6.1-stable review patch. If anyone has any objections, please let me know.
------------------
From: Eric Dumazet edumazet@google.com
[ Upstream commit a853c609504e2d1d83e71285e3622fda1f1451d8 ]
All inet_getpeer() callers except ip4_frag_init() don't need to acquire a permanent refcount on the inetpeer.
They can switch to full RCU protection.
Move the refcount_inc_not_zero() into ip4_frag_init(), so that all the other callers no longer have to perform a pair of expensive atomic operations on a possibly contended cache line.
inet_putpeer() no longer needs to be exported.
After this patch, my DUT can receive 8,400,000 UDP packets per second targeting closed ports, using 50% less cpu cycles than before.
Also change two calls to l3mdev_master_ifindex() by l3mdev_master_ifindex_rcu() (Ido ideas)
Fixes: 8c2bd38b95f7 ("icmp: change the order of rate limits") Signed-off-by: Eric Dumazet edumazet@google.com Link: https://patch.msgid.link/20241215175629.1248773-5-edumazet@google.com Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org --- net/ipv4/icmp.c | 9 ++++----- net/ipv4/inetpeer.c | 8 ++------ net/ipv4/ip_fragment.c | 15 ++++++++++----- net/ipv4/route.c | 15 ++++++++------- net/ipv6/icmp.c | 4 ++-- net/ipv6/ip6_output.c | 4 ++-- net/ipv6/ndisc.c | 6 ++++-- 7 files changed, 32 insertions(+), 29 deletions(-)
diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c index 203734e29d462..a6adf6a2ec4b5 100644 --- a/net/ipv4/icmp.c +++ b/net/ipv4/icmp.c @@ -316,7 +316,6 @@ static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, struct dst_entry *dst = &rt->dst; struct inet_peer *peer; bool rc = true; - int vif;
if (!apply_ratelimit) return true; @@ -325,12 +324,12 @@ static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt, if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) goto out;
- vif = l3mdev_master_ifindex(dst->dev); - peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, vif); + rcu_read_lock(); + peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, + l3mdev_master_ifindex_rcu(dst->dev)); rc = inet_peer_xrlim_allow(peer, READ_ONCE(net->ipv4.sysctl_icmp_ratelimit)); - if (peer) - inet_putpeer(peer); + rcu_read_unlock(); out: if (!rc) __ICMP_INC_STATS(net, ICMP_MIB_RATELIMITHOST); diff --git a/net/ipv4/inetpeer.c b/net/ipv4/inetpeer.c index 596e2c3a8551f..23896b6b8417d 100644 --- a/net/ipv4/inetpeer.c +++ b/net/ipv4/inetpeer.c @@ -112,8 +112,6 @@ static struct inet_peer *lookup(const struct inetpeer_addr *daddr, p = rb_entry(parent, struct inet_peer, rb_node); cmp = inetpeer_addr_cmp(daddr, &p->daddr); if (cmp == 0) { - if (!refcount_inc_not_zero(&p->refcnt)) - break; now = jiffies; if (READ_ONCE(p->dtime) != now) WRITE_ONCE(p->dtime, now); @@ -177,6 +175,7 @@ static void inet_peer_gc(struct inet_peer_base *base, } }
+/* Must be called under RCU : No refcount change is done here. */ struct inet_peer *inet_getpeer(struct inet_peer_base *base, const struct inetpeer_addr *daddr) { @@ -187,10 +186,8 @@ struct inet_peer *inet_getpeer(struct inet_peer_base *base, /* Attempt a lockless lookup first. * Because of a concurrent writer, we might not find an existing entry. */ - rcu_read_lock(); seq = read_seqbegin(&base->lock); p = lookup(daddr, base, seq, NULL, &gc_cnt, &parent, &pp); - rcu_read_unlock();
if (p) return p; @@ -208,7 +205,7 @@ struct inet_peer *inet_getpeer(struct inet_peer_base *base, if (p) { p->daddr = *daddr; p->dtime = (__u32)jiffies; - refcount_set(&p->refcnt, 2); + refcount_set(&p->refcnt, 1); atomic_set(&p->rid, 0); p->metrics[RTAX_LOCK-1] = INETPEER_METRICS_NEW; p->rate_tokens = 0; @@ -236,7 +233,6 @@ void inet_putpeer(struct inet_peer *p) if (refcount_dec_and_test(&p->refcnt)) call_rcu(&p->rcu, inetpeer_free_rcu); } -EXPORT_SYMBOL_GPL(inet_putpeer);
