Commit | Line | Data |
---|---|---|
7c657876 ACM |
1 | /* |
2 | * net/dccp/minisocks.c | |
3 | * | |
4 | * An implementation of the DCCP protocol | |
5 | * Arnaldo Carvalho de Melo <acme@conectiva.com.br> | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
11 | */ | |
12 | ||
7c657876 | 13 | #include <linux/dccp.h> |
8a73cd09 | 14 | #include <linux/kernel.h> |
7c657876 ACM |
15 | #include <linux/skbuff.h> |
16 | #include <linux/timer.h> | |
17 | ||
18 | #include <net/sock.h> | |
19 | #include <net/xfrm.h> | |
20 | #include <net/inet_timewait_sock.h> | |
21 | ||
ae31c339 | 22 | #include "ackvec.h" |
7c657876 ACM |
23 | #include "ccid.h" |
24 | #include "dccp.h" | |
afe00251 | 25 | #include "feat.h" |
7c657876 | 26 | |
64cf1e5d ACM |
27 | struct inet_timewait_death_row dccp_death_row = { |
28 | .sysctl_max_tw_buckets = NR_FILE * 2, | |
29 | .period = DCCP_TIMEWAIT_LEN / INET_TWDR_TWKILL_SLOTS, | |
30 | .death_lock = SPIN_LOCK_UNLOCKED, | |
31 | .hashinfo = &dccp_hashinfo, | |
32 | .tw_timer = TIMER_INITIALIZER(inet_twdr_hangman, 0, | |
33 | (unsigned long)&dccp_death_row), | |
34 | .twkill_work = __WORK_INITIALIZER(dccp_death_row.twkill_work, | |
35 | inet_twdr_twkill_work, | |
36 | &dccp_death_row), | |
37 | /* Short-time timewait calendar */ | |
38 | ||
39 | .twcal_hand = -1, | |
40 | .twcal_timer = TIMER_INITIALIZER(inet_twdr_twcal_tick, 0, | |
41 | (unsigned long)&dccp_death_row), | |
42 | }; | |
43 | ||
f21e68ca ACM |
44 | EXPORT_SYMBOL_GPL(dccp_death_row); |
45 | ||
7c657876 ACM |
46 | void dccp_time_wait(struct sock *sk, int state, int timeo) |
47 | { | |
64cf1e5d | 48 | struct inet_timewait_sock *tw = NULL; |
7c657876 | 49 | |
64cf1e5d ACM |
50 | if (dccp_death_row.tw_count < dccp_death_row.sysctl_max_tw_buckets) |
51 | tw = inet_twsk_alloc(sk, state); | |
52 | ||
53 | if (tw != NULL) { | |
54 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
55 | const int rto = (icsk->icsk_rto << 2) - (icsk->icsk_rto >> 1); | |
3df80d93 ACM |
56 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) |
57 | if (tw->tw_family == PF_INET6) { | |
58 | const struct ipv6_pinfo *np = inet6_sk(sk); | |
59 | struct inet6_timewait_sock *tw6; | |
64cf1e5d | 60 | |
3df80d93 ACM |
61 | tw->tw_ipv6_offset = inet6_tw_offset(sk->sk_prot); |
62 | tw6 = inet6_twsk((struct sock *)tw); | |
63 | ipv6_addr_copy(&tw6->tw_v6_daddr, &np->daddr); | |
64 | ipv6_addr_copy(&tw6->tw_v6_rcv_saddr, &np->rcv_saddr); | |
65 | tw->tw_ipv6only = np->ipv6only; | |
66 | } | |
67 | #endif | |
64cf1e5d ACM |
68 | /* Linkage updates. */ |
69 | __inet_twsk_hashdance(tw, sk, &dccp_hashinfo); | |
70 | ||
71 | /* Get the TIME_WAIT timeout firing. */ | |
72 | if (timeo < rto) | |
73 | timeo = rto; | |
74 | ||
75 | tw->tw_timeout = DCCP_TIMEWAIT_LEN; | |
76 | if (state == DCCP_TIME_WAIT) | |
77 | timeo = DCCP_TIMEWAIT_LEN; | |
78 | ||
79 | inet_twsk_schedule(tw, &dccp_death_row, timeo, | |
