sock_reset_state_flags(xprt);
b70ae915
TM		 1148 smp_mb__after_atomic();
1149}
1150
2118071d
TM		 1151/**
1152 * xs_error_report - callback to handle TCP socket state errors
1153 * @sk: socket
1154 *
1155 * Note: we don't call sock_error() since there may be a rpc_task
1156 * using the socket, and so we don't want to clear sk->sk_err.
1157 */
1158static void xs_error_report(struct sock *sk)
1159{
4f8943f8 1160 struct sock_xprt *transport;
2118071d 1161 struct rpc_xprt *xprt;
2118071d 1162
2118071d 1163 if (!(xprt = xprt_from_sock(sk)))
ea9afca8 1164 return;
2118071d 1165
4f8943f8 1166 transport = container_of(xprt, struct sock_xprt, xprt);
af84537d
BC		 1167 transport->xprt_err = -sk->sk_err;
1168 if (transport->xprt_err == 0)
ea9afca8 1169 return;
2118071d 1170 dprintk("RPC: xs_error_report client %p, error=%d...\n",
af84537d
BC		 1171 xprt, -transport->xprt_err);
1172 trace_rpc_socket_error(xprt, sk->sk_socket, transport->xprt_err);
1173
1174 /* barrier ensures xprt_err is set before XPRT_SOCK_WAKE_ERROR */
1175 smp_mb__before_atomic();
4f8943f8 1176 xs_run_error_worker(transport, XPRT_SOCK_WAKE_ERROR);
2a9e1cfa
TM		 1177}
1178
fe315e76 1179static void xs_reset_transport(struct sock_xprt *transport)
a246b010 1180{
ee0ac0c2
CL		 1181 struct socket *sock = transport->sock;
1182 struct sock *sk = transport->inet;
6cc7e908 1183 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1184 struct file *filp = transport->file;
a246b010 1185
fe315e76
CL		 1186 if (sk == NULL)
1187 return;
f0043206
TM		 1188 /*
1189 * Make sure we're calling this in a context from which it is safe
1190 * to call __fput_sync(). In practice that means rpciod and the
1191 * system workqueue.
1192 */
1193 if (!(current->flags & PF_WQ_WORKER)) {
1194 WARN_ON_ONCE(1);
1195 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
1196 return;
1197 }
9903cd1c 1198
264d1df3
JL		 1199 if (atomic_read(&transport->xprt.swapper))
1200 sk_clear_memalloc(sk);
1201
09939204
TM		 1202 kernel_sock_shutdown(sock, SHUT_RDWR);
1203
edc1b01c 1204 mutex_lock(&transport->recv_mutex);
ea9afca8 1205 lock_sock(sk);
ee0ac0c2
CL		 1206 transport->inet = NULL;
1207 transport->sock = NULL;
a73881c9 1208 transport->file = NULL;
a246b010 1209
9903cd1c 1210 sk->sk_user_data = NULL;
2a9e1cfa
TM		 1211
1212 xs_restore_old_callbacks(transport, sk);
0c78789e 1213 xprt_clear_connected(xprt);
6cc7e908 1214 xs_sock_reset_connection_flags(xprt);
ae053551
TM		 1215 /* Reset stream record info */
1216 xs_stream_reset_connect(transport);
ea9afca8 1217 release_sock(sk);
edc1b01c 1218 mutex_unlock(&transport->recv_mutex);
a246b010 1219
6cc7e908 1220 trace_rpc_socket_close(xprt, sock);
f0043206 1221 __fput_sync(filp);
0445f92c
TM		 1222
1223 xprt_disconnect_done(xprt);
fe315e76
CL		 1224}
1225
1226/**
1227 * xs_close - close a socket
1228 * @xprt: transport
1229 *
1230 * This is used when all requests are complete; ie, no DRC state remains
1231 * on the server we want to save.
f75e6745
TM		 1232 *
1233 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
1234 * xs_reset_transport() zeroing the socket from underneath a writer.
fe315e76
CL		 1235 */
1236static void xs_close(struct rpc_xprt *xprt)
1237{
1238 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1239
1240 dprintk("RPC: xs_close xprt %p\n", xprt);
1241
1242 xs_reset_transport(transport);
61d0a8e6 1243 xprt->reestablish_timeout = 0;
a246b010
CL		 1244}
1245
4a068258
CL		 1246static void xs_inject_disconnect(struct rpc_xprt *xprt)
1247{
1248 dprintk("RPC: injecting transport disconnect on xprt=%p\n",
1249 xprt);
1250 xprt_disconnect_done(xprt);
1251}
1252
315f3812
KM		 1253static void xs_xprt_free(struct rpc_xprt *xprt)
1254{
1255 xs_free_peer_addresses(xprt);
1256 xprt_free(xprt);
1257}
1258
9903cd1c
CL		 1259/**
1260 * xs_destroy - prepare to shutdown a transport
1261 * @xprt: doomed transport
1262 *
1263 */
1264static void xs_destroy(struct rpc_xprt *xprt)
a246b010 1265{
03c78827
TM		 1266 struct sock_xprt *transport = container_of(xprt,
1267 struct sock_xprt, xprt);
46121cf7 1268 dprintk("RPC: xs_destroy xprt %p\n", xprt);
9903cd1c 1269
03c78827 1270 cancel_delayed_work_sync(&transport->connect_worker);
a1311d87 1271 xs_close(xprt);
edc1b01c 1272 cancel_work_sync(&transport->recv_worker);
b5e92419 1273 cancel_work_sync(&transport->error_worker);
315f3812 1274 xs_xprt_free(xprt);
a1311d87 1275 module_put(THIS_MODULE);
a246b010
CL		 1276}
1277
9903cd1c 1278/**
f9b2ee71
TM		 1279 * xs_udp_data_read_skb - receive callback for UDP sockets
1280 * @xprt: transport
1281 * @sk: socket
1282 * @skb: skbuff
9903cd1c 1283 *
a246b010 1284 */
f9b2ee71
TM		 1285static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
1286 struct sock *sk,
1287 struct sk_buff *skb)
a246b010 1288{
9903cd1c 1289 struct rpc_task *task;
a246b010 1290 struct rpc_rqst *rovr;
f9b2ee71 1291 int repsize, copied;
d8ed029d
AD		 1292 u32 _xid;
1293 __be32 *xp;
a246b010 1294
1da8c681 1295 repsize = skb->len;
a246b010 1296 if (repsize < 4) {
46121cf7 1297 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
f9b2ee71 1298 return;
a246b010
CL		 1299 }
1300
1301 /* Copy the XID from the skb... */
1da8c681 1302 xp = skb_header_pointer(skb, 0, sizeof(_xid), &_xid);
a246b010 1303 if (xp == NULL)
f9b2ee71 1304 return;
a246b010
CL		 1305
1306 /* Look up and lock the request corresponding to the given XID */
75c84151 1307 spin_lock(&xprt->queue_lock);
a246b010
CL		 1308 rovr = xprt_lookup_rqst(xprt, *xp);
1309 if (!rovr)
1310 goto out_unlock;
729749bb 1311 xprt_pin_rqst(rovr);
ecd465ee 1312 xprt_update_rtt(rovr->rq_task);
75c84151 1313 spin_unlock(&xprt->queue_lock);
a246b010
CL		 1314 task = rovr->rq_task;
1315
a246b010
CL		 1316 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1317 copied = repsize;
1318
1319 /* Suck it into the iovec, verify checksum if not done by hw. */
1781f7f5 1320 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
75c84151 1321 spin_lock(&xprt->queue_lock);
0afa6b44 1322 __UDPX_INC_STATS(sk, UDP_MIB_INERRORS);
729749bb 1323 goto out_unpin;
1781f7f5
HX		 1324 }
1325
a246b010 1326
b5e92419 1327 spin_lock(&xprt->transport_lock);
6a24dfb6 1328 xprt_adjust_cwnd(xprt, task, copied);
b5e92419 1329 spin_unlock(&xprt->transport_lock);
75c84151 1330 spin_lock(&xprt->queue_lock);
1570c1e4 1331 xprt_complete_rqst(task, copied);
0afa6b44 1332 __UDPX_INC_STATS(sk, UDP_MIB_INDATAGRAMS);
729749bb
TM		 1333out_unpin:
1334 xprt_unpin_rqst(rovr);
a246b010 1335 out_unlock:
75c84151 1336 spin_unlock(&xprt->queue_lock);
f9b2ee71
TM		 1337}
1338
1339static void xs_udp_data_receive(struct sock_xprt *transport)
1340{
1341 struct sk_buff *skb;
1342 struct sock *sk;
1343 int err;
1344
1345 mutex_lock(&transport->recv_mutex);
1346 sk = transport->inet;
1347 if (sk == NULL)
1348 goto out;
1349 for (;;) {
ec095263 1350 skb = skb_recv_udp(sk, MSG_DONTWAIT, &err);
4f546149 1351 if (skb == NULL)
f9b2ee71 1352 break;
4f546149
TM		 1353 xs_udp_data_read_skb(&transport->xprt, sk, skb);
1354 consume_skb(skb);
1355 cond_resched();
f9b2ee71 1356 }
0ffe86f4 1357 xs_poll_check_readable(transport);
f9b2ee71
TM		 1358out:
1359 mutex_unlock(&transport->recv_mutex);
1360}
1361
1362static void xs_udp_data_receive_workfn(struct work_struct *work)
1363{
1364 struct sock_xprt *transport =
1365 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 1366 unsigned int pflags = memalloc_nofs_save();
1367
f9b2ee71 1368 xs_udp_data_receive(transport);
a1231fda 1369 memalloc_nofs_restore(pflags);
f9b2ee71
TM		 1370}
1371
1372/**
f67939e4 1373 * xs_data_ready - "data ready" callback for sockets
f9b2ee71
TM		 1374 * @sk: socket with data to read
1375 *
1376 */
1377static void xs_data_ready(struct sock *sk)
1378{
1379 struct rpc_xprt *xprt;
1380
f9b2ee71
TM		 1381 xprt = xprt_from_sock(sk);
1382 if (xprt != NULL) {
1383 struct sock_xprt *transport = container_of(xprt,
1384 struct sock_xprt, xprt);
f67939e4
CL		 1385
1386 trace_xs_data_ready(xprt);
1387
5157b956
TM		 1388 transport->old_data_ready(sk);
1389 /* Any data means we had a useful conversation, so
1390 * then we don't need to delay the next reconnect
1391 */
