| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <20070501.044317.47815539.yoshfuji@linux-ipv6.org>
Date: Tue, 01 May 2007 04:43:17 +0900 (JST)
From: YOSHIFUJI Hideaki / 吉藤英明
<yoshfuji@...ux-ipv6.org>
To: dada1@...mosbay.com, davem@...emloft.net
Cc: zacco@...hu, baruch@...en.org, netdev@...r.kernel.org,
yoshfuji@...ux-ipv6.org
Subject: Re: many sockets, slow sendto
In article <20070430144715.b0c03c83.dada1@...mosbay.com> (at Mon, 30 Apr 2007 14:47:15 +0200), Eric Dumazet <dada1@...mosbay.com> says:
> Also, I am not sure we need to use all 128 bits of IPV6 address, maybe the 64 low order bits are enough ?
Well, maybe, but in IPv6, auto-configured addresses on an interface have
the same 64-bit LSBs. So, I'd keep as-is so far.
Here's the take 2, mainly for fixing UDP-Lite side.
Regards,
----
[IPV6]: Convert UDP(-Lite} to new 2-pass algos.
Some inputs from Eric Dumazet <dada1@...mosbay.com>.
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@...ux-ipv6.org>
---
diff --git a/include/net/udp.h b/include/net/udp.h
index 98755eb..2c06017 100644
--- a/include/net/udp.h
+++ b/include/net/udp.h
@@ -120,8 +120,12 @@ static inline void udp_lib_close(struct sock *sk, long timeout)
/* net/ipv4/udp.c */
+extern unsigned int udp_hash_port_and_rcvaddr(__u16 port,
+ const struct sock *sk);
extern int udp_get_port(struct sock *sk, unsigned short snum,
- int (*saddr_cmp)(const struct sock *, const struct sock *));
+ int (*saddr_cmp)(const struct sock *, const struct sock *),
+ unsigned int (*hash_port_rcvaddr)(__u16 port,
+ const struct sock *sk));
extern void udp_err(struct sk_buff *, u32);
extern int udp_sendmsg(struct kiocb *iocb, struct sock *sk,
diff --git a/include/net/udplite.h b/include/net/udplite.h
index 635b0ea..6da0d41 100644
--- a/include/net/udplite.h
+++ b/include/net/udplite.h
@@ -120,5 +120,6 @@ static inline __wsum udplite_csum_outgoing(struct sock *sk, struct sk_buff *skb)
extern void udplite4_register(void);
extern int udplite_get_port(struct sock *sk, unsigned short snum,
- int (*scmp)(const struct sock *, const struct sock *));
+ int (*scmp)(const struct sock *, const struct sock *),
+ unsigned int (*uhash)(__u16, const struct sock *));
#endif /* _UDPLITE_H */
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 1449707..9d4293d 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -125,6 +125,12 @@ static inline unsigned int hash_port_and_addr(__u16 port, __be32 addr)
return port ^ addr;
}
+unsigned int udp4_hash_port_and_rcvaddr(__u16 port,
+ const struct sock *sk)
+{
+ return hash_port_and_addr(port, inet_sk(sk)->rcv_saddr);
+}
+
static inline int __udp_lib_port_inuse(unsigned int hash, int port,
__be32 daddr, struct hlist_head udptable[])
{
@@ -156,7 +162,9 @@ static inline int __udp_lib_port_inuse(unsigned int hash, int port,
int __udp_lib_get_port(struct sock *sk, unsigned short snum,
struct hlist_head udptable[], int *port_rover,
int (*saddr_comp)(const struct sock *sk1,
- const struct sock *sk2 ) )
+ const struct sock *sk2),
+ unsigned int (*hash_port_rcvaddr)(__u16 port,
+ const struct sock *sk))
{
struct hlist_node *node;
struct hlist_head *head;
@@ -176,8 +184,7 @@ int __udp_lib_get_port(struct sock *sk, unsigned short snum,
for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
int size;
- hash = hash_port_and_addr(result,
- inet_sk(sk)->rcv_saddr);
+ hash = hash_port_rcvaddr(result, sk);