80 | DCCP_TIMEWAIT_LEN); | |
81 | inet_twsk_put(tw); | |
82 | } else { | |
83 | /* Sorry, if we're out of memory, just CLOSE this | |
84 | * socket up. We've got bigger problems than | |
85 | * non-graceful socket closings. | |
86 | */ | |
59348b19 | 87 | DCCP_WARN("time wait bucket table overflow\n"); |
64cf1e5d ACM |
88 | } |
89 | ||
90 | dccp_done(sk); | |
7c657876 ACM |
91 | } |
92 | ||
93 | struct sock *dccp_create_openreq_child(struct sock *sk, | |
94 | const struct request_sock *req, | |
95 | const struct sk_buff *skb) | |
96 | { | |
97 | /* | |
98 | * Step 3: Process LISTEN state | |
99 | * | |
d83ca5ac GR |
100 | * (* Generate a new socket and switch to that socket *) |
101 | * Set S := new socket for this port pair | |
7c657876 ACM |
102 | */ |
103 | struct sock *newsk = inet_csk_clone(sk, req, GFP_ATOMIC); | |
104 | ||
105 | if (newsk != NULL) { | |
106 | const struct dccp_request_sock *dreq = dccp_rsk(req); | |
107 | struct inet_connection_sock *newicsk = inet_csk(sk); | |
108 | struct dccp_sock *newdp = dccp_sk(newsk); | |
a4bf3902 | 109 | struct dccp_minisock *newdmsk = dccp_msk(newsk); |
7c657876 | 110 | |
67e6b629 | 111 | newdp->dccps_role = DCCP_ROLE_SERVER; |
ae31c339 | 112 | newdp->dccps_hc_rx_ackvec = NULL; |
67e6b629 ACM |
113 | newdp->dccps_service_list = NULL; |
114 | newdp->dccps_service = dreq->dreq_service; | |
115 | newicsk->icsk_rto = DCCP_TIMEOUT_INIT; | |
b0e56780 | 116 | do_gettimeofday(&newdp->dccps_epoch); |
7c657876 | 117 | |
afe00251 AB |
118 | if (dccp_feat_clone(sk, newsk)) |
119 | goto out_free; | |
120 | ||
a4bf3902 | 121 | if (newdmsk->dccpms_send_ack_vector) { |
ae31c339 | 122 | newdp->dccps_hc_rx_ackvec = |
7400d781 | 123 | dccp_ackvec_alloc(GFP_ATOMIC); |
ae31c339 | 124 | if (unlikely(newdp->dccps_hc_rx_ackvec == NULL)) |
7c657876 ACM |
125 | goto out_free; |
126 | } | |
127 | ||
91f0ebf7 | 128 | newdp->dccps_hc_rx_ccid = |
a4bf3902 | 129 | ccid_hc_rx_new(newdmsk->dccpms_rx_ccid, |
91f0ebf7 ACM |
130 | newsk, GFP_ATOMIC); |
131 | newdp->dccps_hc_tx_ccid = | |
a4bf3902 | 132 | ccid_hc_tx_new(newdmsk->dccpms_tx_ccid, |
91f0ebf7 ACM |
133 | newsk, GFP_ATOMIC); |
134 | if (unlikely(newdp->dccps_hc_rx_ccid == NULL || | |
135 | newdp->dccps_hc_tx_ccid == NULL)) { | |
ae31c339 | 136 | dccp_ackvec_free(newdp->dccps_hc_rx_ackvec); |
91f0ebf7 ACM |
137 | ccid_hc_rx_delete(newdp->dccps_hc_rx_ccid, newsk); |
138 | ccid_hc_tx_delete(newdp->dccps_hc_tx_ccid, newsk); | |
7c657876 ACM |
139 | out_free: |
140 | /* It is still raw copy of parent, so invalidate | |
141 | * destructor and make plain sk_free() */ | |
142 | newsk->sk_destruct = NULL; | |
143 | sk_free(newsk); | |
144 | return NULL; | |
145 | } | |
146 | ||
7c657876 ACM |
147 | /* |
148 | * Step 3: Process LISTEN state | |
149 | * | |
d83ca5ac GR |
150 | * Choose S.ISS (initial seqno) or set from Init Cookies |
151 | * Initialize S.GAR := S.ISS | |
152 | * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies | |
7c657876 ACM |
153 | */ |
154 | ||
155 | /* See dccp_v4_conn_request */ | |
a4bf3902 | 156 | newdmsk->dccpms_sequence_window = req->rcv_wnd; |
7c657876 ACM |
157 | |
158 | newdp->dccps_gar = newdp->dccps_isr = dreq->dreq_isr; | |
159 | dccp_update_gsr(newsk, dreq->dreq_isr); | |
160 | ||
161 | newdp->dccps_iss = dreq->dreq_iss; | |
162 | dccp_update_gss(newsk, dreq->dreq_iss); | |
163 | ||
03ace394 ACM |
164 | /* |
165 | * SWL and AWL are initially adjusted so that they are not less than | |
166 | * the initial Sequence Numbers received and sent, respectively: | |
167 | * SWL := max(GSR + 1 - floor(W/4), ISR), | |
168 | * AWL := max(GSS - W' + 1, ISS). | |
169 | * These adjustments MUST be applied only at the beginning of the | |
170 | * connection. | |
171 | */ | |
172 | dccp_set_seqno(&newdp->dccps_swl, | |
173 | max48(newdp->dccps_swl, newdp->dccps_isr)); | |
174 | dccp_set_seqno(&newdp->dccps_awl, | |
175 | max48(newdp->dccps_awl, newdp->dccps_iss)); | |
176 | ||
7c657876 ACM |
177 | dccp_init_xmit_timers(newsk); |
178 | ||
179 | DCCP_INC_STATS_BH(DCCP_MIB_PASSIVEOPENS); | |
180 | } | |
181 | return newsk; | |
182 | } | |
183 | ||
f21e68ca ACM |
184 | EXPORT_SYMBOL_GPL(dccp_create_openreq_child); |
185 | ||
7c657876 ACM |
186 | /* |
187 | * Process an incoming packet for RESPOND sockets represented | |
188 | * as an request_sock. | |
189 | */ | |
190 | struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb, | |
191 | struct request_sock *req, | |
192 | struct request_sock **prev) | |
193 | { | |
194 | struct sock *child = NULL; | |
195 | ||
196 | /* Check for retransmitted REQUEST */ | |
197 | if (dccp_hdr(skb)->dccph_type == DCCP_PKT_REQUEST) { | |
e11d9d30 | 198 | struct dccp_request_sock *dreq = dccp_rsk(req); |
7c657876 | 199 | |
e11d9d30 | 200 | if (after48(DCCP_SKB_CB(skb)->dccpd_seq, dreq->dreq_isr)) { |
7c657876 | 201 | dccp_pr_debug("Retransmitted REQUEST\n"); |
e11d9d30 GR |
202 | dreq->dreq_isr = DCCP_SKB_CB(skb)->dccpd_seq; |
203 | /* | |
204 | * Send another RESPONSE packet | |
205 | * To protect against Request floods, increment retrans | |
206 | * counter (backoff, monitored by dccp_response_timer). | |
207 | */ | |
208 | req->retrans++; | |
7c657876 ACM |
209 | req->rsk_ops->rtx_syn_ack(sk, req, NULL); |
210 | } | |
211 | /* Network Duplicate, discard packet */ | |
212 | return NULL; | |
213 | } | |
214 | ||
215 | DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_PACKET_ERROR; | |
216 | ||
217 | if (dccp_hdr(skb)->dccph_type != DCCP_PKT_ACK && | |
218 | dccp_hdr(skb)->dccph_type != DCCP_PKT_DATAACK) | |
219 | goto drop; | |
220 | ||
221 | /* Invalid ACK */ | |
222 | if (DCCP_SKB_CB(skb)->dccpd_ack_seq != dccp_rsk(req)->dreq_iss) { | |
7690af3f ACM |
223 | dccp_pr_debug("Invalid ACK number: ack_seq=%llu, " |
