[nf-next] netfilter: ctnetlink: add kernel side filtering for dump
diff mbox series

Message ID 20191219103638.20454-1-romain.bellan@wifirst.fr
State Changes Requested
Delegated to: Pablo Neira
Headers show
Series
  • [nf-next] netfilter: ctnetlink: add kernel side filtering for dump
Related show

Commit Message

Romain Bellan Dec. 19, 2019, 10:36 a.m. UTC
Conntrack dump does not support kernel side filtering (only get exists,
but it returns only one entry. And user has to give a full valid tuple)

It means that userspace has to implement filtering after receiving many
irrelevant entries, consuming ressources (conntrack table is sometimes
very huge, much more than a routing table for example).

This patch adds filtering in kernel side. To achieve this goal, we:

 * Add a new CTA_FILTER netlink attributes, actually a flag list to
   parametize filtering
 * Convert some *nlattr_to_tuple() functions, to allow a partial parsing
   of CTA_TUPLE_ORIG and CTA_TUPLE_REPLY (so nf_conntrack_tuple it not
   fully set)

Filtering is now possible on:
 * IP SRC/DST values
 * Ports for TCP and UDP flows
 * IMCP(v6) codes types and IDs

Filtering is done has an "AND" operator. For example, when flags
PROTO_SRC_PORT, PROTO_NUM and IP_SRC are sets, only entries matching all
values are dumped.

Signed-off-by: Romain Bellan <romain.bellan@wifirst.fr>
Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
---
 include/net/netfilter/nf_conntrack_l4proto.h  |   6 +-
 .../linux/netfilter/nfnetlink_conntrack.h     |  41 +++
 net/netfilter/nf_conntrack_core.c             |  19 +-
 net/netfilter/nf_conntrack_netlink.c          | 292 ++++++++++++++++--
 net/netfilter/nf_conntrack_proto_icmp.c       |  41 ++-
 net/netfilter/nf_conntrack_proto_icmpv6.c     |  43 ++-
 6 files changed, 374 insertions(+), 68 deletions(-)

Comments

Pablo Neira Ayuso Dec. 30, 2019, 12:12 p.m. UTC | #1
Hi Romain,

On Thu, Dec 19, 2019 at 11:36:38AM +0100, Romain Bellan wrote:
> Conntrack dump does not support kernel side filtering (only get exists,
> but it returns only one entry. And user has to give a full valid tuple)
> 
> It means that userspace has to implement filtering after receiving many
> irrelevant entries, consuming ressources (conntrack table is sometimes
> very huge, much more than a routing table for example).
> 
> This patch adds filtering in kernel side. To achieve this goal, we:
> 
>  * Add a new CTA_FILTER netlink attributes, actually a flag list to
>    parametize filtering
>  * Convert some *nlattr_to_tuple() functions, to allow a partial parsing
>    of CTA_TUPLE_ORIG and CTA_TUPLE_REPLY (so nf_conntrack_tuple it not
>    fully set)
> 
> Filtering is now possible on:
>  * IP SRC/DST values
>  * Ports for TCP and UDP flows
>  * IMCP(v6) codes types and IDs
> 
> Filtering is done has an "AND" operator. For example, when flags
> PROTO_SRC_PORT, PROTO_NUM and IP_SRC are sets, only entries matching all
> values are dumped.

Thanks for submitting this.

I did not yet have a look at this in detail, will do asap.

However, I would like to know if you would plan to submit userspace
patches for libnetfilter_conntrack for this. Main problem here is
backward compatibility (old conntrack tool and new kernel).
Romain Bellan Jan. 7, 2020, 9:10 a.m. UTC | #2
Hi Pablo,

> I did not yet have a look at this in detail, will do asap.
> 
> However, I would like to know if you would plan to submit userspace
> patches for libnetfilter_conntrack for this. Main problem here is
> backward compatibility (old conntrack tool and new kernel).

Currently I wrote a patch for the pyroute2 python library (to control netlink using Python) whith checks of kernel version for using filtering in kernel or userspace.

I would like to submit a patch for the libnetfilter_conntrack if you think that it is useful, but i didn't have a look on it yet.

About compatibility, currently the only way is to check with the kernel version, but I can add something like NLM_F_DUMP_FILTERED in the netlink reply. What would be the best way for you?

