Skip to contentSkip to conversation

netlink_delinearize: incorrect meta protocol dependency kill

PATCH In review Pablo Neira Ayuso 3 participants 5 messages Aug 26, 2021 – Aug 30, 2021 netfilter

lore
Share & tools
lore

Pablo Neira Ayuso

author maintainer
#0 lore raw

PATCH 1/1 netlink_delinearize: incorrect meta protocol dependency kill

meta protocol is meaningful in bridge, netdev and inet familiiess, do not remove this.

Fixes a1bcf8a34975 ("payload: add payload_may_dependency_kill()")
---
 src/netlink_delinearize.c      | 22 ++++++++++++--
 tests/py/bridge/meta.t         |  3 ++
 tests/py/bridge/meta.t.json    | 54 ++++++++++++++++++++++++++++++++++
 tests/py/bridge/meta.t.payload | 18 ++++++++++++
 tests/py/inet/meta.t           |  4 +++
 tests/py/inet/meta.t.json      | 54 ++++++++++++++++++++++++++++++++++
 tests/py/inet/meta.t.payload   | 18 ++++++++++++
 tests/py/ip/meta.t             |  2 ++
 tests/py/ip/meta.t.json        | 16 ++++++++++
 tests/py/ip/meta.t.payload     |  9 ++++++
 tests/py/ip6/meta.t            |  3 ++
 tests/py/ip6/meta.t.json       | 54 ++++++++++++++++++++++++++++++++++
 tests/py/ip6/meta.t.payload    | 18 ++++++++++++
 13 files changed, 272 insertions(+), 3 deletions(-)
src/netlink_delinearize.c +19 -3
index 5b545701e8b7..92617a46df6f 100644
@@ -1993,7 +1993,7 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 				     const struct expr *expr)
 {
 	struct expr *dep = ctx->pdep->expr;
-	uint16_t l3proto;
+	uint16_t l3proto, protocol;
 	uint8_t l4proto;
 
 	if (ctx->pbase != PROTO_BASE_NETWORK_HDR)
@@ -2005,7 +2005,22 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 	case NFPROTO_BRIDGE:
 		break;
 	default:
-		return true;
+		if (dep->left->etype != EXPR_META ||
+		    dep->right->etype != EXPR_VALUE)
+			return false;
+
+		if (dep->left->meta.key == NFT_META_PROTOCOL) {
+			protocol = mpz_get_uint16(dep->right->value);
+
+			if (family == NFPROTO_IPV4 &&
+			    protocol == ETH_P_IP)
+				return true;
+			else if (family == NFPROTO_IPV6 &&
+				 protocol == ETH_P_IPV6)
+				return true;
+		}
+
+		return false;
 	}
 
 	if (expr->left->meta.key != NFT_META_L4PROTO)
@@ -2015,7 +2030,8 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 
 	switch (dep->left->etype) {
 	case EXPR_META:
-		if (dep->left->meta.key != NFT_META_NFPROTO)
+		if (dep->left->meta.key != NFT_META_NFPROTO &&
+		    dep->left->meta.key != NFT_META_PROTOCOL)
 			return true;
 		break;
 	case EXPR_PAYLOAD:
tests/py/bridge/meta.t +3 -0
index eda7082f02b4..d77ebd899f18 100644
@@ -6,3 +6,6 @@ meta obrname "br0";ok
 meta ibrname "br0";ok
 meta ibrvproto vlan;ok;meta ibrvproto 8021q
 meta ibrpvid 100;ok
+
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
tests/py/bridge/meta.t.json +54 -0

57 lines not on this page · raw message

tests/py/bridge/meta.t.payload +18 -0
index aa8c994bfe58..14177767fdfc 100644
@@ -17,3 +17,21 @@ bridge test-bridge input
 bridge test-bridge input
   [ meta load bri_iifpvid => reg 1 ]
   [ cmp eq reg 1 0x00000064 ]
+
+# meta protocol ip udp dport 67
+bridge
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x00000008 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
+
+# meta protocol ip6 udp dport 67
+bridge
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x0000dd86 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
tests/py/inet/meta.t +4 -0
index 3638898b5dbb..423cc5f32cba 100644
@@ -12,6 +12,10 @@ meta nfproto ipv4 tcp dport 22;ok
 meta nfproto ipv4 ip saddr 1.2.3.4;ok;ip saddr 1.2.3.4
 meta nfproto ipv6 meta l4proto tcp;ok;meta nfproto ipv6 meta l4proto 6
 meta nfproto ipv4 counter ip saddr 1.2.3.4;ok
+
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
+
 meta ipsec exists;ok
 meta secpath missing;ok;meta ipsec missing
 meta ibrname "br0";fail
tests/py/inet/meta.t.json +54 -0

