* -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
@ 2013-10-22 21:41 fengguang.wu
2013-10-22 22:00 ` David Miller
0 siblings, 1 reply; 11+ messages in thread
From: fengguang.wu @ 2013-10-22 21:41 UTC (permalink / raw)
To: Eric W. Biederman; +Cc: netdev, LKML
Hi Eric,
We noticed big netperf throughput regressions
a4fe34bf902b8f709c63 2e685cad57906e19add7
------------------------ ------------------------
707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
and bisected it to
commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
Author: Eric W. Biederman <ebiederm@xmission.com>
Date: Sat Oct 19 16:26:19 2013 -0700
tcp_memcontrol: Kill struct tcp_memcontrol
Replace the pointers in struct cg_proto with actual data fields and kill
struct tcp_memcontrol as it is not fully redundant.
This removes a confusing, unnecessary layer of abstraction.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
:040000 040000 a1896af98145c8ae2765a787845c43c9700c7dc0 02c93b50f66f1d1b34983bf3cc7e9a0dcc7105dc M include
:040000 040000 ebe5d0619b54ddf730224f6581f595491eb36989 cd560b4a6e56cecac931814ba16420e167eb68f6 M mm
:040000 040000 5df01f70484e07fbf98a7d5b8e0a53270777ac3d 1f8d1b340d8810a79691777f4e3ee529027b3c9b M net
bisect run success
# bad: [aec2994e1799312822a30fefc27205e7360fe5af] Merge 'pwm/for-next' into devel-hourly-2013102222
# good: [31d141e3a666269a3b6fcccddb0351caf7454240] Linux 3.12-rc6
git bisect start 'aec2994e1799312822a30fefc27205e7360fe5af' '31d141e3a666269a3b6fcccddb0351caf7454240' '--'
# good: [ef26157747d42254453f6b3ac2bd8bd3c53339c3] batman-adv: tvlv - basic infrastructure
git bisect good ef26157747d42254453f6b3ac2bd8bd3c53339c3
# bad: [cc6a88faebab06b0323818cd102a6aae443cf34a] Merge 'netdev-next/master' into devel-hourly-2013102222
git bisect bad cc6a88faebab06b0323818cd102a6aae443cf34a
# good: [5cda73b68ebf7e08586d61e6777e64e12df23f07] Merge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next
git bisect good 5cda73b68ebf7e08586d61e6777e64e12df23f07
# good: [a8fab0744585c1ab61009bfc1a1958f28e1c864f] x86/jump_label: expect default_nop if static_key gets enabled on boot-up
git bisect good a8fab0744585c1ab61009bfc1a1958f28e1c864f
# good: [21d35d212469c3138f8916f7e47b779313d79751] net: sky2: remove unnecessary pci_set_drvdata()
git bisect good 21d35d212469c3138f8916f7e47b779313d79751
# bad: [61c1db7fae21ed33c614356a43bf6580c5e53118] ipv6: sit: add GSO/TSO support
git bisect bad 61c1db7fae21ed33c614356a43bf6580c5e53118
# good: [a4fe34bf902b8f709c635ab37f1f39de0b86cff2] tcp_memcontrol: Remove the per netns control.
git bisect good a4fe34bf902b8f709c635ab37f1f39de0b86cff2
# bad: [1b66917d6b76db0abe1a1bbf86b2517ba8b91d98] cgxb4: remove duplicate include in cxgb4.h
git bisect bad 1b66917d6b76db0abe1a1bbf86b2517ba8b91d98
# bad: [0a6fa23dcb10eeb21adfd9955f7030f952a8122d] ipv4: Use math to point per net sysctls into the appropriate struct net.
