1 // SPDX-License-Identifier: GPL-2.0-only
3 * (C) 2015 Red Hat GmbH
4 * Author: Florian Westphal <fw@strlen.de>
7 #include <linux/module.h>
8 #include <linux/static_key.h>
9 #include <linux/hash.h>
10 #include <linux/jhash.h>
11 #include <linux/if_vlan.h>
12 #include <linux/init.h>
13 #include <linux/skbuff.h>
14 #include <linux/netlink.h>
15 #include <linux/netfilter.h>
16 #include <linux/netfilter/nfnetlink.h>
17 #include <linux/netfilter/nf_tables.h>
18 #include <net/netfilter/nf_tables_core.h>
19 #include <net/netfilter/nf_tables.h>
21 #define NFT_TRACETYPE_LL_HSIZE 20
22 #define NFT_TRACETYPE_NETWORK_HSIZE 40
23 #define NFT_TRACETYPE_TRANSPORT_HSIZE 20
25 DEFINE_STATIC_KEY_FALSE(nft_trace_enabled
);
26 EXPORT_SYMBOL_GPL(nft_trace_enabled
);
28 static int trace_fill_id(struct sk_buff
*nlskb
, struct sk_buff
*skb
)
32 /* using skb address as ID results in a limited number of
33 * values (and quick reuse).
35 * So we attempt to use as many skb members that will not
36 * change while skb is with netfilter.
38 id
= (__be32
)jhash_2words(hash32_ptr(skb
), skb_get_hash(skb
),
41 return nla_put_be32(nlskb
, NFTA_TRACE_ID
, id
);
44 static int trace_fill_header(struct sk_buff
*nlskb
, u16 type
,
45 const struct sk_buff
*skb
,
46 int off
, unsigned int len
)
53 nla
= nla_reserve(nlskb
, type
, len
);
54 if (!nla
|| skb_copy_bits(skb
, off
, nla_data(nla
), len
))
60 static int nf_trace_fill_ll_header(struct sk_buff
*nlskb
,
61 const struct sk_buff
*skb
)
63 struct vlan_ethhdr veth
;
66 BUILD_BUG_ON(sizeof(veth
) > NFT_TRACETYPE_LL_HSIZE
);
68 off
= skb_mac_header(skb
) - skb
->data
;
72 if (skb_copy_bits(skb
, off
, &veth
, ETH_HLEN
))
75 veth
.h_vlan_proto
= skb
->vlan_proto
;
76 veth
.h_vlan_TCI
= htons(skb_vlan_tag_get(skb
));
77 veth
.h_vlan_encapsulated_proto
= skb
->protocol
;
79 return nla_put(nlskb
, NFTA_TRACE_LL_HEADER
, sizeof(veth
), &veth
);
82 static int nf_trace_fill_dev_info(struct sk_buff
*nlskb
,
83 const struct net_device
*indev
,
84 const struct net_device
*outdev
)
87 if (nla_put_be32(nlskb
, NFTA_TRACE_IIF
,
88 htonl(indev
->ifindex
)))
91 if (nla_put_be16(nlskb
, NFTA_TRACE_IIFTYPE
,
97 if (nla_put_be32(nlskb
, NFTA_TRACE_OIF
,
98 htonl(outdev
->ifindex
)))
101 if (nla_put_be16(nlskb
, NFTA_TRACE_OIFTYPE
,
102 htons(outdev
->type
)))
109 static int nf_trace_fill_pkt_info(struct sk_buff
*nlskb
,
110 const struct nft_pktinfo
*pkt
)
112 const struct sk_buff
*skb
= pkt
->skb
;
113 int off
= skb_network_offset(skb
);
114 unsigned int len
, nh_end
;
116 nh_end
= pkt
->tprot_set
? pkt
->xt
.thoff
: skb
->len
;
117 len
= min_t(unsigned int, nh_end
- skb_network_offset(skb
),
118 NFT_TRACETYPE_NETWORK_HSIZE
);
119 if (trace_fill_header(nlskb
, NFTA_TRACE_NETWORK_HEADER
, skb
, off
, len
))
122 if (pkt
->tprot_set
) {
123 len
= min_t(unsigned int, skb
->len
- pkt
->xt
.thoff
,
124 NFT_TRACETYPE_TRANSPORT_HSIZE
);
125 if (trace_fill_header(nlskb
, NFTA_TRACE_TRANSPORT_HEADER
, skb
,
130 if (!skb_mac_header_was_set(skb
))
133 if (skb_vlan_tag_get(skb
))
134 return nf_trace_fill_ll_header(nlskb
, skb
);
136 off
= skb_mac_header(skb
) - skb
->data
;
137 len
= min_t(unsigned int, -off
, NFT_TRACETYPE_LL_HSIZE
);
138 return trace_fill_header(nlskb
, NFTA_TRACE_LL_HEADER
,
142 static int nf_trace_fill_rule_info(struct sk_buff
*nlskb
,
143 const struct nft_traceinfo
*info
)
148 /* a continue verdict with ->type == RETURN means that this is
149 * an implicit return (end of chain reached).
151 * Since no rule matched, the ->rule pointer is invalid.