head = &udptable[hash & (UDP_HTABLE_SIZE - 1)];
if (hlist_empty(head)) {
if (result > sysctl_local_port_range[1])
@@ -203,8 +210,7 @@ int __udp_lib_get_port(struct sock *sk, unsigned short snum,
result = sysctl_local_port_range[0]
+ ((result - sysctl_local_port_range[0]) &
(UDP_HTABLE_SIZE - 1));
- hash = hash_port_and_addr(result,
- inet_sk(sk)->rcv_saddr);
+ hash = hash_port_rcvaddr(result, sk);
if (! __udp_lib_port_inuse(hash, result,
inet_sk(sk)->rcv_saddr, udptable))
break;
@@ -214,7 +220,7 @@ int __udp_lib_get_port(struct sock *sk, unsigned short snum,
gotit:
*port_rover = snum = result;
} else {
- hash = hash_port_and_addr(snum, inet_sk(sk)->rcv_saddr);
+ hash = hash_port_rcvaddr(snum, sk);
head = &udptable[hash & (UDP_HTABLE_SIZE - 1)];
sk_for_each(sk2, node, head)
@@ -241,9 +247,11 @@ fail:
}
int udp_get_port(struct sock *sk, unsigned short snum,
- int (*scmp)(const struct sock *, const struct sock *))
+ int (*scmp)(const struct sock *, const struct sock *),
+ unsigned int (*uhash)(u16 port, const struct sock *))
{
- return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover, scmp);
+ return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover,
+ scmp, uhash);
}
int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
@@ -257,7 +265,8 @@ int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
{
- return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
+ return udp_get_port(sk, snum, ipv4_rcv_saddr_equal,
+ udp4_hash_port_and_rcvaddr);
}
/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
@@ -328,8 +337,8 @@ found:
}
static inline struct sock *udp_v4_mcast_next(
- struct sock *sk,
- unsigned int hnum, __be16 loc_port, __be32 loc_addr,
+ struct sock *sk, unsigned int hnum,
+ __be16 loc_port, __be32 loc_addr,
__be16 rmt_port, __be32 rmt_addr,
int dif)
{
diff --git a/net/ipv4/udp_impl.h b/net/ipv4/udp_impl.h
index 820a477..d7216c8 100644
--- a/net/ipv4/udp_impl.h
+++ b/net/ipv4/udp_impl.h
@@ -10,7 +10,8 @@ extern void __udp4_lib_err(struct sk_buff *, u32, struct hlist_head []);
extern int __udp_lib_get_port(struct sock *sk, unsigned short snum,
struct hlist_head udptable[], int *port_rover,
- int (*)(const struct sock*,const struct sock*));
+ int (*)(const struct sock*,const struct sock*),
+ unsigned int (*)(__u16, const struct sock*));
extern int ipv4_rcv_saddr_equal(const struct sock *, const struct sock *);
diff --git a/net/ipv4/udplite.c b/net/ipv4/udplite.c
index f34fd68..4c4e0fd 100644
--- a/net/ipv4/udplite.c
+++ b/net/ipv4/udplite.c
@@ -18,15 +18,22 @@ DEFINE_SNMP_STAT(struct udp_mib, udplite_statistics) __read_mostly;
struct hlist_head udplite_hash[UDP_HTABLE_SIZE];
static int udplite_port_rover;
+extern unsigned int udp4_hash_port_and_rcvaddr(__u16 port,
+ const struct sock *sk);
+
int udplite_get_port(struct sock *sk, unsigned short p,
- int (*c)(const struct sock *, const struct sock *))
+ int (*c)(const struct sock *, const struct sock *),
+ unsigned int (*h)(__u16, const struct sock *))
{
- return __udp_lib_get_port(sk, p, udplite_hash, &udplite_port_rover, c);
+ return __udp_lib_get_port(sk, p, udplite_hash, &udplite_port_rover,
+ c, h);
}
static int udplite_v4_get_port(struct sock *sk, unsigned short snum)
{
- return udplite_get_port(sk, snum, ipv4_rcv_saddr_equal);
+ return udplite_get_port(sk, snum,
+ ipv4_rcv_saddr_equal,
+ udp4_hash_port_and_rcvaddr);
