xref: /linux/net/smc/smc_rx.c (revision 02680c23d7b3febe45ea3d4f9818c2b2dc89020a)
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 
25 /* callback implementation to wakeup consumers blocked with smc_rx_wait().
26  * indirectly called by smc_cdc_msg_recv_action().
27  */
28 static void smc_rx_wake_up(struct sock *sk)
29 {
30 	struct socket_wq *wq;
31 
32 	/* derived from sock_def_readable() */
33 	/* called already in smc_listen_work() */
34 	rcu_read_lock();
35 	wq = rcu_dereference(sk->sk_wq);
36 	if (skwq_has_sleeper(wq))
37 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | EPOLLPRI |
38 						EPOLLRDNORM | EPOLLRDBAND);
39 	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
40 	if ((sk->sk_shutdown == SHUTDOWN_MASK) ||
41 	    (sk->sk_state == SMC_CLOSED))
42 		sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
43 	rcu_read_unlock();
44 }
45 
46 /* Update consumer cursor
47  *   @conn   connection to update
48  *   @cons   consumer cursor
49  *   @len    number of Bytes consumed
50  *   Returns:
51  *   1 if we should end our receive, 0 otherwise
52  */
53 static int smc_rx_update_consumer(struct smc_sock *smc,
54 				  union smc_host_cursor cons, size_t len)
55 {
56 	struct smc_connection *conn = &smc->conn;
57 	struct sock *sk = &smc->sk;
58 	bool force = false;
59 	int diff, rc = 0;
60 
61 	smc_curs_add(conn->rmb_desc->len, &cons, len);
62 
63 	/* did we process urgent data? */
64 	if (conn->urg_state == SMC_URG_VALID || conn->urg_rx_skip_pend) {
65 		diff = smc_curs_comp(conn->rmb_desc->len, &cons,
66 				     &conn->urg_curs);
67 		if (sock_flag(sk, SOCK_URGINLINE)) {
68 			if (diff == 0) {
69 				force = true;
70 				rc = 1;
71 				conn->urg_state = SMC_URG_READ;
72 			}
73 		} else {
74 			if (diff == 1) {
75 				/* skip urgent byte */
76 				force = true;
77 				smc_curs_add(conn->rmb_desc->len, &cons, 1);
78 				conn->urg_rx_skip_pend = false;
79 			} else if (diff < -1)
80 				/* we read past urgent byte */
81 				conn->urg_state = SMC_URG_READ;
82 		}
83 	}
84 
85 	smc_curs_copy(&conn->local_tx_ctrl.cons, &cons, conn);
86 
87 	/* send consumer cursor update if required */
88 	/* similar to advertising new TCP rcv_wnd if required */
89 	smc_tx_consumer_update(conn, force);
90 
91 	return rc;
92 }
93 
94 static void smc_rx_update_cons(struct smc_sock *smc, size_t len)
95 {
96 	struct smc_connection *conn = &smc->conn;
97 	union smc_host_cursor cons;
98 
99 	smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
100 	smc_rx_update_consumer(smc, cons, len);
101 }
102 
103 struct smc_spd_priv {
104 	struct smc_sock *smc;
105 	size_t		 len;
106 };
107 
108 static void smc_rx_pipe_buf_release(struct pipe_inode_info *pipe,
109 				    struct pipe_buffer *buf)
110 {
111 	struct smc_spd_priv *priv = (struct smc_spd_priv *)buf->private;
112 	struct smc_sock *smc = priv->smc;
113 	struct smc_connection *conn;
114 	struct sock *sk = &smc->sk;
115 
116 	if (sk->sk_state == SMC_CLOSED ||
117 	    sk->sk_state == SMC_PEERFINCLOSEWAIT ||
118 	    sk->sk_state == SMC_APPFINCLOSEWAIT)
119 		goto out;
120 	conn = &smc->conn;
121 	lock_sock(sk);
122 	smc_rx_update_cons(smc, priv->len);
123 	release_sock(sk);
124 	if (atomic_sub_and_test(priv->len, &conn->splice_pending))
125 		smc_rx_wake_up(sk);
126 out:
127 	kfree(priv);
128 	put_page(buf->page);