1392 if (xprt->reestablish_timeout)
1393 xprt->reestablish_timeout = 0;
42d42a5b 1394 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
40a5f1b1 1395 queue_work(xprtiod_workqueue, &transport->recv_worker);
f9b2ee71 1396 }
a246b010
CL		 1397}
1398
a519fc7a
TM		 1399/*
1400 * Helper function to force a TCP close if the server is sending
1401 * junk and/or it has put us in CLOSE_WAIT
1402 */
1403static void xs_tcp_force_close(struct rpc_xprt *xprt)
1404{
a519fc7a
TM		 1405 xprt_force_disconnect(xprt);
1406}
1407
9e00abc3 1408#if defined(CONFIG_SUNRPC_BACKCHANNEL)
6b26cc8c
CL		 1409static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt)
1410{
1411 return PAGE_SIZE;
1412}
9e00abc3 1413#endif /* CONFIG_SUNRPC_BACKCHANNEL */
44b98efd 1414
efce2d0b
TM		 1415/**
1416 * xs_local_state_change - callback to handle AF_LOCAL socket state changes
1417 * @sk: socket whose state has changed
1418 *
1419 */
1420static void xs_local_state_change(struct sock *sk)
1421{
1422 struct rpc_xprt *xprt;
1423 struct sock_xprt *transport;
1424
1425 if (!(xprt = xprt_from_sock(sk)))
1426 return;
1427 transport = container_of(xprt, struct sock_xprt, xprt);
1428 if (sk->sk_shutdown & SHUTDOWN_MASK) {
1429 clear_bit(XPRT_CONNECTED, &xprt->state);
1430 /* Trigger the socket release */
1431 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
1432 }
1433}
1434
9903cd1c
CL		 1435/**
1436 * xs_tcp_state_change - callback to handle TCP socket state changes
1437 * @sk: socket whose state has changed
1438 *
1439 */
1440static void xs_tcp_state_change(struct sock *sk)
a246b010 1441{
9903cd1c 1442 struct rpc_xprt *xprt;
0fdea1e8 1443 struct sock_xprt *transport;
a246b010 1444
a246b010 1445 if (!(xprt = xprt_from_sock(sk)))
ea9afca8 1446 return;
46121cf7 1447 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
669502ff 1448 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
46121cf7
CL		 1449 sk->sk_state, xprt_connected(xprt),
1450 sock_flag(sk, SOCK_DEAD),
669502ff
AC		 1451 sock_flag(sk, SOCK_ZAPPED),
1452 sk->sk_shutdown);
a246b010 1453
0fdea1e8 1454 transport = container_of(xprt, struct sock_xprt, xprt);
40b5ea0c 1455 trace_rpc_socket_state_change(xprt, sk->sk_socket);
a246b010
CL		 1456 switch (sk->sk_state) {
1457 case TCP_ESTABLISHED:
a246b010 1458 if (!xprt_test_and_set_connected(xprt)) {
8b71798c 1459 xprt->connect_cookie++;
0fdea1e8
TM		 1460 clear_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
1461 xprt_clear_connecting(xprt);
51971139 1462
3968a8a5
CL		 1463 xprt->stat.connect_count++;
1464 xprt->stat.connect_time += (long)jiffies -
1465 xprt->stat.connect_start;
4f8943f8 1466 xs_run_error_worker(transport, XPRT_SOCK_WAKE_PENDING);
a246b010 1467 }
a246b010 1468 break;
3b948ae5
TM		 1469 case TCP_FIN_WAIT1:
1470 /* The client initiated a shutdown of the socket */
7c1d71cf 1471 xprt->connect_cookie++;
663b8858 1472 xprt->reestablish_timeout = 0;
3b948ae5 1473 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1474 smp_mb__before_atomic();
3b948ae5 1475 clear_bit(XPRT_CONNECTED, &xprt->state);
ef803670 1476 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
4e857c58 1477 smp_mb__after_atomic();
a246b010 1478 break;
632e3bdc 1479 case TCP_CLOSE_WAIT:
3b948ae5 1480 /* The server initiated a shutdown of the socket */
7c1d71cf 1481 xprt->connect_cookie++;
d0bea455 1482 clear_bit(XPRT_CONNECTED, &xprt->state);
4f8943f8 1483 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
df561f66 1484 fallthrough;
663b8858
TM		 1485 case TCP_CLOSING:
1486 /*
1487 * If the server closed down the connection, make sure that
1488 * we back off before reconnecting
1489 */
1490 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1491 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3b948ae5
TM		 1492 break;
1493 case TCP_LAST_ACK:
670f9457 1494 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1495 smp_mb__before_atomic();
3b948ae5 1496 clear_bit(XPRT_CONNECTED, &xprt->state);
4e857c58 1497 smp_mb__after_atomic();
3b948ae5
TM		 1498 break;
1499 case TCP_CLOSE:
0fdea1e8
TM		 1500 if (test_and_clear_bit(XPRT_SOCK_CONNECTING,
1501 &transport->sock_state))
1502 xprt_clear_connecting(xprt);
9b30889c 1503 clear_bit(XPRT_CLOSING, &xprt->state);
9b30889c 1504 /* Trigger the socket release */
4f8943f8 1505 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
a246b010 1506 }
a246b010
CL		 1507}
1508
1f0fa154
IJ		 1509static void xs_write_space(struct sock *sk)
1510{
4f8943f8 1511 struct sock_xprt *transport;
1f0fa154
IJ		 1512 struct rpc_xprt *xprt;
1513
13331a55 1514 if (!sk->sk_socket)
1f0fa154 1515 return;
13331a55 1516 clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1f0fa154
IJ		 1517
1518 if (unlikely(!(xprt = xprt_from_sock(sk))))
1519 return;
4f8943f8 1520 transport = container_of(xprt, struct sock_xprt, xprt);
2790a624
TM		 1521 if (!test_and_clear_bit(XPRT_SOCK_NOSPACE, &transport->sock_state))
1522 return;
4f8943f8
TM		 1523 xs_run_error_worker(transport, XPRT_SOCK_WAKE_WRITE);
1524 sk->sk_write_pending--;
1f0fa154
IJ		 1525}
1526
9903cd1c 1527/**
c7b2cae8
CL		 1528 * xs_udp_write_space - callback invoked when socket buffer space
1529 * becomes available
9903cd1c
CL		 1530 * @sk: socket whose state has changed
1531 *
a246b010
CL		 1532 * Called when more output buffer space is available for this socket.
1533 * We try not to wake our writers until they can make "significant"
c7b2cae8 1534 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
a246b010
CL		 1535 * with a bunch of small requests.
1536 */
c7b2cae8 1537static void xs_udp_write_space(struct sock *sk)
a246b010 1538{
c7b2cae8 1539 /* from net/core/sock.c:sock_def_write_space */
1f0fa154
IJ		 1540 if (sock_writeable(sk))
1541 xs_write_space(sk);
c7b2cae8 1542}
a246b010 1543
c7b2cae8
CL		 1544/**
1545 * xs_tcp_write_space - callback invoked when socket buffer space
1546 * becomes available
1547 * @sk: socket whose state has changed
1548 *
1549 * Called when more output buffer space is available for this socket.
1550 * We try not to wake our writers until they can make "significant"
1551 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1552 * with a bunch of small requests.
1553 */
1554static void xs_tcp_write_space(struct sock *sk)
1555{
c7b2cae8 1556 /* from net/core/stream.c:sk_stream_write_space */
64dc6130 1557 if (sk_stream_is_writeable(sk))
1f0fa154 1558 xs_write_space(sk);
a246b010
CL		 1559}
1560
470056c2 1561static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
a246b010 1562{
ee0ac0c2
CL		 1563 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1564 struct sock *sk = transport->inet;
a246b010 1565
7c6e066e 1566 if (transport->rcvsize) {
a246b010 1567 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
7c6e066e 1568 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
a246b010 1569 }
7c6e066e 1570 if (transport->sndsize) {
a246b010 1571 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
7c6e066e 1572 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
a246b010
CL		 1573 sk->sk_write_space(sk);
1574 }
1575}
1576
43118c29 1577/**
470056c2 1578 * xs_udp_set_buffer_size - set send and receive limits
43118c29 1579 * @xprt: generic transport
470056c2
CL		 1580 * @sndsize: requested size of send buffer, in bytes
1581 * @rcvsize: requested size of receive buffer, in bytes
43118c29 1582 *
470056c2 1583 * Set socket send and receive buffer size limits.
43118c29 1584 */
470056c2 1585static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
43118c29 1586{
7c6e066e
CL		 1587 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1588
1589 transport->sndsize = 0;
470056c2 1590 if (sndsize)
7c6e066e
CL		 1591 transport->sndsize = sndsize + 1024;
1592 transport->rcvsize = 0;
470056c2 1593 if (rcvsize)
7c6e066e 1594 transport->rcvsize = rcvsize + 1024;
470056c2
CL		 1595
1596 xs_udp_do_set_buffer_size(xprt);
43118c29
CL		 1597}
1598
46c0ee8b
CL		 1599/**
1600 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
acf0a39f 1601 * @xprt: controlling transport
46c0ee8b
CL		 1602 * @task: task that timed out
1603 *
1604 * Adjust the congestion window after a retransmit timeout has occurred.
