]> Git Repo - linux.git/blame - net/sunrpc/xprtsock.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
SUNRPC: Fix initialisation of struct rpc_xprt_switch
[linux.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 <[email protected]>
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 <[email protected]>
55aa4f58
CL		 15 *
16 * IP socket transport implementation, (C) 2005 Chuck Lever <[email protected]>
8f9d5b1a
CL		 17 *
18 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
19 * <[email protected]>
a246b010
CL		 20 */
21
22#include <linux/types.h>
176e21ee 23#include <linux/string.h>
a246b010 24#include <linux/slab.h>
bc25571e 25#include <linux/module.h>
a246b010 26#include <linux/capability.h>
a246b010
CL		 27#include <linux/pagemap.h>
28#include <linux/errno.h>
29#include <linux/socket.h>
30#include <linux/in.h>
31#include <linux/net.h>
32#include <linux/mm.h>
176e21ee 33#include <linux/un.h>
a246b010
CL		 34#include <linux/udp.h>
35#include <linux/tcp.h>
36#include <linux/sunrpc/clnt.h>
5976687a 37#include <linux/sunrpc/addr.h>
02107148 38#include <linux/sunrpc/sched.h>
4cfc7e60 39#include <linux/sunrpc/svcsock.h>
49c36fcc 40#include <linux/sunrpc/xprtsock.h>
a246b010 41#include <linux/file.h>
9e00abc3 42#ifdef CONFIG_SUNRPC_BACKCHANNEL
44b98efd
RL		 43#include <linux/sunrpc/bc_xprt.h>
44#endif
a246b010
CL		 45
46#include <net/sock.h>
47#include <net/checksum.h>
48#include <net/udp.h>
49#include <net/tcp.h>
277e4ab7 50#include <linux/bvec.h>
6a829eb8 51#include <linux/highmem.h>
277e4ab7 52#include <linux/uio.h>
a1231fda 53#include <linux/sched/mm.h>
a246b010 54
40b5ea0c
TM		 55#include <trace/events/sunrpc.h>
56
4cfc7e60 57#include "sunrpc.h"
176e21ee
CL		 58
59static void xs_close(struct rpc_xprt *xprt);
7196dbb0
TM		 60static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
61 struct socket *sock);
176e21ee 62
c556b754
CL		 63/*
64 * xprtsock tunables
65 */
09acfea5
TM		 66static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
67static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
68static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
c556b754 69
09acfea5
TM		 70static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
71static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
c556b754 72
7d1e8255 73#define XS_TCP_LINGER_TO (15U * HZ)
25fe6142 74static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
7d1e8255 75
fbf76683
CL		 76/*
77 * We can register our own files under /proc/sys/sunrpc by
78 * calling register_sysctl_table() again. The files in that
79 * directory become the union of all files registered there.
80 *
81 * We simply need to make sure that we don't collide with
82 * someone else's file names!
83 */
84
fbf76683
CL		 85static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
86static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
d9ba131d 87static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
fbf76683
CL		 88static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
89static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
90
91static struct ctl_table_header *sunrpc_table_header;
92
93/*
94 * FIXME: changing the UDP slot table size should also resize the UDP
95 * socket buffers for existing UDP transports
96 */
fe2c6338 97static struct ctl_table xs_tunables_table[] = {
fbf76683 98 {
fbf76683
CL		 99 .procname = "udp_slot_table_entries",
100 .data = &xprt_udp_slot_table_entries,
101 .maxlen = sizeof(unsigned int),
102 .mode = 0644,
6d456111 103 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 104 .extra1 = &min_slot_table_size,
105 .extra2 = &max_slot_table_size
106 },
107 {
fbf76683
CL		 108 .procname = "tcp_slot_table_entries",
109 .data = &xprt_tcp_slot_table_entries,
110 .maxlen = sizeof(unsigned int),
111 .mode = 0644,
6d456111 112 .proc_handler = proc_dointvec_minmax,
fbf76683
CL		 113 .extra1 = &min_slot_table_size,
114 .extra2 = &max_slot_table_size
115 },
d9ba131d
TM		 116 {
117 .procname = "tcp_max_slot_table_entries",
118 .data = &xprt_max_tcp_slot_table_entries,
119 .maxlen = sizeof(unsigned int),
120 .mode = 0644,
121 .proc_handler = proc_dointvec_minmax,
122 .extra1 = &min_slot_table_size,
123 .extra2 = &max_tcp_slot_table_limit
124 },
fbf76683 125 {
fbf76683
CL		 126 .procname = "min_resvport",
127 .data = &xprt_min_resvport,
128 .maxlen = sizeof(unsigned int),
129 .mode = 0644,
6d456111 130 .proc_handler = proc_dointvec_minmax,
fbf76683 131 .extra1 = &xprt_min_resvport_limit,
826799e6 132 .extra2 = &xprt_max_resvport_limit
fbf76683
CL		 133 },
134 {
fbf76683
CL		 135 .procname = "max_resvport",
136 .data = &xprt_max_resvport,
137 .maxlen = sizeof(unsigned int),
138 .mode = 0644,
6d456111 139 .proc_handler = proc_dointvec_minmax,
826799e6 140 .extra1 = &xprt_min_resvport_limit,
fbf76683
CL		 141 .extra2 = &xprt_max_resvport_limit
142 },
25fe6142
TM		 143 {
144 .procname = "tcp_fin_timeout",
145 .data = &xs_tcp_fin_timeout,
146 .maxlen = sizeof(xs_tcp_fin_timeout),
147 .mode = 0644,
6d456111 148 .proc_handler = proc_dointvec_jiffies,
fbf76683 149 },
f8572d8f 150 { },
fbf76683
CL		 151};
152
fe2c6338 153static struct ctl_table sunrpc_table[] = {
fbf76683 154 {
fbf76683
CL		 155 .procname = "sunrpc",
156 .mode = 0555,
157 .child = xs_tunables_table
158 },
f8572d8f 159 { },
fbf76683
CL		 160};
161
03bf4b70
CL		 162/*
163 * Wait duration for a reply from the RPC portmapper.
164 */
165#define XS_BIND_TO (60U * HZ)
166
167/*
168 * Delay if a UDP socket connect error occurs. This is most likely some
169 * kind of resource problem on the local host.
170 */
171#define XS_UDP_REEST_TO (2U * HZ)
172
173/*
174 * The reestablish timeout allows clients to delay for a bit before attempting
175 * to reconnect to a server that just dropped our connection.
176 *
177 * We implement an exponential backoff when trying to reestablish a TCP
178 * transport connection with the server. Some servers like to drop a TCP
179 * connection when they are overworked, so we start with a short timeout and
180 * increase over time if the server is down or not responding.
181 */
182#define XS_TCP_INIT_REEST_TO (3U * HZ)
03bf4b70
CL		 183
184/*
185 * TCP idle timeout; client drops the transport socket if it is idle
186 * for this long. Note that we also timeout UDP sockets to prevent
187 * holding port numbers when there is no RPC traffic.
188 */
189#define XS_IDLE_DISC_TO (5U * 60 * HZ)
190
f895b252 191#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
a246b010 192# undef RPC_DEBUG_DATA
9903cd1c 193# define RPCDBG_FACILITY RPCDBG_TRANS
a246b010
CL		 194#endif
195
a246b010 196#ifdef RPC_DEBUG_DATA
9903cd1c 197static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010 198{
9903cd1c
CL		 199 u8 *buf = (u8 *) packet;
200 int j;
a246b010 201
46121cf7 202 dprintk("RPC: %s\n", msg);
a246b010
CL		 203 for (j = 0; j < count && j < 128; j += 4) {
204 if (!(j & 31)) {
205 if (j)
206 dprintk("\n");
207 dprintk("0x%04x ", j);
208 }
209 dprintk("%02x%02x%02x%02x ",
210 buf[j], buf[j+1], buf[j+2], buf[j+3]);
211 }
212 dprintk("\n");
213}
214#else
9903cd1c 215static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
a246b010
CL		 216{
217 /* NOP */
218}
219#endif
220
2118071d
TM		 221static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
222{
223 return (struct rpc_xprt *) sk->sk_user_data;
224}
225
95392c59
CL		 226static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
227{
228 return (struct sockaddr *) &xprt->addr;
229}
230
176e21ee
CL		 231static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
232{
233 return (struct sockaddr_un *) &xprt->addr;
234}
235
95392c59 236static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
edb267a6 237{
95392c59
CL		 238 return (struct sockaddr_in *) &xprt->addr;
239}
240
241static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
242{
243 return (struct sockaddr_in6 *) &xprt->addr;
244}
245
c877b849 246static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
edb267a6 247{
c877b849 248 struct sockaddr *sap = xs_addr(xprt);
9dc3b095
CL		 249 struct sockaddr_in6 *sin6;
250 struct sockaddr_in *sin;
176e21ee 251 struct sockaddr_un *sun;
c877b849 252 char buf[128];
edb267a6 253
9dc3b095 254 switch (sap->sa_family) {
176e21ee
CL		 255 case AF_LOCAL:
256 sun = xs_addr_un(xprt);
257 strlcpy(buf, sun->sun_path, sizeof(buf));
258 xprt->address_strings[RPC_DISPLAY_ADDR] =
259 kstrdup(buf, GFP_KERNEL);
260 break;
9dc3b095 261 case AF_INET:
176e21ee
CL		 262 (void)rpc_ntop(sap, buf, sizeof(buf));
263 xprt->address_strings[RPC_DISPLAY_ADDR] =
264 kstrdup(buf, GFP_KERNEL);
9dc3b095 265 sin = xs_addr_in(xprt);
fc0b5791 266 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
9dc3b095
CL		 267 break;
268 case AF_INET6:
176e21ee
CL		 269 (void)rpc_ntop(sap, buf, sizeof(buf));
270 xprt->address_strings[RPC_DISPLAY_ADDR] =
271 kstrdup(buf, GFP_KERNEL);
9dc3b095 272 sin6 = xs_addr_in6(xprt);
fc0b5791 273 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
9dc3b095
CL		 274 break;
275 default:
276 BUG();
756805e7 277 }
176e21ee 278
9dc3b095 279 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
edb267a6
CL		 280}
281
9dc3b095 282static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
4b6473fb 283{
9dc3b095
CL		 284 struct sockaddr *sap = xs_addr(xprt);
285 char buf[128];
4b6473fb 286
81160e66 287 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
c877b849 288 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
4b6473fb 289
81160e66 290 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
c877b849
CL		 291 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
292}
4b6473fb 293
9dc3b095
CL		 294static void xs_format_peer_addresses(struct rpc_xprt *xprt,
295 const char *protocol,
296 const char *netid)
c877b849 297{
b454ae90 298 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
b454ae90 299 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
c877b849 300 xs_format_common_peer_addresses(xprt);
9dc3b095 301 xs_format_common_peer_ports(xprt);
edb267a6 302}
4b6473fb 303
9dc3b095 304static void xs_update_peer_port(struct rpc_xprt *xprt)
4b6473fb 305{
9dc3b095
CL		 306 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
307 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
4417c8c4 308
9dc3b095 309 xs_format_common_peer_ports(xprt);
edb267a6
CL		 310}
311
312static void xs_free_peer_addresses(struct rpc_xprt *xprt)
313{
33e01dc7
CL		 314 unsigned int i;
315
316 for (i = 0; i < RPC_DISPLAY_MAX; i++)
317 switch (i) {
318 case RPC_DISPLAY_PROTO:
319 case RPC_DISPLAY_NETID:
320 continue;
321 default:
322 kfree(xprt->address_strings[i]);
323 }
edb267a6
CL		 324}
325
277e4ab7
TM		 326static size_t
327xs_alloc_sparse_pages(struct xdr_buf *buf, size_t want, gfp_t gfp)
328{
329 size_t i,n;
330
16e5e90f 331 if (!want || !(buf->flags & XDRBUF_SPARSE_PAGES))
277e4ab7 332 return want;
277e4ab7
TM		 333 n = (buf->page_base + want + PAGE_SIZE - 1) >> PAGE_SHIFT;
334 for (i = 0; i < n; i++) {
335 if (buf->pages[i])
336 continue;
337 buf->bvec[i].bv_page = buf->pages[i] = alloc_page(gfp);
338 if (!buf->pages[i]) {
16e5e90f
TM		 339 i *= PAGE_SIZE;
340 return i > buf->page_base ? i - buf->page_base : 0;
277e4ab7
TM		 341 }
342 }
343 return want;
344}
345
346static ssize_t
347xs_sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags, size_t seek)
348{
349 ssize_t ret;
350 if (seek != 0)
351 iov_iter_advance(&msg->msg_iter, seek);
352 ret = sock_recvmsg(sock, msg, flags);
353 return ret > 0 ? ret + seek : ret;
354}
355
356static ssize_t
357xs_read_kvec(struct socket *sock, struct msghdr *msg, int flags,
358 struct kvec *kvec, size_t count, size_t seek)
359{
0e9b4a82 360 iov_iter_kvec(&msg->msg_iter, READ, kvec, 1, count);
277e4ab7
TM		 361 return xs_sock_recvmsg(sock, msg, flags, seek);
362}
363
364static ssize_t
365xs_read_bvec(struct socket *sock, struct msghdr *msg, int flags,
366 struct bio_vec *bvec, unsigned long nr, size_t count,
367 size_t seek)
368{
0e9b4a82 369 iov_iter_bvec(&msg->msg_iter, READ, bvec, nr, count);
277e4ab7
TM		 370 return xs_sock_recvmsg(sock, msg, flags, seek);
371}
372
373static ssize_t
374xs_read_discard(struct socket *sock, struct msghdr *msg, int flags,
375 size_t count)
376{
b76a5afd
TM		 377 iov_iter_discard(&msg->msg_iter, READ, count);
378 return sock_recvmsg(sock, msg, flags);
277e4ab7
TM		 379}
380
6a829eb8
TM		 381#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
382static void
383xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
384{
385 struct bvec_iter bi = {
386 .bi_size = count,
387 };
388 struct bio_vec bv;
389
390 bvec_iter_advance(bvec, &bi, seek & PAGE_MASK);
391 for_each_bvec(bv, bvec, bi, bi)
392 flush_dcache_page(bv.bv_page);
393}
394#else
395static inline void
396xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek)
397{
398}
399#endif
400
277e4ab7
TM		 401static ssize_t
402xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags,
403 struct xdr_buf *buf, size_t count, size_t seek, size_t *read)
404{
405 size_t want, seek_init = seek, offset = 0;
406 ssize_t ret;
407
e92053a5
TM		 408 want = min_t(size_t, count, buf->head[0].iov_len);
409 if (seek < want) {
277e4ab7
TM		 410 ret = xs_read_kvec(sock, msg, flags, &buf->head[0], want, seek);
411 if (ret <= 0)
412 goto sock_err;
413 offset += ret;
414 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
415 goto out;
416 if (ret != want)
c4433055 417 goto out;
277e4ab7
TM		 418 seek = 0;
419 } else {
e92053a5
TM		 420 seek -= want;
421 offset += want;
277e4ab7 422 }
16e5e90f
TM		 423
424 want = xs_alloc_sparse_pages(buf,
425 min_t(size_t, count - offset, buf->page_len),
12a3ad61 426 GFP_KERNEL);
16e5e90f 427 if (seek < want) {
277e4ab7
TM		 428 ret = xs_read_bvec(sock, msg, flags, buf->bvec,
429 xdr_buf_pagecount(buf),
430 want + buf->page_base,
431 seek + buf->page_base);
432 if (ret <= 0)
433 goto sock_err;
6a829eb8 434 xs_flush_bvec(buf->bvec, ret, seek + buf->page_base);
277e4ab7
TM		 435 offset += ret - buf->page_base;
436 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
437 goto out;
438 if (ret != want)
c4433055 439 goto out;
277e4ab7
TM		 440 seek = 0;
441 } else {
16e5e90f
TM		 442 seek -= want;
443 offset += want;
277e4ab7 444 }
16e5e90f 445
e92053a5
TM		 446 want = min_t(size_t, count - offset, buf->tail[0].iov_len);
447 if (seek < want) {
277e4ab7
TM		 448 ret = xs_read_kvec(sock, msg, flags, &buf->tail[0], want, seek);
449 if (ret <= 0)
450 goto sock_err;
451 offset += ret;