/* * Check transmit rate limitation for given message. diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c index 1427a94fc77a0..0ed999fdca2d7 100644 --- a/net/ipv4/ip_fragment.c +++ b/net/ipv4/ip_fragment.c @@ -82,15 +82,20 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *skb, static void ip4_frag_init(struct inet_frag_queue *q, const void *a) { struct ipq *qp = container_of(q, struct ipq, q); - struct net *net = q->fqdir->net; - const struct frag_v4_compare_key *key = a; + struct net *net = q->fqdir->net; + struct inet_peer *p = NULL;
q->key.v4 = *key; qp->ecn = 0; - qp->peer = q->fqdir->max_dist ? - inet_getpeer_v4(net->ipv4.peers, key->saddr, key->vif) : - NULL; + if (q->fqdir->max_dist) { + rcu_read_lock(); + p = inet_getpeer_v4(net->ipv4.peers, key->saddr, key->vif); + if (p && !refcount_inc_not_zero(&p->refcnt)) + p = NULL; + rcu_read_unlock(); + } + qp->peer = p; }
static void ip4_frag_free(struct inet_frag_queue *q) diff --git a/net/ipv4/route.c b/net/ipv4/route.c index ae83b86fb209d..f877a96fd1eb5 100644 --- a/net/ipv4/route.c +++ b/net/ipv4/route.c @@ -882,11 +882,11 @@ void ip_rt_send_redirect(struct sk_buff *skb) } log_martians = IN_DEV_LOG_MARTIANS(in_dev); vif = l3mdev_master_ifindex_rcu(rt->dst.dev); - rcu_read_unlock();
net = dev_net(rt->dst.dev); peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, vif); if (!peer) { + rcu_read_unlock(); icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt_nexthop(rt, ip_hdr(skb)->daddr)); return; @@ -905,7 +905,7 @@ void ip_rt_send_redirect(struct sk_buff *skb) */ if (peer->n_redirects >= ip_rt_redirect_number) { peer->rate_last = jiffies; - goto out_put_peer; + goto out_unlock; }
/* Check for load limit; set rate_last to the latest sent @@ -926,8 +926,8 @@ void ip_rt_send_redirect(struct sk_buff *skb) &ip_hdr(skb)->saddr, inet_iif(skb), &ip_hdr(skb)->daddr, &gw); } -out_put_peer: - inet_putpeer(peer); +out_unlock: + rcu_read_unlock(); }
static int ip_error(struct sk_buff *skb) @@ -987,9 +987,9 @@ static int ip_error(struct sk_buff *skb) break; }
+ rcu_read_lock(); peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, - l3mdev_master_ifindex(skb->dev)); - + l3mdev_master_ifindex_rcu(skb->dev)); send = true; if (peer) { now = jiffies; @@ -1001,8 +1001,9 @@ static int ip_error(struct sk_buff *skb) peer->rate_tokens -= ip_rt_error_cost; else send = false; - inet_putpeer(peer); } + rcu_read_unlock(); + if (send) icmp_send(skb, ICMP_DEST_UNREACH, code, 0);
diff --git a/net/ipv6/icmp.c b/net/ipv6/icmp.c index ad34482186a9c..7d88fd314c390 100644 --- a/net/ipv6/icmp.c +++ b/net/ipv6/icmp.c @@ -222,10 +222,10 @@ static bool icmpv6_xrlim_allow(struct sock *sk, u8 type, if (rt->rt6i_dst.plen < 128) tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
+ rcu_read_lock(); peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr); res = inet_peer_xrlim_allow(peer, tmo); - if (peer) - inet_putpeer(peer); + rcu_read_unlock(); } if (!res) __ICMP6_INC_STATS(net, ip6_dst_idev(dst), diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c index 5332aeddf9277..d7f7a714bd232 100644 --- a/net/ipv6/ip6_output.c +++ b/net/ipv6/ip6_output.c @@ -610,6 +610,7 @@ int ip6_forward(struct sk_buff *skb) else target = &hdr->daddr;
+ rcu_read_lock(); peer = inet_getpeer_v6(net->ipv6.peers, &hdr->daddr);
/* Limit redirects both by destination (here) @@ -617,8 +618,7 @@ int ip6_forward(struct sk_buff *skb) */ if (inet_peer_xrlim_allow(peer, 1*HZ)) ndisc_send_redirect(skb, target); - if (peer) - inet_putpeer(peer); + rcu_read_unlock(); } else { int addrtype = ipv6_addr_type(&hdr->saddr);
diff --git a/net/ipv6/ndisc.c b/net/ipv6/ndisc.c index d1eb0e324b7c0..44d3e6ab0c7d4 100644 --- a/net/ipv6/ndisc.c +++ b/net/ipv6/ndisc.c @@ -1721,10 +1721,12 @@ void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target) "Redirect: destination is not a neighbour\n"); goto release; } + + rcu_read_lock(); peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr); ret = inet_peer_xrlim_allow(peer, 1*HZ); - if (peer) - inet_putpeer(peer); + rcu_read_unlock(); + if (!ret) goto release;