1605 */
6a24dfb6 1606static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
46c0ee8b 1607{
b5e92419 1608 spin_lock(&xprt->transport_lock);
6a24dfb6 1609 xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
b5e92419 1610 spin_unlock(&xprt->transport_lock);
46c0ee8b
CL		 1611}
1612
826799e6 1613static int xs_get_random_port(void)
b85d8806 1614{
826799e6
BF		 1615 unsigned short min = xprt_min_resvport, max = xprt_max_resvport;
1616 unsigned short range;
1617 unsigned short rand;
1618
1619 if (max < min)
1620 return -EADDRINUSE;
1621 range = max - min + 1;
81895a65 1622 rand = prandom_u32_max(range);
826799e6 1623 return rand + min;
b85d8806
CL		 1624}
1625
4dda9c8a
TM		 1626static unsigned short xs_sock_getport(struct socket *sock)
1627{
1628 struct sockaddr_storage buf;
4dda9c8a
TM		 1629 unsigned short port = 0;
1630
9b2c45d4 1631 if (kernel_getsockname(sock, (struct sockaddr *)&buf) < 0)
4dda9c8a
TM		 1632 goto out;
1633 switch (buf.ss_family) {
1634 case AF_INET6:
1635 port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port);
1636 break;
1637 case AF_INET:
1638 port = ntohs(((struct sockaddr_in *)&buf)->sin_port);
1639 }
1640out:
1641 return port;
1642}
1643
92200412
CL		 1644/**
1645 * xs_set_port - reset the port number in the remote endpoint address
1646 * @xprt: generic transport
1647 * @port: new port number
1648 *
1649 */
1650static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1651{
46121cf7 1652 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
c4efcb1d 1653
9dc3b095
CL		 1654 rpc_set_port(xs_addr(xprt), port);
1655 xs_update_peer_port(xprt);
92200412
CL		 1656}
1657
4dda9c8a
TM		 1658static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
1659{
e6237b6f 1660 if (transport->srcport == 0 && transport->xprt.reuseport)
4dda9c8a
TM		 1661 transport->srcport = xs_sock_getport(sock);
1662}
1663
826799e6 1664static int xs_get_srcport(struct sock_xprt *transport)
67a391d7 1665{
826799e6 1666 int port = transport->srcport;
67a391d7
TM		 1667
1668 if (port == 0 && transport->xprt.resvport)
1669 port = xs_get_random_port();
1670 return port;
1671}
1672
421ab1be 1673static unsigned short xs_sock_srcport(struct rpc_xprt *xprt)
a8482488
OK		 1674{
1675 struct sock_xprt *sock = container_of(xprt, struct sock_xprt, xprt);
b49ea673
N		 1676 unsigned short ret = 0;
1677 mutex_lock(&sock->recv_mutex);
1678 if (sock->sock)
1679 ret = xs_sock_getport(sock->sock);
1680 mutex_unlock(&sock->recv_mutex);
1681 return ret;
a8482488 1682}
421ab1be
TM		 1683
1684static int xs_sock_srcaddr(struct rpc_xprt *xprt, char *buf, size_t buflen)
1685{
1686 struct sock_xprt *sock = container_of(xprt, struct sock_xprt, xprt);
1687 union {
1688 struct sockaddr sa;
1689 struct sockaddr_storage st;
1690 } saddr;
1691 int ret = -ENOTCONN;
1692
1693 mutex_lock(&sock->recv_mutex);
1694 if (sock->sock) {
1695 ret = kernel_getsockname(sock->sock, &saddr.sa);
1696 if (ret >= 0)
1697 ret = snprintf(buf, buflen, "%pISc", &saddr.sa);
1698 }
1699 mutex_unlock(&sock->recv_mutex);
1700 return ret;
1701}
a8482488 1702
baaf4e48 1703static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
67a391d7 1704{
fbfffbd5
CL		 1705 if (transport->srcport != 0)
1706 transport->srcport = 0;
67a391d7
TM		 1707 if (!transport->xprt.resvport)
1708 return 0;
1709 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1710 return xprt_max_resvport;
1711 return --port;
1712}
beb59b68 1713static int xs_bind(struct sock_xprt *transport, struct socket *sock)
a246b010 1714{
beb59b68 1715 struct sockaddr_storage myaddr;
67a391d7 1716 int err, nloop = 0;
826799e6 1717 int port = xs_get_srcport(transport);
67a391d7 1718 unsigned short last;
a246b010 1719
0f7a622c
CP		 1720 /*
1721 * If we are asking for any ephemeral port (i.e. port == 0 &&
1722 * transport->xprt.resvport == 0), don't bind. Let the local
1723 * port selection happen implicitly when the socket is used
1724 * (for example at connect time).
1725 *
1726 * This ensures that we can continue to establish TCP
1727 * connections even when all local ephemeral ports are already
1728 * a part of some TCP connection. This makes no difference
12b20ce3 1729 * for UDP sockets, but also doesn't harm them.
0f7a622c
CP		 1730 *
1731 * If we're asking for any reserved port (i.e. port == 0 &&
1732 * transport->xprt.resvport == 1) xs_get_srcport above will
1733 * ensure that port is non-zero and we will bind as needed.
1734 */
826799e6
BF		 1735 if (port <= 0)
1736 return port;
0f7a622c 1737
beb59b68 1738 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
a246b010 1739 do {
beb59b68
PE		 1740 rpc_set_port((struct sockaddr *)&myaddr, port);
1741 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1742 transport->xprt.addrlen);
a246b010 1743 if (err == 0) {
bc1c56e9
N		 1744 if (transport->xprt.reuseport)
1745 transport->srcport = port;
d3bc9a1d 1746 break;
a246b010 1747 }
67a391d7 1748 last = port;
baaf4e48 1749 port = xs_next_srcport(transport, port);
67a391d7
TM		 1750 if (port > last)
1751 nloop++;
1752 } while (err == -EADDRINUSE && nloop != 2);
90058d37 1753
4232e863 1754 if (myaddr.ss_family == AF_INET)
beb59b68
PE		 1755 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1756 &((struct sockaddr_in *)&myaddr)->sin_addr,
1757 port, err ? "failed" : "ok", err);
1758 else
1759 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1760 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1761 port, err ? "failed" : "ok", err);
a246b010
CL		 1762 return err;
1763}
1764
176e21ee
CL		 1765/*
1766 * We don't support autobind on AF_LOCAL sockets
1767 */
1768static void xs_local_rpcbind(struct rpc_task *task)
1769{
fb43d172 1770 xprt_set_bound(task->tk_xprt);
176e21ee
CL		 1771}
1772
1773static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1774{
1775}
beb59b68 1776
ed07536e 1777#ifdef CONFIG_DEBUG_LOCK_ALLOC
064a9177
N		 1778static struct lock_class_key xs_key[3];
1779static struct lock_class_key xs_slock_key[3];
ed07536e 1780
176e21ee
CL		 1781static inline void xs_reclassify_socketu(struct socket *sock)
1782{
1783 struct sock *sk = sock->sk;
1784
176e21ee 1785 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
064a9177 1786 &xs_slock_key[0], "sk_lock-AF_LOCAL-RPC", &xs_key[0]);
176e21ee
CL		 1787}
1788
8945ee5e 1789static inline void xs_reclassify_socket4(struct socket *sock)
ed07536e
PZ		 1790{
1791 struct sock *sk = sock->sk;
8945ee5e 1792
8945ee5e 1793 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
064a9177 1794 &xs_slock_key[1], "sk_lock-AF_INET-RPC", &xs_key[1]);
8945ee5e 1795}
ed07536e 1796
8945ee5e
CL		 1797static inline void xs_reclassify_socket6(struct socket *sock)
1798{
1799 struct sock *sk = sock->sk;
ed07536e 1800
8945ee5e 1801 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
064a9177 1802 &xs_slock_key[2], "sk_lock-AF_INET6-RPC", &xs_key[2]);
ed07536e 1803}
6bc9638a
PE		 1804
1805static inline void xs_reclassify_socket(int family, struct socket *sock)
1806{
fafc4e1e 1807 if (WARN_ON_ONCE(!sock_allow_reclassification(sock->sk)))
1b7a1819
WAA		 1808 return;
1809
4232e863 1810 switch (family) {
176e21ee
CL		 1811 case AF_LOCAL:
1812 xs_reclassify_socketu(sock);
1813 break;
4232e863 1814 case AF_INET:
6bc9638a 1815 xs_reclassify_socket4(sock);
4232e863
CL		 1816 break;
1817 case AF_INET6:
6bc9638a 1818 xs_reclassify_socket6(sock);
4232e863
CL		 1819 break;
1820 }
6bc9638a 1821}
ed07536e 1822#else
6bc9638a
PE		 1823static inline void xs_reclassify_socket(int family, struct socket *sock)
1824{
1825}
ed07536e
PZ		 1826#endif
1827
93dc41bd
N		 1828static void xs_dummy_setup_socket(struct work_struct *work)
1829{
1830}
1831
6bc9638a 1832static struct socket *xs_create_sock(struct rpc_xprt *xprt,
4dda9c8a
TM		 1833 struct sock_xprt *transport, int family, int type,
1834 int protocol, bool reuseport)
22f79326 1835{
a73881c9 1836 struct file *filp;
22f79326
PE		 1837 struct socket *sock;
1838 int err;
1839
6bc9638a 1840 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
22f79326
PE		 1841 if (err < 0) {
1842 dprintk("RPC: can't create %d transport socket (%d).\n",
1843 protocol, -err);
1844 goto out;
1845 }
6bc9638a 1846 xs_reclassify_socket(family, sock);
22f79326 1847
4dda9c8a 1848 if (reuseport)
fe31a326 1849 sock_set_reuseport(sock->sk);
4dda9c8a 1850
4cea288a
BH		 1851 err = xs_bind(transport, sock);
1852 if (err) {
22f79326
PE		 1853 sock_release(sock);
1854 goto out;
1855 }
1856
a73881c9
TM		 1857 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1858 if (IS_ERR(filp))
1859 return ERR_CAST(filp);
1860 transport->file = filp;
1861
22f79326
PE		 1862 return sock;
1863out:
1864 return ERR_PTR(err);
1865}
1866
176e21ee
CL		 1867static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1868 struct socket *sock)
1869{
1870 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1871 xprt);
1872
1873 if (!transport->inet) {
1874 struct sock *sk = sock->sk;
1875
ea9afca8 1876 lock_sock(sk);
176e21ee
CL		 1877
1878 xs_save_old_callbacks(transport, sk);
1879
1880 sk->sk_user_data = xprt;
a2648094 1881 sk->sk_data_ready = xs_data_ready;
176e21ee 1882 sk->sk_write_space = xs_udp_write_space;
efce2d0b 1883 sk->sk_state_change = xs_local_state_change;
2118071d 1884 sk->sk_error_report = xs_error_report;
176e21ee
CL		 1885
1886 xprt_clear_connected(xprt);
1887
1888 /* Reset to new socket */
1889 transport->sock = sock;
1890 transport->inet = sk;
1891
ea9afca8 1892 release_sock(sk);
176e21ee
CL		 1893 }
1894
ae053551 1895 xs_stream_start_connect(transport);
6c7a64e5 1896
176e21ee
CL		 1897 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1898}
1899
1900/**
1901 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
176e21ee 1902 * @transport: socket transport to connect
176e21ee 1903 */
dc107402 1904static int xs_local_setup_socket(struct sock_xprt *transport)
176e21ee 1905{
176e21ee 1906 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1907 struct file *filp;
176e21ee 1908 struct socket *sock;
68e9a246 1909 int status;
176e21ee 1910
176e21ee
CL		 1911 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1912 SOCK_STREAM, 0, &sock, 1);
1913 if (status < 0) {
1914 dprintk("RPC: can't create AF_LOCAL "
1915 "transport socket (%d).\n", -status);
1916 goto out;
1917 }
d1358917 1918 xs_reclassify_socket(AF_LOCAL, sock);
176e21ee 1919
a73881c9
TM		 1920 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1921 if (IS_ERR(filp)) {
1922 status = PTR_ERR(filp);
1923 goto out;
1924 }
1925 transport->file = filp;
1926
176e21ee
CL		 1927 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1928 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1929
1930 status = xs_local_finish_connecting(xprt, sock);
40b5ea0c 1931 trace_rpc_socket_connect(xprt, sock, status);
176e21ee
CL		 1932 switch (status) {
1933 case 0:
1934 dprintk("RPC: xprt %p connected to %s\n",
1935 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
3968a8a5
CL		 1936 xprt->stat.connect_count++;
1937 xprt->stat.connect_time += (long)jiffies -
1938 xprt->stat.connect_start;
176e21ee 1939 xprt_set_connected(xprt);
93f479d3 1940 break;
3601c4a9 1941 case -ENOBUFS:
176e21ee
CL		 1942 break;
1943 case -ENOENT:
1944 dprintk("RPC: xprt %p: socket %s does not exist\n",
1945 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1946 break;
4a20a988
TM		 1947 case -ECONNREFUSED:
1948 dprintk("RPC: xprt %p: connection refused for %s\n",
1949 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1950 break;
176e21ee
CL		 1951 default:
1952 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
1953 __func__, -status,
1954 xprt->address_strings[RPC_DISPLAY_ADDR]);
1955 }
1956
1957out:
1958 xprt_clear_connecting(xprt);
1959 xprt_wake_pending_tasks(xprt, status);
dc107402
BF		 1960 return status;
1961}
1962
b6669737 1963static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
dc107402 1964{
dc107402
BF		 1965 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1966 int ret;
1967
aad41a7d
TM		 1968 if (transport->file)
1969 goto force_disconnect;
1970
c4f03966 1971 if (RPC_IS_ASYNC(task)) {
dc107402
BF		 1972 /*
1973 * We want the AF_LOCAL connect to be resolved in the
1974 * filesystem namespace of the process making the rpc
1975 * call. Thus we connect synchronously.