Best regards,
Pablo Neira Ayuso Jan. 13, 2020, 11 p.m. UTC | #3
On Tue, Jan 07, 2020 at 10:10:26AM +0100, Romain Bellan wrote:
> Hi Pablo,
> 
> > I did not yet have a look at this in detail, will do asap.
> > 
> > However, I would like to know if you would plan to submit userspace
> > patches for libnetfilter_conntrack for this. Main problem here is
> > backward compatibility (old conntrack tool and new kernel).
> 
> Currently I wrote a patch for the pyroute2 python library (to
> control netlink using Python) whith checks of kernel version for
> using filtering in kernel or userspace.
> 
> I would like to submit a patch for the libnetfilter_conntrack if you
> think that it is useful, but i didn't have a look on it yet.

Please have a look and make an API proposal. Use the
libnetfilter_conntrack libmnl API for this.

> About compatibility, currently the only way is to check with the
> kernel version, but I can add something like NLM_F_DUMP_FILTERED in
> the netlink reply. What would be the best way for you?

Yes, dump filtered is fine to signal userspace.

I was thinking on how conntrack could use this. Problem is that this
depends on the kernel version. Unless there is a way to do some
probing to see if the filtering is in place, the userspace conntrack
utility cannot use this. Could you also have a look into that?

Thanks.

Patch
diff mbox series

diff --git a/include/net/netfilter/nf_conntrack_l4proto.h b/include/net/netfilter/nf_conntrack_l4proto.h
index 4cad1f0a327a..88186b95b3c2 100644
--- a/include/net/netfilter/nf_conntrack_l4proto.h
+++ b/include/net/netfilter/nf_conntrack_l4proto.h
@@ -42,7 +42,8 @@  struct nf_conntrack_l4proto {
 	/* Calculate tuple nlattr size */
 	unsigned int (*nlattr_tuple_size)(void);
 	int (*nlattr_to_tuple)(struct nlattr *tb[],
-			       struct nf_conntrack_tuple *t);
+			       struct nf_conntrack_tuple *t,
+			       u_int32_t flags);
 	const struct nla_policy *nla_policy;
 
 	struct {
@@ -152,7 +153,8 @@  const struct nf_conntrack_l4proto *nf_ct_l4proto_find(u8 l4proto);
 int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb,
 			       const struct nf_conntrack_tuple *tuple);
 int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[],
-			       struct nf_conntrack_tuple *t);
+			       struct nf_conntrack_tuple *t,
+			       u_int32_t flags);
 unsigned int nf_ct_port_nlattr_tuple_size(void);
 extern const struct nla_policy nf_ct_port_nla_policy[];
 
diff --git a/include/uapi/linux/netfilter/nfnetlink_conntrack.h b/include/uapi/linux/netfilter/nfnetlink_conntrack.h
index 1d41810d17e2..8c67b9da2592 100644
--- a/include/uapi/linux/netfilter/nfnetlink_conntrack.h
+++ b/include/uapi/linux/netfilter/nfnetlink_conntrack.h
@@ -55,6 +55,7 @@  enum ctattr_type {
 	CTA_LABELS,
 	CTA_LABELS_MASK,
 	CTA_SYNPROXY,
+	CTA_FILTER,
 	__CTA_MAX
 };
 #define CTA_MAX (__CTA_MAX - 1)
@@ -276,4 +277,44 @@  enum ctattr_expect_stats {
 };
 #define CTA_STATS_EXP_MAX (__CTA_STATS_EXP_MAX - 1)
 
