diff mbox series

[bpf-next,01/13] bpf: add bpf_seq_afinfo in tcp_iter_state

Message ID 20200617211537.1855673-1-yhs@fb.com
State Changes Requested
Delegated to: BPF Maintainers
Headers show
Series implement bpf iterator for tcp and udp sockets | expand

Commit Message

Yonghong Song June 17, 2020, 9:15 p.m. UTC 
A new field bpf_seq_afinfo is added to tcp_iter_state
to provide bpf tcp iterator afinfo. There are two
reasons on why we did this.

First, the current way to get afinfo from PDE_DATA
does not work for bpf iterator as its seq_file
inode does not conform to /proc/net/{tcp,tcp6}
inode structures. More specifically, anonymous
bpf iterator will use an anonymous inode which
is shared in the system and we cannot change inode
private data structure at all.

Second, bpf iterator for tcp/tcp6 wants to
traverse all tcp and tcp6 sockets in one pass
and bpf program can control whether they want
to skip one sk_family or not. Having a different
afinfo with family AF_UNSPEC make it easier
to understand in the code.

This patch does not change /proc/net/{tcp,tcp6} behavior
as the bpf_seq_afinfo will be NULL for these two proc files.

Signed-off-by: Yonghong Song <yhs@fb.com>
---
 include/net/tcp.h   |  1 +
 net/ipv4/tcp_ipv4.c | 30 ++++++++++++++++++++++++------
 2 files changed, 25 insertions(+), 6 deletions(-)

Comments

Martin KaFai Lau June 18, 2020, 6:09 p.m. UTC | #1 
On Wed, Jun 17, 2020 at 02:15:37PM -0700, Yonghong Song wrote:
> A new field bpf_seq_afinfo is added to tcp_iter_state
> to provide bpf tcp iterator afinfo. There are two
> reasons on why we did this.
> 
> First, the current way to get afinfo from PDE_DATA
> does not work for bpf iterator as its seq_file
> inode does not conform to /proc/net/{tcp,tcp6}
> inode structures. More specifically, anonymous
> bpf iterator will use an anonymous inode which
> is shared in the system and we cannot change inode
> private data structure at all.
> 
> Second, bpf iterator for tcp/tcp6 wants to
> traverse all tcp and tcp6 sockets in one pass
> and bpf program can control whether they want
> to skip one sk_family or not. Having a different
> afinfo with family AF_UNSPEC make it easier
> to understand in the code.
Acked-by: Martin KaFai Lau <kafai@fb.com>
diff mbox series

Patch

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 4de9485f73d9..eab1c7d0facb 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1935,6 +1935,7 @@  struct tcp_iter_state {
 	struct seq_net_private	p;
 	enum tcp_seq_states	state;
 	struct sock		*syn_wait_sk;
+	struct tcp_seq_afinfo	*bpf_seq_afinfo;
 	int			bucket, offset, sbucket, num;
 	loff_t			last_pos;
 };
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index ad6435ba6d72..9cb65ee4ec63 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2211,13 +2211,18 @@  EXPORT_SYMBOL(tcp_v4_destroy_sock);
  */
 static void *listening_get_next(struct seq_file *seq, void *cur)
 {
-	struct tcp_seq_afinfo *afinfo = PDE_DATA(file_inode(seq->file));
+	struct tcp_seq_afinfo *afinfo;
 	struct tcp_iter_state *st = seq->private;
 	struct net *net = seq_file_net(seq);
 	struct inet_listen_hashbucket *ilb;
 	struct hlist_nulls_node *node;
 	struct sock *sk = cur;
 
+	if (st->bpf_seq_afinfo)
+		afinfo = st->bpf_seq_afinfo;
+	else
+		afinfo = PDE_DATA(file_inode(seq->file));
+
 	if (!sk) {
 get_head:
 		ilb = &tcp_hashinfo.listening_hash[st->bucket];
@@ -2235,7 +2240,8 @@  static void *listening_get_next(struct seq_file *seq, void *cur)
 	sk_nulls_for_each_from(sk, node) {
 		if (!net_eq(sock_net(sk), net))
 			continue;
-		if (sk->sk_family == afinfo->family)
+		if (afinfo->family == AF_UNSPEC ||
+		    sk->sk_family == afinfo->family)
 			return sk;
 	}
 	spin_unlock(&ilb->lock);
@@ -2272,11 +2278,16 @@  static inline bool empty_bucket(const struct tcp_iter_state *st)
  */
 static void *established_get_first(struct seq_file *seq)
 {
-	struct tcp_seq_afinfo *afinfo = PDE_DATA(file_inode(seq->file));
+	struct tcp_seq_afinfo *afinfo;
 	struct tcp_iter_state *st = seq->private;
 	struct net *net = seq_file_net(seq);
 	void *rc = NULL;
 
+	if (st->bpf_seq_afinfo)
+		afinfo = st->bpf_seq_afinfo;
+	else
+		afinfo = PDE_DATA(file_inode(seq->file));
+
 	st->offset = 0;
 	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
 		struct sock *sk;
@@ -2289,7 +2300,8 @@  static void *established_get_first(struct seq_file *seq)
 
 		spin_lock_bh(lock);
 		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
-			if (sk->sk_family != afinfo->family ||
+			if ((afinfo->family != AF_UNSPEC &&
+			     sk->sk_family != afinfo->family) ||
 			    !net_eq(sock_net(sk), net)) {
 				continue;
 			}
@@ -2304,19 +2316,25 @@  static void *established_get_first(struct seq_file *seq)
 
 static void *established_get_next(struct seq_file *seq, void *cur)
 {
-	struct tcp_seq_afinfo *afinfo = PDE_DATA(file_inode(seq->file));
+	struct tcp_seq_afinfo *afinfo;
 	struct sock *sk = cur;
 	struct hlist_nulls_node *node;
 	struct tcp_iter_state *st = seq->private;
 	struct net *net = seq_file_net(seq);
 
+	if (st->bpf_seq_afinfo)
+		afinfo = st->bpf_seq_afinfo;
+	else
+		afinfo = PDE_DATA(file_inode(seq->file));
+
 	++st->num;
 	++st->offset;
 
 	sk = sk_nulls_next(sk);
 
 	sk_nulls_for_each_from(sk, node) {
-		if (sk->sk_family == afinfo->family &&
+		if ((afinfo->family == AF_UNSPEC ||
+		     sk->sk_family == afinfo->family) &&
 		    net_eq(sock_net(sk), net))
 			return sk;
 	}