}
static int udplite_rcv(struct sk_buff *skb)
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index b083c09..1d05a69 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -52,56 +52,95 @@
DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly;
+static inline unsigned int udp6_hash_port(__u16 port)
+{
+ return port;
+}
+
+static inline unsigned int udp6_hash_port_and_addr(__u16 port,
+ const struct in6_addr *addr)
+{
+ u32 hash = 0;
+ hash = ((__force u32) addr->s6_addr32[0]) ^
+ ((__force u32) addr->s6_addr32[1]) ^
+ ((__force u32) addr->s6_addr32[2]) ^
+ ((__force u32) addr->s6_addr32[3]);
+ hash ^= hash >> 16;
+ hash ^= hash >> 8;
+ return udp6_hash_port(port) ^ hash;
+}
+
+unsigned int udp6_hash_port_and_rcvaddr(__u16 port,
+ const struct sock *sk)
+{
+ return udp6_hash_port_and_addr(port, &inet6_sk(sk)->rcv_saddr);
+}
+
static inline int udp_v6_get_port(struct sock *sk, unsigned short snum)
{
- return udp_get_port(sk, snum, ipv6_rcv_saddr_equal);
+ return udp_get_port(sk, snum,
+ ipv6_rcv_saddr_equal,
+ udp6_hash_port_and_rcvaddr);
}
static struct sock *__udp6_lib_lookup(struct in6_addr *saddr, __be16 sport,
struct in6_addr *daddr, __be16 dport,
int dif, struct hlist_head udptable[])
{
- struct sock *sk, *result = NULL;
+ struct sock *sk = NULL, *result = NULL;
struct hlist_node *node;
- unsigned short hnum = ntohs(dport);
- int badness = -1;
+ unsigned hash, hashwild;
+ int score, best = -1;
+
+ hash = udp6_hash_port_and_addr(ntohs(dport), saddr);
+ hashwild = udp6_hash_port(ntohs(dport));
read_lock(&udp_hash_lock);
- sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
+lookup:
+ sk_for_each(sk, node, &udptable[hash & (UDP_HTABLE_SIZE - 1)]) {
struct inet_sock *inet = inet_sk(sk);
+ struct ipv6_pinfo *np = inet6_sk(sk);
- if (sk->sk_hash == hnum && sk->sk_family == PF_INET6) {
- struct ipv6_pinfo *np = inet6_sk(sk);
- int score = 0;
- if (inet->dport) {
- if (inet->dport != sport)
- continue;
- score++;
- }
- if (!ipv6_addr_any(&np->rcv_saddr)) {
- if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
- continue;
- score++;
- }
- if (!ipv6_addr_any(&np->daddr)) {
- if (!ipv6_addr_equal(&np->daddr, saddr))
- continue;
- score++;
- }
- if (sk->sk_bound_dev_if) {
- if (sk->sk_bound_dev_if != dif)
- continue;
- score++;
- }
- if (score == 4) {
- result = sk;
- break;
- } else if (score > badness) {
- result = sk;
- badness = score;
- }
+ if (sk->sk_hash != hash || sk->sk_family != PF_INET6 ||
+ inet->num != dport)
+ continue;
+
+ score = 0;
+
+ if (inet->dport) {
+ if (inet->dport != sport)
+ continue;
+ score++;
}
+ if (!ipv6_addr_any(&np->rcv_saddr)) {
+ if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
+ continue;
+ score++;
+ }
+ if (!ipv6_addr_any(&np->daddr)) {
+ if (!ipv6_addr_equal(&np->daddr, saddr))
+ continue;
+ score++;
+ }
+ if (sk->sk_bound_dev_if) {
+ if (sk->sk_bound_dev_if != dif)
+ continue;
+ score++;
+ }
+ if (score == 4) {
+ result = sk;
+ goto found;
+ } else if (score > best) {
+ result = sk;
+ best = score;
+ }
+ }
+
+ if (hash != hashwild) {
+ hash = hashwild;
+ goto lookup;
}
+found:
if (result)
sock_hold(result);
read_unlock(&udp_hash_lock);
@@ -302,38 +341,41 @@ drop:
}
static struct sock *udp_v6_mcast_next(struct sock *sk,
+ unsigned int hnum,
__be16 loc_port, struct in6_addr *loc_addr,
__be16 rmt_port, struct in6_addr *rmt_addr,
int dif)
{
struct hlist_node *node;
struct sock *s = sk;
- unsigned short num = ntohs(loc_port);
sk_for_each_from(s, node) {
struct inet_sock *inet = inet_sk(s);
+ struct ipv6_pinfo *np = inet6_sk(s);
- if (s->sk_hash == num && s->sk_family == PF_INET6) {
- struct ipv6_pinfo *np = inet6_sk(s);
- if (inet->dport) {
- if (inet->dport != rmt_port)
- continue;
- }
- if (!ipv6_addr_any(&np->daddr) &&
- !ipv6_addr_equal(&np->daddr, rmt_addr))
- continue;
+ if (s->sk_hash != hnum || s->sk_family != PF_INET6 ||
+ inet->num != loc_port)
+ continue;
- if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
+ if (inet->dport) {
+ if (inet->dport != rmt_port)
continue;
+ }
+ if (!ipv6_addr_any(&np->daddr) &&
+ !ipv6_addr_equal(&np->daddr, rmt_addr))
+ continue;
- if (!ipv6_addr_any(&np->rcv_saddr)) {
- if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
- continue;
- }
- if (!inet6_mc_check(s, loc_addr, rmt_addr))
+ if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
+ continue;
+
+ if (!ipv6_addr_any(&np->rcv_saddr)) {
+ if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
continue;
- return s;
}
+
+ if (!inet6_mc_check(s, loc_addr, rmt_addr))
+ continue;
+ return s;
}
return NULL;
}
@@ -348,20 +390,42 @@ static int __udp6_lib_mcast_deliver(struct sk_buff *skb, struct in6_addr *saddr,
struct sock *sk, *sk2;
const struct udphdr *uh = udp_hdr(skb);
int dif;
+ int hport = ntohs(uh->dest);
+ unsigned int hash = udp6_hash_port_and_addr(ntohs(uh->dest), daddr);
+ unsigned int hashwild = udp6_hash_port(ntohs(uh->dest));
- read_lock(&udp_hash_lock);
- sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
dif = inet6_iif(skb);
- sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
+
+ read_lock(&udp_hash_lock);
+redo:
+ sk = sk_head(&udptable[hash & (UDP_HTABLE_SIZE - 1)]);
+ sk = udp_v6_mcast_next(sk, hash, uh->dest, daddr, uh->source, saddr, dif);
if (!sk) {
+ if (hash != hashwild) {
+ hash = hashwild;
+ goto redo;
+ }
kfree_skb(skb);
goto out;
}
sk2 = sk;
- while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
- uh->source, saddr, dif))) {
- struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC);
+ while(1) {
+ struct sk_buff *buff;
+
+ sk2 = udp_v6_mcast_next(sk_next(sk2), hash, hport, daddr,
+ uh->source, saddr, dif);
+ if (!sk2) {
+ if (hash == hashwild)
+ break;
+ hash = hashwild;
+ sk2 = sk_head(&udptable[hash & (UDP_HTABLE_SIZE - 1)]);
+ sk2 = udp_v6_mcast_next(sk2, hash, uh->dest, daddr, uh->source, saddr, dif);
+ if (!sk2)
+ break;
+ }
+
+ buff = skb_clone(skb, GFP_ATOMIC);
if (buff)
udpv6_queue_rcv_skb(sk2, buff);
}
diff --git a/net/ipv6/udplite.c b/net/ipv6/udplite.c
index f54016a..797d76d 100644
--- a/net/ipv6/udplite.c
+++ b/net/ipv6/udplite.c
@@ -17,6 +17,9 @@
DEFINE_SNMP_STAT(struct udp_mib, udplite_stats_in6) __read_mostly;
+extern unsigned int udp6_hash_port_and_rcvaddr(__u16 port,
+ const struct sock *sk);
+
static int udplitev6_rcv(struct sk_buff **pskb)
{
return __udp6_lib_rcv(pskb, udplite_hash, IPPROTO_UDPLITE);
@@ -37,7 +40,9 @@ static struct inet6_protocol udplitev6_protocol = {
static int udplite_v6_get_port(struct sock *sk, unsigned short snum)
{
- return udplite_get_port(sk, snum, ipv6_rcv_saddr_equal);
+ return udplite_get_port(sk, snum,
+ ipv6_rcv_saddr_equal,
+ udp6_hash_port_and_rcvaddr);
}
struct proto udplitev6_prot = {
--yoshfuji
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists