Patchwork [RFC,v2] netfilter: add connlabel conntrack extension

login
register
mail settings
Submitter Florian Westphal
Date Nov. 2, 2012, 12:43 p.m.
Message ID <1351860231-5434-1-git-send-email-fw@strlen.de>
Download mbox | patch
Permalink /patch/196553/
State Superseded
Headers show

Comments

Florian Westphal - Nov. 2, 2012, 12:43 p.m.
similar to connmarks, except labels are bit-based; i.e.
all labels may be attached to a flow at the same time.

Up to 1024 labels are supported.
Mapping of bit-identifier to label name is done in userspace.

The extension is enabled at run-time once "-m connlabel" netfilter
rules are added.

If connlabel rules have been added, sizeof(void*) bytes are allocated
per conntrack.  This allows to store up to BITS_PER_LONG - 1
(i.e. 31 or 63) labels without allocating additional memory.

When this space no longer suffices, additional memory is allocated on
demand.

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 Still sending this as an RFC as the ctnetlink glue is
 still missing.  I plan to work on ctnetlink integration once
 the extension is deemed to be ok.

 Changes since V1:
  - label <-> bit mapping moved to userspace
  - CONNLABEL target removed in favor of "-m connlabel --label foo --set"

 userspace patch is at
 http://git.breakpoint.cc/cgit.cgi/fw/iptables.git/commit/?h=nfct_ext_clabel_02&id=e1ebe4798b206558541d7e50784ec8214c1d7f75

Current operation:
- extension is enabled at run-time when -m connlabel rules are added
  (uses a per-ns refcnt to decide if extension needs to be allocated).

- instead of allocating the entire possible bit-store area (128 bytes)
  in the extension, only a pointer is allocated. label-bits are encoded into
  the pointer.  Once that size is exceeded (i.e., we're asked to set a
  bit >= BITS_PER_LONG - 1), we kmalloc a larger storage area.

  Lowest bit of pointer determines if the area is kmalloc'd or not.
  I've decided to do this because I think most users of this extension
  won't be using more than, say, 10 labels.

- testing if a label is set is guarded by rcu; setting a label is guarded
  by ct->lock.

Further plans:
 - extend ctnetlink to send a label bit-vector to userspace, or
   remove/attach labels from/to connections.

 This would also require extending libnetfilter_conntrack to provide
 some meaningful abstraction; I'll send a separate email with an API
 proposal before working on this, though.

Deferred/shelved for now:
 Support for "enumerated labels" (ie. labels that cannot overlap)
 is missing, but this could easily be added later on.

 For instance, we could change the pointer-encoding to support
 2^7 (i.e. 127 non-overlapping enumerated labels) plus
 24 (or 56 on 64 bit systems) overlapping labels.

 When more are requested, we can fallback to kmalloc() as we do now,
 there is enough room in struct __nf_conn_labels_rcu_ptr to
 add a "u16 enumerated_label".

 include/net/netfilter/nf_conntrack_extend.h |    4 +
 include/net/netfilter/nf_conntrack_labels.h |   65 ++++++++++++
 include/net/netns/conntrack.h               |    3 +
 include/uapi/linux/netfilter/xt_connlabel.h |   13 +++
 net/netfilter/Kconfig                       |   16 +++
 net/netfilter/Makefile                      |    2 +
 net/netfilter/nf_conntrack_core.c           |   11 ++
 net/netfilter/nf_conntrack_labels.c         |  143 +++++++++++++++++++++++++++
 net/netfilter/nf_conntrack_netlink.c        |    3 +
 net/netfilter/xt_connlabel.c                |   77 ++++++++++++++
 10 files changed, 337 insertions(+), 0 deletions(-)
 create mode 100644 include/net/netfilter/nf_conntrack_labels.h
 create mode 100644 include/uapi/linux/netfilter/xt_connlabel.h
 create mode 100644 net/netfilter/nf_conntrack_labels.c
 create mode 100644 net/netfilter/xt_connlabel.c
Florian Westphal - Nov. 7, 2012, 8:04 p.m.
Florian Westphal <fw@strlen.de> wrote:
> Further plans:
>  - extend ctnetlink to send a label bit-vector to userspace, or
>    remove/attach labels from/to connections.

I've implemented this via CTA_LABELS attribute, which
is a unsigned long[] blob; each bit set indicates that the
connlabel is set on the given connection.

CTA_LABELS is sent to userspace via ctnetlink, it may also be
used to replace the labels currently assigned to a connection
by sending request with CTA_LABELS attribute set to the kernel.

>  This would also require extending libnetfilter_conntrack to provide
>  some meaningful abstraction; I'll send a separate email with an API
>  proposal before working on this, though.

I propose to add following API calls:

int nfct_label_set(struct nf_conntrack *ct, const char *label);

	sets the label 'label' on the ct object.

void nfct_label_unset(struct nf_conntrack *ct, const char *label);

	opposite, label is cleared if it was set.

int nfct_label_get_max(const struct nf_conntrack *ct);

	returns the highest label-bit currently set on the connection,
	or -1 if none is set.

int nfct_label_get(const struct nf_conntrack *ct, int bit, char *buf,
				size_t len);

fills buf (up to size len) with the name of the label identified
by 'bit', if it is currently set on the conntrack.

returns -1 on error (i.e., label was not set), else length of the name.

Can be used together with nfct_label_get_max() to iterate over all the
labels set on the object, e.g. something like

for i = 0; i < nfct_label_get_max(ct); i++
	if (nfct_label_get(ct, i, buf, len) > 0)
		printf("label: %s (bit %d)\n", buf, i);

open question is how the library should do the mapping, i.e.
should it hard-code a path to the mapping file (currently
its /etc/xtables/connlabel.conf in my iptables-patch).

I think exposing it makes no sense since noone could know
where the file would be located on a given system.

Also, should we add calls to iterate of the entire set of
configured labels, e.g. something like

void *nfct_label_open(void);
void nfct_label_close(void *);
int nfct_label_iterate(void *fp, char *buf, size_t buflen);

so you could do
void *h = nfct_label_open();
int bit;
while ((bit = nfct_label_iterate(h, bufm sizeof(buf))) > 0)
	printf("bit %d: %s\n", bit, buf);

?

If there are no objections/suggestions i'll have a stab at adding
this to libnetfilter_conntrack.

Thanks,
Florian
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 12, 2012, 6:44 a.m.
Hi Florian,

On Wed, Nov 07, 2012 at 09:04:28PM +0100, Florian Westphal wrote:
> Florian Westphal <fw@strlen.de> wrote:
> > Further plans:
> >  - extend ctnetlink to send a label bit-vector to userspace, or
> >    remove/attach labels from/to connections.
> 
> I've implemented this via CTA_LABELS attribute, which
> is a unsigned long[] blob; each bit set indicates that the
> connlabel is set on the given connection.
> 
> CTA_LABELS is sent to userspace via ctnetlink, it may also be
> used to replace the labels currently assigned to a connection
> by sending request with CTA_LABELS attribute set to the kernel.
> 
> >  This would also require extending libnetfilter_conntrack to provide
> >  some meaningful abstraction; I'll send a separate email with an API
> >  proposal before working on this, though.
> 
> I propose to add following API calls:
> 
> int nfct_label_set(struct nf_conntrack *ct, const char *label);
> 
> 	sets the label 'label' on the ct object.
> 
> void nfct_label_unset(struct nf_conntrack *ct, const char *label);
> 
> 	opposite, label is cleared if it was set.

Can you use the existing nfct_attr_set/unset API for this?

> int nfct_label_get_max(const struct nf_conntrack *ct);
> 
> 	returns the highest label-bit currently set on the connection,
> 	or -1 if none is set.
> 
> int nfct_label_get(const struct nf_conntrack *ct, int bit, char *buf,
> 				size_t len);
> 
> fills buf (up to size len) with the name of the label identified
> by 'bit', if it is currently set on the conntrack.
> 
> returns -1 on error (i.e., label was not set), else length of the name.
> 
> Can be used together with nfct_label_get_max() to iterate over all the
> labels set on the object, e.g. something like
> 
> for i = 0; i < nfct_label_get_max(ct); i++
> 	if (nfct_label_get(ct, i, buf, len) > 0)
> 		printf("label: %s (bit %d)\n", buf, i);
> 
> open question is how the library should do the mapping, i.e.
> should it hard-code a path to the mapping file (currently
> its /etc/xtables/connlabel.conf in my iptables-patch).

Add a new parameter to nfct_label_open to indicate where the file that
contains the mapping is.

> I think exposing it makes no sense since noone could know
> where the file would be located on a given system.
> 
> Also, should we add calls to iterate of the entire set of
> configured labels, e.g. something like
> 
> void *nfct_label_open(void);
> void nfct_label_close(void *);
> int nfct_label_iterate(void *fp, char *buf, size_t buflen);
> 
> so you could do
> void *h = nfct_label_open();
> int bit;
> while ((bit = nfct_label_iterate(h, bufm sizeof(buf))) > 0)
> 	printf("bit %d: %s\n", bit, buf);
> 
> ?

That seems fine to me.

