diff options
| author | Eric Dumazet <eric.dumazet@gmail.com> | 2010-12-18 18:35:15 +0100 |
|---|---|---|
| committer | Pablo Neira Ayuso <pablo@netfilter.org> | 2011-01-13 12:05:12 +0100 |
| commit | 255d0dc34068a976550ce555e153c0bfcfec7cc6 (patch) | |
| tree | e936c3d55eaf144cbc4edf8f9332d8089719d0d4 /net/ipv4/netfilter | |
| parent | b017900aac4a158b9bf7ffdcb8a369a91115b3e4 (diff) | |
| download | linux-255d0dc34068a976550ce555e153c0bfcfec7cc6.tar.gz linux-255d0dc34068a976550ce555e153c0bfcfec7cc6.tar.bz2 linux-255d0dc34068a976550ce555e153c0bfcfec7cc6.zip | |
netfilter: x_table: speedup compat operations
One iptables invocation with 135000 rules takes 35 seconds of cpu time
on a recent server, using a 32bit distro and a 64bit kernel.
We eventually trigger NMI/RCU watchdog.
INFO: rcu_sched_state detected stall on CPU 3 (t=6000 jiffies)
COMPAT mode has quadratic behavior and consume 16 bytes of memory per
rule.
Switch the xt_compat algos to use an array instead of list, and use a
binary search to locate an offset in the sorted array.
This halves memory need (8 bytes per rule), and removes quadratic
behavior [ O(N*N) -> O(N*log2(N)) ]
Time of iptables goes from 35 s to 150 ms.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Diffstat (limited to 'net/ipv4/netfilter')
| -rw-r--r-- | net/ipv4/netfilter/arp_tables.c | 2 | ||||
| -rw-r--r-- | net/ipv4/netfilter/ip_tables.c | 2 |
2 files changed, 4 insertions, 0 deletions
diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c index 3fac340a28d5..47e5178b998b 100644 --- a/net/ipv4/netfilter/arp_tables.c +++ b/net/ipv4/netfilter/arp_tables.c @@ -883,6 +883,7 @@ static int compat_table_info(const struct xt_table_info *info, memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); newinfo->initial_entries = 0; loc_cpu_entry = info->entries[raw_smp_processor_id()]; + xt_compat_init_offsets(NFPROTO_ARP, info->number); xt_entry_foreach(iter, loc_cpu_entry, info->size) { ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo); if (ret != 0) @@ -1350,6 +1351,7 @@ static int translate_compat_table(const char *name, duprintf("translate_compat_table: size %u\n", info->size); j = 0; xt_compat_lock(NFPROTO_ARP); + xt_compat_init_offsets(NFPROTO_ARP, number); /* Walk through entries, checking offsets. */ xt_entry_foreach(iter0, entry0, total_size) { ret = check_compat_entry_size_and_hooks(iter0, info, &size, diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c index a846d633b3b6..c5a75d70970f 100644 --- a/net/ipv4/netfilter/ip_tables.c +++ b/net/ipv4/netfilter/ip_tables.c @@ -1080,6 +1080,7 @@ static int compat_table_info(const struct xt_table_info *info, memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); newinfo->initial_entries = 0; loc_cpu_entry = info->entries[raw_smp_processor_id()]; + xt_compat_init_offsets(AF_INET, info->number); xt_entry_foreach(iter, loc_cpu_entry, info->size) { ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo); if (ret != 0) @@ -1681,6 +1682,7 @@ translate_compat_table(struct net *net, duprintf("translate_compat_table: size %u\n", info->size); j = 0; xt_compat_lock(AF_INET); + xt_compat_init_offsets(AF_INET, number); /* Walk through entries, checking offsets. */ xt_entry_foreach(iter0, entry0, total_size) { ret = check_compat_entry_size_and_hooks(iter0, info, &size, |