Date:	Wed, 8 Feb 2012 01:27:43 +0100
From:	Pablo Neira Ayuso <pablo@...filter.org>
To:	Hans Schillstrom <hans.schillstrom@...csson.com>
Cc:	kaber@...sh.net, jengelh@...ozas.de,
	netfilter-devel@...r.kernel.org, netdev@...r.kernel.org,
	hans@...illstrom.com
Subject: Re: [v8 PATCH 2/3] NETFILTER module xt_hmark, new target for HASH
 based fwmark

On Fri, Jan 27, 2012 at 03:41:42PM +0100, Hans Schillstrom wrote:
> diff --git a/include/linux/netfilter/xt_hmark.h b/include/linux/netfilter/xt_hmark.h
> new file mode 100644
> index 0000000..f2ac47b
> --- /dev/null
> +++ b/include/linux/netfilter/xt_hmark.h
> @@ -0,0 +1,71 @@
> +#ifndef XT_HMARK_H_
> +#define XT_HMARK_H_
> +
> +#include <linux/types.h>
> +
> +/*
> + * Flags must not start at 0, since it's used as none.

Then, define XT_HMARK_NONE = 0.

Please, once this is done, remove this comment.

> + */
> +enum {
> +	XT_HMARK_SADR_AND = 1,	/* SNAT & DNAT are used by the kernel module */
                                              ^^^^^
I don't understand why that comment is there.

> +	XT_HMARK_DADR_AND,
> +	XT_HMARK_SPI_AND,
> +	XT_HMARK_SPI_OR,
> +	XT_HMARK_SPORT_AND,
> +	XT_HMARK_DPORT_AND,
> +	XT_HMARK_SPORT_OR,
> +	XT_HMARK_DPORT_OR,
> +	XT_HMARK_PROTO_AND,
> +	XT_HMARK_RND,
> +	XT_HMARK_MODULUS,
> +	XT_HMARK_OFFSET,
> +	XT_HMARK_USE_SNAT,
> +	XT_HMARK_USE_DNAT,
> +	XT_HMARK_METHOD_L3,
> +	XT_HMARK_METHOD_L3_4,
> +	XT_F_HMARK_USE_SNAT = 1 << XT_HMARK_USE_SNAT,

You can probably do something like this to beautify this defintion:

XT_F_HMARK_WHATEVER     = (1 << XT_HMARK_BLAH),
XT_F_HMARK_BLOB         = (1 << XT_HMARK_BLAHBLAH),

With some tabs, and so on.

> +	XT_F_HMARK_USE_DNAT = 1 << XT_HMARK_USE_DNAT,
> +	XT_F_HMARK_SADR_AND = 1 << XT_HMARK_SADR_AND,
> +	XT_F_HMARK_DADR_AND = 1 << XT_HMARK_DADR_AND,
> +	XT_F_HMARK_SPI_AND = 1 << XT_HMARK_SPI_AND,
> +	XT_F_HMARK_SPI_OR = 1 << XT_HMARK_SPI_OR,
> +	XT_F_HMARK_SPORT_AND = 1 << XT_HMARK_SPORT_AND,
> +	XT_F_HMARK_DPORT_AND = 1 << XT_HMARK_DPORT_AND,
> +	XT_F_HMARK_SPORT_OR = 1 << XT_HMARK_SPORT_OR,
> +	XT_F_HMARK_DPORT_OR = 1 << XT_HMARK_DPORT_OR,
> +	XT_F_HMARK_PROTO_AND = 1 << XT_HMARK_PROTO_AND,
> +	XT_F_HMARK_RND = 1 << XT_HMARK_RND,
> +	XT_F_HMARK_MODULUS = 1 << XT_HMARK_MODULUS,
> +	XT_F_HMARK_OFFSET = 1 << XT_HMARK_OFFSET,
> +	XT_F_HMARK_METHOD_L3 = 1 << XT_HMARK_METHOD_L3,
> +	XT_F_HMARK_METHOD_L3_4 = 1 << XT_HMARK_METHOD_L3_4,
> +};
> +
> +#define XT_F_HMARK_L4_OPTS  (XT_F_HMARK_SPI_AND | XT_F_HMARK_SPI_OR\
> +			     | XT_F_HMARK_SPORT_AND | XT_F_HMARK_SPORT_OR\
> +			     | XT_F_HMARK_DPORT_AND | XT_F_HMARK_DPORT_OR\
> +			     | XT_F_HMARK_PROTO_AND)

I find nobody using this definition in the kernel. Please, move it
where it belong. 

