From 2e77d89b2fa8e3f8325b8ce7893ec3645f41aff5 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 24 Nov 2008 15:52:46 -0800 Subject: net: avoid a pair of dst_hold()/dst_release() in ip_append_data() We can reduce pressure on dst entry refcount that slowdown UDP transmit path on SMP machines. This pressure is visible on RTP servers when delivering content to mediagateways, especially big ones, handling thousand of streams. Several cpus send UDP frames to the same destination, hence use the same dst entry. This patch makes ip_append_data() eventually steal the refcount its callers had to take on the dst entry. This doesnt avoid all refcounting, but still gives speedups on SMP, on UDP/RAW transmit path Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/ip.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/net/ip.h') diff --git a/include/net/ip.h b/include/net/ip.h index bc026ecb513..ddef10c22e3 100644 --- a/include/net/ip.h +++ b/include/net/ip.h @@ -110,7 +110,7 @@ extern int ip_append_data(struct sock *sk, int odd, struct sk_buff *skb), void *from, int len, int protolen, struct ipcm_cookie *ipc, - struct rtable *rt, + struct rtable **rt, unsigned int flags); extern int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb); extern ssize_t ip_append_page(struct sock *sk, struct page *page, -- cgit v1.2.3-70-g09d2 From b27aeadb5948d400df83db4d29590fb9862ba49d Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Tue, 25 Nov 2008 18:00:48 -0800 Subject: netns xfrm: per-netns sysctls Make net.core.xfrm_aevent_etime net.core.xfrm_acq_expires net.core.xfrm_aevent_rseqth net.core.xfrm_larval_drop sysctls per-netns. For that make net_core_path[] global, register it to prevent two /proc/net/core antries and change initcall position -- xfrm_init() is called from fs_initcall, so this one should be fs_initcall at least. Signed-off-by: Alexey Dobriyan Signed-off-by: David S. Miller --- include/net/ip.h | 1 + include/net/netns/xfrm.h | 10 ++++++ include/net/xfrm.h | 14 +++++--- net/core/sysctl_net_core.c | 42 +++-------------------- net/xfrm/Makefile | 4 +-- net/xfrm/xfrm_policy.c | 10 ++++-- net/xfrm/xfrm_state.c | 16 +++------ net/xfrm/xfrm_sysctl.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++ net/xfrm/xfrm_user.c | 4 +-- 9 files changed, 125 insertions(+), 61 deletions(-) create mode 100644 net/xfrm/xfrm_sysctl.c (limited to 'include/net/ip.h') diff --git a/include/net/ip.h b/include/net/ip.h index ddef10c22e3..10868139e65 100644 --- a/include/net/ip.h +++ b/include/net/ip.h @@ -187,6 +187,7 @@ extern void inet_get_local_port_range(int *low, int *high); extern int sysctl_ip_default_ttl; extern int sysctl_ip_nonlocal_bind; +extern struct ctl_path net_core_path[]; extern struct ctl_path net_ipv4_ctl_path[]; /* From inetpeer.c */ diff --git a/include/net/netns/xfrm.h b/include/net/netns/xfrm.h index 09f3060e9d1..1ba912749ca 100644 --- a/include/net/netns/xfrm.h +++ b/include/net/netns/xfrm.h @@ -6,6 +6,8 @@ #include #include +struct ctl_table_header; + struct xfrm_policy_hash { struct hlist_head *table; unsigned int hmask; @@ -41,6 +43,14 @@ struct netns_xfrm { struct work_struct policy_hash_work; struct sock *nlsk; + + u32 sysctl_aevent_etime; + u32 sysctl_aevent_rseqth; + int sysctl_larval_drop; + u32 sysctl_acq_expires; +#ifdef CONFIG_SYSCTL + struct ctl_table_header *sysctl_hdr; +#endif }; #endif diff --git a/include/net/xfrm.h b/include/net/xfrm.h index 1554ccd0c94..2e9f5c0018a 100644 --- a/include/net/xfrm.h +++ b/include/net/xfrm.h @@ -47,11 +47,6 @@ #define XFRM_INC_STATS_USER(net, field) ((void)(net)) #endif -extern u32 sysctl_xfrm_aevent_etime; -extern u32 sysctl_xfrm_aevent_rseqth; -extern int sysctl_xfrm_larval_drop; -extern u32 sysctl_xfrm_acq_expires; - extern struct mutex xfrm_cfg_mutex; /* Organization of SPD aka "XFRM rules" @@ -1310,6 +1305,15 @@ extern int xfrm_proc_init(struct net *net); extern void xfrm_proc_fini(struct net *net); #endif +extern int xfrm_sysctl_init(struct net *net); +#ifdef CONFIG_SYSCTL +extern void xfrm_sysctl_fini(struct net *net); +#else +static inline void xfrm_sysctl_fini(struct net *net) +{ +} +#endif + extern void xfrm_state_walk_init(struct xfrm_state_walk *walk, u8 proto); extern int xfrm_state_walk(struct net *net, struct xfrm_state_walk *walk, int (*func)(struct xfrm_state *, int, void*), void *); diff --git a/net/core/sysctl_net_core.c b/net/core/sysctl_net_core.c index 2bc0384b044..83d3398559e 100644 --- a/net/core/sysctl_net_core.c +++ b/net/core/sysctl_net_core.c @@ -12,7 +12,6 @@ #include #include #include -#include static struct ctl_table net_core_table[] = { #ifdef CONFIG_NET @@ -89,40 +88,6 @@ static struct ctl_table net_core_table[] = { .mode = 0644, .proc_handler = proc_dointvec }, -#ifdef CONFIG_XFRM - { - .ctl_name = NET_CORE_AEVENT_ETIME, - .procname = "xfrm_aevent_etime", - .data = &sysctl_xfrm_aevent_etime, - .maxlen = sizeof(u32), - .mode = 0644, - .proc_handler = proc_dointvec - }, - { - .ctl_name = NET_CORE_AEVENT_RSEQTH, - .procname = "xfrm_aevent_rseqth", - .data = &sysctl_xfrm_aevent_rseqth, - .maxlen = sizeof(u32), - .mode = 0644, - .proc_handler = proc_dointvec - }, - { - .ctl_name = CTL_UNNUMBERED, - .procname = "xfrm_larval_drop", - .data = &sysctl_xfrm_larval_drop, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = proc_dointvec - }, - { - .ctl_name = CTL_UNNUMBERED, - .procname = "xfrm_acq_expires", - .data = &sysctl_xfrm_acq_expires, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = proc_dointvec - }, -#endif /* CONFIG_XFRM */ #endif /* CONFIG_NET */ { .ctl_name = NET_CORE_BUDGET, @@ -155,7 +120,7 @@ static struct ctl_table netns_core_table[] = { { .ctl_name = 0 } }; -static __net_initdata struct ctl_path net_core_path[] = { +__net_initdata struct ctl_path net_core_path[] = { { .procname = "net", .ctl_name = CTL_NET, }, { .procname = "core", .ctl_name = NET_CORE, }, { }, @@ -207,8 +172,11 @@ static __net_initdata struct pernet_operations sysctl_core_ops = { static __init int sysctl_core_init(void) { + static struct ctl_table empty[1]; + + register_sysctl_paths(net_core_path, empty); register_net_sysctl_rotable(net_core_path, net_core_table); return register_pernet_subsys(&sysctl_core_ops); } -__initcall(sysctl_core_init); +fs_initcall(sysctl_core_init); diff --git a/net/xfrm/Makefile b/net/xfrm/Makefile index 0f439a72cca..c631047e1b2 100644 --- a/net/xfrm/Makefile +++ b/net/xfrm/Makefile @@ -3,8 +3,8 @@ # obj-$(CONFIG_XFRM) := xfrm_policy.o xfrm_state.o xfrm_hash.o \ - xfrm_input.o xfrm_output.o xfrm_algo.o + xfrm_input.o xfrm_output.o xfrm_algo.o \ + xfrm_sysctl.o obj-$(CONFIG_XFRM_STATISTICS) += xfrm_proc.o obj-$(CONFIG_XFRM_USER) += xfrm_user.o obj-$(CONFIG_XFRM_IPCOMP) += xfrm_ipcomp.o - diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c index 38822b34ba7..393cc65dbfa 100644 --- a/net/xfrm/xfrm_policy.c +++ b/net/xfrm/xfrm_policy.c @@ -34,8 +34,6 @@ #include "xfrm_hash.h" -int sysctl_xfrm_larval_drop __read_mostly = 1; - DEFINE_MUTEX(xfrm_cfg_mutex); EXPORT_SYMBOL(xfrm_cfg_mutex); @@ -1671,7 +1669,7 @@ restart: if (unlikely(nx<0)) { err = nx; - if (err == -EAGAIN && sysctl_xfrm_larval_drop) { + if (err == -EAGAIN && net->xfrm.sysctl_larval_drop) { /* EREMOTE tells the caller to generate * a one-shot blackhole route. */ @@ -2504,8 +2502,13 @@ static int __net_init xfrm_net_init(struct net *net) rv = xfrm_policy_init(net); if (rv < 0) goto out_policy; + rv = xfrm_sysctl_init(net); + if (rv < 0) + goto out_sysctl; return 0; +out_sysctl: + xfrm_policy_fini(net); out_policy: xfrm_state_fini(net); out_state: @@ -2516,6 +2519,7 @@ out_statistics: static void __net_exit xfrm_net_exit(struct net *net) { + xfrm_sysctl_fini(net); xfrm_policy_fini(net); xfrm_state_fini(net); xfrm_statistics_fini(net); diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c index 662e47b0bcc..2fd57f8f77c 100644 --- a/net/xfrm/xfrm_state.c +++ b/net/xfrm/xfrm_state.c @@ -24,14 +24,6 @@ #include "xfrm_hash.h" -u32 sysctl_xfrm_aevent_etime __read_mostly = XFRM_AE_ETIME; -EXPORT_SYMBOL(sysctl_xfrm_aevent_etime); - -u32 sysctl_xfrm_aevent_rseqth __read_mostly = XFRM_AE_SEQT_SIZE; -EXPORT_SYMBOL(sysctl_xfrm_aevent_rseqth); - -u32 sysctl_xfrm_acq_expires __read_mostly = 30; - /* Each xfrm_state may be linked to two tables: 1. Hash table by (spi,daddr,ah/esp) to find SA by SPI. (input,ctl) @@ -851,8 +843,8 @@ xfrm_state_find(xfrm_address_t *daddr, xfrm_address_t *saddr, h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, family); hlist_add_head(&x->byspi, net->xfrm.state_byspi+h); } - x->lft.hard_add_expires_seconds = sysctl_xfrm_acq_expires; - x->timer.expires = jiffies + sysctl_xfrm_acq_expires*HZ; + x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires; + x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ; add_timer(&x->timer); net->xfrm.state_num++; xfrm_hash_grow_check(net, x->bydst.next != NULL); @@ -1040,9 +1032,9 @@ static struct xfrm_state *__find_acq_core(struct net *net, unsigned short family x->props.family = family; x->props.mode = mode; x->props.reqid = reqid; - x->lft.hard_add_expires_seconds = sysctl_xfrm_acq_expires; + x->lft.hard_add_expires_seconds = net->xfrm.sysctl_acq_expires; xfrm_state_hold(x); - x->timer.expires = jiffies + sysctl_xfrm_acq_expires*HZ; + x->timer.expires = jiffies + net->xfrm.sysctl_acq_expires*HZ; add_timer(&x->timer); list_add(&x->km.all, &net->xfrm.state_all); hlist_add_head(&x->bydst, net->xfrm.state_bydst+h); diff --git a/net/xfrm/xfrm_sysctl.c b/net/xfrm/xfrm_sysctl.c new file mode 100644 index 00000000000..2e6ffb66f06 --- /dev/null +++ b/net/xfrm/xfrm_sysctl.c @@ -0,0 +1,85 @@ +#include +#include +#include + +static void __xfrm_sysctl_init(struct net *net) +{ + net->xfrm.sysctl_aevent_etime = XFRM_AE_ETIME; + net->xfrm.sysctl_aevent_rseqth = XFRM_AE_SEQT_SIZE; + net->xfrm.sysctl_larval_drop = 1; + net->xfrm.sysctl_acq_expires = 30; +} + +#ifdef CONFIG_SYSCTL +static struct ctl_table xfrm_table[] = { + { + .ctl_name = NET_CORE_AEVENT_ETIME, + .procname = "xfrm_aevent_etime", + .maxlen = sizeof(u32), + .mode = 0644, + .proc_handler = proc_dointvec + }, + { + .ctl_name = NET_CORE_AEVENT_RSEQTH, + .procname = "xfrm_aevent_rseqth", + .maxlen = sizeof(u32), + .mode = 0644, + .proc_handler = proc_dointvec + }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "xfrm_larval_drop", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec + }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "xfrm_acq_expires", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec + }, + {} +}; + +int __net_init xfrm_sysctl_init(struct net *net) +{ + struct ctl_table *table; + + __xfrm_sysctl_init(net); + + table = kmemdup(xfrm_table, sizeof(xfrm_table), GFP_KERNEL); + if (!table) + goto out_kmemdup; + table[0].data = &net->xfrm.sysctl_aevent_etime; + table[1].data = &net->xfrm.sysctl_aevent_rseqth; + table[2].data = &net->xfrm.sysctl_larval_drop; + table[3].data = &net->xfrm.sysctl_acq_expires; + + net->xfrm.sysctl_hdr = register_net_sysctl_table(net, net_core_path, table); + if (!net->xfrm.sysctl_hdr) + goto out_register; + return 0; + +out_register: + kfree(table); +out_kmemdup: + return -ENOMEM; +} + +void xfrm_sysctl_fini(struct net *net) +{ + struct ctl_table *table; + + table = net->xfrm.sysctl_hdr->ctl_table_arg; + unregister_net_sysctl_table(net->xfrm.sysctl_hdr); + kfree(table); +} +#else +int __net_init xfrm_sysctl_init(struct net *net) +{ + __xfrm_sysctl_init(net); + return 0; +} +#endif diff --git a/net/xfrm/xfrm_user.c b/net/xfrm/xfrm_user.c index b7240d5b77a..38ffaf33312 100644 --- a/net/xfrm/xfrm_user.c +++ b/net/xfrm/xfrm_user.c @@ -368,9 +368,9 @@ static struct xfrm_state *xfrm_state_construct(struct net *net, goto error; x->km.seq = p->seq; - x->replay_maxdiff = sysctl_xfrm_aevent_rseqth; + x->replay_maxdiff = net->xfrm.sysctl_aevent_rseqth; /* sysctl_xfrm_aevent_etime is in 100ms units */ - x->replay_maxage = (sysctl_xfrm_aevent_etime*HZ)/XFRM_AE_ETH_M; + x->replay_maxage = (net->xfrm.sysctl_aevent_etime*HZ)/XFRM_AE_ETH_M; x->preplay.bitmap = 0; x->preplay.seq = x->replay.seq+x->replay_maxdiff; x->preplay.oseq = x->replay.oseq +x->replay_maxdiff; -- cgit v1.2.3-70-g09d2