153 if (info
->type
== NFT_TRACETYPE_RETURN
&&
154 info
->verdict
->code
== NFT_CONTINUE
)
157 return nla_put_be64(nlskb
, NFTA_TRACE_RULE_HANDLE
,
158 cpu_to_be64(info
->rule
->handle
),
162 static bool nft_trace_have_verdict_chain(struct nft_traceinfo
*info
)
164 switch (info
->type
) {
165 case NFT_TRACETYPE_RETURN
:
166 case NFT_TRACETYPE_RULE
:
172 switch (info
->verdict
->code
) {
183 void nft_trace_notify(struct nft_traceinfo
*info
)
185 const struct nft_pktinfo
*pkt
= info
->pkt
;
186 struct nfgenmsg
*nfmsg
;
187 struct nlmsghdr
*nlh
;
192 if (!nfnetlink_has_listeners(nft_net(pkt
), NFNLGRP_NFTRACE
))
195 size
= nlmsg_total_size(sizeof(struct nfgenmsg
)) +
196 nla_total_size(strlen(info
->chain
->table
->name
)) +
197 nla_total_size(strlen(info
->chain
->name
)) +
198 nla_total_size_64bit(sizeof(__be64
)) + /* rule handle */
199 nla_total_size(sizeof(__be32
)) + /* trace type */
200 nla_total_size(0) + /* VERDICT, nested */
201 nla_total_size(sizeof(u32
)) + /* verdict code */
202 nla_total_size(sizeof(u32
)) + /* id */
203 nla_total_size(NFT_TRACETYPE_LL_HSIZE
) +
204 nla_total_size(NFT_TRACETYPE_NETWORK_HSIZE
) +
205 nla_total_size(NFT_TRACETYPE_TRANSPORT_HSIZE
) +
206 nla_total_size(sizeof(u32
)) + /* iif */
207 nla_total_size(sizeof(__be16
)) + /* iiftype */
208 nla_total_size(sizeof(u32
)) + /* oif */
209 nla_total_size(sizeof(__be16
)) + /* oiftype */
210 nla_total_size(sizeof(u32
)) + /* mark */
211 nla_total_size(sizeof(u32
)) + /* nfproto */
212 nla_total_size(sizeof(u32
)); /* policy */
214 if (nft_trace_have_verdict_chain(info
))
215 size
+= nla_total_size(strlen(info
->verdict
->chain
->name
)); /* jump target */
217 skb
= nlmsg_new(size
, GFP_ATOMIC
);
221 event
= nfnl_msg_type(NFNL_SUBSYS_NFTABLES
, NFT_MSG_TRACE
);
222 nlh
= nlmsg_put(skb
, 0, 0, event
, sizeof(struct nfgenmsg
), 0);
224 goto nla_put_failure
;
226 nfmsg
= nlmsg_data(nlh
);
227 nfmsg
->nfgen_family
= info
->basechain
->type
->family
;
228 nfmsg
->version
= NFNETLINK_V0
;
231 if (nla_put_be32(skb
, NFTA_TRACE_NFPROTO
, htonl(nft_pf(pkt
))))
232 goto nla_put_failure
;
234 if (nla_put_be32(skb
, NFTA_TRACE_TYPE
, htonl(info
->type
)))
235 goto nla_put_failure
;
237 if (trace_fill_id(skb
, pkt
->skb
))
238 goto nla_put_failure
;
240 if (nla_put_string(skb
, NFTA_TRACE_CHAIN
, info
->chain
->name
))
241 goto nla_put_failure
;
243 if (nla_put_string(skb
, NFTA_TRACE_TABLE
, info
->chain
->table
->name
))
244 goto nla_put_failure
;
246 if (nf_trace_fill_rule_info(skb
, info
))
247 goto nla_put_failure
;
249 switch (info
->type
) {
250 case NFT_TRACETYPE_UNSPEC
:
251 case __NFT_TRACETYPE_MAX
:
253 case NFT_TRACETYPE_RETURN
:
254 case NFT_TRACETYPE_RULE
:
255 if (nft_verdict_dump(skb
, NFTA_TRACE_VERDICT
, info
->verdict
))
256 goto nla_put_failure
;
258 case NFT_TRACETYPE_POLICY
:
259 if (nla_put_be32(skb
, NFTA_TRACE_POLICY
,
260 htonl(info
->basechain
->policy
)))
261 goto nla_put_failure
;
265 if (pkt
->skb
->mark
&&
266 nla_put_be32(skb
, NFTA_TRACE_MARK
, htonl(pkt
->skb
->mark
)))
267 goto nla_put_failure
;
269 if (!info
->packet_dumped
) {
270 if (nf_trace_fill_dev_info(skb
, nft_in(pkt
), nft_out(pkt
)))
271 goto nla_put_failure
;
273 if (nf_trace_fill_pkt_info(skb
, pkt
))
274 goto nla_put_failure
;
275 info
->packet_dumped
= true;
279 nfnetlink_send(skb
, nft_net(pkt
), 0, NFNLGRP_NFTRACE
, 0, GFP_ATOMIC
);
287 void nft_trace_init(struct nft_traceinfo
*info
, const struct nft_pktinfo
*pkt
,
288 const struct nft_verdict
*verdict
,
289 const struct nft_chain
*chain
)
291 info
->basechain
= nft_base_chain(chain
);
293 info
->packet_dumped
= false;
295 info
->verdict
= verdict
;