From def8b4faff5ca349beafbbfeb2c51f3602a6ef3a Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Tue, 28 Oct 2008 13:24:06 -0700 Subject: net: reduce structures when XFRM=n ifdef out * struct sk_buff::sp (pointer) * struct dst_entry::xfrm (pointer) * struct sock::sk_policy (2 pointers) Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index ada50c04d09f..d6b750a25078 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -229,7 +229,9 @@ struct sock { } sk_backlog; wait_queue_head_t *sk_sleep; struct dst_entry *sk_dst_cache; +#ifdef CONFIG_XFRM struct xfrm_policy *sk_policy[2]; +#endif rwlock_t sk_dst_lock; atomic_t sk_rmem_alloc; atomic_t sk_wmem_alloc; -- cgit v1.2.2 From 645ca708f936b2fbeb79e52d7823e3eb2c0905f8 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 01:41:45 -0700 Subject: udp: introduce struct udp_table and multiple spinlocks UDP sockets are hashed in a 128 slots hash table. This hash table is protected by *one* rwlock. This rwlock is readlocked each time an incoming UDP message is handled. This rwlock is writelocked each time a socket must be inserted in hash table (bind time), or deleted from this table (close time) This is not scalable on SMP machines : 1) Even in read mode, lock() and unlock() are atomic operations and must dirty a contended cache line, shared by all cpus. 2) A writer might be starved if many readers are 'in flight'. This can happen on a machine with some NIC receiving many UDP messages. User process can be delayed a long time at socket creation/dismantle time. This patch prepares RCU migration, by introducing 'struct udp_table and struct udp_hslot', and using one spinlock per chain, to reduce contention on central rwlock. Introducing one spinlock per chain reduces latencies, for port randomization on heavily loaded UDP servers. This also speedup bindings to specific ports. udp_lib_unhash() was uninlined, becoming to big. Some cleanups were done to ease review of following patch (RCUification of UDP Unicast lookups) Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index d6b750a25078..d200dfbe1ef6 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -599,7 +599,7 @@ struct proto { union { struct inet_hashinfo *hashinfo; - struct hlist_head *udp_hash; + struct udp_table *udp_table; struct raw_hashinfo *raw_hash; } h; -- cgit v1.2.2 From 271b72c7fa82c2c7a795bc16896149933110672d Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 02:11:14 -0700 Subject: udp: RCU handling for Unicast packets. Goals are : 1) Optimizing handling of incoming Unicast UDP frames, so that no memory writes should happen in the fast path. Note: Multicasts and broadcasts still will need to take a lock, because doing a full lockless lookup in this case is difficult. 2) No expensive operations in the socket bind/unhash phases : - No expensive synchronize_rcu() calls. - No added rcu_head in socket structure, increasing memory needs, but more important, forcing us to use call_rcu() calls, that have the bad property of making sockets structure cold. (rcu grace period between socket freeing and its potential reuse make this socket being cold in CPU cache). David did a previous patch using call_rcu() and noticed a 20% impact on TCP connection rates. Quoting Cristopher Lameter : "Right. That results in cacheline cooldown. You'd want to recycle the object as they are cache hot on a per cpu basis. That is screwed up by the delayed regular rcu processing. We have seen multiple regressions due to cacheline cooldown. The only choice in cacheline hot sensitive areas is to deal with the complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU." - Because udp sockets are allocated from dedicated kmem_cache, use of SLAB_DESTROY_BY_RCU can help here. Theory of operation : --------------------- As the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()), special attention must be taken by readers and writers. Use of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed, reused, inserted in a different chain or in worst case in the same chain while readers could do lookups in the same time. In order to avoid loops, a reader must check each socket found in a chain really belongs to the chain the reader was traversing. If it finds a mismatch, lookup must start again at the begining. This *restart* loop is the reason we had to use rdlock for the multicast case, because we dont want to send same message several times to the same socket. We use RCU only for fast path. Thus, /proc/net/udp still takes spinlocks. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 37 ++++++++++++++++++++++++++++++++++++- 1 file changed, 36 insertions(+), 1 deletion(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index d200dfbe1ef6..0bea25db5471 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -363,6 +363,27 @@ static __inline__ int sk_del_node_init(struct sock *sk) return rc; } +static __inline__ int __sk_del_node_init_rcu(struct sock *sk) +{ + if (sk_hashed(sk)) { + hlist_del_init_rcu(&sk->sk_node); + return 1; + } + return 0; +} + +static __inline__ int sk_del_node_init_rcu(struct sock *sk) +{ + int rc = __sk_del_node_init_rcu(sk); + + if (rc) { + /* paranoid for a while -acme */ + WARN_ON(atomic_read(&sk->sk_refcnt) == 1); + __sock_put(sk); + } + return rc; +} + static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) { hlist_add_head(&sk->sk_node, list); @@ -374,6 +395,17 @@ static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) __sk_add_node(sk, list); } +static __inline__ void __sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +{ + hlist_add_head_rcu(&sk->sk_node, list); +} + +static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +{ + sock_hold(sk); + __sk_add_node_rcu(sk, list); +} + static __inline__ void __sk_del_bind_node(struct sock *sk) { __hlist_del(&sk->sk_bind_node); @@ -387,6 +419,8 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) +#define sk_for_each_rcu(__sk, node, list) \ + hlist_for_each_entry_rcu(__sk, node, list, sk_node) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) @@ -589,8 +623,9 @@ struct proto { int *sysctl_rmem; int max_header; - struct kmem_cache *slab; + struct kmem_cache *slab; unsigned int obj_size; + int slab_flags; atomic_t *orphan_count; -- cgit v1.2.2 From 96631ed16c514cf8b28fab991a076985ce378c26 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 29 Oct 2008 11:19:58 -0700 Subject: udp: introduce sk_for_each_rcu_safenext() Corey Minyard found a race added in commit 271b72c7fa82c2c7a795bc16896149933110672d (udp: RCU handling for Unicast packets.) "If the socket is moved from one list to another list in-between the time the hash is calculated and the next field is accessed, and the socket has moved to the end of the new list, the traversal will not complete properly on the list it should have, since the socket will be on the end of the new list and there's not a way to tell it's on a new list and restart the list traversal. I think that this can be solved by pre-fetching the "next" field (with proper barriers) before checking the hash." This patch corrects this problem, introducing a new sk_for_each_rcu_safenext() macro. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 0bea25db5471..a4f6d3fc0470 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -419,8 +419,8 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) -#define sk_for_each_rcu(__sk, node, list) \ - hlist_for_each_entry_rcu(__sk, node, list, sk_node) +#define sk_for_each_rcu_safenext(__sk, node, list, next) \ + hlist_for_each_entry_rcu_safenext(__sk, node, list, sk_node, next) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) -- cgit v1.2.2 From d5f642384e9da75393160350f75bbb9a527f7c58 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Tue, 4 Nov 2008 14:45:58 -0800 Subject: net: #ifdef ->sk_security Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 941ad7c830a3..08291c1be41e 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -271,7 +271,9 @@ struct sock { struct sk_buff *sk_send_head; __u32 sk_sndmsg_off; int sk_write_pending; +#ifdef CONFIG_SECURITY void *sk_security; +#endif __u32 sk_mark; /* XXX 4 bytes hole on 64 bit */ void (*sk_state_change)(struct sock *sk); -- cgit v1.2.2 From 2378982487c492541d17adc0a870e7e83b07ba43 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Wed, 12 Nov 2008 23:25:32 -0800 Subject: net: ifdef struct sock::sk_async_wait_queue Every user is under CONFIG_NET_DMA already, so ifdef field as well. Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/sock.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 08291c1be41e..8b2b82131b67 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -239,7 +239,9 @@ struct sock { int sk_sndbuf; struct sk_buff_head sk_receive_queue; struct sk_buff_head sk_write_queue; +#ifdef CONFIG_NET_DMA struct sk_buff_head sk_async_wait_queue; +#endif int sk_wmem_queued; int sk_forward_alloc; gfp_t sk_allocation; -- cgit v1.2.2 From 88ab1932eac721c6e7336708558fa5ed02c85c80 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 16 Nov 2008 19:39:21 -0800 Subject: udp: Use hlist_nulls in UDP RCU code This is a straightforward patch, using hlist_nulls infrastructure. RCUification already done on UDP two weeks ago. Using hlist_nulls permits us to avoid some memory barriers, both at lookup time and delete time. Patch is large because it adds new macros to include/net/sock.h. These macros will be used by TCP & DCCP in next patch. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 57 +++++++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 46 insertions(+), 11 deletions(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 8b2b82131b67..0a638948868d 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -42,6 +42,7 @@ #include #include +#include #include #include #include @@ -52,6 +53,7 @@ #include #include +#include #include #include @@ -106,6 +108,7 @@ struct net; * @skc_reuse: %SO_REUSEADDR setting * @skc_bound_dev_if: bound device index if != 0 * @skc_node: main hash linkage for various protocol lookup tables + * @skc_nulls_node: main hash linkage for UDP/UDP-Lite protocol * @skc_bind_node: bind hash linkage for various protocol lookup tables * @skc_refcnt: reference count * @skc_hash: hash value used with various protocol lookup tables @@ -120,7 +123,10 @@ struct sock_common { volatile unsigned char skc_state; unsigned char skc_reuse; int skc_bound_dev_if; - struct hlist_node skc_node; + union { + struct hlist_node skc_node; + struct hlist_nulls_node skc_nulls_node; + }; struct hlist_node skc_bind_node; atomic_t skc_refcnt; unsigned int skc_hash; @@ -206,6 +212,7 @@ struct sock { #define sk_reuse __sk_common.skc_reuse #define sk_bound_dev_if __sk_common.skc_bound_dev_if #define sk_node __sk_common.skc_node +#define sk_nulls_node __sk_common.skc_nulls_node #define sk_bind_node __sk_common.skc_bind_node #define sk_refcnt __sk_common.skc_refcnt #define sk_hash __sk_common.skc_hash @@ -300,12 +307,30 @@ static inline struct sock *sk_head(const struct hlist_head *head) return hlist_empty(head) ? NULL : __sk_head(head); } +static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) +{ + return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); +} + +static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) +{ + return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); +} + static inline struct sock *sk_next(const struct sock *sk) { return sk->sk_node.next ? hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; } +static inline struct sock *sk_nulls_next(const struct sock *sk) +{ + return (!is_a_nulls(sk->sk_nulls_node.next)) ? + hlist_nulls_entry(sk->sk_nulls_node.next, + struct sock, sk_nulls_node) : + NULL; +} + static inline int sk_unhashed(const struct sock *sk) { return hlist_unhashed(&sk->sk_node); @@ -321,6 +346,11 @@ static __inline__ void sk_node_init(struct hlist_node *node) node->pprev = NULL; } +static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) +{ + node->pprev = NULL; +} + static __inline__ void __sk_del_node(struct sock *sk) { __hlist_del(&sk->sk_node); @@ -367,18 +397,18 @@ static __inline__ int sk_del_node_init(struct sock *sk) return rc; } -static __inline__ int __sk_del_node_init_rcu(struct sock *sk) +static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) { if (sk_hashed(sk)) { - hlist_del_init_rcu(&sk->sk_node); + hlist_nulls_del_init_rcu(&sk->sk_nulls_node); return 1; } return 0; } -static __inline__ int sk_del_node_init_rcu(struct sock *sk) +static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) { - int rc = __sk_del_node_init_rcu(sk); + int rc = __sk_nulls_del_node_init_rcu(sk); if (rc) { /* paranoid for a while -acme */ @@ -399,15 +429,15 @@ static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) __sk_add_node(sk, list); } -static __inline__ void __sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) { - hlist_add_head_rcu(&sk->sk_node, list); + hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); } -static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) +static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) { sock_hold(sk); - __sk_add_node_rcu(sk, list); + __sk_nulls_add_node_rcu(sk, list); } static __inline__ void __sk_del_bind_node(struct sock *sk) @@ -423,11 +453,16 @@ static __inline__ void sk_add_bind_node(struct sock *sk, #define sk_for_each(__sk, node, list) \ hlist_for_each_entry(__sk, node, list, sk_node) -#define sk_for_each_rcu_safenext(__sk, node, list, next) \ - hlist_for_each_entry_rcu_safenext(__sk, node, list, sk_node, next) +#define sk_nulls_for_each(__sk, node, list) \ + hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) +#define sk_nulls_for_each_rcu(__sk, node, list) \ + hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) #define sk_for_each_from(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_from(__sk, node, sk_node) +#define sk_nulls_for_each_from(__sk, node) \ + if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ + hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) #define sk_for_each_continue(__sk, node) \ if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ hlist_for_each_entry_continue(__sk, node, sk_node) -- cgit v1.2.2 From 1748376b6626acf59c24e9592ac67b3fe2a0e026 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 25 Nov 2008 21:16:35 -0800 Subject: net: Use a percpu_counter for sockets_allocated Instead of using one atomic_t per protocol, use a percpu_counter for "sockets_allocated", to reduce cache line contention on heavy duty network servers. Note : We revert commit (248969ae31e1b3276fc4399d67ce29a5d81e6fd9 net: af_unix can make unix_nr_socks visbile in /proc), since it is not anymore used after sock_prot_inuse_add() addition Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index 00cd486d362f..a2a3890959c4 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -649,7 +649,7 @@ struct proto { /* Memory pressure */ void (*enter_memory_pressure)(struct sock *sk); atomic_t *memory_allocated; /* Current allocated memory. */ - atomic_t *sockets_allocated; /* Current number of sockets. */ + struct percpu_counter *sockets_allocated; /* Current number of sockets. */ /* * Pressure flag: try to collapse. * Technical note: it is used by multiple contexts non atomically. -- cgit v1.2.2 From dd24c00191d5e4a1ae896aafe33c6b8095ab4bd1 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 25 Nov 2008 21:17:14 -0800 Subject: net: Use a percpu_counter for orphan_count Instead of using one atomic_t per protocol, use a percpu_counter for "orphan_count", to reduce cache line contention on heavy duty network servers. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/net/sock.h') diff --git a/include/net/sock.h b/include/net/sock.h index a2a3890959c4..5a3a151bd730 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -666,7 +666,7 @@ struct proto { unsigned int obj_size; int slab_flags; - atomic_t *orphan_count; + struct percpu_counter *orphan_count; struct request_sock_ops *rsk_prot; struct timewait_sock_ops *twsk_prot; -- cgit v1.2.2