+enum ctattr_filter {
+	CTA_FILTER_UNSPEC,
+	CTA_FILTER_ORIG_FLAGS,
+	CTA_FILTER_REPLY_FLAGS,
+	__CTA_FILTER_MAX
+};
+#define CTA_FILTER_MAX (__CTA_FILTER_MAX - 1)
+
+#define CTA_FILTER_FLAG(x) CTA_FILTER_FLAG_ ## x
+
+/* applied on tuple filters */
+#define CTA_FILTER_FLAG_CTA_IP_SRC		(1 << 0)
+#define CTA_FILTER_FLAG_CTA_IP_DST		(1 << 1)
+#define CTA_FILTER_FLAG_CTA_TUPLE_ZONE		(1 << 2)
+#define CTA_FILTER_FLAG_CTA_PROTO_NUM		(1 << 3)
+#define CTA_FILTER_FLAG_CTA_PROTO_SRC_PORT	(1 << 4)
+#define CTA_FILTER_FLAG_CTA_PROTO_DST_PORT	(1 << 5)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMP_TYPE	(1 << 6)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMP_CODE	(1 << 7)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMP_ID	(1 << 8)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_TYPE	(1 << 9)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_CODE	(1 << 10)
+#define CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_ID	(1 << 11)
+
+#define CTA_FILTER_FLAG_ALL_CTA_PROTO \
+  (CTA_FILTER_FLAG_CTA_PROTO_SRC_PORT | \
+   CTA_FILTER_FLAG_CTA_PROTO_DST_PORT | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMP_TYPE | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMP_CODE | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMP_ID | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_TYPE | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_CODE | \
+   CTA_FILTER_FLAG_CTA_PROTO_ICMPV6_ID)
+#define CTA_FILTER_FLAG_ALL			0xFFFFFFFF
+
+/* applied on filters */
+#define CTA_FILTER_FLAG_CTA_MARK		(1 << 0)
+#define CTA_FILTER_FLAG_CTA_MARK_MASK		(1 << 1)
+
+
 #endif /* _IPCONNTRACK_NETLINK_H */
diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index 0af1898af2b8..790f0c1a46b8 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -1835,13 +1835,22 @@  const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = {
 EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy);
 
 int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[],
-			       struct nf_conntrack_tuple *t)
+			       struct nf_conntrack_tuple *t,
+			       u_int32_t flags)
 {
-	if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT])
-		return -EINVAL;
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_SRC_PORT)) {
+		if (!tb[CTA_PROTO_SRC_PORT])
+			return -EINVAL;
+
+		t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]);
+	}
 
-	t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]);
-	t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]);
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_DST_PORT)) {
+		if (!tb[CTA_PROTO_DST_PORT])
+			return -EINVAL;
+
+		t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]);
+	}
 
 	return 0;
 }
diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c
index d8d33ef52ce0..04507abbebc8 100644
--- a/net/netfilter/nf_conntrack_netlink.c
+++ b/net/netfilter/nf_conntrack_netlink.c
@@ -847,20 +847,62 @@  static int ctnetlink_done(struct netlink_callback *cb)
 }
 
 struct ctnetlink_filter {
+	u_int32_t cta_flags;
 	u8 family;
+
+	u_int32_t orig_flags;
+	struct nf_conntrack_tuple orig;
+
+	u_int32_t reply_flags;
+	struct nf_conntrack_tuple reply;
+
+	struct nf_conntrack_zone zone;
+
 	struct {
 		u_int32_t val;
 		u_int32_t mask;
 	} mark;
 };
 
+
+static const struct nla_policy cta_filter_nla_policy[CTA_FILTER_MAX + 1] = {
+	[CTA_FILTER_ORIG_FLAGS]		= { .type = NLA_U32 },
+	[CTA_FILTER_REPLY_FLAGS]	= { .type = NLA_U32 },
+};
+
+static int ctnetlink_parse_filter(const struct nlattr *attr,
+				  struct ctnetlink_filter *filter)
+{
+	struct nlattr *tb[CTA_FILTER_MAX+1];
+	int ret = 0;
+
+	ret = nla_parse_nested(tb, CTA_FILTER_MAX, attr, cta_filter_nla_policy, NULL);
+	if (ret)
+		return ret;
+
+	if (tb[CTA_FILTER_ORIG_FLAGS])
+		filter->orig_flags = nla_get_u32(tb[CTA_FILTER_ORIG_FLAGS]);
+
+	if (tb[CTA_FILTER_REPLY_FLAGS])
+		filter->reply_flags = nla_get_u32(tb[CTA_FILTER_REPLY_FLAGS]);
+
+	return 0;
+}
+
+static int ctnetlink_parse_zone(const struct nlattr *attr, struct nf_conntrack_zone *zone);
+static int ctnetlink_parse_partial_tuple(const struct nlattr * const cda[],
+					 struct nf_conntrack_tuple *tuple,
+					 u32 type, u_int8_t l3num, struct nf_conntrack_zone *zone,
+					 u_int32_t flags);
+
 static struct ctnetlink_filter *
 ctnetlink_alloc_filter(const struct nlattr * const cda[], u8 family)
 {
 	struct ctnetlink_filter *filter;
+	int err;
 
 #ifndef CONFIG_NF_CONNTRACK_MARK
-	if (cda[CTA_MARK] && cda[CTA_MARK_MASK])
+	if (cda[CTA_MARK] || cda[CTA_MARK_MASK])
 		return ERR_PTR(-EOPNOTSUPP);
 #endif
 
@@ -871,11 +913,53 @@  ctnetlink_alloc_filter(const struct nlattr * const cda[], u8 family)
 	filter->family = family;
 
 #ifdef CONFIG_NF_CONNTRACK_MARK
-	if (cda[CTA_MARK] && cda[CTA_MARK_MASK]) {
+	if (cda[CTA_MARK]) {
 		filter->mark.val = ntohl(nla_get_be32(cda[CTA_MARK]));
-		filter->mark.mask = ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
+		filter->cta_flags |= CTA_FILTER_FLAG(CTA_MARK);
+
+		if (cda[CTA_MARK_MASK]) {
+			filter->mark.mask = ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
+			filter->cta_flags |= CTA_FILTER_FLAG(CTA_MARK_MASK);
+		}
 	}
 #endif
+	if (!cda[CTA_FILTER])
+		return filter;
+
+	err = ctnetlink_parse_zone(cda[CTA_ZONE], &(filter->zone));
+	if (err < 0)
+		return ERR_PTR(err);
+
+	err = ctnetlink_parse_filter(cda[CTA_FILTER], filter);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	if (filter->orig_flags) {
+		if (!cda[CTA_TUPLE_ORIG])
+			return ERR_PTR(-EINVAL);
+
+		err = ctnetlink_parse_partial_tuple(cda, &(filter->orig),
+						    CTA_TUPLE_ORIG,
+						    filter->family,
+						    &(filter->zone),
+						    filter->orig_flags);
+		if (err < 0)
+			return ERR_PTR(err);
+	}
+
+	if (filter->reply_flags) {
+		if (!cda[CTA_TUPLE_REPLY])
+			return ERR_PTR(-EINVAL);
+
+		err = ctnetlink_parse_partial_tuple(cda, &(filter->reply),
+						    CTA_TUPLE_REPLY,
+						    filter->family,
+						    &(filter->zone),
+						    filter->orig_flags);
+		if (err < 0)
+			return ERR_PTR(err);
+	}
+
 	return filter;
 }
 
@@ -886,7 +970,8 @@  static int ctnetlink_start(struct netlink_callback *cb)
 	struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
 	u8 family = nfmsg->nfgen_family;
 
-	if (family || (cda[CTA_MARK] && cda[CTA_MARK_MASK])) {
+	if (family || (cda[CTA_MARK] && cda[CTA_MARK_MASK]) ||
+	    cda[CTA_FILTER]) {
 		filter = ctnetlink_alloc_filter(cda, family);
 		if (IS_ERR(filter))
 			return PTR_ERR(filter);
@@ -896,9 +981,82 @@  static int ctnetlink_start(struct netlink_callback *cb)
 	return 0;
 }
 
+static int ctnetlink_filter_match_tuple(struct nf_conntrack_tuple *filter_tuple,
+					struct nf_conntrack_tuple *ct_tuple,
+					u_int32_t flags, int family)
+{
+	switch (family) {
+	case NFPROTO_IPV4:
+		if (flags & CTA_FILTER_FLAG(CTA_IP_SRC) &&
+		    filter_tuple->src.u3.ip != ct_tuple->src.u3.ip)
+			return  0;
+
+		if (flags & CTA_FILTER_FLAG(CTA_IP_DST) &&
+		    filter_tuple->dst.u3.ip != ct_tuple->dst.u3.ip)
+			return  0;
+		break;
+
+	case NFPROTO_IPV6:
+		if (flags & CTA_FILTER_FLAG(CTA_IP_SRC) &&
+		    !ipv6_addr_cmp(&(filter_tuple->src.u3.in6),
+				   &(ct_tuple->src.u3.in6)))
+			return 0;
+
+		if (flags & CTA_FILTER_FLAG(CTA_IP_DST) &&
+		    !ipv6_addr_cmp(&(filter_tuple->dst.u3.in6),
+				   &(ct_tuple->dst.u3.in6)))
+			return 0;
+		break;
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_NUM) &&
+	    filter_tuple->dst.protonum != ct_tuple->dst.protonum)
+		return 0;
+
+	switch (ct_tuple->dst.protonum) {
+	case IPPROTO_TCP:
+	case IPPROTO_UDP:
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_SRC_PORT) &&
+		    filter_tuple->src.u.tcp.port != ct_tuple->src.u.tcp.port)
+			return 0;
+
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_DST_PORT) &&
+		    filter_tuple->dst.u.tcp.port != ct_tuple->dst.u.tcp.port)
+			return 0;
+		break;
+
+	case IPPROTO_ICMP:
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_TYPE) &&
+		    filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type)
+			return 0;
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_CODE) &&
+		    filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code)
+			return 0;
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_ID) &&
+		    filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id)
+			return 0;
+		break;
+
+	case IPPROTO_ICMPV6:
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_TYPE) &&
+		    filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type)
+			return 0;
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_CODE) &&
+		    filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code)
+			return 0;
+		if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_ID) &&
+		    filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id)
+			return 0;
+		break;
+	}
+
+	return 1;
+}
+
 static int ctnetlink_filter_match(struct nf_conn *ct, void *data)
 {
 	struct ctnetlink_filter *filter = data;
+	struct nf_conntrack_tuple *ct_tuple;
 