git bisect bad 0a6fa23dcb10eeb21adfd9955f7030f952a8122d
# bad: [2e685cad57906e19add7189b5ff49dfb6aaa21d3] tcp_memcontrol: Kill struct tcp_memcontrol
git bisect bad 2e685cad57906e19add7189b5ff49dfb6aaa21d3
# first bad commit: [2e685cad57906e19add7189b5ff49dfb6aaa21d3] tcp_memcontrol: Kill struct tcp_memcontrol
netperf.Throughput_Mbps
750 ++-------------------------------------------------------------------+
*...*... .*... ..*
700 ++ *...*..*...*...*...*...*...*. *...*...*...*...*..*...*. |
| |
650 ++ |
| |
600 ++ |
| |
550 ++ |
| |
500 ++ |
| |
450 ++ |
O O O O O O O O O O O O |
400 ++------O--------------O-------O-------------------------------------+
vmstat.system.in
17200 ++-----------------------------------------------------------------+
17100 ++ ..*..*. *...*.. |
| *...*. .. .. * .*...|
17000 ++ .. . + *. *
16900 ++..*.. . * + .. |
16800 *+ * + * . |
16700 ++ *.. + + * |
| . + + .. |
16600 ++ * + . |
16500 ++ * |
16400 ++ |
16300 ++ |
O O O O O O |
16200 ++ O O O O O O O O O |
16100 ++-----------------------------------------------------------------+
vmstat.system.cs
550000 ++----------------------------------------------------------------+
500000 *+..*..*.. *...*.. *... ..*..*.. .*..*...*.. *...*
| . .. . .. *. . .. . .. |
450000 ++ *...* * * * |
400000 ++ |
350000 ++ |
300000 ++ |
| |
250000 ++ |
200000 ++ |
150000 ++ |
100000 ++ |
| |
50000 O+ O O O O O O O O O O O O O O |
0 ++----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions
150000 ++----------------------------------------------------------------+
| *... ..*... |
140000 ++ + * *. * |
130000 ++ + : : : |
| ..* : : : *...*
120000 ++ .*.. ..*. : : O : .. |
110000 ++. *. : : : .* |
* * : * .. |
100000 ++ *. + : * |
90000 ++ .. + : .. |
| + : . |
80000 ++ O * * |
70000 O+ O O O O |
| O O O O O |
60000 ++-----O-------------------------O--------------O-----------------+
lock_stat.slock-AF_INET.contentions.lock_sock_nested
130000 ++----------------------------------------------------------------+
| |
120000 ++ *... ..*...* |
110000 ++ + * *. : |
| + + : O: *...*
100000 ++ .*.. .*...* + : : .. |
|.. .. + : : .* |
90000 *+ * * : * .. |
| *. + : * |
80000 ++ .. + : .. |
70000 ++ + : . |
| O O O * * |
60000 O+ O O O |
| O O O O O O O |
50000 ++----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions.tcp_v4_rcv
150000 ++----------------------------------------------------------------+
| *... |
140000 ++ + * *...*...* |
130000 ++ + : : : |
| ..* : : : *...*
120000 ++ .*.. ..*. : : O : .. |
110000 ++. *. : : : .* |
* * : .. |
100000 ++ *. * * |
90000 ++ .. .. + .. |
| . + . |
80000 ++ O * * |
70000 ++ O O O O |
O O O O O O |
60000 ++-----O-------------------------O--------------O-----------------+
lock_stat.slock-AF_INET/1.contentions
50000 ++-----------------------------------------------------------------+
| *.. ..*
45000 ++ + *...*.. *. |
40000 ++ + *. + *.. .. |
| .*.. + .. + . ..* |
35000 ++. *... ..* + * *. *. |
* *. * : : .. .. |
30000 ++ : : . |
| : : * |
25000 ++ : : |
20000 ++ * |
| O |
15000 ++ O O |
O O O O O O O O O O O O |
10000 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET/1.contentions.tcp_v4_rcv
50000 ++-----------------------------------------------------------------+
| *.. ..*
45000 ++ + *...*.. *. |
40000 ++ + *. + *.. .. |
| .*.. + .. + . ..* |
35000 ++. *... ..* + * *. *. |
* *. * : : .. .. |
30000 ++ : : . |
| : : * |
25000 ++ : : |
20000 ++ * |
| O |
15000 ++ O O |
O O O O O O O O O O O O |
10000 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET/1.contentions.release_sock
35000 ++-----------------------------------------------------------------*
| *.. *. |
30000 ++ + *...*..*. .. |
