Commit d6cc1d64 authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller

inetpeer: various changes

Try to reduce cache line contentions in peer management, to reduce IP
defragmentation overhead.

- peer_fake_node is marked 'const' to make sure its not modified.
  (tested with CONFIG_DEBUG_RODATA=y)

- Group variables in two structures to reduce number of dirtied cache
lines. One named "peers" for avl tree root, its number of entries, and
associated lock. (candidate for RCU conversion)

- A second one named "unused_peers" for unused list and its lock

- Add a !list_empty() test in unlink_from_unused() to avoid taking lock
when entry is not unused.

- Use atomic_dec_and_lock() in inet_putpeer() to avoid taking lock in
some cases.
Signed-off-by: default avatarEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 6b10de38
...@@ -70,17 +70,25 @@ ...@@ -70,17 +70,25 @@
static struct kmem_cache *peer_cachep __read_mostly; static struct kmem_cache *peer_cachep __read_mostly;
#define node_height(x) x->avl_height #define node_height(x) x->avl_height
static struct inet_peer peer_fake_node = {
.avl_left = &peer_fake_node, #define peer_avl_empty ((struct inet_peer *)&peer_fake_node)
.avl_right = &peer_fake_node, static const struct inet_peer peer_fake_node = {
.avl_left = peer_avl_empty,
.avl_right = peer_avl_empty,
.avl_height = 0 .avl_height = 0
}; };
#define peer_avl_empty (&peer_fake_node)
static struct inet_peer *peer_root = peer_avl_empty; static struct {
static DEFINE_RWLOCK(peer_pool_lock); struct inet_peer *root;
rwlock_t lock;
int total;
} peers = {
.root = peer_avl_empty,
.lock = __RW_LOCK_UNLOCKED(peers.lock),
.total = 0,
};
#define PEER_MAXDEPTH 40 /* sufficient for about 2^27 nodes */ #define PEER_MAXDEPTH 40 /* sufficient for about 2^27 nodes */
static int peer_total;
/* Exported for sysctl_net_ipv4. */ /* Exported for sysctl_net_ipv4. */
int inet_peer_threshold __read_mostly = 65536 + 128; /* start to throw entries more int inet_peer_threshold __read_mostly = 65536 + 128; /* start to throw entries more
* aggressively at this stage */ * aggressively at this stage */
...@@ -89,8 +97,13 @@ int inet_peer_maxttl __read_mostly = 10 * 60 * HZ; /* usual time to live: 10 min ...@@ -89,8 +97,13 @@ int inet_peer_maxttl __read_mostly = 10 * 60 * HZ; /* usual time to live: 10 min
int inet_peer_gc_mintime __read_mostly = 10 * HZ; int inet_peer_gc_mintime __read_mostly = 10 * HZ;
int inet_peer_gc_maxtime __read_mostly = 120 * HZ; int inet_peer_gc_maxtime __read_mostly = 120 * HZ;
static LIST_HEAD(unused_peers); static struct {
static DEFINE_SPINLOCK(inet_peer_unused_lock); struct list_head list;
spinlock_t lock;
} unused_peers = {
.list = LIST_HEAD_INIT(unused_peers.list),
.lock = __SPIN_LOCK_UNLOCKED(unused_peers.lock),
};
static void peer_check_expire(unsigned long dummy); static void peer_check_expire(unsigned long dummy);
static DEFINE_TIMER(peer_periodic_timer, peer_check_expire, 0, 0); static DEFINE_TIMER(peer_periodic_timer, peer_check_expire, 0, 0);
...@@ -131,9 +144,11 @@ void __init inet_initpeers(void) ...@@ -131,9 +144,11 @@ void __init inet_initpeers(void)
/* Called with or without local BH being disabled. */ /* Called with or without local BH being disabled. */
static void unlink_from_unused(struct inet_peer *p) static void unlink_from_unused(struct inet_peer *p)
{ {
spin_lock_bh(&inet_peer_unused_lock); if (!list_empty(&p->unused)) {
list_del_init(&p->unused); spin_lock_bh(&unused_peers.lock);
spin_unlock_bh(&inet_peer_unused_lock); list_del_init(&p->unused);
spin_unlock_bh(&unused_peers.lock);
}
} }
/* /*
...@@ -146,9 +161,9 @@ static void unlink_from_unused(struct inet_peer *p) ...@@ -146,9 +161,9 @@ static void unlink_from_unused(struct inet_peer *p)
struct inet_peer *u, **v; \ struct inet_peer *u, **v; \
if (_stack != NULL) { \ if (_stack != NULL) { \
stackptr = _stack; \ stackptr = _stack; \
*stackptr++ = &peer_root; \ *stackptr++ = &peers.root; \
} \ } \
for (u = peer_root; u != peer_avl_empty; ) { \ for (u = peers.root; u != peer_avl_empty; ) { \