 	if (filter == NULL)
 		goto out;
@@ -910,8 +1068,29 @@  static int ctnetlink_filter_match(struct nf_conn *ct, void *data)
 	if (filter->family && nf_ct_l3num(ct) != filter->family)
 		goto ignore_entry;
 
+	if (filter->orig_flags) {
+		ct_tuple = nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL);
+		if (!ctnetlink_filter_match_tuple(&(filter->orig), ct_tuple,
+						  filter->orig_flags,
+						  filter->family))
+			goto ignore_entry;
+	}
+
+	if (filter->reply_flags) {
+		ct_tuple = nf_ct_tuple(ct, IP_CT_DIR_REPLY);
+		if (!ctnetlink_filter_match_tuple(&(filter->reply), ct_tuple,
+						  filter->reply_flags,
+						  filter->family))
+			goto ignore_entry;
+	}
+
+
 #ifdef CONFIG_NF_CONNTRACK_MARK
-	if ((ct->mark & filter->mark.mask) != filter->mark.val)
+	if ((filter->cta_flags & CTA_FILTER_FLAG(CTA_MARK_MASK)) &&
+	    (ct->mark & filter->mark.mask) != filter->mark.val)
+		goto ignore_entry;
+	else if ((filter->cta_flags & CTA_FILTER_FLAG(CTA_MARK)) &&
+		 ct->mark != filter->mark.val)
 		goto ignore_entry;
 #endif
 
@@ -1014,31 +1193,50 @@  ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 }
 
 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
-				struct nf_conntrack_tuple *t)
+				struct nf_conntrack_tuple *t,
+				u_int32_t flags)
 {
-	if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
-		return -EINVAL;
+	if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) {
+		if (!tb[CTA_IP_V4_SRC])
+			return -EINVAL;
 
-	t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
-	t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
+		t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) {
+		if (!tb[CTA_IP_V4_DST])
+			return -EINVAL;
+
+		t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
+	}
 
 	return 0;
 }
 
 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
-				struct nf_conntrack_tuple *t)
+				struct nf_conntrack_tuple *t,
+				u_int32_t flags)
 {
-	if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
-		return -EINVAL;
+	if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) {
+		if (!tb[CTA_IP_V6_SRC])
+			return -EINVAL;
 
-	t->src.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_SRC]);
-	t->dst.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_DST]);
+		t->src.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_SRC]);
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) {
+		if (!tb[CTA_IP_V6_DST])
+			return -EINVAL;
+
+		t->dst.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_DST]);
+	}
 
 	return 0;
 }
 
 static int ctnetlink_parse_tuple_ip(struct nlattr *attr,
-				    struct nf_conntrack_tuple *tuple)
+				    struct nf_conntrack_tuple *tuple,
+				    u_int32_t flags)
 {
 	struct nlattr *tb[CTA_IP_MAX+1];
 	int ret = 0;
@@ -1054,10 +1252,10 @@  static int ctnetlink_parse_tuple_ip(struct nlattr *attr,
 
 	switch (tuple->src.l3num) {
 	case NFPROTO_IPV4:
-		ret = ipv4_nlattr_to_tuple(tb, tuple);
+		ret = ipv4_nlattr_to_tuple(tb, tuple, flags);
 		break;
 	case NFPROTO_IPV6:
-		ret = ipv6_nlattr_to_tuple(tb, tuple);
+		ret = ipv6_nlattr_to_tuple(tb, tuple, flags);
 		break;
 	}
 
@@ -1069,7 +1267,8 @@  static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
 };
 
 static int ctnetlink_parse_tuple_proto(struct nlattr *attr,
-				       struct nf_conntrack_tuple *tuple)
+				       struct nf_conntrack_tuple *tuple,
+				       u_int32_t flags)
 {
 	const struct nf_conntrack_l4proto *l4proto;
 	struct nlattr *tb[CTA_PROTO_MAX+1];
@@ -1080,8 +1279,12 @@  static int ctnetlink_parse_tuple_proto(struct nlattr *attr,
 	if (ret < 0)
 		return ret;
 
+	if (!(flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)))
+		return 0;
+
 	if (!tb[CTA_PROTO_NUM])
 		return -EINVAL;
+
 	tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
 
 	rcu_read_lock();
@@ -1092,7 +1295,7 @@  static int ctnetlink_parse_tuple_proto(struct nlattr *attr,
 						     l4proto->nla_policy,
 						     NULL);
 		if (ret == 0)
-			ret = l4proto->nlattr_to_tuple(tb, tuple);
+			ret = l4proto->nlattr_to_tuple(tb, tuple, flags);
 	}
 
 	rcu_read_unlock();
@@ -1144,9 +1347,10 @@  static const struct nla_policy tuple_nla_policy[CTA_TUPLE_MAX+1] = {
 };
 
 static int
-ctnetlink_parse_tuple(const struct nlattr * const cda[],
-		      struct nf_conntrack_tuple *tuple, u32 type,
-		      u_int8_t l3num, struct nf_conntrack_zone *zone)
+ctnetlink_parse_partial_tuple(const struct nlattr * const cda[],
+			      struct nf_conntrack_tuple *tuple, u32 type,
+			      u_int8_t l3num, struct nf_conntrack_zone *zone,
+			      u_int32_t flags)
 {
 	struct nlattr *tb[CTA_TUPLE_MAX+1];
 	int err;
@@ -1158,23 +1362,33 @@  ctnetlink_parse_tuple(const struct nlattr * const cda[],
 	if (err < 0)
 		return err;
 
-	if (!tb[CTA_TUPLE_IP])
-		return -EINVAL;
 
 	tuple->src.l3num = l3num;
 
-	err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
-	if (err < 0)
-		return err;
+	if (flags & CTA_FILTER_FLAG(CTA_IP_DST) ||
+	    flags & CTA_FILTER_FLAG(CTA_IP_SRC))
+	{
+		if (!tb[CTA_TUPLE_IP])
+			return -EINVAL;
 
-	if (!tb[CTA_TUPLE_PROTO])
-		return -EINVAL;
+		err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple, flags);
+		if (err < 0)
+			return err;
+	}
 
-	err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
-	if (err < 0)
-		return err;
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)) {
+		if (!tb[CTA_TUPLE_PROTO])
+			return -EINVAL;
+		err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple, flags);
+		if (err < 0)
+			return err;
+	}
+	else if (flags & CTA_FILTER_FLAG(ALL_CTA_PROTO)) {
+		/* Can't manage proto flags without a protonum  */
+		return -EINVAL;
+	}
 
-	if (tb[CTA_TUPLE_ZONE]) {
+	if (flags & CTA_FILTER_FLAG(CTA_TUPLE_ZONE)) {
 		if (!zone)
 			return -EINVAL;
 
@@ -1193,6 +1407,15 @@  ctnetlink_parse_tuple(const struct nlattr * const cda[],
 	return 0;
 }
 