> +
> +union hports {

this is exposed to user-space. Please, use xt_hmark_ports or something
similar to make sure we don't have any clash in the name-space.

Better. Define this structure inside xt_hmark_info since it only seems
to be useful there, I'd say.

> +	struct {
> +		__u16	src;
> +		__u16	dst;
> +	} p16;
> +	__u32	v32;
> +};
> +
> +struct xt_hmark_info {
> +	union nf_inet_addr	smask;		/* Source address mask */
> +	union nf_inet_addr	dmask;		/* Dest address mask */
> +	union hports		pmask;
> +	union hports		pset;
> +	__u32			spimask;
> +	__u32			spiset;
> +	__u16			flags;		/* Print out only */
> +	__u16			prmask;		/* L4 Proto mask */
> +	__u32			hashrnd;
> +	__u32			hmod;		/* Modulus */
> +	__u32			hoffs;		/* Offset */
> +};
> +
> +#endif /* XT_HMARK_H_ */
> diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
> index f8ac4ef..dfe84e1 100644
> --- a/net/netfilter/Kconfig
> +++ b/net/netfilter/Kconfig
> @@ -488,6 +488,23 @@ config NETFILTER_XT_TARGET_HL
>  	since you can easily create immortal packets that loop
>  	forever on the network.
>  
> +config NETFILTER_XT_TARGET_HMARK
> +	tristate '"HMARK" target support'
> +	depends on NETFILTER_ADVANCED
> +	---help---
> +	This option adds the "HMARK" target.
> +
> +	The target allows you to create rules in the "raw" and "mangle" tables
> +	which alter the netfilter mark (nfmark) field within a given range.
> +	First a 32 bit hash value is generated then modulus by <limit> and
> +	finally an offset is added before it's written to nfmark.
> +
> +	Prior to routing, the nfmark can influence the routing method (see
> +	"Use netfilter MARK value as routing key") and can also be used by
> +	other subsystems to change their behavior.
> +
> +	The mark match can also be used to match nfmark produced by this module.
> +
>  config NETFILTER_XT_TARGET_IDLETIMER
>  	tristate  "IDLETIMER target support"
>  	depends on NETFILTER_ADVANCED
> diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
> index 40f4c3d..21bc5e8 100644
> --- a/net/netfilter/Makefile
> +++ b/net/netfilter/Makefile
> @@ -57,6 +57,7 @@ obj-$(CONFIG_NETFILTER_XT_TARGET_CONNSECMARK) += xt_CONNSECMARK.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_CT) += xt_CT.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_DSCP) += xt_DSCP.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_HL) += xt_HL.o
> +obj-$(CONFIG_NETFILTER_XT_TARGET_HMARK) += xt_hmark.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_LED) += xt_LED.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_NFLOG) += xt_NFLOG.o
>  obj-$(CONFIG_NETFILTER_XT_TARGET_NFQUEUE) += xt_NFQUEUE.o
> diff --git a/net/netfilter/xt_hmark.c b/net/netfilter/xt_hmark.c
> new file mode 100644
> index 0000000..c9d6654
> --- /dev/null
> +++ b/net/netfilter/xt_hmark.c
> @@ -0,0 +1,334 @@
> +/*
> + * xt_hmark - Netfilter module to set mark as hash value
> + *
> + * (C) 2011 Hans Schillstrom <hans.schillstrom@...csson.com>
> + *
> + *Description:
> + *	This module calculates a hash value that can be modified by modulus
> + *	and an offset, i.e. it is possible to produce a skb->mark within a range.
> + *	The hash value is based on a direction independent five tuple:
> + *	src & dst addr src & dst ports and protocol.

This description above I think it's sufficient.

