| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <67d634cd-cf16-df21-7b8a-5d865d95e4e6@lab.ntt.co.jp>
Date: Fri, 5 Apr 2019 16:58:58 +0900
From: Toshiaki Makita <makita.toshiaki@....ntt.co.jp>
To: Felix Fietkau <nbd@....name>,
Rafał Miłecki <zajec5@...il.com>
Cc: Toshiaki Makita <toshiaki.makita1@...il.com>,
netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
Stefano Brivio <sbrivio@...hat.com>,
Sabrina Dubroca <sd@...asysnail.net>,
David Ahern <dsahern@...il.com>, Jo-Philipp Wich <jo@...n.io>,
Koen Vandeputte <koen.vandeputte@...ntric.com>
Subject: Re: NAT performance regression caused by vlan GRO support
On 2019/04/05 16:14, Felix Fietkau wrote:
> On 2019-04-05 09:11, Rafał Miłecki wrote:
>> On 05.04.2019 07:48, Rafał Miłecki wrote:
>>> On 05.04.2019 06:26, Toshiaki Makita wrote:
>>>> My test results:
>>>>
>>>> Receiving packets from eth0.10, forwarding them to eth0.20 and applying
>>>> MASQUERADE on eth0.20, using i40e 25G NIC on kernel 4.20.13.
>>>> Disabled rxvlan by ethtool -K to exercise vlan_gro_receive().
>>>> Measured TCP throughput by netperf.
>>>>
>>>> GRO on : 17 Gbps
>>>> GRO off: 5 Gbps
>>>>
>>>> So I failed to reproduce your problem.
>>>
>>> :( Thanks for trying & checking that!
>>>
>>>
>>>> Would you check the CPU usage by "mpstat -P ALL" or similar (like "sar
>>>> -u ALL -P ALL") to check if the traffic is able to consume 100% CPU on
>>>> your machine?
>>>
>>> 1) ethtool -K eth0 gro on + iperf running (577 Mb/s)
>>> root@...nWrt:/# mpstat -P ALL 10 3
>>> Linux 5.1.0-rc3+ (OpenWrt) 03/27/19 _armv7l_ (2 CPU)
>>>
>>> 16:33:40 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:33:50 all 0.00 0.00 0.00 0.00 0.00 58.79 0.00 0.00 41.21
>>> 16:33:50 0 0.00 0.00 0.00 0.00 0.00 100.00 0.00 0.00 0.00
>>> 16:33:50 1 0.00 0.00 0.00 0.00 0.00 17.58 0.00 0.00 82.42
>>>
>>> 16:33:50 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:34:00 all 0.00 0.00 0.05 0.00 0.00 59.44 0.00 0.00 40.51
>>> 16:34:00 0 0.00 0.00 0.10 0.00 0.00 99.90 0.00 0.00 0.00
>>> 16:34:00 1 0.00 0.00 0.00 0.00 0.00 18.98 0.00 0.00 81.02
>>>
>>> 16:34:00 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:34:10 all 0.00 0.00 0.00 0.00 0.00 59.59 0.00 0.00 40.41
>>> 16:34:10 0 0.00 0.00 0.00 0.00 0.00 100.00 0.00 0.00 0.00
>>> 16:34:10 1 0.00 0.00 0.00 0.00 0.00 19.18 0.00 0.00 80.82
>>>
>>> Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> Average: all 0.00 0.00 0.02 0.00 0.00 59.27 0.00 0.00 40.71
>>> Average: 0 0.00 0.00 0.03 0.00 0.00 99.97 0.00 0.00 0.00
>>> Average: 1 0.00 0.00 0.00 0.00 0.00 18.58 0.00 0.00 81.42
>>>
>>>
>>> 2) ethtool -K eth0 gro off + iperf running (941 Mb/s)
>>> root@...nWrt:/# mpstat -P ALL 10 3
>>> Linux 5.1.0-rc3+ (OpenWrt) 03/27/19 _armv7l_ (2 CPU)
>>>
>>> 16:34:39 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:34:49 all 0.00 0.00 0.05 0.00 0.00 86.91 0.00 0.00 13.04
>>> 16:34:49 0 0.00 0.00 0.10 0.00 0.00 78.22 0.00 0.00 21.68
>>> 16:34:49 1 0.00 0.00 0.00 0.00 0.00 95.60 0.00 0.00 4.40
>>>
>>> 16:34:49 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:34:59 all 0.00 0.00 0.10 0.00 0.00 87.06 0.00 0.00 12.84
>>> 16:34:59 0 0.00 0.00 0.20 0.00 0.00 79.72 0.00 0.00 20.08
>>> 16:34:59 1 0.00 0.00 0.00 0.00 0.00 94.41 0.00 0.00 5.59
>>>
>>> 16:34:59 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:35:09 all 0.00 0.00 0.05 0.00 0.00 85.71 0.00 0.00 14.24
>>> 16:35:09 0 0.00 0.00 0.10 0.00 0.00 79.42 0.00 0.00 20.48
>>> 16:35:09 1 0.00 0.00 0.00 0.00 0.00 92.01 0.00 0.00 7.99
>>>
>>> Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> Average: all 0.00 0.00 0.07 0.00 0.00 86.56 0.00 0.00 13.37
>>> Average: 0 0.00 0.00 0.13 0.00 0.00 79.12 0.00 0.00 20.75
>>> Average: 1 0.00 0.00 0.00 0.00 0.00 94.01 0.00 0.00 5.99
>>>
>>>
>>> 3) System idle (no iperf)
>>> root@...nWrt:/# mpstat -P ALL 10 1
>>> Linux 5.1.0-rc3+ (OpenWrt) 03/27/19 _armv7l_ (2 CPU)
>>>
>>> 16:35:31 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> 16:35:41 all 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>> 16:35:41 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>> 16:35:41 1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>>
>>> Average: CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle
>>> Average: all 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>> Average: 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>> Average: 1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
>>>
>>>
>>>> If CPU is 100%, perf may help us analyze your problem. If it's
>>>> available, try running below while testing:
>>>> # perf record -a -g -- sleep 5
>>>>
>>>> And then run this after testing:
>>>> # perf report --no-child
>>>
>>> I can see my CPU 0 is fully loaded when using "gro on". I'll try perf now.
>>
>> I guess its GRO + csum_partial() to be blamed for this performance drop.
>>
>> Maybe csum_partial() is very fast on your powerful machine and few extra calls
>> don't make a difference? I can imagine it affecting much slower home router with
>> ARM cores.
> Most high performance Ethernet devices implement hardware checksum
> offload, which completely gets rid of this overhead.
> Unfortunately, the BCM53xx/47xx Ethernet MAC doesn't have this, which is
> why you're getting such crappy performance.
Hmm... now I disabled rx checksum and tried the test again, and indeed I
see csum_partial from GRO path. But I also see csum_partial even without
GRO from nf_conntrack_in -> tcp_packet -> __skb_checksum_complete.
Probably Rafał disabled nf_conntrack_checksum sysctl knob?
But anyway even with disabling rx csum offload my machine has better
performance with GRO. I'm sure in some cases GRO should be disabled, but
I guess it's difficult to determine whether we should disable GRO or not
automatically when csum offload is not available.
--
Toshiaki Makita
Powered by blists - more mailing lists