224 | "dreq_iss=%llu\n", | |
f6ccf554 DM |
225 | (unsigned long long) |
226 | DCCP_SKB_CB(skb)->dccpd_ack_seq, | |
227 | (unsigned long long) | |
228 | dccp_rsk(req)->dreq_iss); | |
7c657876 ACM |
229 | goto drop; |
230 | } | |
231 | ||
57cca05a | 232 | child = inet_csk(sk)->icsk_af_ops->syn_recv_sock(sk, skb, req, NULL); |
7c657876 ACM |
233 | if (child == NULL) |
234 | goto listen_overflow; | |
235 | ||
236 | /* FIXME: deal with options */ | |
237 | ||
238 | inet_csk_reqsk_queue_unlink(sk, req, prev); | |
239 | inet_csk_reqsk_queue_removed(sk, req); | |
240 | inet_csk_reqsk_queue_add(sk, req, child); | |
241 | out: | |
242 | return child; | |
243 | listen_overflow: | |
244 | dccp_pr_debug("listen_overflow!\n"); | |
245 | DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY; | |
246 | drop: | |
247 | if (dccp_hdr(skb)->dccph_type != DCCP_PKT_RESET) | |
cfb6eeb4 | 248 | req->rsk_ops->send_reset(sk, skb); |
7c657876 ACM |
249 | |
250 | inet_csk_reqsk_queue_drop(sk, req, prev); | |
251 | goto out; | |
252 | } | |
253 | ||
f21e68ca ACM |
254 | EXPORT_SYMBOL_GPL(dccp_check_req); |
255 | ||
7c657876 ACM |
256 | /* |
257 | * Queue segment on the new socket if the new socket is active, | |
258 | * otherwise we just shortcircuit this and continue with | |
259 | * the new socket. | |
260 | */ | |
261 | int dccp_child_process(struct sock *parent, struct sock *child, | |
262 | struct sk_buff *skb) | |
263 | { | |
264 | int ret = 0; | |
265 | const int state = child->sk_state; | |
266 | ||
267 | if (!sock_owned_by_user(child)) { | |
7690af3f ACM |
268 | ret = dccp_rcv_state_process(child, skb, dccp_hdr(skb), |
269 | skb->len); | |
7c657876 ACM |
270 | |
271 | /* Wakeup parent, send SIGIO */ | |
272 | if (state == DCCP_RESPOND && child->sk_state != state) | |
273 | parent->sk_data_ready(parent, 0); | |
274 | } else { | |
275 | /* Alas, it is possible again, because we do lookup | |
276 | * in main socket hash table and lock on listening | |
277 | * socket does not protect us more. | |
278 | */ | |
279 | sk_add_backlog(child, skb); | |
280 | } | |
281 | ||
282 | bh_unlock_sock(child); | |
283 | sock_put(child); | |
284 | return ret; | |
285 | } | |
f21e68ca ACM |
286 | |
287 | EXPORT_SYMBOL_GPL(dccp_child_process); | |
8a73cd09 GR |
288 | |
289 | void dccp_reqsk_send_ack(struct sk_buff *skb, struct request_sock *rsk) | |
290 | { | |
59348b19 | 291 | DCCP_BUG("DCCP-ACK packets are never sent in LISTEN/RESPOND state"); |
8a73cd09 GR |
292 | } |
293 | ||
294 | EXPORT_SYMBOL_GPL(dccp_reqsk_send_ack); | |
cf557926 GR |
295 | |
296 | void dccp_reqsk_init(struct request_sock *req, struct sk_buff *skb) | |
297 | { | |
298 | inet_rsk(req)->rmt_port = dccp_hdr(skb)->dccph_sport; | |
299 | inet_rsk(req)->acked = 0; | |
afb0a34d | 300 | req->rcv_wnd = sysctl_dccp_feat_sequence_window; |
cf557926 GR |
301 | } |
302 | ||
303 | EXPORT_SYMBOL_GPL(dccp_reqsk_init); |