I think you can remove this header below. The documentation will be
available through the manpage.

> + *	There is two distinct modes for hash calculation:
> + *
> + *	MODE_L3:
> + *	In this mode ONLY src & dst addresses can be used in hash calc.
> + *	src-mask & dst-mask is the only valid masks.
> + *	In this mode no special care for fragments is necessary.
> + *
> + *	MODE_L3_4:
> + *	All five fields L4-proto, ports and addresses can be used in calc.
> + *	ESP and AH don't have ports so SPI will be used instead.
> + *	AH will not use ports even if it might be possible.
> + *	Tunnels - only the outer saddr and daddr will be used,
> + *
> + *	For ICMP error messages the hash mark values will be calculated on
> + *	the source packet i.e. the packet caused the error (If sufficient
> + *	amount of data exists).
> + *
> + *	Fragments is not handled in this mode, (if they reach us)
> + *	i.e.  fw-mark will be updated.
> + *
> + *	This program is free software; you can redistribute it and/or modify
> + *	it under the terms of the GNU General Public License version 2 as
> + *	published by the Free Software Foundation.
> + */
> +
> +#include <linux/module.h>
> +#include <linux/skbuff.h>
> +#include <net/ip.h>
> +#include <linux/icmp.h>
> +
> +#include <linux/netfilter/xt_hmark.h>
> +#include <linux/netfilter/x_tables.h>
> +#if defined(CONFIG_NF_NAT)
> +#include <net/netfilter/nf_nat.h>
> +#endif
> +#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
> +#	define WITH_IPV6 1
> +#include <net/ipv6.h>
> +#include <linux/netfilter_ipv6/ip6_tables.h>
> +#endif
> +
> +MODULE_LICENSE("GPL");
> +MODULE_AUTHOR("Hans Schillstrom <hans.schillstrom@...csson.com>");
> +MODULE_DESCRIPTION("Xtables: Packet range mark operations by Hash value");
> +MODULE_ALIAS("ipt_HMARK");
> +MODULE_ALIAS("ip6t_HMARK");
> +
> +/*
> + * ICMP, get header offset if icmp error
> + */
> +static int get_inner_hdr(struct sk_buff *skb, int iphsz, int nhoff)
> +{
> +	const struct icmphdr *icmph;
> +	struct icmphdr _ih;
> +
> +	/* Not enough header? */
> +	icmph = skb_header_pointer(skb, nhoff + iphsz, sizeof(_ih), &_ih);
> +	if (icmph == NULL)
> +		return nhoff;
> +
> +	if (icmph->type > NR_ICMP_TYPES)
> +		return nhoff;
> +
> +	/* Error message? */
> +	if (icmph->type != ICMP_DEST_UNREACH &&
> +	    icmph->type != ICMP_SOURCE_QUENCH &&
> +	    icmph->type != ICMP_TIME_EXCEEDED &&
> +	    icmph->type != ICMP_PARAMETERPROB &&
> +	    icmph->type != ICMP_REDIRECT)
> +		return nhoff;
> +
> +	return nhoff + iphsz + sizeof(_ih);
> +}
> +
> +#ifdef WITH_IPV6
> +/*
> + * Get ipv6 header offset if icmp error
> + */
> +static int get_inner6_hdr(struct sk_buff *skb, int *offset)
> +{
> +	struct icmp6hdr *icmp6h, _ih6;
> +
> +	icmp6h = skb_header_pointer(skb, *offset, sizeof(_ih6), &_ih6);
> +	if (icmp6h == NULL)
> +		return 0;
> +
> +	if (icmp6h->icmp6_type && icmp6h->icmp6_type < 128) {
> +		*offset +=  sizeof(struct icmp6hdr);
> +		return 1;
> +	}
> +	return 0;
> +}
> +/*
> + * Calculate hash based fw-mark, on the five tuple if possible.
> + * special cases :
> + *  - Fragments do not use ports not even on the first fragment,
> + *    nf_defrag_ipv6.ko don't defrag for us like it do in ipv4.
> + *    This might be changed in the future.
> + *  - On ICMP errors the inner header will be used.
> + *  - Tunnels no ports
> + *  - ESP & AH uses SPI
> + * @returns XT_CONTINUE
> + */
> +static unsigned int
> +hmark_v6(struct sk_buff *skb, const struct xt_action_param *par)
> +{
> +	struct xt_hmark_info *info = (struct xt_hmark_info *)par->targinfo;
> +	struct ipv6hdr *ip6, _ip6;
> +	int poff, flag = IP6T_FH_F_AUTH; /* Ports offset, find_hdr flags */
> +	u32 addr1, addr2, hash, nhoffs = 0;
> +	u8 nexthdr;
> +	union hports uports = { .v32 = 0 };
> +	unsigned short fragoff = 0;
> +
> +	ip6 = (struct ipv6hdr *) (skb->data + skb_network_offset(skb));
> +
> +	/* Try to get transport header */

