[PATCH bpf-next v3 1/2] net: netfilter: Make ct zone opts configurable for bpf ct helpers

[PATCH bpf-next v3 1/2] net: netfilter: Make ct zone opts configurable for bpf ct helpers

[Brad Cowie]

Add ct zone id and direction to bpf_ct_opts so that arbitrary ct zones
can be used for xdp/tc bpf ct helper functions bpf_{xdp,skb}_ct_alloc
and bpf_{xdp,skb}_ct_lookup.

Signed-off-by: Brad Cowie <brad@faucet.nz>
---
v2 -> v3:
  - Remove whitespace changes
  - Add reserved padding options
  - If ct_zone_id is set when opts__sz isn't 16 return -EINVAL
  - Remove ct_zone_flags, not used by nf_conntrack_alloc
    or nf_conntrack_find_get
  - Update comments to reflect opts__sz can be 16 or 12

v1 -> v2:
  - Make ct zone flags/dir configurable
---
 net/netfilter/nf_conntrack_bpf.c | 68 ++++++++++++++++++++++++++------
 1 file changed, 55 insertions(+), 13 deletions(-)

diff --git a/net/netfilter/nf_conntrack_bpf.c b/net/netfilter/nf_conntrack_bpf.c
index d2492d050fe6..4a136fc3a9c0 100644
--- a/net/netfilter/nf_conntrack_bpf.c
+++ b/net/netfilter/nf_conntrack_bpf.c
@@ -32,7 +32,9 @@
  *		   -EINVAL - Passed NULL for bpf_tuple pointer
  *		   -EINVAL - opts->reserved is not 0
  *		   -EINVAL - netns_id is less than -1
- *		   -EINVAL - opts__sz isn't NF_BPF_CT_OPTS_SZ (12)
+ *		   -EINVAL - opts__sz isn't NF_BPF_CT_OPTS_SZ (16) or 12
+ *		   -EINVAL - opts->ct_zone_id set when
+			     opts__sz isn't NF_BPF_CT_OPTS_SZ (16)
  *		   -EPROTO - l4proto isn't one of IPPROTO_TCP or IPPROTO_UDP
  *		   -ENONET - No network namespace found for netns_id
  *		   -ENOENT - Conntrack lookup could not find entry for tuple
@@ -42,6 +44,8 @@
  *		 Values:
  *		   IPPROTO_TCP, IPPROTO_UDP
  * @dir:       - connection tracking tuple direction.
+ * @ct_zone_id - connection tracking zone id.
+ * @ct_zone_dir - connection tracking zone direction.
  * @reserved   - Reserved member, will be reused for more options in future
  *		 Values:
  *		   0
@@ -51,11 +55,13 @@ struct bpf_ct_opts {
 	s32 error;
 	u8 l4proto;
 	u8 dir;
-	u8 reserved[2];
+	u16 ct_zone_id;
+	u8 ct_zone_dir;
+	u8 reserved[3];
 };
 
 enum {
-	NF_BPF_CT_OPTS_SZ = 12,
+	NF_BPF_CT_OPTS_SZ = 16,
 };
 
 static int bpf_nf_ct_tuple_parse(struct bpf_sock_tuple *bpf_tuple,
@@ -104,12 +110,21 @@ __bpf_nf_ct_alloc_entry(struct net *net, struct bpf_sock_tuple *bpf_tuple,
 			u32 timeout)
 {
 	struct nf_conntrack_tuple otuple, rtuple;
+	struct nf_conntrack_zone ct_zone;
 	struct nf_conn *ct;
 	int err;
 
-	if (!opts || !bpf_tuple || opts->reserved[0] || opts->reserved[1] ||
-	    opts_len != NF_BPF_CT_OPTS_SZ)
+	if (!opts || !bpf_tuple)
 		return ERR_PTR(-EINVAL);
+	if (!(opts_len == NF_BPF_CT_OPTS_SZ || opts_len == 12))
+		return ERR_PTR(-EINVAL);
+	if (opts_len == NF_BPF_CT_OPTS_SZ) {
+		if (opts->reserved[0] || opts->reserved[1] || opts->reserved[2])
+			return ERR_PTR(-EINVAL);
+	} else {
+		if (opts->ct_zone_id)
+			return ERR_PTR(-EINVAL);
+	}
 
 	if (unlikely(opts->netns_id < BPF_F_CURRENT_NETNS))
 		return ERR_PTR(-EINVAL);
@@ -130,7 +145,16 @@ __bpf_nf_ct_alloc_entry(struct net *net, struct bpf_sock_tuple *bpf_tuple,
 			return ERR_PTR(-ENONET);
 	}
 
-	ct = nf_conntrack_alloc(net, &nf_ct_zone_dflt, &otuple, &rtuple,
+	if (opts_len == NF_BPF_CT_OPTS_SZ) {
+		if (opts->ct_zone_dir == 0)
+			opts->ct_zone_dir = NF_CT_DEFAULT_ZONE_DIR;
+		nf_ct_zone_init(&ct_zone,
+				opts->ct_zone_id, opts->ct_zone_dir, 0);
+	} else {
+		ct_zone = nf_ct_zone_dflt;
+	}
+
+	ct = nf_conntrack_alloc(net, &ct_zone, &otuple, &rtuple,
 				GFP_ATOMIC);
 	if (IS_ERR(ct))
 		goto out;
@@ -152,12 +176,21 @@ static struct nf_conn *__bpf_nf_ct_lookup(struct net *net,
 {
 	struct nf_conntrack_tuple_hash *hash;
 	struct nf_conntrack_tuple tuple;
+	struct nf_conntrack_zone ct_zone;
 	struct nf_conn *ct;
 	int err;
 
-	if (!opts || !bpf_tuple || opts->reserved[0] || opts->reserved[1] ||
-	    opts_len != NF_BPF_CT_OPTS_SZ)
+	if (!opts || !bpf_tuple)
 		return ERR_PTR(-EINVAL);
+	if (!(opts_len == NF_BPF_CT_OPTS_SZ || opts_len == 12))
+		return ERR_PTR(-EINVAL);
+	if (opts_len == NF_BPF_CT_OPTS_SZ) {
+		if (opts->reserved[0] || opts->reserved[1] || opts->reserved[2])
+			return ERR_PTR(-EINVAL);
+	} else {
+		if (opts->ct_zone_id)
+			return ERR_PTR(-EINVAL);
+	}
 	if (unlikely(opts->l4proto != IPPROTO_TCP && opts->l4proto != IPPROTO_UDP))
 		return ERR_PTR(-EPROTO);
 	if (unlikely(opts->netns_id < BPF_F_CURRENT_NETNS))
@@ -174,7 +207,16 @@ static struct nf_conn *__bpf_nf_ct_lookup(struct net *net,
 			return ERR_PTR(-ENONET);
 	}
 
-	hash = nf_conntrack_find_get(net, &nf_ct_zone_dflt, &tuple);
+	if (opts_len == NF_BPF_CT_OPTS_SZ) {
+		if (opts->ct_zone_dir == 0)
+			opts->ct_zone_dir = NF_CT_DEFAULT_ZONE_DIR;
+		nf_ct_zone_init(&ct_zone,
+				opts->ct_zone_id, opts->ct_zone_dir, 0);
+	} else {
+		ct_zone = nf_ct_zone_dflt;
+	}
+
+	hash = nf_conntrack_find_get(net, &ct_zone, &tuple);
 	if (opts->netns_id >= 0)
 		put_net(net);
 	if (!hash)
@@ -245,7 +287,7 @@ __bpf_kfunc_start_defs();
  * @opts	- Additional options for allocation (documented above)
  *		    Cannot be NULL
  * @opts__sz	- Length of the bpf_ct_opts structure
- *		    Must be NF_BPF_CT_OPTS_SZ (12)
+ *		    Must be NF_BPF_CT_OPTS_SZ (16) or 12
  */
 __bpf_kfunc struct nf_conn___init *
 bpf_xdp_ct_alloc(struct xdp_md *xdp_ctx, struct bpf_sock_tuple *bpf_tuple,
@@ -279,7 +321,7 @@ bpf_xdp_ct_alloc(struct xdp_md *xdp_ctx, struct bpf_sock_tuple *bpf_tuple,
  * @opts	- Additional options for lookup (documented above)
  *		    Cannot be NULL
  * @opts__sz	- Length of the bpf_ct_opts structure
- *		    Must be NF_BPF_CT_OPTS_SZ (12)
+ *		    Must be NF_BPF_CT_OPTS_SZ (16) or 12
  */
 __bpf_kfunc struct nf_conn *
 bpf_xdp_ct_lookup(struct xdp_md *xdp_ctx, struct bpf_sock_tuple *bpf_tuple,
@@ -312,7 +354,7 @@ bpf_xdp_ct_lookup(struct xdp_md *xdp_ctx, struct bpf_sock_tuple *bpf_tuple,
  * @opts	- Additional options for allocation (documented above)
  *		    Cannot be NULL
  * @opts__sz	- Length of the bpf_ct_opts structure
- *		    Must be NF_BPF_CT_OPTS_SZ (12)
+ *		    Must be NF_BPF_CT_OPTS_SZ (16) or 12
  */
 __bpf_kfunc struct nf_conn___init *
 bpf_skb_ct_alloc(struct __sk_buff *skb_ctx, struct bpf_sock_tuple *bpf_tuple,
@@ -347,7 +389,7 @@ bpf_skb_ct_alloc(struct __sk_buff *skb_ctx, struct bpf_sock_tuple *bpf_tuple,
  * @opts	- Additional options for lookup (documented above)
  *		    Cannot be NULL
  * @opts__sz	- Length of the bpf_ct_opts structure
- *		    Must be NF_BPF_CT_OPTS_SZ (12)
+ *		    Must be NF_BPF_CT_OPTS_SZ (16) or 12
  */
 __bpf_kfunc struct nf_conn *
 bpf_skb_ct_lookup(struct __sk_buff *skb_ctx, struct bpf_sock_tuple *bpf_tuple,
-- 
2.34.1

[PATCH bpf-next v3 2/2] selftests/bpf: Update tests for new ct zone opts for nf_conntrack kfuncs

[Brad Cowie]

Add test for allocating and looking up ct entry in a
non-default ct zone with kfuncs bpf_{xdp,skb}_ct_alloc
and bpf_{xdp,skb}_ct_lookup.

Add negative tests for looking up ct entry in a different
ct zone to where it was allocated and with a different
direction.

Signed-off-by: Brad Cowie <brad@faucet.nz>
---
v2 -> v3:
  - Test both old and new bpf_ct_opts struct definitions
  - Restore test for reserved options
  - Add test for ct_zone_dir

v1 -> v2:
  - Separate test changes into different patch
  - Add test for allocating/looking up entry in non-default ct zone
---
 tools/testing/selftests/bpf/config            |   1 +
 .../testing/selftests/bpf/prog_tests/bpf_nf.c |   6 +
 .../testing/selftests/bpf/progs/test_bpf_nf.c | 117 ++++++++++++++++--
 3 files changed, 114 insertions(+), 10 deletions(-)

diff --git a/tools/testing/selftests/bpf/config b/tools/testing/selftests/bpf/config
index eeabd798bc3a..2fb16da78dce 100644
--- a/tools/testing/selftests/bpf/config
+++ b/tools/testing/selftests/bpf/config
@@ -80,6 +80,7 @@ CONFIG_NETFILTER_XT_TARGET_CT=y
 CONFIG_NETKIT=y
 CONFIG_NF_CONNTRACK=y
 CONFIG_NF_CONNTRACK_MARK=y
+CONFIG_NF_CONNTRACK_ZONES=y
 CONFIG_NF_DEFRAG_IPV4=y
 CONFIG_NF_DEFRAG_IPV6=y
 CONFIG_NF_NAT=y
diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_nf.c b/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
index b30ff6b3b81a..b73401a71e4a 100644
--- a/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
+++ b/tools/testing/selftests/bpf/prog_tests/bpf_nf.c
@@ -122,6 +122,12 @@ static void test_bpf_nf_ct(int mode)
 	ASSERT_EQ(skel->bss->test_exist_lookup_mark, 43, "Test existing connection lookup ctmark");
 	ASSERT_EQ(skel->data->test_snat_addr, 0, "Test for source natting");
 	ASSERT_EQ(skel->data->test_dnat_addr, 0, "Test for destination natting");
+	ASSERT_EQ(skel->data->test_ct_zone_id_alloc_entry, 0, "Test for alloc new entry in specified ct zone");
+	ASSERT_EQ(skel->data->test_ct_zone_id_insert_entry, 0, "Test for insert new entry in specified ct zone");
+	ASSERT_EQ(skel->data->test_ct_zone_id_succ_lookup, 0, "Test for successful lookup in specified ct_zone");
+	ASSERT_EQ(skel->bss->test_ct_zone_dir_enoent_lookup, -ENOENT, "Test ENOENT for lookup with wrong ct zone dir");
+	ASSERT_EQ(skel->bss->test_ct_zone_id_enoent_lookup, -ENOENT, "Test ENOENT for lookup in wrong ct zone");
+
 end:
 	if (client_fd != -1)
 		close(client_fd);
diff --git a/tools/testing/selftests/bpf/progs/test_bpf_nf.c b/tools/testing/selftests/bpf/progs/test_bpf_nf.c
index 77ad8adf68da..f57cd5d6d548 100644
--- a/tools/testing/selftests/bpf/progs/test_bpf_nf.c
+++ b/tools/testing/selftests/bpf/progs/test_bpf_nf.c
@@ -9,6 +9,9 @@
 #define EINVAL 22
 #define ENOENT 2
 
+#define NF_CT_ZONE_DIR_ORIG (1 << IP_CT_DIR_ORIGINAL)
+#define NF_CT_ZONE_DIR_REPL (1 << IP_CT_DIR_REPLY)
+
 extern unsigned long CONFIG_HZ __kconfig;
 
 int test_einval_bpf_tuple = 0;
@@ -22,6 +25,11 @@ int test_eafnosupport = 0;
 int test_alloc_entry = -EINVAL;
 int test_insert_entry = -EAFNOSUPPORT;
 int test_succ_lookup = -ENOENT;
+int test_ct_zone_id_alloc_entry = -EINVAL;
+int test_ct_zone_id_insert_entry = -EAFNOSUPPORT;
+int test_ct_zone_id_succ_lookup = -ENOENT;
+int test_ct_zone_dir_enoent_lookup = 0;
+int test_ct_zone_id_enoent_lookup = 0;
 u32 test_delta_timeout = 0;
 u32 test_status = 0;
 u32 test_insert_lookup_mark = 0;
@@ -48,6 +56,16 @@ struct bpf_ct_opts___local {
 	u8 reserved[3];
 } __attribute__((preserve_access_index));
 
+struct bpf_ct_opts___new {
+	s32 netns_id;
+	s32 error;
+	u8 l4proto;
+	u8 dir;
+	u16 ct_zone_id;
+	u8 ct_zone_dir;
+	u8 reserved[3];
+} __attribute__((preserve_access_index));
+
 struct nf_conn *bpf_xdp_ct_alloc(struct xdp_md *, struct bpf_sock_tuple *, u32,
 				 struct bpf_ct_opts___local *, u32) __ksym;
 struct nf_conn *bpf_xdp_ct_lookup(struct xdp_md *, struct bpf_sock_tuple *, u32,
@@ -84,16 +102,6 @@ nf_ct_test(struct nf_conn *(*lookup_fn)(void *, struct bpf_sock_tuple *, u32,
 	else
 		test_einval_bpf_tuple = opts_def.error;
 
-	opts_def.reserved[0] = 1;
-	ct = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,
-		       sizeof(opts_def));
-	opts_def.reserved[0] = 0;
-	opts_def.l4proto = IPPROTO_TCP;
-	if (ct)
-		bpf_ct_release(ct);
-	else
-		test_einval_reserved = opts_def.error;
-
 	opts_def.netns_id = -2;
 	ct = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,
 		       sizeof(opts_def));
@@ -220,10 +228,98 @@ nf_ct_test(struct nf_conn *(*lookup_fn)(void *, struct bpf_sock_tuple *, u32,
 	}
 }
 
+static __always_inline void
+nf_ct_opts_new_test(struct nf_conn *(*lookup_fn)(void *, struct bpf_sock_tuple *, u32,
+						 struct bpf_ct_opts___new *, u32),
+		    struct nf_conn *(*alloc_fn)(void *, struct bpf_sock_tuple *, u32,
+						struct bpf_ct_opts___new *, u32),
+		    void *ctx)
+{
+	struct bpf_ct_opts___new opts_def = { .l4proto = IPPROTO_TCP, .netns_id = -1 };
+	struct bpf_sock_tuple bpf_tuple;
+	struct nf_conn *ct;
+
+	__builtin_memset(&bpf_tuple, 0, sizeof(bpf_tuple.ipv4));
+
+	opts_def.reserved[0] = 1;
+	ct = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,
+		       sizeof(opts_def));
+	opts_def.reserved[0] = 0;
+	opts_def.l4proto = IPPROTO_TCP;
+	if (ct)
+		bpf_ct_release(ct);
+	else
+		test_einval_reserved = opts_def.error;
+
+	bpf_tuple.ipv4.saddr = bpf_get_prandom_u32(); /* src IP */
+	bpf_tuple.ipv4.daddr = bpf_get_prandom_u32(); /* dst IP */
+	bpf_tuple.ipv4.sport = bpf_get_prandom_u32(); /* src port */
+	bpf_tuple.ipv4.dport = bpf_get_prandom_u32(); /* dst port */
+
+	/* use non-default ct zone */
+	opts_def.ct_zone_id = 10;
+	opts_def.ct_zone_dir = NF_CT_ZONE_DIR_ORIG;
+	ct = alloc_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,
+		      sizeof(opts_def));
+	if (ct) {
+		__u16 sport = bpf_get_prandom_u32();
+		__u16 dport = bpf_get_prandom_u32();
+		union nf_inet_addr saddr = {};
+		union nf_inet_addr daddr = {};
+		struct nf_conn *ct_ins;
+
+		bpf_ct_set_timeout(ct, 10000);
+
+		/* snat */
+		saddr.ip = bpf_get_prandom_u32();
+		bpf_ct_set_nat_info(ct, &saddr, sport, NF_NAT_MANIP_SRC___local);
+		/* dnat */
+		daddr.ip = bpf_get_prandom_u32();
+		bpf_ct_set_nat_info(ct, &daddr, dport, NF_NAT_MANIP_DST___local);
+
+		ct_ins = bpf_ct_insert_entry(ct);
+		if (ct_ins) {
+			struct nf_conn *ct_lk;
+
+			/* entry should exist in same ct zone we inserted it */
+			ct_lk = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4),
+					  &opts_def, sizeof(opts_def));
+			if (ct_lk) {
+				bpf_ct_release(ct_lk);
+				test_ct_zone_id_succ_lookup = 0;
+			}
+
+			/* entry should not exist with wrong direction */
+			opts_def.ct_zone_dir = NF_CT_ZONE_DIR_REPL;
+			ct_lk = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4),
+					  &opts_def, sizeof(opts_def));
+			opts_def.ct_zone_dir = NF_CT_ZONE_DIR_ORIG;
+			if (ct_lk)
+				bpf_ct_release(ct_lk);
+			else
+				test_ct_zone_dir_enoent_lookup = opts_def.error;
+
+			/* entry should not exist in default ct zone */
+			opts_def.ct_zone_id = 0;
+			ct_lk = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4),
+					  &opts_def, sizeof(opts_def));
+			if (ct_lk)
+				bpf_ct_release(ct_lk);
+			else
+				test_ct_zone_id_enoent_lookup = opts_def.error;
+
+			bpf_ct_release(ct_ins);
+			test_ct_zone_id_insert_entry = 0;
+		}
+		test_ct_zone_id_alloc_entry = 0;
+	}
+}
+
 SEC("xdp")
 int nf_xdp_ct_test(struct xdp_md *ctx)
 {
 	nf_ct_test((void *)bpf_xdp_ct_lookup, (void *)bpf_xdp_ct_alloc, ctx);
+	nf_ct_opts_new_test((void *)bpf_xdp_ct_lookup, (void *)bpf_xdp_ct_alloc, ctx);
 	return 0;
 }
 
@@ -231,6 +327,7 @@ SEC("tc")
 int nf_skb_ct_test(struct __sk_buff *ctx)
 {
 	nf_ct_test((void *)bpf_skb_ct_lookup, (void *)bpf_skb_ct_alloc, ctx);
+	nf_ct_opts_new_test((void *)bpf_skb_ct_lookup, (void *)bpf_skb_ct_alloc, ctx);
 	return 0;
 }
 
-- 
2.34.1

Re: [PATCH bpf-next v3 2/2] selftests/bpf: Update tests for new ct zone opts for nf_conntrack kfuncs

[Martin KaFai Lau]

On 5/7/24 10:04 PM, Brad Cowie wrote:
> @@ -84,16 +102,6 @@ nf_ct_test(struct nf_conn *(*lookup_fn)(void *, struct bpf_sock_tuple *, u32,
>   	else
>   		test_einval_bpf_tuple = opts_def.error;
>   
> -	opts_def.reserved[0] = 1;
> -	ct = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,
> -		       sizeof(opts_def));
> -	opts_def.reserved[0] = 0;
> -	opts_def.l4proto = IPPROTO_TCP;
> -	if (ct)
> -		bpf_ct_release(ct);
> -	else
> -		test_einval_reserved = opts_def.error;
> -
>   	opts_def.netns_id = -2;
>   	ct = lookup_fn(ctx, &bpf_tuple, sizeof(bpf_tuple.ipv4), &opts_def,

This non-zero reserved[0] test is still valid and useful. How about create a new 
test_einval_reserved_new for testing the new struct?

pw-bot: cr

[ Sorry for the delay. I have some backlog. ].