452 if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
453 goto out;
454 if (ret != want)
c4433055 455 goto out;
9734ad57 456 } else if (offset < seek_init)
e92053a5 457 offset = seek_init;
277e4ab7 458 ret = -EMSGSIZE;
277e4ab7
TM		 459out:
460 *read = offset - seek_init;
461 return ret;
277e4ab7
TM		 462sock_err:
463 offset += seek;
464 goto out;
465}
466
467static void
468xs_read_header(struct sock_xprt *transport, struct xdr_buf *buf)
469{
470 if (!transport->recv.copied) {
471 if (buf->head[0].iov_len >= transport->recv.offset)
472 memcpy(buf->head[0].iov_base,
473 &transport->recv.xid,
474 transport->recv.offset);
475 transport->recv.copied = transport->recv.offset;
476 }
477}
478
479static bool
480xs_read_stream_request_done(struct sock_xprt *transport)
481{
482 return transport->recv.fraghdr & cpu_to_be32(RPC_LAST_STREAM_FRAGMENT);
483}
484
e92053a5
TM		 485static void
486xs_read_stream_check_eor(struct sock_xprt *transport,
487 struct msghdr *msg)
488{
489 if (xs_read_stream_request_done(transport))
490 msg->msg_flags |= MSG_EOR;
491}
492
277e4ab7
TM		 493static ssize_t
494xs_read_stream_request(struct sock_xprt *transport, struct msghdr *msg,
495 int flags, struct rpc_rqst *req)
496{
497 struct xdr_buf *buf = &req->rq_private_buf;
01f2f5b8
AH		 498 size_t want, uninitialized_var(read);
499 ssize_t uninitialized_var(ret);
277e4ab7
TM		 500
501 xs_read_header(transport, buf);
502
503 want = transport->recv.len - transport->recv.offset;
e92053a5
TM		 504 if (want != 0) {
505 ret = xs_read_xdr_buf(transport->sock, msg, flags, buf,
506 transport->recv.copied + want,
507 transport->recv.copied,
508 &read);
509 transport->recv.offset += read;
510 transport->recv.copied += read;
727fcc64 511 }
e92053a5 512
727fcc64 513 if (transport->recv.offset == transport->recv.len)
e92053a5 514 xs_read_stream_check_eor(transport, msg);
277e4ab7 515
e92053a5
TM		 516 if (want == 0)
517 return 0;
518
277e4ab7 519 switch (ret) {
c4433055
TM		 520 default:
521 break;
26781eab 522 case -EFAULT:
277e4ab7 523 case -EMSGSIZE:
26781eab 524 msg->msg_flags |= MSG_TRUNC;
c4433055 525 return read;
277e4ab7
TM		 526 case 0:
527 return -ESHUTDOWN;
277e4ab7 528 }
c4433055 529 return ret < 0 ? ret : read;
277e4ab7
TM		 530}
531
532static size_t
533xs_read_stream_headersize(bool isfrag)
534{
535 if (isfrag)
536 return sizeof(__be32);
537 return 3 * sizeof(__be32);
538}
539
540static ssize_t
541xs_read_stream_header(struct sock_xprt *transport, struct msghdr *msg,
542 int flags, size_t want, size_t seek)
543{
544 struct kvec kvec = {
545 .iov_base = &transport->recv.fraghdr,
546 .iov_len = want,
547 };
548 return xs_read_kvec(transport->sock, msg, flags, &kvec, want, seek);
549}
550
551#if defined(CONFIG_SUNRPC_BACKCHANNEL)
552static ssize_t
553xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
554{
555 struct rpc_xprt *xprt = &transport->xprt;
556 struct rpc_rqst *req;
557 ssize_t ret;
558
559 /* Look up and lock the request corresponding to the given XID */
560 req = xprt_lookup_bc_request(xprt, transport->recv.xid);
561 if (!req) {
562 printk(KERN_WARNING "Callback slot table overflowed\n");
563 return -ESHUTDOWN;
564 }
565
566 ret = xs_read_stream_request(transport, msg, flags, req);
567 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 568 xprt_complete_bc_request(req, transport->recv.copied);
277e4ab7
TM		 569
570 return ret;
571}
572#else /* CONFIG_SUNRPC_BACKCHANNEL */
573static ssize_t
574xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags)
575{
576 return -ESHUTDOWN;
577}
578#endif /* CONFIG_SUNRPC_BACKCHANNEL */
579
580static ssize_t
581xs_read_stream_reply(struct sock_xprt *transport, struct msghdr *msg, int flags)
582{
583 struct rpc_xprt *xprt = &transport->xprt;
584 struct rpc_rqst *req;
585 ssize_t ret = 0;
586
587 /* Look up and lock the request corresponding to the given XID */
588 spin_lock(&xprt->queue_lock);
589 req = xprt_lookup_rqst(xprt, transport->recv.xid);
590 if (!req) {
591 msg->msg_flags |= MSG_TRUNC;
592 goto out;
593 }
594 xprt_pin_rqst(req);
595 spin_unlock(&xprt->queue_lock);
596
597 ret = xs_read_stream_request(transport, msg, flags, req);
598
599 spin_lock(&xprt->queue_lock);
600 if (msg->msg_flags & (MSG_EOR|MSG_TRUNC))
c4433055 601 xprt_complete_rqst(req->rq_task, transport->recv.copied);
277e4ab7
TM		 602 xprt_unpin_rqst(req);
603out:
604 spin_unlock(&xprt->queue_lock);
605 return ret;
606}
607
608static ssize_t
609xs_read_stream(struct sock_xprt *transport, int flags)
610{
611 struct msghdr msg = { 0 };
612 size_t want, read = 0;
613 ssize_t ret = 0;
614
615 if (transport->recv.len == 0) {
616 want = xs_read_stream_headersize(transport->recv.copied != 0);
617 ret = xs_read_stream_header(transport, &msg, flags, want,
618 transport->recv.offset);
619 if (ret <= 0)
620 goto out_err;
621 transport->recv.offset = ret;
c4433055
TM		 622 if (transport->recv.offset != want)
623 return transport->recv.offset;
277e4ab7
TM		 624 transport->recv.len = be32_to_cpu(transport->recv.fraghdr) &
625 RPC_FRAGMENT_SIZE_MASK;
626 transport->recv.offset -= sizeof(transport->recv.fraghdr);
627 read = ret;
628 }
629
630 switch (be32_to_cpu(transport->recv.calldir)) {
c4433055
TM		 631 default:
632 msg.msg_flags |= MSG_TRUNC;
633 break;
277e4ab7
TM		 634 case RPC_CALL:
635 ret = xs_read_stream_call(transport, &msg, flags);
636 break;
637 case RPC_REPLY:
638 ret = xs_read_stream_reply(transport, &msg, flags);
639 }
640 if (msg.msg_flags & MSG_TRUNC) {
641 transport->recv.calldir = cpu_to_be32(-1);
642 transport->recv.copied = -1;
643 }
644 if (ret < 0)
645 goto out_err;
646 read += ret;
647 if (transport->recv.offset < transport->recv.len) {
c4433055
TM		 648 if (!(msg.msg_flags & MSG_TRUNC))
649 return read;
b76a5afd 650 msg.msg_flags = 0;
277e4ab7
TM		 651 ret = xs_read_discard(transport->sock, &msg, flags,
652 transport->recv.len - transport->recv.offset);
653 if (ret <= 0)
654 goto out_err;
655 transport->recv.offset += ret;
656 read += ret;
657 if (transport->recv.offset != transport->recv.len)
c4433055 658 return read;
277e4ab7
TM		 659 }
660 if (xs_read_stream_request_done(transport)) {
c50b8ee0 661 trace_xs_stream_read_request(transport);
277e4ab7
TM		 662 transport->recv.copied = 0;
663 }
664 transport->recv.offset = 0;
665 transport->recv.len = 0;
666 return read;
667out_err:
79462857 668 return ret != 0 ? ret : -ESHUTDOWN;
277e4ab7
TM		 669}
670
0ffe86f4
TM		 671static __poll_t xs_poll_socket(struct sock_xprt *transport)
672{
a73881c9
TM		 673 return transport->sock->ops->poll(transport->file, transport->sock,
674 NULL);
0ffe86f4
TM		 675}
676
677static bool xs_poll_socket_readable(struct sock_xprt *transport)
678{
679 __poll_t events = xs_poll_socket(transport);
680
681 return (events & (EPOLLIN | EPOLLRDNORM)) && !(events & EPOLLRDHUP);
682}
683
684static void xs_poll_check_readable(struct sock_xprt *transport)
685{
686
687 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
688 if (!xs_poll_socket_readable(transport))
689 return;
690 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
691 queue_work(xprtiod_workqueue, &transport->recv_worker);
692}
693
c50b8ee0
TM		 694static void xs_stream_data_receive(struct sock_xprt *transport)
695{
696 size_t read = 0;
697 ssize_t ret = 0;
698
699 mutex_lock(&transport->recv_mutex);
700 if (transport->sock == NULL)
701 goto out;
c50b8ee0
TM		 702 for (;;) {
703 ret = xs_read_stream(transport, MSG_DONTWAIT);
c4433055 704 if (ret < 0)
c50b8ee0
TM		 705 break;
706 read += ret;
707 cond_resched();
708 }
5f52a9d4
TM		 709 if (ret == -ESHUTDOWN)
710 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
711 else
712 xs_poll_check_readable(transport);
c50b8ee0
TM		 713out:
714 mutex_unlock(&transport->recv_mutex);
715 trace_xs_stream_read_data(&transport->xprt, ret, read);
716}
717
718static void xs_stream_data_receive_workfn(struct work_struct *work)
719{
720 struct sock_xprt *transport =
721 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 722 unsigned int pflags = memalloc_nofs_save();
723
c50b8ee0 724 xs_stream_data_receive(transport);
a1231fda 725 memalloc_nofs_restore(pflags);
c50b8ee0
TM		 726}
727
550aebfe
TM		 728static void
729xs_stream_reset_connect(struct sock_xprt *transport)
730{
731 transport->recv.offset = 0;
732 transport->recv.len = 0;
733 transport->recv.copied = 0;
734 transport->xmit.offset = 0;
ae053551
TM		 735}
736
737static void
738xs_stream_start_connect(struct sock_xprt *transport)
739{
550aebfe
TM		 740 transport->xprt.stat.connect_count++;
741 transport->xprt.stat.connect_start = jiffies;
742}
743
b4b5cc85
CL		 744#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
745
e791f8e9 746static int xs_sendmsg(struct socket *sock, struct msghdr *msg, size_t seek)
b4b5cc85 747{
e791f8e9
TM		 748 if (seek)
749 iov_iter_advance(&msg->msg_iter, seek);
750 return sock_sendmsg(sock, msg);
751}
b4b5cc85 752
e791f8e9
TM		 753static int xs_send_kvec(struct socket *sock, struct msghdr *msg, struct kvec *vec, size_t seek)
754{
e791f8e9
TM		 755 iov_iter_kvec(&msg->msg_iter, WRITE, vec, 1, vec->iov_len);
756 return xs_sendmsg(sock, msg, seek);
b4b5cc85
CL		 757}
758
0472e476 759static int xs_send_pagedata(struct socket *sock, struct msghdr *msg, struct xdr_buf *xdr, size_t base)
b4b5cc85 760{
f279cd00 761 int err;
24c5684b 762
0472e476
TM		 763 err = xdr_alloc_bvec(xdr, GFP_KERNEL);
764 if (err < 0)
765 return err;
b4b5cc85 766
0472e476
TM		 767 iov_iter_bvec(&msg->msg_iter, WRITE, xdr->bvec,
768 xdr_buf_pagecount(xdr),
769 xdr->page_len + xdr->page_base);
770 return xs_sendmsg(sock, msg, base + xdr->page_base);
b4b5cc85
CL		 771}
772
06b5fc3a
TM		 773#define xs_record_marker_len() sizeof(rpc_fraghdr)
774
067fb11b
CL		 775/* Common case:
776 * - stream transport
777 * - sending from byte 0 of the message
778 * - the message is wholly contained in @xdr's head iovec
779 */
06b5fc3a
TM		 780static int xs_send_rm_and_kvec(struct socket *sock, struct msghdr *msg,
781 rpc_fraghdr marker, struct kvec *vec, size_t base)
067fb11b 782{
067fb11b 783 struct kvec iov[2] = {
06b5fc3a 784 [0] = {
067fb11b
CL		 785 .iov_base = &marker,
786 .iov_len = sizeof(marker)
787 },
06b5fc3a 788 [1] = *vec,
b4b5cc85 789 };
06b5fc3a 790 size_t len = iov[0].iov_len + iov[1].iov_len;
b4b5cc85 791
06b5fc3a
TM		 792 iov_iter_kvec(&msg->msg_iter, WRITE, iov, 2, len);
793 return xs_sendmsg(sock, msg, base);
b4b5cc85
CL		 794}
795
9903cd1c
CL		 796/**
797 * xs_sendpages - write pages directly to a socket
798 * @sock: socket to send on
799 * @addr: UDP only -- address of destination
800 * @addrlen: UDP only -- length of destination address
801 * @xdr: buffer containing this request
802 * @base: starting position in the buffer
06b5fc3a 803 * @rm: stream record marker field
f279cd00 804 * @sent_p: return the total number of bytes successfully queued for sending
9903cd1c 805 *
a246b010 806 */
06b5fc3a 807static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, rpc_fraghdr rm, int *sent_p)
a246b010 808{
e791f8e9
TM		 809 struct msghdr msg = {
810 .msg_name = addr,
811 .msg_namelen = addrlen,
812 .msg_flags = XS_SENDMSG_FLAGS | MSG_MORE,
813 };
06b5fc3a
TM		 814 unsigned int rmsize = rm ? sizeof(rm) : 0;
815 unsigned int remainder = rmsize + xdr->len - base;
816 unsigned int want;
f279cd00 817 int err = 0;
a246b010 818
262965f5 819 if (unlikely(!sock))
fba91afb 820 return -ENOTSOCK;
262965f5 821
06b5fc3a
TM		 822 want = xdr->head[0].iov_len + rmsize;
823 if (base < want) {
824 unsigned int len = want - base;
24c5684b 825 remainder -= len;
e791f8e9
TM		 826 if (remainder == 0)
827 msg.msg_flags &= ~MSG_MORE;
06b5fc3a
TM		 828 if (rmsize)
829 err = xs_send_rm_and_kvec(sock, &msg, rm,
830 &xdr->head[0], base);
067fb11b 831 else
06b5fc3a 832 err = xs_send_kvec(sock, &msg, &xdr->head[0], base);
24c5684b 833 if (remainder == 0 || err != len)
a246b010 834 goto out;
f279cd00 835 *sent_p += err;
a246b010
CL		 836 base = 0;
837 } else
06b5fc3a 838 base -= want;
a246b010 839
24c5684b
TM		 840 if (base < xdr->page_len) {
841 unsigned int len = xdr->page_len - base;
842 remainder -= len;
0472e476
TM		 843 if (remainder == 0)
844 msg.msg_flags &= ~MSG_MORE;
845 err = xs_send_pagedata(sock, &msg, xdr, base);
846 if (remainder == 0 || err != len)
a246b010 847 goto out;
0472e476 848 *sent_p += err;
a246b010 849 base = 0;
24c5684b
TM		 850 } else
851 base -= xdr->page_len;
852
853 if (base >= xdr->tail[0].iov_len)
f279cd00 854 return 0;
e791f8e9
TM		 855 msg.msg_flags &= ~MSG_MORE;
856 err = xs_send_kvec(sock, &msg, &xdr->tail[0], base);
a246b010 857out:
f279cd00
JB		 858 if (err > 0) {
859 *sent_p += err;
860 err = 0;
861 }
862 return err;
a246b010
CL		 863}
864
9903cd1c 865/**
c544577d 866 * xs_nospace - handle transmit was incomplete
50f484e2 867 * @req: pointer to RPC request
9903cd1c 868 *
a246b010 869 */
c544577d 870static int xs_nospace(struct rpc_rqst *req)
a246b010 871{
262965f5 872 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 873 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
06ea0bfe 874 struct sock *sk = transport->inet;
24ca9a84 875 int ret = -EAGAIN;
a246b010 876
46121cf7 877 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
c544577d
TM		 878 req->rq_task->tk_pid,
879 req->rq_slen - transport->xmit.offset,
262965f5
CL		 880 req->rq_slen);
881
b6ddf64f 882 /* Protect against races with write_space */
b5e92419 883 spin_lock(&xprt->transport_lock);
b6ddf64f
TM		 884
885 /* Don't race with disconnect */
886 if (xprt_connected(xprt)) {
13331a55
TM		 887 /* wait for more buffer space */
888 sk->sk_write_pending++;
c544577d 889 xprt_wait_for_buffer_space(xprt);
13331a55 890 } else
5e3771ce 891 ret = -ENOTCONN;
262965f5 892
b5e92419 893 spin_unlock(&xprt->transport_lock);
06ea0bfe
TM		 894
895 /* Race breaker in case memory is freed before above code is called */
d48f9ce7
DV		 896 if (ret == -EAGAIN) {
897 struct socket_wq *wq;
898
899 rcu_read_lock();
900 wq = rcu_dereference(sk->sk_wq);
901 set_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags);
902 rcu_read_unlock();
903
904 sk->sk_write_space(sk);
905 }
5e3771ce 906 return ret;
262965f5
CL		 907}
908
277e4ab7
TM		 909static void
910xs_stream_prepare_request(struct rpc_rqst *req)
911{
12a3ad61 912 req->rq_task->tk_status = xdr_alloc_bvec(&req->rq_rcv_buf, GFP_KERNEL);
277e4ab7
TM		 913}
914
4cd34e7c
TM		 915/*
916 * Determine if the previous message in the stream was aborted before it
917 * could complete transmission.
918 */
919static bool
920xs_send_request_was_aborted(struct sock_xprt *transport, struct rpc_rqst *req)
921{
922 return transport->xmit.offset != 0 && req->rq_bytes_sent == 0;
923}
924
61677eee 925/*
06b5fc3a 926 * Return the stream record marker field for a record of length < 2^31-1
61677eee 927 */
06b5fc3a
TM		 928static rpc_fraghdr
929xs_stream_record_marker(struct xdr_buf *xdr)
61677eee 930{
06b5fc3a
TM		 931 if (!xdr->len)
932 return 0;
933 return cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | (u32)xdr->len);
61677eee
CL		 934}
935
176e21ee
CL		 936/**
937 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
50f484e2 938 * @req: pointer to RPC request
176e21ee
CL		 939 *
940 * Return values:
941 * 0: The request has been sent
942 * EAGAIN: The socket was blocked, please call again later to
943 * complete the request
944 * ENOTCONN: Caller needs to invoke connect logic then call again
945 * other: Some other error occured, the request was not sent
946 */
adfa7144 947static int xs_local_send_request(struct rpc_rqst *req)
176e21ee 948{
176e21ee
CL		 949 struct rpc_xprt *xprt = req->rq_xprt;
950 struct sock_xprt *transport =
951 container_of(xprt, struct sock_xprt, xprt);
952 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 953 rpc_fraghdr rm = xs_stream_record_marker(xdr);
954 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
176e21ee 955 int status;
f279cd00 956 int sent = 0;
176e21ee 957
4cd34e7c
TM		 958 /* Close the stream if the previous transmission was incomplete */
959 if (xs_send_request_was_aborted(transport, req)) {
960 xs_close(xprt);
961 return -ENOTCONN;
962 }
963
176e21ee
CL		 964 xs_pktdump("packet data:",
965 req->rq_svec->iov_base, req->rq_svec->iov_len);
966
78215759 967 req->rq_xtime = ktime_get();
6c7a64e5 968 status = xs_sendpages(transport->sock, NULL, 0, xdr,
7e3d3620 969 transport->xmit.offset, rm, &sent);
176e21ee 970 dprintk("RPC: %s(%u) = %d\n",
6c7a64e5 971 __func__, xdr->len - transport->xmit.offset, status);
743c69e7
N		 972
973 if (status == -EAGAIN && sock_writeable(transport->inet))
974 status = -ENOBUFS;
975
f279cd00 976 if (likely(sent > 0) || status == 0) {
6c7a64e5
TM		 977 transport->xmit.offset += sent;
978 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 979 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 980 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 981 transport->xmit.offset = 0;
176e21ee
CL		 982 return 0;
983 }
984 status = -EAGAIN;
985 }
986
987 switch (status) {
3601c4a9 988 case -ENOBUFS:
b5872f0c 989 break;
176e21ee 990 case -EAGAIN:
c544577d 991 status = xs_nospace(req);
176e21ee
CL		 992 break;
993 default:
994 dprintk("RPC: sendmsg returned unrecognized error %d\n",
995 -status);
e9d47639 996 /* fall through */
176e21ee
CL		 997 case -EPIPE:
998 xs_close(xprt);
999 status = -ENOTCONN;
1000 }
1001
1002 return status;
1003}
1004
262965f5
CL		 1005/**
1006 * xs_udp_send_request - write an RPC request to a UDP socket
50f484e2 1007 * @req: pointer to RPC request
262965f5
CL		 1008 *
1009 * Return values:
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
262965f5 1015 */
adfa7144 1016static int xs_udp_send_request(struct rpc_rqst *req)
262965f5 1017{
262965f5 1018 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 1019 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 1020 struct xdr_buf *xdr = &req->rq_snd_buf;
f279cd00 1021 int sent = 0;
262965f5 1022 int status;
a246b010 1023
9903cd1c 1024 xs_pktdump("packet data:",
a246b010
CL		 1025 req->rq_svec->iov_base,
1026 req->rq_svec->iov_len);
1027
01d37c42
TM		 1028 if (!xprt_bound(xprt))
1029 return -ENOTCONN;
75891f50
TM		 1030
1031 if (!xprt_request_get_cong(xprt, req))
1032 return -EBADSLT;
1033
78215759 1034 req->rq_xtime = ktime_get();
f279cd00 1035 status = xs_sendpages(transport->sock, xs_addr(xprt), xprt->addrlen,
06b5fc3a 1036 xdr, 0, 0, &sent);
a246b010 1037
46121cf7 1038 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
6c7a64e5 1039 xdr->len, status);
a246b010 1040
3dedbb5c
JB		 1041 /* firewall is blocking us, don't return -EAGAIN or we end up looping */
1042 if (status == -EPERM)
1043 goto process_status;
1044
743c69e7
N		 1045 if (status == -EAGAIN && sock_writeable(transport->inet))
1046 status = -ENOBUFS;
1047
f279cd00
JB		 1048 if (sent > 0 || status == 0) {
1049 req->rq_xmit_bytes_sent += sent;
1050 if (sent >= req->rq_slen)
2199700f
TM		 1051 return 0;
1052 /* Still some bytes left; set up for a retry later. */
262965f5 1053 status = -EAGAIN;
2199700f 1054 }
a246b010 1055
3dedbb5c 1056process_status:
262965f5 1057 switch (status) {
fba91afb
TM		 1058 case -ENOTSOCK:
1059 status = -ENOTCONN;
1060 /* Should we call xs_close() here? */
1061 break;
b6ddf64f 1062 case -EAGAIN:
c544577d 1063 status = xs_nospace(req);
b6ddf64f 1064 break;
262965f5 1065 case -ENETUNREACH:
3601c4a9 1066 case -ENOBUFS:
262965f5 1067 case -EPIPE:
a246b010 1068 case -ECONNREFUSED:
3dedbb5c 1069 case -EPERM:
a246b010 1070 /* When the server has died, an ICMP port unreachable message
9903cd1c 1071 * prompts ECONNREFUSED. */
13331a55
TM		 1072 break;
1073 default:
1074 dprintk("RPC: sendmsg returned unrecognized error %d\n",
1075 -status);
a246b010 1076 }
5fe46e9d 1077
262965f5 1078 return status;
a246b010
CL		 1079}
1080
9903cd1c 1081/**
262965f5 1082 * xs_tcp_send_request - write an RPC request to a TCP socket
50f484e2 1083 * @req: pointer to RPC request
9903cd1c
CL		 1084 *
1085 * Return values:
262965f5
CL		 1086 * 0: The request has been sent
1087 * EAGAIN: The socket was blocked, please call again later to
1088 * complete the request
1089 * ENOTCONN: Caller needs to invoke connect logic then call again
25985edc 1090 * other: Some other error occurred, the request was not sent
9903cd1c
CL		 1091 *
1092 * XXX: In the case of soft timeouts, should we eventually give up
262965f5 1093 * if sendmsg is not able to make progress?
9903cd1c 1094 */
adfa7144 1095static int xs_tcp_send_request(struct rpc_rqst *req)
a246b010 1096{
a246b010 1097 struct rpc_xprt *xprt = req->rq_xprt;
ee0ac0c2 1098 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262965f5 1099 struct xdr_buf *xdr = &req->rq_snd_buf;
7e3d3620
TM		 1100 rpc_fraghdr rm = xs_stream_record_marker(xdr);
1101 unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen;
9ffadfbc 1102 bool vm_wait = false;
b595bb15 1103 int status;
f279cd00 1104 int sent;
a246b010 1105
4cd34e7c
TM		 1106 /* Close the stream if the previous transmission was incomplete */
1107 if (xs_send_request_was_aborted(transport, req)) {
1108 if (transport->sock != NULL)
1109 kernel_sock_shutdown(transport->sock, SHUT_RDWR);
1110 return -ENOTCONN;
1111 }
1112
262965f5
CL		 1113 xs_pktdump("packet data:",
1114 req->rq_svec->iov_base,
1115 req->rq_svec->iov_len);
a246b010 1116
7196dbb0
TM		 1117 if (test_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state))
1118 xs_tcp_set_socket_timeouts(xprt, transport->sock);
1119
a246b010
CL		 1120 /* Continue transmitting the packet/record. We must be careful
1121 * to cope with writespace callbacks arriving _after_ we have
262965f5 1122 * called sendmsg(). */
78215759 1123 req->rq_xtime = ktime_get();
a246b010 1124 while (1) {
f279cd00
JB		 1125 sent = 0;
1126 status = xs_sendpages(transport->sock, NULL, 0, xdr,
7e3d3620 1127 transport->xmit.offset, rm, &sent);
a246b010 1128
46121cf7 1129 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
6c7a64e5 1130 xdr->len - transport->xmit.offset, status);
a246b010 1131
262965f5
CL		 1132 /* If we've sent the entire packet, immediately
1133 * reset the count of bytes sent. */
6c7a64e5
TM		 1134 transport->xmit.offset += sent;
1135 req->rq_bytes_sent = transport->xmit.offset;
7e3d3620 1136 if (likely(req->rq_bytes_sent >= msglen)) {
6c7a64e5 1137 req->rq_xmit_bytes_sent += transport->xmit.offset;
6c7a64e5 1138 transport->xmit.offset = 0;
262965f5
CL		 1139 return 0;
1140 }
a246b010 1141
9ffadfbc
TM		 1142 WARN_ON_ONCE(sent == 0 && status == 0);
1143
1144 if (status == -EAGAIN ) {
1145 /*
1146 * Return EAGAIN if we're sure we're hitting the
1147 * socket send buffer limits.
1148 */
1149 if (test_bit(SOCK_NOSPACE, &transport->sock->flags))
1150 break;
1151 /*
1152 * Did we hit a memory allocation failure?
1153 */
1154 if (sent == 0) {
1155 status = -ENOBUFS;
1156 if (vm_wait)
1157 break;
1158 /* Retry, knowing now that we're below the
1159 * socket send buffer limit
1160 */
1161 vm_wait = true;
1162 }
1163 continue;
1164 }
f580dd04
TM		 1165 if (status < 0)
1166 break;
9ffadfbc 1167 vm_wait = false;
a246b010
CL		 1168 }
1169
262965f5 1170 switch (status) {
fba91afb
TM		 1171 case -ENOTSOCK:
1172 status = -ENOTCONN;
1173 /* Should we call xs_close() here? */
1174 break;
262965f5 1175 case -EAGAIN:
c544577d 1176 status = xs_nospace(req);
262965f5 1177 break;
262965f5 1178 case -ECONNRESET:
2a9e1cfa 1179 case -ECONNREFUSED:
262965f5 1180 case -ENOTCONN:
3913c78c 1181 case -EADDRINUSE:
b5872f0c 1182 case -ENOBUFS:
b9d2bb2e 1183 case -EPIPE:
13331a55
TM		 1184 break;
1185 default:
1186 dprintk("RPC: sendmsg returned unrecognized error %d\n",
1187 -status);
a246b010 1188 }
5fe46e9d 1189
a246b010
CL		 1190 return status;
1191}
1192
2a9e1cfa
TM		 1193static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1194{
1195 transport->old_data_ready = sk->sk_data_ready;
1196 transport->old_state_change = sk->sk_state_change;
1197 transport->old_write_space = sk->sk_write_space;
2118071d 1198 transport->old_error_report = sk->sk_error_report;
2a9e1cfa
TM		 1199}
1200
1201static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
1202{
1203 sk->sk_data_ready = transport->old_data_ready;
1204 sk->sk_state_change = transport->old_state_change;
1205 sk->sk_write_space = transport->old_write_space;
2118071d
TM		 1206 sk->sk_error_report = transport->old_error_report;
1207}
1208
42d42a5b
TM		 1209static void xs_sock_reset_state_flags(struct rpc_xprt *xprt)
1210{
1211 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1212
1213 clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
4f8943f8
TM		 1214 clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state);
1215 clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state);
1216 clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state);
1217}
1218
1219static void xs_run_error_worker(struct sock_xprt *transport, unsigned int nr)
1220{
1221 set_bit(nr, &transport->sock_state);
1222 queue_work(xprtiod_workqueue, &transport->error_worker);
42d42a5b
TM		 1223}
1224
b70ae915
TM		 1225static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
1226{
1227 smp_mb__before_atomic();
1228 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1229 clear_bit(XPRT_CLOSING, &xprt->state);
42d42a5b 1230 xs_sock_reset_state_flags(xprt);
b70ae915
TM		 1231 smp_mb__after_atomic();
1232}
1233
2118071d
TM		 1234/**
1235 * xs_error_report - callback to handle TCP socket state errors
1236 * @sk: socket
1237 *
1238 * Note: we don't call sock_error() since there may be a rpc_task
1239 * using the socket, and so we don't want to clear sk->sk_err.
1240 */
1241static void xs_error_report(struct sock *sk)
1242{
4f8943f8 1243 struct sock_xprt *transport;
2118071d
TM		 1244 struct rpc_xprt *xprt;
1245 int err;
1246
1247 read_lock_bh(&sk->sk_callback_lock);
1248 if (!(xprt = xprt_from_sock(sk)))
1249 goto out;
1250
4f8943f8 1251 transport = container_of(xprt, struct sock_xprt, xprt);
2118071d
TM		 1252 err = -sk->sk_err;
1253 if (err == 0)
1254 goto out;
1255 dprintk("RPC: xs_error_report client %p, error=%d...\n",
1256 xprt, -err);
e8353c76 1257 trace_rpc_socket_error(xprt, sk->sk_socket, err);
4f8943f8 1258 xs_run_error_worker(transport, XPRT_SOCK_WAKE_ERROR);
2118071d
TM		 1259 out:
1260 read_unlock_bh(&sk->sk_callback_lock);
2a9e1cfa
TM		 1261}
1262
fe315e76 1263static void xs_reset_transport(struct sock_xprt *transport)
a246b010 1264{
ee0ac0c2
CL		 1265 struct socket *sock = transport->sock;
1266 struct sock *sk = transport->inet;
6cc7e908 1267 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1268 struct file *filp = transport->file;
a246b010 1269
fe315e76
CL		 1270 if (sk == NULL)
1271 return;
9903cd1c 1272
264d1df3
JL		 1273 if (atomic_read(&transport->xprt.swapper))
1274 sk_clear_memalloc(sk);
1275
09939204
TM		 1276 kernel_sock_shutdown(sock, SHUT_RDWR);
1277
edc1b01c 1278 mutex_lock(&transport->recv_mutex);
a246b010 1279 write_lock_bh(&sk->sk_callback_lock);
ee0ac0c2
CL		 1280 transport->inet = NULL;
1281 transport->sock = NULL;
a73881c9 1282 transport->file = NULL;
a246b010 1283
9903cd1c 1284 sk->sk_user_data = NULL;
2a9e1cfa
TM		 1285
1286 xs_restore_old_callbacks(transport, sk);
0c78789e 1287 xprt_clear_connected(xprt);
a246b010 1288 write_unlock_bh(&sk->sk_callback_lock);
6cc7e908 1289 xs_sock_reset_connection_flags(xprt);
ae053551
TM		 1290 /* Reset stream record info */
1291 xs_stream_reset_connect(transport);
edc1b01c 1292 mutex_unlock(&transport->recv_mutex);
a246b010 1293
6cc7e908 1294 trace_rpc_socket_close(xprt, sock);
a73881c9 1295 fput(filp);
0445f92c
TM		 1296
1297 xprt_disconnect_done(xprt);
fe315e76
CL		 1298}
1299
1300/**
1301 * xs_close - close a socket
1302 * @xprt: transport
1303 *
1304 * This is used when all requests are complete; ie, no DRC state remains
1305 * on the server we want to save.
