Commit 2a24444f authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller

ipv6: reduce percpu needs for icmpv6msg mibs

Reading /proc/net/snmp6 on a machine with a lot of cpus is very
expensive (can be ~88000 us).

This is because ICMPV6MSG MIB uses 4096 bytes per cpu, and folding
values for all possible cpus can read 16 Mbytes of memory (32MBytes on
non x86 arches)

ICMP messages are not considered as fast path on a typical server, and
eventually few cpus handle them anyway. We can afford an atomic
operation instead of using percpu data.

This saves 4096 bytes per cpu and per network namespace.
Signed-off-by: default avatarEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 3d249d4c
No related merge requests found
...@@ -132,6 +132,15 @@ extern struct ctl_path net_ipv6_ctl_path[]; ...@@ -132,6 +132,15 @@ extern struct ctl_path net_ipv6_ctl_path[];
SNMP_INC_STATS##modifier((net)->mib.statname##_statistics, (field));\ SNMP_INC_STATS##modifier((net)->mib.statname##_statistics, (field));\
}) })
/* per device and per net counters are atomic_long_t */
#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \
({ \
struct inet6_dev *_idev = (idev); \
if (likely(_idev != NULL)) \
SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\
})
#define _DEVADD(net, statname, modifier, idev, field, val) \ #define _DEVADD(net, statname, modifier, idev, field, val) \
({ \ ({ \
struct inet6_dev *_idev = (idev); \ struct inet6_dev *_idev = (idev); \
...@@ -168,11 +177,11 @@ extern struct ctl_path net_ipv6_ctl_path[]; ...@@ -168,11 +177,11 @@ extern struct ctl_path net_ipv6_ctl_path[];
_DEVINCATOMIC(net, icmpv6, _BH, idev, field) _DEVINCATOMIC(net, icmpv6, _BH, idev, field)
#define ICMP6MSGOUT_INC_STATS(net, idev, field) \ #define ICMP6MSGOUT_INC_STATS(net, idev, field) \
_DEVINCATOMIC(net, icmpv6msg, , idev, field +256) _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
#define ICMP6MSGOUT_INC_STATS_BH(net, idev, field) \ #define ICMP6MSGOUT_INC_STATS_BH(net, idev, field) \
_DEVINCATOMIC(net, icmpv6msg, _BH, idev, field +256) _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
#define ICMP6MSGIN_INC_STATS_BH(net, idev, field) \ #define ICMP6MSGIN_INC_STATS_BH(net, idev, field) \
_DEVINCATOMIC(net, icmpv6msg, _BH, idev, field) _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field)
struct ip6_ra_chain { struct ip6_ra_chain {
struct ip6_ra_chain *next; struct ip6_ra_chain *next;
......
...@@ -18,7 +18,7 @@ struct netns_mib { ...@@ -18,7 +18,7 @@ struct netns_mib {
DEFINE_SNMP_STAT(struct udp_mib, udplite_stats_in6); DEFINE_SNMP_STAT(struct udp_mib, udplite_stats_in6);
DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics); DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics);
DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics); DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics);
DEFINE_SNMP_STAT(struct icmpv6msg_mib, icmpv6msg_statistics); DEFINE_SNMP_STAT_ATOMIC(struct icmpv6msg_mib, icmpv6msg_statistics);
#endif #endif
#ifdef CONFIG_XFRM_STATISTICS #ifdef CONFIG_XFRM_STATISTICS
DEFINE_SNMP_STAT(struct linux_xfrm_mib, xfrm_statistics); DEFINE_SNMP_STAT(struct linux_xfrm_mib, xfrm_statistics);
......
...@@ -84,7 +84,7 @@ struct icmpv6_mib_device { ...@@ -84,7 +84,7 @@ struct icmpv6_mib_device {
#define ICMP6MSG_MIB_MAX __ICMP6MSG_MIB_MAX #define ICMP6MSG_MIB_MAX __ICMP6MSG_MIB_MAX
/* per network ns counters */ /* per network ns counters */
struct icmpv6msg_mib { struct icmpv6msg_mib {
unsigned long mibs[ICMP6MSG_MIB_MAX]; atomic_long_t mibs[ICMP6MSG_MIB_MAX];
}; };
/* per device counters, (shared on all cpus) */ /* per device counters, (shared on all cpus) */
struct icmpv6msg_mib_device { struct icmpv6msg_mib_device {
......
...@@ -985,9 +985,9 @@ static int __net_init ipv6_init_mibs(struct net *net) ...@@ -985,9 +985,9 @@ static int __net_init ipv6_init_mibs(struct net *net)
sizeof(struct icmpv6_mib), sizeof(struct icmpv6_mib),
__alignof__(struct icmpv6_mib)) < 0) __alignof__(struct icmpv6_mib)) < 0)
goto err_icmp_mib; goto err_icmp_mib;
if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics, net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
sizeof(struct icmpv6msg_mib), GFP_KERNEL);
__alignof__(struct icmpv6msg_mib)) < 0) if (!net->mib.icmpv6msg_statistics)
goto err_icmpmsg_mib; goto err_icmpmsg_mib;
return 0; return 0;
...@@ -1008,7 +1008,7 @@ static void ipv6_cleanup_mibs(struct net *net) ...@@ -1008,7 +1008,7 @@ static void ipv6_cleanup_mibs(struct net *net)
snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6); snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
snmp_mib_free((void __percpu **)net->mib.ipv6_statistics); snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics); snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics); kfree(net->mib.icmpv6msg_statistics);
} }
static int __net_init inet6_net_init(struct net *net) static int __net_init inet6_net_init(struct net *net)
......
...@@ -142,11 +142,7 @@ static const struct snmp_mib snmp6_udplite6_list[] = { ...@@ -142,11 +142,7 @@ static const struct snmp_mib snmp6_udplite6_list[] = {
SNMP_MIB_SENTINEL SNMP_MIB_SENTINEL
}; };
/* can be called either with percpu mib (pcpumib != NULL), static void snmp6_seq_show_icmpv6msg(struct seq_file *seq, atomic_long_t *smib)
* or shared one (smib != NULL)
*/
static void snmp6_seq_show_icmpv6msg(struct seq_file *seq, void __percpu **pcpumib,
atomic_long_t *smib)
{ {
char name[32]; char name[32];
int i; int i;
...@@ -163,14 +159,14 @@ static void snmp6_seq_show_icmpv6msg(struct seq_file *seq, void __percpu **pcpum ...@@ -163,14 +159,14 @@ static void snmp6_seq_show_icmpv6msg(struct seq_file *seq, void __percpu **pcpum
snprintf(name, sizeof(name), "Icmp6%s%s", snprintf(name, sizeof(name), "Icmp6%s%s",
i & 0x100 ? "Out" : "In", p); i & 0x100 ? "Out" : "In", p);
seq_printf(seq, "%-32s\t%lu\n", name, seq_printf(seq, "%-32s\t%lu\n", name,
pcpumib ? snmp_fold_field(pcpumib, i) : atomic_long_read(smib + i)); atomic_long_read(smib + i));
} }
/* print by number (nonzero only) - ICMPMsgStat format */ /* print by number (nonzero only) - ICMPMsgStat format */
for (i = 0; i < ICMP6MSG_MIB_MAX; i++) { for (i = 0; i < ICMP6MSG_MIB_MAX; i++) {
unsigned long val; unsigned long val;
val = pcpumib ? snmp_fold_field(pcpumib, i) : atomic_long_read(smib + i); val = atomic_long_read(smib + i);
if (!val) if (!val)
continue; continue;
snprintf(name, sizeof(name), "Icmp6%sType%u", snprintf(name, sizeof(name), "Icmp6%sType%u",
...@@ -215,8 +211,7 @@ static int snmp6_seq_show(struct seq_file *seq, void *v) ...@@ -215,8 +211,7 @@ static int snmp6_seq_show(struct seq_file *seq, void *v)
snmp6_ipstats_list, offsetof(struct ipstats_mib, syncp)); snmp6_ipstats_list, offsetof(struct ipstats_mib, syncp));
snmp6_seq_show_item(seq, (void __percpu **)net->mib.icmpv6_statistics, snmp6_seq_show_item(seq, (void __percpu **)net->mib.icmpv6_statistics,
NULL, snmp6_icmp6_list); NULL, snmp6_icmp6_list);
snmp6_seq_show_icmpv6msg(seq, snmp6_seq_show_icmpv6msg(seq, net->mib.icmpv6msg_statistics->mibs);
(void __percpu **)net->mib.icmpv6msg_statistics, NULL);
snmp6_seq_show_item(seq, (void __percpu **)net->mib.udp_stats_in6, snmp6_seq_show_item(seq, (void __percpu **)net->mib.udp_stats_in6,
NULL, snmp6_udp6_list); NULL, snmp6_udp6_list);
snmp6_seq_show_item(seq, (void __percpu **)net->mib.udplite_stats_in6, snmp6_seq_show_item(seq, (void __percpu **)net->mib.udplite_stats_in6,
...@@ -246,7 +241,7 @@ static int snmp6_dev_seq_show(struct seq_file *seq, void *v) ...@@ -246,7 +241,7 @@ static int snmp6_dev_seq_show(struct seq_file *seq, void *v)
snmp6_ipstats_list); snmp6_ipstats_list);
snmp6_seq_show_item(seq, NULL, idev->stats.icmpv6dev->mibs, snmp6_seq_show_item(seq, NULL, idev->stats.icmpv6dev->mibs,
snmp6_icmp6_list); snmp6_icmp6_list);
snmp6_seq_show_icmpv6msg(seq, NULL, idev->stats.icmpv6msgdev->mibs); snmp6_seq_show_icmpv6msg(seq, idev->stats.icmpv6msgdev->mibs);
return 0; return 0;
} }
......
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