aboutsummaryrefslogtreecommitdiffstats
path: root/net/rds/tcp_recv.c
blob: 49c9660107882b0d1f2d8de973babf6102ea7e24 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
/*
 * Copyright (c) 2006 Oracle.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */
#include <linux/kernel.h>
#include <linux/slab.h>
#include <net/tcp.h>

#include "rds.h"
#include "tcp.h"

static struct kmem_cache *rds_tcp_incoming_slab;

static void rds_tcp_inc_purge(struct rds_incoming *inc)
{
	struct rds_tcp_incoming *tinc;
	tinc = container_of(inc, struct rds_tcp_incoming, ti_inc);
	rdsdebug("purging tinc %p inc %p\n", tinc, inc);
	skb_queue_purge(&tinc->ti_skb_list);
}

void rds_tcp_inc_free(struct rds_incoming *inc)
{
	struct rds_tcp_incoming *tinc;
	tinc = container_of(inc, struct rds_tcp_incoming, ti_inc);
	rds_tcp_inc_purge(inc);
	rdsdebug("freeing tinc %p inc %p\n", tinc, inc);
	kmem_cache_free(rds_tcp_incoming_slab, tinc);
}

/*
 * this is pretty lame, but, whatever.
 */
int rds_tcp_inc_copy_to_user(struct rds_incoming *inc, struct iovec *first_iov,
			     size_t size)
{
	struct rds_tcp_incoming *tinc;
	struct iovec *iov, tmp;
	struct sk_buff *skb;
	unsigned long to_copy, skb_off;
	int ret = 0;

	if (size == 0)
		goto out;

	tinc = container_of(inc, struct rds_tcp_incoming, ti_inc);
	iov = first_iov;
	tmp = *iov;

	skb_queue_walk(&tinc->ti_skb_list, skb) {
		skb_off = 0;
		while (skb_off < skb->len) {
			while (tmp.iov_len == 0) {
				iov++;
				tmp = *iov;
			}

			to_copy = min(tmp.iov_len, size);
			to_copy = min(to_copy, skb->len - skb_off);

			rdsdebug("ret %d size %zu skb %p skb_off %lu "
				 "skblen %d iov_base %p iov_len %zu cpy %lu\n",
				 ret, size, skb, skb_off, skb->len,
				 tmp.iov_base, tmp.iov_len, to_copy);

			/* modifies tmp as it copies */
			if (skb_copy_datagram_iovec(skb, skb_off, &tmp,
						    to_copy)) {
				ret = -EFAULT;
				goto out;
			}

			rds_stats_add(s_copy_to_user, to_copy);
			size -= to_copy;
			ret += to_copy;
			skb_off += to_copy;
			if (size == 0)
				goto out;
		}
	}
out:
	return ret;
}

/*
 * We have a series of skbs that have fragmented pieces of the congestion
 * bitmap.  They must add up to the exact size of the congestion bitmap.  We
 * use the skb helpers to copy those into the pages that make up the in-memory
 * congestion bitmap for the remote address of this connection.  We then tell
 * the congestion core that the bitmap has been changed so that it can wake up
 * sleepers.
 *
 * This is racing with sending paths which are using test_bit to see if the
 * bitmap indicates that their recipient is congested.
 */

static void rds_tcp_cong_recv(struct rds_connection *conn,
			      struct rds_tcp_incoming *tinc)
{
	struct sk_buff *skb;
	unsigned int to_copy, skb_off;
	unsigned int map_off;
	unsigned int map_page;
	struct rds_cong_map *map;
	int ret;

	/* catch completely corrupt packets */
	if (be32_to_cpu(tinc->ti_inc.i_hdr.h_len) != RDS_CONG_MAP_BYTES)
		return;

	map_page = 0;
	map_off = 0;
	map = conn->c_fcong;

	skb_queue_walk(&tinc->ti_skb_list, skb) {
		skb_off = 0;
		while (skb_off < skb->len) {
			to_copy = min_t(unsigned int, PAGE_SIZE - map_off,
					skb->len - skb_off);

			BUG_ON(map_page >= RDS_CONG_MAP_PAGES);

			/* only returns 0 or -error */
			ret = skb_copy_bits(skb, skb_off,
				(void *)map->m_page_addrs[map_page] + map_off,
				to_copy);
			BUG_ON(ret != 0);

			skb_off += to_copy;
			map_off += to_copy;
			if (map_off == PAGE_SIZE) {
				map_off = 0;
				map_page++;
			}
		}
	}

	rds_cong_map_updated(map, ~(u64) 0);
}

struct rds_tcp_desc_arg {
	struct rds_connection *conn;
	gfp_t gfp;
	enum km_type km;
};

static int rds_tcp_data_recv(read_descriptor_t *desc, struct sk_buff *skb,
			     unsigned int offset, size_t len)
{
	struct rds_tcp_desc_arg *arg = desc->arg.data;
	struct rds_connection *conn = arg->conn;
	struct rds_tcp_connection *tc = conn->c_transport_data;
	struct rds_tcp_incoming *tinc = tc->t_tinc;
	struct sk_buff *clone;
	size_t left = len, to_copy;

	rdsdebug("tcp data tc %p skb %p offset %u len %zu\n", tc, skb, offset,
		 len);

