2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* INETPEER - A storage for permanent information about peers
|
|
|
|
*
|
|
|
|
* Authors: Andrey V. Savochkin <saw@msu.ru>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#ifndef _NET_INETPEER_H
|
|
|
|
#define _NET_INETPEER_H
|
|
|
|
|
|
|
|
#include <linux/types.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/jiffies.h>
|
|
|
|
#include <linux/spinlock.h>
|
2011-01-27 12:55:53 +08:00
|
|
|
#include <linux/rtnetlink.h>
|
2010-12-01 04:20:00 +08:00
|
|
|
#include <net/ipv6.h>
|
2011-07-27 07:09:06 +08:00
|
|
|
#include <linux/atomic.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2011-02-10 06:30:26 +08:00
|
|
|
struct inetpeer_addr_base {
|
2010-12-01 03:53:55 +08:00
|
|
|
union {
|
2011-02-10 06:30:26 +08:00
|
|
|
__be32 a4;
|
|
|
|
__be32 a6[4];
|
2010-12-01 03:53:55 +08:00
|
|
|
};
|
2011-02-10 06:30:26 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
struct inetpeer_addr {
|
|
|
|
struct inetpeer_addr_base addr;
|
|
|
|
__u16 family;
|
2010-12-02 09:28:18 +08:00
|
|
|
};
|
2010-12-01 03:53:55 +08:00
|
|
|
|
2009-11-03 11:26:03 +08:00
|
|
|
struct inet_peer {
|
2006-10-20 15:28:35 +08:00
|
|
|
/* group together avl_left,avl_right,v4daddr to speedup lookups */
|
2010-10-26 07:55:38 +08:00
|
|
|
struct inet_peer __rcu *avl_left, *avl_right;
|
2010-12-02 09:28:18 +08:00
|
|
|
struct inetpeer_addr daddr;
|
2009-11-12 17:33:09 +08:00
|
|
|
__u32 avl_height;
|
2011-06-09 14:31:27 +08:00
|
|
|
|
|
|
|
u32 metrics[RTAX_MAX];
|
|
|
|
u32 rate_tokens; /* rate limiting for ICMP */
|
|
|
|
unsigned long rate_last;
|
2012-06-05 11:00:18 +08:00
|
|
|
union {
|
|
|
|
struct list_head gc_list;
|
|
|
|
struct rcu_head gc_rcu;
|
|
|
|
};
|
2010-06-16 12:52:13 +08:00
|
|
|
/*
|
inetpeer: get rid of ip_id_count
Ideally, we would need to generate IP ID using a per destination IP
generator.
linux kernels used inet_peer cache for this purpose, but this had a huge
cost on servers disabling MTU discovery.
1) each inet_peer struct consumes 192 bytes
2) inetpeer cache uses a binary tree of inet_peer structs,
with a nominal size of ~66000 elements under load.
3) lookups in this tree are hitting a lot of cache lines, as tree depth
is about 20.
4) If server deals with many tcp flows, we have a high probability of
not finding the inet_peer, allocating a fresh one, inserting it in
the tree with same initial ip_id_count, (cf secure_ip_id())
5) We garbage collect inet_peer aggressively.
IP ID generation do not have to be 'perfect'
Goal is trying to avoid duplicates in a short period of time,
so that reassembly units have a chance to complete reassembly of
fragments belonging to one message before receiving other fragments
with a recycled ID.
We simply use an array of generators, and a Jenkin hash using the dst IP
as a key.
ipv6_select_ident() is put back into net/ipv6/ip6_output.c where it
belongs (it is only used from this file)
secure_ip_id() and secure_ipv6_id() no longer are needed.
