<xback>
robimarko: Ansuel: I got a question regarding DSA
<xback>
i'm playing around with a Mikrotik ipq40xx (wap ac) board, and noticed that a simple ping to ethernet ports already translates to 2.2ms of latency
<xback>
on other boards, laten is easily <1ms
<xback>
Those other boards are not DSA
<xback>
is it normal that DSA adds additional latency?
<robimarko>
Well, I guess it ads a bit of latency as you need to add/strip the tag
<Znevna>
but 2.2ms is a bit much
<robimarko>
But likely, ipq40xx ethernet and DSA drivers have room for improvement
* f00b4r0
gets bitten by the "ipset" dnsmasq configuration tab in luci that does nothing on default 22.03 (with fw4)
SpectreDev_01 has quit [Quit: Connection closed for inactivity]
<\x>
xback: what if you raise minimum cpu frequency to 200MHz?
rz has quit [Remote host closed the connection]
rz has joined #openwrt-devel
robimarko has quit [Remote host closed the connection]
robimarko has joined #openwrt-devel
<xback>
\x: i'll give it a try
Daanct12 has quit [Quit: WeeChat 4.0.4]
colo has quit [Quit: Bye.]
colo has joined #openwrt-devel
<xback>
\x: it works :-)
<xback>
when stressing the cpu to force it increasing clock freq, the latency drops to 0.6ms and stays steady
<xback>
also simply setting performance governor does the trick :-)
SpectreDev_01 has joined #openwrt-devel
<SpectreDev_01>
robimarko: if the pull request is all good can you merge it
<robimarko>
We are not at that point yet
<robimarko>
Nor can I merge it
<robimarko>
I am still waiting on explanation on the partition layout
<robimarko>
BDF needs to be merged first as well
<SpectreDev_01>
Hm I don't know who can explain it, I really don't understand what Linksys has done
torv is now known as Guest2595
torv has joined #openwrt-devel
torv has quit [Remote host closed the connection]
torv has joined #openwrt-devel
Guest2595 has quit [Remote host closed the connection]
<robimarko>
Well, in its current form it makes zero sense
<SpectreDev_01>
I don't know any other options, ubi mounts rootfs and alt_rootfs (although alt_rootfs requires ubiformat)
Danct12 has quit [Read error: Connection reset by peer]
<mrkiko>
robimarko: I am happy about it since it solves a real problem, and seems quite nice and elegant to me. However, I was thinking we where still trying to solve the problem in ath10k itself or something
<robimarko>
Well, nobody even has a clue where the issue is
<robimarko>
Its some kind of counter issue in the driver but where
goliath has quit [Quit: SIGSEGV]
<mrkiko>
robimarko: ehe, I can imagine it not being eqasy
<mrkiko>
robimarko: but as I said, this solution is nice in my perspective; also because there are other situations where the process may need some time to exit, so this solution has the advantage of being more generic
<robimarko>
Well, previous solution with just 20 loops without delay wasnt really doing anything
<robimarko>
Any modern CPU would just fly through them
<mrkiko>
robimarko: and after all, at least with SIGTERM a process can handle the signal so some time to do so is right
<mrkiko>
I was encountering these kind of failures when sysupgrading with MM running sometimes
<robimarko>
Well, MM needs some time to stop
<mrkiko>
I don't know if the failure was excactly the samer, but in the end it failed ot kill all porocesses and didn't upgrade
<mrkiko>
robimarko: yeah
xavifr has quit [Ping timeout: 480 seconds]
<SpectreDev_01>
Anything I can do about the partitions
Danct12 has joined #openwrt-devel
<SpectreDev_01>
Damn it sysupgrade from mtd24 to mtd22 causes a lock up says waiting for root device
goliath has joined #openwrt-devel
<SpectreDev_01>
robimarko: does this mean syscfg is ubifs