129 	sock_put(sk);
130 }
131 
132 static const struct pipe_buf_operations smc_pipe_ops = {
133 	.release = smc_rx_pipe_buf_release,
134 	.get = generic_pipe_buf_get
135 };
136 
137 static void smc_rx_spd_release(struct splice_pipe_desc *spd,
138 			       unsigned int i)
139 {
140 	put_page(spd->pages[i]);
141 }
142 
143 static int smc_rx_splice(struct pipe_inode_info *pipe, char *src, size_t len,
144 			 struct smc_sock *smc)
145 {
146 	struct splice_pipe_desc spd;
147 	struct partial_page partial;
148 	struct smc_spd_priv *priv;
149 	int bytes;
150 
151 	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
152 	if (!priv)
153 		return -ENOMEM;
154 	priv->len = len;
155 	priv->smc = smc;
156 	partial.offset = src - (char *)smc->conn.rmb_desc->cpu_addr;
157 	partial.len = len;
158 	partial.private = (unsigned long)priv;
159 
160 	spd.nr_pages_max = 1;
161 	spd.nr_pages = 1;
162 	spd.pages = &smc->conn.rmb_desc->pages;
163 	spd.partial = &partial;
164 	spd.ops = &smc_pipe_ops;
165 	spd.spd_release = smc_rx_spd_release;
166 
167 	bytes = splice_to_pipe(pipe, &spd);
168 	if (bytes > 0) {
169 		sock_hold(&smc->sk);
170 		get_page(smc->conn.rmb_desc->pages);
171 		atomic_add(bytes, &smc->conn.splice_pending);
172 	}
173 
174 	return bytes;
175 }
176 
177 static int smc_rx_data_available_and_no_splice_pend(struct smc_connection *conn)
178 {
179 	return atomic_read(&conn->bytes_to_rcv) &&
180 	       !atomic_read(&conn->splice_pending);
181 }
182 
183 /* blocks rcvbuf consumer until >=len bytes available or timeout or interrupted
184  *   @smc    smc socket
185  *   @timeo  pointer to max seconds to wait, pointer to value 0 for no timeout
186  *   @fcrit  add'l criterion to evaluate as function pointer
187  * Returns:
188  * 1 if at least 1 byte available in rcvbuf or if socket error/shutdown.
189  * 0 otherwise (nothing in rcvbuf nor timeout, e.g. interrupted).
190  */
191 int smc_rx_wait(struct smc_sock *smc, long *timeo,
192 		int (*fcrit)(struct smc_connection *conn))
193 {
194 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
195 	struct smc_connection *conn = &smc->conn;
196 	struct smc_cdc_conn_state_flags *cflags =
197 					&conn->local_tx_ctrl.conn_state_flags;
198 	struct sock *sk = &smc->sk;
199 	int rc;
200 
201 	if (fcrit(conn))
202 		return 1;
203 	sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
204 	add_wait_queue(sk_sleep(sk), &wait);
205 	rc = sk_wait_event(sk, timeo,
206 			   sk->sk_err ||
207 			   cflags->peer_conn_abort ||
208 			   sk->sk_shutdown & RCV_SHUTDOWN ||
209 			   conn->killed ||
210 			   fcrit(conn),
211 			   &wait);
212 	remove_wait_queue(sk_sleep(sk), &wait);
213 	sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
214 	return rc;
215 }
216 
217 static int smc_rx_recv_urg(struct smc_sock *smc, struct msghdr *msg, int len,
218 			   int flags)
219 {
220 	struct smc_connection *conn = &smc->conn;
221 	union smc_host_cursor cons;
222 	struct sock *sk = &smc->sk;
223 	int rc = 0;
224 
225 	if (sock_flag(sk, SOCK_URGINLINE) ||
226 	    !(conn->urg_state == SMC_URG_VALID) ||
227 	    conn->urg_state == SMC_URG_READ)
228 		return -EINVAL;
229 
230 	if (conn->urg_state == SMC_URG_VALID) {
231 		if (!(flags & MSG_PEEK))
232 			smc->conn.urg_state = SMC_URG_READ;
233 		msg->msg_flags |= MSG_OOB;
234 		if (len > 0) {
235 			if (!(flags & MSG_TRUNC))
236 				rc = memcpy_to_msg(msg, &conn->urg_rx_byte, 1);
237 			len = 1;
238 			smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
239 			if (smc_curs_diff(conn->rmb_desc->len, &cons,
240 					  &conn->urg_curs) > 1)
241 				conn->urg_rx_skip_pend = true;
242 			/* Urgent Byte was already accounted for, but trigger
243 			 * skipping the urgent byte in non-inline case
244 			 */
245 			if (!(flags & MSG_PEEK))
246 				smc_rx_update_consumer(smc, cons, 0);
247 		} else {
248 			msg->msg_flags |= MSG_TRUNC;
249 		}
250 
251 		return rc ? -EFAULT : len;
252 	}
253 
254 	if (sk->sk_state == SMC_CLOSED || sk->sk_shutdown & RCV_SHUTDOWN)
255 		return 0;
256 
257 	return -EAGAIN;
258 }
259 
260 static bool smc_rx_recvmsg_data_available(struct smc_sock *smc)
261 {
262 	struct smc_connection *conn = &smc->conn;
263 
264 	if (smc_rx_data_available(conn))
265 		return true;
266 	else if (conn->urg_state == SMC_URG_VALID)
267 		/* we received a single urgent Byte - skip */
268 		smc_rx_update_cons(smc, 0);
269 	return false;
270 }
271 
272 /* smc_rx_recvmsg - receive data from RMBE
273  * @msg:	copy data to receive buffer
274  * @pipe:	copy data to pipe if set - indicates splice() call
275  *
276  * rcvbuf consumer: main API called by socket layer.
277  * Called under sk lock.
278  */
279 int smc_rx_recvmsg(struct smc_sock *smc, struct msghdr *msg,
280 		   struct pipe_inode_info *pipe, size_t len, int flags)
281 {
282 	size_t copylen, read_done = 0, read_remaining = len;
283 	size_t chunk_len, chunk_off, chunk_len_sum;
284 	struct smc_connection *conn = &smc->conn;
285 	int (*func)(struct smc_connection *conn);
286 	union smc_host_cursor cons;
287 	int readable, chunk;
288 	char *rcvbuf_base;
289 	struct sock *sk;
290 	int splbytes;
291 	long timeo;
292 	int target;		/* Read at least these many bytes */
293 	int rc;
294 
295 	if (unlikely(flags & MSG_ERRQUEUE))
296 		return -EINVAL; /* future work for sk.sk_family == AF_SMC */
297 
298 	sk = &smc->sk;
299 	if (sk->sk_state == SMC_LISTEN)
300 		return -ENOTCONN;
301 	if (flags & MSG_OOB)
302 		return smc_rx_recv_urg(smc, msg, len, flags);
303 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
304 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
305 
306 	/* we currently use 1 RMBE per RMB, so RMBE == RMB base addr */
307 	rcvbuf_base = conn->rx_off + conn->rmb_desc->cpu_addr;
308 
309 	do { /* while (read_remaining) */
310 		if (read_done >= target || (pipe && read_done))
311 			break;
312 
313 		if (conn->killed)
314 			break;
315 
316 		if (smc_rx_recvmsg_data_available(smc))
317 			goto copy;
318 
319 		if (sk->sk_shutdown & RCV_SHUTDOWN) {
320 			/* smc_cdc_msg_recv_action() could have run after
321 			 * above smc_rx_recvmsg_data_available()
322 			 */
323 			if (smc_rx_recvmsg_data_available(smc))
324 				goto copy;
325 			break;
326 		}
327 
328 		if (read_done) {
329 			if (sk->sk_err ||
330 			    sk->sk_state == SMC_CLOSED ||
331 			    !timeo ||
332 			    signal_pending(current))
333 				break;
334 		} else {
335 			if (sk->sk_err) {
336 				read_done = sock_error(sk);
337 				break;
338 			}
339 			if (sk->sk_state == SMC_CLOSED) {
340 				if (!sock_flag(sk, SOCK_DONE)) {
341 					/* This occurs when user tries to read
342 					 * from never connected socket.
343 					 */
344 					read_done = -ENOTCONN;
345 					break;
346 				}
347 				break;
348 			}
349 			if (signal_pending(current)) {
350 				read_done = sock_intr_errno(timeo);
351 				break;
352 			}
353 			if (!timeo)
354 				return -EAGAIN;
355 		}
356 
357 		if (!smc_rx_data_available(conn)) {
358 			smc_rx_wait(smc, &timeo, smc_rx_data_available);
359 			continue;
360 		}
361 
362 copy:
363 		/* initialize variables for 1st iteration of subsequent loop */
364 		/* could be just 1 byte, even after waiting on data above */
365 		readable = atomic_read(&conn->bytes_to_rcv);
366 		splbytes = atomic_read(&conn->splice_pending);
367 		if (!readable || (msg && splbytes)) {
368 			if (splbytes)
369 				func = smc_rx_data_available_and_no_splice_pend;
370 			else
371 				func = smc_rx_data_available;
372 			smc_rx_wait(smc, &timeo, func);
373 			continue;
374 		}
375 
376 		smc_curs_copy(&cons, &conn->local_tx_ctrl.cons, conn);
377 		/* subsequent splice() calls pick up where previous left */
378 		if (splbytes)
379 			smc_curs_add(conn->rmb_desc->len, &cons, splbytes);
380 		if (conn->urg_state == SMC_URG_VALID &&
381 		    sock_flag(&smc->sk, SOCK_URGINLINE) &&
382 		    readable > 1)
383 			readable--;	/* always stop at urgent Byte */
384 		/* not more than what user space asked for */
385 		copylen = min_t(size_t, read_remaining, readable);
386 		/* determine chunks where to read from rcvbuf */
387 		/* either unwrapped case, or 1st chunk of wrapped case */
388 		chunk_len = min_t(size_t, copylen, conn->rmb_desc->len -
389 				  cons.count);
390 		chunk_len_sum = chunk_len;
391 		chunk_off = cons.count;
392 		smc_rmb_sync_sg_for_cpu(conn);
393 		for (chunk = 0; chunk < 2; chunk++) {
394 			if (!(flags & MSG_TRUNC)) {
395 				if (msg) {
396 					rc = memcpy_to_msg(msg, rcvbuf_base +
397 							   chunk_off,
398 							   chunk_len);
399 				} else {
400 					rc = smc_rx_splice(pipe, rcvbuf_base +
401 							chunk_off, chunk_len,
402 							smc);
403 				}
404 				if (rc < 0) {
405 					if (!read_done)
406 						read_done = -EFAULT;
407 					smc_rmb_sync_sg_for_device(conn);
408 					goto out;
409 				}
410 			}
411 			read_remaining -= chunk_len;
412 			read_done += chunk_len;
413 
414 			if (chunk_len_sum == copylen)
415 				break; /* either on 1st or 2nd iteration */
416 			/* prepare next (== 2nd) iteration */
417 			chunk_len = copylen - chunk_len; /* remainder */
418 			chunk_len_sum += chunk_len;
419 			chunk_off = 0; /* modulo offset in recv ring buffer */
420 		}
421 		smc_rmb_sync_sg_for_device(conn);
422 
423 		/* update cursors */
424 		if (!(flags & MSG_PEEK)) {
425 			/* increased in recv tasklet smc_cdc_msg_rcv() */
426 			smp_mb__before_atomic();
427 			atomic_sub(copylen, &conn->bytes_to_rcv);
428 			/* guarantee 0 <= bytes_to_rcv <= rmb_desc->len */
429 			smp_mb__after_atomic();
430 			if (msg && smc_rx_update_consumer(smc, cons, copylen))
431 				goto out;
432 		}
433 	} while (read_remaining);
434 out:
435 	return read_done;
436 }
437 
438 /* Initialize receive properties on connection establishment. NB: not __init! */
439 void smc_rx_init(struct smc_sock *smc)
440 {
441 	smc->sk.sk_data_ready = smc_rx_wake_up;
442 	atomic_set(&smc->conn.splice_pending, 0);
443 	smc->conn.urg_state = SMC_URG_READ;
444 }
445