Files
linux/net/ipv4
Eric Dumazet 27b75c95f1 net: avoid RCU for NOCACHE dst
There is no point using RCU for dst we allocate for a very short time
(used once).

Change dst_release() to take DST_NOCACHE into account, but also change
skb_dst_set_noref() to force a refcount increment for such dst.

This is a _huge_ gain, because we dont waste memory to store xx thousand
of dsts. Instead of queueing them to RCU, we can free them instantly.

CPU caches can stay hot, re-using same memory blocks to hold temporary
dsts.

Note : remove unneeded smp_mb__before_atomic_dec(); in dst_release(),
since atomic_dec_return() implies a full memory barrier.

Stress test, 160.000.000 udp frames sent, IP route cache disabled
(DDOS).

Before:

real    0m38.091s
user    0m13.189s
sys     7m53.018s

After:

real	0m29.946s
user	0m12.157s
sys	7m40.605s

For reference, if IP route cache was enabled :

real	0m32.030s
user	0m10.521s
sys	8m15.243s

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2010-10-20 03:02:23 -07:00
..
2010-09-23 14:33:39 -07:00
2010-02-22 16:20:22 -08:00
2010-10-17 13:53:16 -07:00
2010-10-17 13:53:16 -07:00
2010-10-03 21:50:51 -07:00
2010-08-19 00:08:30 -07:00
2010-10-19 03:50:48 -07:00
2010-09-23 14:33:39 -07:00
2010-09-23 14:33:39 -07:00
2010-06-28 23:24:29 -07:00
2010-02-22 16:20:22 -08:00
2010-10-03 21:50:53 -07:00
2010-06-30 13:31:19 -07:00
2010-08-19 00:08:30 -07:00
2010-10-20 03:02:23 -07:00
2010-06-26 22:00:03 -07:00
…
2010-10-17 13:46:33 -07:00
…
2010-09-23 14:33:39 -07:00
2010-09-23 14:33:39 -07:00
2010-09-23 14:33:39 -07:00
2010-07-12 12:57:54 -07:00
2010-07-12 12:57:54 -07:00