From 1a24e04e4b50939daa3041682b38b82c896ca438 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 15 May 2015 12:39:25 -0700 Subject: net: fix sk_mem_reclaim_partial() sk_mem_reclaim_partial() goal is to ensure each socket has one SK_MEM_QUANTUM forward allocation. This is needed both for performance and better handling of memory pressure situations in follow up patches. SK_MEM_QUANTUM is currently a page, but might be reduced to 4096 bytes as some arches have 64KB pages. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/sock.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/net/sock.h b/include/net/sock.h index d882f4c8e438..4581a60636f8 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1368,7 +1368,7 @@ static inline struct inode *SOCK_INODE(struct socket *socket) * Functions for memory accounting */ int __sk_mem_schedule(struct sock *sk, int size, int kind); -void __sk_mem_reclaim(struct sock *sk); +void __sk_mem_reclaim(struct sock *sk, int amount); #define SK_MEM_QUANTUM ((int)PAGE_SIZE) #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) @@ -1409,7 +1409,7 @@ static inline void sk_mem_reclaim(struct sock *sk) if (!sk_has_account(sk)) return; if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) - __sk_mem_reclaim(sk); + __sk_mem_reclaim(sk, sk->sk_forward_alloc); } static inline void sk_mem_reclaim_partial(struct sock *sk) @@ -1417,7 +1417,7 @@ static inline void sk_mem_reclaim_partial(struct sock *sk) if (!sk_has_account(sk)) return; if (sk->sk_forward_alloc > SK_MEM_QUANTUM) - __sk_mem_reclaim(sk); + __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1); } static inline void sk_mem_charge(struct sock *sk, int size) -- cgit v1.3-18-gd494 From a6c5ea4ccf0033591e6e476d7a273c0074c07aa7 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 15 May 2015 12:39:26 -0700 Subject: tcp: rename sk_forced_wmem_schedule() to sk_forced_mem_schedule() We plan to use sk_forced_wmem_schedule() in input path as well, so make it non static and rename it. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/tcp.h | 2 ++ net/ipv4/tcp_output.c | 6 ++++-- 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/net/tcp.h b/include/net/tcp.h index 7ace6acbf5fd..841691a296dc 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -311,6 +311,8 @@ static inline bool tcp_out_of_memory(struct sock *sk) return false; } +void sk_forced_mem_schedule(struct sock *sk, int size); + static inline bool tcp_too_many_orphans(struct sock *sk, int shift) { struct percpu_counter *ocp = sk->sk_prot->orphan_count; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 7386d32cd670..bac1a950d087 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -2816,8 +2816,10 @@ begin_fwd: * connection tear down and (memory) recovery. * Otherwise tcp_send_fin() could be tempted to either delay FIN * or even be forced to close flow without any FIN. + * In general, we want to allow one skb per socket to avoid hangs + * with edge trigger epoll() */ -static void sk_forced_wmem_schedule(struct sock *sk, int size) +void sk_forced_mem_schedule(struct sock *sk, int size) { int amt, status; @@ -2864,7 +2866,7 @@ coalesce: return; } skb_reserve(skb, MAX_TCP_HEADER); - sk_forced_wmem_schedule(sk, skb->truesize); + sk_forced_mem_schedule(sk, skb->truesize); /* FIN eats a sequence byte, write_seq advanced by tcp_queue_skb(). */ tcp_init_nondata_skb(skb, tp->write_seq, TCPHDR_ACK | TCPHDR_FIN); -- cgit v1.3-18-gd494 From b8da51ebb1aa93908350f95efae73aecbc2e266c Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 15 May 2015 12:39:27 -0700 Subject: tcp: introduce tcp_under_memory_pressure() Introduce an optimized version of sk_under_memory_pressure() for TCP. Our intent is to use it in fast paths. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/tcp.h | 8 ++++++++ net/ipv4/tcp_input.c | 8 ++++---- net/ipv4/tcp_output.c | 4 ++-- net/ipv4/tcp_timer.c | 2 +- 4 files changed, 15 insertions(+), 7 deletions(-) (limited to 'include') diff --git a/include/net/tcp.h b/include/net/tcp.h index 841691a296dc..0d85223efa4c 100644 --- a/include/net/tcp.h +++ b/include/net/tcp.h @@ -286,6 +286,14 @@ extern atomic_long_t tcp_memory_allocated; extern struct percpu_counter tcp_sockets_allocated; extern int tcp_memory_pressure; +/* optimized version of sk_under_memory_pressure() for TCP sockets */ +static inline bool tcp_under_memory_pressure(const struct sock *sk) +{ + if (mem_cgroup_sockets_enabled && sk->sk_cgrp) + return !!sk->sk_cgrp->memory_pressure; + + return tcp_memory_pressure; +} /* * The next routines deal with comparing 32 bit unsigned ints * and worry about wraparound (automatic with unsigned arithmetic). diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index cf8b20ff6658..093779f7e893 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -359,7 +359,7 @@ static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb) /* Check #1 */ if (tp->rcv_ssthresh < tp->window_clamp && (int)tp->rcv_ssthresh < tcp_space(sk) && - !sk_under_memory_pressure(sk)) { + !tcp_under_memory_pressure(sk)) { int incr; /* Check #2. Increase window, if skb with such overhead @@ -446,7 +446,7 @@ static void tcp_clamp_window(struct sock *sk) if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] && !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) && - !sk_under_memory_pressure(sk) && + !tcp_under_memory_pressure(sk) && sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) { sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc), sysctl_tcp_rmem[2]); @@ -4781,7 +4781,7 @@ static int tcp_prune_queue(struct sock *sk) if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) tcp_clamp_window(sk); - else if (sk_under_memory_pressure(sk)) + else if (tcp_under_memory_pressure(sk)) tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss); tcp_collapse_ofo_queue(sk); @@ -4825,7 +4825,7 @@ static bool tcp_should_expand_sndbuf(const struct sock *sk) return false; /* If we are under global TCP memory pressure, do not expand. */ - if (sk_under_memory_pressure(sk)) + if (tcp_under_memory_pressure(sk)) return false; /* If we are under soft global TCP memory pressure, do not expand. */ diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index bac1a950d087..08c2cc40b26d 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -2392,7 +2392,7 @@ u32 __tcp_select_window(struct sock *sk) if (free_space < (full_space >> 1)) { icsk->icsk_ack.quick = 0; - if (sk_under_memory_pressure(sk)) + if (tcp_under_memory_pressure(sk)) tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss); @@ -2843,7 +2843,7 @@ void tcp_send_fin(struct sock *sk) * Note: in the latter case, FIN packet will be sent after a timeout, * as TCP stack thinks it has already been transmitted. */ - if (tskb && (tcp_send_head(sk) || sk_under_memory_pressure(sk))) { + if (tskb && (tcp_send_head(sk) || tcp_under_memory_pressure(sk))) { coalesce: TCP_SKB_CB(tskb)->tcp_flags |= TCPHDR_FIN; TCP_SKB_CB(tskb)->end_seq++; diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 65bf670e8714..5b752f58a900 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -247,7 +247,7 @@ void tcp_delack_timer_handler(struct sock *sk) } out: - if (sk_under_memory_pressure(sk)) + if (tcp_under_memory_pressure(sk)) sk_mem_reclaim(sk); } -- cgit v1.3-18-gd494