diff options
author | Eric Dumazet <eric.dumazet@gmail.com> | 2010-10-08 06:37:34 +0000 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2010-10-11 13:06:53 -0700 |
commit | fc66f95c68b6d4535a0ea2ea15d5cf626e310956 (patch) | |
tree | ac3a7f08ad741a67ff683bf93e5669ddcae95ed7 /include/net/dst_ops.h | |
parent | 0ed8ddf4045fcfcac36bad753dc4046118c603ec (diff) | |
download | op-kernel-dev-fc66f95c68b6d4535a0ea2ea15d5cf626e310956.zip op-kernel-dev-fc66f95c68b6d4535a0ea2ea15d5cf626e310956.tar.gz |
net dst: use a percpu_counter to track entries
struct dst_ops tracks number of allocated dst in an atomic_t field,
subject to high cache line contention in stress workload.
Switch to a percpu_counter, to reduce number of time we need to dirty a
central location. Place it on a separate cache line to avoid dirtying
read only fields.
Stress test :
(Sending 160.000.000 UDP frames,
IP route cache disabled, dual E5540 @2.53GHz,
32bit kernel, FIB_TRIE, SLUB/NUMA)
Before:
real 0m51.179s
user 0m15.329s
sys 10m15.942s
After:
real 0m45.570s
user 0m15.525s
sys 9m56.669s
With a small reordering of struct neighbour fields, subject of a
following patch, (to separate refcnt from other read mostly fields)
real 0m41.841s
user 0m15.261s
sys 8m45.949s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'include/net/dst_ops.h')
-rw-r--r-- | include/net/dst_ops.h | 37 |
1 files changed, 36 insertions, 1 deletions
diff --git a/include/net/dst_ops.h b/include/net/dst_ops.h index d1ff9b7..1fa5306 100644 --- a/include/net/dst_ops.h +++ b/include/net/dst_ops.h @@ -1,6 +1,7 @@ #ifndef _NET_DST_OPS_H #define _NET_DST_OPS_H #include <linux/types.h> +#include <linux/percpu_counter.h> struct dst_entry; struct kmem_cachep; @@ -22,7 +23,41 @@ struct dst_ops { void (*update_pmtu)(struct dst_entry *dst, u32 mtu); int (*local_out)(struct sk_buff *skb); - atomic_t entries; struct kmem_cache *kmem_cachep; + + struct percpu_counter pcpuc_entries ____cacheline_aligned_in_smp; }; + +static inline int dst_entries_get_fast(struct dst_ops *dst) +{ + return percpu_counter_read_positive(&dst->pcpuc_entries); +} + +static inline int dst_entries_get_slow(struct dst_ops *dst) +{ + int res; + + local_bh_disable(); + res = percpu_counter_sum_positive(&dst->pcpuc_entries); + local_bh_enable(); + return res; +} + +static inline void dst_entries_add(struct dst_ops *dst, int val) +{ + local_bh_disable(); + percpu_counter_add(&dst->pcpuc_entries, val); + local_bh_enable(); +} + +static inline int dst_entries_init(struct dst_ops *dst) +{ + return percpu_counter_init(&dst->pcpuc_entries, 0); +} + +static inline void dst_entries_destroy(struct dst_ops *dst) +{ + percpu_counter_destroy(&dst->pcpuc_entries); +} + #endif |