> If there are no objections/suggestions i'll have a stab at adding
> this to libnetfilter_conntrack.
> 
> Thanks,
> Florian
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 12, 2012, 6:50 a.m.
Hi Florian,

On Fri, Nov 02, 2012 at 01:43:51PM +0100, Florian Westphal wrote:
> similar to connmarks, except labels are bit-based; i.e.
> all labels may be attached to a flow at the same time.
> 
> Up to 1024 labels are supported.
> Mapping of bit-identifier to label name is done in userspace.
> 
> The extension is enabled at run-time once "-m connlabel" netfilter
> rules are added.
> 
> If connlabel rules have been added, sizeof(void*) bytes are allocated
> per conntrack.  This allows to store up to BITS_PER_LONG - 1
> (i.e. 31 or 63) labels without allocating additional memory.
> 
> When this space no longer suffices, additional memory is allocated on
> demand.
> 
> Signed-off-by: Florian Westphal <fw@strlen.de>
> ---
>  Still sending this as an RFC as the ctnetlink glue is
>  still missing.  I plan to work on ctnetlink integration once
>  the extension is deemed to be ok.
> 
>  Changes since V1:
>   - label <-> bit mapping moved to userspace
>   - CONNLABEL target removed in favor of "-m connlabel --label foo --set"
> 
>  userspace patch is at
>  http://git.breakpoint.cc/cgit.cgi/fw/iptables.git/commit/?h=nfct_ext_clabel_02&id=e1ebe4798b206558541d7e50784ec8214c1d7f75
> 
> Current operation:
> - extension is enabled at run-time when -m connlabel rules are added
>   (uses a per-ns refcnt to decide if extension needs to be allocated).
> 
> - instead of allocating the entire possible bit-store area (128 bytes)
>   in the extension, only a pointer is allocated. label-bits are encoded into
>   the pointer.  Once that size is exceeded (i.e., we're asked to set a
>   bit >= BITS_PER_LONG - 1), we kmalloc a larger storage area.
> 
>   Lowest bit of pointer determines if the area is kmalloc'd or not.
>   I've decided to do this because I think most users of this extension
>   won't be using more than, say, 10 labels.
> 
> - testing if a label is set is guarded by rcu; setting a label is guarded
>   by ct->lock.
> 
> Further plans:
>  - extend ctnetlink to send a label bit-vector to userspace, or
>    remove/attach labels from/to connections.
> 
>  This would also require extending libnetfilter_conntrack to provide
>  some meaningful abstraction; I'll send a separate email with an API
>  proposal before working on this, though.
> 
> Deferred/shelved for now:
>  Support for "enumerated labels" (ie. labels that cannot overlap)
>  is missing, but this could easily be added later on.
> 
>  For instance, we could change the pointer-encoding to support
>  2^7 (i.e. 127 non-overlapping enumerated labels) plus
>  24 (or 56 on 64 bit systems) overlapping labels.
> 
>  When more are requested, we can fallback to kmalloc() as we do now,
>  there is enough room in struct __nf_conn_labels_rcu_ptr to
>  add a "u16 enumerated_label".
> 
>  include/net/netfilter/nf_conntrack_extend.h |    4 +
>  include/net/netfilter/nf_conntrack_labels.h |   65 ++++++++++++
>  include/net/netns/conntrack.h               |    3 +
>  include/uapi/linux/netfilter/xt_connlabel.h |   13 +++
>  net/netfilter/Kconfig                       |   16 +++
>  net/netfilter/Makefile                      |    2 +
>  net/netfilter/nf_conntrack_core.c           |   11 ++
>  net/netfilter/nf_conntrack_labels.c         |  143 +++++++++++++++++++++++++++
>  net/netfilter/nf_conntrack_netlink.c        |    3 +
>  net/netfilter/xt_connlabel.c                |   77 ++++++++++++++
>  10 files changed, 337 insertions(+), 0 deletions(-)
>  create mode 100644 include/net/netfilter/nf_conntrack_labels.h
>  create mode 100644 include/uapi/linux/netfilter/xt_connlabel.h
>  create mode 100644 net/netfilter/nf_conntrack_labels.c
>  create mode 100644 net/netfilter/xt_connlabel.c
> 
> diff --git a/include/net/netfilter/nf_conntrack_extend.h b/include/net/netfilter/nf_conntrack_extend.h
> index 8b4d1fc2..977bc8a 100644
> --- a/include/net/netfilter/nf_conntrack_extend.h
> +++ b/include/net/netfilter/nf_conntrack_extend.h
> @@ -23,6 +23,9 @@ enum nf_ct_ext_id {
>  #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
>  	NF_CT_EXT_TIMEOUT,
>  #endif
> +#ifdef CONFIG_NF_CONNTRACK_LABELS
> +	NF_CT_EXT_LABELS,
> +#endif
>  	NF_CT_EXT_NUM,
>  };
>  
> @@ -33,6 +36,7 @@ enum nf_ct_ext_id {
>  #define NF_CT_EXT_ZONE_TYPE struct nf_conntrack_zone
>  #define NF_CT_EXT_TSTAMP_TYPE struct nf_conn_tstamp
>  #define NF_CT_EXT_TIMEOUT_TYPE struct nf_conn_timeout
> +#define NF_CT_EXT_LABELS_TYPE struct nf_conn_labels
>  
>  /* Extensions: optional stuff which isn't permanently in struct. */
>  struct nf_ct_ext {
> diff --git a/include/net/netfilter/nf_conntrack_labels.h b/include/net/netfilter/nf_conntrack_labels.h
> new file mode 100644
> index 0000000..dfcad1c
> --- /dev/null
> +++ b/include/net/netfilter/nf_conntrack_labels.h
> @@ -0,0 +1,65 @@
> +#include <linux/types.h>
> +#include <net/net_namespace.h>
> +#include <linux/netfilter/nf_conntrack_common.h>
> +#include <linux/netfilter/nf_conntrack_tuple_common.h>
> +#include <net/netfilter/nf_conntrack.h>
> +#include <net/netfilter/nf_conntrack_extend.h>
> +
> +#include <uapi/linux/netfilter/xt_connlabel.h>
> +
> +#define NF_CONNLABEL_NOT_ALLOCATED 1
> +
> +struct __nf_conn_labels_rcu_ptr {
> +	struct rcu_head rcu;
> +	u16 maxbit;
> +	unsigned long bits[];
> +};
> +
> +struct nf_conn_labels {
> +	struct __nf_conn_labels_rcu_ptr __rcu *label_bits;
> +};
> +
> +static inline bool __nf_ct_labels_allocated(struct __nf_conn_labels_rcu_ptr *p)
> +{
> +	return ((unsigned long) p & NF_CONNLABEL_NOT_ALLOCATED) == 0;
> +}
> +
> +static inline u16 __nf_ct_labels_get_maxbit(struct __nf_conn_labels_rcu_ptr *p)
> +{
> +	if (__nf_ct_labels_allocated(p))
> +		return p->maxbit;
> +	return BITS_PER_LONG - 1;
> +}
> +
> +static inline struct nf_conn_labels *nf_ct_labels_find(const struct nf_conn *ct)
> +{
> +#ifdef CONFIG_NF_CONNTRACK_LABELS
> +	return nf_ct_ext_find(ct, NF_CT_EXT_LABELS);
> +#endif
> +}
> +
> +static inline struct nf_conn_labels *nf_ct_labels_ext_add(struct nf_conn *ct)
> +{
> +#ifdef CONFIG_NF_CONNTRACK_LABELS
> +	struct nf_conn_labels *cl_ext;
> +	struct net *net = nf_ct_net(ct);
> +
> +	if (net->ct.labels_used == 0)
> +		return NULL;
> +
> +	cl_ext = nf_ct_ext_add(ct, NF_CT_EXT_LABELS, GFP_ATOMIC);
> +	if (cl_ext == NULL)
> +		return NULL;
> +
> +	RCU_INIT_POINTER(cl_ext->label_bits, (void *) NF_CONNLABEL_NOT_ALLOCATED);
> +	return cl_ext;
> +#else
> +	return NULL;
> +#endif
> +}
> +
> +void nf_conntrack_labels_fini(struct net *net);
> +int nf_conntrack_labels_init(struct net *net);
> +
> +bool nf_connlabel_match(const struct nf_conn *ct, u16 bit);
> +int nf_connlabel_set(struct nf_conn *ct, u16 bit);
> diff --git a/include/net/netns/conntrack.h b/include/net/netns/conntrack.h
> index a1d83cc..e1f0935 100644
> --- a/include/net/netns/conntrack.h
> +++ b/include/net/netns/conntrack.h
> @@ -83,6 +83,9 @@ struct netns_ct {
>  	int			sysctl_auto_assign_helper;
>  	bool			auto_assign_helper_warned;
>  	struct nf_ip_net	nf_ct_proto;
> +#if defined(CONFIG_NF_CONNTRACK_LABELS)
> +	unsigned int labels_used;
> +#endif
>  #ifdef CONFIG_NF_NAT_NEEDED
>  	struct hlist_head	*nat_bysource;
>  	unsigned int		nat_htable_size;
> diff --git a/include/uapi/linux/netfilter/xt_connlabel.h b/include/uapi/linux/netfilter/xt_connlabel.h
> new file mode 100644
> index 0000000..768739b
> --- /dev/null
> +++ b/include/uapi/linux/netfilter/xt_connlabel.h
> @@ -0,0 +1,13 @@
> +#include <linux/types.h>
> +
> +#define XT_CONNLABEL_MAXBIT 1023
> +enum xt_connlabel_mtopts {
> +	XT_CONNLABEL_OP_INVERT = 1 << 0,
> +	XT_CONNLABEL_OP_SET    = 1 << 1,
> +};
> +
> +struct xt_connlabel_mtinfo {
> +	__u16 bit;
> +	__u16 options;
> +};
> +
> diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
> index fefa514..80a0fdd 100644
> --- a/net/netfilter/Kconfig
> +++ b/net/netfilter/Kconfig
> @@ -124,6 +124,15 @@ config NF_CONNTRACK_TIMESTAMP
>  
>  	  If unsure, say `N'.
>  
> +config NF_CONNTRACK_LABELS
> +	bool  'Connection tracking labels'
> +	depends on NETFILTER_ADVANCED
> +	help
> +	  This option enables support for assigning user-defined text strings
> +	  to connection tracking entries.
> +
> +	  If unsure, say `N'.
> +
>  config NF_CT_PROTO_DCCP
>  	tristate 'DCCP protocol connection tracking support (EXPERIMENTAL)'
>  	depends on EXPERIMENTAL
> @@ -472,6 +481,13 @@ config NETFILTER_XT_SET
>  
>  	  To compile it as a module, choose M here.  If unsure, say N.
>  
> +config NETFILTER_XT_CONNLABELS
> +	tristate '"connlimit" target and match support"'
> +	depends on NF_CONNTRACK_LABELS
> +	---help---
> +	  This match allows you to match against connlabels assigned to the
> +	  connection.
> +
>  # alphabetically ordered list of targets
>  
>  comment "Xtables targets"
> diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
> index 3259697..0e7139c 100644
> --- a/net/netfilter/Makefile
> +++ b/net/netfilter/Makefile
> @@ -4,6 +4,7 @@ nf_conntrack-y	:= nf_conntrack_core.o nf_conntrack_standalone.o nf_conntrack_exp
>  nf_conntrack-$(CONFIG_NF_CONNTRACK_TIMEOUT) += nf_conntrack_timeout.o
>  nf_conntrack-$(CONFIG_NF_CONNTRACK_TIMESTAMP) += nf_conntrack_timestamp.o
>  nf_conntrack-$(CONFIG_NF_CONNTRACK_EVENTS) += nf_conntrack_ecache.o
> +nf_conntrack-$(CONFIG_NF_CONNTRACK_LABELS) += nf_conntrack_labels.o
>  
>  obj-$(CONFIG_NETFILTER) = netfilter.o
>  
> @@ -70,6 +71,7 @@ obj-$(CONFIG_NETFILTER_XTABLES) += x_tables.o xt_tcpudp.o
>  obj-$(CONFIG_NETFILTER_XT_MARK) += xt_mark.o
>  obj-$(CONFIG_NETFILTER_XT_CONNMARK) += xt_connmark.o
>  obj-$(CONFIG_NETFILTER_XT_SET) += xt_set.o
> +obj-$(CONFIG_NETFILTER_XT_CONNLABELS) += xt_connlabel.o
>  obj-$(CONFIG_NF_NAT) += xt_nat.o
>  
>  # targets
> diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
> index 0f241be..c19e25c 100644
> --- a/net/netfilter/nf_conntrack_core.c
> +++ b/net/netfilter/nf_conntrack_core.c
> @@ -45,6 +45,7 @@
>  #include <net/netfilter/nf_conntrack_zones.h>
>  #include <net/netfilter/nf_conntrack_timestamp.h>
>  #include <net/netfilter/nf_conntrack_timeout.h>
> +#include <net/netfilter/nf_conntrack_labels.h>
>  #include <net/netfilter/nf_nat.h>
>  #include <net/netfilter/nf_nat_core.h>
>  
> @@ -770,6 +771,7 @@ void nf_conntrack_free(struct nf_conn *ct)
>  }
>  EXPORT_SYMBOL_GPL(nf_conntrack_free);
>  
> +
>  /* Allocate a new conntrack: we return -ENOMEM if classification
>     failed due to stress.  Otherwise it really is unclassifiable. */
>  static struct nf_conntrack_tuple_hash *
> @@ -816,6 +818,7 @@ init_conntrack(struct net *net, struct nf_conn *tmpl,
>  
>  	nf_ct_acct_ext_add(ct, GFP_ATOMIC);
>  	nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
> +	nf_ct_labels_ext_add(ct);
>  
>  	ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL;
>  	nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0,
> @@ -1585,7 +1588,15 @@ static int nf_conntrack_init_net(struct net *net)
>  	ret = nf_conntrack_helper_init(net);
>  	if (ret < 0)
>  		goto err_helper;
> +
> +	ret = nf_conntrack_labels_init(net);
> +	if (ret < 0)
> +		goto err_labels;
> +
>  	return 0;
> +
> +err_labels:
> +	nf_conntrack_helper_fini(net);
>  err_helper:
>  	nf_conntrack_timeout_fini(net);
>  err_timeout:
> diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
> new file mode 100644
> index 0000000..eab398b
> --- /dev/null
> +++ b/net/netfilter/nf_conntrack_labels.c
> @@ -0,0 +1,143 @@
> +#include <linux/ctype.h>
> +#include <linux/export.h>
> +#include <linux/jhash.h>
> +#include <linux/spinlock.h>
> +#include <linux/types.h>
> +#include <linux/slab.h>
> +
> +#include <net/netfilter/nf_conntrack_ecache.h>
> +#include <net/netfilter/nf_conntrack_labels.h>
> +
> +static int labels_set_realloc(struct nf_conn_labels *l,
> +			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)

I think we can simplify this code if we use the CT target to set the
number of labels that we'll use, so we skip allocations in runtime and
possible reallocation.

... -t raw -j CT --labels 32

> +{
> +	struct __nf_conn_labels_rcu_ptr *ptr;
> +	u32 word, i, size;
> +	u16 maxbit;
> +
> +	size = round_up(bit+1, BITS_PER_LONG) / BITS_PER_LONG;
> +	size *= sizeof(long);
> +
> +	ptr = kmalloc(sizeof(*ptr) + size, GFP_ATOMIC);
> +	if (!ptr)
> +		return -ENOMEM;
> +
> +	if (__nf_ct_labels_allocated(oldptr)) {
> +		word = BIT_WORD(oldptr->maxbit);
> +		for (i = 0; i <= word; i++)
> +			ptr->bits[i] = oldptr->bits[i];
> +		kfree_rcu(oldptr, rcu);
> +	} else {
> +		unsigned long tmp = (unsigned long) oldptr;
> +		ptr->bits[0] = tmp >> NF_CONNLABEL_NOT_ALLOCATED;
> +		i = 1;
> +	}
> +
> +	maxbit = (size * BITS_PER_BYTE) - 1;
> +	word = BIT_WORD(maxbit);
> +	for (;i <= word; i++)
> +		ptr->bits[i] = 0;
> +
> +	set_bit(bit, ptr->bits);
> +	ptr->maxbit = maxbit;
> +	rcu_assign_pointer(l->label_bits, ptr);
> +	return 0;
> +}
> +
> +static bool __nf_connlabel_match(struct nf_conn_labels *labels, u16 bit)
> +{
> +	struct __nf_conn_labels_rcu_ptr *ptr = rcu_dereference(labels->label_bits);
> +
> +	if (!__nf_ct_labels_allocated(ptr)) {
> +		unsigned long tmp = (unsigned long) ptr >> NF_CONNLABEL_NOT_ALLOCATED;
> +		return bit < BITS_PER_LONG && test_bit(bit, &tmp);
> +	}
> +
> +	return bit <= ptr->maxbit && test_bit(bit, ptr->bits);
> +}
> +
> +bool nf_connlabel_match(const struct nf_conn *ct, u16 bit)
> +{
> +	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
> +	bool ret;
> +
> +	if (!labels)
> +		return false;
> +
> +	rcu_read_lock();
> +	ret = __nf_connlabel_match(labels, bit);
> +	rcu_read_unlock();
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(nf_connlabel_match);
> +
> +int nf_connlabel_set(struct nf_conn *ct, u16 bit)
> +{
> +	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
> +	struct __nf_conn_labels_rcu_ptr *ptr;
> +	int err;
> +
> +	if (!labels)
> +		return -ENOSPC;
> +
> +	rcu_read_lock();
> +	if (__nf_connlabel_match(labels, bit)) {
> +		rcu_read_unlock();
> +		return 0; /* fast path: already set */
> +	}
> +	rcu_read_unlock();
> +
> +	spin_lock(&ct->lock);
> +
> +	ptr = rcu_dereference_protected(labels->label_bits, 1);
> +	if (__nf_ct_labels_allocated(ptr)) {
> +		if (bit <= ptr->maxbit) {
> +			set_bit(bit, ptr->bits);
> +			err = 0;
> +		} else
> +			err = labels_set_realloc(labels, ptr, bit);
> +	} else if (bit < (BITS_PER_LONG - NF_CONNLABEL_NOT_ALLOCATED)) {
> +		unsigned long tmp = (unsigned long) ptr;
> +		bit = 1 << (bit + NF_CONNLABEL_NOT_ALLOCATED);
> +		rcu_assign_pointer(labels->label_bits, (void *) (tmp|bit));
> +		err = 0;
> +	} else {
> +		err = labels_set_realloc(labels, ptr, bit);
> +	}
> +
> +	spin_unlock(&ct->lock);
> +	return err;
> +}
> +EXPORT_SYMBOL_GPL(nf_connlabel_set);
> +
> +static void nf_ct_ext_label_destroy(struct nf_conn *ct)
> +{
> +	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
> +	if (labels) {
> +		struct __nf_conn_labels_rcu_ptr *p = rcu_dereference(labels->label_bits);
> +		if (__nf_ct_labels_allocated(p))
> +			kfree(p);
> +	}
> +}
> +
> +static struct nf_ct_ext_type labels_extend __read_mostly = {
> +	.len    = sizeof(struct nf_conn_labels),
> +	.align  = __alignof__(struct nf_conn_labels),
> +	.destroy = nf_ct_ext_label_destroy,
> +	.id     = NF_CT_EXT_LABELS,
> +};
> +
> +int nf_conntrack_labels_init(struct net *net)
> +{
> +	if (net_eq(net, &init_net))
> +		return nf_ct_extend_register(&labels_extend);
> +	return 0;
> +}
> +
> +void nf_conntrack_labels_fini(struct net *net)
> +{
> +	if (net_eq(net, &init_net))
> +		nf_ct_extend_unregister(&labels_extend);
> +}
> +
> diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c
> index 7bbfb3d..cf22362 100644
> --- a/net/netfilter/nf_conntrack_netlink.c
> +++ b/net/netfilter/nf_conntrack_netlink.c
> @@ -43,6 +43,7 @@
>  #include <net/netfilter/nf_conntrack_acct.h>
>  #include <net/netfilter/nf_conntrack_zones.h>
>  #include <net/netfilter/nf_conntrack_timestamp.h>
> +#include <net/netfilter/nf_conntrack_labels.h>
>  #ifdef CONFIG_NF_NAT_NEEDED
>  #include <net/netfilter/nf_nat_core.h>
>  #include <net/netfilter/nf_nat_l4proto.h>
> @@ -1489,6 +1490,8 @@ ctnetlink_create_conntrack(struct net *net, u16 zone,
>  	nf_ct_acct_ext_add(ct, GFP_ATOMIC);
>  	nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
>  	nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC);
> +	nf_ct_labels_ext_add(ct);
> +
>  	/* we must add conntrack extensions before confirmation. */
>  	ct->status |= IPS_CONFIRMED;
>  
> diff --git a/net/netfilter/xt_connlabel.c b/net/netfilter/xt_connlabel.c
> new file mode 100644
> index 0000000..40c92f3
> --- /dev/null
> +++ b/net/netfilter/xt_connlabel.c
> @@ -0,0 +1,77 @@
> +#include <linux/module.h>
> +#include <linux/skbuff.h>
> +#include <net/netfilter/nf_conntrack.h>
> +#include <net/netfilter/nf_conntrack_labels.h>
> +#include <linux/netfilter/x_tables.h>
> +
> +static bool
> +connlabel_mt(const struct sk_buff *skb, struct xt_action_param *par)
> +{
> +        const struct xt_connlabel_mtinfo *info = par->matchinfo;
> +	enum ip_conntrack_info ctinfo;
> +	struct nf_conn *ct;
> +	bool invert = info->options & XT_CONNLABEL_OP_INVERT;
> +
> +	ct = nf_ct_get(skb, &ctinfo);
> +	if (ct == NULL || nf_ct_is_untracked(ct))
> +		return invert;
> +
> +	if (info->options & XT_CONNLABEL_OP_SET)
> +		return (nf_connlabel_set(ct, info->bit) == 0) ^ invert;
> +
> +	return nf_connlabel_match(ct, info->bit) ^ invert;
> +}
> +
> +static int connlabel_mt_check(const struct xt_mtchk_param *par)
> +{
> +	const int options = XT_CONNLABEL_OP_INVERT |
> +			    XT_CONNLABEL_OP_SET;
> +	struct xt_connlabel_mtinfo *info = par->matchinfo;
> +	int ret;
> +
> +	if (info->bit > XT_CONNLABEL_MAXBIT)
> +		return -ERANGE;
> +
> +	if (info->options & ~options) {
> +		pr_err("Unknown options in mask %x\n", info->options);
> +		return -EINVAL;
> +	}
> +
> +	ret = nf_ct_l3proto_try_module_get(par->family);
> +	if (ret < 0) {
> +		pr_info("cannot load conntrack support for proto=%u\n", par->family);
> +		return ret;
> +	}
> +
> +	par->net->ct.labels_used++;
> +	return ret;
> +}
> +
> +static void connlabel_mt_destroy(const struct xt_mtdtor_param *par)
> +{
> +	par->net->ct.labels_used--;
> +	nf_ct_l3proto_module_put(par->family);
> +}
> +
> +static struct xt_match connlabels_mt_reg __read_mostly = {
> +        .name           = "connlabel",
> +        .family         = NFPROTO_UNSPEC,
> +        .checkentry     = connlabel_mt_check,
> +        .match          = connlabel_mt,
> +        .matchsize      = sizeof(struct xt_connlabel_mtinfo),
> +        .destroy        = connlabel_mt_destroy,
> +        .me             = THIS_MODULE,
> +};
> +
> +static int __init connlabel_mt_init(void)
> +{
> +	return xt_register_match(&connlabels_mt_reg);
> +}
> +
> +static void __exit connlabel_mt_exit(void)
> +{
> +	xt_unregister_match(&connlabels_mt_reg);
> +}
> +
> +module_init(connlabel_mt_init);
> +module_exit(connlabel_mt_exit);
> -- 
> 1.7.8.6
> 
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Florian Westphal - Nov. 12, 2012, 12:30 p.m.
Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > int nfct_label_set(struct nf_conntrack *ct, const char *label);
> > 
> > 	sets the label 'label' on the ct object.
> > 
> > void nfct_label_unset(struct nf_conntrack *ct, const char *label);
> > 
> > 	opposite, label is cleared if it was set.
> 
> Can you use the existing nfct_attr_set/unset API for this?