Rename ip_select_ident_more() to ip_select_ident_segs() to avoid
unnecessary decrement/increment of the number of segments.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-06-02 20:26:03 +08:00
|
|
|
* Once inet_peer is queued for deletion (refcnt == -1), following field
|
|
|
|
* is not available: rid
|
2011-01-27 12:55:53 +08:00
|
|
|
* We can share memory with rcu_head to help keep inet_peer small.
|
2010-06-16 12:52:13 +08:00
|
|
|
*/
|
|
|
|
union {
|
|
|
|
struct {
|
2011-02-10 07:36:47 +08:00
|
|
|
atomic_t rid; /* Frag reception counter */
|
2010-06-16 12:52:13 +08:00
|
|
|
};
|
|
|
|
struct rcu_head rcu;
|
2011-06-08 21:35:34 +08:00
|
|
|
struct inet_peer *gc_next;
|
2010-06-16 12:52:13 +08:00
|
|
|
};
|
2011-06-09 14:31:27 +08:00
|
|
|
|
|
|
|
/* following fields might be frequently dirtied */
|
|
|
|
__u32 dtime; /* the time of last use of not referenced entries */
|
|
|
|
atomic_t refcnt;
|
2005-04-17 06:20:36 +08:00
|
|
|
};
|
|
|
|
|
2012-06-10 07:27:05 +08:00
|
|
|
struct inet_peer_base {
|
|
|
|
struct inet_peer __rcu *root;
|
|
|
|
seqlock_t lock;
|
|
|
|
int total;
|
|
|
|
};
|
|
|
|
|
2012-06-10 13:36:36 +08:00
|
|
|
#define INETPEER_BASE_BIT 0x1UL
|
|
|
|
|
|
|
|
static inline struct inet_peer *inetpeer_ptr(unsigned long val)
|
|
|
|
{
|
|
|
|
BUG_ON(val & INETPEER_BASE_BIT);
|
|
|
|
return (struct inet_peer *) val;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct inet_peer_base *inetpeer_base_ptr(unsigned long val)
|
|
|
|
{
|
|
|
|
if (!(val & INETPEER_BASE_BIT))
|
|
|
|
return NULL;
|
|
|
|
val &= ~INETPEER_BASE_BIT;
|
|
|
|
return (struct inet_peer_base *) val;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool inetpeer_ptr_is_peer(unsigned long val)
|
|
|
|
{
|
|
|
|
return !(val & INETPEER_BASE_BIT);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void __inetpeer_ptr_set_peer(unsigned long *val, struct inet_peer *peer)
|
|
|
|
{
|
|
|
|
/* This implicitly clears INETPEER_BASE_BIT */
|
|
|
|
*val = (unsigned long) peer;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline bool inetpeer_ptr_set_peer(unsigned long *ptr, struct inet_peer *peer)
|
|
|
|
{
|
|
|
|
unsigned long val = (unsigned long) peer;
|
|
|
|
unsigned long orig = *ptr;
|
|
|
|
|
2012-06-11 19:13:57 +08:00
|
|
|
if (!(orig & INETPEER_BASE_BIT) ||
|
2012-06-10 13:36:36 +08:00
|
|
|
cmpxchg(ptr, orig, val) != orig)
|
|
|
|
return false;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void inetpeer_init_ptr(unsigned long *ptr, struct inet_peer_base *base)
|
|
|
|
{
|
|
|
|
*ptr = (unsigned long) base | INETPEER_BASE_BIT;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void inetpeer_transfer_peer(unsigned long *to, unsigned long *from)
|
|
|
|
{
|
|
|
|
unsigned long val = *from;
|
|
|
|
|
|
|
|
*to = val;
|
|
|
|
if (inetpeer_ptr_is_peer(val)) {
|
|
|
|
struct inet_peer *peer = inetpeer_ptr(val);
|
|
|
|
atomic_inc(&peer->refcnt);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-09-22 01:22:41 +08:00
|
|
|
void inet_peer_base_init(struct inet_peer_base *);
|
2012-06-10 07:27:05 +08:00
|
|
|
|
2013-09-22 01:22:41 +08:00
|
|
|
void inet_initpeers(void) __init;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2011-01-28 05:52:16 +08:00
|
|
|
#define INETPEER_METRICS_NEW (~(u32) 0)
|
|
|
|
|
|
|
|
static inline bool inet_metrics_new(const struct inet_peer *p)
|
|
|
|
{
|
|
|
|
return p->metrics[RTAX_LOCK-1] == INETPEER_METRICS_NEW;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* can be called with or without local BH being disabled */
|
2012-06-10 10:12:36 +08:00
|
|
|
struct inet_peer *inet_getpeer(struct inet_peer_base *base,
|
2012-06-08 09:20:41 +08:00
|
|
|
const struct inetpeer_addr *daddr,
|
|
|
|
int create);
|
2010-12-01 03:54:19 +08:00
|
|
|
|
2012-06-10 10:12:36 +08:00
|
|
|
static inline struct inet_peer *inet_getpeer_v4(struct inet_peer_base *base,
|
2012-06-08 09:21:40 +08:00
|
|
|
__be32 v4daddr,
|
|
|
|
int create)
|
2010-12-01 03:54:19 +08:00
|
|
|
{
|
2010-12-02 09:28:18 +08:00
|
|
|
struct inetpeer_addr daddr;
|
2010-12-01 03:54:19 +08:00
|
|
|
|
2011-02-10 06:30:26 +08:00
|
|
|
daddr.addr.a4 = v4daddr;
|
2010-12-01 03:54:19 +08:00
|
|
|
daddr.family = AF_INET;
|
2012-06-10 10:12:36 +08:00
|
|
|
return inet_getpeer(base, &daddr, create);
|
2010-12-01 03:54:19 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2012-06-10 10:12:36 +08:00
|
|
|
static inline struct inet_peer *inet_getpeer_v6(struct inet_peer_base *base,
|
2012-06-08 09:21:40 +08:00
|
|
|
const struct in6_addr *v6daddr,
|
|
|
|
int create)
|
2010-12-01 04:20:00 +08:00
|
|
|
{
|
2010-12-02 09:28:18 +08:00
|
|
|
struct inetpeer_addr daddr;
|
2010-12-01 04:20:00 +08:00
|
|
|
|
2011-11-21 11:39:03 +08:00
|
|
|
*(struct in6_addr *)daddr.addr.a6 = *v6daddr;
|
2010-12-01 04:20:00 +08:00
|
|
|
daddr.family = AF_INET6;
|
2012-06-10 10:12:36 +08:00
|
|
|
return inet_getpeer(base, &daddr, create);
|
2010-12-01 04:20:00 +08:00
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* can be called from BH context or outside */
|
2013-09-22 01:22:41 +08:00
|
|
|
void inet_putpeer(struct inet_peer *p);
|
|
|
|
bool inet_peer_xrlim_allow(struct inet_peer *peer, int timeout);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-09-22 01:22:41 +08:00
|
|
|
void inetpeer_invalidate_tree(struct inet_peer_base *);
|
2012-03-07 05:20:26 +08:00
|
|
|
|
2010-06-16 12:52:13 +08:00
|
|
|
/*
|
inetpeer: get rid of ip_id_count
Ideally, we would need to generate IP ID using a per destination IP
generator.
linux kernels used inet_peer cache for this purpose, but this had a huge
cost on servers disabling MTU discovery.
1) each inet_peer struct consumes 192 bytes
2) inetpeer cache uses a binary tree of inet_peer structs,
with a nominal size of ~66000 elements under load.
3) lookups in this tree are hitting a lot of cache lines, as tree depth
is about 20.
4) If server deals with many tcp flows, we have a high probability of
not finding the inet_peer, allocating a fresh one, inserting it in
the tree with same initial ip_id_count, (cf secure_ip_id())
5) We garbage collect inet_peer aggressively.
IP ID generation do not have to be 'perfect'
Goal is trying to avoid duplicates in a short period of time,
so that reassembly units have a chance to complete reassembly of
fragments belonging to one message before receiving other fragments
with a recycled ID.
We simply use an array of generators, and a Jenkin hash using the dst IP
as a key.
ipv6_select_ident() is put back into net/ipv6/ip6_output.c where it
belongs (it is only used from this file)
secure_ip_id() and secure_ipv6_id() no longer are needed.
Rename ip_select_ident_more() to ip_select_ident_segs() to avoid
unnecessary decrement/increment of the number of segments.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-06-02 20:26:03 +08:00
|
|
|
* temporary check to make sure we dont access rid, tcp_ts,
|
2010-06-16 12:52:13 +08:00
|
|
|
* tcp_ts_stamp if no refcount is taken on inet_peer
|
|
|
|
*/
|
|
|
|
static inline void inet_peer_refcheck(const struct inet_peer *p)
|
|
|
|
{
|
|
|
|
WARN_ON_ONCE(atomic_read(&p->refcnt) <= 0);
|
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif /* _NET_INETPEER_H */
|