57 lines not on this page · raw message

tests/py/inet/meta.t.payload +18 -0
index 6ccf6d24210a..f2c861378c61 100644
@@ -79,3 +79,21 @@ inet test-inet input
   [ ct load mark => reg 1 ]
   [ bitwise reg 1 = ( reg 1 >> 0x00000008 ) ]
   [ meta set mark with reg 1 ]
+
+# meta protocol ip udp dport 67
+inet
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x00000008 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
+
+# meta protocol ip6 udp dport 67
+inet
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x0000dd86 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
tests/py/ip/meta.t +2 -0
index f733d22de2c3..fecd0caf71a7 100644
@@ -8,6 +8,8 @@ meta l4proto ipv6-icmp icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-adv
 meta l4proto 58 icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-advert
 icmpv6 type nd-router-advert;ok
 
+meta protocol ip udp dport 67;ok
+
 meta ibrname "br0";fail
 meta obrname "br0";fail
 
tests/py/ip/meta.t.json +16 -0
index f83864f672d5..3df31ce381fc 100644
@@ -140,3 +140,19 @@
         "accept": null
     }
 ]
+
+# meta protocol ip udp dport 67
+[
+    {
+        "match": {
+            "left": {
+                "payload": {
+                    "field": "dport",
+                    "protocol": "udp"
+                }
+            },
+            "op": "==",
+            "right": 67
+        }
+    }
+]
tests/py/ip/meta.t.payload +9 -0
index 7bc69a290d24..ebff0e4bb3b0 100644
@@ -44,3 +44,12 @@ ip6 test-ip4 input
   [ meta load sdifname => reg 1 ]
   [ cmp neq reg 1 0x31667276 0x00000000 0x00000000 0x00000000 ]
   [ immediate reg 0 accept ]
+
+# meta protocol ip udp dport 67
+ip
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x00000008 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
tests/py/ip6/meta.t +3 -0
index dce97f5b0fd0..2c1aee2309a9 100644
@@ -9,5 +9,8 @@ meta l4proto icmp icmp type echo-request;ok;icmp type echo-request
 meta l4proto 1 icmp type echo-request;ok;icmp type echo-request
 icmp type echo-request;ok
 
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
+
 meta sdif "lo" accept;ok
 meta sdifname != "vrf1" accept;ok
tests/py/ip6/meta.t.json +54 -0

57 lines not on this page · raw message

tests/py/ip6/meta.t.payload +18 -0
index be04816eeec2..d654b019e7a7 100644
@@ -44,3 +44,21 @@ ip6 test-ip6 input
   [ meta load sdifname => reg 1 ]
   [ cmp neq reg 1 0x31667276 0x00000000 0x00000000 0x00000000 ]
   [ immediate reg 0 accept ]
+
+# meta protocol ip udp dport 67
+ip6
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x00000008 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
+
+# meta protocol ip6 udp dport 67
+ip6
+  [ meta load protocol => reg 1 ]
+  [ cmp eq reg 1 0x0000dd86 ]
+  [ meta load l4proto => reg 1 ]
+  [ cmp eq reg 1 0x00000011 ]
+  [ payload load 2b @ transport header + 2 => reg 1 ]
+  [ cmp eq reg 1 0x00004300 ]
Signature · 1 line
2.20.1
maintainer
#1 lore raw
Quoted 4 lines · Pablo Neira Ayuso

meta protocol is meaningful in bridge, netdev and inet familiiess, do not remove this.

Fixes a1bcf8a34975 ("payload: add payload_may_dependency_kill()")

Looks like its a regression from "netlink_delinearize: Refactor meta_may_dependency_kill()", 056aaa3e6dc65aced5e552233ac3e7f89fb81f86.

reviewer
#2 lore raw
In reply to Pablo Neira Ayuso

Hi,

Quoted 2 lines · Pablo Neira Ayuso

meta protocol is meaningful in bridge, netdev and inet familiiess, do not remove this.

So you want to avoid dependency killing in above families, but:

[...]

Quoted 4 lines
src/netlink_delinearize.c +0 -0
index 5b545701e8b7..92617a46df6f 100644

[...]

Quoted 14 lines
src/netlink_delinearize.c +9 -1
@@ -2005,7 +2005,22 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 	case NFPROTO_BRIDGE:
 		break;
 	default:
-		return true;
+		if (dep->left->etype != EXPR_META ||
+		    dep->right->etype != EXPR_VALUE)
+			return false;
+
+		if (dep->left->meta.key == NFT_META_PROTOCOL) {
+			protocol = mpz_get_uint16(dep->right->value);
+
+			if (family == NFPROTO_IPV4 &&
+			    protocol == ETH_P_IP)

src/netlink_delinearize.c static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx, · line 2026

Quoted 10 lines
src/netlink_delinearize.c +7 -0
+				return true;
+			else if (family == NFPROTO_IPV6 &&
+				 protocol == ETH_P_IPV6)
+				return true;
+		}
+
+		return false;
 	}
 
 	if (expr->left->meta.key != NFT_META_L4PROTO)

The above code only applies to families other than inet, netdev or bridge?! Am I missing something?

AFAIU, dependency killing defaults to true and examines the cases where it should not happen. So the old behaviour was for unexpected dependency expressions to just drop them instead of keeping them. The code you add above does the opposite. That's not wrong per se, but I assume there was a reason why dependency elimination was implemented "greedy", therefore I kept it that way when refactoring the code.

src/netlink_delinearize.c static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx, · line 2037

Quoted 10 lines
src/netlink_delinearize.c +2 -1
@@ -2015,7 +2030,8 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 
 	switch (dep->left->etype) {
 	case EXPR_META:
-		if (dep->left->meta.key != NFT_META_NFPROTO)
+		if (dep->left->meta.key != NFT_META_NFPROTO &&
+		    dep->left->meta.key != NFT_META_PROTOCOL)
 			return true;
 		break;
 	case EXPR_PAYLOAD:

If we continue evaluation for 'meta protocol' payload dependencies, RHS values need adjustment: with 'meta nfproto' we'll see NFPROTO_IPV4 or NFPROTO_IPV6 while with 'meta protocol' we'll see ETH_P_IP or ETH_P_IPV6. The EXPR_PAYLOAD case has such conversion code.

[...]

tests/py/ip/meta.t meta l4proto ipv6-icmp icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-adv · line 12

Quoted 10 lines
tests/py/ip/meta.t +2 -0
index f733d22de2c3..fecd0caf71a7 100644
@@ -8,6 +8,8 @@ meta l4proto ipv6-icmp icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-adv
 meta l4proto 58 icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-advert
 icmpv6 type nd-router-advert;ok
 
+meta protocol ip udp dport 67;ok
+

Hmm. Shouldn't this drop the dependency, i.e. list as 'udp dport 67' as the family is defined by table family already?

There should also be a case for 'meta protocol ip6 udp dport 67' which makes sure the dependency is not dropped, right?

Quoted 17 lines
tests/py/ip/meta.t +0 -0
 meta ibrname "br0";fail
 meta obrname "br0";fail
 
tests/py/ip/meta.t.json +6 -0
index f83864f672d5..3df31ce381fc 100644
@@ -140,3 +140,19 @@
         "accept": null
     }
 ]
+
+# meta protocol ip udp dport 67
+[
+    {
+        "match": {
+            "left": {

tests/py/ip/meta.t.json line 158

Quoted 10 lines
tests/py/ip/meta.t.json +10 -0
+                "payload": {
+                    "field": "dport",
+                    "protocol": "udp"
+                }
+            },
+            "op": "==",
+            "right": 67
+        }
+    }
+]

This translation is not correct, the 'meta protocol' match is missing in JSON. Even though it is not present when listing ruleset, the idea is that JSON is identical to standard syntax input to avoid any "short cuts" (I guess it's just a typo, other spots are fine).

[...]

tests/py/ip6/meta.t meta l4proto icmp icmp type echo-request;ok;icmp type echo-request · line 14

Quoted 11 lines
tests/py/ip6/meta.t +3 -0
index dce97f5b0fd0..2c1aee2309a9 100644
@@ -9,5 +9,8 @@ meta l4proto icmp icmp type echo-request;ok;icmp type echo-request
 meta l4proto 1 icmp type echo-request;ok;icmp type echo-request
 icmp type echo-request;ok
 
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
+

Here, 'meta protocol ip6' should be dropped.

Pablo, do you want to have another look at your patch or should I try to fix things up?

Cheers, Phil

author maintainer
#3 lore raw

Hi Phil,

Quoted 37 lines · Phil Sutter

Hi,

Quoted 2 lines · Pablo Neira Ayuso

meta protocol is meaningful in bridge, netdev and inet familiiess, do not remove this.

So you want to avoid dependency killing in above families, but:

[...]

Quoted 4 lines
src/netlink_delinearize.c +0 -0
index 5b545701e8b7..92617a46df6f 100644

[...]

Quoted 23 lines
src/netlink_delinearize.c +16 -1
@@ -2005,7 +2005,22 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 	case NFPROTO_BRIDGE:
 		break;
 	default:
-		return true;
+		if (dep->left->etype != EXPR_META ||
+		    dep->right->etype != EXPR_VALUE)
+			return false;
+
+		if (dep->left->meta.key == NFT_META_PROTOCOL) {
+			protocol = mpz_get_uint16(dep->right->value);
+
+			if (family == NFPROTO_IPV4 &&
+			    protocol == ETH_P_IP)
+				return true;
+			else if (family == NFPROTO_IPV6 &&
+				 protocol == ETH_P_IPV6)
+				return true;
+		}
+
+		return false;
 	}
 
Quoted 4 lines

src/netlink_delinearize.c static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx, · line 2026

Quoted 1 line
src/netlink_delinearize.c +0 -0
 	if (expr->left->meta.key != NFT_META_L4PROTO)

The above code only applies to families other than inet, netdev or bridge?! Am I missing something?

Yes, this code only applies to ip and ip6, this removes "meta protocol" in this cases:

rule ip x y meta protocol ip udp dport 67 => udp dport 67 rule ip6 x y meta protocol ip6 udp dport 67 => udp dport 67

I should have keep it in a separated patch actually since it is not related to bridge/inet/netdev. Sorry for the noise.

Quoted 9 lines

AFAIU, dependency killing defaults to true and examines the cases where it should not happen. So the old behaviour was for unexpected dependency expressions to just drop them instead of keeping them. The code you add above does the opposite. That's not wrong per se, but I assume there was a reason why dependency elimination was implemented "greedy", therefore I kept it that way when refactoring the code.

Quoted 2 lines
src/netlink_delinearize.c +0 -0
@@ -2015,7 +2030,8 @@ static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx,
 
Quoted 13 lines

src/netlink_delinearize.c static bool meta_may_dependency_kill(struct payload_dep_ctx *ctx, · line 2037

Quoted 8 lines
src/netlink_delinearize.c +2 -1
 	switch (dep->left->etype) {
 	case EXPR_META:
-		if (dep->left->meta.key != NFT_META_NFPROTO)
+		if (dep->left->meta.key != NFT_META_NFPROTO &&
+		    dep->left->meta.key != NFT_META_PROTOCOL)
 			return true;
 		break;
 	case EXPR_PAYLOAD:

If we continue evaluation for 'meta protocol' payload dependencies, RHS values need adjustment: with 'meta nfproto' we'll see NFPROTO_IPV4 or NFPROTO_IPV6 while with 'meta protocol' we'll see ETH_P_IP or ETH_P_IPV6. The EXPR_PAYLOAD case has such conversion code.

The chunk above the actual fix for the reported issue.

Yes, checking for ETH_P_IP and ETH_P_IPV6 would be good to narrow down the dependency killing more precisely.

