1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16#define KMSG_COMPONENT "IPVS"
17#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18
19#include <linux/in.h>
20#include <linux/ip.h>
21#include <linux/kernel.h>
22#include <linux/netfilter.h>
23#include <linux/netfilter_ipv4.h>
24#include <linux/udp.h>
25
26#include <net/ip_vs.h>
27#include <net/ip.h>
28#include <net/ip6_checksum.h>
29
30static struct ip_vs_conn *
31udp_conn_in_get(int af, const struct sk_buff *skb, struct ip_vs_protocol *pp,
32 const struct ip_vs_iphdr *iph, unsigned int proto_off,
33 int inverse)
34{
35 struct ip_vs_conn *cp;
36 __be16 _ports[2], *pptr;
37
38 pptr = skb_header_pointer(skb, proto_off, sizeof(_ports), _ports);
39 if (pptr == NULL)
40 return NULL;
41
42 if (likely(!inverse)) {
43 cp = ip_vs_conn_in_get(af, iph->protocol,
44 &iph->saddr, pptr[0],
45 &iph->daddr, pptr[1]);
46 } else {
47 cp = ip_vs_conn_in_get(af, iph->protocol,
48 &iph->daddr, pptr[1],
49 &iph->saddr, pptr[0]);
50 }
51
52 return cp;
53}
54
55
56static struct ip_vs_conn *
57udp_conn_out_get(int af, const struct sk_buff *skb, struct ip_vs_protocol *pp,
58 const struct ip_vs_iphdr *iph, unsigned int proto_off,
59 int inverse)
60{
61 struct ip_vs_conn *cp;
62 __be16 _ports[2], *pptr;
63
64 pptr = skb_header_pointer(skb, proto_off, sizeof(_ports), _ports);
65 if (pptr == NULL)
66 return NULL;
67
68 if (likely(!inverse)) {
69 cp = ip_vs_conn_out_get(af, iph->protocol,
70 &iph->saddr, pptr[0],
71 &iph->daddr, pptr[1]);
72 } else {
73 cp = ip_vs_conn_out_get(af, iph->protocol,
74 &iph->daddr, pptr[1],
75 &iph->saddr, pptr[0]);
76 }
77
78 return cp;
79}
80
81
82static int
83udp_conn_schedule(int af, struct sk_buff *skb, struct ip_vs_protocol *pp,
84 int *verdict, struct ip_vs_conn **cpp)
85{
86 struct ip_vs_service *svc;
87 struct udphdr _udph, *uh;
88 struct ip_vs_iphdr iph;
89
90 ip_vs_fill_iphdr(af, skb_network_header(skb), &iph);
91
92 uh = skb_header_pointer(skb, iph.len, sizeof(_udph), &_udph);
93 if (uh == NULL) {
94 *verdict = NF_DROP;
95 return 0;
96 }
97
98 svc = ip_vs_service_get(af, skb->mark, iph.protocol,
99 &iph.daddr, uh->dest);
100 if (svc) {
101 if (ip_vs_todrop()) {
102
103
104
105
106 ip_vs_service_put(svc);
107 *verdict = NF_DROP;
108 return 0;
109 }
110
111
112
113
114
115 *cpp = ip_vs_schedule(svc, skb);
116 if (!*cpp) {
117 *verdict = ip_vs_leave(svc, skb, pp);
118 return 0;
119 }
120 ip_vs_service_put(svc);
121 }
122 return 1;
123}
124
125
126static inline void
127udp_fast_csum_update(int af, struct udphdr *uhdr,
128 const union nf_inet_addr *oldip,
129 const union nf_inet_addr *newip,
130 __be16 oldport, __be16 newport)
131{
132#ifdef CONFIG_IP_VS_IPV6
133 if (af == AF_INET6)
134 uhdr->check =
135 csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
136 ip_vs_check_diff2(oldport, newport,
137 ~csum_unfold(uhdr->check))));
138 else
139#endif
140 uhdr->check =
141 csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
142 ip_vs_check_diff2(oldport, newport,
143 ~csum_unfold(uhdr->check))));
144 if (!uhdr->check)
145 uhdr->check = CSUM_MANGLED_0;