| + *.. .. .. ..* |
| .*.. + . . *. *. |
25000 ++. *...*...* * * : .. .. |
* + : . |
20000 ++ + : * |
| + : |
15000 ++ * |
| |
| O |
10000 O+ O O O O O O |
| O O O O O O O |
5000 ++-----------------------------------------------------------------+
lock_stat.&rq->lock.contentions
45000 ++-----------------------------------------------------------------+
| .*.. *... |
40000 ++ .. *.. .. *.. |
35000 ++ .*.. ..* . . *. .*...*
|.. *...*. * .. .*. |
30000 *+ .. |
25000 ++ *.. *... ..* |
| . .. *. |
20000 ++ * |
15000 ++ |
| |
10000 ++ |
5000 O+ O O O O O O O O O O O O O |
| O |
0 ++-----------------------------------------------------------------+
lock_stat.&rq->lock.contentions.try_to_wake_up
35000 ++-----------------------------------------------------------------+
| |
30000 ++ .*.. *.. |
| * .. *.. : |
25000 ++.. : * . : *. * |
|. : : *. : .. :: |
20000 *+ : : .. : .* : : |
| : : : *... .. + : : |
15000 ++ : : * * * + : :|
| *... : + .. + : *
10000 ++ * + . * |
| * |
5000 ++ O |
O O O O O O O O O O O O |
0 ++------------------------O--------------O-------------------------+
lock_stat.&rq->lock.contentions.__schedule
35000 ++-----------------------------------------------------------------+
| |
30000 ++ *.. |
| .. .*.. |
25000 ++ .* . *... .. *. |
|.. + .* *...* * : ..|
20000 *+ + .. + : |
| *...* + ..*... : *
15000 ++ + ..*. * |
| *...*..*. |
10000 ++ |
| |
5000 ++ O |
O O O O O O O O O O O O O O |
0 ++-----------------------------------------------------------------+
lock_stat.&(&base->lock)->rlock.contentions.lock_timer_base
18000 ++----------------------------*------------------------------------+
| ..*.. + *.. .*...*
16000 ++ ..*. *.. + *... .*. |
14000 ++ .*..*...*. . + * *. .. |
|.. * + : .. .* |
12000 *+ + : .. |
10000 ++ + : * |
| * |
8000 ++ |
6000 ++ |
| |
4000 ++ |
2000 ++ O |
O O O O O O O O O O O O O O |
0 ++-----------------------------------------------------------------+
lock_stat.&(&n->list_lock)->rlock.contentions
140000 ++----------------------------------------------------------------+
| O |
120000 O+ O O O O O O O |
| O O O O O O |
100000 ++ |
| |
80000 ++ |
| |
60000 ++ |
| |
40000 ++ |
| |
20000 ++ |
| |
0 *+--*--*---*---*--*---*---*--*---*---*--*---*---*--*---*---*--*---*
lock_stat.&(&n->list_lock)->rlock.contentions.get_partial_node
250000 ++----------------------------------------------------------------+
| |
O O |
200000 ++ O O O O O O O O O O O O |
| O |
| |
150000 ++ |
| |
100000 ++ |
| |
| |
50000 ++ |
| |
| |
0 *+--*--*---*---*--*---*---*--*---*---*--*---*---*--*---*---*--*---*
lock_stat.&(&n->list_lock)->rlock.contentions.unfreeze_partials
45000 ++-----------------------------------------------------------------+
O O |
40000 ++ O O O O |
35000 ++ O O O O O O O O |
| O |
30000 ++ |
25000 ++ |
| |
20000 ++ |
15000 ++ |
| |
10000 ++ |
5000 ++ |
| |
0 *+--*--*---*---*---*--*---*---*---*--*---*---*--*---*---*---*--*---*
iostat.cpu.user
1.8 ++-------------------------------------------------------------------+
*...*...*. *...*. *...*.. |
1.6 ++ .. .. .. .. *...*.. .*...*...* *...*
| . . . .. + .. |
| *..* * * + . |
1.4 ++ * |
| |
1.2 ++ |
| |
1 ++ |
| |
| |
0.8 ++ O O O O |
O O O O O O O O O O O |
0.6 ++-------------------------------------------------------------------+
iostat.cpu.system
96.6 ++------------------------------------------------------------------+
| O O O O O O O |
96.4 O+ O O O O O O |
| |
| |
96.2 ++ |
| |
96 ++ |
| |
95.8 ++ O |
| ..*. *. *.. *.. |
| *. .. + .. .. . .. . ..*
95.6 ++ .. + ..*...* *... ..* *. |
| ..* *...* *...*. *. |
95.4 *+--*---------------------------------------------------------------+
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-22 21:41 -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol" fengguang.wu
@ 2013-10-22 22:00 ` David Miller
2013-10-23 4:38 ` Eric W. Biederman
0 siblings, 1 reply; 11+ messages in thread
From: David Miller @ 2013-10-22 22:00 UTC (permalink / raw)
To: fengguang.wu; +Cc: ebiederm, netdev, linux-kernel
From: fengguang.wu@intel.com
Date: Tue, 22 Oct 2013 22:41:29 +0100
> We noticed big netperf throughput regressions
>
> a4fe34bf902b8f709c63 2e685cad57906e19add7
> ------------------------ ------------------------
> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
> 3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
>
> and bisected it to
>
> commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
> Author: Eric W. Biederman <ebiederm@xmission.com>
> Date: Sat Oct 19 16:26:19 2013 -0700
>
> tcp_memcontrol: Kill struct tcp_memcontrol
Eric please look into this, I'd rather have a fix to apply than revert your
work.
Thanks.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-22 22:00 ` David Miller
@ 2013-10-23 4:38 ` Eric W. Biederman
[not found] ` <20131023061019.GA15698@localhost>
2013-10-23 12:25 ` Christoph Paasch
0 siblings, 2 replies; 11+ messages in thread
From: Eric W. Biederman @ 2013-10-23 4:38 UTC (permalink / raw)
To: David Miller; +Cc: fengguang.wu, netdev, linux-kernel
David Miller <davem@davemloft.net> writes:
> From: fengguang.wu@intel.com
> Date: Tue, 22 Oct 2013 22:41:29 +0100
>
>> We noticed big netperf throughput regressions
>>
>> a4fe34bf902b8f709c63 2e685cad57906e19add7
>> ------------------------ ------------------------
>> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
>> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
>> 3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
>>
>> and bisected it to
>>
>> commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
>> Author: Eric W. Biederman <ebiederm@xmission.com>
>> Date: Sat Oct 19 16:26:19 2013 -0700
>>
>> tcp_memcontrol: Kill struct tcp_memcontrol
>
> Eric please look into this, I'd rather have a fix to apply than revert your
> work.
Will do I expect some ordering changed, and that changed the cache line
behavior.
If I can't find anything we can revert this one particular patch without
affecting anything else, but it would be nice to keep the data structure
smaller.
Fengguag what would I need to do to reproduce this?
Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
[not found] ` <20131023061019.GA15698@localhost>
@ 2013-10-23 9:43 ` Eric W. Biederman
2013-10-23 11:46 ` Fengguang Wu
0 siblings, 1 reply; 11+ messages in thread
From: Eric W. Biederman @ 2013-10-23 9:43 UTC (permalink / raw)
To: Fengguang Wu; +Cc: David Miller, netdev, linux-kernel
Fengguang Wu <fengguang.wu@intel.com> writes:
> On Tue, Oct 22, 2013 at 09:38:10PM -0700, Eric W. Biederman wrote:
>> David Miller <davem@davemloft.net> writes:
>>
>> > From: fengguang.wu@intel.com
>> > Date: Tue, 22 Oct 2013 22:41:29 +0100
>> >
>> >> We noticed big netperf throughput regressions
>> >>
>> >> a4fe34bf902b8f709c63 2e685cad57906e19add7
>> >> ------------------------ ------------------------
>> >> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
>> >> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
>> >> 3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
>> >>
>> >> and bisected it to
>> >>
>> >> commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
>> >> Author: Eric W. Biederman <ebiederm@xmission.com>
>> >> Date: Sat Oct 19 16:26:19 2013 -0700
>> >>
>> >> tcp_memcontrol: Kill struct tcp_memcontrol
>> >
>> > Eric please look into this, I'd rather have a fix to apply than revert your
>> > work.
>>
>> Will do I expect some ordering changed, and that changed the cache line
>> behavior.
>>
>> If I can't find anything we can revert this one particular patch without
>> affecting anything else, but it would be nice to keep the data structure
>> smaller.
>>
>> Fengguag what would I need to do to reproduce this?
>
> Eric, attached is the kernel config.
>
> We used these commands in the test:
>
> netserver
> netperf -t TCP_STREAM -c -C -l 120 # repeat 64 times and get average
>
> btw, we've got more complete change set (attached) and also noticed
> performance increase in the TCP_SENDFILE case:
>
> a4fe34bf902b8f709c63 2e685cad57906e19add7
> ------------------------ ------------------------
> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
> 2572.20 -17.7% 2116.20 lkp-sb03/micro/netperf/120s-200%-TCP_MAERTS
> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
> 1006.60 -54.4% 459.40 lkp-sbx04/micro/netperf/120s-200%-TCP_STREAM
> 3278.60 -25.2% 2453.80 lkp-t410/micro/netperf/120s-200%-TCP_MAERTS
> 1902.80 +21.7% 2315.00 lkp-t410/micro/netperf/120s-200%-TCP_SENDFILE
> 3345.40 -26.7% 2451.00 lkp-t410/micro/netperf/120s-200%-TCP_STREAM
> 15588.60 -20.9% 12331.40 TOTAL netperf.Throughput_Mbps
I have a second question. Do you mount the cgroup filesystem? Do you
set memory.kmem.tcp.limit_in_bytes?
If you aren't setting any memory cgroup limits or creating any groups
this change should not have had any effect whatsoever. And you haven't
mentioned it so I don't expect you are enabling the memory cgroup limits
explicitly.
If you have enabled the memory cgroups can you please describe your
configuration as that may play a significant role.
Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-23 9:43 ` Eric W. Biederman
@ 2013-10-23 11:46 ` Fengguang Wu
0 siblings, 0 replies; 11+ messages in thread
From: Fengguang Wu @ 2013-10-23 11:46 UTC (permalink / raw)
To: Eric W. Biederman; +Cc: David Miller, netdev, linux-kernel
On Wed, Oct 23, 2013 at 02:43:14AM -0700, Eric W. Biederman wrote:
> Fengguang Wu <fengguang.wu@intel.com> writes:
>
> > On Tue, Oct 22, 2013 at 09:38:10PM -0700, Eric W. Biederman wrote:
> >> David Miller <davem@davemloft.net> writes:
> >>
> >> > From: fengguang.wu@intel.com
> >> > Date: Tue, 22 Oct 2013 22:41:29 +0100
> >> >
> >> >> We noticed big netperf throughput regressions
> >> >>
> >> >> a4fe34bf902b8f709c63 2e685cad57906e19add7
> >> >> ------------------------ ------------------------
> >> >> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
> >> >> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
> >> >> 3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
> >> >>
> >> >> and bisected it to
> >> >>
> >> >> commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
> >> >> Author: Eric W. Biederman <ebiederm@xmission.com>
> >> >> Date: Sat Oct 19 16:26:19 2013 -0700
> >> >>
> >> >> tcp_memcontrol: Kill struct tcp_memcontrol
> >> >
> >> > Eric please look into this, I'd rather have a fix to apply than revert your
> >> > work.
> >>
> >> Will do I expect some ordering changed, and that changed the cache line
> >> behavior.
> >>
> >> If I can't find anything we can revert this one particular patch without
> >> affecting anything else, but it would be nice to keep the data structure
> >> smaller.
> >>
> >> Fengguag what would I need to do to reproduce this?
> >
> > Eric, attached is the kernel config.
> >
> > We used these commands in the test:
> >
> > netserver
> > netperf -t TCP_STREAM -c -C -l 120 # repeat 64 times and get average
Sorry it's not about repeating, but running 64 netperf in parallel.
The number 64 is 2 times the number of logical CPUs.
> > btw, we've got more complete change set (attached) and also noticed
> > performance increase in the TCP_SENDFILE case:
> >
> > a4fe34bf902b8f709c63 2e685cad57906e19add7
> > ------------------------ ------------------------
> > 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
> > 2572.20 -17.7% 2116.20 lkp-sb03/micro/netperf/120s-200%-TCP_MAERTS
> > 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
> > 1006.60 -54.4% 459.40 lkp-sbx04/micro/netperf/120s-200%-TCP_STREAM
> > 3278.60 -25.2% 2453.80 lkp-t410/micro/netperf/120s-200%-TCP_MAERTS
> > 1902.80 +21.7% 2315.00 lkp-t410/micro/netperf/120s-200%-TCP_SENDFILE
> > 3345.40 -26.7% 2451.00 lkp-t410/micro/netperf/120s-200%-TCP_STREAM
> > 15588.60 -20.9% 12331.40 TOTAL netperf.Throughput_Mbps
>
> I have a second question. Do you mount the cgroup filesystem? Do you
> set memory.kmem.tcp.limit_in_bytes?
No I didn't mount cgroup at all.
> If you aren't setting any memory cgroup limits or creating any groups
> this change should not have had any effect whatsoever. And you haven't
> mentioned it so I don't expect you are enabling the memory cgroup limits
> explicitly.
>
> If you have enabled the memory cgroups can you please describe your
> configuration as that may play a significant role.
>
> Eric
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-23 4:38 ` Eric W. Biederman
[not found] ` <20131023061019.GA15698@localhost>
@ 2013-10-23 12:25 ` Christoph Paasch
2013-10-23 13:02 ` Eric Dumazet
2013-10-23 22:07 ` -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol" Fengguang Wu
1 sibling, 2 replies; 11+ messages in thread
From: Christoph Paasch @ 2013-10-23 12:25 UTC (permalink / raw)
To: Eric W. Biederman; +Cc: David Miller, fengguang.wu, netdev, linux-kernel
Hello,
On 22/10/13 - 21:38:10, Eric W. Biederman wrote:
> David Miller <davem@davemloft.net> writes:
> > From: fengguang.wu@intel.com
> > Date: Tue, 22 Oct 2013 22:41:29 +0100
> >
> >> We noticed big netperf throughput regressions
> >>
> >> a4fe34bf902b8f709c63 2e685cad57906e19add7
> >> ------------------------ ------------------------
> >> 707.40 -40.7% 419.60 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
> >> 2775.60 -23.7% 2116.40 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
> >> 3483.00 -27.2% 2536.00 TOTAL netperf.Throughput_Mbps
> >>
> >> and bisected it to
> >>
> >> commit 2e685cad57906e19add7189b5ff49dfb6aaa21d3
> >> Author: Eric W. Biederman <ebiederm@xmission.com>
> >> Date: Sat Oct 19 16:26:19 2013 -0700
> >>
> >> tcp_memcontrol: Kill struct tcp_memcontrol
> >
> > Eric please look into this, I'd rather have a fix to apply than revert your
> > work.
>
> Will do I expect some ordering changed, and that changed the cache line
> behavior.
may it be the below?
Cheers,
Christoph
----
From: Christoph Paasch <christoph.paasch@uclouvain.be>
Subject: [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure
2e685cad57 (tcp_memcontrol: Kill struct tcp_memcontrol) falsly modified
the access to memory_pressure of sk->sk_prot->memory_pressure. The patch
did modify the memory_pressure-field of struct cg_proto, but not the one
of struct proto.