1976 *
1977 * If we want to support asynchronous AF_LOCAL calls,
1978 * we'll need to figure out how to pass a namespace to
1979 * connect.
1980 */
5ad64b36 1981 task->tk_rpc_status = -ENOTCONN;
dc107402 1982 rpc_exit(task, -ENOTCONN);
aad41a7d 1983 goto out_wake;
dc107402
BF		 1984 }
1985 ret = xs_local_setup_socket(transport);
1986 if (ret && !RPC_IS_SOFTCONN(task))
1987 msleep_interruptible(15000);
aad41a7d
TM		 1988 return;
1989force_disconnect:
1990 xprt_force_disconnect(xprt);
1991out_wake:
1992 xprt_clear_connecting(xprt);
1993 xprt_wake_pending_tasks(xprt, -ENOTCONN);
176e21ee
CL		 1994}
1995
3c87ef6e 1996#if IS_ENABLED(CONFIG_SUNRPC_SWAP)
d6e971d8 1997/*
693486d5
N		 1998 * Note that this should be called with XPRT_LOCKED held, or recv_mutex
1999 * held, or when we otherwise know that we have exclusive access to the
2000 * socket, to guard against races with xs_reset_transport.
d6e971d8 2001 */
a564b8f0
MG		 2002static void xs_set_memalloc(struct rpc_xprt *xprt)
2003{
2004 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
2005 xprt);
2006
d6e971d8
JL		 2007 /*
2008 * If there's no sock, then we have nothing to set. The
2009 * reconnecting process will get it for us.
2010 */
2011 if (!transport->inet)
2012 return;
8e228133 2013 if (atomic_read(&xprt->swapper))
a564b8f0
MG		 2014 sk_set_memalloc(transport->inet);
2015}
2016
2017/**
d67fa4d8 2018 * xs_enable_swap - Tag this transport as being used for swap.
a564b8f0 2019 * @xprt: transport to tag
a564b8f0 2020 *
8e228133
JL		 2021 * Take a reference to this transport on behalf of the rpc_clnt, and
2022 * optionally mark it for swapping if it wasn't already.
a564b8f0 2023 */
d67fa4d8
JL		 2024static int
2025xs_enable_swap(struct rpc_xprt *xprt)
a564b8f0 2026{
d6e971d8 2027 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
a564b8f0 2028
693486d5
N		 2029 mutex_lock(&xs->recv_mutex);
2030 if (atomic_inc_return(&xprt->swapper) == 1 &&
2031 xs->inet)
d6e971d8 2032 sk_set_memalloc(xs->inet);
693486d5 2033 mutex_unlock(&xs->recv_mutex);
8e228133
JL		 2034 return 0;
2035}
a564b8f0 2036
8e228133 2037/**
d67fa4d8 2038 * xs_disable_swap - Untag this transport as being used for swap.
8e228133
JL		 2039 * @xprt: transport to tag
2040 *
2041 * Drop a "swapper" reference to this xprt on behalf of the rpc_clnt. If the
2042 * swapper refcount goes to 0, untag the socket as a memalloc socket.
2043 */
d67fa4d8
JL		 2044static void
2045xs_disable_swap(struct rpc_xprt *xprt)
8e228133 2046{
d6e971d8 2047 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
8e228133 2048
693486d5
N		 2049 mutex_lock(&xs->recv_mutex);
2050 if (atomic_dec_and_test(&xprt->swapper) &&
2051 xs->inet)
d6e971d8 2052 sk_clear_memalloc(xs->inet);
693486d5 2053 mutex_unlock(&xs->recv_mutex);
a564b8f0 2054}
a564b8f0
MG		 2055#else
2056static void xs_set_memalloc(struct rpc_xprt *xprt)
2057{
2058}
d67fa4d8
JL		 2059
2060static int
2061xs_enable_swap(struct rpc_xprt *xprt)
2062{
2063 return -EINVAL;
2064}
2065
2066static void
2067xs_disable_swap(struct rpc_xprt *xprt)
2068{
2069}
a564b8f0
MG		 2070#endif
2071
16be2d20
CL		 2072static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
2073{
2074 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2075
2076 if (!transport->inet) {
2077 struct sock *sk = sock->sk;
2078
ea9afca8 2079 lock_sock(sk);
16be2d20 2080
2a9e1cfa
TM		 2081 xs_save_old_callbacks(transport, sk);
2082
16be2d20 2083 sk->sk_user_data = xprt;
f9b2ee71 2084 sk->sk_data_ready = xs_data_ready;
16be2d20 2085 sk->sk_write_space = xs_udp_write_space;
16be2d20
CL		 2086
2087 xprt_set_connected(xprt);
2088
2089 /* Reset to new socket */
2090 transport->sock = sock;
2091 transport->inet = sk;
2092
a564b8f0
MG		 2093 xs_set_memalloc(xprt);
2094
ea9afca8 2095 release_sock(sk);
16be2d20
CL		 2096 }
2097 xs_udp_do_set_buffer_size(xprt);
02910177
TM		 2098
2099 xprt->stat.connect_start = jiffies;
16be2d20
CL		 2100}
2101
8c14ff2a 2102static void xs_udp_setup_socket(struct work_struct *work)
a246b010 2103{
34161db6
TM		 2104 struct sock_xprt *transport =
2105 container_of(work, struct sock_xprt, connect_worker.work);
c8475461 2106 struct rpc_xprt *xprt = &transport->xprt;
d099b8af 2107 struct socket *sock;
b65c0310 2108 int status = -EIO;
8db55a03 2109 unsigned int pflags = current->flags;
9903cd1c 2110
8db55a03
N		 2111 if (atomic_read(&xprt->swapper))
2112 current->flags |= PF_MEMALLOC;
8c14ff2a 2113 sock = xs_create_sock(xprt, transport,
4dda9c8a
TM		 2114 xs_addr(xprt)->sa_family, SOCK_DGRAM,
2115 IPPROTO_UDP, false);
b65c0310 2116 if (IS_ERR(sock))
b0d93ad5 2117 goto out;
68e220bd 2118
c740eff8
CL		 2119 dprintk("RPC: worker connecting xprt %p via %s to "
2120 "%s (port %s)\n", xprt,
2121 xprt->address_strings[RPC_DISPLAY_PROTO],
2122 xprt->address_strings[RPC_DISPLAY_ADDR],
2123 xprt->address_strings[RPC_DISPLAY_PORT]);
68e220bd
CL		 2124
2125 xs_udp_finish_connecting(xprt, sock);
40b5ea0c 2126 trace_rpc_socket_connect(xprt, sock, 0);
b0d93ad5
CL		 2127 status = 0;
2128out:
b0d93ad5 2129 xprt_clear_connecting(xprt);
cf76785d 2130 xprt_unlock_connect(xprt, transport);
7d1e8255 2131 xprt_wake_pending_tasks(xprt, status);
8db55a03 2132 current_restore_flags(pflags, PF_MEMALLOC);
a246b010
CL		 2133}
2134
4876cc77
TM		 2135/**
2136 * xs_tcp_shutdown - gracefully shut down a TCP socket
2137 * @xprt: transport
2138 *
2139 * Initiates a graceful shutdown of the TCP socket by calling the
2140 * equivalent of shutdown(SHUT_RDWR);
2141 */
2142static void xs_tcp_shutdown(struct rpc_xprt *xprt)
2143{
2144 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2145 struct socket *sock = transport->sock;
9b30889c 2146 int skst = transport->inet ? transport->inet->sk_state : TCP_CLOSE;
4876cc77
TM		 2147
2148 if (sock == NULL)
2149 return;
0a6ff58e
TM		 2150 if (!xprt->reuseport) {
2151 xs_close(xprt);
2152 return;
2153 }
9b30889c 2154 switch (skst) {
7c81e6a9
TM		 2155 case TCP_FIN_WAIT1:
2156 case TCP_FIN_WAIT2:
2157 break;
2158 case TCP_ESTABLISHED:
2159 case TCP_CLOSE_WAIT:
4876cc77
TM		 2160 kernel_sock_shutdown(sock, SHUT_RDWR);
2161 trace_rpc_socket_shutdown(xprt, sock);
9b30889c 2162 break;
7c81e6a9 2163 default:
4876cc77 2164 xs_reset_transport(transport);
9b30889c 2165 }
4876cc77
TM		 2166}
2167
8d1b8c62
TM		 2168static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
2169 struct socket *sock)
2170{
7196dbb0
TM		 2171 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2172 unsigned int keepidle;
2173 unsigned int keepcnt;
8d1b8c62
TM		 2174 unsigned int timeo;
2175
b5e92419 2176 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2177 keepidle = DIV_ROUND_UP(xprt->timeout->to_initval, HZ);
2178 keepcnt = xprt->timeout->to_retries + 1;
2179 timeo = jiffies_to_msecs(xprt->timeout->to_initval) *
2180 (xprt->timeout->to_retries + 1);
2181 clear_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2182 spin_unlock(&xprt->transport_lock);
7196dbb0 2183
8d1b8c62 2184 /* TCP Keepalive options */
ce3d9544 2185 sock_set_keepalive(sock->sk);
71c48eb8 2186 tcp_sock_set_keepidle(sock->sk, keepidle);
d41ecaac 2187 tcp_sock_set_keepintvl(sock->sk, keepidle);
480aeb96 2188 tcp_sock_set_keepcnt(sock->sk, keepcnt);
8d1b8c62
TM		 2189
2190 /* TCP user timeout (see RFC5482) */
c488aead 2191 tcp_sock_set_user_timeout(sock->sk, timeo);
8d1b8c62
TM		 2192}
2193
7196dbb0
TM		 2194static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
2195 unsigned long connect_timeout,
2196 unsigned long reconnect_timeout)
2197{
2198 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2199 struct rpc_timeout to;
2200 unsigned long initval;
2201
b5e92419 2202 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2203 if (reconnect_timeout < xprt->max_reconnect_timeout)
2204 xprt->max_reconnect_timeout = reconnect_timeout;
2205 if (connect_timeout < xprt->connect_timeout) {
2206 memcpy(&to, xprt->timeout, sizeof(to));
2207 initval = DIV_ROUND_UP(connect_timeout, to.to_retries + 1);
2208 /* Arbitrary lower limit */
2209 if (initval < XS_TCP_INIT_REEST_TO << 1)
2210 initval = XS_TCP_INIT_REEST_TO << 1;
2211 to.to_initval = initval;
2212 to.to_maxval = initval;
2213 memcpy(&transport->tcp_timeout, &to,
2214 sizeof(transport->tcp_timeout));
2215 xprt->timeout = &transport->tcp_timeout;
2216 xprt->connect_timeout = connect_timeout;
2217 }
2218 set_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2219 spin_unlock(&xprt->transport_lock);
7196dbb0
TM		 2220}
2221
16be2d20 2222static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
a246b010 2223{
16be2d20 2224 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
edb267a6 2225
ee0ac0c2 2226 if (!transport->inet) {
b0d93ad5 2227 struct sock *sk = sock->sk;
7f260e85 2228
d88e4d82
N		 2229 /* Avoid temporary address, they are bad for long-lived
2230 * connections such as NFS mounts.
