| 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
| ||
|
Date: Thu, 8 Nov 2018 02:13:29 +0100
From: Paweł Staszewski <pstaszewski@...are.pl>
To: Jesper Dangaard Brouer <brouer@...hat.com>
Cc: David Ahern <dsahern@...il.com>, netdev <netdev@...r.kernel.org>,
Yoel Caspersen <yoel@...knet.dk>
Subject: Re: Kernel 4.19 network performance - forwarding/routing normal users
traffic
W dniu 08.11.2018 o 01:59, Paweł Staszewski pisze:
>
>
> W dniu 05.11.2018 o 21:17, Jesper Dangaard Brouer pisze:
>> On Sun, 4 Nov 2018 01:24:03 +0100 Paweł Staszewski
>> <pstaszewski@...are.pl> wrote:
>>
>>> And today again after allpy patch for page allocator - reached again
>>> 64/64 Gbit/s
>>>
>>> with only 50-60% cpu load
>> Great.
>>
>>> today no slowpath hit for netwoking :)
>>>
>>> But again dropped pckt at 64GbitRX and 64TX ....
>>> And as it should not be pcie express limit -i think something more is
>> Well, this does sounds like a PCIe bandwidth limit to me.
>>
>> See the PCIe BW here: https://en.wikipedia.org/wiki/PCI_Express
>>
>> You likely have PCIe v3, where 1-lane have 984.6 MBytes/s or 7.87 Gbit/s
>> Thus, x16-lanes have 15.75 GBytes or 126 Gbit/s. It does say "in each
>> direction", but you are also forwarding this RX->TX on both (dual) ports
>> NIC that is sharing the same PCIe slot.
> Network controller changed from 2-port 100G connectx4 to 2 separate
> cards 100G connectx5
>
>
> PerfTop: 92239 irqs/sec kernel:99.4% exact: 0.0% [4000Hz
> cycles], (all, 56 CPUs)
> ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>
> 6.65% [kernel] [k] irq_entries_start
> 5.57% [kernel] [k] tasklet_action_common.isra.21
> 4.60% [kernel] [k] mlx5_eq_int
> 4.04% [kernel] [k] mlx5e_skb_from_cqe_mpwrq_linear
> 3.66% [kernel] [k] _raw_spin_lock_irqsave
> 3.58% [kernel] [k] mlx5e_sq_xmit
> 2.66% [kernel] [k] fib_table_lookup
> 2.52% [kernel] [k] _raw_spin_lock
> 2.51% [kernel] [k] build_skb
> 2.50% [kernel] [k] _raw_spin_lock_irq
> 2.04% [kernel] [k] try_to_wake_up
> 1.83% [kernel] [k] queued_spin_lock_slowpath
> 1.81% [kernel] [k] mlx5e_poll_tx_cq
> 1.65% [kernel] [k] do_idle
> 1.50% [kernel] [k] mlx5e_poll_rx_cq
> 1.34% [kernel] [k] __sched_text_start
> 1.32% [kernel] [k] cmd_exec
> 1.30% [kernel] [k] cmd_work_handler
> 1.16% [kernel] [k] vlan_do_receive
> 1.15% [kernel] [k] memcpy_erms
> 1.15% [kernel] [k] __dev_queue_xmit
> 1.07% [kernel] [k] mlx5_cmd_comp_handler
> 1.06% [kernel] [k] sched_ttwu_pending
> 1.00% [kernel] [k] ipt_do_table
> 0.98% [kernel] [k] ip_finish_output2
> 0.92% [kernel] [k] pfifo_fast_dequeue
> 0.88% [kernel] [k] mlx5e_handle_rx_cqe_mpwrq
> 0.78% [kernel] [k] dev_gro_receive
> 0.78% [kernel] [k] mlx5e_napi_poll
> 0.76% [kernel] [k] mlx5e_post_rx_mpwqes
> 0.70% [kernel] [k] process_one_work
> 0.67% [kernel] [k] __netif_receive_skb_core
> 0.65% [kernel] [k] __build_skb
> 0.63% [kernel] [k] llist_add_batch
> 0.62% [kernel] [k] tcp_gro_receive
> 0.60% [kernel] [k] inet_gro_receive
> 0.59% [kernel] [k] ip_route_input_rcu
> 0.59% [kernel] [k] rcu_irq_exit
> 0.56% [kernel] [k] napi_complete_done
> 0.52% [kernel] [k] kmem_cache_alloc
> 0.48% [kernel] [k] __softirqentry_text_start
> 0.48% [kernel] [k] mlx5e_xmit
> 0.47% [kernel] [k] __queue_work
> 0.46% [kernel] [k] memset_erms
> 0.46% [kernel] [k] dev_hard_start_xmit
> 0.45% [kernel] [k] insert_work
> 0.45% [kernel] [k] enqueue_task_fair
> 0.44% [kernel] [k] __wake_up_common
> 0.43% [kernel] [k] finish_task_switch
> 0.43% [kernel] [k] kmem_cache_free_bulk
> 0.42% [kernel] [k] ip_forward
> 0.42% [kernel] [k] worker_thread
> 0.41% [kernel] [k] schedule
> 0.41% [kernel] [k] _raw_spin_unlock_irqrestore
> 0.40% [kernel] [k] netif_skb_features
> 0.40% [kernel] [k] queue_work_on
> 0.40% [kernel] [k] pfifo_fast_enqueue
> 0.39% [kernel] [k] vlan_dev_hard_start_xmit
> 0.39% [kernel] [k] page_frag_free
> 0.36% [kernel] [k] swiotlb_map_page
> 0.36% [kernel] [k] update_cfs_rq_h_load
> 0.35% [kernel] [k] validate_xmit_skb.isra.142
> 0.35% [kernel] [k] dev_ifconf
> 0.35% [kernel] [k] check_preempt_curr
> 0.34% [kernel] [k] _raw_spin_trylock
> 0.34% [kernel] [k] rcu_idle_exit
> 0.33% [kernel] [k] ip_rcv_core.isra.20.constprop.25
> 0.33% [kernel] [k] __qdisc_run
> 0.33% [kernel] [k] skb_release_data
> 0.32% [kernel] [k] native_sched_clock
> 0.30% [kernel] [k] add_interrupt_randomness
> 0.29% [kernel] [k] interrupt_entry
> 0.28% [kernel] [k] skb_gro_receive
> 0.26% [kernel] [k] read_tsc
> 0.26% [kernel] [k] __get_xps_queue_idx
> 0.26% [kernel] [k] inet_gifconf
> 0.26% [kernel] [k] skb_segment
> 0.25% [kernel] [k] __tasklet_schedule_common
> 0.25% [kernel] [k] smpboot_thread_fn
> 0.23% [kernel] [k] __update_load_avg_se
> 0.22% [kernel] [k] tcp4_gro_receive
>
>
> Not much traffic now:
> bwm-ng v0.6.1 (probing every 0.500s), press 'h' for help
> input: /proc/net/dev type: rate
> | iface Rx Tx Total
> ==============================================================================
>
> enp175s0: 6.95 Gb/s 4.20 Gb/s
> 11.15 Gb/s
> enp216s0: 4.23 Gb/s 6.98 Gb/s
> 11.21 Gb/s
> ------------------------------------------------------------------------------
>
> total: 11.18 Gb/s 11.18 Gb/s
> 22.37 Gb/s
>
> bwm-ng v0.6.1 (probing every 1.000s), press 'h' for help
> input: /proc/net/dev type: rate
> | iface Rx Tx Total
> ==============================================================================
>
> enp175s0: 700264.50 P/s 923890.25 P/s 1624154.75
> P/s
> enp216s0: 932598.81 P/s 708771.50 P/s 1641370.25
> P/s
> ------------------------------------------------------------------------------
>
> total: 1632863.38 P/s 1632661.75 P/s 3265525.00
> P/s
>
>
>
>
Also is that normal that some kworker procs takes 10%+ of cpu ?
below top
2913 root 20 0 0 0 0 I 10.3 0.0 6:58.29
kworker/u112:1-
7 root 20 0 0 0 0 I 8.6 0.0 6:17.18
kworker/u112:0-
10289 root 20 0 0 0 0 I 6.6 0.0 6:33.90
kworker/u112:4-
2939 root 20 0 0 0 0 R 3.6 0.0 7:37.68
kworker/u112:2-
4557 root 20 0 0 0 0 I 1.3 0.0 0:08.82
kworker/45:4-ev
6775 root 20 0 0 0 0 I 1.3 0.0 0:26.30
kworker/50:4-ev
6833 root 20 0 0 0 0 D 1.3 0.0 0:04.96
kworker/15:0+ev
6840 root 20 0 0 0 0 I 1.3 0.0 0:09.32
kworker/55:2-ev
6874 root 20 0 0 0 0 D 1.3 0.0 0:08.51
kworker/53:0+ev
7710 root 20 0 0 0 0 I 1.3 0.0 0:07.78
kworker/14:1-ev
12075 root 20 0 0 0 0 I 1.3 0.0 1:19.22
kworker/23:3-ev
31209 root 20 0 0 0 0 I 1.3 0.0 0:07.02
kworker/20:1-ev
32351 root 20 0 0 0 0 R 1.3 0.0 0:06.99
kworker/51:2+ev
39869 root 20 0 0 0 0 D 1.3 0.0 0:06.15
kworker/42:0+ev
39959 root 20 0 0 0 0 I 1.3 0.0 0:16.23
kworker/51:1-ev
42858 root 20 0 0 0 0 I 1.3 0.0 0:47.72
kworker/27:2-ev
43281 root 20 0 0 0 0 I 1.3 0.0 0:14.99
kworker/14:4-ev
43282 root 20 0 0 0 0 I 1.3 0.0 0:13.38
kworker/16:1-ev
43389 root 20 0 0 0 0 D 1.3 0.0 0:08.92
kworker/54:2+ev
45214 root 20 0 0 0 0 I 1.3 0.0 0:05.82
kworker/55:0-ev
46894 root 20 0 0 0 0 I 1.3 0.0 0:04.11
kworker/46:1-ev
47027 root 20 0 0 0 0 D 1.3 0.0 0:03.79
kworker/47:1+ev
47129 root 20 0 0 0 0 D 1.3 0.0 0:03.15
kworker/52:0+ev
47133 root 20 0 0 0 0 I 1.3 0.0 0:03.19
kworker/49:1-ev
47179 root 20 0 0 0 0 I 1.3 0.0 0:02.83
kworker/17:3-ev
48062 root 20 0 0 0 0 I 1.3 0.0 0:02.54
kworker/44:1-ev
48158 root 20 0 0 0 0 I 1.3 0.0 0:02.17
kworker/16:2-ev
48168 root 20 0 0 0 0 I 1.3 0.0 0:02.13
kworker/27:3-ev
48247 root 20 0 0 0 0 I 1.3 0.0 0:01.83
kworker/22:0-ev
48337 root 20 0 0 0 0 I 1.3 0.0 0:01.57
kworker/15:1-ev
48345 root 20 0 0 0 0 I 1.3 0.0 0:01.49
kworker/24:3-ev
49302 root 20 0 0 0 0 I 1.3 0.0 0:00.71
kworker/54:1-ev
49366 root 20 0 0 0 0 I 1.3 0.0 0:00.38
kworker/20:3-ev
49400 root 20 0 0 0 0 I 1.3 0.0 0:00.31
kworker/26:2-ev
49430 root 20 0 0 0 0 I 1.3 0.0 0:00.21
kworker/42:2-ev
49463 root 20 0 0 0 0 D 1.3 0.0 0:00.08
kworker/50:2+ev
51698 root 20 0 0 0 0 D 1.3 0.0 0:14.85
kworker/46:2+ev
54238 root 20 0 0 0 0 I 1.3 0.0 0:23.73
kworker/52:1-ev
2507 root 20 0 0 0 0 I 1.0 0.0 0:09.60
kworker/44:2-ev
4525 root 20 0 0 0 0 I 1.0 0.0 0:08.07
kworker/26:1-ev
4556 root 20 0 0 0 0 I 1.0 0.0 0:05.15
kworker/48:0-ev
4604 root 20 0 0 0 0 I 1.0 0.0 0:10.90
kworker/19:0-ev
5789 root 20 0 0 0 0 I 1.0 0.0 0:08.24
kworker/18:0-ev
6868 root 20 0 0 0 0 I 1.0 0.0 0:09.68
kworker/47:0-ev
6900 root 20 0 0 0 0 I 1.0 0.0 0:28.83
kworker/18:1-ev
7764 root 20 0 0 0 0 I 1.0 0.0 0:03.00
kworker/49:2-ev
12045 root 20 0 0 0 0 I 1.0 0.0 1:16.98
kworker/24:2-ev
32218 root 20 0 0 0 0 I 1.0 0.0 0:04.13
kworker/45:2-ev
34082 root 20 0 0 0 0 I 1.0 0.0 0:06.29
kworker/17:1-ev
39791 root 20 0 0 0 0 I 1.0 0.0 0:19.51
kworker/21:4-ev
39973 root 20 0 0 0 0 I 1.0 0.0 0:17.12
kworker/53:2-ev
43223 root 20 0 0 0 0 I 1.0 0.0 0:07.88
kworker/25:0-ev
43295 root 20 0 0 0 0 I 1.0 0.0 0:10.89
kworker/22:4-ev
46055 root 20 0 0 0 0 I 1.0 0.0 0:04.00
kworker/21:2-ev
46077 root 20 0 0 0 0 I 1.0 0.0 0:04.62
kworker/19:1-ev
47204 root 20 0 0 0 0 I 1.0 0.0 0:03.03
kworker/25:2-ev
47989 root 20 0 0 0 0 I 1.0 0.0 0:02.65
kworker/43:1-ev
49127 root 20 0 0 0 0 I 1.0 0.0 0:01.10
kworker/48:2-ev
49317 root 20 0 0 0 0 I 1.0 0.0 0:00.56
kworker/23:1-ev
54191 root 20 0 0 0 0 R 1.0 0.0 0:30.27
kworker/43:2+ev
81 root 20 0 0 0 0 S 0.7 0.0 0:50.27
ksoftirqd/14
87 root 20 0 0 0 0 S 0.7 0.0 1:02.92
ksoftirqd/15
102 root 20 0 0 0 0 S 0.7 0.0 0:29.78
ksoftirqd/18
117 root 20 0 0 0 0 S 0.7 0.0 0:30.73
ksoftirqd/21
127 root 20 0 0 0 0 S 0.7 0.0 0:24.45
ksoftirqd/23
137 root 20 0 0 0 0 S 0.7 0.0 0:24.94
ksoftirqd/25
142 root 20 0 0 0 0 S 0.7 0.0 0:21.74
ksoftirqd/26
222 root 20 0 0 0 0 S 0.7 0.0 0:27.83
ksoftirqd/42
227 root 20 0 0 0 0 S 0.7 0.0 0:25.35
ksoftirqd/43
242 root 20 0 0 0 0 S 0.7 0.0 0:21.40
ksoftirqd/46
267 root 20 0 0 0 0 S 0.7 0.0 0:08.62
ksoftirqd/51
5174 root 20 0 0 0 0 I 0.7 0.0 5:57.10
kworker/u112:3-
>
>
>>
>>
>>> going on there - and hard to catch - cause perf top doestn chenged
>>> besides there is no queued slowpath hit now
>>>
>>> I ordered now also intel cards to compare - but 3 weeks eta
>>> Faster - cause 3 days - i will have mellanox connectx 5 - so can
>>> separate traffic to two different x16 pcie busses
>> I do think you need to separate traffic to two different x16 PCIe
>> slots. I have found that the ConnectX-5 is significantly faster
>> packet-per-sec performance than ConnectX-4, but that is not your
>> use-case (max BW). I've not tested these NICs for maximum
>> _bidirectional_ bandwidth limits, I've only made sure I can do 100G
>> unidirectional, which can hit some funny motherboard memory limits
>> (remember to equip motherboard with 4 RAM blocks for full memory BW).
>>
> Yes memory channels are separated and there are 4 modules per cpu :)
>
>
Powered by blists - more mailing lists