2 * Checksum updating actions
4 * Copyright (c) 2010 Gregoire Baron <baronchon@n7mm.org>
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the Free
8 * Software Foundation; either version 2 of the License, or (at your option)
13 #include <linux/types.h>
14 #include <linux/init.h>
15 #include <linux/kernel.h>
16 #include <linux/module.h>
17 #include <linux/spinlock.h>
19 #include <linux/netlink.h>
20 #include <net/netlink.h>
21 #include <linux/rtnetlink.h>
23 #include <linux/skbuff.h>
28 #include <linux/icmpv6.h>
29 #include <linux/igmp.h>
32 #include <net/ip6_checksum.h>
34 #include <net/act_api.h>
36 #include <linux/tc_act/tc_csum.h>
37 #include <net/tc_act/tc_csum.h>
39 #define CSUM_TAB_MASK 15
41 static const struct nla_policy csum_policy
[TCA_CSUM_MAX
+ 1] = {
42 [TCA_CSUM_PARMS
] = { .len
= sizeof(struct tc_csum
), },
45 static int tcf_csum_init(struct net
*n
, struct nlattr
*nla
, struct nlattr
*est
,
46 struct tc_action
*a
, int ovr
, int bind
)
48 struct nlattr
*tb
[TCA_CSUM_MAX
+ 1];
56 err
= nla_parse_nested(tb
, TCA_CSUM_MAX
, nla
, csum_policy
);
60 if (tb
[TCA_CSUM_PARMS
] == NULL
)
62 parm
= nla_data(tb
[TCA_CSUM_PARMS
]);
64 if (!tcf_hash_check(parm
->index
, a
, bind
)) {
65 ret
= tcf_hash_create(parm
->index
, est
, a
, sizeof(*p
),
71 if (bind
)/* dont override defaults */
73 tcf_hash_release(a
, bind
);
79 spin_lock_bh(&p
->tcf_lock
);
80 p
->tcf_action
= parm
->action
;
81 p
->update_flags
= parm
->update_flags
;
82 spin_unlock_bh(&p
->tcf_lock
);
84 if (ret
== ACT_P_CREATED
)
91 * tcf_csum_skb_nextlayer - Get next layer pointer
92 * @skb: sk_buff to use
93 * @ihl: previous summed headers length
94 * @ipl: complete packet length
95 * @jhl: next header length
97 * Check the expected next layer availability in the specified sk_buff.
98 * Return the next layer pointer if pass, NULL otherwise.
100 static void *tcf_csum_skb_nextlayer(struct sk_buff
*skb
,
101 unsigned int ihl
, unsigned int ipl
,
104 int ntkoff
= skb_network_offset(skb
);
107 if (!pskb_may_pull(skb
, ipl
+ ntkoff
) || (ipl
< hl
) ||
108 skb_try_make_writable(skb
, hl
+ ntkoff
))
111 return (void *)(skb_network_header(skb
) + ihl
);
114 static int tcf_csum_ipv4_icmp(struct sk_buff
*skb
,
115 unsigned int ihl
, unsigned int ipl
)
117 struct icmphdr
*icmph
;
119 icmph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*icmph
));
124 skb
->csum
= csum_partial(icmph
, ipl
- ihl
, 0);
125 icmph
->checksum
= csum_fold(skb
->csum
);
127 skb
->ip_summed
= CHECKSUM_NONE
;
132 static int tcf_csum_ipv4_igmp(struct sk_buff
*skb
,
133 unsigned int ihl
, unsigned int ipl
)
135 struct igmphdr
*igmph
;
137 igmph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*igmph
));
142 skb