2231 * RFC4941, section 3.6 suggests that:
2232 * Individual applications, which have specific
2233 * knowledge about the normal duration of connections,
2234 * MAY override this as appropriate.
2235 */
18d5ad62
CH		 2236 if (xs_addr(xprt)->sa_family == PF_INET6) {
2237 ip6_sock_set_addr_preferences(sk,
2238 IPV6_PREFER_SRC_PUBLIC);
2239 }
d88e4d82 2240
8d1b8c62 2241 xs_tcp_set_socket_timeouts(xprt, sock);
d737e5d4 2242 tcp_sock_set_nodelay(sk);
775f06ab 2243
ea9afca8 2244 lock_sock(sk);
b0d93ad5 2245
2a9e1cfa
TM		 2246 xs_save_old_callbacks(transport, sk);
2247
b0d93ad5 2248 sk->sk_user_data = xprt;
5157b956 2249 sk->sk_data_ready = xs_data_ready;
b0d93ad5
CL		 2250 sk->sk_state_change = xs_tcp_state_change;
2251 sk->sk_write_space = xs_tcp_write_space;
2118071d 2252 sk->sk_error_report = xs_error_report;
3167e12c
CL		 2253
2254 /* socket options */
3167e12c 2255 sock_reset_flag(sk, SOCK_LINGER);
b0d93ad5
CL		 2256
2257 xprt_clear_connected(xprt);
2258
2259 /* Reset to new socket */
ee0ac0c2
CL		 2260 transport->sock = sock;
2261 transport->inet = sk;
b0d93ad5 2262
ea9afca8 2263 release_sock(sk);
b0d93ad5
CL		 2264 }
2265
01d37c42 2266 if (!xprt_bound(xprt))
280254b6 2267 return -ENOTCONN;
01d37c42 2268
a564b8f0
MG		 2269 xs_set_memalloc(xprt);
2270
ae053551 2271 xs_stream_start_connect(transport);
e1806c7b 2272
b0d93ad5 2273 /* Tell the socket layer to start connecting... */
0fdea1e8 2274 set_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
280254b6 2275 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
16be2d20
CL		 2276}
2277
9903cd1c 2278/**
b61d59ff 2279 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
acf0a39f 2280 * @work: queued work item
9903cd1c
CL		 2281 *
2282 * Invoked by a work queue tasklet.
a246b010 2283 */
cdd518d5 2284static void xs_tcp_setup_socket(struct work_struct *work)
a246b010 2285{
cdd518d5
PE		 2286 struct sock_xprt *transport =
2287 container_of(work, struct sock_xprt, connect_worker.work);
ee0ac0c2 2288 struct socket *sock = transport->sock;
a9f5f0f7 2289 struct rpc_xprt *xprt = &transport->xprt;
280254b6 2290 int status;
8db55a03 2291 unsigned int pflags = current->flags;
a246b010 2292
8db55a03
N		 2293 if (atomic_read(&xprt->swapper))
2294 current->flags |= PF_MEMALLOC;
89f42494
TM		 2295
2296 if (xprt_connected(xprt))
2297 goto out;
2298 if (test_and_clear_bit(XPRT_SOCK_CONNECT_SENT,
2299 &transport->sock_state) ||
2300 !sock) {
2301 xs_reset_transport(transport);
2302 sock = xs_create_sock(xprt, transport, xs_addr(xprt)->sa_family,
2303 SOCK_STREAM, IPPROTO_TCP, true);
b61d59ff 2304 if (IS_ERR(sock)) {
280254b6 2305 xprt_wake_pending_tasks(xprt, PTR_ERR(sock));
3167e12c
CL		 2306 goto out;
2307 }
7d1e8255 2308 }
a246b010 2309
c740eff8
CL		 2310 dprintk("RPC: worker connecting xprt %p via %s to "
2311 "%s (port %s)\n", xprt,
2312 xprt->address_strings[RPC_DISPLAY_PROTO],
2313 xprt->address_strings[RPC_DISPLAY_ADDR],
2314 xprt->address_strings[RPC_DISPLAY_PORT]);
edb267a6 2315
16be2d20 2316 status = xs_tcp_finish_connecting(xprt, sock);
40b5ea0c 2317 trace_rpc_socket_connect(xprt, sock, status);
46121cf7
CL		 2318 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2319 xprt, -status, xprt_connected(xprt),
2320 sock->sk->sk_state);
2a491991
TM		 2321 switch (status) {
2322 case 0:
2323 case -EINPROGRESS:
280254b6 2324 /* SYN_SENT! */
89f42494 2325 set_bit(XPRT_SOCK_CONNECT_SENT, &transport->sock_state);
280254b6
TM		 2326 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2327 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2328 fallthrough;
2a491991 2329 case -EALREADY:
280254b6
TM		 2330 goto out_unlock;
2331 case -EADDRNOTAVAIL:
2332 /* Source port number is unavailable. Try a new one! */
2333 transport->srcport = 0;
2334 status = -EAGAIN;
2335 break;
9fcfe0c8
TM		 2336 case -EINVAL:
2337 /* Happens, for instance, if the user specified a link
2338 * local IPv6 address without a scope-id.
2339 */
3ed5e2a2
TM		 2340 case -ECONNREFUSED:
2341 case -ECONNRESET:
eb5b46fa 2342 case -ENETDOWN:
3ed5e2a2 2343 case -ENETUNREACH:
4ba161a7 2344 case -EHOSTUNREACH:
3913c78c 2345 case -EADDRINUSE:
3601c4a9 2346 case -ENOBUFS:
280254b6
TM		 2347 break;
2348 default:
2349 printk("%s: connect returned unhandled error %d\n",
2350 __func__, status);
2351 status = -EAGAIN;
a246b010 2352 }
280254b6
TM		 2353
2354 /* xs_tcp_force_close() wakes tasks with a fixed error code.
2355 * We need to wake them first to ensure the correct error code.
2356 */
2357 xprt_wake_pending_tasks(xprt, status);
2358 xs_tcp_force_close(xprt);
a246b010 2359out:
68e220bd 2360 xprt_clear_connecting(xprt);
280254b6 2361out_unlock:
cf76785d 2362 xprt_unlock_connect(xprt, transport);
8db55a03 2363 current_restore_flags(pflags, PF_MEMALLOC);
68e220bd 2364}
b0d93ad5 2365
9903cd1c
CL		 2366/**
2367 * xs_connect - connect a socket to a remote endpoint
1b092092 2368 * @xprt: pointer to transport structure
9903cd1c
CL		 2369 * @task: address of RPC task that manages state of connect request
2370 *
2371 * TCP: If the remote end dropped the connection, delay reconnecting.
03bf4b70
CL		 2372 *
2373 * UDP socket connects are synchronous, but we use a work queue anyway
2374 * to guarantee that even unprivileged user processes can set up a
2375 * socket on a privileged port.
2376 *
2377 * If a UDP socket connect fails, the delay behavior here prevents
2378 * retry floods (hard mounts).
9903cd1c 2379 */
1b092092 2380static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
a246b010 2381{
ee0ac0c2 2382 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
02910177 2383 unsigned long delay = 0;
a246b010 2384
718ba5b8
TM		 2385 WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport));
2386
89f42494 2387 if (transport->sock != NULL) {
46121cf7 2388 dprintk("RPC: xs_connect delayed xprt %p for %lu "
89f42494 2389 "seconds\n", xprt, xprt->reestablish_timeout / HZ);
99b1a4c3 2390
675dd90a
CL		 2391 delay = xprt_reconnect_delay(xprt);
2392 xprt_reconnect_backoff(xprt, XS_TCP_INIT_REEST_TO);
02910177 2393
02910177 2394 } else
46121cf7 2395 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
02910177
TM		 2396
2397 queue_delayed_work(xprtiod_workqueue,
2398 &transport->connect_worker,
2399 delay);
a246b010
CL		 2400}
2401
4f8943f8
TM		 2402static void xs_wake_disconnect(struct sock_xprt *transport)
2403{
2404 if (test_and_clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state))
2405 xs_tcp_force_close(&transport->xprt);
2406}
2407
2408static void xs_wake_write(struct sock_xprt *transport)
2409{
2410 if (test_and_clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state))
2411 xprt_write_space(&transport->xprt);
2412}
2413
2414static void xs_wake_error(struct sock_xprt *transport)
2415{
2416 int sockerr;
4f8943f8
TM		 2417
2418 if (!test_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2419 return;
2420 mutex_lock(&transport->recv_mutex);
2421 if (transport->sock == NULL)
2422 goto out;
2423 if (!test_and_clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2424 goto out;
af84537d 2425 sockerr = xchg(&transport->xprt_err, 0);
4f8943f8
TM		 2426 if (sockerr < 0)
2427 xprt_wake_pending_tasks(&transport->xprt, sockerr);
2428out:
2429 mutex_unlock(&transport->recv_mutex);
2430}
2431
2432static void xs_wake_pending(struct sock_xprt *transport)
2433{
2434 if (test_and_clear_bit(XPRT_SOCK_WAKE_PENDING, &transport->sock_state))
2435 xprt_wake_pending_tasks(&transport->xprt, -EAGAIN);
2436}
2437
2438static void xs_error_handle(struct work_struct *work)
2439{
2440 struct sock_xprt *transport = container_of(work,
2441 struct sock_xprt, error_worker);
2442
2443 xs_wake_disconnect(transport);
2444 xs_wake_write(transport);
2445 xs_wake_error(transport);
2446 xs_wake_pending(transport);
2447}
2448
176e21ee 2449/**
12b20ce3 2450 * xs_local_print_stats - display AF_LOCAL socket-specific stats
176e21ee
CL		 2451 * @xprt: rpc_xprt struct containing statistics
2452 * @seq: output file
2453 *
2454 */
2455static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2456{
2457 long idle_time = 0;
2458
2459 if (xprt_connected(xprt))
2460 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2461
2462 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
15a45206 2463 "%llu %llu %lu %llu %llu\n",
176e21ee
CL		 2464 xprt->stat.bind_count,
2465 xprt->stat.connect_count,
8440a886 2466 xprt->stat.connect_time / HZ,
176e21ee
CL		 2467 idle_time,
2468 xprt->stat.sends,
2469 xprt->stat.recvs,
2470 xprt->stat.bad_xids,
2471 xprt->stat.req_u,
15a45206
AA		 2472 xprt->stat.bklog_u,
2473 xprt->stat.max_slots,
2474 xprt->stat.sending_u,
2475 xprt->stat.pending_u);
176e21ee
CL		 2476}
2477
262ca07d 2478/**
12b20ce3 2479 * xs_udp_print_stats - display UDP socket-specific stats
262ca07d
CL		 2480 * @xprt: rpc_xprt struct containing statistics
2481 * @seq: output file
2482 *
2483 */
2484static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2485{
c8475461
CL		 2486 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2487
15a45206
AA		 2488 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2489 "%lu %llu %llu\n",
fbfffbd5 2490 transport->srcport,
262ca07d
CL		 2491 xprt->stat.bind_count,
2492 xprt->stat.sends,
2493 xprt->stat.recvs,
2494 xprt->stat.bad_xids,
2495 xprt->stat.req_u,
15a45206
AA		 2496 xprt->stat.bklog_u,
2497 xprt->stat.max_slots,
2498 xprt->stat.sending_u,
2499 xprt->stat.pending_u);
262ca07d
CL		 2500}
2501
2502/**
12b20ce3 2503 * xs_tcp_print_stats - display TCP socket-specific stats
262ca07d
CL		 2504 * @xprt: rpc_xprt struct containing statistics
2505 * @seq: output file
2506 *
2507 */
2508static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2509{
c8475461 2510 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262ca07d
CL		 2511 long idle_time = 0;
2512
2513 if (xprt_connected(xprt))
2514 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2515
15a45206
AA		 2516 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2517 "%llu %llu %lu %llu %llu\n",
fbfffbd5 2518 transport->srcport,
262ca07d
CL		 2519 xprt->stat.bind_count,
2520 xprt->stat.connect_count,
8440a886 2521 xprt->stat.connect_time / HZ,
262ca07d
CL		 2522 idle_time,
2523 xprt->stat.sends,
2524 xprt->stat.recvs,
2525 xprt->stat.bad_xids,
2526 xprt->stat.req_u,
15a45206
AA		 2527 xprt->stat.bklog_u,
2528 xprt->stat.max_slots,
2529 xprt->stat.sending_u,
2530 xprt->stat.pending_u);
262ca07d
CL		 2531}
2532
4cfc7e60
RI		 2533/*
2534 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2535 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2536 * to use the server side send routines.
2537 */
5fe6eaa1 2538static int bc_malloc(struct rpc_task *task)
4cfc7e60 2539{
5fe6eaa1
CL		 2540 struct rpc_rqst *rqst = task->tk_rqstp;
2541 size_t size = rqst->rq_callsize;
4cfc7e60
RI		 2542 struct page *page;
2543 struct rpc_buffer *buf;
2544
5fe6eaa1
CL		 2545 if (size > PAGE_SIZE - sizeof(struct rpc_buffer)) {
2546 WARN_ONCE(1, "xprtsock: large bc buffer request (size %zu)\n",
2547 size);
2548 return -EINVAL;
2549 }
4cfc7e60 2550
b2648015 2551 page = alloc_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN);
4cfc7e60 2552 if (!page)
5fe6eaa1 2553 return -ENOMEM;
4cfc7e60
RI		 2554
2555 buf = page_address(page);
2556 buf->len = PAGE_SIZE;
2557
5fe6eaa1 2558 rqst->rq_buffer = buf->data;
18e601d6 2559 rqst->rq_rbuffer = (char *)rqst->rq_buffer + rqst->rq_callsize;
5fe6eaa1 2560 return 0;
4cfc7e60
RI		 2561}
2562
2563/*
2564 * Free the space allocated in the bc_alloc routine
2565 */
3435c74a 2566static void bc_free(struct rpc_task *task)
4cfc7e60 2567{
3435c74a 2568 void *buffer = task->tk_rqstp->rq_buffer;
4cfc7e60
RI		 2569 struct rpc_buffer *buf;
2570
4cfc7e60
RI		 2571 buf = container_of(buffer, struct rpc_buffer, data);
2572 free_page((unsigned long)buf);
2573}
2574
4cfc7e60
RI		 2575static int bc_sendto(struct rpc_rqst *req)
2576{
da1661b9 2577 struct xdr_buf *xdr = &req->rq_snd_buf;
4cfc7e60 2578 struct sock_xprt *transport =
067fb11b 2579 container_of(req->rq_xprt, struct sock_xprt, xprt);
067fb11b 2580 struct msghdr msg = {
da1661b9 2581 .msg_flags = 0,
067fb11b
CL		 2582 };
2583 rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT |
da1661b9
CL		 2584 (u32)xdr->len);
2585 unsigned int sent = 0;
2586 int err;
4cfc7e60 2587
8729aaba 2588 req->rq_xtime = ktime_get();
ff053dbb
TM		 2589 err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
2590 if (err < 0)
2591 return err;
da1661b9
CL		 2592 err = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, marker, &sent);
2593 xdr_free_bvec(xdr);
2594 if (err < 0 || sent != (xdr->len + sizeof(marker)))
067fb11b 2595 return -EAGAIN;
da1661b9 2596 return sent;
4cfc7e60
RI		 2597}
2598
ca4faf54
CL		 2599/**
2600 * bc_send_request - Send a backchannel Call on a TCP socket
2601 * @req: rpc_rqst containing Call message to be sent
2602 *
2603 * xpt_mutex ensures @rqstp's whole message is written to the socket
2604 * without interruption.
2605 *
2606 * Return values:
2607 * %0 if the message was sent successfully
2608 * %ENOTCONN if the message was not sent
4cfc7e60 2609 */
adfa7144 2610static int bc_send_request(struct rpc_rqst *req)
4cfc7e60 2611{
4cfc7e60 2612 struct svc_xprt *xprt;
7fc56136 2613 int len;
4cfc7e60 2614
4cfc7e60
RI		 2615 /*
2616 * Get the server socket associated with this callback xprt
2617 */
2618 xprt = req->rq_xprt->bc_xprt;
4cfc7e60
RI		 2619
2620 /*
2621 * Grab the mutex to serialize data as the connection is shared
2622 * with the fore channel
2623 */
c544577d 2624 mutex_lock(&xprt->xpt_mutex);
4cfc7e60
RI		 2625 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2626 len = -ENOTCONN;
2627 else
2628 len = bc_sendto(req);
2629 mutex_unlock(&xprt->xpt_mutex);
2630
2631 if (len > 0)
2632 len = 0;
2633
2634 return len;
2635}
2636
2637/*
2638 * The close routine. Since this is client initiated, we do nothing
2639 */
2640
2641static void bc_close(struct rpc_xprt *xprt)
2642{
6221f1d9 2643 xprt_disconnect_done(xprt);
4cfc7e60
RI		 2644}
2645
2646/*
2647 * The xprt destroy routine. Again, because this connection is client
2648 * initiated, we do nothing
2649 */
2650
2651static void bc_destroy(struct rpc_xprt *xprt)
2652{
47f72efa
KM		 2653 dprintk("RPC: bc_destroy xprt %p\n", xprt);
2654
2655 xs_xprt_free(xprt);
2656 module_put(THIS_MODULE);
4cfc7e60
RI		 2657}
2658
d31ae254 2659static const struct rpc_xprt_ops xs_local_ops = {
176e21ee 2660 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2661 .release_xprt = xprt_release_xprt,
f39c1bfb 2662 .alloc_slot = xprt_alloc_slot,
a9cde23a 2663 .free_slot = xprt_free_slot,
176e21ee
CL		 2664 .rpcbind = xs_local_rpcbind,
2665 .set_port = xs_local_set_port,
dc107402 2666 .connect = xs_local_connect,
176e21ee
CL		 2667 .buf_alloc = rpc_malloc,
2668 .buf_free = rpc_free,
550aebfe 2669 .prepare_request = xs_stream_prepare_request,
176e21ee 2670 .send_request = xs_local_send_request,
8ba6a92d 2671 .wait_for_reply_request = xprt_wait_for_reply_request_def,
176e21ee 2672 .close = xs_close,
a1311d87 2673 .destroy = xs_destroy,
176e21ee 2674 .print_stats = xs_local_print_stats,
d67fa4d8
JL		 2675 .enable_swap = xs_enable_swap,
2676 .disable_swap = xs_disable_swap,
176e21ee
CL		 2677};
2678
d31ae254 2679static const struct rpc_xprt_ops xs_udp_ops = {
43118c29 2680 .set_buffer_size = xs_udp_set_buffer_size,
12a80469 2681 .reserve_xprt = xprt_reserve_xprt_cong,
49e9a890 2682 .release_xprt = xprt_release_xprt_cong,
f39c1bfb 2683 .alloc_slot = xprt_alloc_slot,
a9cde23a 2684 .free_slot = xprt_free_slot,
45160d62 2685 .rpcbind = rpcb_getport_async,
92200412 2686 .set_port = xs_set_port,
262965f5 2687 .connect = xs_connect,
421ab1be
TM		 2688 .get_srcaddr = xs_sock_srcaddr,
2689 .get_srcport = xs_sock_srcport,
02107148
CL		 2690 .buf_alloc = rpc_malloc,
2691 .buf_free = rpc_free,
262965f5 2692 .send_request = xs_udp_send_request,
8ba6a92d 2693 .wait_for_reply_request = xprt_wait_for_reply_request_rtt,
46c0ee8b 2694 .timer = xs_udp_timer,
a58dd398 2695 .release_request = xprt_release_rqst_cong,
262965f5
CL		 2696 .close = xs_close,
2697 .destroy = xs_destroy,
262ca07d 2698 .print_stats = xs_udp_print_stats,
d67fa4d8
JL		 2699 .enable_swap = xs_enable_swap,
2700 .disable_swap = xs_disable_swap,
4a068258 2701 .inject_disconnect = xs_inject_disconnect,
262965f5
CL		 2702};
2703
d31ae254 2704static const struct rpc_xprt_ops xs_tcp_ops = {
12a80469 2705 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2706 .release_xprt = xprt_release_xprt,
36bd7de9 2707 .alloc_slot = xprt_alloc_slot,
a9cde23a 2708 .free_slot = xprt_free_slot,
45160d62 2709 .rpcbind = rpcb_getport_async,
92200412 2710 .set_port = xs_set_port,
0b9e7943 2711 .connect = xs_connect,
421ab1be
TM		 2712 .get_srcaddr = xs_sock_srcaddr,
2713 .get_srcport = xs_sock_srcport,
02107148
CL		 2714 .buf_alloc = rpc_malloc,
2715 .buf_free = rpc_free,
277e4ab7 2716 .prepare_request = xs_stream_prepare_request,
262965f5 2717 .send_request = xs_tcp_send_request,
8ba6a92d 2718 .wait_for_reply_request = xprt_wait_for_reply_request_def,
c627d31b 2719 .close = xs_tcp_shutdown,
9903cd1c 2720 .destroy = xs_destroy,
7196dbb0 2721 .set_connect_timeout = xs_tcp_set_connect_timeout,
262ca07d 2722 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2723 .enable_swap = xs_enable_swap,
2724 .disable_swap = xs_disable_swap,
4a068258 2725 .inject_disconnect = xs_inject_disconnect,
42e5c3e2
CL		 2726#ifdef CONFIG_SUNRPC_BACKCHANNEL
2727 .bc_setup = xprt_setup_bc,
6b26cc8c 2728 .bc_maxpayload = xs_tcp_bc_maxpayload,
7402a4fe 2729 .bc_num_slots = xprt_bc_max_slots,
42e5c3e2
CL		 2730 .bc_free_rqst = xprt_free_bc_rqst,
2731 .bc_destroy = xprt_destroy_bc,
2732#endif
a246b010
CL		 2733};
2734
4cfc7e60
RI		 2735/*
2736 * The rpc_xprt_ops for the server backchannel
2737 */
2738
d31ae254 2739static const struct rpc_xprt_ops bc_tcp_ops = {
4cfc7e60
RI		 2740 .reserve_xprt = xprt_reserve_xprt,
2741 .release_xprt = xprt_release_xprt,
84e28a30 2742 .alloc_slot = xprt_alloc_slot,
a9cde23a 2743 .free_slot = xprt_free_slot,
4cfc7e60
RI		 2744 .buf_alloc = bc_malloc,
2745 .buf_free = bc_free,
2746 .send_request = bc_send_request,
8ba6a92d 2747 .wait_for_reply_request = xprt_wait_for_reply_request_def,
4cfc7e60
RI		 2748 .close = bc_close,
2749 .destroy = bc_destroy,
2750 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2751 .enable_swap = xs_enable_swap,
2752 .disable_swap = xs_disable_swap,
4a068258 2753 .inject_disconnect = xs_inject_disconnect,
4cfc7e60
RI		 2754};
2755
92476850
CL		 2756static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2757{
2758 static const struct sockaddr_in sin = {
2759 .sin_family = AF_INET,
2760 .sin_addr.s_addr = htonl(INADDR_ANY),
2761 };
2762 static const struct sockaddr_in6 sin6 = {
2763 .sin6_family = AF_INET6,
2764 .sin6_addr = IN6ADDR_ANY_INIT,
2765 };
2766
2767 switch (family) {
176e21ee
CL		 2768 case AF_LOCAL:
2769 break;
92476850
CL		 2770 case AF_INET:
2771 memcpy(sap, &sin, sizeof(sin));
2772 break;
2773 case AF_INET6:
2774 memcpy(sap, &sin6, sizeof(sin6));
2775 break;
2776 default:
2777 dprintk("RPC: %s: Bad address family\n", __func__);
2778 return -EAFNOSUPPORT;
2779 }
2780 return 0;
2781}
2782
3c341b0b 2783static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
d9ba131d
TM		 2784 unsigned int slot_table_size,
2785 unsigned int max_slot_table_size)
c8541ecd
CL		 2786{
2787 struct rpc_xprt *xprt;
ffc2e518 2788 struct sock_xprt *new;
c8541ecd 2789
96802a09 2790 if (args->addrlen > sizeof(xprt->addr)) {
46121cf7 2791 dprintk("RPC: xs_setup_xprt: address too large\n");
c8541ecd
CL		 2792 return ERR_PTR(-EBADF);
2793 }
2794
d9ba131d
TM		 2795 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2796 max_slot_table_size);
bd1722d4 2797 if (xprt == NULL) {
46121cf7
CL		 2798 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2799 "rpc_xprt\n");
c8541ecd
CL		 2800 return ERR_PTR(-ENOMEM);
2801 }
c8541ecd 2802
bd1722d4 2803 new = container_of(xprt, struct sock_xprt, xprt);
edc1b01c 2804 mutex_init(&new->recv_mutex);
96802a09
FM		 2805 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2806 xprt->addrlen = args->addrlen;
d3bc9a1d 2807 if (args->srcaddr)
fbfffbd5 2808 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
92476850
CL		 2809 else {
2810 int err;
2811 err = xs_init_anyaddr(args->dstaddr->sa_family,
2812 (struct sockaddr *)&new->srcaddr);
2aa13531
SK		 2813 if (err != 0) {
2814 xprt_free(xprt);
92476850 2815 return ERR_PTR(err);
2aa13531 2816 }
92476850 2817 }
c8541ecd
CL		 2818
2819 return xprt;
2820}
2821
176e21ee
CL		 2822static const struct rpc_timeout xs_local_default_timeout = {
2823 .to_initval = 10 * HZ,
2824 .to_maxval = 10 * HZ,
2825 .to_retries = 2,
2826};
2827
2828/**
2829 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2830 * @args: rpc transport creation arguments
2831 *
2832 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2833 */
2834static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2835{
2836 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2837 struct sock_xprt *transport;
2838 struct rpc_xprt *xprt;
2839 struct rpc_xprt *ret;
2840
d9ba131d
TM		 2841 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2842 xprt_max_tcp_slot_table_entries);
176e21ee
CL		 2843 if (IS_ERR(xprt))
2844 return xprt;
2845 transport = container_of(xprt, struct sock_xprt, xprt);
2846
2847 xprt->prot = 0;
d3abc739 2848 xprt->xprt_class = &xs_local_transport;
176e21ee
CL		 2849 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2850
2851 xprt->bind_timeout = XS_BIND_TO;
2852 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2853 xprt->idle_timeout = XS_IDLE_DISC_TO;
2854
2855 xprt->ops = &xs_local_ops;
2856 xprt->timeout = &xs_local_default_timeout;
2857
550aebfe 2858 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 2859 INIT_WORK(&transport->error_worker, xs_error_handle);
550aebfe 2860 INIT_DELAYED_WORK(&transport->connect_worker, xs_dummy_setup_socket);
93dc41bd 2861
176e21ee
CL		 2862 switch (sun->sun_family) {
2863 case AF_LOCAL:
2864 if (sun->sun_path[0] != '/') {
2865 dprintk("RPC: bad AF_LOCAL address: %s\n",
2866 sun->sun_path);
2867 ret = ERR_PTR(-EINVAL);
2868 goto out_err;
2869 }
2870 xprt_set_bound(xprt);
176e21ee
CL		 2871 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
2872 break;
2873 default:
2874 ret = ERR_PTR(-EAFNOSUPPORT);
2875 goto out_err;
2876 }
2877
2878 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2879 xprt->address_strings[RPC_DISPLAY_ADDR]);
2880
2881 if (try_module_get(THIS_MODULE))
2882 return xprt;
2883 ret = ERR_PTR(-EINVAL);
2884out_err:
315f3812 2885 xs_xprt_free(xprt);
176e21ee
CL		 2886 return ret;
2887}
2888
2881ae74
TM		 2889static const struct rpc_timeout xs_udp_default_timeout = {
2890 .to_initval = 5 * HZ,
2891 .to_maxval = 30 * HZ,
2892 .to_increment = 5 * HZ,
2893 .to_retries = 5,
2894};
2895
9903cd1c
CL		 2896/**
2897 * xs_setup_udp - Set up transport to use a UDP socket
96802a09 2898 * @args: rpc transport creation arguments
9903cd1c
CL		 2899 *
2900 */
483066d6 2901static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
a246b010 2902{
8f9d5b1a 2903 struct sockaddr *addr = args->dstaddr;
c8541ecd 2904 struct rpc_xprt *xprt;
c8475461 2905 struct sock_xprt *transport;
0a68b0be 2906 struct rpc_xprt *ret;
a246b010 2907
d9ba131d
TM		 2908 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2909 xprt_udp_slot_table_entries);
c8541ecd
CL		 2910 if (IS_ERR(xprt))
2911 return xprt;
c8475461 2912 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 2913
ec739ef0 2914 xprt->prot = IPPROTO_UDP;
d3abc739 2915 xprt->xprt_class = &xs_udp_transport;
a246b010
CL		 2916 /* XXX: header size can vary due to auth type, IPv6, etc. */
2917 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2918
03bf4b70 2919 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 2920 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2921 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 2922
262965f5 2923 xprt->ops = &xs_udp_ops;
a246b010 2924
ba7392bb 2925 xprt->timeout = &xs_udp_default_timeout;
a246b010 2926
f9b2ee71 2927 INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn);
4f8943f8 2928 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 2929 INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket);
2930
8f9d5b1a
CL		 2931 switch (addr->sa_family) {
2932 case AF_INET:
2933 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2934 xprt_set_bound(xprt);
2935
9dc3b095 2936 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
8f9d5b1a
CL		 2937 break;
2938 case AF_INET6:
2939 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2940 xprt_set_bound(xprt);
2941
9dc3b095 2942 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
8f9d5b1a
CL		 2943 break;
2944 default:
0a68b0be
BF		 2945 ret = ERR_PTR(-EAFNOSUPPORT);
2946 goto out_err;
8f9d5b1a
CL		 2947 }
2948
c740eff8
CL		 2949 if (xprt_bound(xprt))
2950 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2951 xprt->address_strings[RPC_DISPLAY_ADDR],
2952 xprt->address_strings[RPC_DISPLAY_PORT],
2953 xprt->address_strings[RPC_DISPLAY_PROTO]);
2954 else
2955 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2956 xprt->address_strings[RPC_DISPLAY_ADDR],
2957 xprt->address_strings[RPC_DISPLAY_PROTO]);
edb267a6 2958
bc25571e
TT		 2959 if (try_module_get(THIS_MODULE))
2960 return xprt;
0a68b0be
BF		 2961 ret = ERR_PTR(-EINVAL);
2962out_err:
315f3812 2963 xs_xprt_free(xprt);
0a68b0be 2964 return ret;
a246b010
CL		 2965}
2966
2881ae74
TM		 2967static const struct rpc_timeout xs_tcp_default_timeout = {
2968 .to_initval = 60 * HZ,
2969 .to_maxval = 60 * HZ,
2970 .to_retries = 2,
2971};
2972
9903cd1c
CL		 2973/**
2974 * xs_setup_tcp - Set up transport to use a TCP socket
96802a09 2975 * @args: rpc transport creation arguments
9903cd1c
CL		 2976 *
2977 */
483066d6 2978static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
a246b010 2979{
8f9d5b1a 2980 struct sockaddr *addr = args->dstaddr;
c8541ecd 2981 struct rpc_xprt *xprt;
c8475461 2982 struct sock_xprt *transport;
0a68b0be 2983 struct rpc_xprt *ret;
b7993ceb
TM		 2984 unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
2985
2986 if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
2987 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
a246b010 2988
d9ba131d 2989 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
b7993ceb 2990 max_slot_table_size);
c8541ecd
CL		 2991 if (IS_ERR(xprt))
2992 return xprt;
c8475461 2993 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 2994
ec739ef0 2995 xprt->prot = IPPROTO_TCP;
d3abc739 2996 xprt->xprt_class = &xs_tcp_transport;
808012fb 2997 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
a246b010 2998
03bf4b70 2999 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 3000 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3001 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 3002
262965f5 3003 xprt->ops = &xs_tcp_ops;
ba7392bb 3004 xprt->timeout = &xs_tcp_default_timeout;
a246b010 3005
3851f1cd 3006 xprt->max_reconnect_timeout = xprt->timeout->to_maxval;
7196dbb0
TM		 3007 xprt->connect_timeout = xprt->timeout->to_initval *
3008 (xprt->timeout->to_retries + 1);
3851f1cd 3009
c50b8ee0 3010 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 3011 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 3012 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket);
3013
8f9d5b1a
CL		 3014 switch (addr->sa_family) {
3015 case AF_INET:
3016 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
3017 xprt_set_bound(xprt);
3018
9dc3b095 3019 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
8f9d5b1a
CL		 3020 break;
3021 case AF_INET6:
3022 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
3023 xprt_set_bound(xprt);
3024
9dc3b095 3025 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
8f9d5b1a
CL		 3026 break;
3027 default:
0a68b0be
BF		 3028 ret = ERR_PTR(-EAFNOSUPPORT);
3029 goto out_err;
8f9d5b1a
CL		 3030 }
3031
c740eff8
CL		 3032 if (xprt_bound(xprt))
3033 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3034 xprt->address_strings[RPC_DISPLAY_ADDR],
3035 xprt->address_strings[RPC_DISPLAY_PORT],
3036 xprt->address_strings[RPC_DISPLAY_PROTO]);
3037 else
3038 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
3039 xprt->address_strings[RPC_DISPLAY_ADDR],
3040 xprt->address_strings[RPC_DISPLAY_PROTO]);
3041
bc25571e
TT		 3042 if (try_module_get(THIS_MODULE))
3043 return xprt;
0a68b0be
BF		 3044 ret = ERR_PTR(-EINVAL);
3045out_err:
315f3812 3046 xs_xprt_free(xprt);
0a68b0be 3047 return ret;
a246b010 3048}
282b32e1 3049
f300baba
AB		 3050/**
3051 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
3052 * @args: rpc transport creation arguments
3053 *
3054 */
3055static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
3056{
3057 struct sockaddr *addr = args->dstaddr;
3058 struct rpc_xprt *xprt;
3059 struct sock_xprt *transport;
3060 struct svc_sock *bc_sock;
0a68b0be 3061 struct rpc_xprt *ret;
f300baba 3062
d9ba131d
TM		 3063 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
3064 xprt_tcp_slot_table_entries);
f300baba
AB		 3065 if (IS_ERR(xprt))
3066 return xprt;
3067 transport = container_of(xprt, struct sock_xprt, xprt);
3068
3069 xprt->prot = IPPROTO_TCP;
d3abc739 3070 xprt->xprt_class = &xs_bc_tcp_transport;
f300baba
AB		 3071 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
3072 xprt->timeout = &xs_tcp_default_timeout;
3073
3074 /* backchannel */
3075 xprt_set_bound(xprt);
3076 xprt->bind_timeout = 0;
f300baba
AB		 3077 xprt->reestablish_timeout = 0;
3078 xprt->idle_timeout = 0;
3079
f300baba
AB		 3080 xprt->ops = &bc_tcp_ops;
3081
3082 switch (addr->sa_family) {
3083 case AF_INET:
3084 xs_format_peer_addresses(xprt, "tcp",
3085 RPCBIND_NETID_TCP);
3086 break;
3087 case AF_INET6:
3088 xs_format_peer_addresses(xprt, "tcp",
3089 RPCBIND_NETID_TCP6);
3090 break;
3091 default:
0a68b0be
BF		 3092 ret = ERR_PTR(-EAFNOSUPPORT);
3093 goto out_err;
f300baba
AB		 3094 }
3095
50fa0d40
PE		 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]);
f300baba 3100
99de8ea9
BF		 3101 /*
3102 * Once we've associated a backchannel xprt with a connection,
28303ca3
WM		 3103 * we want to keep it around as long as the connection lasts,
3104 * in case we need to start using it for a backchannel again;
3105 * this reference won't be dropped until bc_xprt is destroyed.
99de8ea9
BF		 3106 */
3107 xprt_get(xprt);
3108 args->bc_xprt->xpt_bc_xprt = xprt;
3109 xprt->bc_xprt = args->bc_xprt;
3110 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
3111 transport->sock = bc_sock->sk_sock;
3112 transport->inet = bc_sock->sk_sk;
3113
f300baba
AB		 3114 /*
3115 * Since we don't want connections for the backchannel, we set
3116 * the xprt status to connected
3117 */
3118 xprt_set_connected(xprt);
3119
f300baba
AB		 3120 if (try_module_get(THIS_MODULE))
3121 return xprt;
642aab58
KM		 3122
3123 args->bc_xprt->xpt_bc_xprt = NULL;
39a9beab 3124 args->bc_xprt->xpt_bc_xps = NULL;
99de8ea9 3125 xprt_put(xprt);
0a68b0be
BF		 3126 ret = ERR_PTR(-EINVAL);
3127out_err:
315f3812 3128 xs_xprt_free(xprt);
0a68b0be 3129 return ret;
f300baba
AB		 3130}
3131
176e21ee
CL		 3132static struct xprt_class xs_local_transport = {
3133 .list = LIST_HEAD_INIT(xs_local_transport.list),
3134 .name = "named UNIX socket",
3135 .owner = THIS_MODULE,
3136 .ident = XPRT_TRANSPORT_LOCAL,
3137 .setup = xs_setup_local,
d5aa6b22 3138 .netid = { "" },
176e21ee
CL		 3139};
3140
bc25571e
TT		 3141static struct xprt_class xs_udp_transport = {
3142 .list = LIST_HEAD_INIT(xs_udp_transport.list),
3143 .name = "udp",
3144 .owner = THIS_MODULE,
f300baba 3145 .ident = XPRT_TRANSPORT_UDP,
bc25571e 3146 .setup = xs_setup_udp,
d5aa6b22 3147 .netid = { "udp", "udp6", "" },
bc25571e
TT		 3148};
3149
3150static struct xprt_class xs_tcp_transport = {
3151 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
3152 .name = "tcp",
3153 .owner = THIS_MODULE,
f300baba 3154 .ident = XPRT_TRANSPORT_TCP,
bc25571e 3155 .setup = xs_setup_tcp,
d5aa6b22 3156 .netid = { "tcp", "tcp6", "" },
bc25571e
TT		 3157};
3158
f300baba
AB		 3159static struct xprt_class xs_bc_tcp_transport = {
3160 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
3161 .name = "tcp NFSv4.1 backchannel",
3162 .owner = THIS_MODULE,
3163 .ident = XPRT_TRANSPORT_BC_TCP,
3164 .setup = xs_setup_bc_tcp,
d5aa6b22 3165 .netid = { "" },
f300baba
AB		 3166};
3167
282b32e1 3168/**
bc25571e 3169 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
282b32e1
CL		 3170 *
3171 */
3172int init_socket_xprt(void)
3173{
2b1bec5f 3174 if (!sunrpc_table_header)
0b4d4147 3175 sunrpc_table_header = register_sysctl_table(sunrpc_table);
fbf76683 3176
176e21ee 3177 xprt_register_transport(&xs_local_transport);
bc25571e
TT		 3178 xprt_register_transport(&xs_udp_transport);
3179 xprt_register_transport(&xs_tcp_transport);
f300baba 3180 xprt_register_transport(&xs_bc_tcp_transport);
bc25571e 3181
282b32e1
CL		 3182 return 0;
3183}
3184
3185/**
bc25571e 3186 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
282b32e1
CL		 3187 *
3188 */
3189void cleanup_socket_xprt(void)
3190{
fbf76683
CL		 3191 if (sunrpc_table_header) {
3192 unregister_sysctl_table(sunrpc_table_header);
3193 sunrpc_table_header = NULL;
3194 }
bc25571e 3195
176e21ee 3196 xprt_unregister_transport(&xs_local_transport);
bc25571e
TT		 3197 xprt_unregister_transport(&xs_udp_transport);
3198 xprt_unregister_transport(&xs_tcp_transport);
f300baba 3199 xprt_unregister_transport(&xs_bc_tcp_transport);
282b32e1 3200}
cbf11071 3201
9bbb9e5a 3202static int param_set_portnr(const char *val, const struct kernel_param *kp)
cbf11071 3203{
ffb6ca33 3204 return param_set_uint_minmax(val, kp,
826799e6 3205 RPC_MIN_RESVPORT,
cbf11071
TM		 3206 RPC_MAX_RESVPORT);
3207}
3208
9c27847d 3209static const struct kernel_param_ops param_ops_portnr = {
9bbb9e5a
RR		 3210 .set = param_set_portnr,
3211 .get = param_get_uint,
3212};
3213
cbf11071
TM		 3214#define param_check_portnr(name, p) \
3215 __param_check(name, p, unsigned int);
3216
3217module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3218module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3219
9bbb9e5a
RR		 3220static int param_set_slot_table_size(const char *val,
3221 const struct kernel_param *kp)
cbf11071
TM		 3222{
3223 return param_set_uint_minmax(val, kp,
3224 RPC_MIN_SLOT_TABLE,
3225 RPC_MAX_SLOT_TABLE);
3226}
3227
9c27847d 3228static const struct kernel_param_ops param_ops_slot_table_size = {
9bbb9e5a
RR		 3229 .set = param_set_slot_table_size,
3230 .get = param_get_uint,
3231};
3232
cbf11071
TM		 3233#define param_check_slot_table_size(name, p) \
3234 __param_check(name, p, unsigned int);
3235
d9ba131d
TM		 3236static int param_set_max_slot_table_size(const char *val,
3237 const struct kernel_param *kp)
3238{
3239 return param_set_uint_minmax(val, kp,
3240 RPC_MIN_SLOT_TABLE,
3241 RPC_MAX_SLOT_TABLE_LIMIT);
3242}
3243
9c27847d 3244static const struct kernel_param_ops param_ops_max_slot_table_size = {
d9ba131d
TM		 3245 .set = param_set_max_slot_table_size,
3246 .get = param_get_uint,
3247};
3248
3249#define param_check_max_slot_table_size(name, p) \
3250 __param_check(name, p, unsigned int);
3251
cbf11071
TM		 3252module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3253 slot_table_size, 0644);
d9ba131d
TM		 3254module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3255 max_slot_table_size, 0644);
cbf11071
TM		 3256module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3257 slot_table_size, 0644);
Linux 6.x block layer and io_uring tree(s)