	/*
	 * tcp_read_sock() interprets partial progress as an indication to stop
	 * processing.
	 */
	while (left) {
		if (!tinc) {
			tinc = kmem_cache_alloc(rds_tcp_incoming_slab,
					        arg->gfp);
			if (!tinc) {
				desc->error = -ENOMEM;
				goto out;
			}
			tc->t_tinc = tinc;
			rdsdebug("alloced tinc %p\n", tinc);
			rds_inc_init(&tinc->ti_inc, conn, conn->c_faddr);
			/*
			 * XXX * we might be able to use the __ variants when
			 * we've already serialized at a higher level.
			 */
			skb_queue_head_init(&tinc->ti_skb_list);
		}

		if (left && tc->t_tinc_hdr_rem) {
			to_copy = min(tc->t_tinc_hdr_rem, left);
			rdsdebug("copying %zu header from skb %p\n", to_copy,
				 skb);
			skb_copy_bits(skb, offset,
				      (char *)&tinc->ti_inc.i_hdr +
						sizeof(struct rds_header) -
						tc->t_tinc_hdr_rem,
				      to_copy);
			tc->t_tinc_hdr_rem -= to_copy;
			left -= to_copy;
			offset += to_copy;

			if (tc->t_tinc_hdr_rem == 0) {
				/* could be 0 for a 0 len message */
				tc->t_tinc_data_rem =
					be32_to_cpu(tinc->ti_inc.i_hdr.h_len);
			}
		}

		if (left && tc->t_tinc_data_rem) {
			clone = skb_clone(skb, arg->gfp);
			if (!clone) {
				desc->error = -ENOMEM;
				goto out;
			}

			to_copy = min(tc->t_tinc_data_rem, left);
			pskb_pull(clone, offset);
			pskb_trim(clone, to_copy);
			skb_queue_tail(&tinc->ti_skb_list, clone);

			rdsdebug("skb %p data %p len %d off %u to_copy %zu -> "
				 "clone %p data %p len %d\n",
				 skb, skb->data, skb->len, offset, to_copy,
				 clone, clone->data, clone->len);

			tc->t_tinc_data_rem -= to_copy;
			left -= to_copy;
			offset += to_copy;
		}

		if (tc->t_tinc_hdr_rem == 0 && tc->t_tinc_data_rem == 0) {
			if (tinc->ti_inc.i_hdr.h_flags == RDS_FLAG_CONG_BITMAP)
				rds_tcp_cong_recv(conn, tinc);
			else
				rds_recv_incoming(conn, conn->c_faddr,
						  conn->c_laddr, &tinc->ti_inc,
						  arg->gfp, arg->km);

			tc->t_tinc_hdr_rem = sizeof(struct rds_header);
			tc->t_tinc_data_rem = 0;
			tc->t_tinc = NULL;
			rds_inc_put(&tinc->ti_inc);
			tinc = NULL;
		}
	}
out:
	rdsdebug("returning len %zu left %zu skb len %d rx queue depth %d\n",
		 len, left, skb->len,
		 skb_queue_len(&tc->t_sock->sk->sk_receive_queue));
	return len - left;
}

/* the caller has to hold the sock lock */
int rds_tcp_read_sock(struct rds_connection *conn, gfp_t gfp, enum km_type km)
{
	struct rds_tcp_connection *tc = conn->c_transport_data;
	struct socket *sock = tc->t_sock;
	read_descriptor_t desc;
	struct rds_tcp_desc_arg arg;

	/* It's like glib in the kernel! */
	arg.conn = conn;
	arg.gfp = gfp;
	arg.km = km;
	desc.arg.data = &arg;
	desc.error = 0;
	desc.count = 1; /* give more than one skb per call */

	tcp_read_sock(sock->sk, &desc, rds_tcp_data_recv);
	rdsdebug("tcp_read_sock for tc %p gfp 0x%x returned %d\n", tc, gfp,
		 desc.error);

	return desc.error;
}

/*
 * We hold the sock lock to serialize our rds_tcp_recv->tcp_read_sock from
 * data_ready.
 *
 * if we fail to allocate we're in trouble.. blindly wait some time before
 * trying again to see if the VM can free up something for us.
 */
int rds_tcp_recv(struct rds_connection *conn)
{
	struct rds_tcp_connection *tc = conn->c_transport_data;
	struct socket *sock = tc->t_sock;
	int ret = 0;

	rdsdebug("recv worker conn %p tc %p sock %p\n", conn, tc, sock);

	lock_sock(sock->sk);
	ret = rds_tcp_read_sock(conn, GFP_KERNEL, KM_USER0);
	release_sock(sock->sk);

	return ret;
}

void rds_tcp_data_ready(struct sock *sk, int bytes)
{
	void (*ready)(struct sock *sk, int bytes);
	struct rds_connection *conn;
	struct rds_tcp_connection *tc;

	rdsdebug("data ready sk %p bytes %d\n", sk, bytes);

	read_lock(&sk->sk_callback_lock);
	conn = sk->sk_user_data;
	if (!conn) { /* check for teardown race */
		ready = sk->sk_data_ready;
		goto out;
	}

	tc = conn->c_transport_data;
	ready = tc->t_orig_data_ready;
	rds_tcp_stats_inc(s_tcp_data_ready_calls);

	if (rds_tcp_read_sock(conn, GFP_ATOMIC, KM_SOFTIRQ0) == -ENOMEM)
		queue_delayed_work(rds_wq, &conn->c_recv_w, 0);
out:
	read_unlock(&sk->sk_callback_lock);
	ready(sk, bytes);
}

int __init rds_tcp_recv_init(void)
{
	rds_tcp_incoming_slab = kmem_cache_create("rds_tcp_incoming",
					sizeof(struct rds_tcp_incoming),
					0, 0, NULL);
	if (!rds_tcp_incoming_slab)
		return -ENOMEM;
	return 0;
}

void rds_tcp_recv_exit(void)
{
	kmem_cache_destroy(rds_tcp_incoming_slab);
}