Don't see how, nfct_attr_unset clears the entire attribute.
I admit that it would be nice if one could re-use existing api...

Right now nfct_attr_set_l(.. ATTR_CONNLABELS..) can be used to assign
the bit-vector directly (and _get can be used to retrieve the
label bit-vector).

> > open question is how the library should do the mapping, i.e.
> > should it hard-code a path to the mapping file (currently
> > its /etc/xtables/connlabel.conf in my iptables-patch).
> 
> Add a new parameter to nfct_label_open to indicate where the file that
> contains the mapping is.

I think that we should avoid it; else this might become a
portability nightmare where applications end up
trying half a zillion names (/etc/xtables, /usr/etc, /usr/local...)

> > void *nfct_label_open(void);
> > void nfct_label_close(void *);
> > int nfct_label_iterate(void *fp, char *buf, size_t buflen);
> > 
> > so you could do
> > void *h = nfct_label_open();
> > int bit;
> > while ((bit = nfct_label_iterate(h, bufm sizeof(buf))) > 0)
> > 	printf("bit %d: %s\n", bit, buf);
> > 
> > ?
> 
> That seems fine to me.

Thanks.  I now have the following public API:

/**
 * nfct_label_set_bit - set a bit in the label bitvector
 * \param ct pointer to a valid conntrack object
 * \param bit bit to set
 * returns 0 if the bit is now set
 */
int nfct_label_set_bit(struct nf_conntrack *ct, uint16_t bit)
{
        return __label_set_bit(ct, bit);
}

/**
 * nfct_label_unset_bit - clear label bit on a conntrack object
 * \param ct pointer to a valid conntrack object
 * \param bit bit to clear
 */
void nfct_label_unset(struct nf_conntrack *ct, uint16_t bit)
{
        __label_unset_bit(ct, bit);
}

/**
 * nfct_label_test_bit - determine wheter bit is set
 * \param ct pointer to a valid conntrack object
 * \param bit bit to query
 *
 * returns 0 on sucess and -1 if bit is not set.
 */
int nfct_label_test_bit(struct nf_conntrack *ct, uint16_t bit)
{
        return __label_test_bit(ct, bit);
}

/**
 * nfct_label_get_max - get highest bit set in label vector
 * \param ct pointer to a valid conntrack object
 * returns highest bit set, or -1 if no bits are set.
 */
int nfct_label_get_max(struct nf_conntrack *ct)
{
        return __label_get_max(ct);
}

/**
 * nfct_label_open
 * \param open the default bit-name mapping file for reading.
 *
 * returns a FILE handle to be passed to nfct_label_iterate.
 */
void *nfct_label_open(void);

/**
 * nfct_label_iterate
 * \param handle handle obtained from nfct_label_open
 * \param buf buffer where name will be stored
 * \param len size of the buffer
 *
 * returns the bit the name is mapped to, or -1 when no more
 * labels are defined.  The handle should be closed via
 * fclose() once it is of no more interest.
 */
int nfct_label_iterate(void *handle, char *buf, size_t len);
/**
 * @}
 */

Since nfct_label_open() now returns FILE* userspace can provide other
mapping files as well [ nfct_label_open just looks at the default location ].

I've killed the functions that took a "const char *labelname" as
argument; it's error prone to fiddle with hidden file-descriptor
in the library (not thread-safe...).

We could later add another library that provides more sophisticated
stuff, such as a label<->bit cache.

The patch that i currently have (untested, so I won't send it yet...)
adds about 300 LOC, i would like to avoid bloating libnetfilter_conntrack
more than that.

Thanks,
Florian
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Florian Westphal - Nov. 12, 2012, 12:47 p.m.
Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
> > new file mode 100644
> > index 0000000..eab398b
> > --- /dev/null
> > +++ b/net/netfilter/nf_conntrack_labels.c
> > @@ -0,0 +1,143 @@
> > +#include <linux/ctype.h>
> > +#include <linux/export.h>
> > +#include <linux/jhash.h>
> > +#include <linux/spinlock.h>
> > +#include <linux/types.h>
> > +#include <linux/slab.h>
> > +
> > +#include <net/netfilter/nf_conntrack_ecache.h>
> > +#include <net/netfilter/nf_conntrack_labels.h>
> > +
> > +static int labels_set_realloc(struct nf_conn_labels *l,
> > +			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)
> 
> I think we can simplify this code if we use the CT target to set the
> number of labels that we'll use, so we skip allocations in runtime and
> possible reallocation.
> 
> ... -t raw -j CT --labels 32

I'm not convinced yet ;-)

I think we should avoid to make users fiddle with CT target options
just to get certain functionality working.

Alternative would be to keep track of highest bit requested in a "-m connlabel"
rule to figure out the needed size.

In any case, it would require adding "u16 len" to the extension area; else
we can't figure out how many bytes are valid, i.e.:

struct nf_conn_labels {
+       u16 size;		/* length of label storage */
+	unsigned long bits[];	/* variable-sized label storage */
+};

it would increase minimum length needed but it would avoid
the rcu dances done by the current scheme.

It this is ok for you I'll make this change to see how many LOC are
saved by this.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 12, 2012, 4:24 p.m.
Hi Florian,

On Mon, Nov 12, 2012 at 01:30:53PM +0100, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > > int nfct_label_set(struct nf_conntrack *ct, const char *label);
> > > 
> > > 	sets the label 'label' on the ct object.
> > > 
> > > void nfct_label_unset(struct nf_conntrack *ct, const char *label);
> > > 
> > > 	opposite, label is cleared if it was set.
> > 
> > Can you use the existing nfct_attr_set/unset API for this?
> 
> Don't see how, nfct_attr_unset clears the entire attribute.
> I admit that it would be nice if one could re-use existing api...
> 
> Right now nfct_attr_set_l(.. ATTR_CONNLABELS..) can be used to assign
> the bit-vector directly (and _get can be used to retrieve the
> label bit-vector).

You could use a nfct_bitmask object and methods to modify it:

struct nfct_bitmask *nfct_bitmask_alloc(void)
nfct_bitmask_set_bit(struct nfct_bitmask *, uint16_t bit)
nfct_bitmask_test_bit(struct nfct_bitmask *, uint16_t bit)
nfct_bitmask_unset_bit(struct nfct_bitmask *, uint16_t bit)

Then you can attach the bitmask to the object:

nfct_set_attr(ct, ATTR_CONNLABELS, bitmask);

The setter would do the special handling internally by attaching the
bitmask label to the ct object.

> > > open question is how the library should do the mapping, i.e.
> > > should it hard-code a path to the mapping file (currently
> > > its /etc/xtables/connlabel.conf in my iptables-patch).
> > 
> > Add a new parameter to nfct_label_open to indicate where the file that
> > contains the mapping is.
> 
> I think that we should avoid it; else this might become a
> portability nightmare where applications end up
> trying half a zillion names (/etc/xtables, /usr/etc, /usr/local...)

We can provide some constant to recommend a path. My only concern is
that the hardcoded path may not exist in some filesystem, thus,
forcing to recompile the library.

Anyway, I think the main user for this will be the conntrack utility
and any netfilter utilities.

One more question below:

> > > void *nfct_label_open(void);
> > > void nfct_label_close(void *);
> > > int nfct_label_iterate(void *fp, char *buf, size_t buflen);
> > > 
> > > so you could do
> > > void *h = nfct_label_open();
> > > int bit;
> > > while ((bit = nfct_label_iterate(h, bufm sizeof(buf))) > 0)
> > > 	printf("bit %d: %s\n", bit, buf);
> > > 
> > > ?
> > 
> > That seems fine to me.
> 
> Thanks.  I now have the following public API:
> 
> /**
>  * nfct_label_set_bit - set a bit in the label bitvector
>  * \param ct pointer to a valid conntrack object
>  * \param bit bit to set
>  * returns 0 if the bit is now set
>  */
> int nfct_label_set_bit(struct nf_conntrack *ct, uint16_t bit)
> {
>         return __label_set_bit(ct, bit);
> }
> 
> /**
>  * nfct_label_unset_bit - clear label bit on a conntrack object
>  * \param ct pointer to a valid conntrack object
>  * \param bit bit to clear
>  */
> void nfct_label_unset(struct nf_conntrack *ct, uint16_t bit)
> {
>         __label_unset_bit(ct, bit);
> }
> 
> /**
>  * nfct_label_test_bit - determine wheter bit is set
>  * \param ct pointer to a valid conntrack object
>  * \param bit bit to query
>  *
>  * returns 0 on sucess and -1 if bit is not set.
>  */
> int nfct_label_test_bit(struct nf_conntrack *ct, uint16_t bit)
> {
>         return __label_test_bit(ct, bit);
> }
> 
> /**
>  * nfct_label_get_max - get highest bit set in label vector
>  * \param ct pointer to a valid conntrack object
>  * returns highest bit set, or -1 if no bits are set.
>  */
> int nfct_label_get_max(struct nf_conntrack *ct)
> {
>         return __label_get_max(ct);
> }
> 
> /**
>  * nfct_label_open
>  * \param open the default bit-name mapping file for reading.
>  *
>  * returns a FILE handle to be passed to nfct_label_iterate.
>  */
> void *nfct_label_open(void);
> 
> /**
>  * nfct_label_iterate
>  * \param handle handle obtained from nfct_label_open
>  * \param buf buffer where name will be stored
>  * \param len size of the buffer
>  *
>  * returns the bit the name is mapped to, or -1 when no more
>  * labels are defined.  The handle should be closed via
>  * fclose() once it is of no more interest.
>  */
> int nfct_label_iterate(void *handle, char *buf, size_t len);

One question, what is void *handle? Is it hidding a FILE * or a cache
containing the bitmask?
 
You can define some:

struct nfct_label_handle;

in libnetfilter_conntrack.h, and keep the layout private by defining
it in some internal header.

I'd go for the cache-based approach, ie. store a lookup array in
memory extracted from the file.

> /**
>  * @}
>  */
> 
> Since nfct_label_open() now returns FILE* userspace can provide other
> mapping files as well [ nfct_label_open just looks at the default location ].
> 
> I've killed the functions that took a "const char *labelname" as
> argument; it's error prone to fiddle with hidden file-descriptor
> in the library (not thread-safe...).
> 
> We could later add another library that provides more sophisticated
> stuff, such as a label<->bit cache.
> 
> The patch that i currently have (untested, so I won't send it yet...)
> adds about 300 LOC, i would like to avoid bloating libnetfilter_conntrack
> more than that.

Examples files usually help to see how the API looks in action, if you
can attach one in the next round, in would be great.

Regards.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Florian Westphal - Nov. 12, 2012, 4:32 p.m.
Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > Don't see how, nfct_attr_unset clears the entire attribute.
> > I admit that it would be nice if one could re-use existing api...
> > 
> > Right now nfct_attr_set_l(.. ATTR_CONNLABELS..) can be used to assign
> > the bit-vector directly (and _get can be used to retrieve the
> > label bit-vector).
> 
> You could use a nfct_bitmask object and methods to modify it:
> 
> struct nfct_bitmask *nfct_bitmask_alloc(void)
> nfct_bitmask_set_bit(struct nfct_bitmask *, uint16_t bit)
> nfct_bitmask_test_bit(struct nfct_bitmask *, uint16_t bit)
> nfct_bitmask_unset_bit(struct nfct_bitmask *, uint16_t bit)

$ git grep nfct_bitmask
zsh: exit 1

What library is that? My search-engine fu fails.

> Then you can attach the bitmask to the object:
> 
> nfct_set_attr(ct, ATTR_CONNLABELS, bitmask);
> 
> The setter would do the special handling internally by attaching the
> bitmask label to the ct object.

Sounds good.

> > I think that we should avoid it; else this might become a
> > portability nightmare where applications end up
> > trying half a zillion names (/etc/xtables, /usr/etc, /usr/local...)
> 
> We can provide some constant to recommend a path. My only concern is
> that the hardcoded path may not exist in some filesystem, thus,
> forcing to recompile the library.
> 
> Anyway, I think the main user for this will be the conntrack utility
> and any netfilter utilities.

True.

> I'd go for the cache-based approach, ie. store a lookup array in
> memory extracted from the file.

Oh.  I wanted to avoid this to not bloat the library.
I'll add a cache.

> Examples files usually help to see how the API looks in action, if you
> can attach one in the next round, in would be great.

Will do.

Thanks a lot for your help!
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 12, 2012, 7:02 p.m.
On Mon, Nov 12, 2012 at 05:32:08PM +0100, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > > Don't see how, nfct_attr_unset clears the entire attribute.
> > > I admit that it would be nice if one could re-use existing api...
> > > 
> > > Right now nfct_attr_set_l(.. ATTR_CONNLABELS..) can be used to assign
> > > the bit-vector directly (and _get can be used to retrieve the
> > > label bit-vector).
> > 
> > You could use a nfct_bitmask object and methods to modify it:
> > 
> > struct nfct_bitmask *nfct_bitmask_alloc(void)
> > nfct_bitmask_set_bit(struct nfct_bitmask *, uint16_t bit)
> > nfct_bitmask_test_bit(struct nfct_bitmask *, uint16_t bit)
> > nfct_bitmask_unset_bit(struct nfct_bitmask *, uint16_t bit)
> 
> $ git grep nfct_bitmask
> zsh: exit 1
> 
> What library is that? My search-engine fu fails.

It doesn't exist so far :-), but I think it should not require much
effort by using the existing include/internal/bitops.h in
libnetfilter_conntrack.

> > Then you can attach the bitmask to the object:
> > 
> > nfct_set_attr(ct, ATTR_CONNLABELS, bitmask);
> > 
> > The setter would do the special handling internally by attaching the
> > bitmask label to the ct object.
> 
> Sounds good.
> 
> > > I think that we should avoid it; else this might become a
> > > portability nightmare where applications end up
> > > trying half a zillion names (/etc/xtables, /usr/etc, /usr/local...)
> > 
> > We can provide some constant to recommend a path. My only concern is
> > that the hardcoded path may not exist in some filesystem, thus,
> > forcing to recompile the library.
> > 
> > Anyway, I think the main user for this will be the conntrack utility
> > and any netfilter utilities.
> 
> True.
> 
> > I'd go for the cache-based approach, ie. store a lookup array in
> > memory extracted from the file.
> 
> Oh.  I wanted to avoid this to not bloat the library.
> I'll add a cache.
> 
> > Examples files usually help to see how the API looks in action, if you
> > can attach one in the next round, in would be great.
> 
> Will do.
> 
> Thanks a lot for your help!

Regards.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 15, 2012, 12:13 p.m.
Hi Florian,

On Mon, Nov 12, 2012 at 01:47:05PM +0100, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > > diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
> > > new file mode 100644
> > > index 0000000..eab398b
> > > --- /dev/null
> > > +++ b/net/netfilter/nf_conntrack_labels.c
> > > @@ -0,0 +1,143 @@
> > > +#include <linux/ctype.h>
> > > +#include <linux/export.h>
> > > +#include <linux/jhash.h>
> > > +#include <linux/spinlock.h>
> > > +#include <linux/types.h>
> > > +#include <linux/slab.h>
> > > +
> > > +#include <net/netfilter/nf_conntrack_ecache.h>
> > > +#include <net/netfilter/nf_conntrack_labels.h>
> > > +
> > > +static int labels_set_realloc(struct nf_conn_labels *l,
> > > +			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)
> > 
> > I think we can simplify this code if we use the CT target to set the
> > number of labels that we'll use, so we skip allocations in runtime and
> > possible reallocation.
> > 
> > ... -t raw -j CT --labels 32
> 
> I'm not convinced yet ;-)
> 
> I think we should avoid to make users fiddle with CT target options
> just to get certain functionality working.

I agree that we should try to keep things easy for users.

Still, since the conntrack helper discussion during the last workshop,
I think that users should explicitly enable conntrack features they
want via iptables.

In that direction, I've been toying with some patches to explicitly
enable connectiong tracking via the CT target, ie. instead of tracking
everything by default and using NOTRACK to say what you don't what
(like we do now), tell what you want to track via some explict rule.
PF people are doing it that way.

Still that's an important semantic change so we'll have to keep some
compatibility mode for some time (we would do the conntrack lookup in
the raw table, and follow-up rules will find the ct object). That also
would change the assumption that "you have no ct object in the raw
table" to "you get the ct object just after the -j CT
--enable-tracking rule".

Note that we would have to rename (via aliasing) the raw table to
make it become a filter table in the PREROUTING chain once that
semantic change happens.

As result of that change, it would allow us to remove the current
template logic from:

1) attach the template
2) use the template in nf_conntrack_in() to set on features we requested

to:

1) lookup/create ct object and attach features requested by the user
   all at once.

I'm going to find some spare time to send a RFC patch for this.

> Alternative would be to keep track of highest bit requested in a "-m connlabel"
> rule to figure out the needed size.
> 
> In any case, it would require adding "u16 len" to the extension area; else
> we can't figure out how many bytes are valid, i.e.:
> 
> struct nf_conn_labels {
> +       u16 size;		/* length of label storage */
> +	unsigned long bits[];	/* variable-sized label storage */
> +};
> 
> it would increase minimum length needed but it would avoid
> the rcu dances done by the current scheme.
> 
> It this is ok for you I'll make this change to see how many LOC are
> saved by this.

If we simplify the current connlabel code it would be great. And so
far, the only way I can think to obtain that is to explictly specify
via the CT target the length of the label.

Regards.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Florian Westphal - Nov. 15, 2012, 12:50 p.m.
Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> On Mon, Nov 12, 2012 at 01:47:05PM +0100, Florian Westphal wrote:
> > Alternative would be to keep track of highest bit requested in a "-m connlabel"
> > rule to figure out the needed size.
> > 
> > In any case, it would require adding "u16 len" to the extension area; else
> > we can't figure out how many bytes are valid, i.e.:
> > 
> > struct nf_conn_labels {
> > +       u16 size;		/* length of label storage */
> > +	unsigned long bits[];	/* variable-sized label storage */
> > +};
> > 
> > it would increase minimum length needed but it would avoid
> > the rcu dances done by the current scheme.
> > 
> > It this is ok for you I'll make this change to see how many LOC are
> > saved by this.
> 
> If we simplify the current connlabel code it would be great. And so
> far, the only way I can think to obtain that is to explictly specify
> via the CT target the length of the label.

I've turned it into a var-length extension, so all the rcu dances
and runtime reallocs are gone.

I'll be sending the first non-rfc version of the patchset soon.

Minor drawback: I had to reduce it to 128 labels max, else the
extension offset array (u8 offset[]) will overflow if too many
extensions are enabled.

At the moment the 128 labels max constraint is no problem at all,
and we can increase it later if we reduce total possible extension
size (or use u16 for length/offset tracking).
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Stephen Clark - Nov. 15, 2012, 12:52 p.m.
On 11/15/2012 07:13 AM, Pablo Neira Ayuso wrote:
> Hi Florian,
>
> On Mon, Nov 12, 2012 at 01:47:05PM +0100, Florian Westphal wrote:
>> Pablo Neira Ayuso<pablo@netfilter.org>  wrote:
>>>> diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
>>>> new file mode 100644
>>>> index 0000000..eab398b
>>>> --- /dev/null
>>>> +++ b/net/netfilter/nf_conntrack_labels.c
>>>> @@ -0,0 +1,143 @@
>>>> +#include<linux/ctype.h>
>>>> +#include<linux/export.h>
>>>> +#include<linux/jhash.h>
>>>> +#include<linux/spinlock.h>
>>>> +#include<linux/types.h>
>>>> +#include<linux/slab.h>
>>>> +
>>>> +#include<net/netfilter/nf_conntrack_ecache.h>
>>>> +#include<net/netfilter/nf_conntrack_labels.h>
>>>> +
>>>> +static int labels_set_realloc(struct nf_conn_labels *l,
>>>> +			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)
>>> I think we can simplify this code if we use the CT target to set the
>>> number of labels that we'll use, so we skip allocations in runtime and
>>> possible reallocation.
>>>
>>> ... -t raw -j CT --labels 32
>> I'm not convinced yet ;-)
>>
>> I think we should avoid to make users fiddle with CT target options
>> just to get certain functionality working.
> I agree that we should try to keep things easy for users.
>
> Still, since the conntrack helper discussion during the last workshop,
> I think that users should explicitly enable conntrack features they
> want via iptables.
>
> In that direction, I've been toying with some patches to explicitly
> enable connectiong tracking via the CT target, ie. instead of tracking
> everything by default and using NOTRACK to say what you don't what
> (like we do now), tell what you want to track via some explict rule.
> PF people are doing it that way.
>
> Still that's an important semantic change so we'll have to keep some
> compatibility mode for some time
Yeah, like forever!! Do you realize what a drastic change this would be? How many
users actually use NOTRACK, and if they do it is for a very specific case. Most 
users
expect CONNTRACK to happen.

> (we would do the conntrack lookup in
> the raw table, and follow-up rules will find the ct object). That also
> would change the assumption that "you have no ct object in the raw
> table" to "you get the ct object just after the -j CT
> --enable-tracking rule".
>
> Note that we would have to rename (via aliasing) the raw table to
> make it become a filter table in the PREROUTING chain once that
> semantic change happens.
>
> As result of that change, it would allow us to remove the current
> template logic from:
>
> 1) attach the template
> 2) use the template in nf_conntrack_in() to set on features we requested
>
> to:
>
> 1) lookup/create ct object and attach features requested by the user
>     all at once.
>
> I'm going to find some spare time to send a RFC patch for this.
>
>> Alternative would be to keep track of highest bit requested in a "-m connlabel"
>> rule to figure out the needed size.
>>
>> In any case, it would require adding "u16 len" to the extension area; else
>> we can't figure out how many bytes are valid, i.e.:
>>
>> struct nf_conn_labels {
>> +       u16 size;		/* length of label storage */
>> +	unsigned long bits[];	/* variable-sized label storage */
>> +};
>>
>> it would increase minimum length needed but it would avoid
>> the rcu dances done by the current scheme.
>>
>> It this is ok for you I'll make this change to see how many LOC are
>> saved by this.
> If we simplify the current connlabel code it would be great. And so
> far, the only way I can think to obtain that is to explictly specify
> via the CT target the length of the label.
>
> Regards.
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
Pablo Neira - Nov. 15, 2012, 1:06 p.m.
On Thu, Nov 15, 2012 at 07:52:53AM -0500, Stephen Clark wrote:
> On 11/15/2012 07:13 AM, Pablo Neira Ayuso wrote:
> >Hi Florian,
> >
> >On Mon, Nov 12, 2012 at 01:47:05PM +0100, Florian Westphal wrote:
> >>Pablo Neira Ayuso<pablo@netfilter.org>  wrote:
> >>>>diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
> >>>>new file mode 100644
> >>>>index 0000000..eab398b
> >>>>--- /dev/null
> >>>>+++ b/net/netfilter/nf_conntrack_labels.c
> >>>>@@ -0,0 +1,143 @@
> >>>>+#include<linux/ctype.h>
> >>>>+#include<linux/export.h>
> >>>>+#include<linux/jhash.h>
> >>>>+#include<linux/spinlock.h>
> >>>>+#include<linux/types.h>
> >>>>+#include<linux/slab.h>
> >>>>+
> >>>>+#include<net/netfilter/nf_conntrack_ecache.h>
> >>>>+#include<net/netfilter/nf_conntrack_labels.h>
> >>>>+
> >>>>+static int labels_set_realloc(struct nf_conn_labels *l,
> >>>>+			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)
> >>>I think we can simplify this code if we use the CT target to set the
> >>>number of labels that we'll use, so we skip allocations in runtime and
> >>>possible reallocation.
> >>>
> >>>... -t raw -j CT --labels 32
> >>I'm not convinced yet ;-)
> >>
> >>I think we should avoid to make users fiddle with CT target options
> >>just to get certain functionality working.
> >I agree that we should try to keep things easy for users.
> >
> >Still, since the conntrack helper discussion during the last workshop,
> >I think that users should explicitly enable conntrack features they
> >want via iptables.
> >
> >In that direction, I've been toying with some patches to explicitly
> >enable connectiong tracking via the CT target, ie. instead of tracking
> >everything by default and using NOTRACK to say what you don't what
> >(like we do now), tell what you want to track via some explict rule.
> >PF people are doing it that way.
> >
> >Still that's an important semantic change so we'll have to keep some
> >compatibility mode for some time
>
> Yeah, like forever!! Do you realize what a drastic change this would
> be? How many users actually use NOTRACK, and if they do it is for a
> very specific case. Most users expect CONNTRACK to happen.

Aware of it. I already mentioned that we would need to add some
compatibility mode to have dual working mode, relying on the compat
behaviour by default, so noone would be affected.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pablo Neira - Nov. 15, 2012, 1:09 p.m.
On Thu, Nov 15, 2012 at 01:50:02PM +0100, Florian Westphal wrote:
> Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > On Mon, Nov 12, 2012 at 01:47:05PM +0100, Florian Westphal wrote:
> > > Alternative would be to keep track of highest bit requested in a "-m connlabel"
> > > rule to figure out the needed size.
> > > 
> > > In any case, it would require adding "u16 len" to the extension area; else
> > > we can't figure out how many bytes are valid, i.e.:
> > > 
> > > struct nf_conn_labels {
> > > +       u16 size;		/* length of label storage */
> > > +	unsigned long bits[];	/* variable-sized label storage */
> > > +};
> > > 
> > > it would increase minimum length needed but it would avoid
> > > the rcu dances done by the current scheme.
> > > 
> > > It this is ok for you I'll make this change to see how many LOC are
> > > saved by this.
> > 
> > If we simplify the current connlabel code it would be great. And so
> > far, the only way I can think to obtain that is to explictly specify
> > via the CT target the length of the label.
> 
> I've turned it into a var-length extension, so all the rcu dances
> and runtime reallocs are gone.
> 
> I'll be sending the first non-rfc version of the patchset soon.
> 
> Minor drawback: I had to reduce it to 128 labels max, else the
> extension offset array (u8 offset[]) will overflow if too many
> extensions are enabled.

Good point. That reminds me that we should add some bugtrap to
net/netfilter/nf_conntrack_extend.c to check for such possible
overflows.

> At the moment the 128 labels max constraint is no problem at all,
> and we can increase it later if we reduce total possible extension
> size (or use u16 for length/offset tracking).

Sure.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/include/net/netfilter/nf_conntrack_extend.h b/include/net/netfilter/nf_conntrack_extend.h
index 8b4d1fc2..977bc8a 100644
--- a/include/net/netfilter/nf_conntrack_extend.h
+++ b/include/net/netfilter/nf_conntrack_extend.h
@@ -23,6 +23,9 @@  enum nf_ct_ext_id {
 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
 	NF_CT_EXT_TIMEOUT,
 #endif
+#ifdef CONFIG_NF_CONNTRACK_LABELS
+	NF_CT_EXT_LABELS,
+#endif
 	NF_CT_EXT_NUM,
 };
 
@@ -33,6 +36,7 @@  enum nf_ct_ext_id {
 #define NF_CT_EXT_ZONE_TYPE struct nf_conntrack_zone
 #define NF_CT_EXT_TSTAMP_TYPE struct nf_conn_tstamp
 #define NF_CT_EXT_TIMEOUT_TYPE struct nf_conn_timeout
+#define NF_CT_EXT_LABELS_TYPE struct nf_conn_labels
 
 /* Extensions: optional stuff which isn't permanently in struct. */
 struct nf_ct_ext {
diff --git a/include/net/netfilter/nf_conntrack_labels.h b/include/net/netfilter/nf_conntrack_labels.h
new file mode 100644
index 0000000..dfcad1c
--- /dev/null
+++ b/include/net/netfilter/nf_conntrack_labels.h
@@ -0,0 +1,65 @@ 
+#include <linux/types.h>
+#include <net/net_namespace.h>
+#include <linux/netfilter/nf_conntrack_common.h>
+#include <linux/netfilter/nf_conntrack_tuple_common.h>
+#include <net/netfilter/nf_conntrack.h>
+#include <net/netfilter/nf_conntrack_extend.h>
+
+#include <uapi/linux/netfilter/xt_connlabel.h>
+
+#define NF_CONNLABEL_NOT_ALLOCATED 1
+
+struct __nf_conn_labels_rcu_ptr {
+	struct rcu_head rcu;
+	u16 maxbit;
+	unsigned long bits[];
+};
+
+struct nf_conn_labels {
+	struct __nf_conn_labels_rcu_ptr __rcu *label_bits;
+};
+
+static inline bool __nf_ct_labels_allocated(struct __nf_conn_labels_rcu_ptr *p)
+{
+	return ((unsigned long) p & NF_CONNLABEL_NOT_ALLOCATED) == 0;
+}
+
+static inline u16 __nf_ct_labels_get_maxbit(struct __nf_conn_labels_rcu_ptr *p)
+{
+	if (__nf_ct_labels_allocated(p))
+		return p->maxbit;
+	return BITS_PER_LONG - 1;
+}
+
+static inline struct nf_conn_labels *nf_ct_labels_find(const struct nf_conn *ct)
+{
+#ifdef CONFIG_NF_CONNTRACK_LABELS
+	return nf_ct_ext_find(ct, NF_CT_EXT_LABELS);
+#endif
+}
+
+static inline struct nf_conn_labels *nf_ct_labels_ext_add(struct nf_conn *ct)
+{
+#ifdef CONFIG_NF_CONNTRACK_LABELS
+	struct nf_conn_labels *cl_ext;
+	struct net *net = nf_ct_net(ct);
+
+	if (net->ct.labels_used == 0)
+		return NULL;
+
+	cl_ext = nf_ct_ext_add(ct, NF_CT_EXT_LABELS, GFP_ATOMIC);
+	if (cl_ext == NULL)
+		return NULL;
+
+	RCU_INIT_POINTER(cl_ext->label_bits, (void *) NF_CONNLABEL_NOT_ALLOCATED);
+	return cl_ext;
+#else
+	return NULL;
+#endif
+}
+
+void nf_conntrack_labels_fini(struct net *net);
+int nf_conntrack_labels_init(struct net *net);
+
+bool nf_connlabel_match(const struct nf_conn *ct, u16 bit);
+int nf_connlabel_set(struct nf_conn *ct, u16 bit);
diff --git a/include/net/netns/conntrack.h b/include/net/netns/conntrack.h
index a1d83cc..e1f0935 100644
--- a/include/net/netns/conntrack.h
+++ b/include/net/netns/conntrack.h
@@ -83,6 +83,9 @@  struct netns_ct {
 	int			sysctl_auto_assign_helper;
 	bool			auto_assign_helper_warned;
 	struct nf_ip_net	nf_ct_proto;
+#if defined(CONFIG_NF_CONNTRACK_LABELS)
+	unsigned int labels_used;
+#endif
 #ifdef CONFIG_NF_NAT_NEEDED
 	struct hlist_head	*nat_bysource;
 	unsigned int		nat_htable_size;
diff --git a/include/uapi/linux/netfilter/xt_connlabel.h b/include/uapi/linux/netfilter/xt_connlabel.h
new file mode 100644
index 0000000..768739b
--- /dev/null
+++ b/include/uapi/linux/netfilter/xt_connlabel.h
@@ -0,0 +1,13 @@ 
+#include <linux/types.h>
+
+#define XT_CONNLABEL_MAXBIT 1023
+enum xt_connlabel_mtopts {
+	XT_CONNLABEL_OP_INVERT = 1 << 0,
+	XT_CONNLABEL_OP_SET    = 1 << 1,
+};
+
+struct xt_connlabel_mtinfo {
+	__u16 bit;
+	__u16 options;
+};
+
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index fefa514..80a0fdd 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -124,6 +124,15 @@  config NF_CONNTRACK_TIMESTAMP
 
 	  If unsure, say `N'.
 
+config NF_CONNTRACK_LABELS
+	bool  'Connection tracking labels'
+	depends on NETFILTER_ADVANCED
+	help
+	  This option enables support for assigning user-defined text strings
+	  to connection tracking entries.
+
+	  If unsure, say `N'.
+
 config NF_CT_PROTO_DCCP
 	tristate 'DCCP protocol connection tracking support (EXPERIMENTAL)'
 	depends on EXPERIMENTAL
@@ -472,6 +481,13 @@  config NETFILTER_XT_SET
 
 	  To compile it as a module, choose M here.  If unsure, say N.
 
+config NETFILTER_XT_CONNLABELS
+	tristate '"connlimit" target and match support"'
+	depends on NF_CONNTRACK_LABELS
+	---help---
+	  This match allows you to match against connlabels assigned to the
+	  connection.
+
 # alphabetically ordered list of targets
 
 comment "Xtables targets"
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 3259697..0e7139c 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -4,6 +4,7 @@  nf_conntrack-y	:= nf_conntrack_core.o nf_conntrack_standalone.o nf_conntrack_exp
 nf_conntrack-$(CONFIG_NF_CONNTRACK_TIMEOUT) += nf_conntrack_timeout.o
 nf_conntrack-$(CONFIG_NF_CONNTRACK_TIMESTAMP) += nf_conntrack_timestamp.o
 nf_conntrack-$(CONFIG_NF_CONNTRACK_EVENTS) += nf_conntrack_ecache.o
+nf_conntrack-$(CONFIG_NF_CONNTRACK_LABELS) += nf_conntrack_labels.o
 
 obj-$(CONFIG_NETFILTER) = netfilter.o
 
@@ -70,6 +71,7 @@  obj-$(CONFIG_NETFILTER_XTABLES) += x_tables.o xt_tcpudp.o
 obj-$(CONFIG_NETFILTER_XT_MARK) += xt_mark.o
 obj-$(CONFIG_NETFILTER_XT_CONNMARK) += xt_connmark.o
 obj-$(CONFIG_NETFILTER_XT_SET) += xt_set.o
+obj-$(CONFIG_NETFILTER_XT_CONNLABELS) += xt_connlabel.o
 obj-$(CONFIG_NF_NAT) += xt_nat.o
 
 # targets
diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index 0f241be..c19e25c 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -45,6 +45,7 @@ 
 #include <net/netfilter/nf_conntrack_zones.h>
 #include <net/netfilter/nf_conntrack_timestamp.h>
 #include <net/netfilter/nf_conntrack_timeout.h>
+#include <net/netfilter/nf_conntrack_labels.h>
 #include <net/netfilter/nf_nat.h>
 #include <net/netfilter/nf_nat_core.h>
 
@@ -770,6 +771,7 @@  void nf_conntrack_free(struct nf_conn *ct)
 }
 EXPORT_SYMBOL_GPL(nf_conntrack_free);
 
+
 /* Allocate a new conntrack: we return -ENOMEM if classification
    failed due to stress.  Otherwise it really is unclassifiable. */
 static struct nf_conntrack_tuple_hash *
@@ -816,6 +818,7 @@  init_conntrack(struct net *net, struct nf_conn *tmpl,
 
 	nf_ct_acct_ext_add(ct, GFP_ATOMIC);
 	nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
+	nf_ct_labels_ext_add(ct);
 