+static int
+ctnetlink_parse_tuple(const struct nlattr * const cda[],
+		      struct nf_conntrack_tuple *tuple, u32 type,
+		      u_int8_t l3num, struct nf_conntrack_zone *zone)
+{
+	return ctnetlink_parse_partial_tuple(cda, tuple, type, l3num, zone,
+					     CTA_FILTER_FLAG(ALL));
+}
+
 static const struct nla_policy help_nla_policy[CTA_HELP_MAX+1] = {
 	[CTA_HELP_NAME]		= { .type = NLA_NUL_STRING,
 				    .len = NF_CT_HELPER_NAME_LEN - 1 },
@@ -1240,6 +1463,7 @@  static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
 				    .len = NF_CT_LABELS_MAX_SIZE },
 	[CTA_LABELS_MASK]	= { .type = NLA_BINARY,
 				    .len = NF_CT_LABELS_MAX_SIZE },
+	[CTA_FILTER]		= { .type = NLA_NESTED },
 };
 
 static int ctnetlink_flush_iterate(struct nf_conn *ct, void *data)
diff --git a/net/netfilter/nf_conntrack_proto_icmp.c b/net/netfilter/nf_conntrack_proto_icmp.c
index c2e3dff773bc..78092466192c 100644
--- a/net/netfilter/nf_conntrack_proto_icmp.c
+++ b/net/netfilter/nf_conntrack_proto_icmp.c
@@ -271,20 +271,35 @@  static const struct nla_policy icmp_nla_policy[CTA_PROTO_MAX+1] = {
 };
 
 static int icmp_nlattr_to_tuple(struct nlattr *tb[],
-				struct nf_conntrack_tuple *tuple)
+				struct nf_conntrack_tuple *tuple,
+				u_int32_t flags)
 {
-	if (!tb[CTA_PROTO_ICMP_TYPE] ||
-	    !tb[CTA_PROTO_ICMP_CODE] ||
-	    !tb[CTA_PROTO_ICMP_ID])
-		return -EINVAL;
-
-	tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]);
-	tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]);
-	tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]);
-
-	if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
-	    !invmap[tuple->dst.u.icmp.type])
-		return -EINVAL;
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_TYPE)) {
+
+		if (!tb[CTA_PROTO_ICMP_TYPE])
+			return -EINVAL;
+
+		tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]);
+		if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
+		    !invmap[tuple->dst.u.icmp.type])
+			return -EINVAL;
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_CODE)) {
+
+		if (!tb[CTA_PROTO_ICMP_CODE])
+			return -EINVAL;
+
+		tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]);
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_ID)) {
+
+		if (!tb[CTA_PROTO_ICMP_ID])
+			return -EINVAL;
+
+		tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]);
+	}
 
 	return 0;
 }
diff --git a/net/netfilter/nf_conntrack_proto_icmpv6.c b/net/netfilter/nf_conntrack_proto_icmpv6.c
index 6f9144e1f1c1..3bdccff83ec3 100644
--- a/net/netfilter/nf_conntrack_proto_icmpv6.c
+++ b/net/netfilter/nf_conntrack_proto_icmpv6.c
@@ -193,21 +193,36 @@  static const struct nla_policy icmpv6_nla_policy[CTA_PROTO_MAX+1] = {
 };
 
 static int icmpv6_nlattr_to_tuple(struct nlattr *tb[],
-				struct nf_conntrack_tuple *tuple)
+				struct nf_conntrack_tuple *tuple,
+				u_int32_t flags)
 {
-	if (!tb[CTA_PROTO_ICMPV6_TYPE] ||
-	    !tb[CTA_PROTO_ICMPV6_CODE] ||
-	    !tb[CTA_PROTO_ICMPV6_ID])
-		return -EINVAL;
-
-	tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMPV6_TYPE]);
-	tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMPV6_CODE]);
-	tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMPV6_ID]);
-
-	if (tuple->dst.u.icmp.type < 128 ||
-	    tuple->dst.u.icmp.type - 128 >= sizeof(invmap) ||
-	    !invmap[tuple->dst.u.icmp.type - 128])
-		return -EINVAL;
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_TYPE)) {
+
+		if (!tb[CTA_PROTO_ICMPV6_TYPE])
+			return -EINVAL;
+
+		tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMPV6_TYPE]);
+		if (tuple->dst.u.icmp.type < 128 ||
+		    tuple->dst.u.icmp.type - 128 >= sizeof(invmap) ||
+		    !invmap[tuple->dst.u.icmp.type - 128])
+			return -EINVAL;
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_CODE)) {
+
+		if (!tb[CTA_PROTO_ICMPV6_CODE])
+			return -EINVAL;
+
+		tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMPV6_CODE]);
+	}
+
+	if (flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_ID)) {
+
+		if (!tb[CTA_PROTO_ICMPV6_ID])
+			return -EINVAL;
+
+		tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMPV6_ID]);
+	}
 
 	return 0;
 }