blob: 59795e42c8b61464512184990c3a12698c9a184f [file] [log] [blame]
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001/*
2 * NET Generic infrastructure for Network protocols.
3 *
4 * Definitions for request_sock
5 *
6 * Authors: Arnaldo Carvalho de Melo <acme@conectiva.com.br>
7 *
8 * From code originally in include/net/tcp.h
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 */
15#ifndef _REQUEST_SOCK_H
16#define _REQUEST_SOCK_H
17
18#include <linux/slab.h>
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070019#include <linux/spinlock.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070020#include <linux/types.h>
Ilpo Järvinen547b7922008-07-25 21:43:18 -070021#include <linux/bug.h>
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070022
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070023#include <net/sock.h>
24
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070025struct request_sock;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070026struct sk_buff;
27struct dst_entry;
28struct proto;
29
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070030struct request_sock_ops {
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070031 int family;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070032 int obj_size;
Christoph Lametere18b8902006-12-06 20:33:20 -080033 struct kmem_cache *slab;
Catalin Marinas7e56b5d2008-11-21 16:45:22 -080034 char *slab_name;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070035 int (*rtx_syn_ack)(struct sock *sk,
Christoph Paasch1a2c6182013-03-17 08:23:34 +000036 struct request_sock *req);
Gui Jianfeng6edafaa2008-08-06 23:50:04 -070037 void (*send_ack)(struct sock *sk, struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070038 struct request_sock *req);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -080039 void (*send_reset)(struct sock *sk,
40 struct sk_buff *skb);
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070041 void (*destructor)(struct request_sock *req);
Octavian Purdila72659ec2010-01-17 19:09:39 -080042 void (*syn_ack_timeout)(struct sock *sk,
43 struct request_sock *req);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070044};
45
Eric Dumazete6c022a2012-10-27 23:16:46 +000046extern int inet_rtx_syn_ack(struct sock *parent, struct request_sock *req);
47
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070048/* struct request_sock - mini sock to represent a connection request
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070049 */
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070050struct request_sock {
Eric Dumazet3fb62c52013-04-19 14:29:25 -070051 struct request_sock *dl_next;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070052 u16 mss;
Eric Dumazete6c022a2012-10-27 23:16:46 +000053 u8 num_retrans; /* number of retransmits */
54 u8 cookie_ts:1; /* syncookie: encode tcpopts in timestamp */
55 u8 num_timeout:7; /* number of timeouts */
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070056 /* The following two fields can be easily recomputed I think -AK */
57 u32 window_clamp; /* window clamp at creation time */
58 u32 rcv_wnd; /* rcv_wnd offered first time */
59 u32 ts_recent;
60 unsigned long expires;
Eric Dumazet72a3eff2006-11-16 02:30:37 -080061 const struct request_sock_ops *rsk_ops;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070062 struct sock *sk;
Venkat Yekkirala4237c752006-07-24 23:32:50 -070063 u32 secid;
Venkat Yekkirala6b877692006-11-08 17:04:09 -060064 u32 peer_secid;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070065};
66
Eric Dumazet72a3eff2006-11-16 02:30:37 -080067static inline struct request_sock *reqsk_alloc(const struct request_sock_ops *ops)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070068{
Christoph Lameter54e6ecb2006-12-06 20:33:16 -080069 struct request_sock *req = kmem_cache_alloc(ops->slab, GFP_ATOMIC);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070070
71 if (req != NULL)
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070072 req->rsk_ops = ops;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070073
74 return req;
75}
76
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070077static inline void __reqsk_free(struct request_sock *req)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070078{
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070079 kmem_cache_free(req->rsk_ops->slab, req);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070080}
81
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070082static inline void reqsk_free(struct request_sock *req)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070083{
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -070084 req->rsk_ops->destructor(req);
85 __reqsk_free(req);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070086}
87
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070088extern int sysctl_max_syn_backlog;
89
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -070090/** struct listen_sock - listen state
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070091 *
92 * @max_qlen_log - log_2 of maximal queued SYNs/REQUESTs
93 */
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -070094struct listen_sock {
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070095 u8 max_qlen_log;
Eric Dumazet946cedc2011-08-30 03:21:44 +000096 u8 synflood_warned;
97 /* 2 bytes hole, try to use */
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -070098 int qlen;
99 int qlen_young;
100 int clock_hand;
101 u32 hash_rnd;
Arnaldo Carvalho de Melo83e36092005-08-09 19:33:31 -0700102 u32 nr_table_entries;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700103 struct request_sock *syn_table[0];
104};
105
Jerry Chu10467162012-08-31 12:29:11 +0000106/*
107 * For a TCP Fast Open listener -
108 * lock - protects the access to all the reqsk, which is co-owned by
109 * the listener and the child socket.
110 * qlen - pending TFO requests (still in TCP_SYN_RECV).
111 * max_qlen - max TFO reqs allowed before TFO is disabled.
112 *
113 * XXX (TFO) - ideally these fields can be made as part of "listen_sock"
114 * structure above. But there is some implementation difficulty due to
115 * listen_sock being part of request_sock_queue hence will be freed when
116 * a listener is stopped. But TFO related fields may continue to be
117 * accessed even after a listener is closed, until its sk_refcnt drops
118 * to 0 implying no more outstanding TFO reqs. One solution is to keep
119 * listen_opt around until sk_refcnt drops to 0. But there is some other
120 * complexity that needs to be resolved. E.g., a listener can be disabled
121 * temporarily through shutdown()->tcp_disconnect(), and re-enabled later.
122 */
123struct fastopen_queue {
124 struct request_sock *rskq_rst_head; /* Keep track of past TFO */
125 struct request_sock *rskq_rst_tail; /* requests that caused RST.
126 * This is part of the defense
127 * against spoofing attack.
128 */
129 spinlock_t lock;
130 int qlen; /* # of pending (TCP_SYN_RECV) reqs */
131 int max_qlen; /* != 0 iff TFO is currently enabled */
132};
133
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700134/** struct request_sock_queue - queue of request_socks
135 *
136 * @rskq_accept_head - FIFO head of established children
137 * @rskq_accept_tail - FIFO tail of established children
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700138 * @rskq_defer_accept - User waits for some data after accept()
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700139 * @syn_wait_lock - serializer
140 *
141 * %syn_wait_lock is necessary only to avoid proc interface having to grab the main
142 * lock sock while browsing the listening hash (otherwise it's deadlock prone).
143 *
144 * This lock is acquired in read mode only from listening_get_next() seq_file
145 * op and it's acquired in write mode _only_ from code that is actively
146 * changing rskq_accept_head. All readers that are holding the master sock lock
147 * don't need to grab this lock in read mode too as rskq_accept_head. writes
148 * are always protected from the main sock lock.
149 */
150struct request_sock_queue {
151 struct request_sock *rskq_accept_head;
152 struct request_sock *rskq_accept_tail;
153 rwlock_t syn_wait_lock;
David S. Millerec0a1962008-06-12 16:31:35 -0700154 u8 rskq_defer_accept;
155 /* 3 bytes hole, try to pack */
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -0700156 struct listen_sock *listen_opt;
Jerry Chu10467162012-08-31 12:29:11 +0000157 struct fastopen_queue *fastopenq; /* This is non-NULL iff TFO has been
158 * enabled on this listener. Check
159 * max_qlen != 0 in fastopen_queue
160 * to determine if TFO is enabled
161 * right at this moment.