if (_daddr == u->v4daddr) \ if (_daddr == u->v4daddr) \
break; \ break; \
if ((__force __u32)_daddr < (__force __u32)u->v4daddr) \ if ((__force __u32)_daddr < (__force __u32)u->v4daddr) \
...@@ -262,7 +277,7 @@ do { \ ...@@ -262,7 +277,7 @@ do { \
n->avl_right = peer_avl_empty; \ n->avl_right = peer_avl_empty; \
**--stackptr = n; \ **--stackptr = n; \
peer_avl_rebalance(stack, stackptr); \ peer_avl_rebalance(stack, stackptr); \
} while(0) } while (0)
/* May be called with local BH enabled. */ /* May be called with local BH enabled. */
static void unlink_from_pool(struct inet_peer *p) static void unlink_from_pool(struct inet_peer *p)
...@@ -271,7 +286,7 @@ static void unlink_from_pool(struct inet_peer *p) ...@@ -271,7 +286,7 @@ static void unlink_from_pool(struct inet_peer *p)
do_free = 0; do_free = 0;
write_lock_bh(&peer_pool_lock); write_lock_bh(&peers.lock);
/* Check the reference counter. It was artificially incremented by 1 /* Check the reference counter. It was artificially incremented by 1
* in cleanup() function to prevent sudden disappearing. If the * in cleanup() function to prevent sudden disappearing. If the
* reference count is still 1 then the node is referenced only as `p' * reference count is still 1 then the node is referenced only as `p'
...@@ -303,10 +318,10 @@ static void unlink_from_pool(struct inet_peer *p) ...@@ -303,10 +318,10 @@ static void unlink_from_pool(struct inet_peer *p)
delp[1] = &t->avl_left; /* was &p->avl_left */ delp[1] = &t->avl_left; /* was &p->avl_left */
} }
peer_avl_rebalance(stack, stackptr); peer_avl_rebalance(stack, stackptr);
peer_total--; peers.total--;
do_free = 1; do_free = 1;
} }
write_unlock_bh(&peer_pool_lock); write_unlock_bh(&peers.lock);
if (do_free) if (do_free)
kmem_cache_free(peer_cachep, p); kmem_cache_free(peer_cachep, p);
...@@ -326,16 +341,16 @@ static int cleanup_once(unsigned long ttl) ...@@ -326,16 +341,16 @@ static int cleanup_once(unsigned long ttl)
struct inet_peer *p = NULL; struct inet_peer *p = NULL;
/* Remove the first entry from the list of unused nodes. */ /* Remove the first entry from the list of unused nodes. */
spin_lock_bh(&inet_peer_unused_lock); spin_lock_bh(&unused_peers.lock);
if (!list_empty(&unused_peers)) { if (!list_empty(&unused_peers.list)) {
__u32 delta; __u32 delta;
p = list_first_entry(&unused_peers, struct inet_peer, unused); p = list_first_entry(&unused_peers.list, struct inet_peer, unused);
delta = (__u32)jiffies - p->dtime; delta = (__u32)jiffies - p->dtime;
if (delta < ttl) { if (delta < ttl) {
/* Do not prune fresh entries. */ /* Do not prune fresh entries. */
spin_unlock_bh(&inet_peer_unused_lock); spin_unlock_bh(&unused_peers.lock);
return -1; return -1;
} }
...@@ -345,7 +360,7 @@ static int cleanup_once(unsigned long ttl) ...@@ -345,7 +360,7 @@ static int cleanup_once(unsigned long ttl)
* before unlink_from_pool() call. */ * before unlink_from_pool() call. */
atomic_inc(&p->refcnt); atomic_inc(&p->refcnt);
} }
spin_unlock_bh(&inet_peer_unused_lock); spin_unlock_bh(&unused_peers.lock);
if (p == NULL) if (p == NULL)
/* It means that the total number of USED entries has /* It means that the total number of USED entries has
...@@ -364,11 +379,11 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create) ...@@ -364,11 +379,11 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create)
struct inet_peer **stack[PEER_MAXDEPTH], ***stackptr; struct inet_peer **stack[PEER_MAXDEPTH], ***stackptr;
/* Look up for the address quickly. */ /* Look up for the address quickly. */
read_lock_bh(&peer_pool_lock); read_lock_bh(&peers.lock);
p = lookup(daddr, NULL); p = lookup(daddr, NULL);
if (p != peer_avl_empty) if (p != peer_avl_empty)
atomic_inc(&p->refcnt); atomic_inc(&p->refcnt);
read_unlock_bh(&peer_pool_lock); read_unlock_bh(&peers.lock);
if (p != peer_avl_empty) { if (p != peer_avl_empty) {
/* The existing node has been found. */ /* The existing node has been found. */
...@@ -390,7 +405,7 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create) ...@@ -390,7 +405,7 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create)
atomic_set(&n->ip_id_count, secure_ip_id(daddr)); atomic_set(&n->ip_id_count, secure_ip_id(daddr));
n->tcp_ts_stamp = 0; n->tcp_ts_stamp = 0;
write_lock_bh(&peer_pool_lock); write_lock_bh(&peers.lock);
/* Check if an entry has suddenly appeared. */ /* Check if an entry has suddenly appeared. */
p = lookup(daddr, stack); p = lookup(daddr, stack);
if (p != peer_avl_empty) if (p != peer_avl_empty)
...@@ -399,10 +414,10 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create) ...@@ -399,10 +414,10 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create)
/* Link the node. */ /* Link the node. */
link_to_pool(n); link_to_pool(n);
INIT_LIST_HEAD(&n->unused); INIT_LIST_HEAD(&n->unused);
peer_total++; peers.total++;
write_unlock_bh(&peer_pool_lock); write_unlock_bh(&peers.lock);
if (peer_total >= inet_peer_threshold) if (peers.total >= inet_peer_threshold)
/* Remove one less-recently-used entry. */ /* Remove one less-recently-used entry. */
cleanup_once(0); cleanup_once(0);
...@@ -411,7 +426,7 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create) ...@@ -411,7 +426,7 @@ struct inet_peer *inet_getpeer(__be32 daddr, int create)
out_free: out_free:
/* The appropriate node is already in the pool. */ /* The appropriate node is already in the pool. */
atomic_inc(&p->refcnt); atomic_inc(&p->refcnt);
write_unlock_bh(&peer_pool_lock); write_unlock_bh(&peers.lock);
/* Remove the entry from unused list if it was there. */ /* Remove the entry from unused list if it was there. */
unlink_from_unused(p); unlink_from_unused(p);
/* Free preallocated the preallocated node. */ /* Free preallocated the preallocated node. */
...@@ -425,12 +440,12 @@ static void peer_check_expire(unsigned long dummy) ...@@ -425,12 +440,12 @@ static void peer_check_expire(unsigned long dummy)
unsigned long now = jiffies; unsigned long now = jiffies;
int ttl; int ttl;
if (peer_total >= inet_peer_threshold) if (peers.total >= inet_peer_threshold)
ttl = inet_peer_minttl; ttl = inet_peer_minttl;
else else
ttl = inet_peer_maxttl ttl = inet_peer_maxttl
- (inet_peer_maxttl - inet_peer_minttl) / HZ * - (inet_peer_maxttl - inet_peer_minttl) / HZ *
peer_total / inet_peer_threshold * HZ; peers.total / inet_peer_threshold * HZ;
while (!cleanup_once(ttl)) { while (!cleanup_once(ttl)) {
if (jiffies != now) if (jiffies != now)
break; break;
...@@ -439,22 +454,25 @@ static void peer_check_expire(unsigned long dummy) ...@@ -439,22 +454,25 @@ static void peer_check_expire(unsigned long dummy)
/* Trigger the timer after inet_peer_gc_mintime .. inet_peer_gc_maxtime /* Trigger the timer after inet_peer_gc_mintime .. inet_peer_gc_maxtime
* interval depending on the total number of entries (more entries, * interval depending on the total number of entries (more entries,
* less interval). */ * less interval). */
if (peer_total >= inet_peer_threshold) if (peers.total >= inet_peer_threshold)
peer_periodic_timer.expires = jiffies + inet_peer_gc_mintime; peer_periodic_timer.expires = jiffies + inet_peer_gc_mintime;
else else
peer_periodic_timer.expires = jiffies peer_periodic_timer.expires = jiffies
+ inet_peer_gc_maxtime + inet_peer_gc_maxtime
- (inet_peer_gc_maxtime - inet_peer_gc_mintime) / HZ * - (inet_peer_gc_maxtime - inet_peer_gc_mintime) / HZ *
peer_total / inet_peer_threshold * HZ; peers.total / inet_peer_threshold * HZ;
add_timer(&peer_periodic_timer); add_timer(&peer_periodic_timer);
} }
void inet_putpeer(struct inet_peer *p) void inet_putpeer(struct inet_peer *p)
{ {
spin_lock_bh(&inet_peer_unused_lock); local_bh_disable();
if (atomic_dec_and_test(&p->refcnt)) {
list_add_tail(&p->unused, &unused_peers); if (atomic_dec_and_lock(&p->refcnt, &unused_peers.lock)) {
list_add_tail(&p->unused, &unused_peers.list);
p->dtime = (__u32)jiffies; p->dtime = (__u32)jiffies;
spin_unlock(&unused_peers.lock);
} }
spin_unlock_bh(&inet_peer_unused_lock);
local_bh_enable();
} }
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment