Files
ubports_kernel_google_msm/include/linux
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
..
2010-08-24 14:57:09 -07:00
2010-08-12 06:38:24 +10:00
2010-09-23 14:33:39 -07:00
2010-08-12 08:43:30 -07:00
2010-08-12 08:43:30 -07:00
2010-08-12 08:43:30 -07:00
2010-08-18 01:09:01 -04:00
2010-09-09 18:57:23 -07:00
2010-08-12 04:14:05 +01:00
2010-09-23 14:33:39 -07:00
2010-08-18 08:35:46 -04:00
2010-08-11 09:23:22 +02:00
2010-08-11 18:21:02 +02:00
2010-09-21 18:04:47 -07:00
2010-08-29 21:57:55 -07:00
2010-08-19 17:16:23 -07:00
2010-09-30 21:16:05 -07:00
2010-08-12 11:27:58 +02:00
2010-09-08 18:16:52 -07:00
2010-09-09 18:57:24 -07:00
2010-09-09 09:05:06 -07:00
2010-08-11 00:29:47 -04:00
2010-08-21 23:40:14 -07:00
2010-09-08 18:16:55 -07:00
2010-10-05 00:29:48 -07:00
2010-09-08 15:04:10 +02:00
2010-10-20 03:02:23 -07:00
2010-09-27 20:24:54 -07:00
2010-09-24 21:27:41 -07:00
2010-08-30 13:23:33 -07:00
2010-08-18 08:35:47 -04:00
2010-08-11 08:59:02 -07:00
2010-09-13 10:26:52 +02:00