So, the access to sk_prot->memory_pressure should not be changed.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
---
include/net/sock.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index c93542f..e3a18ff 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1137,7 +1137,7 @@ static inline bool sk_under_memory_pressure(const struct sock *sk)
if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
return !!sk->sk_cgrp->memory_pressure;
- return !!sk->sk_prot->memory_pressure;
+ return !!*sk->sk_prot->memory_pressure;
}
static inline void sk_leave_memory_pressure(struct sock *sk)
--
1.8.3.2
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-23 12:25 ` Christoph Paasch
@ 2013-10-23 13:02 ` Eric Dumazet
2013-10-23 19:55 ` [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure Eric W. Biederman
2013-10-23 19:58 ` Eric W. Biederman
2013-10-23 22:07 ` -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol" Fengguang Wu
1 sibling, 2 replies; 11+ messages in thread
From: Eric Dumazet @ 2013-10-23 13:02 UTC (permalink / raw)
To: Christoph Paasch
Cc: Eric W. Biederman, David Miller, fengguang.wu, netdev, linux-kernel
On Wed, 2013-10-23 at 14:25 +0200, Christoph Paasch wrote:
> may it be the below?
>
>
> Cheers,
> Christoph
>
> ----
> From: Christoph Paasch <christoph.paasch@uclouvain.be>
> Subject: [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure
>
> 2e685cad57 (tcp_memcontrol: Kill struct tcp_memcontrol) falsly modified
> the access to memory_pressure of sk->sk_prot->memory_pressure. The patch
> did modify the memory_pressure-field of struct cg_proto, but not the one
> of struct proto.
>
> So, the access to sk_prot->memory_pressure should not be changed.
>
> Reported-by: Fengguang Wu <fengguang.wu@intel.com>
> Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
> ---
> include/net/sock.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/include/net/sock.h b/include/net/sock.h
> index c93542f..e3a18ff 100644
> --- a/include/net/sock.h
> +++ b/include/net/sock.h
> @@ -1137,7 +1137,7 @@ static inline bool sk_under_memory_pressure(const struct sock *sk)
> if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
> return !!sk->sk_cgrp->memory_pressure;
>
> - return !!sk->sk_prot->memory_pressure;
> + return !!*sk->sk_prot->memory_pressure;
> }
>
> static inline void sk_leave_memory_pressure(struct sock *sk)
Nice catch, thanks !
Acked-by: Eric Dumazet <edumazet@google.com>
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure
2013-10-23 13:02 ` Eric Dumazet
@ 2013-10-23 19:55 ` Eric W. Biederman
2013-10-23 20:15 ` David Miller
2013-10-23 19:58 ` Eric W. Biederman
1 sibling, 1 reply; 11+ messages in thread
From: Eric W. Biederman @ 2013-10-23 19:55 UTC (permalink / raw)
To: Eric Dumazet
Cc: Christoph Paasch, fengguang.wu, netdev, linux-kernel, Eric Dumazet
From: Christoph Paasch <christoph.paasch@uclouvain.be>
Date: Wed, 23 Oct 2013 12:49:21 -0700
2e685cad57 (tcp_memcontrol: Kill struct tcp_memcontrol) falsly modified
the access to memory_pressure of sk->sk_prot->memory_pressure. The patch
did modify the memory_pressure-field of struct cg_proto, but not the one
of struct proto.
So, the access to sk_prot->memory_pressure should not be changed.
Acked-by: Eric Dumazet <edumazet@google.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
include/net/sock.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index c93542f92420..e3a18ff0c38b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1137,7 +1137,7 @@ static inline bool sk_under_memory_pressure(const struct sock *sk)
if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
return !!sk->sk_cgrp->memory_pressure;
- return !!sk->sk_prot->memory_pressure;
+ return !!*sk->sk_prot->memory_pressure;
}
static inline void sk_leave_memory_pressure(struct sock *sk)
--
1.7.5.4
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure
2013-10-23 13:02 ` Eric Dumazet
2013-10-23 19:55 ` [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure Eric W. Biederman
@ 2013-10-23 19:58 ` Eric W. Biederman
1 sibling, 0 replies; 11+ messages in thread
From: Eric W. Biederman @ 2013-10-23 19:58 UTC (permalink / raw)
To: David Miller
Cc: Christoph Paasch, fengguang.wu, netdev, linux-kernel, Eric Dumazet
From: Christoph Paasch <christoph.paasch@uclouvain.be>
Date: Wed, 23 Oct 2013 12:49:21 -0700
2e685cad57 (tcp_memcontrol: Kill struct tcp_memcontrol) falsly modified
the access to memory_pressure of sk->sk_prot->memory_pressure. The patch
did modify the memory_pressure-field of struct cg_proto, but not the one
of struct proto.
So, the access to sk_prot->memory_pressure should not be changed.
Acked-by: Eric Dumazet <edumazet@google.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
---
Resent because I fat fingered and deleted Dave by accident.
include/net/sock.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index c93542f92420..e3a18ff0c38b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1137,7 +1137,7 @@ static inline bool sk_under_memory_pressure(const struct sock *sk)
if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
return !!sk->sk_cgrp->memory_pressure;
- return !!sk->sk_prot->memory_pressure;
+ return !!*sk->sk_prot->memory_pressure;
}
static inline void sk_leave_memory_pressure(struct sock *sk)
--
1.7.5.4
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure
2013-10-23 19:55 ` [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure Eric W. Biederman
@ 2013-10-23 20:15 ` David Miller
0 siblings, 0 replies; 11+ messages in thread
From: David Miller @ 2013-10-23 20:15 UTC (permalink / raw)
To: ebiederm
Cc: eric.dumazet, christoph.paasch, fengguang.wu, netdev, linux-kernel
From: ebiederm@xmission.com (Eric W. Biederman)
Date: Wed, 23 Oct 2013 12:55:18 -0700
> From: Christoph Paasch <christoph.paasch@uclouvain.be>
> Date: Wed, 23 Oct 2013 12:49:21 -0700
>
> 2e685cad57 (tcp_memcontrol: Kill struct tcp_memcontrol) falsly modified
> the access to memory_pressure of sk->sk_prot->memory_pressure. The patch
> did modify the memory_pressure-field of struct cg_proto, but not the one
> of struct proto.
>
> So, the access to sk_prot->memory_pressure should not be changed.
>
> Acked-by: Eric Dumazet <edumazet@google.com>
> Reported-by: Fengguang Wu <fengguang.wu@intel.com>
> Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
> Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Applied, but I replaced "Fix: " with "net: " in the commit header line.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol"
2013-10-23 12:25 ` Christoph Paasch
2013-10-23 13:02 ` Eric Dumazet
@ 2013-10-23 22:07 ` Fengguang Wu
1 sibling, 0 replies; 11+ messages in thread
From: Fengguang Wu @ 2013-10-23 22:07 UTC (permalink / raw)
To: Christoph Paasch; +Cc: Eric W. Biederman, David Miller, netdev, linux-kernel
> - return !!sk->sk_prot->memory_pressure;
> + return !!*sk->sk_prot->memory_pressure;
Good catch, Christoph! With no surprise, it restores the performance:
a4fe34bf902b8f709c63 2e685cad57906e19add7 a235435d612680e595ea
------------------------ ------------------------ ------------------------
707.40 -41.0% 417.50 -8.8% 645.00 lkp-nex04/micro/netperf/120s-200%-TCP_STREAM
2775.60 -23.7% 2116.50 +2.1% 2834.00 lkp-sb03/micro/netperf/120s-200%-TCP_STREAM
3483.00 -27.2% 2534.00 -0.1% 3479.00 TOTAL netperf.Throughput_Mbps
It's a bit late, but
Tested-by: Fengguang Wu <fengguang.wu@intel.com>
Thanks,
Fengguang
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2013-10-23 22:08 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-10-22 21:41 -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol" fengguang.wu
2013-10-22 22:00 ` David Miller
2013-10-23 4:38 ` Eric W. Biederman
[not found] ` <20131023061019.GA15698@localhost>
2013-10-23 9:43 ` Eric W. Biederman
2013-10-23 11:46 ` Fengguang Wu
2013-10-23 12:25 ` Christoph Paasch
2013-10-23 13:02 ` Eric Dumazet
2013-10-23 19:55 ` [PATCH] Fix: Dereference pointer-value of sk_prot->memory_pressure Eric W. Biederman
2013-10-23 20:15 ` David Miller
2013-10-23 19:58 ` Eric W. Biederman
2013-10-23 22:07 ` -27% netperf TCP_STREAM regression by "tcp_memcontrol: Kill struct tcp_memcontrol" Fengguang Wu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).