netfilter: exthdr: add support for tcp option removal
Subsystems (2)
Florian Westphal
maintainerPATCH 1/1 netfilter: exthdr: add support for tcp option removal
This allows to replace a tcp option with nop padding to selectively disable a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
---
proposed userspace syntax is:
nft add rule f in delete tcp option sack-perm
net/netfilter/nft_exthdr.c | 88 +++++++++++++++++++++++++++++++++++++- 1 file changed, 87 insertions(+), 1 deletion(-)
net/netfilter/nft_exthdr.c +87 -1
@@ -308,6 +308,55 @@ static void nft_exthdr_tcp_set_eval(const struct nft_expr *expr, regs->verdict.code =NFT_BREAK ; } +static void nft_exthdr_tcp_strip_eval (const struct nft_expr *expr, +struct nft_regs *regs, +const struct nft_pktinfo *pkt) +{ +u8 buff[sizeof (struct tcphdr ) +MAX_TCP_OPTION_SPACE ]; +struct nft_exthdr *priv =nft_expr_priv (expr); +unsigned int i, tcphdr_len, optl; +struct tcphdr *tcph; +u8 *opt; + + tcph =nft_tcp_header_pointer (pkt,sizeof (buff), buff, &tcphdr_len); +if (!tcph) + goto err; + +if (skb_ensure_writable (pkt->skb,nft_thoff (pkt) + tcphdr_len)) + goto err; + + opt = (u8 *)nft_tcp_header_pointer (pkt,sizeof (buff), buff, &tcphdr_len); +if (!opt) + goto err; +for (i =sizeof (*tcph); i < tcphdr_len -1 ; i += optl) { +unsigned int j; + + optl =optlen (opt, i); +if (priv->type != opt[i]) +continue ; + +if (i + optl > tcphdr_len) + goto err; + +for (j =0 ; j < optl; ++j) { +u16 n =TCPOPT_NOP ; +u16 o = opt[i+j]; + +if ((i + j) %2 ==0 ) { + o <<=8 ; + n <<=8 ; + } +inet_proto_csum_replace2 (&tcph->check, pkt->skb,htons (o), +htons (n), false); + } +memset (opt + i,TCPOPT_NOP , optl); +return ; + } +return ; +err: + regs->verdict.code =NFT_BREAK ; +} +static void nft_exthdr_sctp_eval (const struct nft_expr *expr,struct nft_regs *regs,const struct nft_pktinfo *pkt)
@@ -457,6 +506,28 @@ static int nft_exthdr_tcp_set_init(const struct nft_ctx *ctx, priv->len); } +static int nft_exthdr_tcp_strip_init (const struct nft_ctx *ctx, +const struct nft_expr *expr, +const struct nlattr *const tb[]) +{ +struct nft_exthdr *priv =nft_expr_priv (expr); + +if (tb[NFTA_EXTHDR_SREG ] || + tb[NFTA_EXTHDR_DREG ] || + tb[NFTA_EXTHDR_FLAGS ] || + tb[NFTA_EXTHDR_OFFSET ] || + tb[NFTA_EXTHDR_LEN ]) +return -EINVAL ; + +if (!tb[NFTA_EXTHDR_TYPE ]) +return -EINVAL ; + + priv->type =nla_get_u8 (tb[NFTA_EXTHDR_TYPE ]); + priv->op =NFT_EXTHDR_OP_TCPOPT ; + +return 0 ; +} +static int nft_exthdr_ipv4_init (const struct nft_ctx *ctx,const struct nft_expr *expr,const struct nlattr *const tb[])
@@ -517,6 +588,13 @@ static int nft_exthdr_dump_set(struct sk_buff *skb, const struct nft_expr *expr)return nft_exthdr_dump_common (skb, priv); } +static int nft_exthdr_dump_strip (struct sk_buff *skb,const struct nft_expr *expr) +{ +const struct nft_exthdr *priv =nft_expr_priv (expr); + +return nft_exthdr_dump_common (skb, priv); +} +static const struct nft_expr_ops nft_exthdr_ipv6_ops = { .type = &nft_exthdr_type, .size =NFT_EXPR_SIZE (sizeof (struct nft_exthdr )),
@@ -549,6 +627,14 @@ static const struct nft_expr_ops nft_exthdr_tcp_set_ops = { .dump = nft_exthdr_dump_set, }; +static const struct nft_expr_ops nft_exthdr_tcp_strip_ops = { + .type = &nft_exthdr_type, + .size =NFT_EXPR_SIZE (sizeof (struct nft_exthdr )), + .eval = nft_exthdr_tcp_strip_eval, + .init = nft_exthdr_tcp_strip_init, + .dump = nft_exthdr_dump_strip, +}; +static const struct nft_expr_ops nft_exthdr_sctp_ops = { .type = &nft_exthdr_type, .size =NFT_EXPR_SIZE (sizeof (struct nft_exthdr )),
@@ -576,7 +662,7 @@ nft_exthdr_select_ops(const struct nft_ctx *ctx,return &nft_exthdr_tcp_set_ops;if (tb[NFTA_EXTHDR_DREG ])return &nft_exthdr_tcp_ops; -break ; +return &nft_exthdr_tcp_strip_ops;case NFT_EXTHDR_OP_IPV6 :if (tb[NFTA_EXTHDR_DREG ])return &nft_exthdr_ipv6_ops;
Signature · 1 line
2.33.1
Pablo Neira Ayuso
maintainerQuoted 7 lines · Florian Westphal
This allows to replace a tcp option with nop padding to selectively disable a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
Is it worth to retain the bitmap approach?
