blob: 403b369c68f3d911f05152e4d373ad643f5a3ae0 [file] [log] [blame]
Dave Hudsondcd08fb2013-11-22 09:25:16 -06001/*
2 * sfe-cm.c
3 * Shortcut forwarding engine connection manager.
4 *
Matthew McClintocka3221942014-01-16 11:44:26 -06005 * Copyright (c) 2013 Qualcomm Atheros, Inc.
6 *
7 * All Rights Reserved.
8 * Qualcomm Atheros Confidential and Proprietary.
Dave Hudsondcd08fb2013-11-22 09:25:16 -06009 */
Matthew McClintocka3221942014-01-16 11:44:26 -060010
Dave Hudsondcd08fb2013-11-22 09:25:16 -060011#include <linux/module.h>
12#include <linux/sysfs.h>
13#include <linux/skbuff.h>
14#include <net/route.h>
15#include <linux/inetdevice.h>
16#include <linux/netfilter_bridge.h>
17#include <net/netfilter/nf_conntrack_acct.h>
18#include <net/netfilter/nf_conntrack_helper.h>
19#include <net/netfilter/nf_conntrack_zones.h>
20#include <net/netfilter/nf_conntrack_core.h>
Matthew McClintockbf6b5bc2014-02-24 12:27:14 -060021#include <linux/if_bridge.h>
Dave Hudsondcd08fb2013-11-22 09:25:16 -060022
23#include "sfe.h"
24#include "sfe_ipv4.h"
25
26/*
27 * Per-module structure.
28 */
29struct sfe_cm {
30 spinlock_t lock; /* Lock for SMP correctness */
31
32 /*
33 * Control state.
34 */
35 struct kobject *sys_sfe_cm; /* sysfs linkage */
36
37 /*
38 * Callback notifiers.
39 */
40 struct notifier_block dev_notifier;
41 /* Device notifier */
42 struct notifier_block inet_notifier;
43 /* IP notifier */
44};
45
46struct sfe_cm __sc;
47
48/*
49 * Expose the hook for the receive processing.
50 */
51extern int (*athrs_fast_nat_recv)(struct sk_buff *skb);
52
53/*
54 * Expose what should be a static flag in the TCP connection tracker.
55 */
56extern int nf_ct_tcp_no_window_check;
57
58/*
59 * sfe_cm_recv()
60 * Handle packet receives.
61 *
62 * Returns 1 if the packet is forwarded or 0 if it isn't.
63 */
64int sfe_cm_recv(struct sk_buff *skb)
65{
66 struct net_device *dev;
67#if (SFE_HOOK_ABOVE_BRIDGE)
68 struct in_device *in_dev;
69#endif
70
71 /*
72 * We know that for the vast majority of packets we need the transport
73 * layer header so we may as well start to fetch it now!
74 */
75 prefetch(skb->data + 32);
76 barrier();
77
78 dev = skb->dev;
79
80#if (SFE_HOOK_ABOVE_BRIDGE)
81 /*
82 * Does our input device support IP processing?
83 */
84 in_dev = (struct in_device *)dev->ip_ptr;
85 if (unlikely(!in_dev)) {
86 DEBUG_TRACE("no IP processing for device: %s\n", dev->name);
87 return 0;
88 }
89
90 /*
91 * Does it have an IP address? If it doesn't then we can't do anything
92 * interesting here!
93 */
94 if (unlikely(!in_dev->ifa_list)) {
95 DEBUG_TRACE("no IP address for device: %s\n", dev->name);
96 return 0;
97 }
98#endif
99
100 /*
101 * We're only interested in IP packets.
102 */
103 if (likely(htons(ETH_P_IP) == skb->protocol)) {
104 return sfe_ipv4_recv(dev, skb);
105 }
106
Matthew McClintocka8ad7962014-01-16 16:49:30 -0600107 DEBUG_TRACE("not IP packet\n");
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600108 return 0;
109}
110
111/*
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600112 * sfe_cm_find_dev_and_mac_addr()
113 * Find the device and MAC address for a given IPv4 address.
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600114 *
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600115 * Returns true if we find the device and MAC address, otherwise false.
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600116 *
117 * We look up the rtable entry for the address and, from its neighbour
118 * structure, obtain the hardware address. This means this function also
119 * works if the neighbours are routers too.
120 */
Matthew McClintock6508efb2014-03-11 15:50:58 -0500121static bool sfe_cm_find_dev_and_mac_addr(uint32_t daddr, uint32_t saddr, struct net_device **dev, uint8_t *mac_addr)
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600122{
123 struct neighbour *neigh;
124 struct rtable *rt;
125 struct dst_entry *dst;
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600126 struct net_device *mac_dev;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600127
128 /*
129 * Look up the rtable entry for the IP address then get the hardware
130 * address from its neighbour structure. This means this work when the
131 * neighbours are routers too.
132 */
Matthew McClintock6508efb2014-03-11 15:50:58 -0500133 rt = ip_route_output(&init_net, daddr, saddr, 0, 0);
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600134 if (unlikely(IS_ERR(rt))) {
135 return false;
136 }
137
138 dst = (struct dst_entry *)rt;
139
140 rcu_read_lock();
141 neigh = dst_get_neighbour_noref(dst);
142 if (unlikely(!neigh)) {
143 rcu_read_unlock();
144 dst_release(dst);
145 return false;
146 }
147
148 if (unlikely(!(neigh->nud_state & NUD_VALID))) {
149 rcu_read_unlock();
150 dst_release(dst);
151 return false;
152 }
153
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600154 mac_dev = neigh->dev;
155 if (!mac_dev) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600156 rcu_read_unlock();
157 dst_release(dst);
158 return false;
159 }
160
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600161 memcpy(mac_addr, neigh->ha, (size_t)mac_dev->addr_len);
162
163 dev_hold(mac_dev);
164 *dev = mac_dev;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600165 rcu_read_unlock();
166
167 dst_release(dst);
168
169 /*
170 * We're only interested in unicast MAC addresses - if it's not a unicast
171 * address then our IP address mustn't be unicast either.
172 */
173 if (is_multicast_ether_addr(mac_addr)) {
174 DEBUG_TRACE("MAC is non-unicast - ignoring\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600175 dev_put(mac_dev);
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600176 return false;
177 }
178
179 return true;
180}
181
182/*
183 * sfe_cm_ipv4_post_routing_hook()
184 * Called for packets about to leave the box - either locally generated or forwarded from another interface
185 */
186static unsigned int sfe_cm_ipv4_post_routing_hook(unsigned int hooknum,
187 struct sk_buff *skb,
188 const struct net_device *in_unused,
189 const struct net_device *out,
190 int (*okfn)(struct sk_buff *))
191{
192 struct sfe_ipv4_create sic;
193 struct net_device *in;
194 struct nf_conn *ct;
195 enum ip_conntrack_info ctinfo;
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600196 struct net_device *dev;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600197 struct net_device *src_dev;
198 struct net_device *dest_dev;
199 struct net_device *src_br_dev = NULL;
200 struct net_device *dest_br_dev = NULL;
201 struct nf_conntrack_tuple orig_tuple;
202 struct nf_conntrack_tuple reply_tuple;
203
204 /*
205 * Don't process broadcast or multicast packets.
206 */
207 if (unlikely(skb->pkt_type == PACKET_BROADCAST)) {
208 DEBUG_TRACE("broadcast, ignoring\n");
209 return NF_ACCEPT;
210 }
211 if (unlikely(skb->pkt_type == PACKET_MULTICAST)) {
212 DEBUG_TRACE("multicast, ignoring\n");
213 return NF_ACCEPT;
214 }
215
216 /*
217 * Don't process packets that are not being forwarded.
218 */
219 in = dev_get_by_index(&init_net, skb->skb_iif);
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600220 if (!in) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600221 DEBUG_TRACE("packet not forwarding\n");
222 return NF_ACCEPT;
223 }
224
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600225 dev_put(in);
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600226
227 /*
228 * Don't process packets that aren't being tracked by conntrack.
229 */
230 ct = nf_ct_get(skb, &ctinfo);
231 if (unlikely(!ct)) {
232 DEBUG_TRACE("no conntrack connection, ignoring\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600233 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600234 }
235
236 /*
237 * Don't process untracked connections.
238 */
239 if (unlikely(ct == &nf_conntrack_untracked)) {
240 DEBUG_TRACE("untracked connection\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600241 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600242 }
243
244 /*
245 * Don't process connections that require support from a 'helper' (typically a NAT ALG).
246 */
247 if (unlikely(nfct_help(ct))) {
248 DEBUG_TRACE("connection has helper\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600249 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600250 }
251
252 /*
253 * Look up the details of our connection in conntrack.
254 *
255 * Note that the data we get from conntrack is for the "ORIGINAL" direction
256 * but our packet may actually be in the "REPLY" direction.
257 */
258 orig_tuple = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
259 reply_tuple = ct->tuplehash[IP_CT_DIR_REPLY].tuple;
260 sic.protocol = (int32_t)orig_tuple.dst.protonum;
261
262 /*
263 * Get addressing information, non-NAT first
264 */
265 sic.src_ip = (__be32)orig_tuple.src.u3.ip;
266 sic.dest_ip = (__be32)orig_tuple.dst.u3.ip;
267
268 /*
269 * NAT'ed addresses - note these are as seen from the 'reply' direction
270 * When NAT does not apply to this connection these will be identical to the above.
271 */
272 sic.src_ip_xlate = (__be32)reply_tuple.dst.u3.ip;
273 sic.dest_ip_xlate = (__be32)reply_tuple.src.u3.ip;
274
275 sic.flags = 0;
276
277 switch (sic.protocol) {
278 case IPPROTO_TCP:
279 sic.src_port = orig_tuple.src.u.tcp.port;
280 sic.dest_port = orig_tuple.dst.u.tcp.port;
281 sic.src_port_xlate = reply_tuple.dst.u.tcp.port;
282 sic.dest_port_xlate = reply_tuple.src.u.tcp.port;
283 sic.src_td_window_scale = ct->proto.tcp.seen[0].td_scale;
284 sic.src_td_max_window = ct->proto.tcp.seen[0].td_maxwin;
285 sic.src_td_end = ct->proto.tcp.seen[0].td_end;
286 sic.src_td_max_end = ct->proto.tcp.seen[0].td_maxend;
287 sic.dest_td_window_scale = ct->proto.tcp.seen[1].td_scale;
288 sic.dest_td_max_window = ct->proto.tcp.seen[1].td_maxwin;
289 sic.dest_td_end = ct->proto.tcp.seen[1].td_end;
290 sic.dest_td_max_end = ct->proto.tcp.seen[1].td_maxend;
291 if (nf_ct_tcp_no_window_check
292 || (ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_BE_LIBERAL)
293 || (ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_BE_LIBERAL)) {
294 sic.flags |= SFE_IPV4_CREATE_FLAG_NO_SEQ_CHECK;
295 }
296
297 /*
298 * Don't try to manage a non-established connection.
299 */
300 if (!test_bit(IPS_ASSURED_BIT, &ct->status)) {
301 DEBUG_TRACE("non-established connection\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600302 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600303 }
304
305 /*
306 * If the connection is shutting down do not manage it.
307 * state can not be SYN_SENT, SYN_RECV because connection is assured
308 * Not managed states: FIN_WAIT, CLOSE_WAIT, LAST_ACK, TIME_WAIT, CLOSE.
309 */
310 spin_lock_bh(&ct->lock);
311 if (ct->proto.tcp.state != TCP_CONNTRACK_ESTABLISHED) {
312 spin_unlock_bh(&ct->lock);
313 DEBUG_TRACE("connection in termination state: %#x, s: %pI4:%u, d: %pI4:%u\n",
314 ct->proto.tcp.state, &sic.src_ip, ntohs(sic.src_port),
315 &sic.dest_ip, ntohs(sic.dest_port));
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600316 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600317 }
318 spin_unlock_bh(&ct->lock);
319 break;
320
321 case IPPROTO_UDP:
322 sic.src_port = orig_tuple.src.u.udp.port;
323 sic.dest_port = orig_tuple.dst.u.udp.port;
324 sic.src_port_xlate = reply_tuple.dst.u.udp.port;
325 sic.dest_port_xlate = reply_tuple.src.u.udp.port;
326 break;
327
328 default:
329 DEBUG_TRACE("unhandled protocol %d\n", sic.protocol);
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600330 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600331 }
332
333 /*
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600334 * Get the net device and MAC addresses that correspond to the various source and
335 * destination host addresses.
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600336 */
Matthew McClintock6508efb2014-03-11 15:50:58 -0500337 if (!sfe_cm_find_dev_and_mac_addr(sic.src_ip, 0, &src_dev, sic.src_mac)) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600338 DEBUG_TRACE("failed to find MAC address for src IP: %pI4\n", &sic.src_ip);
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600339 return NF_ACCEPT;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600340 }
341
Matthew McClintock6508efb2014-03-11 15:50:58 -0500342 if (!sfe_cm_find_dev_and_mac_addr(sic.src_ip_xlate, 0, &dev, sic.src_mac_xlate)) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600343 DEBUG_TRACE("failed to find MAC address for xlate src IP: %pI4\n", &sic.src_ip_xlate);
344 goto done1;
345 }
346
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600347 dev_put(dev);
348
Matthew McClintock6508efb2014-03-11 15:50:58 -0500349 if (!sfe_cm_find_dev_and_mac_addr(sic.dest_ip, 0, &dev, sic.dest_mac)) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600350 DEBUG_TRACE("failed to find MAC address for dest IP: %pI4\n", &sic.dest_ip);
351 goto done1;
352 }
353
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600354 dev_put(dev);
355
Matthew McClintock6508efb2014-03-11 15:50:58 -0500356 if (!sfe_cm_find_dev_and_mac_addr(sic.dest_ip_xlate, sic.src_ip_xlate, &dest_dev, sic.dest_mac_xlate)) {
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600357 DEBUG_TRACE("failed to find MAC address for xlate dest IP: %pI4\n", &sic.dest_ip_xlate);
358 goto done1;
359 }
360
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600361#if (!SFE_HOOK_ABOVE_BRIDGE)
362 /*
363 * Now our devices may actually be a bridge interface. If that's
364 * the case then we need to hunt down the underlying interface.
365 */
366 if (src_dev->priv_flags & IFF_EBRIDGE) {
367 src_br_dev = br_port_dev_get(src_dev, sic.src_mac);
368 if (!src_br_dev) {
369 DEBUG_TRACE("no port found on bridge\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600370 goto done2;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600371 }
372
373 src_dev = src_br_dev;
374 }
375
376 if (dest_dev->priv_flags & IFF_EBRIDGE) {
377 dest_br_dev = br_port_dev_get(dest_dev, sic.dest_mac_xlate);
378 if (!dest_br_dev) {
379 DEBUG_TRACE("no port found on bridge\n");
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600380 goto done3;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600381 }
382
383 dest_dev = dest_br_dev;
384 }
385#else
386 /*
387 * Our devices may actually be part of a bridge interface. If that's
388 * the case then find the bridge interface instead.
389 */
390 if (src_dev->priv_flags & IFF_BRIDGE_PORT) {
391 src_br_dev = src_dev->master;
392 if (!src_br_dev) {
393 DEBUG_TRACE("no bridge found for: %s\n", src_dev->name);
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600394 goto done2;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600395 }
396
397 dev_hold(src_br_dev);
398 src_dev = src_br_dev;
399 }
400
401 if (dest_dev->priv_flags & IFF_BRIDGE_PORT) {
402 dest_br_dev = dest_dev->master;
403 if (!dest_br_dev) {
404 DEBUG_TRACE("no bridge found for: %s\n", dest_dev->name);
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600405 goto done3;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600406 }
407
408 dev_hold(dest_br_dev);
409 dest_dev = dest_br_dev;
410 }
411#endif
412
413 sic.src_dev = src_dev;
414 sic.dest_dev = dest_dev;
415
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600416 sic.src_mtu = src_dev->mtu;
417 sic.dest_mtu = dest_dev->mtu;
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600418
419 sfe_ipv4_create_rule(&sic);
420
421 /*
422 * If we had bridge ports then release them too.
423 */
424 if (dest_br_dev) {
425 dev_put(dest_br_dev);
426 }
427
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600428done3:
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600429 if (src_br_dev) {
430 dev_put(src_br_dev);
431 }
432
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600433done2:
434 dev_put(dest_dev);
435
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600436done1:
Matthew McClintockdb5ac512014-01-16 17:01:40 -0600437 dev_put(src_dev);
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600438
439 return NF_ACCEPT;
440}
441
442#ifdef CONFIG_NF_CONNTRACK_EVENTS
443/*
444 * sfe_cm_conntrack_event()
445 * Callback event invoked when a conntrack connection's state changes.
446 */
Matthew McClintock0680e9f2013-11-26 15:43:10 -0600447#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS
448static int sfe_cm_conntrack_event(struct notifier_block *this,
Matthew McClintock0f5c0592014-02-12 11:17:11 -0600449 unsigned long events, void *ptr)
Matthew McClintock0680e9f2013-11-26 15:43:10 -0600450#else
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600451static int sfe_cm_conntrack_event(unsigned int events, struct nf_ct_event *item)
Matthew McClintock0680e9f2013-11-26 15:43:10 -0600452#endif
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600453{
Matthew McClintock0f5c0592014-02-12 11:17:11 -0600454#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS
455 struct nf_ct_event *item = ptr;
456#endif
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600457 struct sfe_ipv4_destroy sid;
458 struct nf_conn *ct = item->ct;
459 struct nf_conntrack_tuple orig_tuple;
460
461 /*
462 * If we don't have a conntrack entry then we're done.
463 */
464 if (unlikely(!ct)) {
465 DEBUG_WARN("no ct in conntrack event callback\n");
466 return NOTIFY_DONE;
467 }
468
469 /*
470 * If this is an untracked connection then we can't have any state either.
471 */
472 if (unlikely(ct == &nf_conntrack_untracked)) {
473 DEBUG_TRACE("ignoring untracked conn\n");
474 return NOTIFY_DONE;
475 }
476
477 /*
478 * Ignore anything other than IPv4 connections.
479 */
480 if (unlikely(nf_ct_l3num(ct) != AF_INET)) {
481 DEBUG_TRACE("ignoring non-IPv4 conn\n");
482 return NOTIFY_DONE;
483 }
484
485 /*
486 * We're only interested in destroy events.
487 */
488 if (unlikely(!(events & (1 << IPCT_DESTROY)))) {
489 DEBUG_TRACE("ignoring non-destroy event\n");
490 return NOTIFY_DONE;
491 }
492
493 orig_tuple = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
494 sid.protocol = (int32_t)orig_tuple.dst.protonum;
495
496 /*
497 * Extract information from the conntrack connection. We're only interested
498 * in nominal connection information (i.e. we're ignoring any NAT information).
499 */
500 sid.src_ip = (__be32)orig_tuple.src.u3.ip;
501 sid.dest_ip = (__be32)orig_tuple.dst.u3.ip;
502
503 switch (sid.protocol) {
504 case IPPROTO_TCP:
505 sid.src_port = orig_tuple.src.u.tcp.port;
506 sid.dest_port = orig_tuple.dst.u.tcp.port;
507 break;
508
509 case IPPROTO_UDP:
510 sid.src_port = orig_tuple.src.u.udp.port;
511 sid.dest_port = orig_tuple.dst.u.udp.port;
512 break;
513
514 default:
515 DEBUG_TRACE("unhandled protocol: %d\n", sid.protocol);
516 return NOTIFY_DONE;
517 }
518
519
520 sfe_ipv4_destroy_rule(&sid);
521 return NOTIFY_DONE;
522}
523
524/*
525 * Netfilter conntrack event system to monitor connection tracking changes
526 */
Matthew McClintock0680e9f2013-11-26 15:43:10 -0600527#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS
528static struct notifier_block sfe_cm_conntrack_notifier = {
529 .notifier_call = sfe_cm_conntrack_event,
530};
531#else
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600532static struct nf_ct_event_notifier sfe_cm_conntrack_notifier = {
533 .fcn = sfe_cm_conntrack_event,
534};
535#endif
Matthew McClintock0680e9f2013-11-26 15:43:10 -0600536#endif
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600537
538/*
539 * Structure to establish a hook into the post routing netfilter point - this
540 * will pick up local outbound and packets going from one interface to another.
541 *
542 * Note: see include/linux/netfilter_ipv4.h for info related to priority levels.
543 * We want to examine packets after NAT translation and any ALG processing.
544 */
545static struct nf_hook_ops sfe_cm_ipv4_ops_post_routing[] __read_mostly = {
546 {
547 .hook = sfe_cm_ipv4_post_routing_hook,
548 .owner = THIS_MODULE,
549 .pf = PF_INET,
550 .hooknum = NF_INET_POST_ROUTING,
551 .priority = NF_IP_PRI_NAT_SRC + 1,
552 },
553};
554
555/*
556 * sfe_cm_sync_rule()
557 * Synchronize a connection's state.
558 */
559static void sfe_cm_sync_rule(struct sfe_ipv4_sync *sis)
560{
561 struct nf_conntrack_tuple_hash *h;
562 struct nf_conntrack_tuple tuple;
563 struct nf_conn *ct;
564 struct nf_conn_counter *acct;
565
566 /*
567 * Create a tuple so as to be able to look up a connection
568 */
569 memset(&tuple, 0, sizeof(tuple));
570 tuple.src.u3.ip = sis->src_ip;
571 tuple.src.u.all = (__be16)sis->src_port;
572 tuple.src.l3num = AF_INET;
573
574 tuple.dst.u3.ip = sis->dest_ip;
575 tuple.dst.dir = IP_CT_DIR_ORIGINAL;
576 tuple.dst.protonum = (uint8_t)sis->protocol;
577 tuple.dst.u.all = (__be16)sis->dest_port;
578
579 DEBUG_TRACE("update connection - p: %d, s: %pI4:%u, d: %pI4:%u\n",
580 (int)tuple.dst.protonum,
581 &tuple.src.u3.ip, (unsigned int)ntohs(tuple.src.u.all),
582 &tuple.dst.u3.ip, (unsigned int)ntohs(tuple.dst.u.all));
583
584 /*
585 * Look up conntrack connection
586 */
587 h = nf_conntrack_find_get(&init_net, NF_CT_DEFAULT_ZONE, &tuple);
588 if (unlikely(!h)) {
589 DEBUG_TRACE("no connection found\n");
590 return;
591 }
592
593 ct = nf_ct_tuplehash_to_ctrack(h);
594 NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct);
595
596 /*
597 * Only update if this is not a fixed timeout
598 */
599 if (!test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) {
600 ct->timeout.expires += sis->delta_jiffies;
601 }
602
603 acct = nf_conn_acct_find(ct);
604 if (acct) {
605 spin_lock_bh(&ct->lock);
Matthew McClintock704b7a62013-12-19 16:13:01 -0600606 atomic64_set(&acct[IP_CT_DIR_ORIGINAL].packets, sis->src_packet_count);
607 atomic64_set(&acct[IP_CT_DIR_ORIGINAL].bytes, sis->src_byte_count);
608 atomic64_set(&acct[IP_CT_DIR_REPLY].packets, sis->dest_packet_count);
609 atomic64_set(&acct[IP_CT_DIR_REPLY].bytes, sis->dest_byte_count);
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600610 spin_unlock_bh(&ct->lock);
611 }
612
613 switch (sis->protocol) {
614 case IPPROTO_TCP:
615 spin_lock_bh(&ct->lock);
616 if (ct->proto.tcp.seen[0].td_maxwin < sis->src_td_max_window) {
617 ct->proto.tcp.seen[0].td_maxwin = sis->src_td_max_window;
618 }
619 if ((int32_t)(ct->proto.tcp.seen[0].td_end - sis->src_td_end) < 0) {
620 ct->proto.tcp.seen[0].td_end = sis->src_td_end;
621 }
622 if ((int32_t)(ct->proto.tcp.seen[0].td_maxend - sis->src_td_max_end) < 0) {
623 ct->proto.tcp.seen[0].td_maxend = sis->src_td_max_end;
624 }
625 if (ct->proto.tcp.seen[1].td_maxwin < sis->dest_td_max_window) {
626 ct->proto.tcp.seen[1].td_maxwin = sis->dest_td_max_window;
627 }
628 if ((int32_t)(ct->proto.tcp.seen[1].td_end - sis->dest_td_end) < 0) {
629 ct->proto.tcp.seen[1].td_end = sis->dest_td_end;
630 }
631 if ((int32_t)(ct->proto.tcp.seen[1].td_maxend - sis->dest_td_max_end) < 0) {
632 ct->proto.tcp.seen[1].td_maxend = sis->dest_td_max_end;
633 }
634 spin_unlock_bh(&ct->lock);
635 break;
636 }
637
638 /*
639 * Release connection
640 */
641 nf_ct_put(ct);
642}
643
644/*
645 * sfe_cm_device_event()
646 */
647static int sfe_cm_device_event(struct notifier_block *this, unsigned long event, void *ptr)
648{
649 struct net_device *dev = (struct net_device *)ptr;
650
651 switch (event) {
652 case NETDEV_DOWN:
653 if (dev) {
654 sfe_ipv4_destroy_all_rules_for_dev(dev);
655 }
656 break;
657 }
658
659 return NOTIFY_DONE;
660}
661
662/*
663 * sfe_cm_inet_event()
664 */
665static int sfe_cm_inet_event(struct notifier_block *this, unsigned long event, void *ptr)
666{
667 struct net_device *dev = ((struct in_ifaddr *)ptr)->ifa_dev->dev;
668 return sfe_cm_device_event(this, event, dev);
669}
670
671/*
672 * sfe_cm_init()
673 */
674static int __init sfe_cm_init(void)
675{
676 struct sfe_cm *sc = &__sc;
677 int result = -1;
678
679 DEBUG_INFO("SFE CM init\n");
680
681 /*
682 * Create sys/sfe_cm
683 */
684 sc->sys_sfe_cm = kobject_create_and_add("sfe_cm", NULL);
685 if (!sc->sys_sfe_cm) {
686 DEBUG_ERROR("failed to register sfe_cm\n");
687 goto exit1;
688 }
689
690 sc->dev_notifier.notifier_call = sfe_cm_device_event;
691 sc->dev_notifier.priority = 1;
692 register_netdevice_notifier(&sc->dev_notifier);
693
694 sc->inet_notifier.notifier_call = sfe_cm_inet_event;
695 sc->inet_notifier.priority = 1;
696 register_inetaddr_notifier(&sc->inet_notifier);
697
698 /*
699 * Register our netfilter hooks.
