netfilter: conntrack: switch to siphash
Subsystems (3)
Florian Westphal
maintainercover netfilter: conntrack: switch to siphash
Two recent commits switched inet rt and nexthop exception hashes from jhash to siphash.
If those two spots are problematic then conntrack is affected as well, so switch voer to siphash too.
While at it, add a hard upper limit on chain lengths and reject insertion if this is hit.
Florian Westphal (3):
netfilter: conntrack: sanitize table size default settings
netfilter: conntrack: switch to siphash
netfilter: conntrack: refuse insertion if chain has grown too large
.../networking/nf_conntrack-sysctl.rst | 13 ++- include/linux/netfilter/nf_conntrack_common.h | 1 + .../linux/netfilter/nfnetlink_conntrack.h | 1 + net/netfilter/nf_conntrack_core.c | 103 ++++++++++++------ net/netfilter/nf_conntrack_expect.c | 25 +++-- net/netfilter/nf_conntrack_netlink.c | 4 +- net/netfilter/nf_conntrack_standalone.c | 4 +- net/netfilter/nf_nat_core.c | 18 ++- 8 files changed, 114 insertions(+), 55 deletions(-)
Signature · 1 line
2.31.1
Florian Westphal
maintainerPATCH 1/3 netfilter: conntrack: sanitize table size default settings
conntrack has two distinct table size settings:nf_conntrack_max and nf_conntrack_buckets.
The former limits how many conntrack objects are allowed to exist in each namespace.
The second sets the size of the hashtable.
As all entries are inserted twice (once for original direction, once for reply), there should be at least twice as many buckets in the table than the maximum number of conntrack objects that can exist at the same time.
Change the default multiplier to 1 and increase the chosen bucket sizes. This results in the same nf_conntrack_max settings as before but reduces the average bucket list length.
--- .../networking/nf_conntrack-sysctl.rst | 13 ++++---- net/netfilter/nf_conntrack_core.c | 30 +++++++++---------- 2 files changed, 22 insertions(+), 21 deletions(-)
Documentation/networking/nf_conntrack-sysctl.rst +8 -5
@@ -17,9 +17,8 @@ nf_conntrack_acct - BOOLEAN nf_conntrack_buckets - INTEGER Size of hash table. If not specified as parameter during module loading, the default size is calculated by dividing total memory - by 16384 to determine the number of buckets but the hash table will - never have fewer than 32 and limited to 16384 buckets. For systems - with more than 4GB of memory it will be 65536 buckets. + by 16384 to determine the number of buckets. The hash table will + never have fewer than 1024 and never more than 262144 buckets. This sysctl is only writeable in the initial net namespace. nf_conntrack_checksum - BOOLEAN
@@ -100,8 +99,12 @@ nf_conntrack_log_invalid - INTEGER Log invalid packets of a type specified by value. nf_conntrack_max - INTEGER - Size of connection tracking table. Default value is - nf_conntrack_buckets value * 4. + Maximum number of allowed connection tracking entries. This value is set + to nf_conntrack_buckets by default. + Note that connection tracking entries are added to the table twice -- once + for the original direction and once for the reply direction (i.e., with + the reversed address). This means that with default settings a maxed-out + table will have a average hash chain length of 2, not 1. nf_conntrack_tcp_be_liberal - BOOLEAN - 0 - disabled (default)
net/netfilter/nf_conntrack_core.c +14 -16
@@ -2594,26 +2594,24 @@ int nf_conntrack_init_start(void)spin_lock_init (&nf_conntrack_locks[i]);if (!nf_conntrack_htable_size) { -/* Idea from tcp.c: use 1/16384 of memory. -* On i386: 32MB machine has 512 buckets. -* >= 1GB machines have 16384 buckets. -* >= 4GB machines have 65536 buckets. -*/ nf_conntrack_htable_size = (((nr_pages <<PAGE_SHIFT ) /16384 ) /sizeof (struct hlist_head )); -if (nr_pages > (4 * (1024 *1024 *1024 /PAGE_SIZE ))) - nf_conntrack_htable_size =65536 ; +if (BITS_PER_LONG >=64 && + nr_pages > (4 * (1024 *1024 *1024 /PAGE_SIZE ))) + nf_conntrack_htable_size =262144 ;else if (nr_pages > (1024 *1024 *1024 /PAGE_SIZE )) - nf_conntrack_htable_size =16384 ; -if (nf_conntrack_htable_size <32 ) - nf_conntrack_htable_size =32 ; - -/* Use a max. factor of four by default to get the same max as -* with the old struct list_heads. When a table size is given -* we use the old value of 8 to avoid reducing the max. -* entries. */ - max_factor =4 ; + nf_conntrack_htable_size =65536 ; + +if (nf_conntrack_htable_size <1024 ) + nf_conntrack_htable_size =1024 ; +/* Use a max. factor of one by default to keep the average +* hash chain length at 2 entries. Each entry has to be added +* twice (once for original direction, once for reply). +* When a table size is given we use the old value of 8 to +* avoid implicit reduction of the max entries setting. +*/ + max_factor =1 ; } nf_conntrack_hash =nf_ct_alloc_hashtable (&nf_conntrack_htable_size,1 );
Signature · 1 line
2.31.1
Florian Westphal
maintainerPATCH 2/3 netfilter: conntrack: switch to siphash
In reply to Florian WestphalReplace jhash in conntrack and nat core with siphash.
While at it, use the netns mix value as part of the input key rather than abuse the seed value.
--- net/netfilter/nf_conntrack_core.c | 31 +++++++++++++++++------------ net/netfilter/nf_conntrack_expect.c | 25 ++++++++++++++++------- net/netfilter/nf_nat_core.c | 18 +++++++++++++---- 3 files changed, 50 insertions(+), 24 deletions(-)
net/netfilter/nf_conntrack_core.c +18 -13
@@ -21,7 +21,6 @@#include <linux/stddef.h> #include <linux/slab.h> #include <linux/random.h> -#include <linux/jhash.h> #include <linux/siphash.h> #include <linux/err.h> #include <linux/percpu.h>
@@ -184,25 +183,31 @@ EXPORT_SYMBOL_GPL(nf_conntrack_htable_size);unsigned int nf_conntrack_max__read_mostly ;EXPORT_SYMBOL_GPL (nf_conntrack_max);seqcount_spinlock_t nf_conntrack_generation__read_mostly ; -static unsigned int nf_conntrack_hash_rnd__read_mostly ; +static siphash_key_t nf_conntrack_hash_rnd__read_mostly ;static u32 hash_conntrack_raw (const struct nf_conntrack_tuple *tuple,const struct net *net) { -unsigned int n; -u32 seed; +struct { +struct nf_conntrack_man src; +union nf_inet_addr dst_addr; +u32 net_mix; +u16 dport; +u16 proto; + }__aligned(SIPHASH_ALIGNMENT) combined;get_random_once (&nf_conntrack_hash_rnd,sizeof (nf_conntrack_hash_rnd)); -/* The direction must be ignored, so we hash everything up to the -* destination ports (which is a multiple of 4) and treat the last -* three bytes manually. -*/ - seed = nf_conntrack_hash_rnd ^net_hash_mix (net); - n = (sizeof (tuple->src) +sizeof (tuple->dst.u3)) /sizeof (u32); -return jhash2 ((u32 *)tuple, n, seed ^ - (((__force __u16 )tuple->dst.u.all <<16 ) | - tuple->dst.protonum)); +memset (&combined,0 ,sizeof (combined)); + +/* The direction must be ignored, so handle usable members manually. */ + combined.src = tuple->src; + combined.dst_addr = tuple->dst.u3; + combined.net_mix =net_hash_mix (net); + combined.dport = (__force __u16 )tuple->dst.u.all; + combined.proto = tuple->dst.protonum; + +return (u32 )siphash (&combined,sizeof (combined), &nf_conntrack_hash_rnd); }static u32 scale_hash (u32 hash)
net/netfilter/nf_conntrack_expect.c +18 -7
@@ -17,7 +17,7 @@#include <linux/err.h> #include <linux/percpu.h> #include <linux/kernel.h> -#include <linux/jhash.h> +#include <linux/siphash.h> #include <linux/moduleparam.h> #include <linux/export.h> #include <net/net_namespace.h>
@@ -41,7 +41,7 @@ EXPORT_SYMBOL_GPL(nf_ct_expect_hash);unsigned int nf_ct_expect_max__read_mostly ;static struct kmem_cache *nf_ct_expect_cachep__read_mostly ; -static unsigned int nf_ct_expect_hashrnd__read_mostly ; +static siphash_key_t nf_ct_expect_hashrnd__read_mostly ;/* nf_conntrack_expect helper functions */ void nf_ct_unlink_expect_report(struct nf_conntrack_expect *exp,
@@ -81,15 +81,26 @@ static void nf_ct_expectation_timed_out(struct timer_list *t)static unsigned int nf_ct_expect_dst_hash (const struct net *n,const struct nf_conntrack_tuple *tuple) { -unsigned int hash, seed; +struct { +union nf_inet_addr dst_addr; +u32 net_mix; +u16 dport; +u8 l3num; +u8 protonum; + }__aligned(SIPHASH_ALIGNMENT) combined; +u32 hash;get_random_once (&nf_ct_expect_hashrnd,sizeof (nf_ct_expect_hashrnd)); - seed = nf_ct_expect_hashrnd ^net_hash_mix (n); +memset (&combined,0 ,sizeof (combined)); - hash =jhash2 (tuple->dst.u3.all,ARRAY_SIZE (tuple->dst.u3.all), - (((tuple->dst.protonum ^ tuple->src.l3num) <<16 ) | - (__force __u16 )tuple->dst.u.all) ^ seed); + combined.dst_addr = tuple->dst.u3; + combined.net_mix =net_hash_mix (n); + combined.dport = (__force __u16 )tuple->dst.u.all; + combined.l3num = tuple->src.l3num; + combined.protonum = tuple->dst.protonum; + + hash =siphash (&combined,sizeof (combined), &nf_ct_expect_hashrnd);return reciprocal_scale (hash, nf_ct_expect_hsize); }
net/netfilter/nf_nat_core.c +14 -4
@@ -13,7 +13,7 @@#include <linux/skbuff.h> #include <linux/gfp.h> #include <net/xfrm.h> -#include <linux/jhash.h> +#include <linux/siphash.h> #include <linux/rtnetlink.h> #include <net/netfilter/nf_conntrack.h>
@@ -34,7 +34,7 @@ static unsigned int nat_net_id __read_mostly;static struct hlist_head *nf_nat_bysource__read_mostly ;static unsigned int nf_nat_htable_size__read_mostly ; -static unsigned int nf_nat_hash_rnd__read_mostly ; +static siphash_key_t nf_nat_hash_rnd__read_mostly ;struct nf_nat_lookup_hook_priv {struct nf_hook_entries __rcu *entries;
@@ -153,12 +153,22 @@ static unsigned int hash_by_src(const struct net *n, const struct nf_conntrack_tuple *tuple) { unsigned int hash; +struct { +struct nf_conntrack_man src; +u32 net_mix; +u32 protonum; + }__aligned(SIPHASH_ALIGNMENT) combined;get_random_once (&nf_nat_hash_rnd,sizeof (nf_nat_hash_rnd)); +memset (&combined,0 ,sizeof (combined)); +/* Original src, to ensure we map it consistently if poss. */ - hash =jhash2 ((u32 *)&tuple->src,sizeof (tuple->src) /sizeof (u32), - tuple->dst.protonum ^ nf_nat_hash_rnd ^net_hash_mix (n)); + combined.src = tuple->src; + combined.net_mix =net_hash_mix (n); + combined.protonum = tuple->dst.protonum; + + hash =siphash (&combined,sizeof (combined), &nf_nat_hash_rnd);return reciprocal_scale (hash, nf_nat_htable_size); }
Signature · 1 line
2.31.1
Florian Westphal
maintainerPATCH 3/3 netfilter: conntrack: refuse insertion if chain has grown too large
In reply to Florian WestphalAssuming the old default size of 16536 buckets and max hash occupancy of 64k, this results in 128k insertions (origin+reply), so ~8 entries per chain on average.
The revised settings in the earlier change result in about two entries per bucket on average.
This enforces a hard-limit of 64.
This is not tunable at the moment, but its possible to either increase nf_conntrack_buckets or decrease nf_conntrack_max to reduce average lengths.
A new stat counter for this gets exported both via old /proc interface and ctnetlink.
--- include/linux/netfilter/nf_conntrack_common.h | 1 + .../linux/netfilter/nfnetlink_conntrack.h | 1 + net/netfilter/nf_conntrack_core.c | 42 +++++++++++++++---- net/netfilter/nf_conntrack_netlink.c | 4 +- net/netfilter/nf_conntrack_standalone.c | 4 +- 5 files changed, 42 insertions(+), 10 deletions(-)
include/linux/netfilter/nf_conntrack_common.h +1 -0
@@ -18,6 +18,7 @@ struct ip_conntrack_stat {unsigned int expect_create;unsigned int expect_delete;unsigned int search_restart; +unsigned int chaintoolong; };#define NFCT_INFOMASK 7UL
include/uapi/linux/netfilter/nfnetlink_conntrack.h +1 -0
@@ -257,6 +257,7 @@ enum ctattr_stats_cpu {CTA_STATS_ERROR ,CTA_STATS_SEARCH_RESTART ,CTA_STATS_CLASH_RESOLVE , +CTA_STATS_CHAIN_TOOLONG , __CTA_STATS_MAX, };#define CTA_STATS_MAX (__CTA_STATS_MAX - 1)
net/netfilter/nf_conntrack_core.c +35 -7
102 lines not on this page · raw message
net/netfilter/nf_conntrack_netlink.c +3 -1
@@ -2484,7 +2484,9 @@ ctnetlink_ct_stat_cpu_fill_info(struct sk_buff *skb, u32 portid, u32 seq,nla_put_be32 (skb,CTA_STATS_SEARCH_RESTART ,htonl (st->search_restart)) ||nla_put_be32 (skb,CTA_STATS_CLASH_RESOLVE , - htonl(st->clash_resolve))) +htonl (st->clash_resolve)) || +nla_put_be32 (skb,CTA_STATS_CHAIN_TOOLONG , + htonl(st->chaintoolong))) goto nla_put_failure;nlmsg_end (skb, nlh);
net/netfilter/nf_conntrack_standalone.c +2 -2
@@ -429,7 +429,7 @@ static int ct_cpu_seq_show(struct seq_file *seq, void *v)unsigned int nr_conntracks;if (v ==SEQ_START_TOKEN ) { -seq_puts (seq,"entries clashres found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n" ); +seq_puts (seq,"entries clashres found new invalid ignore delete chainlength insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n" );return 0 ; }
@@ -444,7 +444,7 @@ static int ct_cpu_seq_show(struct seq_file *seq, void *v) st->invalid,0 ,0 , -0 , + st->chaintoolong, st->insert, st->insert_failed, st->drop,
Signature · 1 line
2.31.1
Florian Westphal
maintainerPATCH 3/3 netfilter: refuse insertion if chain has grown too large
In reply to Florian WestphalAlso add a stat counter for this that gets exported both via old /proc interface and ctnetlink.
Assuming the old default size of 16536 buckets and max hash occupancy of 64k, this results in 128k insertions (origin+reply), so ~8 entries per chain on average.
The revised settings in this series will result in about two entries per bucket on average.
This allows a hard-limit ceiling of 64.
This is not tunable at the moment, but its possible to either increase nf_conntrack_buckets or decrease nf_conntrack_max to reduce average lengths.
--- include/linux/netfilter/nf_conntrack_common.h | 1 + .../linux/netfilter/nfnetlink_conntrack.h | 1 + net/netfilter/nf_conntrack_core.c | 42 +++++++++++++++---- net/netfilter/nf_conntrack_netlink.c | 4 +- net/netfilter/nf_conntrack_standalone.c | 4 +- 5 files changed, 42 insertions(+), 10 deletions(-)
include/linux/netfilter/nf_conntrack_common.h +1 -0
@@ -18,6 +18,7 @@ struct ip_conntrack_stat {unsigned int expect_create;unsigned int expect_delete;unsigned int search_restart; +unsigned int chaintoolong; };#define NFCT_INFOMASK 7UL
include/uapi/linux/netfilter/nfnetlink_conntrack.h +1 -0
@@ -257,6 +257,7 @@ enum ctattr_stats_cpu {CTA_STATS_ERROR ,CTA_STATS_SEARCH_RESTART ,CTA_STATS_CLASH_RESOLVE , +CTA_STATS_CHAIN_TOOLONG , __CTA_STATS_MAX, };#define CTA_STATS_MAX (__CTA_STATS_MAX - 1)
net/netfilter/nf_conntrack_core.c +35 -7
102 lines not on this page · raw message
net/netfilter/nf_conntrack_netlink.c +3 -1
@@ -2484,7 +2484,9 @@ ctnetlink_ct_stat_cpu_fill_info(struct sk_buff *skb, u32 portid, u32 seq,nla_put_be32 (skb,CTA_STATS_SEARCH_RESTART ,htonl (st->search_restart)) ||nla_put_be32 (skb,CTA_STATS_CLASH_RESOLVE , - htonl(st->clash_resolve))) +htonl (st->clash_resolve)) || +nla_put_be32 (skb,CTA_STATS_CHAIN_TOOLONG , + htonl(st->chaintoolong))) goto nla_put_failure;nlmsg_end (skb, nlh);
net/netfilter/nf_conntrack_standalone.c +2 -2
@@ -429,7 +429,7 @@ static int ct_cpu_seq_show(struct seq_file *seq, void *v)unsigned int nr_conntracks;if (v ==SEQ_START_TOKEN ) { -seq_puts (seq,"entries clashres found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n" ); +seq_puts (seq,"entries clashres found new invalid ignore delete chainlength insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n" );return 0 ; }
@@ -444,7 +444,7 @@ static int ct_cpu_seq_show(struct seq_file *seq, void *v) st->invalid,0 ,0 , -0 , + st->chaintoolong, st->insert, st->insert_failed, st->drop,
Signature · 1 line
2.31.1
Quoted 8 lines · Florian Westphal
Two recent commits switched inet rt and nexthop exception hashes from jhash to siphash.
If those two spots are problematic then conntrack is affected as well, so switch voer to siphash too.
While at it, add a hard upper limit on chain lengths and reject insertion if this is hit.
Applied, thanks.
I've taken this 3/3 version:
https://patchwork.ozlabs.org/project/netfilter-devel/patch/[email protected]/