mirrors/linux-stable
1
0
Fork 0
mirror of https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git synced 2024-09-29 13:53:33 +00:00
Code Issues Releases Wiki Activity
linux-stable/include/net/route.h
Wangyang Guo d288a162dd net: dst: Prevent false sharing vs. dst_entry:: __refcnt 
dst_entry::__refcnt is highly contended in scenarios where many connections
happen from and to the same IP. The reference count is an atomic_t, so the
reference count operations have to take the cache-line exclusive.

Aside of the unavoidable reference count contention there is another
significant problem which is caused by that: False sharing.

perf top identified two affected read accesses. dst_entry::lwtstate and
rtable::rt_genid.

dst_entry:__refcnt is located at offset 64 of dst_entry, which puts it into
a seperate cacheline vs. the read mostly members located at the beginning
of the struct.

That prevents false sharing vs. the struct members in the first 64
bytes of the structure, but there is also

  dst_entry::lwtstate

which is located after the reference count and in the same cache line. This
member is read after a reference count has been acquired.

struct rtable embeds a struct dst_entry at offset 0. struct dst_entry has a
size of 112 bytes, which means that the struct members of rtable which
follow the dst member share the same cache line as dst_entry::__refcnt.
Especially

  rtable::rt_genid

is also read by the contexts which have a reference count acquired
already.

When dst_entry:__refcnt is incremented or decremented via an atomic
operation these read accesses stall. This was found when analysing the
memtier benchmark in 1:100 mode, which amplifies the problem extremly.

Move the rt[6i]_uncached[_list] members out of struct rtable and struct
rt6_info into struct dst_entry to provide padding and move the lwtstate
member after that so it ends up in the same cache line.

The resulting improvement depends on the micro-architecture and the number
of CPUs. It ranges from +20% to +120% with a localhost memtier/memcached
benchmark.

[ tglx: Rearrange struct ]

Signed-off-by: Wangyang Guo <wangyang.guo@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20230323102800.042297517@linutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2023-03-28 18:52:22 -07:00

399 lines
11 KiB
C
Raw Blame History

/* SPDX-License-Identifier: GPL-2.0-or-later */
/*
* INET An implementation of the TCP/IP protocol suite for the LINUX
* operating system. INET is implemented using the BSD Socket
* interface as the means of communication with the user level.
*
* Definitions for the IP router.
*
* Version: @(#)route.h 1.0.4 05/27/93
*
* Authors: Ross Biro
* Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
* Fixes:
* Alan Cox : Reformatted. Added ip_rt_local()
* Alan Cox : Support for TCP parameters.
* Alexey Kuznetsov: Major changes for new routing code.
* Mike McLagan : Routing by source
* Robert Olsson : Added rt_cache statistics
*/
#ifndef _ROUTE_H
#define _ROUTE_H
#include <net/dst.h>
#include <net/inetpeer.h>
#include <net/flow.h>
#include <net/inet_sock.h>
#include <net/ip_fib.h>
#include <net/arp.h>
#include <net/ndisc.h>
#include <linux/in_route.h>
#include <linux/rtnetlink.h>
#include <linux/rcupdate.h>
#include <linux/route.h>
#include <linux/ip.h>
#include <linux/cache.h>
#include <linux/security.h>
#define RTO_ONLINK 0x01
#define RT_CONN_FLAGS(sk) (RT_TOS(inet_sk(sk)->tos) | sock_flag(sk, SOCK_LOCALROUTE))
#define RT_CONN_FLAGS_TOS(sk,tos) (RT_TOS(tos) | sock_flag(sk, SOCK_LOCALROUTE))
static inline __u8 ip_sock_rt_scope(const struct sock *sk)
{
if (sock_flag(sk, SOCK_LOCALROUTE))
return RT_SCOPE_LINK;
return RT_SCOPE_UNIVERSE;
}
static inline __u8 ip_sock_rt_tos(const struct sock *sk)
{
return RT_TOS(inet_sk(sk)->tos);
}
struct ip_tunnel_info;
struct fib_nh;
struct fib_info;
struct uncached_list;
struct rtable {
struct dst_entry dst;
int rt_genid;
unsigned int rt_flags;
__u16 rt_type;
__u8 rt_is_input;
__u8 rt_uses_gateway;
int rt_iif;
u8 rt_gw_family;
/* Info on neighbour */
union {
__be32 rt_gw4;
struct in6_addr rt_gw6;
};
/* Miscellaneous cached information */
u32 rt_mtu_locked:1,
rt_pmtu:31;
};
static inline bool rt_is_input_route(const struct rtable *rt)
{
return rt->rt_is_input != 0;
}
static inline bool rt_is_output_route(const struct rtable *rt)
{
return rt->rt_is_input == 0;
}
static inline __be32 rt_nexthop(const struct rtable *rt, __be32 daddr)
{
if (rt->rt_gw_family == AF_INET)
return rt->rt_gw4;
return daddr;
}
struct ip_rt_acct {
__u32 o_bytes;
__u32 o_packets;
__u32 i_bytes;
__u32 i_packets;
};
struct rt_cache_stat {
unsigned int in_slow_tot;
unsigned int in_slow_mc;
unsigned int in_no_route;
unsigned int in_brd;
unsigned int in_martian_dst;
unsigned int in_martian_src;
unsigned int out_slow_tot;
unsigned int out_slow_mc;
};
extern struct ip_rt_acct __percpu *ip_rt_acct;
struct in_device;
int ip_rt_init(void);
void rt_cache_flush(struct net *net);
void rt_flush_dev(struct net_device *dev);
struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *flp,
const struct sk_buff *skb);
struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *flp,
struct fib_result *res,
const struct sk_buff *skb);
static inline struct rtable *__ip_route_output_key(struct net *net,
struct flowi4 *flp)
{
return ip_route_output_key_hash(net, flp, NULL);
}
struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
const struct sock *sk);
struct rtable *ip_route_output_tunnel(struct sk_buff *skb,
struct net_device *dev,
struct net *net, __be32 *saddr,
const struct ip_tunnel_info *info,
u8 protocol, bool use_cache);
struct dst_entry *ipv4_blackhole_route(struct net *net,
struct dst_entry *dst_orig);
static inline struct rtable *ip_route_output_key(struct net *net, struct flowi4 *flp)
{
return ip_route_output_flow(net, flp, NULL);
}
static inline struct rtable *ip_route_output(struct net *net, __be32 daddr,
__be32 saddr, u8 tos, int oif)
{
struct flowi4 fl4 = {
.flowi4_oif = oif,
.flowi4_tos = tos,
.daddr = daddr,
.saddr = saddr,
};
return ip_route_output_key(net, &fl4);
}
static inline struct rtable *ip_route_output_ports(struct net *net, struct flowi4 *fl4,
struct sock *sk,
__be32 daddr, __be32 saddr,
__be16 dport, __be16 sport,
__u8 proto, __u8 tos, int oif)
{
flowi4_init_output(fl4, oif, sk ? sk->sk_mark : 0, tos,
RT_SCOPE_UNIVERSE, proto,
sk ? inet_sk_flowi_flags(sk) : 0,
daddr, saddr, dport, sport, sock_net_uid(net, sk));
if (sk)
security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
return ip_route_output_flow(net, fl4, sk);
}
static inline struct rtable *ip_route_output_gre(struct net *net, struct flowi4 *fl4,
__be32 daddr, __be32 saddr,
__be32 gre_key, __u8 tos, int oif)
{
memset(fl4, 0, sizeof(*fl4));
fl4->flowi4_oif = oif;
fl4->daddr = daddr;
fl4->saddr = saddr;
fl4->flowi4_tos = tos;
fl4->flowi4_proto = IPPROTO_GRE;
fl4->fl4_gre_key = gre_key;
return ip_route_output_key(net, fl4);
}
int ip_mc_validate_source(struct sk_buff *skb, __be32 daddr, __be32 saddr,
u8 tos, struct net_device *dev,
struct in_device *in_dev, u32 *itag);
int ip_route_input_noref(struct sk_buff *skb, __be32 dst, __be32 src,
u8 tos, struct net_device *devin);
int ip_route_use_hint(struct sk_buff *skb, __be32 dst, __be32 src,
u8 tos, struct net_device *devin,
const struct sk_buff *hint);
static inline int ip_route_input(struct sk_buff *skb, __be32 dst, __be32 src,
u8 tos, struct net_device *devin)
{
int err;
rcu_read_lock();
err = ip_route_input_noref(skb, dst, src, tos, devin);
if (!err) {
skb_dst_force(skb);
if (!skb_dst(skb))
err = -EINVAL;
}
rcu_read_unlock();
return err;
}
void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, int oif,
u8 protocol);
void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu);
void ipv4_redirect(struct sk_buff *skb, struct net *net, int oif, u8 protocol);
void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk);
void ip_rt_send_redirect(struct sk_buff *skb);
unsigned int inet_addr_type(struct net *net, __be32 addr);
unsigned int inet_addr_type_table(struct net *net, __be32 addr, u32 tb_id);
unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
__be32 addr);
unsigned int inet_addr_type_dev_table(struct net *net,
const struct net_device *dev,
__be32 addr);
void ip_rt_multicast_event(struct in_device *);
int ip_rt_ioctl(struct net *, unsigned int cmd, struct rtentry *rt);
void ip_rt_get_source(u8 *src, struct sk_buff *skb, struct rtable *rt);
struct rtable *rt_dst_alloc(struct net_device *dev,
unsigned int flags, u16 type, bool noxfrm);
struct rtable *rt_dst_clone(struct net_device *dev, struct rtable *rt);
struct in_ifaddr;
void fib_add_ifaddr(struct in_ifaddr *);
void fib_del_ifaddr(struct in_ifaddr *, struct in_ifaddr *);
void fib_modify_prefix_metric(struct in_ifaddr *ifa, u32 new_metric);
void rt_add_uncached_list(struct rtable *rt);
void rt_del_uncached_list(struct rtable *rt);
int fib_dump_info_fnhe(struct sk_buff *skb, struct netlink_callback *cb,
u32 table_id, struct fib_info *fi,
int *fa_index, int fa_start, unsigned int flags);
static inline void ip_rt_put(struct rtable *rt)
{
/* dst_release() accepts a NULL parameter.
* We rely on dst being first structure in struct rtable
*/
BUILD_BUG_ON(offsetof(struct rtable, dst) != 0);
dst_release(&rt->dst);
}
#define IPTOS_RT_MASK (IPTOS_TOS_MASK & ~3)
extern const __u8 ip_tos2prio[16];
static inline char rt_tos2priority(u8 tos)
{
return ip_tos2prio[IPTOS_TOS(tos)>>1];
}
/* ip_route_connect() and ip_route_newports() work in tandem whilst
* binding a socket for a new outgoing connection.
*
* In order to use IPSEC properly, we must, in the end, have a
* route that was looked up using all available keys including source
* and destination ports.
*
* However, if a source port needs to be allocated (the user specified
* a wildcard source port) we need to obtain addressing information
* in order to perform that allocation.
*
* So ip_route_connect() looks up a route using wildcarded source and
* destination ports in the key, simply so that we can get a pair of
* addresses to use for port allocation.
*
* Later, once the ports are allocated, ip_route_newports() will make
* another route lookup if needed to make sure we catch any IPSEC
* rules keyed on the port information.
*
* The callers allocate the flow key on their stack, and must pass in
* the same flowi4 object to both the ip_route_connect() and the
* ip_route_newports() calls.
*/
static inline void ip_route_connect_init(struct flowi4 *fl4, __be32 dst,
__be32 src, int oif, u8 protocol,
__be16 sport, __be16 dport,
const struct sock *sk)
{
__u8 flow_flags = 0;
if (inet_sk(sk)->transparent)
flow_flags |= FLOWI_FLAG_ANYSRC;
flowi4_init_output(fl4, oif, sk->sk_mark, ip_sock_rt_tos(sk),
ip_sock_rt_scope(sk), protocol, flow_flags, dst,
src, dport, sport, sk->sk_uid);
}
static inline struct rtable *ip_route_connect(struct flowi4 *fl4, __be32 dst,
__be32 src, int oif, u8 protocol,
__be16 sport, __be16 dport,
struct sock *sk)
{
struct net *net = sock_net(sk);
struct rtable *rt;
ip_route_connect_init(fl4, dst, src, oif, protocol, sport, dport, sk);
if (!dst || !src) {
rt = __ip_route_output_key(net, fl4);
if (IS_ERR(rt))
return rt;
ip_rt_put(rt);
flowi4_update_output(fl4, oif, fl4->flowi4_tos, fl4->daddr,
fl4->saddr);
}
security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
return ip_route_output_flow(net, fl4, sk);
}
static inline struct rtable *ip_route_newports(struct flowi4 *fl4, struct rtable *rt,
__be16 orig_sport, __be16 orig_dport,
__be16 sport, __be16 dport,
struct sock *sk)
{
if (sport != orig_sport || dport != orig_dport) {
fl4->fl4_dport = dport;
fl4->fl4_sport = sport;
ip_rt_put(rt);
flowi4_update_output(fl4, sk->sk_bound_dev_if,
RT_CONN_FLAGS(sk), fl4->daddr,
fl4->saddr);
security_sk_classify_flow(sk, flowi4_to_flowi_common(fl4));
return ip_route_output_flow(sock_net(sk), fl4, sk);
}
return rt;
}
static inline int inet_iif(const struct sk_buff *skb)
{
struct rtable *rt = skb_rtable(skb);
if (rt && rt->rt_iif)
return rt->rt_iif;
return skb->skb_iif;
}
static inline int ip4_dst_hoplimit(const struct dst_entry *dst)
{
int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
struct net *net = dev_net(dst->dev);
if (hoplimit == 0)
hoplimit = READ_ONCE(net->ipv4.sysctl_ip_default_ttl);
return hoplimit;
}
static inline struct neighbour *ip_neigh_gw4(struct net_device *dev,
__be32 daddr)
{
struct neighbour *neigh;
neigh = __ipv4_neigh_lookup_noref(dev, (__force u32)daddr);
if (unlikely(!neigh))
neigh = __neigh_create(&arp_tbl, &daddr, dev, false);
return neigh;
}
static inline struct neighbour *ip_neigh_for_gw(struct rtable *rt,
struct sk_buff *skb,
bool *is_v6gw)
{
struct net_device *dev = rt->dst.dev;
struct neighbour *neigh;
if (likely(rt->rt_gw_family == AF_INET)) {
neigh = ip_neigh_gw4(dev, rt->rt_gw4);
} else if (rt->rt_gw_family == AF_INET6) {
neigh = ip_neigh_gw6(dev, &rt->rt_gw6);
*is_v6gw = true;
} else {
neigh = ip_neigh_gw4(dev, ip_hdr(skb)->daddr);
}
return neigh;
}
#endif /* _ROUTE_H */
Reference in a new issue View git blame Copy permalink