 	ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL;
 	nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0,
@@ -1585,7 +1588,15 @@  static int nf_conntrack_init_net(struct net *net)
 	ret = nf_conntrack_helper_init(net);
 	if (ret < 0)
 		goto err_helper;
+
+	ret = nf_conntrack_labels_init(net);
+	if (ret < 0)
+		goto err_labels;
+
 	return 0;
+
+err_labels:
+	nf_conntrack_helper_fini(net);
 err_helper:
 	nf_conntrack_timeout_fini(net);
 err_timeout:
diff --git a/net/netfilter/nf_conntrack_labels.c b/net/netfilter/nf_conntrack_labels.c
new file mode 100644
index 0000000..eab398b
--- /dev/null
+++ b/net/netfilter/nf_conntrack_labels.c
@@ -0,0 +1,143 @@ 
+#include <linux/ctype.h>
+#include <linux/export.h>
+#include <linux/jhash.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+
+#include <net/netfilter/nf_conntrack_ecache.h>
+#include <net/netfilter/nf_conntrack_labels.h>
+
+static int labels_set_realloc(struct nf_conn_labels *l,
+			      struct __nf_conn_labels_rcu_ptr *oldptr, u16 bit)
+{
+	struct __nf_conn_labels_rcu_ptr *ptr;
+	u32 word, i, size;
+	u16 maxbit;
+
+	size = round_up(bit+1, BITS_PER_LONG) / BITS_PER_LONG;
+	size *= sizeof(long);
+
+	ptr = kmalloc(sizeof(*ptr) + size, GFP_ATOMIC);
+	if (!ptr)
+		return -ENOMEM;
+
+	if (__nf_ct_labels_allocated(oldptr)) {
+		word = BIT_WORD(oldptr->maxbit);
+		for (i = 0; i <= word; i++)
+			ptr->bits[i] = oldptr->bits[i];
+		kfree_rcu(oldptr, rcu);
+	} else {
+		unsigned long tmp = (unsigned long) oldptr;
+		ptr->bits[0] = tmp >> NF_CONNLABEL_NOT_ALLOCATED;
+		i = 1;
+	}
+
+	maxbit = (size * BITS_PER_BYTE) - 1;
+	word = BIT_WORD(maxbit);
+	for (;i <= word; i++)
+		ptr->bits[i] = 0;
+
+	set_bit(bit, ptr->bits);
+	ptr->maxbit = maxbit;
+	rcu_assign_pointer(l->label_bits, ptr);
+	return 0;
+}
+
+static bool __nf_connlabel_match(struct nf_conn_labels *labels, u16 bit)
+{
+	struct __nf_conn_labels_rcu_ptr *ptr = rcu_dereference(labels->label_bits);
+
+	if (!__nf_ct_labels_allocated(ptr)) {
+		unsigned long tmp = (unsigned long) ptr >> NF_CONNLABEL_NOT_ALLOCATED;
+		return bit < BITS_PER_LONG && test_bit(bit, &tmp);
+	}
+
+	return bit <= ptr->maxbit && test_bit(bit, ptr->bits);
+}
+
+bool nf_connlabel_match(const struct nf_conn *ct, u16 bit)
+{
+	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
+	bool ret;
+
+	if (!labels)
+		return false;
+
+	rcu_read_lock();
+	ret = __nf_connlabel_match(labels, bit);
+	rcu_read_unlock();
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(nf_connlabel_match);
+
+int nf_connlabel_set(struct nf_conn *ct, u16 bit)
+{
+	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
+	struct __nf_conn_labels_rcu_ptr *ptr;
+	int err;
+
+	if (!labels)
+		return -ENOSPC;
+
+	rcu_read_lock();
+	if (__nf_connlabel_match(labels, bit)) {
+		rcu_read_unlock();
+		return 0; /* fast path: already set */
+	}
+	rcu_read_unlock();
+
+	spin_lock(&ct->lock);
+
+	ptr = rcu_dereference_protected(labels->label_bits, 1);
+	if (__nf_ct_labels_allocated(ptr)) {
+		if (bit <= ptr->maxbit) {
+			set_bit(bit, ptr->bits);
+			err = 0;
+		} else
+			err = labels_set_realloc(labels, ptr, bit);
+	} else if (bit < (BITS_PER_LONG - NF_CONNLABEL_NOT_ALLOCATED)) {
+		unsigned long tmp = (unsigned long) ptr;
+		bit = 1 << (bit + NF_CONNLABEL_NOT_ALLOCATED);
+		rcu_assign_pointer(labels->label_bits, (void *) (tmp|bit));
+		err = 0;
+	} else {
+		err = labels_set_realloc(labels, ptr, bit);
+	}
+
+	spin_unlock(&ct->lock);
+	return err;
+}
+EXPORT_SYMBOL_GPL(nf_connlabel_set);
+
+static void nf_ct_ext_label_destroy(struct nf_conn *ct)
+{
+	struct nf_conn_labels *labels = nf_ct_labels_find(ct);
+	if (labels) {
+		struct __nf_conn_labels_rcu_ptr *p = rcu_dereference(labels->label_bits);
+		if (__nf_ct_labels_allocated(p))
+			kfree(p);
+	}
+}
+
+static struct nf_ct_ext_type labels_extend __read_mostly = {
+	.len    = sizeof(struct nf_conn_labels),
+	.align  = __alignof__(struct nf_conn_labels),
+	.destroy = nf_ct_ext_label_destroy,
+	.id     = NF_CT_EXT_LABELS,
+};
+
+int nf_conntrack_labels_init(struct net *net)
+{
+	if (net_eq(net, &init_net))
+		return nf_ct_extend_register(&labels_extend);
+	return 0;
+}
+
+void nf_conntrack_labels_fini(struct net *net)
+{
+	if (net_eq(net, &init_net))
+		nf_ct_extend_unregister(&labels_extend);
+}
+
diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c
index 7bbfb3d..cf22362 100644
--- a/net/netfilter/nf_conntrack_netlink.c
+++ b/net/netfilter/nf_conntrack_netlink.c
@@ -43,6 +43,7 @@ 
 #include <net/netfilter/nf_conntrack_acct.h>
 #include <net/netfilter/nf_conntrack_zones.h>
 #include <net/netfilter/nf_conntrack_timestamp.h>
+#include <net/netfilter/nf_conntrack_labels.h>
 #ifdef CONFIG_NF_NAT_NEEDED
 #include <net/netfilter/nf_nat_core.h>
 #include <net/netfilter/nf_nat_l4proto.h>
@@ -1489,6 +1490,8 @@  ctnetlink_create_conntrack(struct net *net, u16 zone,
 	nf_ct_acct_ext_add(ct, GFP_ATOMIC);
 	nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
 	nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC);
+	nf_ct_labels_ext_add(ct);
+
 	/* we must add conntrack extensions before confirmation. */
 	ct->status |= IPS_CONFIRMED;
 
diff --git a/net/netfilter/xt_connlabel.c b/net/netfilter/xt_connlabel.c
new file mode 100644
index 0000000..40c92f3
--- /dev/null
+++ b/net/netfilter/xt_connlabel.c
@@ -0,0 +1,77 @@ 
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <net/netfilter/nf_conntrack.h>
+#include <net/netfilter/nf_conntrack_labels.h>
+#include <linux/netfilter/x_tables.h>
+
+static bool
+connlabel_mt(const struct sk_buff *skb, struct xt_action_param *par)
+{
+        const struct xt_connlabel_mtinfo *info = par->matchinfo;
+	enum ip_conntrack_info ctinfo;
+	struct nf_conn *ct;
+	bool invert = info->options & XT_CONNLABEL_OP_INVERT;
+
+	ct = nf_ct_get(skb, &ctinfo);
+	if (ct == NULL || nf_ct_is_untracked(ct))
+		return invert;
+
+	if (info->options & XT_CONNLABEL_OP_SET)
+		return (nf_connlabel_set(ct, info->bit) == 0) ^ invert;
+
+	return nf_connlabel_match(ct, info->bit) ^ invert;
+}
+
+static int connlabel_mt_check(const struct xt_mtchk_param *par)
+{
+	const int options = XT_CONNLABEL_OP_INVERT |
+			    XT_CONNLABEL_OP_SET;
+	struct xt_connlabel_mtinfo *info = par->matchinfo;
+	int ret;
+
+	if (info->bit > XT_CONNLABEL_MAXBIT)
+		return -ERANGE;
+
+	if (info->options & ~options) {
+		pr_err("Unknown options in mask %x\n", info->options);
+		return -EINVAL;
+	}
+
+	ret = nf_ct_l3proto_try_module_get(par->family);
+	if (ret < 0) {
+		pr_info("cannot load conntrack support for proto=%u\n", par->family);
+		return ret;
+	}
+
+	par->net->ct.labels_used++;
+	return ret;
+}
+
+static void connlabel_mt_destroy(const struct xt_mtdtor_param *par)
+{
+	par->net->ct.labels_used--;
+	nf_ct_l3proto_module_put(par->family);
+}
+
+static struct xt_match connlabels_mt_reg __read_mostly = {
+        .name           = "connlabel",
+        .family         = NFPROTO_UNSPEC,
+        .checkentry     = connlabel_mt_check,
+        .match          = connlabel_mt,
+        .matchsize      = sizeof(struct xt_connlabel_mtinfo),
+        .destroy        = connlabel_mt_destroy,
+        .me             = THIS_MODULE,
+};
+
+static int __init connlabel_mt_init(void)
+{
+	return xt_register_match(&connlabels_mt_reg);
+}
+
+static void __exit connlabel_mt_exit(void)
+{
+	xt_unregister_match(&connlabels_mt_reg);
+}
+
+module_init(connlabel_mt_init);
+module_exit(connlabel_mt_exit);