1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Shared Memory Communications over RDMA (SMC-R) and RoCE
4 *
5 * Manage RMBE
6 * copy new RMBE data into user space
7 *
8 * Copyright IBM Corp. 2016
9 *
10 * Author(s): Ursula Braun <ubraun@linux.vnet.ibm.com>
11 */
12
13 #include <linux/net.h>
14 #include <linux/rcupdate.h>
15 #include <linux/sched/signal.h>
16
17 #include <net/sock.h>
18
19 #include "smc.h"
20 #include "smc_core.h"
21 #include "smc_cdc.h"
22 #include "smc_tx.h" /* smc_tx_consumer_update() */
23 #include "smc_rx.h"
24 #include "smc_stats.h"
25 #include "smc_tracepoint.h"
26
27 /* callback implementation to wakeup consumers blocked with smc_rx_wait().
28 * indirectly called by smc_cdc_msg_recv_action().
29 */
smc_rx_wake_up(struct sock * sk)30 static void smc_rx_wake_up(struct sock *sk)
31 {
32 struct socket_wq *wq;
33
34 /* derived from sock_def_readable() */
35 /* called already in smc_listen_work() */
36 rcu_read_lock();
37 wq = rcu_dereference(sk->sk_wq);
38 if (skwq_has_sleeper(wq))
39 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | EPOLLPRI |
40 EPOLLRDNORM | EPOLLRDBAND);
41 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
42 if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
43 (sk->sk_state == SMC_CLOSED))
44 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
45 rcu_read_unlock();
46 }
47
48 /* Update consumer cursor
49 * @conn connection to update
50 * @cons consumer cursor
51 * @len number of Bytes consumed
52 * Returns:
53 * 1 if we should end our receive, 0 otherwise
54 */
smc_rx_update_consumer(struct smc_sock * smc,union smc_host_cursor cons,size_t len)55 static int smc_rx_update_consumer(struct smc_sock *smc,
56 union smc_host_cursor cons, size_t len)
57 {
58 struct smc_connection *conn = &smc->conn;
59 struct sock *sk = &smc->sk;
60 bool force = false;
61 int diff, rc = 0;
62
63 smc_curs_add(conn->rmb_desc->len, &cons, len);
64
65 /* did we process urgent data? */
66 if (conn->urg_state == SMC_URG_VALID || conn->urg_rx_skip_pend) {
67 diff = smc_curs_comp(conn->rmb_desc->len, &cons,
68 &conn->urg_curs);
69 if (sock_flag(sk, SOCK_URGINLINE)) {
70 if (diff == 0) {
71 force = true;
72 rc = 1;
73 conn->urg_state = SMC_URG_READ;
74 }
75 } else {
76 if (diff == 1) {
77 /* skip urgent byte */
78 force = true;
79 smc_curs_add(conn->rmb_desc->len, &cons, 1);
80 conn->urg_rx_skip_pend = false;
81 } else if (diff < -1)
82 /* we read past urgent byte */
83 conn->urg_state = SMC_URG_READ;
84 }
85 }
86
87 smc_curs_copy(&conn->local_tx_ctrl.cons, &cons, conn);
88
89 /* send consumer cursor update if required */
90 /* similar to advertising new TCP rcv_wnd if required */
91 smc_tx_consumer_update(conn, force);
92
93 return rc;
94 }
95
smc_rx_update_cons(struct smc_sock * smc,size_t len)96 static void smc_rx_update_cons(struct smc_sock *smc, size_t len)
97 {
98 struct smc_connection *conn = &smc->conn;
99 union smc_host_cursor cons;
100
101 smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
102 smc_rx_update_consumer(smc, cons, len);
103 }
104
105 struct smc_spd_priv {
106 struct smc_sock *smc;
107 size_t len;
108 };
109
smc_rx_pipe_buf_release(struct pipe_inode_info * pipe,struct pipe_buffer * buf)110 static void smc_rx_pipe_buf_release(struct pipe_inode_info *pipe,
111 struct pipe_buffer *buf)
112 {
113 struct smc_spd_priv *priv = (struct smc_spd_priv *)buf->private;
114 struct smc_sock *smc = priv->smc;
115 struct smc_connection *conn;
116 struct sock *sk = &smc->sk;
117
118 if (sk->sk_state == SMC_CLOSED ||
119 sk->sk_state == SMC_PEERFINCLOSEWAIT ||
120 sk->sk_state == SMC_APPFINCLOSEWAIT)
121 goto out;
122 conn = &smc->conn;
123 lock_sock(sk);
124 smc_rx_update_cons(smc, priv->len);
125 release_sock(sk);
126 if (atomic_sub_and_test(priv->len, &conn->splice_pending))
127 smc_rx_wake_up(sk);
128 out:
129 kfree(priv);
130 put_page(buf->page);
131 sock_put(sk);
132 }
133
134 static const struct pipe_buf_operations smc_pipe_ops = {
135 .release = smc_rx_pipe_buf_release,
136 .get = generic_pipe_buf_get
137 };
138
smc_rx_spd_release(struct splice_pipe_desc * spd,unsigned int i)139 static void smc_rx_spd_release(struct splice_pipe_desc *spd,
140 unsigned int i)
141 {
142 put_page(spd->pages[i]);
143 }
144
smc_rx_splice(struct pipe_inode_info * pipe,char * src,size_t len,struct smc_sock * smc)145 static int smc_rx_splice(struct pipe_inode_info *pipe, char *src, size_t len,
146 struct smc_sock *smc)
147 {
148 struct splice_pipe_desc spd;
149 struct partial_page partial;
150 struct smc_spd_priv *priv;
151 int bytes;
152
153 priv = kzalloc(sizeof(*priv), GFP_KERNEL);
154 if (!priv)
155 return -ENOMEM;
156 priv->len = len;
157 priv->smc = smc;
158 partial.offset = src - (char *)smc->conn.rmb_desc->cpu_addr;
159 partial.len = len;
160 partial.private = (unsigned long)priv;
161
162 spd.nr_pages_max = 1;
163 spd.nr_pages = 1;
164 spd.pages = &smc->conn.rmb_desc->pages;
165 spd.partial = &partial;
166 spd.ops = &smc_pipe_ops;
167 spd.spd_release = smc_rx_spd_release;
168
169 bytes = splice_to_pipe(pipe, &spd);
170 if (bytes > 0) {
171 sock_hold(&smc->sk);
172 get_page(smc->conn.rmb_desc->pages);
173 atomic_add(bytes, &smc->conn.splice_pending);
174 }
175
176 return bytes;
177 }
178
smc_rx_data_available_and_no_splice_pend(struct smc_connection * conn)179 static int smc_rx_data_available_and_no_splice_pend(struct smc_connection *conn)
180 {
181 return atomic_read(&conn->bytes_to_rcv) &&
182 !atomic_read(&conn->splice_pending);
183 }
184
185 /* blocks rcvbuf consumer until >=len bytes available or timeout or interrupted
186 * @smc smc socket
187 * @timeo pointer to max seconds to wait, pointer to value 0 for no timeout
188 * @fcrit add'l criterion to evaluate as function pointer
189 * Returns:
190 * 1 if at least 1 byte available in rcvbuf or if socket error/shutdown.
191 * 0 otherwise (nothing in rcvbuf nor timeout, e.g. interrupted).
192 */
smc_rx_wait(struct smc_sock * smc,long * timeo,int (* fcrit)(struct smc_connection * conn))193 int smc_rx_wait(struct smc_sock *smc, long *timeo,
194 int (*fcrit)(struct smc_connection *conn))
195 {
196 DEFINE_WAIT_FUNC(wait, woken_wake_function);
197 struct smc_connection *conn = &smc->conn;
198 struct smc_cdc_conn_state_flags *cflags =
199 &conn->local_tx_ctrl.conn_state_flags;
200 struct sock *sk = &smc->sk;
201 int rc;
202
203 if (fcrit(conn))
204 return 1;
205 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
206 add_wait_queue(sk_sleep(sk), &wait);
207 rc = sk_wait_event(sk, timeo,
208 sk->sk_err ||
209 cflags->peer_conn_abort ||
210 sk->sk_shutdown & RCV_SHUTDOWN ||
211 conn->killed ||
212 fcrit(conn),
213 &wait);
214 remove_wait_queue(sk_sleep(sk), &wait);
215 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
216 return rc;
217 }
218
smc_rx_recv_urg(struct smc_sock * smc,struct msghdr * msg,int len,int flags)219 static int smc_rx_recv_urg(struct smc_sock *smc, struct msghdr *msg, int len,
220 int flags)
221 {
222 struct smc_connection *conn = &smc->conn;
223 union smc_host_cursor cons;
224 struct sock *sk = &smc->sk;
225 int rc = 0;
226
227 if (sock_flag(sk, SOCK_URGINLINE) ||
228 !(conn->urg_state == SMC_URG_VALID) ||
229 conn->urg_state == SMC_URG_READ)
230 return -EINVAL;
231
232 SMC_STAT_INC(smc, urg_data_cnt);
233 if (conn->urg_state == SMC_URG_VALID) {
234 if (!(flags & MSG_PEEK))
235 smc->conn.urg_state = SMC_URG_READ;
236 msg->msg_flags |= MSG_OOB;
237 if (len > 0) {
238 if (!(flags & MSG_TRUNC))
239 rc = memcpy_to_msg(msg, &conn->urg_rx_byte, 1);
240 len = 1;
241 smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
242 if (smc_curs_diff(conn->rmb_desc->len, &cons,
243 &conn->urg_curs) > 1)
244 conn->urg_rx_skip_pend = true;
245 /* Urgent Byte was already accounted for, but trigger
246 * skipping the urgent byte in non-inline case
247 */
248 if (!(flags & MSG_PEEK))
249 smc_rx_update_consumer(smc, cons, 0);
250 } else {
251 msg->msg_flags |= MSG_TRUNC;
252 }
253
254 return rc ? -EFAULT : len;
255 }
256
257 if (sk->sk_state == SMC_CLOSED || sk->sk_shutdown & RCV_SHUTDOWN)
258 return 0;
259
260 return -EAGAIN;
261 }
262
smc_rx_recvmsg_data_available(struct smc_sock * smc)263 static bool smc_rx_recvmsg_data_available(struct smc_sock *smc)
264 {
265 struct smc_connection *conn = &smc->conn;
266
267 if (smc_rx_data_available(conn))
268 return true;
269 else if (conn->urg_state == SMC_URG_VALID)
270 /* we received a single urgent Byte - skip */
271 smc_rx_update_cons(smc, 0);
272 return false;
273 }
274
275 /* smc_rx_recvmsg - receive data from RMBE
276 * @msg: copy data to receive buffer
277 * @pipe: copy data to pipe if set - indicates splice() call
278 *
279 * rcvbuf consumer: main API called by socket layer.
280 * Called under sk lock.
281 */
smc_rx_recvmsg(struct smc_sock * smc,struct msghdr * msg,struct pipe_inode_info * pipe,size_t len,int flags)282 int smc_rx_recvmsg(struct smc_sock *smc, struct msghdr *msg,
283 struct pipe_inode_info *pipe, size_t len, int flags)
284 {
285 size_t copylen, read_done = 0, read_remaining = len;
286 size_t chunk_len, chunk_off, chunk_len_sum;
287 struct smc_connection *conn = &smc->conn;
288 int (*func)(struct smc_connection *conn);
289 union smc_host_cursor cons;
290 int readable, chunk;
291 char *rcvbuf_base;
292 struct sock *sk;
293 int splbytes;
294 long timeo;
295 int target; /* Read at least these many bytes */
296 int rc;
297
298 if (unlikely(flags & MSG_ERRQUEUE))
299 return -EINVAL; /* future work for sk.sk_family == AF_SMC */
300
301 sk = &smc->sk;
302 if (sk->sk_state == SMC_LISTEN)
303 return -ENOTCONN;
304 if (flags & MSG_OOB)
305 return smc_rx_recv_urg(smc, msg, len, flags);
306 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
307 target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
308
309 readable = atomic_read(&conn->bytes_to_rcv);
310 if (readable >= conn->rmb_desc->len)
311 SMC_STAT_RMB_RX_FULL(smc, !conn->lnk);
312
313 if (len < readable)
314 SMC_STAT_RMB_RX_SIZE_SMALL(smc, !conn->lnk);
315 /* we currently use 1 RMBE per RMB, so RMBE == RMB base addr */
316 rcvbuf_base = conn->rx_off + conn->rmb_desc->cpu_addr;
317
318 do { /* while (read_remaining) */
319 if (read_done >= target || (pipe && read_done))
320 break;
321
322 if (conn->killed)
323 break;
324
325 if (smc_rx_recvmsg_data_available(smc))
326 goto copy;
327
328 if (sk->sk_shutdown & RCV_SHUTDOWN) {
329 /* smc_cdc_msg_recv_action() could have run after
330 * above smc_rx_recvmsg_data_available()
331 */
332 if (smc_rx_recvmsg_data_available(smc))
333 goto copy;
334 break;
335 }
336
337 if (read_done) {
338 if (sk->sk_err ||
339 sk->sk_state == SMC_CLOSED ||
340 !timeo ||
341 signal_pending(current))
342 break;
343 } else {
344 if (sk->sk_err) {
345 read_done = sock_error(sk);
346 break;
347 }
348 if (sk->sk_state == SMC_CLOSED) {
349 if (!sock_flag(sk, SOCK_DONE)) {
350 /* This occurs when user tries to read
351 * from never connected socket.
352 */
353 read_done = -ENOTCONN;
354 break;
355 }
356 break;
357 }
358 if (!timeo)
359 return -EAGAIN;
360 if (signal_pending(current)) {
361 read_done = sock_intr_errno(timeo);
362 break;
363 }
364 }
365
366 if (!smc_rx_data_available(conn)) {
367 smc_rx_wait(smc, &timeo, smc_rx_data_available);
368 continue;
369 }
370
371 copy:
372 /* initialize variables for 1st iteration of subsequent loop */
373 /* could be just 1 byte, even after waiting on data above */
374 readable = atomic_read(&conn->bytes_to_rcv);
375 splbytes = atomic_read(&conn->splice_pending);
376 if (!readable || (msg && splbytes)) {
377 if (splbytes)
378 func = smc_rx_data_available_and_no_splice_pend;
379 else
380 func = smc_rx_data_available;
381 smc_rx_wait(smc, &timeo, func);
382 continue;
383 }
384
385 smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
386 /* subsequent splice() calls pick up where previous left */
387 if (splbytes)
388 smc_curs_add(conn->rmb_desc->len, &cons, splbytes);
389 if (conn->urg_state == SMC_URG_VALID &&
390 sock_flag(&smc->sk, SOCK_URGINLINE) &&
391 readable > 1)
392 readable--; /* always stop at urgent Byte */
393 /* not more than what user space asked for */
394 copylen = min_t(size_t, read_remaining, readable);
395 /* determine chunks where to read from rcvbuf */
396 /* either unwrapped case, or 1st chunk of wrapped case */
397 chunk_len = min_t(size_t, copylen, conn->rmb_desc->len -
398 cons.count);
399 chunk_len_sum = chunk_len;
400 chunk_off = cons.count;
401 smc_rmb_sync_sg_for_cpu(conn);
402 for (chunk = 0; chunk < 2; chunk++) {
403 if (!(flags & MSG_TRUNC)) {
404 if (msg) {
405 rc = memcpy_to_msg(msg, rcvbuf_base +
406 chunk_off,
407 chunk_len);
408 } else {
409 rc = smc_rx_splice(pipe, rcvbuf_base +
410 chunk_off, chunk_len,
411 smc);
412 }
413 if (rc < 0) {
414 if (!read_done)
415 read_done = -EFAULT;
416 smc_rmb_sync_sg_for_device(conn);
417 goto out;
418 }
419 }
420 read_remaining -= chunk_len;
421 read_done += chunk_len;
422
423 if (chunk_len_sum == copylen)
424 break; /* either on 1st or 2nd iteration */
425 /* prepare next (== 2nd) iteration */
426 chunk_len = copylen - chunk_len; /* remainder */
427 chunk_len_sum += chunk_len;
428 chunk_off = 0; /* modulo offset in recv ring buffer */
429 }
430 smc_rmb_sync_sg_for_device(conn);
431
432 /* update cursors */
433 if (!(flags & MSG_PEEK)) {
434 /* increased in recv tasklet smc_cdc_msg_rcv() */
435 smp_mb__before_atomic();
436 atomic_sub(copylen, &conn->bytes_to_rcv);
437 /* guarantee 0 <= bytes_to_rcv <= rmb_desc->len */
438 smp_mb__after_atomic();
439 if (msg && smc_rx_update_consumer(smc, cons, copylen))
440 goto out;
441 }
442
443 trace_smc_rx_recvmsg(smc, copylen);
444 } while (read_remaining);
445 out:
446 return read_done;
447 }
448
449 /* Initialize receive properties on connection establishment. NB: not __init! */
smc_rx_init(struct smc_sock * smc)450 void smc_rx_init(struct smc_sock *smc)
451 {
452 smc->sk.sk_data_ready = smc_rx_wake_up;
453 atomic_set(&smc->conn.splice_pending, 0);
454 smc->conn.urg_state = SMC_URG_READ;
455 }
456