162 */
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700163};
164
165extern int reqsk_queue_alloc(struct request_sock_queue *queue,
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800166 unsigned int nr_table_entries);
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700167
Pavel Emelyanovdab6ba32007-11-15 02:57:06 -0800168extern void __reqsk_queue_destroy(struct request_sock_queue *queue);
Arnaldo Carvalho de Melo83e36092005-08-09 19:33:31 -0700169extern void reqsk_queue_destroy(struct request_sock_queue *queue);
Jerry Chu10467162012-08-31 12:29:11 +0000170extern void reqsk_fastopen_remove(struct sock *sk,
171 struct request_sock *req, bool reset);
Arnaldo Carvalho de Melo83e36092005-08-09 19:33:31 -0700172
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700173static inline struct request_sock *
174 reqsk_queue_yank_acceptq(struct request_sock_queue *queue)
175{
176 struct request_sock *req = queue->rskq_accept_head;
177
Norbert Kiesel3eb48012006-03-26 17:39:55 -0800178 queue->rskq_accept_head = NULL;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700179 return req;
180}
181
182static inline int reqsk_queue_empty(struct request_sock_queue *queue)
183{
184 return queue->rskq_accept_head == NULL;
185}
186
187static inline void reqsk_queue_unlink(struct request_sock_queue *queue,
188 struct request_sock *req,
189 struct request_sock **prev_req)
190{
191 write_lock(&queue->syn_wait_lock);
192 *prev_req = req->dl_next;
193 write_unlock(&queue->syn_wait_lock);
194}
195
196static inline void reqsk_queue_add(struct request_sock_queue *queue,
197 struct request_sock *req,
198 struct sock *parent,
199 struct sock *child)
200{
201 req->sk = child;
202 sk_acceptq_added(parent);
203
204 if (queue->rskq_accept_head == NULL)
205 queue->rskq_accept_head = req;
206 else
207 queue->rskq_accept_tail->dl_next = req;
208
209 queue->rskq_accept_tail = req;
210 req->dl_next = NULL;
211}
212
213static inline struct request_sock *reqsk_queue_remove(struct request_sock_queue *queue)
214{
215 struct request_sock *req = queue->rskq_accept_head;
216
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700217 WARN_ON(req == NULL);
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700218
219 queue->rskq_accept_head = req->dl_next;
220 if (queue->rskq_accept_head == NULL)
221 queue->rskq_accept_tail = NULL;
222
223 return req;
224}
225
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700226static inline int reqsk_queue_removed(struct request_sock_queue *queue,
227 struct request_sock *req)
228{
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -0700229 struct listen_sock *lopt = queue->listen_opt;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700230
Eric Dumazete6c022a2012-10-27 23:16:46 +0000231 if (req->num_timeout == 0)
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700232 --lopt->qlen_young;
233
234 return --lopt->qlen;
235}
236
237static inline int reqsk_queue_added(struct request_sock_queue *queue)
238{
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -0700239 struct listen_sock *lopt = queue->listen_opt;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700240 const int prev_qlen = lopt->qlen;
241
242 lopt->qlen_young++;
243 lopt->qlen++;
244 return prev_qlen;
245}
246
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700247static inline int reqsk_queue_len(const struct request_sock_queue *queue)
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700248{
249 return queue->listen_opt != NULL ? queue->listen_opt->qlen : 0;
250}
251
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700252static inline int reqsk_queue_len_young(const struct request_sock_queue *queue)
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700253{
254 return queue->listen_opt->qlen_young;
255}
256
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700257static inline int reqsk_queue_is_full(const struct request_sock_queue *queue)
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700258{
259 return queue->listen_opt->qlen >> queue->listen_opt->max_qlen_log;
260}
261
262static inline void reqsk_queue_hash_req(struct request_sock_queue *queue,
263 u32 hash, struct request_sock *req,
Arnaldo Carvalho de Melo81297652005-12-13 23:15:24 -0800264 unsigned long timeout)
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700265{
Arnaldo Carvalho de Melo2ad69c52005-06-18 22:48:55 -0700266 struct listen_sock *lopt = queue->listen_opt;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700267
268 req->expires = jiffies + timeout;
Eric Dumazete6c022a2012-10-27 23:16:46 +0000269 req->num_retrans = 0;
270 req->num_timeout = 0;
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700271 req->sk = NULL;
272 req->dl_next = lopt->syn_table[hash];
273
274 write_lock(&queue->syn_wait_lock);
275 lopt->syn_table[hash] = req;
276 write_unlock(&queue->syn_wait_lock);
277}
278
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700279#endif /* _REQUEST_SOCK_H */