Quoted 5 lines
Signed-off-by Florian Westphal <[email protected]>---
proposed userspace syntax is:nft add rule f in delete tcp option sack-perm
nft add rule f in tcp option reset sack-perm,...
Pablo Neira Ayuso
maintainerQuoted 10 lines · Pablo Neira Ayuso
Quoted 7 lines · Florian Westphal
This allows to replace a tcp option with nop padding to selectively disable a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
Is it worth to retain the bitmap approach?
Probably a new nested attribute to store the list of types that you would like to strip:
NFTA_EXTHDR_TYPES
NFTA_EXTHDR_TYPE
NFTA_EXTHDR_TYPE
From the kernel, you could build a bitmap (just like TCPOPTSTRIP) based on this list.
From the dump path, you can iterate over the bitmap to check for bitset to build this nest.
Quoted 7 lines
Quoted 5 lines
Signed-off-by Florian Westphal <[email protected]>---
proposed userspace syntax is:nft add rule f in delete tcp option sack-perm
nft add rule f in tcp option reset sack-perm,...
Florian Westphal
maintainerQuoted 10 lines · Pablo Neira Ayuso
Quoted 7 lines · Florian Westphal
This allows to replace a tcp option with nop padding to selectively disable a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
Is it worth to retain the bitmap approach?
I don't think so. For TCPOPTSTRIP it makes sense because you can't use multiple targets in one rule.
I'd rework this to not set BREAK if the option wasn't present in the first place, so you could do
delete tcp option sack-perm delete tcp option timestamp ...
and so on.
Let me know if you disagree.
I could also rework it so that option comes from sreg instead of imm, but i could not find a use-case where having the option number coming from a map lookup would make sense.
Quoted 7 lines
Quoted 5 lines
Signed-off-by Florian Westphal <[email protected]>---
proposed userspace syntax is:nft add rule f in delete tcp option sack-perm
nft add rule f in tcp option reset sack-perm,...
Why 'reset'? My initial version had 'remove' but 'delete' already exists as a token so it was simpler.
Pablo Neira Ayuso
maintainerQuoted 15 lines · Florian Westphal
Quoted 10 lines · Pablo Neira Ayuso
Quoted 7 lines · Florian Westphal
This allows to replace a tcp option with nop padding to selectively disable a particular tcp option.
Optstrip mode is chosen when userspace passes the exthdr expression with neither a source nor a destination register attribute.
This is identical to xtables TCPOPTSTRIP extension.
Is it worth to retain the bitmap approach?
I don't think so. For TCPOPTSTRIP it makes sense because you can't use multiple targets in one rule.
Quoted 8 lines
I'd rework this to not set BREAK if the option wasn't present in the first place, so you could do
delete tcp option sack-perm delete tcp option timestamp ...
and so on.
Let me know if you disagree.
It's OK if you prefer this way. I can see references on the web to reseting multiple options, not sure if it is actually useful in practise, in such you can to parse the packet several times.
Quoted 8 lines
I could also rework it so that option comes from sreg instead of imm, but i could not find a use-case where having the option number coming from a map lookup would make sense.
Quoted 4 lines
Quoted 4 lines
Signed-off-by Florian Westphal <[email protected]>---
proposed userspace syntax is:
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
nft add rule f in delete tcp option sack-perm
nft add rule f in tcp option reset sack-perm,...
Why 'reset'? My initial version had 'remove' but 'delete' already exists as a token so it was simpler.
'reset' also exists as a token. This is setting to nop, I just though reset might make more sense, there might be a nice to really remove TCP options in the future (costful but paranoid scenario, an observer can spot options that have been nop)