700 */
701 result = nf_register_hooks(sfe_cm_ipv4_ops_post_routing, ARRAY_SIZE(sfe_cm_ipv4_ops_post_routing));
702 if (result < 0) {
703 DEBUG_ERROR("can't register nf post routing hook: %d\n", result);
704 goto exit6;
705 }
706
707#ifdef CONFIG_NF_CONNTRACK_EVENTS
708 /*
709 * Register a notifier hook to get fast notifications of expired connections.
710 */
711 result = nf_conntrack_register_notifier(&init_net, &sfe_cm_conntrack_notifier);
712 if (result < 0) {
713 DEBUG_ERROR("can't register nf notifier hook: %d\n", result);
714 goto exit7;
715 }
716#endif
717
718 spin_lock_init(&sc->lock);
719
720 /*
721 * Hook the receive path in the network stack.
722 */
723 BUG_ON(athrs_fast_nat_recv != NULL);
724 RCU_INIT_POINTER(athrs_fast_nat_recv, sfe_cm_recv);
725
726 /*
727 * Hook the shortcut sync callback.
728 */
729 sfe_ipv4_register_sync_rule_callback(sfe_cm_sync_rule);
730 return 0;
731
732#ifdef CONFIG_NF_CONNTRACK_EVENTS
733exit7:
734#endif
735 nf_unregister_hooks(sfe_cm_ipv4_ops_post_routing, ARRAY_SIZE(sfe_cm_ipv4_ops_post_routing));
736
737exit6:
738 unregister_inetaddr_notifier(&sc->inet_notifier);
739 unregister_netdevice_notifier(&sc->dev_notifier);
740 kobject_put(sc->sys_sfe_cm);
741
742exit1:
743 return result;
744}
745
746/*
747 * sfe_cm_exit()
748 */
749static void __exit sfe_cm_exit(void)
750{
751 struct sfe_cm *sc = &__sc;
752
753 DEBUG_INFO("SFE CM exit\n");
754
755 /*
756 * Unregister our sync callback.
757 */
758 sfe_ipv4_register_sync_rule_callback(NULL);
759
760 /*
761 * Unregister our receive callback.
762 */
763 RCU_INIT_POINTER(athrs_fast_nat_recv, NULL);
764
765 /*
766 * Wait for all callbacks to complete.
767 */
768 rcu_barrier();
769
770 /*
771 * Destroy all connections.
772 */
773 sfe_ipv4_destroy_all_rules_for_dev(NULL);
774
775// XXX - this is where we need to unregister with any lower level offload services.
776
777#ifdef CONFIG_NF_CONNTRACK_EVENTS
778 nf_conntrack_unregister_notifier(&init_net, &sfe_cm_conntrack_notifier);
779
780#endif
781 nf_unregister_hooks(sfe_cm_ipv4_ops_post_routing, ARRAY_SIZE(sfe_cm_ipv4_ops_post_routing));
782
783 unregister_inetaddr_notifier(&sc->inet_notifier);
784 unregister_netdevice_notifier(&sc->dev_notifier);
785
786 kobject_put(sc->sys_sfe_cm);
787
788}
789
790module_init(sfe_cm_init)
791module_exit(sfe_cm_exit)
792
793MODULE_AUTHOR("Qualcomm Atheros Inc.");
794MODULE_DESCRIPTION("Shortcut Forwarding Engine - Connection Manager");
Matthew McClintocka3221942014-01-16 11:44:26 -0600795MODULE_LICENSE("Dual BSD/GPL");
Dave Hudsondcd08fb2013-11-22 09:25:16 -0600796