->csum
= csum_partial(igmph
, ipl
- ihl
, 0);
143 igmph
->csum
= csum_fold(skb
->csum
);
145 skb
->ip_summed
= CHECKSUM_NONE
;
150 static int tcf_csum_ipv6_icmp(struct sk_buff
*skb
,
151 unsigned int ihl
, unsigned int ipl
)
153 struct icmp6hdr
*icmp6h
;
154 const struct ipv6hdr
*ip6h
;
156 icmp6h
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*icmp6h
));
160 ip6h
= ipv6_hdr(skb
);
161 icmp6h
->icmp6_cksum
= 0;
162 skb
->csum
= csum_partial(icmp6h
, ipl
- ihl
, 0);
163 icmp6h
->icmp6_cksum
= csum_ipv6_magic(&ip6h
->saddr
, &ip6h
->daddr
,
164 ipl
- ihl
, IPPROTO_ICMPV6
,
167 skb
->ip_summed
= CHECKSUM_NONE
;
172 static int tcf_csum_ipv4_tcp(struct sk_buff
*skb
,
173 unsigned int ihl
, unsigned int ipl
)
176 const struct iphdr
*iph
;
178 tcph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*tcph
));
184 skb
->csum
= csum_partial(tcph
, ipl
- ihl
, 0);
185 tcph
->check
= tcp_v4_check(ipl
- ihl
,
186 iph
->saddr
, iph
->daddr
, skb
->csum
);
188 skb
->ip_summed
= CHECKSUM_NONE
;
193 static int tcf_csum_ipv6_tcp(struct sk_buff
*skb
,
194 unsigned int ihl
, unsigned int ipl
)
197 const struct ipv6hdr
*ip6h
;
199 tcph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*tcph
));
203 ip6h
= ipv6_hdr(skb
);
205 skb
->csum
= csum_partial(tcph
, ipl
- ihl
, 0);
206 tcph
->check
= csum_ipv6_magic(&ip6h
->saddr
, &ip6h
->daddr
,
207 ipl
- ihl
, IPPROTO_TCP
,
210 skb
->ip_summed
= CHECKSUM_NONE
;
215 static int tcf_csum_ipv4_udp(struct sk_buff
*skb
,
216 unsigned int ihl
, unsigned int ipl
, int udplite
)
219 const struct iphdr
*iph
;
223 * Support both UDP and UDPLITE checksum algorithms, Don't use
224 * udph->len to get the real length without any protocol check,
225 * UDPLITE uses udph->len for another thing,
226 * Use iph->tot_len, or just ipl.
229 udph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*udph
));
234 ul
= ntohs(udph
->len
);
236 if (udplite
|| udph
->check
) {
242 skb
->csum
= csum_partial(udph
, ipl
- ihl
, 0);
243 else if ((ul
>= sizeof(*udph
)) && (ul
<= ipl
- ihl
))
244 skb
->csum
= csum_partial(udph
, ul
, 0);
246 goto ignore_obscure_skb
;
249 goto ignore_obscure_skb
;
251 skb
->csum
= csum_partial(udph
, ul
, 0);
254 udph
->check
= csum_tcpudp_magic(iph
->saddr
, iph
->daddr
,
259 udph
->check
= CSUM_MANGLED_0
;
262 skb
->ip_summed
= CHECKSUM_NONE
;
268 static int tcf_csum_ipv6_udp(struct sk_buff
*skb
,
269 unsigned int ihl
, unsigned int ipl
, int udplite
)
272 const struct ipv6hdr
*ip6h
;
276 * Support both UDP and UDPLITE checksum algorithms, Don't use
277 * udph->len to get the real length without any protocol check,
278 * UDPLITE uses udph->len for another thing,
279 * Use ip6h->payload_len + sizeof(*ip6h) ... , or just ipl.
282 udph
= tcf_csum_skb_nextlayer(skb
, ihl
, ipl
, sizeof(*udph
));
286 ip6h
= ipv6_hdr(skb
);
287 ul
= ntohs(udph
->len
);
293 skb
->csum
= csum_partial(udph
, ipl
- ihl
, 0);
295 else if ((ul
>= sizeof(*udph
)) && (ul
<= ipl
- ihl
))
296 skb
->csum
= csum_partial(udph
, ul
, 0);
299 goto ignore_obscure_skb
;
302 goto ignore_obscure_skb
;
304 skb
->csum
= csum_partial(udph
, ul
, 0);
307 udph
->check
= csum_ipv6_magic(&ip6h
->saddr
, &ip6h
->daddr
, ul
,
308 udplite
? IPPROTO_UDPLITE
: IPPROTO_UDP
,
312 udph
->check
= CSUM_MANGLED_0
;
314 skb
->ip_summed
= CHECKSUM_NONE
;
320 static int tcf_csum_ipv4(struct sk_buff
*skb
, u32 update_flags
)
322 const struct iphdr
*iph
;
325 ntkoff
= skb_network_offset(skb
);
327 if (!pskb_may_pull(skb
, sizeof(*iph
) + ntkoff
))
332 switch (iph
->frag_off
& htons(IP_OFFSET
) ? 0 : iph
->protocol
) {
334 if (update_flags
& TCA_CSUM_UPDATE_FLAG_ICMP
)
335 if (!tcf_csum_ipv4_icmp(skb
, iph
->ihl
* 4,
336 ntohs(iph
->tot_len
)))
340 if (update_flags
& TCA_CSUM_UPDATE_FLAG_IGMP
)
341 if (!tcf_csum_ipv4_igmp(skb
, iph
->ihl
* 4,
342 ntohs(iph
->tot_len
)))
346 if (update_flags
& TCA_CSUM_UPDATE_FLAG_TCP
)
347 if (!tcf_csum_ipv4_tcp(skb
, iph
->ihl
* 4,
348 ntohs(iph
->tot_len
)))
352 if (update_flags
& TCA_CSUM_UPDATE_FLAG_UDP
)
353 if (!tcf_csum_ipv4_udp(skb
, iph
->ihl
* 4,
354 ntohs(iph
->tot_len
), 0))
357 case IPPROTO_UDPLITE
:
358 if (update_flags
& TCA_CSUM_UPDATE_FLAG_UDPLITE
)
359 if (!tcf_csum_ipv4_udp(skb
, iph
->ihl
* 4,
360 ntohs(iph
->tot_len
), 1))
365 if (update_flags
& TCA_CSUM_UPDATE_FLAG_IPV4HDR
) {
366 if (skb_try_make_writable(skb
, sizeof(*iph
) + ntkoff
))
369 ip_send_check(ip_hdr(skb
));
378 static int tcf_csum_ipv6_hopopts(struct ipv6_opt_hdr
*ip6xh
,
379 unsigned int ixhl
, unsigned int *pl
)
381 int off
, len
, optlen
;
382 unsigned char *xh
= (void *)ip6xh
;
384 off
= sizeof(*ip6xh
);
393 optlen
= xh
[off
+ 1] + 2;
394 if (optlen
!= 6 || len
< 6 || (off
& 3) != 2)
395 /* wrong jumbo option length/alignment */
397 *pl
= ntohl(*(__be32
*)(xh
+ off
+ 2));
400 optlen
= xh
[off
+ 1] + 2;
402 /* ignore obscure options */
414 static int tcf_csum_ipv6(struct sk_buff
*skb
, u32 update_flags
)
416 struct ipv6hdr
*ip6h
;
417 struct ipv6_opt_hdr
*ip6xh
;
418 unsigned int hl
, ixhl
;
423 ntkoff
= skb_network_offset(skb
);
427 if (!pskb_may_pull(skb
, hl
+ ntkoff
))
430 ip6h
= ipv6_hdr(skb
);
432 pl
= ntohs(ip6h
->payload_len
);
433 nexthdr
= ip6h
->nexthdr
;
437 case NEXTHDR_FRAGMENT
:
439 case NEXTHDR_ROUTING
:
442 if (!pskb_may_pull(skb
, hl
+ sizeof(*ip6xh
) + ntkoff
))
444 ip6xh
= (void *)(skb_network_header(skb
) + hl
);
445 ixhl
= ipv6_optlen(ip6xh
);
446 if (!pskb_may_pull(skb
, hl
+ ixhl
+ ntkoff
))
448 ip6xh
= (void *)(skb_network_header(skb
) + hl
);
449 if ((nexthdr
== NEXTHDR_HOP
) &&
450 !(tcf_csum_ipv6_hopopts(ip6xh
, ixhl
, &pl
)))
452 nexthdr
= ip6xh
->nexthdr
;
456 if (update_flags
& TCA_CSUM_UPDATE_FLAG_ICMP
)
457 if (!tcf_csum_ipv6_icmp(skb
,
458 hl
, pl
+ sizeof(*ip6h
)))
462 if (update_flags
& TCA_CSUM_UPDATE_FLAG_TCP
)
463 if (!tcf_csum_ipv6_tcp(skb
,
464 hl
, pl
+ sizeof(*ip6h
)))
468 if (update_flags
& TCA_CSUM_UPDATE_FLAG_UDP
)
469 if (!tcf_csum_ipv6_udp(skb
, hl
,
470 pl
+ sizeof(*ip6h
), 0))
473 case IPPROTO_UDPLITE
:
474 if (update_flags
& TCA_CSUM_UPDATE_FLAG_UDPLITE
)
475 if (!tcf_csum_ipv6_udp(skb
, hl
,
476 pl
+ sizeof(*ip6h
), 1))
482 } while (pskb_may_pull(skb
, hl
+ 1 + ntkoff
));
492 static int tcf_csum(struct sk_buff
*skb
,
493 const struct tc_action
*a
, struct tcf_result
*res
)
495 struct tcf_csum
*p
= a
->priv
;
499 spin_lock(&p
->tcf_lock
);
500 p
->tcf_tm
.lastuse
= jiffies
;
501 bstats_update(&p
->tcf_bstats
, skb
);
502 action
= p
->tcf_action
;
503 update_flags
= p
->update_flags
;
504 spin_unlock(&p
->tcf_lock
);
506 if (unlikely(action
== TC_ACT_SHOT
))
509 switch (tc_skb_protocol(skb
)) {
510 case cpu_to_be16(ETH_P_IP
):
511 if (!tcf_csum_ipv4(skb
, update_flags
))
514 case cpu_to_be16(ETH_P_IPV6
):
515 if (!tcf_csum_ipv6(skb
, update_flags
))
523 spin_lock(&p
->tcf_lock
);
524 p
->tcf_qstats
.drops
++;
525 spin_unlock(&p
->tcf_lock
);
529 static int tcf_csum_dump(struct sk_buff
*skb
,
530 struct tc_action
*a
, int bind
, int ref
)
532 unsigned char *b
= skb_tail_pointer(skb
);
533 struct tcf_csum
*p
= a
->priv
;
534 struct tc_csum opt
= {
535 .update_flags
= p
->update_flags
,
536 .index
= p
->tcf_index
,
537 .action
= p
->tcf_action
,
538 .refcnt
= p
->tcf_refcnt
- ref
,
539 .bindcnt
= p
->tcf_bindcnt
- bind
,
543 if (nla_put(skb
, TCA_CSUM_PARMS
, sizeof(opt
), &opt
))
544 goto nla_put_failure
;
545 t
.install
= jiffies_to_clock_t(jiffies
- p
->tcf_tm
.install
);
546 t
.lastuse
= jiffies_to_clock_t(jiffies
- p
->tcf_tm
.lastuse
);
547 t
.expires
= jiffies_to_clock_t(p
->tcf_tm
.expires
);
548 if (nla_put(skb
, TCA_CSUM_TM
, sizeof(t
), &t
))
549 goto nla_put_failure
;
558 static struct tc_action_ops act_csum_ops
= {
560 .type
= TCA_ACT_CSUM
,
561 .owner
= THIS_MODULE
,
563 .dump
= tcf_csum_dump
,
564 .init
= tcf_csum_init
,
567 MODULE_DESCRIPTION("Checksum updating actions");
568 MODULE_LICENSE("GPL");
570 static int __init
csum_init_module(void)
572 return tcf_register_action(&act_csum_ops
, CSUM_TAB_MASK
);
575 static void __exit
csum_cleanup_module(void)
577 tcf_unregister_action(&act_csum_ops
);
580 module_init(csum_init_module
);
581 module_exit(csum_cleanup_module
);