Quoted 9 lines

[...]

Quoted 8 lines
tests/py/ip/meta.t +0 -0
index f733d22de2c3..fecd0caf71a7 100644
@@ -8,6 +8,8 @@ meta l4proto ipv6-icmp icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-adv
 meta l4proto 58 icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-advert
 icmpv6 type nd-router-advert;ok
 
Quoted 5 lines

tests/py/ip/meta.t meta l4proto ipv6-icmp icmpv6 type nd-router-advert;ok;icmpv6 type nd-router-adv · line 12

Quoted 2 lines
tests/py/ip/meta.t +2 -0
+meta protocol ip udp dport 67;ok
+

Hmm. Shouldn't this drop the dependency, i.e. list as 'udp dport 67' as the family is defined by table family already?

It does actually, this spews a warning if you run tests/py.

Quoted 2 lines

There should also be a case for 'meta protocol ip6 udp dport 67' which makes sure the dependency is not dropped, right?

Yes, I forgot to add it to ip/meta.t

I added it to ip6/meta.t

tests/py/ip6/meta.t +2 -0
@@ -9,5 +9,8 @@ meta l4proto icmp icmp type echo-request;ok;icmp type echo-request
 meta l4proto 1 icmp type echo-request;ok;icmp type echo-request
 icmp type echo-request;ok
 
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
 
Quoted 24 lines
Quoted 24 lines
tests/py/ip/meta.t +0 -0
 meta ibrname "br0";fail
 meta obrname "br0";fail
 
tests/py/ip/meta.t.json +13 -0
index f83864f672d5..3df31ce381fc 100644
@@ -140,3 +140,19 @@
         "accept": null
     }
 ]
+
+# meta protocol ip udp dport 67
+[
+    {
+        "match": {
+            "left": {
+                "payload": {
+                    "field": "dport",
+                    "protocol": "udp"
+                }
+            },
+            "op": "==",
+            "right": 67
Quoted 8 lines

tests/py/ip/meta.t.json line 158

Quoted 3 lines
tests/py/ip/meta.t.json +3 -0
+        }
+    }
+]

This translation is not correct, the 'meta protocol' match is missing in JSON. Even though it is not present when listing ruleset, the idea is that JSON is identical to standard syntax input to avoid any "short cuts" (I guess it's just a typo, other spots are fine).

If I understand correctly the json tests, this actually checking for the listing output, which has already removed the 'meta protocol' match.

IIRC, if I add 'meta protocol' here, the json tests report a warning.

Quoted 9 lines

[...]

Quoted 8 lines
tests/py/ip6/meta.t +0 -0
index dce97f5b0fd0..2c1aee2309a9 100644
@@ -9,5 +9,8 @@ meta l4proto icmp icmp type echo-request;ok;icmp type echo-request
 meta l4proto 1 icmp type echo-request;ok;icmp type echo-request
 icmp type echo-request;ok
 
Quoted 5 lines

tests/py/ip6/meta.t meta l4proto icmp icmp type echo-request;ok;icmp type echo-request · line 14

Quoted 3 lines
tests/py/ip6/meta.t +3 -0
+meta protocol ip udp dport 67;ok
+meta protocol ip6 udp dport 67;ok
+

Here, 'meta protocol ip6' should be dropped.

No, it is checking that this is removed. Note that this spews a warning in tests/py when running tests.

Look: If we want to fix this right(tm), we should update payload_try_merge() to actually remove this redundant dependency at rule load time, instead of doing this lazy dependency removal at listing time.

Quoted 2 lines

Pablo, do you want to have another look at your patch or should I try to fix things up?

Sure, you can follow up to refine. You're welcome.

Thanks for reviewing.

author maintainer
#4 lore raw

On Mon, Aug 30, 2021 at 07:21:14PM +0200, Pablo Neira Ayuso wrote:
[...]

Quoted 4 lines

Look: If we want to fix this right(tm), we should update payload_try_merge() to actually remove this redundant dependency at rule load time, instead of doing this lazy dependency removal at listing time.

Or actually, this match statement could be just canceled from the evaluation phase for these cases:

ip family meta protocol ip
ip6 family meta protocol ip6