f75e6745
TM		 1306 *
1307 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
1308 * xs_reset_transport() zeroing the socket from underneath a writer.
fe315e76
CL		 1309 */
1310static void xs_close(struct rpc_xprt *xprt)
1311{
1312 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1313
1314 dprintk("RPC: xs_close xprt %p\n", xprt);
1315
1316 xs_reset_transport(transport);
61d0a8e6 1317 xprt->reestablish_timeout = 0;
a246b010
CL		 1318}
1319
4a068258
CL		 1320static void xs_inject_disconnect(struct rpc_xprt *xprt)
1321{
1322 dprintk("RPC: injecting transport disconnect on xprt=%p\n",
1323 xprt);
1324 xprt_disconnect_done(xprt);
1325}
1326
315f3812
KM		 1327static void xs_xprt_free(struct rpc_xprt *xprt)
1328{
1329 xs_free_peer_addresses(xprt);
1330 xprt_free(xprt);
1331}
1332
9903cd1c
CL		 1333/**
1334 * xs_destroy - prepare to shutdown a transport
1335 * @xprt: doomed transport
1336 *
1337 */
1338static void xs_destroy(struct rpc_xprt *xprt)
a246b010 1339{
03c78827
TM		 1340 struct sock_xprt *transport = container_of(xprt,
1341 struct sock_xprt, xprt);
46121cf7 1342 dprintk("RPC: xs_destroy xprt %p\n", xprt);
9903cd1c 1343
03c78827 1344 cancel_delayed_work_sync(&transport->connect_worker);
a1311d87 1345 xs_close(xprt);
edc1b01c 1346 cancel_work_sync(&transport->recv_worker);
b5e92419 1347 cancel_work_sync(&transport->error_worker);
315f3812 1348 xs_xprt_free(xprt);
a1311d87 1349 module_put(THIS_MODULE);
a246b010
CL		 1350}
1351
9903cd1c 1352/**
f9b2ee71
TM		 1353 * xs_udp_data_read_skb - receive callback for UDP sockets
1354 * @xprt: transport
1355 * @sk: socket
1356 * @skb: skbuff
9903cd1c 1357 *
a246b010 1358 */
f9b2ee71
TM		 1359static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
1360 struct sock *sk,
1361 struct sk_buff *skb)
a246b010 1362{
9903cd1c 1363 struct rpc_task *task;
a246b010 1364 struct rpc_rqst *rovr;
f9b2ee71 1365 int repsize, copied;
d8ed029d
AD		 1366 u32 _xid;
1367 __be32 *xp;
a246b010 1368
1da8c681 1369 repsize = skb->len;
a246b010 1370 if (repsize < 4) {
46121cf7 1371 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
f9b2ee71 1372 return;
a246b010
CL		 1373 }
1374
1375 /* Copy the XID from the skb... */
1da8c681 1376 xp = skb_header_pointer(skb, 0, sizeof(_xid), &_xid);
a246b010 1377 if (xp == NULL)
f9b2ee71 1378 return;
a246b010
CL		 1379
1380 /* Look up and lock the request corresponding to the given XID */
75c84151 1381 spin_lock(&xprt->queue_lock);
a246b010
CL		 1382 rovr = xprt_lookup_rqst(xprt, *xp);
1383 if (!rovr)
1384 goto out_unlock;
729749bb 1385 xprt_pin_rqst(rovr);
ecd465ee 1386 xprt_update_rtt(rovr->rq_task);
75c84151 1387 spin_unlock(&xprt->queue_lock);
a246b010
CL		 1388 task = rovr->rq_task;
1389
a246b010
CL		 1390 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1391 copied = repsize;
1392
1393 /* Suck it into the iovec, verify checksum if not done by hw. */
1781f7f5 1394 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
75c84151 1395 spin_lock(&xprt->queue_lock);
0afa6b44 1396 __UDPX_INC_STATS(sk, UDP_MIB_INERRORS);
729749bb 1397 goto out_unpin;
1781f7f5
HX		 1398 }
1399
a246b010 1400
b5e92419 1401 spin_lock(&xprt->transport_lock);
6a24dfb6 1402 xprt_adjust_cwnd(xprt, task, copied);
b5e92419 1403 spin_unlock(&xprt->transport_lock);
75c84151 1404 spin_lock(&xprt->queue_lock);
1570c1e4 1405 xprt_complete_rqst(task, copied);
0afa6b44 1406 __UDPX_INC_STATS(sk, UDP_MIB_INDATAGRAMS);
729749bb
TM		 1407out_unpin:
1408 xprt_unpin_rqst(rovr);
a246b010 1409 out_unlock:
75c84151 1410 spin_unlock(&xprt->queue_lock);
f9b2ee71
TM		 1411}
1412
1413static void xs_udp_data_receive(struct sock_xprt *transport)
1414{
1415 struct sk_buff *skb;
1416 struct sock *sk;
1417 int err;
1418
1419 mutex_lock(&transport->recv_mutex);
1420 sk = transport->inet;
1421 if (sk == NULL)
1422 goto out;
1423 for (;;) {
7c13f97f 1424 skb = skb_recv_udp(sk, 0, 1, &err);
4f546149 1425 if (skb == NULL)
f9b2ee71 1426 break;
4f546149
TM		 1427 xs_udp_data_read_skb(&transport->xprt, sk, skb);
1428 consume_skb(skb);
1429 cond_resched();
f9b2ee71 1430 }
0ffe86f4 1431 xs_poll_check_readable(transport);
f9b2ee71
TM		 1432out:
1433 mutex_unlock(&transport->recv_mutex);
1434}
1435
1436static void xs_udp_data_receive_workfn(struct work_struct *work)
1437{
1438 struct sock_xprt *transport =
1439 container_of(work, struct sock_xprt, recv_worker);
a1231fda
TM		 1440 unsigned int pflags = memalloc_nofs_save();
1441
f9b2ee71 1442 xs_udp_data_receive(transport);
a1231fda 1443 memalloc_nofs_restore(pflags);
f9b2ee71
TM		 1444}
1445
1446/**
1447 * xs_data_ready - "data ready" callback for UDP sockets
1448 * @sk: socket with data to read
1449 *
1450 */
1451static void xs_data_ready(struct sock *sk)
1452{
1453 struct rpc_xprt *xprt;
1454
1455 read_lock_bh(&sk->sk_callback_lock);
1456 dprintk("RPC: xs_data_ready...\n");
1457 xprt = xprt_from_sock(sk);
1458 if (xprt != NULL) {
1459 struct sock_xprt *transport = container_of(xprt,
1460 struct sock_xprt, xprt);
5157b956
TM		 1461 transport->old_data_ready(sk);
1462 /* Any data means we had a useful conversation, so
1463 * then we don't need to delay the next reconnect
1464 */
1465 if (xprt->reestablish_timeout)
1466 xprt->reestablish_timeout = 0;
42d42a5b 1467 if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
40a5f1b1 1468 queue_work(xprtiod_workqueue, &transport->recv_worker);
f9b2ee71 1469 }
f064af1e 1470 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1471}
1472
a519fc7a
TM		 1473/*
1474 * Helper function to force a TCP close if the server is sending
1475 * junk and/or it has put us in CLOSE_WAIT
1476 */
1477static void xs_tcp_force_close(struct rpc_xprt *xprt)
1478{
a519fc7a
TM		 1479 xprt_force_disconnect(xprt);
1480}
1481
9e00abc3 1482#if defined(CONFIG_SUNRPC_BACKCHANNEL)
6b26cc8c
CL		 1483static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt)
1484{
1485 return PAGE_SIZE;
1486}
9e00abc3 1487#endif /* CONFIG_SUNRPC_BACKCHANNEL */
44b98efd 1488
9903cd1c
CL		 1489/**
1490 * xs_tcp_state_change - callback to handle TCP socket state changes
1491 * @sk: socket whose state has changed
1492 *
1493 */
1494static void xs_tcp_state_change(struct sock *sk)
a246b010 1495{
9903cd1c 1496 struct rpc_xprt *xprt;
0fdea1e8 1497 struct sock_xprt *transport;
a246b010 1498
f064af1e 1499 read_lock_bh(&sk->sk_callback_lock);
a246b010
CL		 1500 if (!(xprt = xprt_from_sock(sk)))
1501 goto out;
46121cf7 1502 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
669502ff 1503 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
46121cf7
CL		 1504 sk->sk_state, xprt_connected(xprt),
1505 sock_flag(sk, SOCK_DEAD),
669502ff
AC		 1506 sock_flag(sk, SOCK_ZAPPED),
1507 sk->sk_shutdown);
a246b010 1508
0fdea1e8 1509 transport = container_of(xprt, struct sock_xprt, xprt);
40b5ea0c 1510 trace_rpc_socket_state_change(xprt, sk->sk_socket);
a246b010
CL		 1511 switch (sk->sk_state) {
1512 case TCP_ESTABLISHED:
a246b010 1513 if (!xprt_test_and_set_connected(xprt)) {
8b71798c 1514 xprt->connect_cookie++;
0fdea1e8
TM		 1515 clear_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
1516 xprt_clear_connecting(xprt);
51971139 1517
3968a8a5
CL		 1518 xprt->stat.connect_count++;
1519 xprt->stat.connect_time += (long)jiffies -
1520 xprt->stat.connect_start;
4f8943f8 1521 xs_run_error_worker(transport, XPRT_SOCK_WAKE_PENDING);
a246b010 1522 }
a246b010 1523 break;
3b948ae5
TM		 1524 case TCP_FIN_WAIT1:
1525 /* The client initiated a shutdown of the socket */
7c1d71cf 1526 xprt->connect_cookie++;
663b8858 1527 xprt->reestablish_timeout = 0;
3b948ae5 1528 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1529 smp_mb__before_atomic();
3b948ae5 1530 clear_bit(XPRT_CONNECTED, &xprt->state);
ef803670 1531 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
4e857c58 1532 smp_mb__after_atomic();
a246b010 1533 break;
632e3bdc 1534 case TCP_CLOSE_WAIT:
3b948ae5 1535 /* The server initiated a shutdown of the socket */
7c1d71cf 1536 xprt->connect_cookie++;
d0bea455 1537 clear_bit(XPRT_CONNECTED, &xprt->state);
4f8943f8 1538 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
e9d47639 1539 /* fall through */
663b8858
TM		 1540 case TCP_CLOSING:
1541 /*
1542 * If the server closed down the connection, make sure that
1543 * we back off before reconnecting
1544 */
1545 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1546 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3b948ae5
TM		 1547 break;
1548 case TCP_LAST_ACK:
670f9457 1549 set_bit(XPRT_CLOSING, &xprt->state);
4e857c58 1550 smp_mb__before_atomic();
3b948ae5 1551 clear_bit(XPRT_CONNECTED, &xprt->state);
4e857c58 1552 smp_mb__after_atomic();
3b948ae5
TM		 1553 break;
1554 case TCP_CLOSE:
0fdea1e8
TM		 1555 if (test_and_clear_bit(XPRT_SOCK_CONNECTING,
1556 &transport->sock_state))
1557 xprt_clear_connecting(xprt);
9b30889c 1558 clear_bit(XPRT_CLOSING, &xprt->state);
9b30889c 1559 /* Trigger the socket release */
4f8943f8 1560 xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT);
a246b010
CL		 1561 }
1562 out:
f064af1e 1563 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1564}
1565
1f0fa154
IJ		 1566static void xs_write_space(struct sock *sk)
1567{
13331a55 1568 struct socket_wq *wq;
4f8943f8 1569 struct sock_xprt *transport;
1f0fa154
IJ		 1570 struct rpc_xprt *xprt;
1571
13331a55 1572 if (!sk->sk_socket)
1f0fa154 1573 return;
13331a55 1574 clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1f0fa154
IJ		 1575
1576 if (unlikely(!(xprt = xprt_from_sock(sk))))
1577 return;
4f8943f8 1578 transport = container_of(xprt, struct sock_xprt, xprt);
13331a55
TM		 1579 rcu_read_lock();
1580 wq = rcu_dereference(sk->sk_wq);
1581 if (!wq || test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags) == 0)
1582 goto out;
1f0fa154 1583
4f8943f8
TM		 1584 xs_run_error_worker(transport, XPRT_SOCK_WAKE_WRITE);
1585 sk->sk_write_pending--;
13331a55
TM		 1586out:
1587 rcu_read_unlock();
1f0fa154
IJ		 1588}
1589
9903cd1c 1590/**
c7b2cae8
CL		 1591 * xs_udp_write_space - callback invoked when socket buffer space
1592 * becomes available
9903cd1c
CL		 1593 * @sk: socket whose state has changed
1594 *
a246b010
CL		 1595 * Called when more output buffer space is available for this socket.
1596 * We try not to wake our writers until they can make "significant"
c7b2cae8 1597 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
a246b010
CL		 1598 * with a bunch of small requests.
1599 */
c7b2cae8 1600static void xs_udp_write_space(struct sock *sk)
a246b010 1601{
f064af1e 1602 read_lock_bh(&sk->sk_callback_lock);
a246b010 1603
c7b2cae8 1604 /* from net/core/sock.c:sock_def_write_space */
1f0fa154
IJ		 1605 if (sock_writeable(sk))
1606 xs_write_space(sk);
a246b010 1607
f064af1e 1608 read_unlock_bh(&sk->sk_callback_lock);
c7b2cae8 1609}
a246b010 1610
c7b2cae8
CL		 1611/**
1612 * xs_tcp_write_space - callback invoked when socket buffer space
1613 * becomes available
1614 * @sk: socket whose state has changed
1615 *
1616 * Called when more output buffer space is available for this socket.
1617 * We try not to wake our writers until they can make "significant"
1618 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1619 * with a bunch of small requests.
1620 */
1621static void xs_tcp_write_space(struct sock *sk)
1622{
f064af1e 1623 read_lock_bh(&sk->sk_callback_lock);
c7b2cae8
CL		 1624
1625 /* from net/core/stream.c:sk_stream_write_space */
64dc6130 1626 if (sk_stream_is_writeable(sk))
1f0fa154 1627 xs_write_space(sk);
b6ddf64f 1628
f064af1e 1629 read_unlock_bh(&sk->sk_callback_lock);
a246b010
CL		 1630}
1631
470056c2 1632static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
a246b010 1633{
ee0ac0c2
CL		 1634 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1635 struct sock *sk = transport->inet;
a246b010 1636
7c6e066e 1637 if (transport->rcvsize) {
a246b010 1638 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
7c6e066e 1639 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
a246b010 1640 }
7c6e066e 1641 if (transport->sndsize) {
a246b010 1642 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
7c6e066e 1643 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
a246b010
CL		 1644 sk->sk_write_space(sk);
1645 }
1646}
1647
43118c29 1648/**
470056c2 1649 * xs_udp_set_buffer_size - set send and receive limits
43118c29 1650 * @xprt: generic transport
470056c2
CL		 1651 * @sndsize: requested size of send buffer, in bytes
1652 * @rcvsize: requested size of receive buffer, in bytes
43118c29 1653 *
470056c2 1654 * Set socket send and receive buffer size limits.
43118c29 1655 */
470056c2 1656static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
43118c29 1657{
7c6e066e
CL		 1658 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1659
1660 transport->sndsize = 0;
470056c2 1661 if (sndsize)
7c6e066e
CL		 1662 transport->sndsize = sndsize + 1024;
1663 transport->rcvsize = 0;
470056c2 1664 if (rcvsize)
7c6e066e 1665 transport->rcvsize = rcvsize + 1024;
470056c2
CL		 1666
1667 xs_udp_do_set_buffer_size(xprt);
43118c29
CL		 1668}
1669
46c0ee8b
CL		 1670/**
1671 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
acf0a39f 1672 * @xprt: controlling transport
46c0ee8b
CL		 1673 * @task: task that timed out
1674 *
1675 * Adjust the congestion window after a retransmit timeout has occurred.
1676 */
6a24dfb6 1677static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
46c0ee8b 1678{
b5e92419 1679 spin_lock(&xprt->transport_lock);
6a24dfb6 1680 xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
b5e92419 1681 spin_unlock(&xprt->transport_lock);
46c0ee8b
CL		 1682}
1683
826799e6 1684static int xs_get_random_port(void)
b85d8806 1685{
826799e6
BF		 1686 unsigned short min = xprt_min_resvport, max = xprt_max_resvport;
1687 unsigned short range;
1688 unsigned short rand;
1689
1690 if (max < min)
1691 return -EADDRINUSE;
1692 range = max - min + 1;
1693 rand = (unsigned short) prandom_u32() % range;
1694 return rand + min;
b85d8806
CL		 1695}
1696
4dda9c8a
TM		 1697/**
1698 * xs_set_reuseaddr_port - set the socket's port and address reuse options
1699 * @sock: socket
1700 *
1701 * Note that this function has to be called on all sockets that share the
1702 * same port, and it must be called before binding.
1703 */
1704static void xs_sock_set_reuseport(struct socket *sock)
1705{
402e23b4 1706 int opt = 1;
4dda9c8a 1707
402e23b4
TM		 1708 kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
1709 (char *)&opt, sizeof(opt));
4dda9c8a
TM		 1710}
1711
1712static unsigned short xs_sock_getport(struct socket *sock)
1713{
1714 struct sockaddr_storage buf;
4dda9c8a
TM		 1715 unsigned short port = 0;
1716
9b2c45d4 1717 if (kernel_getsockname(sock, (struct sockaddr *)&buf) < 0)
4dda9c8a
TM		 1718 goto out;
1719 switch (buf.ss_family) {
1720 case AF_INET6:
1721 port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port);
1722 break;
1723 case AF_INET:
1724 port = ntohs(((struct sockaddr_in *)&buf)->sin_port);
1725 }
1726out:
1727 return port;
1728}
1729
92200412
CL		 1730/**
1731 * xs_set_port - reset the port number in the remote endpoint address
1732 * @xprt: generic transport
1733 * @port: new port number
1734 *
1735 */
1736static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1737{
46121cf7 1738 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
c4efcb1d 1739
9dc3b095
CL		 1740 rpc_set_port(xs_addr(xprt), port);
1741 xs_update_peer_port(xprt);
92200412
CL		 1742}
1743
4dda9c8a
TM		 1744static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
1745{
1746 if (transport->srcport == 0)
1747 transport->srcport = xs_sock_getport(sock);
1748}
1749
826799e6 1750static int xs_get_srcport(struct sock_xprt *transport)
67a391d7 1751{
826799e6 1752 int port = transport->srcport;
67a391d7
TM		 1753
1754 if (port == 0 && transport->xprt.resvport)
1755 port = xs_get_random_port();
1756 return port;
1757}
1758
baaf4e48 1759static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
67a391d7 1760{
fbfffbd5
CL		 1761 if (transport->srcport != 0)
1762 transport->srcport = 0;
67a391d7
TM		 1763 if (!transport->xprt.resvport)
1764 return 0;
1765 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1766 return xprt_max_resvport;
1767 return --port;
1768}
beb59b68 1769static int xs_bind(struct sock_xprt *transport, struct socket *sock)
a246b010 1770{
beb59b68 1771 struct sockaddr_storage myaddr;
67a391d7 1772 int err, nloop = 0;
826799e6 1773 int port = xs_get_srcport(transport);
67a391d7 1774 unsigned short last;
a246b010 1775
0f7a622c
CP		 1776 /*
1777 * If we are asking for any ephemeral port (i.e. port == 0 &&
1778 * transport->xprt.resvport == 0), don't bind. Let the local
1779 * port selection happen implicitly when the socket is used
1780 * (for example at connect time).
1781 *
1782 * This ensures that we can continue to establish TCP
1783 * connections even when all local ephemeral ports are already
1784 * a part of some TCP connection. This makes no difference
1785 * for UDP sockets, but also doens't harm them.
1786 *
1787 * If we're asking for any reserved port (i.e. port == 0 &&
1788 * transport->xprt.resvport == 1) xs_get_srcport above will
1789 * ensure that port is non-zero and we will bind as needed.
1790 */
826799e6
BF		 1791 if (port <= 0)
1792 return port;
0f7a622c 1793
beb59b68 1794 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
a246b010 1795 do {
beb59b68
PE		 1796 rpc_set_port((struct sockaddr *)&myaddr, port);
1797 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1798 transport->xprt.addrlen);
a246b010 1799 if (err == 0) {
fbfffbd5 1800 transport->srcport = port;
d3bc9a1d 1801 break;
a246b010 1802 }
67a391d7 1803 last = port;
baaf4e48 1804 port = xs_next_srcport(transport, port);
67a391d7
TM		 1805 if (port > last)
1806 nloop++;
1807 } while (err == -EADDRINUSE && nloop != 2);
90058d37 1808
4232e863 1809 if (myaddr.ss_family == AF_INET)
beb59b68
PE		 1810 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1811 &((struct sockaddr_in *)&myaddr)->sin_addr,
1812 port, err ? "failed" : "ok", err);
1813 else
1814 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1815 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1816 port, err ? "failed" : "ok", err);
a246b010
CL		 1817 return err;
1818}
1819
176e21ee
CL		 1820/*
1821 * We don't support autobind on AF_LOCAL sockets
1822 */
1823static void xs_local_rpcbind(struct rpc_task *task)
1824{
fb43d172 1825 xprt_set_bound(task->tk_xprt);
176e21ee
CL		 1826}
1827
1828static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1829{
1830}
beb59b68 1831
ed07536e
PZ		 1832#ifdef CONFIG_DEBUG_LOCK_ALLOC
1833static struct lock_class_key xs_key[2];
1834static struct lock_class_key xs_slock_key[2];
1835
176e21ee
CL		 1836static inline void xs_reclassify_socketu(struct socket *sock)
1837{
1838 struct sock *sk = sock->sk;
1839
176e21ee
CL		 1840 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1841 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1842}
1843
8945ee5e 1844static inline void xs_reclassify_socket4(struct socket *sock)
ed07536e
PZ		 1845{
1846 struct sock *sk = sock->sk;
8945ee5e 1847
8945ee5e
CL		 1848 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1849 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1850}
ed07536e 1851
8945ee5e
CL		 1852static inline void xs_reclassify_socket6(struct socket *sock)
1853{
1854 struct sock *sk = sock->sk;
ed07536e 1855
8945ee5e
CL		 1856 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1857 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
ed07536e 1858}
6bc9638a
PE		 1859
1860static inline void xs_reclassify_socket(int family, struct socket *sock)
1861{
fafc4e1e 1862 if (WARN_ON_ONCE(!sock_allow_reclassification(sock->sk)))
1b7a1819
WAA		 1863 return;
1864
4232e863 1865 switch (family) {
176e21ee
CL		 1866 case AF_LOCAL:
1867 xs_reclassify_socketu(sock);
1868 break;
4232e863 1869 case AF_INET:
6bc9638a 1870 xs_reclassify_socket4(sock);
4232e863
CL		 1871 break;
1872 case AF_INET6:
6bc9638a 1873 xs_reclassify_socket6(sock);
4232e863
CL		 1874 break;
1875 }
6bc9638a 1876}
ed07536e 1877#else
6bc9638a
PE		 1878static inline void xs_reclassify_socket(int family, struct socket *sock)
1879{
1880}
ed07536e
PZ		 1881#endif
1882
93dc41bd
N		 1883static void xs_dummy_setup_socket(struct work_struct *work)
1884{
1885}
1886
6bc9638a 1887static struct socket *xs_create_sock(struct rpc_xprt *xprt,
4dda9c8a
TM		 1888 struct sock_xprt *transport, int family, int type,
1889 int protocol, bool reuseport)
22f79326 1890{
a73881c9 1891 struct file *filp;
22f79326
PE		 1892 struct socket *sock;
1893 int err;
1894
6bc9638a 1895 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
22f79326
PE		 1896 if (err < 0) {
1897 dprintk("RPC: can't create %d transport socket (%d).\n",
1898 protocol, -err);
1899 goto out;
1900 }
6bc9638a 1901 xs_reclassify_socket(family, sock);
22f79326 1902
4dda9c8a
TM		 1903 if (reuseport)
1904 xs_sock_set_reuseport(sock);
1905
4cea288a
BH		 1906 err = xs_bind(transport, sock);
1907 if (err) {
22f79326
PE		 1908 sock_release(sock);
1909 goto out;
1910 }
1911
a73881c9
TM		 1912 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1913 if (IS_ERR(filp))
1914 return ERR_CAST(filp);
1915 transport->file = filp;
1916
22f79326
PE		 1917 return sock;
1918out:
1919 return ERR_PTR(err);
1920}
1921
176e21ee
CL		 1922static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1923 struct socket *sock)
1924{
1925 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1926 xprt);
1927
1928 if (!transport->inet) {
1929 struct sock *sk = sock->sk;
1930
1931 write_lock_bh(&sk->sk_callback_lock);
1932
1933 xs_save_old_callbacks(transport, sk);
1934
1935 sk->sk_user_data = xprt;
a2648094 1936 sk->sk_data_ready = xs_data_ready;
176e21ee 1937 sk->sk_write_space = xs_udp_write_space;
b4411457 1938 sock_set_flag(sk, SOCK_FASYNC);
2118071d 1939 sk->sk_error_report = xs_error_report;
176e21ee
CL		 1940
1941 xprt_clear_connected(xprt);
1942
1943 /* Reset to new socket */
1944 transport->sock = sock;
1945 transport->inet = sk;
1946
1947 write_unlock_bh(&sk->sk_callback_lock);
1948 }
1949
ae053551 1950 xs_stream_start_connect(transport);
6c7a64e5 1951
176e21ee
CL		 1952 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1953}
1954
1955/**
1956 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
176e21ee 1957 * @transport: socket transport to connect
176e21ee 1958 */
dc107402 1959static int xs_local_setup_socket(struct sock_xprt *transport)
176e21ee 1960{
176e21ee 1961 struct rpc_xprt *xprt = &transport->xprt;
a73881c9 1962 struct file *filp;
176e21ee
CL		 1963 struct socket *sock;
1964 int status = -EIO;
1965
176e21ee
CL		 1966 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1967 SOCK_STREAM, 0, &sock, 1);
1968 if (status < 0) {
1969 dprintk("RPC: can't create AF_LOCAL "
1970 "transport socket (%d).\n", -status);
1971 goto out;
1972 }
d1358917 1973 xs_reclassify_socket(AF_LOCAL, sock);
176e21ee 1974
a73881c9
TM		 1975 filp = sock_alloc_file(sock, O_NONBLOCK, NULL);
1976 if (IS_ERR(filp)) {
1977 status = PTR_ERR(filp);
1978 goto out;
1979 }
1980 transport->file = filp;
1981
176e21ee
CL		 1982 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1983 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1984
1985 status = xs_local_finish_connecting(xprt, sock);
40b5ea0c 1986 trace_rpc_socket_connect(xprt, sock, status);
176e21ee
CL		 1987 switch (status) {
1988 case 0:
1989 dprintk("RPC: xprt %p connected to %s\n",
1990 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
3968a8a5
CL		 1991 xprt->stat.connect_count++;
1992 xprt->stat.connect_time += (long)jiffies -
1993 xprt->stat.connect_start;
176e21ee 1994 xprt_set_connected(xprt);
3601c4a9 1995 case -ENOBUFS:
176e21ee
CL		 1996 break;
1997 case -ENOENT:
1998 dprintk("RPC: xprt %p: socket %s does not exist\n",
1999 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2000 break;
4a20a988
TM		 2001 case -ECONNREFUSED:
2002 dprintk("RPC: xprt %p: connection refused for %s\n",
2003 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2004 break;
176e21ee
CL		 2005 default:
2006 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
2007 __func__, -status,
2008 xprt->address_strings[RPC_DISPLAY_ADDR]);
2009 }
2010
2011out:
2012 xprt_clear_connecting(xprt);
2013 xprt_wake_pending_tasks(xprt, status);
dc107402
BF		 2014 return status;
2015}
2016
b6669737 2017static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
dc107402 2018{
dc107402
BF		 2019 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2020 int ret;
2021
2022 if (RPC_IS_ASYNC(task)) {
2023 /*
2024 * We want the AF_LOCAL connect to be resolved in the
2025 * filesystem namespace of the process making the rpc
2026 * call. Thus we connect synchronously.
2027 *
2028 * If we want to support asynchronous AF_LOCAL calls,
2029 * we'll need to figure out how to pass a namespace to
2030 * connect.
2031 */
5ad64b36 2032 task->tk_rpc_status = -ENOTCONN;
dc107402
BF		 2033 rpc_exit(task, -ENOTCONN);
2034 return;
2035 }
2036 ret = xs_local_setup_socket(transport);
2037 if (ret && !RPC_IS_SOFTCONN(task))
2038 msleep_interruptible(15000);
176e21ee
CL		 2039}
2040
3c87ef6e 2041#if IS_ENABLED(CONFIG_SUNRPC_SWAP)
d6e971d8
JL		 2042/*
2043 * Note that this should be called with XPRT_LOCKED held (or when we otherwise
2044 * know that we have exclusive access to the socket), to guard against
2045 * races with xs_reset_transport.
2046 */
a564b8f0
MG		 2047static void xs_set_memalloc(struct rpc_xprt *xprt)
2048{
2049 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
2050 xprt);
2051
d6e971d8
JL		 2052 /*
2053 * If there's no sock, then we have nothing to set. The
2054 * reconnecting process will get it for us.
2055 */
2056 if (!transport->inet)
2057 return;
8e228133 2058 if (atomic_read(&xprt->swapper))
a564b8f0
MG		 2059 sk_set_memalloc(transport->inet);
2060}
2061
2062/**
d67fa4d8 2063 * xs_enable_swap - Tag this transport as being used for swap.
a564b8f0 2064 * @xprt: transport to tag
a564b8f0 2065 *
8e228133
JL		 2066 * Take a reference to this transport on behalf of the rpc_clnt, and
2067 * optionally mark it for swapping if it wasn't already.
a564b8f0 2068 */
d67fa4d8
JL		 2069static int
2070xs_enable_swap(struct rpc_xprt *xprt)
a564b8f0 2071{
d6e971d8 2072 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
a564b8f0 2073
d6e971d8
JL		 2074 if (atomic_inc_return(&xprt->swapper) != 1)
2075 return 0;
2076 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2077 return -ERESTARTSYS;
2078 if (xs->inet)
2079 sk_set_memalloc(xs->inet);
2080 xprt_release_xprt(xprt, NULL);
8e228133
JL		 2081 return 0;
2082}
a564b8f0 2083
8e228133 2084/**
d67fa4d8 2085 * xs_disable_swap - Untag this transport as being used for swap.
8e228133
JL		 2086 * @xprt: transport to tag
2087 *
2088 * Drop a "swapper" reference to this xprt on behalf of the rpc_clnt. If the
2089 * swapper refcount goes to 0, untag the socket as a memalloc socket.
2090 */
d67fa4d8
JL		 2091static void
2092xs_disable_swap(struct rpc_xprt *xprt)
8e228133 2093{
d6e971d8 2094 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
8e228133 2095
d6e971d8
JL		 2096 if (!atomic_dec_and_test(&xprt->swapper))
2097 return;
2098 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2099 return;
2100 if (xs->inet)
2101 sk_clear_memalloc(xs->inet);
2102 xprt_release_xprt(xprt, NULL);
a564b8f0 2103}
a564b8f0
MG		 2104#else
2105static void xs_set_memalloc(struct rpc_xprt *xprt)
2106{
2107}
d67fa4d8
JL		 2108
2109static int
2110xs_enable_swap(struct rpc_xprt *xprt)
2111{
2112 return -EINVAL;
2113}
2114
2115static void
2116xs_disable_swap(struct rpc_xprt *xprt)
2117{
2118}
a564b8f0
MG		 2119#endif
2120
16be2d20
CL		 2121static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
2122{
2123 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2124
2125 if (!transport->inet) {
2126 struct sock *sk = sock->sk;
2127
2128 write_lock_bh(&sk->sk_callback_lock);
2129
2a9e1cfa
TM		 2130 xs_save_old_callbacks(transport, sk);
2131
16be2d20 2132 sk->sk_user_data = xprt;
f9b2ee71 2133 sk->sk_data_ready = xs_data_ready;
16be2d20 2134 sk->sk_write_space = xs_udp_write_space;
b4411457 2135 sock_set_flag(sk, SOCK_FASYNC);
16be2d20
CL		 2136
2137 xprt_set_connected(xprt);
2138
2139 /* Reset to new socket */
2140 transport->sock = sock;
2141 transport->inet = sk;
2142
a564b8f0
MG		 2143 xs_set_memalloc(xprt);
2144
16be2d20
CL		 2145 write_unlock_bh(&sk->sk_callback_lock);
2146 }
2147 xs_udp_do_set_buffer_size(xprt);
02910177
TM		 2148
2149 xprt->stat.connect_start = jiffies;
16be2d20
CL		 2150}
2151
8c14ff2a 2152static void xs_udp_setup_socket(struct work_struct *work)
a246b010 2153{
34161db6
TM		 2154 struct sock_xprt *transport =
2155 container_of(work, struct sock_xprt, connect_worker.work);
c8475461 2156 struct rpc_xprt *xprt = &transport->xprt;
d099b8af 2157 struct socket *sock;
b65c0310 2158 int status = -EIO;
9903cd1c 2159
8c14ff2a 2160 sock = xs_create_sock(xprt, transport,
4dda9c8a
TM		 2161 xs_addr(xprt)->sa_family, SOCK_DGRAM,
2162 IPPROTO_UDP, false);
b65c0310 2163 if (IS_ERR(sock))
b0d93ad5 2164 goto out;
68e220bd 2165
c740eff8
CL		 2166 dprintk("RPC: worker connecting xprt %p via %s to "
2167 "%s (port %s)\n", xprt,
2168 xprt->address_strings[RPC_DISPLAY_PROTO],
2169 xprt->address_strings[RPC_DISPLAY_ADDR],
2170 xprt->address_strings[RPC_DISPLAY_PORT]);
68e220bd
CL		 2171
2172 xs_udp_finish_connecting(xprt, sock);
40b5ea0c 2173 trace_rpc_socket_connect(xprt, sock, 0);
b0d93ad5
CL		 2174 status = 0;
2175out:
b0d93ad5 2176 xprt_clear_connecting(xprt);
cf76785d 2177 xprt_unlock_connect(xprt, transport);
7d1e8255 2178 xprt_wake_pending_tasks(xprt, status);
a246b010
CL		 2179}
2180
4876cc77
TM		 2181/**
2182 * xs_tcp_shutdown - gracefully shut down a TCP socket
2183 * @xprt: transport
2184 *
2185 * Initiates a graceful shutdown of the TCP socket by calling the
2186 * equivalent of shutdown(SHUT_RDWR);
2187 */
2188static void xs_tcp_shutdown(struct rpc_xprt *xprt)
2189{
2190 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2191 struct socket *sock = transport->sock;
9b30889c 2192 int skst = transport->inet ? transport->inet->sk_state : TCP_CLOSE;
4876cc77
TM		 2193
2194 if (sock == NULL)
2195 return;
9b30889c
TM		 2196 switch (skst) {
2197 default:
4876cc77
TM		 2198 kernel_sock_shutdown(sock, SHUT_RDWR);
2199 trace_rpc_socket_shutdown(xprt, sock);
9b30889c
TM		 2200 break;
2201 case TCP_CLOSE:
2202 case TCP_TIME_WAIT:
4876cc77 2203 xs_reset_transport(transport);
9b30889c 2204 }
4876cc77
TM		 2205}
2206
8d1b8c62
TM		 2207static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
2208 struct socket *sock)
2209{
7196dbb0
TM		 2210 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2211 unsigned int keepidle;
2212 unsigned int keepcnt;
8d1b8c62
TM		 2213 unsigned int opt_on = 1;
2214 unsigned int timeo;
2215
b5e92419 2216 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2217 keepidle = DIV_ROUND_UP(xprt->timeout->to_initval, HZ);
2218 keepcnt = xprt->timeout->to_retries + 1;
2219 timeo = jiffies_to_msecs(xprt->timeout->to_initval) *
2220 (xprt->timeout->to_retries + 1);
2221 clear_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2222 spin_unlock(&xprt->transport_lock);
7196dbb0 2223
8d1b8c62
TM		 2224 /* TCP Keepalive options */
2225 kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
2226 (char *)&opt_on, sizeof(opt_on));
2227 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
2228 (char *)&keepidle, sizeof(keepidle));
2229 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
2230 (char *)&keepidle, sizeof(keepidle));
2231 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
2232 (char *)&keepcnt, sizeof(keepcnt));
2233
2234 /* TCP user timeout (see RFC5482) */
8d1b8c62
TM		 2235 kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
2236 (char *)&timeo, sizeof(timeo));
2237}
2238
7196dbb0
TM		 2239static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
2240 unsigned long connect_timeout,
2241 unsigned long reconnect_timeout)
2242{
2243 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2244 struct rpc_timeout to;
2245 unsigned long initval;
2246
b5e92419 2247 spin_lock(&xprt->transport_lock);
7196dbb0
TM		 2248 if (reconnect_timeout < xprt->max_reconnect_timeout)
2249 xprt->max_reconnect_timeout = reconnect_timeout;
2250 if (connect_timeout < xprt->connect_timeout) {
2251 memcpy(&to, xprt->timeout, sizeof(to));
2252 initval = DIV_ROUND_UP(connect_timeout, to.to_retries + 1);
2253 /* Arbitrary lower limit */
2254 if (initval < XS_TCP_INIT_REEST_TO << 1)
2255 initval = XS_TCP_INIT_REEST_TO << 1;
2256 to.to_initval = initval;
2257 to.to_maxval = initval;
2258 memcpy(&transport->tcp_timeout, &to,
2259 sizeof(transport->tcp_timeout));
2260 xprt->timeout = &transport->tcp_timeout;
2261 xprt->connect_timeout = connect_timeout;
2262 }
2263 set_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state);
b5e92419 2264 spin_unlock(&xprt->transport_lock);
7196dbb0
TM		 2265}
2266
16be2d20 2267static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
a246b010 2268{
16be2d20 2269 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
fe19a96b 2270 int ret = -ENOTCONN;
edb267a6 2271
ee0ac0c2 2272 if (!transport->inet) {
b0d93ad5 2273 struct sock *sk = sock->sk;
d88e4d82 2274 unsigned int addr_pref = IPV6_PREFER_SRC_PUBLIC;
7f260e85 2275
d88e4d82
N		 2276 /* Avoid temporary address, they are bad for long-lived
2277 * connections such as NFS mounts.
2278 * RFC4941, section 3.6 suggests that:
2279 * Individual applications, which have specific
2280 * knowledge about the normal duration of connections,
2281 * MAY override this as appropriate.
2282 */
2283 kernel_setsockopt(sock, SOL_IPV6, IPV6_ADDR_PREFERENCES,
2284 (char *)&addr_pref, sizeof(addr_pref));
2285
8d1b8c62 2286 xs_tcp_set_socket_timeouts(xprt, sock);
775f06ab 2287
b0d93ad5
CL		 2288 write_lock_bh(&sk->sk_callback_lock);
2289
2a9e1cfa
TM		 2290 xs_save_old_callbacks(transport, sk);
2291
b0d93ad5 2292 sk->sk_user_data = xprt;
5157b956 2293 sk->sk_data_ready = xs_data_ready;
b0d93ad5
CL		 2294 sk->sk_state_change = xs_tcp_state_change;
2295 sk->sk_write_space = xs_tcp_write_space;
b4411457 2296 sock_set_flag(sk, SOCK_FASYNC);
2118071d 2297 sk->sk_error_report = xs_error_report;
3167e12c
CL		 2298
2299 /* socket options */
3167e12c 2300 sock_reset_flag(sk, SOCK_LINGER);
3167e12c 2301 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
b0d93ad5
CL		 2302
2303 xprt_clear_connected(xprt);
2304
2305 /* Reset to new socket */
ee0ac0c2
CL		 2306 transport->sock = sock;
2307 transport->inet = sk;
b0d93ad5
CL		 2308
2309 write_unlock_bh(&sk->sk_callback_lock);
2310 }
2311
01d37c42 2312 if (!xprt_bound(xprt))
fe19a96b 2313 goto out;
01d37c42 2314
a564b8f0
MG		 2315 xs_set_memalloc(xprt);
2316
ae053551 2317 xs_stream_start_connect(transport);
e1806c7b 2318
b0d93ad5 2319 /* Tell the socket layer to start connecting... */
0fdea1e8 2320 set_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
fe19a96b
TM		 2321 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2322 switch (ret) {
2323 case 0:
4dda9c8a 2324 xs_set_srcport(transport, sock);
e9d47639 2325 /* fall through */
fe19a96b
TM		 2326 case -EINPROGRESS:
2327 /* SYN_SENT! */
fe19a96b
TM		 2328 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2329 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
1f4c17a0
TM		 2330 break;
2331 case -EADDRNOTAVAIL:
2332 /* Source port number is unavailable. Try a new one! */
2333 transport->srcport = 0;
fe19a96b
TM		 2334 }
2335out:
2336 return ret;
16be2d20
CL		 2337}
2338
9903cd1c 2339/**
b61d59ff 2340 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
acf0a39f 2341 * @work: queued work item
9903cd1c
CL		 2342 *
2343 * Invoked by a work queue tasklet.
a246b010 2344 */
cdd518d5 2345static void xs_tcp_setup_socket(struct work_struct *work)
a246b010 2346{
cdd518d5
PE		 2347 struct sock_xprt *transport =
2348 container_of(work, struct sock_xprt, connect_worker.work);
ee0ac0c2 2349 struct socket *sock = transport->sock;
a9f5f0f7 2350 struct rpc_xprt *xprt = &transport->xprt;
b61d59ff 2351 int status = -EIO;
a246b010 2352
ee0ac0c2 2353 if (!sock) {
cdd518d5 2354 sock = xs_create_sock(xprt, transport,
4dda9c8a
TM		 2355 xs_addr(xprt)->sa_family, SOCK_STREAM,
2356 IPPROTO_TCP, true);
b61d59ff
TM		 2357 if (IS_ERR(sock)) {
2358 status = PTR_ERR(sock);
3167e12c
CL		 2359 goto out;
2360 }
7d1e8255 2361 }
a246b010 2362
c740eff8
CL		 2363 dprintk("RPC: worker connecting xprt %p via %s to "
2364 "%s (port %s)\n", xprt,
2365 xprt->address_strings[RPC_DISPLAY_PROTO],
2366 xprt->address_strings[RPC_DISPLAY_ADDR],
2367 xprt->address_strings[RPC_DISPLAY_PORT]);
edb267a6 2368
16be2d20 2369 status = xs_tcp_finish_connecting(xprt, sock);
40b5ea0c 2370 trace_rpc_socket_connect(xprt, sock, status);
46121cf7
CL		 2371 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2372 xprt, -status, xprt_connected(xprt),
2373 sock->sk->sk_state);
2a491991 2374 switch (status) {
f75e6745
TM		 2375 default:
2376 printk("%s: connect returned unhandled error %d\n",
2377 __func__, status);
e9d47639 2378 /* fall through */
f75e6745
TM		 2379 case -EADDRNOTAVAIL:
2380 /* We're probably in TIME_WAIT. Get rid of existing socket,
2381 * and retry
2382 */
a519fc7a 2383 xs_tcp_force_close(xprt);
88b5ed73 2384 break;
2a491991
TM		 2385 case 0:
2386 case -EINPROGRESS:
2387 case -EALREADY:
718ba5b8 2388 xprt_unlock_connect(xprt, transport);
7d1e8255 2389 return;
9fcfe0c8
TM		 2390 case -EINVAL:
2391 /* Happens, for instance, if the user specified a link
2392 * local IPv6 address without a scope-id.
2393 */
3ed5e2a2
TM		 2394 case -ECONNREFUSED:
2395 case -ECONNRESET:
eb5b46fa 2396 case -ENETDOWN:
3ed5e2a2 2397 case -ENETUNREACH:
4ba161a7 2398 case -EHOSTUNREACH:
3913c78c 2399 case -EADDRINUSE:
3601c4a9 2400 case -ENOBUFS:
6ea44adc
N		 2401 /*
2402 * xs_tcp_force_close() wakes tasks with -EIO.
2403 * We need to wake them first to ensure the
2404 * correct error code.
2405 */
2406 xprt_wake_pending_tasks(xprt, status);
4efdd92c 2407 xs_tcp_force_close(xprt);
9fcfe0c8 2408 goto out;
a246b010 2409 }
2a491991 2410 status = -EAGAIN;
a246b010 2411out:
68e220bd 2412 xprt_clear_connecting(xprt);
cf76785d 2413 xprt_unlock_connect(xprt, transport);
7d1e8255 2414 xprt_wake_pending_tasks(xprt, status);
68e220bd 2415}
b0d93ad5 2416
9903cd1c
CL		 2417/**
2418 * xs_connect - connect a socket to a remote endpoint
1b092092 2419 * @xprt: pointer to transport structure
9903cd1c
CL		 2420 * @task: address of RPC task that manages state of connect request
2421 *
2422 * TCP: If the remote end dropped the connection, delay reconnecting.
03bf4b70
CL		 2423 *
2424 * UDP socket connects are synchronous, but we use a work queue anyway
2425 * to guarantee that even unprivileged user processes can set up a
2426 * socket on a privileged port.
2427 *
2428 * If a UDP socket connect fails, the delay behavior here prevents
2429 * retry floods (hard mounts).
9903cd1c 2430 */
1b092092 2431static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
a246b010 2432{
ee0ac0c2 2433 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
02910177 2434 unsigned long delay = 0;
a246b010 2435
718ba5b8
TM		 2436 WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport));
2437
99b1a4c3 2438 if (transport->sock != NULL) {
46121cf7
CL		 2439 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2440 "seconds\n",
03bf4b70 2441 xprt, xprt->reestablish_timeout / HZ);
99b1a4c3
TM		 2442
2443 /* Start by resetting any existing state */
2444 xs_reset_transport(transport);
2445
675dd90a
CL		 2446 delay = xprt_reconnect_delay(xprt);
2447 xprt_reconnect_backoff(xprt, XS_TCP_INIT_REEST_TO);
02910177 2448
02910177 2449 } else
46121cf7 2450 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
02910177
TM		 2451
2452 queue_delayed_work(xprtiod_workqueue,
2453 &transport->connect_worker,
2454 delay);
a246b010
CL		 2455}
2456
4f8943f8
TM		 2457static void xs_wake_disconnect(struct sock_xprt *transport)
2458{
2459 if (test_and_clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state))
2460 xs_tcp_force_close(&transport->xprt);
2461}
2462
2463static void xs_wake_write(struct sock_xprt *transport)
2464{
2465 if (test_and_clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state))
2466 xprt_write_space(&transport->xprt);
2467}
2468
2469static void xs_wake_error(struct sock_xprt *transport)
2470{
2471 int sockerr;
2472 int sockerr_len = sizeof(sockerr);
2473
2474 if (!test_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2475 return;
2476 mutex_lock(&transport->recv_mutex);
2477 if (transport->sock == NULL)
2478 goto out;
2479 if (!test_and_clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
2480 goto out;
2481 if (kernel_getsockopt(transport->sock, SOL_SOCKET, SO_ERROR,
2482 (char *)&sockerr, &sockerr_len) != 0)
2483 goto out;
2484 if (sockerr < 0)
2485 xprt_wake_pending_tasks(&transport->xprt, sockerr);
2486out:
2487 mutex_unlock(&transport->recv_mutex);
2488}
2489
2490static void xs_wake_pending(struct sock_xprt *transport)
2491{
2492 if (test_and_clear_bit(XPRT_SOCK_WAKE_PENDING, &transport->sock_state))
2493 xprt_wake_pending_tasks(&transport->xprt, -EAGAIN);
2494}
2495
2496static void xs_error_handle(struct work_struct *work)
2497{
2498 struct sock_xprt *transport = container_of(work,
2499 struct sock_xprt, error_worker);
2500
2501 xs_wake_disconnect(transport);
2502 xs_wake_write(transport);
2503 xs_wake_error(transport);
2504 xs_wake_pending(transport);
2505}
2506
176e21ee
CL		 2507/**
2508 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2509 * @xprt: rpc_xprt struct containing statistics
2510 * @seq: output file
2511 *
2512 */
2513static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2514{
2515 long idle_time = 0;
2516
2517 if (xprt_connected(xprt))
2518 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2519
2520 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
15a45206 2521 "%llu %llu %lu %llu %llu\n",
176e21ee
CL		 2522 xprt->stat.bind_count,
2523 xprt->stat.connect_count,
8440a886 2524 xprt->stat.connect_time / HZ,
176e21ee
CL		 2525 idle_time,
2526 xprt->stat.sends,
2527 xprt->stat.recvs,
2528 xprt->stat.bad_xids,
2529 xprt->stat.req_u,
15a45206
AA		 2530 xprt->stat.bklog_u,
2531 xprt->stat.max_slots,
2532 xprt->stat.sending_u,
2533 xprt->stat.pending_u);
176e21ee
CL		 2534}
2535
262ca07d
CL		 2536/**
2537 * xs_udp_print_stats - display UDP socket-specifc stats
2538 * @xprt: rpc_xprt struct containing statistics
2539 * @seq: output file
2540 *
2541 */
2542static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2543{
c8475461
CL		 2544 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2545
15a45206
AA		 2546 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2547 "%lu %llu %llu\n",
fbfffbd5 2548 transport->srcport,
262ca07d
CL		 2549 xprt->stat.bind_count,
2550 xprt->stat.sends,
2551 xprt->stat.recvs,
2552 xprt->stat.bad_xids,
2553 xprt->stat.req_u,
15a45206
AA		 2554 xprt->stat.bklog_u,
2555 xprt->stat.max_slots,
2556 xprt->stat.sending_u,
2557 xprt->stat.pending_u);
262ca07d
CL		 2558}
2559
2560/**
2561 * xs_tcp_print_stats - display TCP socket-specifc stats
2562 * @xprt: rpc_xprt struct containing statistics
2563 * @seq: output file
2564 *
2565 */
2566static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2567{
c8475461 2568 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
262ca07d
CL		 2569 long idle_time = 0;
2570
2571 if (xprt_connected(xprt))
2572 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2573
15a45206
AA		 2574 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2575 "%llu %llu %lu %llu %llu\n",
fbfffbd5 2576 transport->srcport,
262ca07d
CL		 2577 xprt->stat.bind_count,
2578 xprt->stat.connect_count,
8440a886 2579 xprt->stat.connect_time / HZ,
262ca07d
CL		 2580 idle_time,
2581 xprt->stat.sends,
2582 xprt->stat.recvs,
2583 xprt->stat.bad_xids,
2584 xprt->stat.req_u,
15a45206
AA		 2585 xprt->stat.bklog_u,
2586 xprt->stat.max_slots,
2587 xprt->stat.sending_u,
2588 xprt->stat.pending_u);
262ca07d
CL		 2589}
2590
4cfc7e60
RI		 2591/*
2592 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2593 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2594 * to use the server side send routines.
2595 */
5fe6eaa1 2596static int bc_malloc(struct rpc_task *task)
4cfc7e60 2597{
5fe6eaa1
CL		 2598 struct rpc_rqst *rqst = task->tk_rqstp;
2599 size_t size = rqst->rq_callsize;
4cfc7e60
RI		 2600 struct page *page;
2601 struct rpc_buffer *buf;
2602
5fe6eaa1
CL		 2603 if (size > PAGE_SIZE - sizeof(struct rpc_buffer)) {
2604 WARN_ONCE(1, "xprtsock: large bc buffer request (size %zu)\n",
2605 size);
2606 return -EINVAL;
2607 }
4cfc7e60 2608
b8a13d03 2609 page = alloc_page(GFP_KERNEL);
4cfc7e60 2610 if (!page)
5fe6eaa1 2611 return -ENOMEM;
4cfc7e60
RI		 2612
2613 buf = page_address(page);
2614 buf->len = PAGE_SIZE;
2615
5fe6eaa1 2616 rqst->rq_buffer = buf->data;
18e601d6 2617 rqst->rq_rbuffer = (char *)rqst->rq_buffer + rqst->rq_callsize;
5fe6eaa1 2618 return 0;
4cfc7e60
RI		 2619}
2620
2621/*
2622 * Free the space allocated in the bc_alloc routine
2623 */
3435c74a 2624static void bc_free(struct rpc_task *task)
4cfc7e60 2625{
3435c74a 2626 void *buffer = task->tk_rqstp->rq_buffer;
4cfc7e60
RI		 2627 struct rpc_buffer *buf;
2628
4cfc7e60
RI		 2629 buf = container_of(buffer, struct rpc_buffer, data);
2630 free_page((unsigned long)buf);
2631}
2632
2633/*
2634 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2635 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2636 */
2637static int bc_sendto(struct rpc_rqst *req)
2638{
2639 int len;
2640 struct xdr_buf *xbufp = &req->rq_snd_buf;
4cfc7e60 2641 struct sock_xprt *transport =
067fb11b 2642 container_of(req->rq_xprt, struct sock_xprt, xprt);
4cfc7e60
RI		 2643 unsigned long headoff;
2644 unsigned long tailoff;
067fb11b
CL		 2645 struct page *tailpage;
2646 struct msghdr msg = {
2647 .msg_flags = MSG_MORE
2648 };
2649 rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT |
2650 (u32)xbufp->len);
2651 struct kvec iov = {
2652 .iov_base = &marker,
2653 .iov_len = sizeof(marker),
2654 };
4cfc7e60 2655
067fb11b
CL		 2656 len = kernel_sendmsg(transport->sock, &msg, &iov, 1, iov.iov_len);
2657 if (len != iov.iov_len)
2658 return -EAGAIN;
4cfc7e60 2659
067fb11b
CL		 2660 tailpage = NULL;
2661 if (xbufp->tail[0].iov_len)
2662 tailpage = virt_to_page(xbufp->tail[0].iov_base);
4cfc7e60
RI		 2663 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2664 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
067fb11b 2665 len = svc_send_common(transport->sock, xbufp,
4cfc7e60 2666 virt_to_page(xbufp->head[0].iov_base), headoff,
067fb11b
CL		 2667 tailpage, tailoff);
2668 if (len != xbufp->len)
2669 return -EAGAIN;
4cfc7e60
RI		 2670 return len;
2671}
2672
2673/*
2674 * The send routine. Borrows from svc_send
2675 */
adfa7144 2676static int bc_send_request(struct rpc_rqst *req)
4cfc7e60 2677{
4cfc7e60 2678 struct svc_xprt *xprt;
7fc56136 2679 int len;
4cfc7e60
RI		 2680
2681 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2682 /*
2683 * Get the server socket associated with this callback xprt
2684 */
2685 xprt = req->rq_xprt->bc_xprt;
4cfc7e60
RI		 2686
2687 /*
2688 * Grab the mutex to serialize data as the connection is shared
2689 * with the fore channel
2690 */
c544577d 2691 mutex_lock(&xprt->xpt_mutex);
4cfc7e60
RI		 2692 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2693 len = -ENOTCONN;
2694 else
2695 len = bc_sendto(req);
2696 mutex_unlock(&xprt->xpt_mutex);
2697
2698 if (len > 0)
2699 len = 0;
2700
2701 return len;
2702}
2703
2704/*
2705 * The close routine. Since this is client initiated, we do nothing
2706 */
2707
2708static void bc_close(struct rpc_xprt *xprt)
2709{
4cfc7e60
RI		 2710}
2711
2712/*
2713 * The xprt destroy routine. Again, because this connection is client
2714 * initiated, we do nothing
2715 */
2716
2717static void bc_destroy(struct rpc_xprt *xprt)
2718{
47f72efa
KM		 2719 dprintk("RPC: bc_destroy xprt %p\n", xprt);
2720
2721 xs_xprt_free(xprt);
2722 module_put(THIS_MODULE);
4cfc7e60
RI		 2723}
2724
d31ae254 2725static const struct rpc_xprt_ops xs_local_ops = {
176e21ee 2726 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2727 .release_xprt = xprt_release_xprt,
f39c1bfb 2728 .alloc_slot = xprt_alloc_slot,
a9cde23a 2729 .free_slot = xprt_free_slot,
176e21ee
CL		 2730 .rpcbind = xs_local_rpcbind,
2731 .set_port = xs_local_set_port,
dc107402 2732 .connect = xs_local_connect,
176e21ee
CL		 2733 .buf_alloc = rpc_malloc,
2734 .buf_free = rpc_free,
550aebfe 2735 .prepare_request = xs_stream_prepare_request,
176e21ee 2736 .send_request = xs_local_send_request,
8ba6a92d 2737 .wait_for_reply_request = xprt_wait_for_reply_request_def,
176e21ee 2738 .close = xs_close,
a1311d87 2739 .destroy = xs_destroy,
176e21ee 2740 .print_stats = xs_local_print_stats,
d67fa4d8
JL		 2741 .enable_swap = xs_enable_swap,
2742 .disable_swap = xs_disable_swap,
176e21ee
CL		 2743};
2744
d31ae254 2745static const struct rpc_xprt_ops xs_udp_ops = {
43118c29 2746 .set_buffer_size = xs_udp_set_buffer_size,
12a80469 2747 .reserve_xprt = xprt_reserve_xprt_cong,
49e9a890 2748 .release_xprt = xprt_release_xprt_cong,
f39c1bfb 2749 .alloc_slot = xprt_alloc_slot,
a9cde23a 2750 .free_slot = xprt_free_slot,
45160d62 2751 .rpcbind = rpcb_getport_async,
92200412 2752 .set_port = xs_set_port,
262965f5 2753 .connect = xs_connect,
02107148
CL		 2754 .buf_alloc = rpc_malloc,
2755 .buf_free = rpc_free,
262965f5 2756 .send_request = xs_udp_send_request,
8ba6a92d 2757 .wait_for_reply_request = xprt_wait_for_reply_request_rtt,
46c0ee8b 2758 .timer = xs_udp_timer,
a58dd398 2759 .release_request = xprt_release_rqst_cong,
262965f5
CL		 2760 .close = xs_close,
2761 .destroy = xs_destroy,
262ca07d 2762 .print_stats = xs_udp_print_stats,
d67fa4d8
JL		 2763 .enable_swap = xs_enable_swap,
2764 .disable_swap = xs_disable_swap,
4a068258 2765 .inject_disconnect = xs_inject_disconnect,
262965f5
CL		 2766};
2767
d31ae254 2768static const struct rpc_xprt_ops xs_tcp_ops = {
12a80469 2769 .reserve_xprt = xprt_reserve_xprt,
4cd34e7c 2770 .release_xprt = xprt_release_xprt,
36bd7de9 2771 .alloc_slot = xprt_alloc_slot,
a9cde23a 2772 .free_slot = xprt_free_slot,
45160d62 2773 .rpcbind = rpcb_getport_async,
92200412 2774 .set_port = xs_set_port,
0b9e7943 2775 .connect = xs_connect,
02107148
CL		 2776 .buf_alloc = rpc_malloc,
2777 .buf_free = rpc_free,
277e4ab7 2778 .prepare_request = xs_stream_prepare_request,
262965f5 2779 .send_request = xs_tcp_send_request,
8ba6a92d 2780 .wait_for_reply_request = xprt_wait_for_reply_request_def,
c627d31b 2781 .close = xs_tcp_shutdown,
9903cd1c 2782 .destroy = xs_destroy,
7196dbb0 2783 .set_connect_timeout = xs_tcp_set_connect_timeout,
262ca07d 2784 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2785 .enable_swap = xs_enable_swap,
2786 .disable_swap = xs_disable_swap,
4a068258 2787 .inject_disconnect = xs_inject_disconnect,
42e5c3e2
CL		 2788#ifdef CONFIG_SUNRPC_BACKCHANNEL
2789 .bc_setup = xprt_setup_bc,
6b26cc8c 2790 .bc_maxpayload = xs_tcp_bc_maxpayload,
42e5c3e2
CL		 2791 .bc_free_rqst = xprt_free_bc_rqst,
2792 .bc_destroy = xprt_destroy_bc,
2793#endif
a246b010
CL		 2794};
2795
4cfc7e60
RI		 2796/*
2797 * The rpc_xprt_ops for the server backchannel
2798 */
2799
d31ae254 2800static const struct rpc_xprt_ops bc_tcp_ops = {
4cfc7e60
RI		 2801 .reserve_xprt = xprt_reserve_xprt,
2802 .release_xprt = xprt_release_xprt,
84e28a30 2803 .alloc_slot = xprt_alloc_slot,
a9cde23a 2804 .free_slot = xprt_free_slot,
4cfc7e60
RI		 2805 .buf_alloc = bc_malloc,
2806 .buf_free = bc_free,
2807 .send_request = bc_send_request,
8ba6a92d 2808 .wait_for_reply_request = xprt_wait_for_reply_request_def,
4cfc7e60
RI		 2809 .close = bc_close,
2810 .destroy = bc_destroy,
2811 .print_stats = xs_tcp_print_stats,
d67fa4d8
JL		 2812 .enable_swap = xs_enable_swap,
2813 .disable_swap = xs_disable_swap,
4a068258 2814 .inject_disconnect = xs_inject_disconnect,
4cfc7e60
RI		 2815};
2816
92476850
CL		 2817static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2818{
2819 static const struct sockaddr_in sin = {
2820 .sin_family = AF_INET,
2821 .sin_addr.s_addr = htonl(INADDR_ANY),
2822 };
2823 static const struct sockaddr_in6 sin6 = {
2824 .sin6_family = AF_INET6,
2825 .sin6_addr = IN6ADDR_ANY_INIT,
2826 };
2827
2828 switch (family) {
176e21ee
CL		 2829 case AF_LOCAL:
2830 break;
92476850
CL		 2831 case AF_INET:
2832 memcpy(sap, &sin, sizeof(sin));
2833 break;
2834 case AF_INET6:
2835 memcpy(sap, &sin6, sizeof(sin6));
2836 break;
2837 default:
2838 dprintk("RPC: %s: Bad address family\n", __func__);
2839 return -EAFNOSUPPORT;
2840 }
2841 return 0;
2842}
2843
3c341b0b 2844static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
d9ba131d
TM		 2845 unsigned int slot_table_size,
2846 unsigned int max_slot_table_size)
c8541ecd
CL		 2847{
2848 struct rpc_xprt *xprt;
ffc2e518 2849 struct sock_xprt *new;
c8541ecd 2850
96802a09 2851 if (args->addrlen > sizeof(xprt->addr)) {
46121cf7 2852 dprintk("RPC: xs_setup_xprt: address too large\n");
c8541ecd
CL		 2853 return ERR_PTR(-EBADF);
2854 }
2855
d9ba131d
TM		 2856 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2857 max_slot_table_size);
bd1722d4 2858 if (xprt == NULL) {
46121cf7
CL		 2859 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2860 "rpc_xprt\n");
c8541ecd
CL		 2861 return ERR_PTR(-ENOMEM);
2862 }
c8541ecd 2863
bd1722d4 2864 new = container_of(xprt, struct sock_xprt, xprt);
edc1b01c 2865 mutex_init(&new->recv_mutex);
96802a09
FM		 2866 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2867 xprt->addrlen = args->addrlen;
d3bc9a1d 2868 if (args->srcaddr)
fbfffbd5 2869 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
92476850
CL		 2870 else {
2871 int err;
2872 err = xs_init_anyaddr(args->dstaddr->sa_family,
2873 (struct sockaddr *)&new->srcaddr);
2aa13531
SK		 2874 if (err != 0) {
2875 xprt_free(xprt);
92476850 2876 return ERR_PTR(err);
2aa13531 2877 }
92476850 2878 }
c8541ecd
CL		 2879
2880 return xprt;
2881}
2882
176e21ee
CL		 2883static const struct rpc_timeout xs_local_default_timeout = {
2884 .to_initval = 10 * HZ,
2885 .to_maxval = 10 * HZ,
2886 .to_retries = 2,
2887};
2888
2889/**
2890 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2891 * @args: rpc transport creation arguments
2892 *
2893 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2894 */
2895static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2896{
2897 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2898 struct sock_xprt *transport;
2899 struct rpc_xprt *xprt;
2900 struct rpc_xprt *ret;
2901
d9ba131d
TM		 2902 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2903 xprt_max_tcp_slot_table_entries);
176e21ee
CL		 2904 if (IS_ERR(xprt))
2905 return xprt;
2906 transport = container_of(xprt, struct sock_xprt, xprt);
2907
2908 xprt->prot = 0;
176e21ee
CL		 2909 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2910
2911 xprt->bind_timeout = XS_BIND_TO;
2912 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2913 xprt->idle_timeout = XS_IDLE_DISC_TO;
2914
2915 xprt->ops = &xs_local_ops;
2916 xprt->timeout = &xs_local_default_timeout;
2917
550aebfe 2918 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 2919 INIT_WORK(&transport->error_worker, xs_error_handle);
550aebfe 2920 INIT_DELAYED_WORK(&transport->connect_worker, xs_dummy_setup_socket);
93dc41bd 2921
176e21ee
CL		 2922 switch (sun->sun_family) {
2923 case AF_LOCAL:
2924 if (sun->sun_path[0] != '/') {
2925 dprintk("RPC: bad AF_LOCAL address: %s\n",
2926 sun->sun_path);
2927 ret = ERR_PTR(-EINVAL);
2928 goto out_err;
2929 }
2930 xprt_set_bound(xprt);
176e21ee 2931 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
7073ea87
BF		 2932 ret = ERR_PTR(xs_local_setup_socket(transport));
2933 if (ret)
2934 goto out_err;
176e21ee
CL		 2935 break;
2936 default:
2937 ret = ERR_PTR(-EAFNOSUPPORT);
2938 goto out_err;
2939 }
2940
2941 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2942 xprt->address_strings[RPC_DISPLAY_ADDR]);
2943
2944 if (try_module_get(THIS_MODULE))
2945 return xprt;
2946 ret = ERR_PTR(-EINVAL);
2947out_err:
315f3812 2948 xs_xprt_free(xprt);
176e21ee
CL		 2949 return ret;
2950}
2951
2881ae74
TM		 2952static const struct rpc_timeout xs_udp_default_timeout = {
2953 .to_initval = 5 * HZ,
2954 .to_maxval = 30 * HZ,
2955 .to_increment = 5 * HZ,
2956 .to_retries = 5,
2957};
2958
9903cd1c
CL		 2959/**
2960 * xs_setup_udp - Set up transport to use a UDP socket
96802a09 2961 * @args: rpc transport creation arguments
9903cd1c
CL		 2962 *
2963 */
483066d6 2964static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
a246b010 2965{
8f9d5b1a 2966 struct sockaddr *addr = args->dstaddr;
c8541ecd 2967 struct rpc_xprt *xprt;
c8475461 2968 struct sock_xprt *transport;
0a68b0be 2969 struct rpc_xprt *ret;
a246b010 2970
d9ba131d
TM		 2971 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2972 xprt_udp_slot_table_entries);
c8541ecd
CL		 2973 if (IS_ERR(xprt))
2974 return xprt;
c8475461 2975 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 2976
ec739ef0 2977 xprt->prot = IPPROTO_UDP;
a246b010
CL		 2978 /* XXX: header size can vary due to auth type, IPv6, etc. */
2979 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2980
03bf4b70 2981 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 2982 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2983 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 2984
262965f5 2985 xprt->ops = &xs_udp_ops;
a246b010 2986
ba7392bb 2987 xprt->timeout = &xs_udp_default_timeout;
a246b010 2988
f9b2ee71 2989 INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn);
4f8943f8 2990 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 2991 INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket);
2992
8f9d5b1a
CL		 2993 switch (addr->sa_family) {
2994 case AF_INET:
2995 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2996 xprt_set_bound(xprt);
2997
9dc3b095 2998 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
8f9d5b1a
CL		 2999 break;
3000 case AF_INET6:
3001 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
3002 xprt_set_bound(xprt);
3003
9dc3b095 3004 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
8f9d5b1a
CL		 3005 break;
3006 default:
0a68b0be
BF		 3007 ret = ERR_PTR(-EAFNOSUPPORT);
3008 goto out_err;
8f9d5b1a
CL		 3009 }
3010
c740eff8
CL		 3011 if (xprt_bound(xprt))
3012 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3013 xprt->address_strings[RPC_DISPLAY_ADDR],
3014 xprt->address_strings[RPC_DISPLAY_PORT],
3015 xprt->address_strings[RPC_DISPLAY_PROTO]);
3016 else
3017 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
3018 xprt->address_strings[RPC_DISPLAY_ADDR],
3019 xprt->address_strings[RPC_DISPLAY_PROTO]);
edb267a6 3020
bc25571e
TT		 3021 if (try_module_get(THIS_MODULE))
3022 return xprt;
0a68b0be
BF		 3023 ret = ERR_PTR(-EINVAL);
3024out_err:
315f3812 3025 xs_xprt_free(xprt);
0a68b0be 3026 return ret;
a246b010
CL		 3027}
3028
2881ae74
TM		 3029static const struct rpc_timeout xs_tcp_default_timeout = {
3030 .to_initval = 60 * HZ,
3031 .to_maxval = 60 * HZ,
3032 .to_retries = 2,
3033};
3034
9903cd1c
CL		 3035/**
3036 * xs_setup_tcp - Set up transport to use a TCP socket
96802a09 3037 * @args: rpc transport creation arguments
9903cd1c
CL		 3038 *
3039 */
483066d6 3040static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
a246b010 3041{
8f9d5b1a 3042 struct sockaddr *addr = args->dstaddr;
c8541ecd 3043 struct rpc_xprt *xprt;
c8475461 3044 struct sock_xprt *transport;
0a68b0be 3045 struct rpc_xprt *ret;
b7993ceb
TM		 3046 unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
3047
3048 if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
3049 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
a246b010 3050
d9ba131d 3051 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
b7993ceb 3052 max_slot_table_size);
c8541ecd
CL		 3053 if (IS_ERR(xprt))
3054 return xprt;
c8475461 3055 transport = container_of(xprt, struct sock_xprt, xprt);
a246b010 3056
ec739ef0 3057 xprt->prot = IPPROTO_TCP;
808012fb 3058 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
a246b010 3059
03bf4b70 3060 xprt->bind_timeout = XS_BIND_TO;
03bf4b70
CL		 3061 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
3062 xprt->idle_timeout = XS_IDLE_DISC_TO;
a246b010 3063
262965f5 3064 xprt->ops = &xs_tcp_ops;
ba7392bb 3065 xprt->timeout = &xs_tcp_default_timeout;
a246b010 3066
3851f1cd 3067 xprt->max_reconnect_timeout = xprt->timeout->to_maxval;
7196dbb0
TM		 3068 xprt->connect_timeout = xprt->timeout->to_initval *
3069 (xprt->timeout->to_retries + 1);
3851f1cd 3070
c50b8ee0 3071 INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn);
4f8943f8 3072 INIT_WORK(&transport->error_worker, xs_error_handle);
edc1b01c
TM		 3073 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket);
3074
8f9d5b1a
CL		 3075 switch (addr->sa_family) {
3076 case AF_INET:
3077 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
3078 xprt_set_bound(xprt);
3079
9dc3b095 3080 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
8f9d5b1a
CL		 3081 break;
3082 case AF_INET6:
3083 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
3084 xprt_set_bound(xprt);
3085
9dc3b095 3086 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
8f9d5b1a
CL		 3087 break;
3088 default:
0a68b0be
BF		 3089 ret = ERR_PTR(-EAFNOSUPPORT);
3090 goto out_err;
8f9d5b1a
CL		 3091 }
3092
c740eff8
CL		 3093 if (xprt_bound(xprt))
3094 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3095 xprt->address_strings[RPC_DISPLAY_ADDR],
3096 xprt->address_strings[RPC_DISPLAY_PORT],
3097 xprt->address_strings[RPC_DISPLAY_PROTO]);
3098 else
3099 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
3100 xprt->address_strings[RPC_DISPLAY_ADDR],
3101 xprt->address_strings[RPC_DISPLAY_PROTO]);
3102
bc25571e
TT		 3103 if (try_module_get(THIS_MODULE))
3104 return xprt;
0a68b0be
BF		 3105 ret = ERR_PTR(-EINVAL);
3106out_err:
315f3812 3107 xs_xprt_free(xprt);
0a68b0be 3108 return ret;
a246b010 3109}
282b32e1 3110
f300baba
AB		 3111/**
3112 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
3113 * @args: rpc transport creation arguments
3114 *
3115 */
3116static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
3117{
3118 struct sockaddr *addr = args->dstaddr;
3119 struct rpc_xprt *xprt;
3120 struct sock_xprt *transport;
3121 struct svc_sock *bc_sock;
0a68b0be 3122 struct rpc_xprt *ret;
f300baba 3123
d9ba131d
TM		 3124 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
3125 xprt_tcp_slot_table_entries);
f300baba
AB		 3126 if (IS_ERR(xprt))
3127 return xprt;
3128 transport = container_of(xprt, struct sock_xprt, xprt);
3129
3130 xprt->prot = IPPROTO_TCP;
f300baba
AB		 3131 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
3132 xprt->timeout = &xs_tcp_default_timeout;
3133
3134 /* backchannel */
3135 xprt_set_bound(xprt);
3136 xprt->bind_timeout = 0;
f300baba
AB		 3137 xprt->reestablish_timeout = 0;
3138 xprt->idle_timeout = 0;
3139
f300baba
AB		 3140 xprt->ops = &bc_tcp_ops;
3141
3142 switch (addr->sa_family) {
3143 case AF_INET:
3144 xs_format_peer_addresses(xprt, "tcp",
3145 RPCBIND_NETID_TCP);
3146 break;
3147 case AF_INET6:
3148 xs_format_peer_addresses(xprt, "tcp",
3149 RPCBIND_NETID_TCP6);
3150 break;
3151 default:
0a68b0be
BF		 3152 ret = ERR_PTR(-EAFNOSUPPORT);
3153 goto out_err;
f300baba
AB		 3154 }
3155
50fa0d40
PE		 3156 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3157 xprt->address_strings[RPC_DISPLAY_ADDR],
3158 xprt->address_strings[RPC_DISPLAY_PORT],
3159 xprt->address_strings[RPC_DISPLAY_PROTO]);
f300baba 3160
99de8ea9
BF		 3161 /*
3162 * Once we've associated a backchannel xprt with a connection,
28303ca3
WM		 3163 * we want to keep it around as long as the connection lasts,
3164 * in case we need to start using it for a backchannel again;
3165 * this reference won't be dropped until bc_xprt is destroyed.
99de8ea9
BF		 3166 */
3167 xprt_get(xprt);
3168 args->bc_xprt->xpt_bc_xprt = xprt;
3169 xprt->bc_xprt = args->bc_xprt;
3170 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
3171 transport->sock = bc_sock->sk_sock;
3172 transport->inet = bc_sock->sk_sk;
3173
f300baba
AB		 3174 /*
3175 * Since we don't want connections for the backchannel, we set
3176 * the xprt status to connected
3177 */
3178 xprt_set_connected(xprt);
3179
f300baba
AB		 3180 if (try_module_get(THIS_MODULE))
3181 return xprt;
642aab58
KM		 3182
3183 args->bc_xprt->xpt_bc_xprt = NULL;
39a9beab 3184 args->bc_xprt->xpt_bc_xps = NULL;
99de8ea9 3185 xprt_put(xprt);
0a68b0be
BF		 3186 ret = ERR_PTR(-EINVAL);
3187out_err:
315f3812 3188 xs_xprt_free(xprt);
0a68b0be 3189 return ret;
f300baba
AB		 3190}
3191
176e21ee
CL		 3192static struct xprt_class xs_local_transport = {
3193 .list = LIST_HEAD_INIT(xs_local_transport.list),
3194 .name = "named UNIX socket",
3195 .owner = THIS_MODULE,
3196 .ident = XPRT_TRANSPORT_LOCAL,
3197 .setup = xs_setup_local,
3198};
3199
bc25571e
TT		 3200static struct xprt_class xs_udp_transport = {
3201 .list = LIST_HEAD_INIT(xs_udp_transport.list),
3202 .name = "udp",
3203 .owner = THIS_MODULE,
f300baba 3204 .ident = XPRT_TRANSPORT_UDP,
bc25571e
TT		 3205 .setup = xs_setup_udp,
3206};
3207
3208static struct xprt_class xs_tcp_transport = {
3209 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
3210 .name = "tcp",
3211 .owner = THIS_MODULE,
f300baba 3212 .ident = XPRT_TRANSPORT_TCP,
bc25571e
TT		 3213 .setup = xs_setup_tcp,
3214};
3215
f300baba
AB		 3216static struct xprt_class xs_bc_tcp_transport = {
3217 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
3218 .name = "tcp NFSv4.1 backchannel",
3219 .owner = THIS_MODULE,
3220 .ident = XPRT_TRANSPORT_BC_TCP,
3221 .setup = xs_setup_bc_tcp,
3222};
3223
282b32e1 3224/**
bc25571e 3225 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
282b32e1
CL		 3226 *
3227 */
3228int init_socket_xprt(void)
3229{
2b1bec5f 3230 if (!sunrpc_table_header)
0b4d4147 3231 sunrpc_table_header = register_sysctl_table(sunrpc_table);
fbf76683 3232
176e21ee 3233 xprt_register_transport(&xs_local_transport);
bc25571e
TT		 3234 xprt_register_transport(&xs_udp_transport);
3235 xprt_register_transport(&xs_tcp_transport);
f300baba 3236 xprt_register_transport(&xs_bc_tcp_transport);
bc25571e 3237
282b32e1
CL		 3238 return 0;
3239}
3240
3241/**
bc25571e 3242 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
282b32e1
CL		 3243 *
3244 */
3245void cleanup_socket_xprt(void)
3246{
fbf76683
CL		 3247 if (sunrpc_table_header) {
3248 unregister_sysctl_table(sunrpc_table_header);
3249 sunrpc_table_header = NULL;
3250 }
bc25571e 3251
176e21ee 3252 xprt_unregister_transport(&xs_local_transport);
bc25571e
TT		 3253 xprt_unregister_transport(&xs_udp_transport);
3254 xprt_unregister_transport(&xs_tcp_transport);
f300baba 3255 xprt_unregister_transport(&xs_bc_tcp_transport);
282b32e1 3256}
cbf11071 3257
9bbb9e5a
RR		 3258static int param_set_uint_minmax(const char *val,
3259 const struct kernel_param *kp,
cbf11071
TM		 3260 unsigned int min, unsigned int max)
3261{
00cfaa94 3262 unsigned int num;
cbf11071
TM		 3263 int ret;
3264
3265 if (!val)
3266 return -EINVAL;
00cfaa94 3267 ret = kstrtouint(val, 0, &num);
9761a246
DC		 3268 if (ret)
3269 return ret;
3270 if (num < min || num > max)
cbf11071
TM		 3271 return -EINVAL;
3272 *((unsigned int *)kp->arg) = num;
3273 return 0;
3274}
3275
9bbb9e5a 3276static int param_set_portnr(const char *val, const struct kernel_param *kp)
cbf11071 3277{
ffb6ca33 3278 return param_set_uint_minmax(val, kp,
826799e6 3279 RPC_MIN_RESVPORT,
cbf11071
TM		 3280 RPC_MAX_RESVPORT);
3281}
3282
9c27847d 3283static const struct kernel_param_ops param_ops_portnr = {
9bbb9e5a
RR		 3284 .set = param_set_portnr,
3285 .get = param_get_uint,
3286};
3287
cbf11071
TM		 3288#define param_check_portnr(name, p) \
3289 __param_check(name, p, unsigned int);
3290
3291module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3292module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3293
9bbb9e5a
RR		 3294static int param_set_slot_table_size(const char *val,
3295 const struct kernel_param *kp)
cbf11071
TM		 3296{
3297 return param_set_uint_minmax(val, kp,
3298 RPC_MIN_SLOT_TABLE,
3299 RPC_MAX_SLOT_TABLE);
3300}
3301
9c27847d 3302static const struct kernel_param_ops param_ops_slot_table_size = {
9bbb9e5a
RR		 3303 .set = param_set_slot_table_size,
3304 .get = param_get_uint,
3305};
3306
cbf11071
TM		 3307#define param_check_slot_table_size(name, p) \
3308 __param_check(name, p, unsigned int);
3309
d9ba131d
TM		 3310static int param_set_max_slot_table_size(const char *val,
3311 const struct kernel_param *kp)
3312{
3313 return param_set_uint_minmax(val, kp,
3314 RPC_MIN_SLOT_TABLE,
3315 RPC_MAX_SLOT_TABLE_LIMIT);
3316}
3317
9c27847d 3318static const struct kernel_param_ops param_ops_max_slot_table_size = {
d9ba131d
TM		 3319 .set = param_set_max_slot_table_size,
3320 .get = param_get_uint,
3321};
3322
3323#define param_check_max_slot_table_size(name, p) \
3324 __param_check(name, p, unsigned int);
3325
cbf11071
TM		 3326module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3327 slot_table_size, 0644);
d9ba131d
TM		 3328module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3329 max_slot_table_size, 0644);
cbf11071
TM		 3330module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3331 slot_table_size, 0644);
Empty description
This page took 1.688611 seconds and 4 git commands to generate.