mmind00 changed the topic of #linux-rockchip to: Rockchip development discussion | public log at https://libera.irclog.whitequark.org/linux-rockchip
diederik has quit [Remote host closed the connection]
diederik has joined #linux-rockchip
kevery1 has joined #linux-rockchip
kevery has quit [Remote host closed the connection]
kevery1 is now known as kevery
diederik has quit [Quit: Going to see what happens IRL, see ya!]
kevery has quit [Ping timeout: 256 seconds]
diederik has joined #linux-rockchip
kevery has joined #linux-rockchip
vagrantc has quit [Quit: leaving]
stikonas has quit [Ping timeout: 255 seconds]
lxcang has joined #linux-rockchip
<lxcang> Good evening/morning, I was here earlier asking about who could offer me some help on how to crank up the throughput of ipsec on RK3568.
psydroid2 has quit [Quit: KVIrc 5.0.0 Aria http://www.kvirc.net/]
crabbedhaloablut has joined #linux-rockchip
lxcang has quit [Remote host closed the connection]
lxcang has joined #linux-rockchip
warpme has joined #linux-rockchip
cmjx has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
warpme has joined #linux-rockchip
<linkmauve> lxcang, have you profiled where it spends time? I usually use perf for that.
DarkNeutrino has joined #linux-rockchip
DarkNeutrino has quit [Changing host]
DarkNeutrino has joined #linux-rockchip
DarkNeutrino has quit [Ping timeout: 264 seconds]
paulk has quit [Read error: Connection reset by peer]
paulk-bis has joined #linux-rockchip
DarkNeutrino has joined #linux-rockchip
DarkNeutrino has quit [Changing host]
DarkNeutrino has joined #linux-rockchip
dsimic has quit [Ping timeout: 256 seconds]
dsimic has joined #linux-rockchip
mripard has quit [Ping timeout: 255 seconds]
chewitt has joined #linux-rockchip
psydroid2 has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
warpme has joined #linux-rockchip
mripard has joined #linux-rockchip
DarkNeutrino has quit [Ping timeout: 256 seconds]
paulk-bis has quit [Quit: WeeChat 3.0]
paulk has joined #linux-rockchip
Stat_headcrabed has joined #linux-rockchip
<lxcang> linkmauve: I will test now. Yesterday someone in the Linux points me to look at the outputs of top.
<lxcang> linkmauve: one also points out that I was testing via a WiFi WAN, without ipsec is 120Mbps, with ipsec is 70Mbps, so this is suspicously not even a magnitude apart.
sL1pKn07 has quit [Remote host closed the connection]
lxcang1 has joined #linux-rockchip
<phh> uh yeah it's pretty suspicious, that CPU should laugh at 120Mbps encryption
<phh> could be a cpufreq scheduler issue, have you tried performance governor just to check?
<CounterPillow> highly doubt that
<CounterPillow> it's enterprise bloat VPN software, of course it's slow, poorly adapted to the platform it runs on, and impossible to debug.
<phh> ipsec, entreprise bloat vpn? uhh... okay...
<phh> it's literally a 6?8? bytes header, then encrypted content, then hmac
<phh> and uh
<phh> that's all
lxcang has quit [Ping timeout: 255 seconds]
<phh> (ah sorry, checking wikipedia it's 8 bytes, then the hmac, then encrypted content)
<CounterPillow> I'm not getting at the protocol.
<phh> at what then? Did I miss a part where lxcang they are not using Linux mainline IPSec support?
<CounterPillow> they're using whatever "strongswan" is
<phh> strongswan handles only the security key negociation, it has nothing to do with the performance
<phh> you can kill -9 strongswan and the ipsec tunnel will happily remain alive for quite some time
<phh> err, let's say kill -STOP in case the kernel frees resources
lxcang1 is now known as lxcang
<lxcang> CounterPillow: what? ipsec is enterprise bloat software?
<lxcang> phh: cisco says an ESP packet should contain 24 bytes of bloats, so that brings a reduction of bandwidth around 10Mbps
<lxcang> phh: still much better than what am I getting
<phh> what? 24 bytes at 1500MTU is 1%, whcih makes 1Mbps for 100Mbps base
<lxcang> phh: my point is I am getting slower than theoretical, then more interesting question is how to troubleshoot
<phh> yup, hence i suggested trying changing cpufreq scheduler
<lxcang> Someone here told me they are getting 1GB/s on ipsec on the very same hardware, I am wishing to replicate this
lxcang1 has joined #linux-rockchip
lxcang has quit [Read error: Connection reset by peer]
lxcang1 has quit [Read error: Connection reset by peer]
sL1pKn07 has joined #linux-rockchip
lxcang has joined #linux-rockchip
lxcang1 has joined #linux-rockchip
lxcang has quit [Read error: Connection reset by peer]
lxcang1 is now known as lxcang
vagrantc has joined #linux-rockchip
Stat_headcrabed has quit [Quit: Stat_headcrabed]
sL1pKn07 has quit [Remote host closed the connection]
sL1pKn07 has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
warpme has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lxcang has quit [Remote host closed the connection]
warpme has joined #linux-rockchip
lxcang has joined #linux-rockchip
<lxcang> Ok finally made a way to make the secondary WAN connection to the BananaPi
<lxcang> linkmauve: ok connected on ipsec, run iperf3, and lo and behold and lol, the iperf3 is having 99.3% cpu, not anything from irq
<lxcang> linkmauve: I think the top is wrongly attributing the CPU usage to iperf3, iperf3 when running without ipsec is not saturating any core to 100%
lxcang1 has joined #linux-rockchip
<lxcang1> When running iperf3 from a downstream device, the CPU usage goes to ksoftirqd/3
lxcang has quit [Ping timeout: 264 seconds]
lxcang1 is now known as lxcang
jakllsch has quit [Ping timeout: 240 seconds]
jakllsch has joined #linux-rockchip
Perflosopher has quit [Quit: Ping timeout (120 seconds)]
Perflosopher has joined #linux-rockchip
lxcang1 has joined #linux-rockchip
lxcang has quit [Ping timeout: 256 seconds]
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lxcang1 is now known as lxcang
warpme has joined #linux-rockchip
stikonas has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
warpme has joined #linux-rockchip
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lxcang has quit [Remote host closed the connection]
lxcang has joined #linux-rockchip
lxcang1 has joined #linux-rockchip
lxcang1 has quit [Remote host closed the connection]
lxcang1 has joined #linux-rockchip
lxcang1 has quit [Remote host closed the connection]
lxcang1 has joined #linux-rockchip
lxcang has quit [Ping timeout: 264 seconds]
lxcang1 is now known as lxcang
<lxcang> Played around the encryption algos a bit, aes-128 can crank up to 288Mbps on upload and that's about it.
<lxcang> I am curious on those said you could get 1GBps = 8.0Gbps on a single core of A55, how did you do that? And whether you are using ipsec too or some other applications require encryption
lxcang1 has joined #linux-rockchip
lxcang has quit [Ping timeout: 256 seconds]
lxcang1 is now known as lxcang
<phh> lxcang: in arm64, there are AES instructions, so you pretty much have RAM speed AES encryption. limitation will actually be hmac
<lxcang> You meant the DH stuff.
bitbang has quit [Ping timeout: 245 seconds]
bitstuff has quit [Ping timeout: 260 seconds]
psydroid2 has quit [Quit: KVIrc 5.0.0 Aria http://www.kvirc.net/]
<phh> I don't? DH is only computed at connection initialisation, not once tunnel is setup
crabbedhaloablut has quit []