I like comments, but this is completely unnecessary, it's clear what
the function below does. Please, use comments only when necessary, ie.
in case that there's something which is not evident to the person that
is reading the code.

> +	nexthdr = ipv6_find_hdr(skb, &nhoffs, -1, &fragoff, &flag);
> +	if (nexthdr < 0)
> +		return XT_CONTINUE;
> +	/* don't check for icmp on fragments */
> +	if ((flag & IP6T_FH_F_FRAG) || (nexthdr != IPPROTO_ICMPV6))
> +		goto noicmp;
> +	/* ICMP: if an error then move ptr to inner header */
> +	if (get_inner6_hdr(skb, &nhoffs)) {
> +		/* Get IPv6 header ptr just to get the saddr & daddr later */
> +		ip6 = skb_header_pointer(skb, nhoffs, sizeof(_ip6), &_ip6);
> +		if (!ip6)
> +			return XT_CONTINUE;
> +		/* Treat AH as ESP */
> +		flag = IP6T_FH_F_AUTH;
> +		nexthdr = ipv6_find_hdr(skb, &nhoffs, -1, &fragoff, &flag);
> +		if (nexthdr < 0)
> +			return XT_CONTINUE;
> +	}
> +noicmp:
> +	/* Mask of the address and xor it into a u32 */
> +	addr1 = (__force u32)
> +		(ip6->saddr.s6_addr32[0] & info->smask.in6.s6_addr32[0]) ^
> +		(ip6->saddr.s6_addr32[1] & info->smask.in6.s6_addr32[1]) ^
> +		(ip6->saddr.s6_addr32[2] & info->smask.in6.s6_addr32[2]) ^
> +		(ip6->saddr.s6_addr32[3] & info->smask.in6.s6_addr32[3]);
> +	addr2 = (__force u32)
> +		(ip6->daddr.s6_addr32[0] & info->dmask.in6.s6_addr32[0]) ^
> +		(ip6->daddr.s6_addr32[1] & info->dmask.in6.s6_addr32[1]) ^
> +		(ip6->daddr.s6_addr32[2] & info->dmask.in6.s6_addr32[2]) ^
> +		(ip6->daddr.s6_addr32[3] & info->dmask.in6.s6_addr32[3]);
> +
> +	/* user space tool ensures that prmask is zero when method is L3*/

You can to double check this in checkentry.

