Commit ef34c0f6 authored by David S. Miller's avatar David S. Miller
Browse files

Merge branch 'snmp-stat-aggregation' 

Raghavendra K T says:

====================
Optimize the snmp stat aggregation for large cpus

While creating 1000 containers, perf is showing lot of time spent in
snmp_fold_field on a large cpu system.

The current patch tries to improve by reordering the statistics gathering.

Please note that similar overhead was also reported while creating
veth pairs  https://lkml.org/lkml/2013/3/19/556



Changes in V4:
 - remove 'item' variable and use IPSTATS_MIB_MAX to avoid sparse
   warning (Eric) also remove 'item' parameter (Joe)
 - add missing memset of padding.

Changes in V3:
 - use memset to initialize temp buffer in leaf function. (David)
 - use memcpy to copy the buffer data to stat instead of unalign_pu (Joe)
 - Move buffer definition to leaf function __snmp6_fill_stats64() (Eric)
 -
Changes in V2:
 - Allocate the stat calculation buffer in stack. (Eric)

Setup:
160 cpu (20 core) baremetal powerpc system with 1TB memory

1000 docker containers was created with command
docker run -itd  ubuntu:15.04  /bin/bash in loop

observation:
Docker container creation linearly increased from around 1.6 sec to 7.5 sec
(at 1000 containers) perf data showed, creating veth interfaces resulting in
the below code path was taking more time.

rtnl_fill_ifinfo
  -> inet6_fill_link_af
    -> inet6_fill_ifla6_attrs
      -> snmp_fold_field

proposed idea:
 currently __snmp6_fill_stats64 calls snmp_fold_field that walks
through per cpu data to of an item (iteratively for around 36 items).
 The patch tries to aggregate the statistics by going through
all the items of each cpu sequentially which is reducing cache
misses.

Performance of docker creation improved by around more than 2x
after the patch.

before the patch:
================
3f45ba571a42e925c4ec4aaee0e48d7610a9ed82a4c931f83324d41822cf6617
real	0m6.836s
user	0m0.095s
sys	0m0.011s

perf record -a docker run -itd  ubuntu:15.04  /bin/bash
=======================================================
    50.73%  docker           [kernel.kallsyms]       [k] snmp_fold_field
     9.07%  swapper          [kernel.kallsyms]       [k] snooze_loop
     3.49%  docker           [kernel.kallsyms]       [k] veth_stats_one
     2.85%  swapper          [kernel.kallsyms]       [k] _raw_spin_lock
     1.37%  docker           docker                  [.] backtrace_qsort
     1.31%  docker           docker                  [.] strings.FieldsFunc

  cache-misses:  2.7%

after the patch:
=============
9178273e9df399c8290b6c196e4aef9273be2876225f63b14a60cf97eacfafb5
real	0m3.249s
user	0m0.088s
sys	0m0.020s

perf record -a docker run -itd  ubuntu:15.04  /bin/bash
=======================================================
    10.57%  docker           docker                [.] scanblock
     8.37%  swapper          [kernel.kallsyms]     [k] snooze_loop
     6.91%  docker           [kernel.kallsyms]     [k] snmp_get_cpu_field
     6.67%  docker           [kernel.kallsyms]     [k] veth_stats_one
     3.96%  docker           docker                [.] runtime_MSpan_Sweep
     2.47%  docker           docker                [.] strings.FieldsFunc

cache-misses: 1.41 %

Please let me know if you have suggestions/comments.
Thanks Eric, Joe and David for the comments.
====================

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 06fb4e70 a3a77372

include/net/ip.h

+10 −0
Original line number Diff line number Diff line
@@ -202,10 +202,20 @@ void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb, 
#define NET_ADD_STATS_BH(net, field, adnd) SNMP_ADD_STATS_BH((net)->mib.net_statistics, field, adnd)

#define NET_ADD_STATS_USER(net, field, adnd) SNMP_ADD_STATS_USER((net)->mib.net_statistics, field, adnd)



u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);

unsigned long snmp_fold_field(void __percpu *mib, int offt);

#if BITS_PER_LONG==32

u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,

			 size_t syncp_offset);

u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);

#else

static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,

					size_t syncp_offset)

{

	return snmp_get_cpu_field(mib, cpu, offct);



}



static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)

{

	return snmp_fold_field(mib, offt);

net/ipv4/af_inet.c

+27 −14
Original line number Diff line number Diff line
@@ -1452,25 +1452,28 @@ int inet_ctl_sock_create(struct sock **sk, unsigned short family, 
}

EXPORT_SYMBOL_GPL(inet_ctl_sock_create);



u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)

{

	return  *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);

}

EXPORT_SYMBOL_GPL(snmp_get_cpu_field);



unsigned long snmp_fold_field(void __percpu *mib, int offt)

{

	unsigned long res = 0;

	int i;



	for_each_possible_cpu(i)

		res += *(((unsigned long *) per_cpu_ptr(mib, i)) + offt);

		res += snmp_get_cpu_field(mib, i, offt);

	return res;

}

EXPORT_SYMBOL_GPL(snmp_fold_field);



#if BITS_PER_LONG==32



u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)

u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,

			 size_t syncp_offset)

{

	u64 res = 0;

	int cpu;



	for_each_possible_cpu(cpu) {

	void *bhptr;

	struct u64_stats_sync *syncp;

	u64 v;
@@ -1483,7 +1486,17 @@ u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) 
		v = *(((u64 *)bhptr) + offt);

	} while (u64_stats_fetch_retry_irq(syncp, start));



		res += v;

	return v;

}

EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);



u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)

{

	u64 res = 0;

	int cpu;



	for_each_possible_cpu(cpu) {

		res += snmp_get_cpu_field(mib, cpu, offct, syncp_offset);

	}

	return res;

}

net/ipv6/addrconf.c

+16 −10
Original line number Diff line number Diff line
@@ -4726,18 +4726,24 @@ static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 
}



static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,

				      int items, int bytes, size_t syncpoff)

					int bytes, size_t syncpoff)

{

	int i;

	int pad = bytes - sizeof(u64) * items;

	int i, c;

	u64 buff[IPSTATS_MIB_MAX];

	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;



	BUG_ON(pad < 0);



	/* Use put_unaligned() because stats may not be aligned for u64. */

	put_unaligned(items, &stats[0]);

	for (i = 1; i < items; i++)

		put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);

	memset(buff, 0, sizeof(buff));

	buff[0] = IPSTATS_MIB_MAX;



	memset(&stats[items], 0, pad);

	for_each_possible_cpu(c) {

		for (i = 1; i < IPSTATS_MIB_MAX; i++)

			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);

	}



	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));

	memset(&stats[IPSTATS_MIB_MAX], 0, pad);

}



static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
@@ -4745,8 +4751,8 @@ static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 
{

	switch (attrtype) {

	case IFLA_INET6_STATS:

		__snmp6_fill_stats64(stats, idev->stats.ipv6,

				     IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));

		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,

				     offsetof(struct ipstats_mib, syncp));

		break;

	case IFLA_INET6_ICMP6STATS:

		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);

CRA Git | Maintained and supported by SUSTech CRA and CCSE