146}
147
148static inline void
149udp_partial_csum_update(int af, struct udphdr *uhdr,
150 const union nf_inet_addr *oldip,
151 const union nf_inet_addr *newip,
152 __be16 oldlen, __be16 newlen)
153{
154#ifdef CONFIG_IP_VS_IPV6
155 if (af == AF_INET6)
156 uhdr->check =
157 csum_fold(ip_vs_check_diff16(oldip->ip6, newip->ip6,
158 ip_vs_check_diff2(oldlen, newlen,
159 ~csum_unfold(uhdr->check))));
160 else
161#endif
162 uhdr->check =
163 csum_fold(ip_vs_check_diff4(oldip->ip, newip->ip,
164 ip_vs_check_diff2(oldlen, newlen,
165 ~csum_unfold(uhdr->check))));
166}
167
168
169static int
170udp_snat_handler(struct sk_buff *skb,
171 struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
172{
173 struct udphdr *udph;
174 unsigned int udphoff;
175 int oldlen;
176
177#ifdef CONFIG_IP_VS_IPV6
178 if (cp->af == AF_INET6)
179 udphoff = sizeof(struct ipv6hdr);
180 else
181#endif
182 udphoff = ip_hdrlen(skb);
183 oldlen = skb->len - udphoff;
184
185
186 if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
187 return 0;
188
189 if (unlikely(cp->app != NULL)) {
190
191 if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
192 return 0;
193
194
195
196
197 if (!ip_vs_app_pkt_out(cp, skb))
198 return 0;
199 }
200
201 udph = (void *)skb_network_header(skb) + udphoff;
202 udph->source = cp->vport;
203
204
205
206
207 if (skb->ip_summed == CHECKSUM_PARTIAL) {
208 udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
209 htons(oldlen),
210 htons(skb->len - udphoff));
211 } else if (!cp->app && (udph->check != 0)) {
212
213 udp_fast_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
214 cp->dport, cp->vport);
215 if (skb->ip_summed == CHECKSUM_COMPLETE)
216 skb->ip_summed = CHECKSUM_NONE;
217 } else {
218
219 udph->check = 0;
220 skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
221#ifdef CONFIG_IP_VS_IPV6
222 if (cp->af == AF_INET6)
223 udph->check = csum_ipv6_magic(&cp->vaddr.in6,
224 &cp->caddr.in6,
225 skb->len - udphoff,
226 cp->protocol, skb->csum);
227 else
228#endif
229 udph->check = csum_tcpudp_magic(cp->vaddr.ip,
230 cp->caddr.ip,
231 skb->len - udphoff,
232 cp->protocol,
233 skb->csum);
234 if (udph->check == 0)
235 udph->check = CSUM_MANGLED_0;
236 IP_VS_DBG(11, "O-pkt: %s O-csum=%d (+%zd)\n",
237 pp->name, udph->check,
238 (char*)&(udph->check) - (char*)udph);
239 }
240 return 1;
241}
242
243
244static int
245udp_dnat_handler(struct sk_buff *skb,
246 struct ip_vs_protocol *pp, struct ip_vs_conn *cp)
247{
248 struct udphdr *udph;
249 unsigned int udphoff;
250 int oldlen;
251
252#ifdef CONFIG_IP_VS_IPV6
253 if (cp->af == AF_INET6)
254 udphoff = sizeof(struct ipv6hdr);
255 else
256#endif
257 udphoff = ip_hdrlen(skb);
258 oldlen = skb->len - udphoff;
259
260
261 if (!skb_make_writable(skb, udphoff+sizeof(*udph)))
262 return 0;
263
264 if (unlikely(cp->app != NULL)) {
265
266 if (pp->csum_check && !pp->csum_check(cp->af, skb, pp))
267 return 0;
268
269
270
271
272
273 if (!ip_vs_app_pkt_in(cp, skb))
274 return 0;
275 }
276
277 udph = (void *)skb_network_header(skb) + udphoff;
278 udph->dest = cp->dport;
279
280
281
282
283 if (skb->ip_summed == CHECKSUM_PARTIAL) {
284 udp_partial_csum_update(cp->af, udph, &cp->daddr, &cp->vaddr,
285 htons(oldlen),
286 htons(skb->len - udphoff));
287 } else if (!cp->app && (udph->check != 0)) {
288
289 udp_fast_csum_update(cp->af, udph, &cp->vaddr, &cp->daddr,
290 cp->vport, cp->dport);
291 if (skb->ip_summed == CHECKSUM_COMPLETE)
292 skb->ip_summed = CHECKSUM_NONE;
293 } else {
294
295 udph->check = 0;
296 skb->csum = skb_checksum(skb, udphoff, skb->len - udphoff, 0);
297#ifdef CONFIG_IP_VS_IPV6
298 if (cp->af == AF_INET6)
299 udph->check = csum_ipv6_magic(&cp->caddr.in6,
300 &cp->daddr.in6,
301 skb->len - udphoff,
302 cp->protocol, skb->csum);
303 else
304#endif
305 udph->check = csum_tcpudp_magic(cp->caddr.ip,
306 cp->daddr.ip,
307 skb->len - udphoff,
308 cp->protocol,
309 skb->csum);
310 if (udph->check == 0)
311 udph->check = CSUM_MANGLED_0;
312 skb->ip_summed = CHECKSUM_UNNECESSARY;
313 }
314 return 1;
315}
316
317
318static int
319udp_csum_check(int af, struct sk_buff *skb, struct ip_vs_protocol *pp)
320{
321 struct udphdr _udph, *uh;
322 unsigned int udphoff;
323
324#ifdef CONFIG_IP_VS_IPV6
325 if (af == AF_INET6)
326 udphoff = sizeof(struct ipv6hdr);
327 else
328#endif
329 udphoff = ip_hdrlen(skb);
330
331 uh = skb_header_pointer(skb, udphoff, sizeof(_udph), &_udph);
332 if (uh == NULL)
333 return 0;
334
335 if (uh->check != 0) {
336 switch (skb->ip_summed) {
337 case CHECKSUM_NONE:
338 skb->csum = skb_checksum(skb, udphoff,
339 skb->len - udphoff, 0);
340 case CHECKSUM_COMPLETE:
341#ifdef CONFIG_IP_VS_IPV6
342 if (af == AF_INET6) {
343 if (csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
344 &ipv6_hdr(skb)->daddr,
345 skb->len - udphoff,
346 ipv6_hdr(skb)->nexthdr,
347 skb->csum)) {
348 IP_VS_DBG_RL_PKT(0, pp, skb, 0,
349 "Failed checksum for");
350 return 0;
351 }
352 } else
353#endif
354 if (csum_tcpudp_magic(ip_hdr(skb)->saddr,
355 ip_hdr(skb)->daddr,
356 skb->len - udphoff,
357 ip_hdr(skb)->protocol,
358 skb->csum)) {
359 IP_VS_DBG_RL_PKT(0, pp, skb, 0,
360 "Failed checksum for");
361 return 0;
362 }
363 break;
364 default:
365
366 break;
367 }
368 }
369 return 1;
370}
371
372
373
374
375
376
377
378#define UDP_APP_TAB_BITS 4
379#define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
380#define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
381
382static struct list_head udp_apps[UDP_APP_TAB_SIZE];
383static DEFINE_SPINLOCK(udp_app_lock);
384
385static inline __u16 udp_app_hashkey(__be16 port)
386{
387 return (((__force u16)port >> UDP_APP_TAB_BITS) ^ (__force u16)port)
388 & UDP_APP_TAB_MASK;
389}
390
391
392static int udp_register_app(struct ip_vs_app *inc)
393{
394 struct ip_vs_app *i;
395 __u16 hash;
396 __be16 port = inc->port;
397 int ret = 0;
398
399 hash = udp_app_hashkey(port);
400
401
402 spin_lock_bh(&udp_app_lock);
403 list_for_each_entry(i, &udp_apps[hash], p_list) {
404 if (i->port == port) {
405 ret = -EEXIST;
406 goto out;
407 }
408 }
409 list_add(&inc->p_list, &udp_apps[hash]);
410 atomic_inc(&ip_vs_protocol_udp.appcnt);
411
412 out:
413 spin_unlock_bh(&udp_app_lock);
414 return ret;
415}
416
417
418static void
419udp_unregister_app(struct ip_vs_app *inc)
420{
421 spin_lock_bh(&udp_app_lock);
422 atomic_dec(&ip_vs_protocol_udp.appcnt);
423 list_del(&inc->p_list);
424 spin_unlock_bh(&udp_app_lock);
425}
426
427
428static int udp_app_conn_bind(struct ip_vs_conn *cp)
429{
430 int hash;
431 struct ip_vs_app *inc;
432 int result = 0;
433
434
435 if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)
436 return 0;
437
438
439 hash = udp_app_hashkey(cp->vport);
440
441 spin_lock(&udp_app_lock);
442 list_for_each_entry(inc, &udp_apps[hash], p_list) {
443 if (inc->port == cp->vport) {
444 if (unlikely(!ip_vs_app_inc_get(inc)))
445 break;
446 spin_unlock(&udp_app_lock);
447
448 IP_VS_DBG_BUF(9, "%s(): Binding conn %s:%u->"
449 "%s:%u to app %s on port %u\n",
450 __func__,
451 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
452 ntohs(cp->cport),
453 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
454 ntohs(cp->vport),
455 inc->name, ntohs(inc->port));
456
457 cp->app = inc;
458 if (inc->init_conn)
459 result = inc->init_conn(inc, cp);
460 goto out;
461 }
462 }
463 spin_unlock(&udp_app_lock);
464
465 out:
466 return result;
467}
468
469
470static int udp_timeouts[IP_VS_UDP_S_LAST+1] = {
471 [IP_VS_UDP_S_NORMAL] = 5*60*HZ,
472 [IP_VS_UDP_S_LAST] = 2*HZ,
473};
474
475static const char *const udp_state_name_table[IP_VS_UDP_S_LAST+1] = {
476 [IP_VS_UDP_S_NORMAL] = "UDP",
477 [IP_VS_UDP_S_LAST] = "BUG!",
478};
479
480
481static int
482udp_set_state_timeout(struct ip_vs_protocol *pp, char *sname, int to)
483{
484 return ip_vs_set_state_timeout(pp->timeout_table, IP_VS_UDP_S_LAST,
485 udp_state_name_table, sname, to);
486}
487
488static const char * udp_state_name(int state)
489{
490 if (state >= IP_VS_UDP_S_LAST)
491 return "ERR!";
492 return udp_state_name_table[state] ? udp_state_name_table[state] : "?";
493}
494
495static int
496udp_state_transition(struct ip_vs_conn *cp, int direction,
497 const struct sk_buff *skb,
498 struct ip_vs_protocol *pp)
499{
500 cp->timeout = pp->timeout_table[IP_VS_UDP_S_NORMAL];
501 return 1;
502}
503
504static void udp_init(struct ip_vs_protocol *pp)
505{
506 IP_VS_INIT_HASH_TABLE(udp_apps);
507 pp->timeout_table = udp_timeouts;
508}
509
510static void udp_exit(struct ip_vs_protocol *pp)
511{
512}
513
514
515struct ip_vs_protocol ip_vs_protocol_udp = {
516 .name = "UDP",
517 .protocol = IPPROTO_UDP,
518 .num_states = IP_VS_UDP_S_LAST,
519 .dont_defrag = 0,
520 .init = udp_init,
521 .exit = udp_exit,
522 .conn_schedule = udp_conn_schedule,
523 .conn_in_get = udp_conn_in_get,
524 .conn_out_get = udp_conn_out_get,
525 .snat_handler = udp_snat_handler,
526 .dnat_handler = udp_dnat_handler,
527 .csum_check = udp_csum_check,
528 .state_transition = udp_state_transition,
529 .state_name = udp_state_name,
530 .register_app = udp_register_app,
531 .unregister_app = udp_unregister_app,
532 .app_conn_bind = udp_app_conn_bind,
533 .debug_packet = ip_vs_tcpudp_debug_packet,
534 .timeout_change = NULL,
535 .set_state_timeout = udp_set_state_timeout,
536};
537