> +	if ((info->flags & XT_F_HMARK_METHOD_L3) ||
> +	    (nexthdr == IPPROTO_ICMPV6))
> +		goto no6ports;
> +
> +	/* Is next header valid for port or SPI calculation ? */
> +	poff = proto_ports_offset(nexthdr);
> +	if ((flag & IP6T_FH_F_FRAG) || poff < 0)
> +		return XT_CONTINUE;
> +
> +	nhoffs += poff;
> +	/* Since uports is modified, skb_header_pointer() can't be used */
> +	if (!pskb_may_pull(skb, nhoffs + 4))
> +		return XT_CONTINUE;
> +	uports.v32 = * (__force u32 *) (skb->data + nhoffs);
> +
> +	if ((nexthdr == IPPROTO_ESP) || (nexthdr == IPPROTO_AH))
> +		uports.v32 = (uports.v32 & info->spimask) | info->spiset;
> +	else {
> +		uports.v32 = (uports.v32 & info->pmask.v32) | info->pset.v32;
> +		/* get a consistent hash (same value on both flow directions) */
> +		if (uports.p16.dst < uports.p16.src)
> +			swap(uports.p16.dst, uports.p16.src);
> +	}
> +
> +no6ports:
> +	nexthdr &= info->prmask;
> +	/* get a consistent hash (same value on both flow directions) */
> +	if (addr2 < addr1)
> +		swap(addr1, addr2);
> +
> +	hash = jhash_3words(addr1, addr2, uports.v32, info->hashrnd) ^ nexthdr;
> +	skb->mark = (hash % info->hmod) + info->hoffs;
> +	return XT_CONTINUE;
> +}
> +#endif
> +/*
> + * Calculate hash based fw-mark, on the five tuple if possible.
> + * special cases :
> + *  - Fragments do not use ports not even on the first fragment,
> + *    unless nf_defrag_xx.ko is used.
> + *  - On ICMP errors the inner header will be used.
> + *  - Tunnels no ports
> + *  - ESP & AH uses SPI
> + * @returns XT_CONTINUE
> + */
> +static unsigned int
> +hmark_v4(struct sk_buff *skb, const struct xt_action_param *par)
> +{
> +	struct xt_hmark_info *info = (struct xt_hmark_info *)par->targinfo;
> +	int nhoff, poff, frag = 0;
> +	struct iphdr *ip, _ip;
> +	u8 ip_proto;
> +	u32 addr1, addr2, hash;
> +	u16 snatport = 0, dnatport = 0;
> +	union hports uports;
> +#if defined(CONFIG_NF_NAT)

remove this #if defined, not required at all.

> +	enum ip_conntrack_info ctinfo;
> +	struct nf_conn *ct = ct = nf_ct_get(skb, &ctinfo);

                       ^^^^^^^^
please, this is redundant, no need for it. Remove it.

> +#endif
> +
> +	nhoff = skb_network_offset(skb);
> +	uports.v32 = 0;
> +
> +	ip = (struct iphdr *) (skb->data + nhoff);
> +	if (ip->protocol == IPPROTO_ICMP) {
> +		/* calc hash on inner header if an icmp error */
> +		nhoff = get_inner_hdr(skb, ip->ihl * 4, nhoff);
> +		ip = skb_header_pointer(skb, nhoff, sizeof(_ip), &_ip);
> +		if (!ip)
> +			return XT_CONTINUE;
> +	}
> +
> +	ip_proto = ip->protocol;
> +	if (ip->frag_off & htons(IP_MF | IP_OFFSET))
> +		frag = 1;
> +
> +	addr1 = (__force u32) ip->saddr & info->smask.ip;
> +	addr2 = (__force u32) ip->daddr & info->dmask.ip;
> +
> +#if defined(CONFIG_NF_NAT)
> +	if (ct && test_bit(IP_CT_IS_REPLY, &ct->status)) {
> +		struct nf_conntrack_tuple *otuple;
> +
> +		otuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
> +		/*
> +		 * On the "return flow", to get the original address
> +		 */
> +		if ((ct->status & IPS_DST_NAT) &&
> +			(info->flags & XT_HMARK_USE_DNAT)) {
> +			addr1 = (__force u32) otuple->dst.u3.in.s_addr;
> +			dnatport = otuple->dst.u.udp.port;
> +		}
> +		if ((ct->status & IPS_SRC_NAT) &&
> +			(info->flags & XT_HMARK_USE_SNAT)) {
> +			addr2 = (__force u32) otuple->src.u3.in.s_addr;
> +			snatport = otuple->src.u.udp.port;
> +		}

You can make this much more simple.

Allow the user to tell your HMARK target to use the conntrack
information instead.

My opinion is that the user must have total control on the target
behaviour through the configuration options. The number of internal
by-default decisions have to be kept up to the minimum, otherwise
the behaviour of the target may seem obscure.

> +	}
> +#endif
> +	/* user space tool ensures that prmask is zero when method is L3*/

No, you have to double check this in checkentry() in kernel-space to
make sure that user-space.

I think we need another round for this.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists