summaryrefslogtreecommitdiff
path: root/net/ipv4/inet_diag.c
diff options
context:
space:
mode:
authorEric Dumazet <edumazet@google.com>2015-10-02 21:43:32 +0300
committerDavid S. Miller <davem@davemloft.net>2015-10-03 14:32:41 +0300
commit079096f103faca2dd87342cca6f23d4b34da8871 (patch)
treefa3fb0fdc064f1611c464384e70a7a402179808f /net/ipv4/inet_diag.c
parent2feda34192a379f8b35a7c6c5826b2f23e884f32 (diff)
downloadlinux-079096f103faca2dd87342cca6f23d4b34da8871.tar.xz
tcp/dccp: install syn_recv requests into ehash table
In this patch, we insert request sockets into TCP/DCCP regular ehash table (where ESTABLISHED and TIMEWAIT sockets are) instead of using the per listener hash table. ACK packets find SYN_RECV pseudo sockets without having to find and lock the listener. In nominal conditions, this halves pressure on listener lock. Note that this will allow for SO_REUSEPORT refinements, so that we can select a listener using cpu/numa affinities instead of the prior 'consistent hash', since only SYN packets will apply this selection logic. We will shrink listen_sock in the following patch to ease code review. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ying Cai <ycai@google.com> Cc: Willem de Bruijn <willemb@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/ipv4/inet_diag.c')
-rw-r--r--net/ipv4/inet_diag.c96
1 files changed, 8 insertions, 88 deletions
diff --git a/net/ipv4/inet_diag.c b/net/ipv4/inet_diag.c
index 0ac1d68dc8a6..ab9f8a66615d 100644
--- a/net/ipv4/inet_diag.c
+++ b/net/ipv4/inet_diag.c
@@ -730,91 +730,21 @@ static void twsk_build_assert(void)
#endif
}
-static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
- struct netlink_callback *cb,
- const struct inet_diag_req_v2 *r,
- const struct nlattr *bc)
-{
- struct inet_connection_sock *icsk = inet_csk(sk);
- struct inet_sock *inet = inet_sk(sk);
- struct inet_diag_entry entry;
- int j, s_j, reqnum, s_reqnum;
- struct listen_sock *lopt;
- int err = 0;
-
- s_j = cb->args[3];
- s_reqnum = cb->args[4];
-
- if (s_j > 0)
- s_j--;
-
- entry.family = sk->sk_family;
-
- spin_lock(&icsk->icsk_accept_queue.syn_wait_lock);
-
- lopt = icsk->icsk_accept_queue.listen_opt;
- if (!lopt || !reqsk_queue_len(&icsk->icsk_accept_queue))
- goto out;
-
- if (bc) {
- entry.sport = inet->inet_num;
- entry.userlocks = sk->sk_userlocks;
- }
-
- for (j = s_j; j < lopt->nr_table_entries; j++) {
- struct request_sock *req, *head = lopt->syn_table[j];
-
- reqnum = 0;
- for (req = head; req; reqnum++, req = req->dl_next) {
- struct inet_request_sock *ireq = inet_rsk(req);
-
- if (reqnum < s_reqnum)
- continue;
- if (r->id.idiag_dport != ireq->ir_rmt_port &&
- r->id.idiag_dport)
- continue;
-
- if (bc) {
- /* Note: entry.sport and entry.userlocks are already set */
- entry_fill_addrs(&entry, req_to_sk(req));
- entry.dport = ntohs(ireq->ir_rmt_port);
-
- if (!inet_diag_bc_run(bc, &entry))
- continue;
- }
-
- err = inet_req_diag_fill(req_to_sk(req), skb,
- NETLINK_CB(cb->skb).portid,
- cb->nlh->nlmsg_seq,
- NLM_F_MULTI, cb->nlh);
- if (err < 0) {
- cb->args[3] = j + 1;
- cb->args[4] = reqnum;
- goto out;
- }
- }
-
- s_reqnum = 0;
- }
-
-out:
- spin_unlock(&icsk->icsk_accept_queue.syn_wait_lock);
-
- return err;
-}
-
void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
struct netlink_callback *cb,
const struct inet_diag_req_v2 *r, struct nlattr *bc)
{
struct net *net = sock_net(skb->sk);
int i, num, s_i, s_num;
+ u32 idiag_states = r->idiag_states;
+ if (idiag_states & TCPF_SYN_RECV)
+ idiag_states |= TCPF_NEW_SYN_RECV;
s_i = cb->args[1];
s_num = num = cb->args[2];
if (cb->args[0] == 0) {
- if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
+ if (!(idiag_states & TCPF_LISTEN))
goto skip_listen_ht;
for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
@@ -844,21 +774,11 @@ void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
r->id.idiag_sport)
goto next_listen;
- if (!(r->idiag_states & TCPF_LISTEN) ||
- r->id.idiag_dport ||
+ if (r->id.idiag_dport ||
cb->args[3] > 0)
- goto syn_recv;
-
- if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
- spin_unlock_bh(&ilb->lock);
- goto done;
- }
-
-syn_recv:
- if (!(r->idiag_states & TCPF_SYN_RECV))
goto next_listen;
- if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) {
+ if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
spin_unlock_bh(&ilb->lock);
goto done;
}
@@ -879,7 +799,7 @@ skip_listen_ht:
s_i = num = s_num = 0;
}
- if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
+ if (!(idiag_states & ~TCPF_LISTEN))
goto out;
for (i = s_i; i <= hashinfo->ehash_mask; i++) {
@@ -906,7 +826,7 @@ skip_listen_ht:
goto next_normal;
state = (sk->sk_state == TCP_TIME_WAIT) ?
inet_twsk(sk)->tw_substate : sk->sk_state;
- if (!(r->idiag_states & (1 << state)))
+ if (!(idiag_states & (1 << state)))
goto next_normal;
if (r->sdiag_family != AF_UNSPEC &&
sk->sk_family != r->sdiag_family)