<\x>
dhewg: about that crash youre seeing on mt7921au with mt76, yeaaah, I had that happen too, and that crash also happens on intel systems like ivy and sandy bridge. Its really puzzling
<\x>
ethernet tx and rx, that way when youre routing / bridging two cores gets used
<\x>
blindly spreading/balancing them is worse perf
<dhewg>
yeah, but is it that much? This thing mostly does wifi, wired is used but not that relevant
<\x>
gigabit codel can be achieved
<\x>
so yeah up to ya
<dhewg>
the only wired client is an old espressobin sbc :P
<\x>
cake still wont budge its like 220 Mbps hehe
<\x>
oh yeah on mine maybe I should also split that ath10k_ahb hehe
<\x>
so like uhhh
<\x>
cpu=2 ; for irq in $(grep "ath10k_ahb" /proc/interrupts | cut -d: -f1 | sed 's/ //') ; do echo "$cpu > /proc/irq/$irq/smp_affinity" ; cpu=$((cpu >> 1)) ; [ $cpu = 0 ] && cpu=2 ; done
<\x>
seems nvme isnt used much anyway
<PaulFertser>
\x: you do not need to remove spaces with that sed there.
<\x>
oh
<\x>
ah yeah that werks hehe
<\x>
dhewg: got any other device with 6Ghz there? I do wonder how does it bench with 6GHz
cbeznea has joined #openwrt-devel
MaxSoniX has joined #openwrt-devel
<dhewg>
yeah, my rz608 arrived
<dhewg>
that's in my notebook now, and the 200mbit/s is measured on the 6g band
csrf has quit [Ping timeout: 480 seconds]
<dhewg>
it's not much more than on 5g, but there's so much traffic on 5g here that you only reach that in the middle of the night or so
<\x>
and it cant use DFS channels on 5G
<\x>
so itll be on crowded freqs
<dhewg>
exactly, for HT80 it's just channel 36 and 149 iirc
danitool has joined #openwrt-devel
Tapper has joined #openwrt-devel
csrf has joined #openwrt-devel
csrf has quit [Read error: Connection reset by peer]
bbezak has quit [Read error: Connection reset by peer]
bbezak has joined #openwrt-devel
Acinonyx has joined #openwrt-devel
Acinonyx_ has quit [Ping timeout: 480 seconds]
danitool has quit [Read error: No route to host]
danitool has joined #openwrt-devel
shoragan has quit [Read error: Network is unreachable]
shoragan has joined #openwrt-devel
robimarko has joined #openwrt-devel
<f00b4r0>
does cell_density have an effect on 5GHz as well? The help and wiki wording seem to suggest it has ties to 2.4 only
Tapper has quit [Quit: Tapper]
<\x>
yes
<\x>
it sets the minimum bitrate for clients
<f00b4r0>
ok thx
<Znevna>
what can cause broken initramfs builds? o0
EyitopeION has joined #openwrt-devel
EyitopeION has quit [Remote host closed the connection]
EyitopeION has joined #openwrt-devel
<EyitopeION>
d
<robimarko>
Znevna: Broken in what sense?
<Znevna>
rb750gr3 doesn't wanna boot it
<Znevna>
it loads it, but ends up booting whats in flash
<Znevna>
and this darn thing doesn't have serial output from routerboot >.>
<robimarko>
Maybe size has increased past some limit?
<Znevna>
oooooh xD I haven't tried that. I only added luci to it
<Znevna>
i wanted to make it easier to pinpoint the guilty commit for the current broken mt7621 without flashing every time, welp, it's gonna take a while. i just did a distclean >.>
<f00b4r0>
can you try reverting 62b72eafe49d2eecd3692691152ed86a0327fcb0 and see if it works again?
<f00b4r0>
oh wait
<f00b4r0>
rb750gr3 is mtk
<Znevna>
yes
<f00b4r0>
ah
* f00b4r0
checks something
<EyitopeION>
[1] Regards to the folks building OpenWRT. I'm porting OpenWRT to a new platform under development and have a problem. The number in square brackets indicate the sequence to read the message.
<EyitopeION>
[2] # The problem - We cannot disable our procd services at build time. - Yes, we read the manual and it reads as below. ``` The build system will attempt to “enable” and/or “disable” initscripts during package install and removal by itself (refer to default_postinst() and default_prerm() in /lib/functions.sh from package base-files – this thing is utterly undocumented, including how to AVOID the automatic behavior where unwanted)"
<f00b4r0>
Znevna: ok I see, I'll prepare a patch for you to try
<f00b4r0>
just a sec
<EyitopeION>
[4] # What we have tried which did not disable the services --Defined the `DISABLED_SERVICES` variable, because we found that `$(3)` in `prepare_rootfs` in `include/rootfs.mk` is `DISABLED_SERVICES`. -- We defined a `Package/postinst` in the makefile which explicitly disables the packages
<EyitopeION>
[5] # The question we have - "What is the appropriate method (in the package makefile) to disable services at build time on v21.02.05?"
<Znevna>
f00b4r0: I can only say so far that with a build before this commit: eaba63cddee44e0761cee1b0f82e3dd359e2dac1 (kernel: fix regression on mt7986 ) i was getting this: https://paste.debian.net/hidden/10524b5d/
<f00b4r0>
that's unrelated
<Habbie>
EyitopeIOn, have you considered using a newer version?
<Znevna>
but after that commit only icmp seems to work, everything else doesn't load, even though I can see packets coming back from the router but there's something wrong with them. webui doesn't load, ssd, nothing.
<f00b4r0>
it's completely untested so it may fail to build, but if it doesn't, I suspect it'll fix your booting problem
<Znevna>
i'll try soon as the toolchain finishes compiling, thank you
<f00b4r0>
you're welcome
<f00b4r0>
robimarko: seems we will eventually need to find a way to build an install-only trimmed down initramfs. I suspect we are hitting bootloader limits on all targets now
<EyitopeION>
Habble, our client requires v21.02.5, which I understand (I just double-checked) to be the latest OpenWRT 21 release. We are building our software stack for our client's platform.
<robimarko>
f00b4r0: unfortunately, yeah we are hitting stupid RouterBoot limits
<f00b4r0>
*nod*
<robimarko>
EyitopeIOn: 22.03 is the latest one
<f00b4r0>
i think he needs/wants 21.x
<robimarko>
Well, that is unfortunate
<EyitopeION>
robimarko, OpenWRT 21. Our requirement is to use OpenWRT 21.02.5 . I had just looked at the releases on GitHub.
<f00b4r0>
i use IB with 21.x and there it's easy to install disabled services. No idea when building from sources though.
<dhewg>
why do you want to install something disabled though? the ususal way is to avoid dependencies if it's unwanted, hence not installing it. Or packages having an 'enabled' config switch, which defaults to 0/false, so which the rc.d script is executed, no daemon or whatever actually is
<dhewg>
s/which/while/
<EyitopeION>
f00b4r0, Yes, `v21.02.5` . We build from source.
<f00b4r0>
dhewg: there are some valid reasons to do this. Usually very specific ones :)
<EyitopeION>
dhewg, The services are to be disabled because
<EyitopeION>
dhewg, It's a client requirement. They are disabled becasue there are specific instances where they are required to run. At every other time, they are turned off. It's a use case. #
<EyitopeION>
dhewg, There is the obvious option of disabling the services manually after booting the system, but that defeats the purpose of defining a `Package/postinst`.
<EyitopeION>
dhewg, Which goes back to the main question: What is the appropriate method (in the package makefile) to disable services at build time on v21.02.05?
<Znevna>
<Znevna> even the downloadable snapshot is bad <-- nope, this is fine. my fault when testing.
<dhewg>
EyitopeIOn: I don't know if there's way, but Package/postinst is to create the postinst-pkg script, not replace postinst
<dhewg>
but looking at the default one, I guess you dont package files in /etc/init.d/. instead package them in another path and then use Package/postinst to copy it over and enable it if required?
<EyitopeION>
dhewg, Yes, at build time, files are copied to `/etc/init.d/` using package makefile of the respective service. Then we define the `Package/postinst` to disable them. The alternative method you describe will not work because **all** services are enabled in `prepare_rootfs` in `include/rootfs.mk`, just after the postinst script runs. The OpenWRT documentation, as I already pointed out above, explicitly says this behaviour and how to stop it is NOT
sadwjp_ has joined #openwrt-devel
<EyitopeION>
dhewg, Hence, why I am asking how exactly it is supposed to be done here i.e. how to you tell the build system NOT to enable all services in `/etc/init.d/`?
sadwjp_ has quit []
MaxS0niX has joined #openwrt-devel
<Znevna>
f00b4r0: I've just tried booting the 22.03.2 initramfs and it works fine, so it's not a size issue, I'm doing something else wrong apparently
<f00b4r0>
Znevna: the initramfs you successfully booted was smaller than the one that failed?
<Znevna>
it was actually bigger
<f00b4r0>
ok
<f00b4r0>
wait; your backtrace shows 5.15 running
<f00b4r0>
the kernel may actually be bigger in master
<Znevna>
that's unrelated
<f00b4r0>
yes it is
MaxSoniX has quit [Ping timeout: 480 seconds]
<robimarko>
Znevna: Have you though about enabling UART in hard_config?
<Znevna>
wait, lemme explain.
<f00b4r0>
if the kernel is bigger, it may cause the boot to fail
<f00b4r0>
just try the patch and see if it makes any difference
<Znevna>
robimarko: I don't know how
<robimarko>
Its just one bit in the hard_config
<f00b4r0>
trying the patch is going to be easier...
<Znevna>
f00b4r0: I was trying to pinpoint a broken commit in 5.15 and decided to test using initramfs since I don't have to flash it. Then I noticed my builds don't boot
<Znevna>
and i've tried even clean 22.03.2 build just now, nada. (without luci).
<Znevna>
and those are on 5.10
<f00b4r0>
that's my point.
<f00b4r0>
please try the patch.
<dhewg>
EyitopeIOn: okay, so I dunno then. add a way, send a patch, backport? :)
<tmn505>
EyitopeIOn: that DISABLED_SERVICES is a IB thing only https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=71ab2c9d179652d9b528817ec04d3f1dd3b4b1df. You would need to pass that as a third parameter to prepare_rootfs when building from source.
<tmn505>
as workaround, build also IB when building from source and then reuse IB to build final images.
Lechu has quit [Ping timeout: 480 seconds]
Lechu has joined #openwrt-devel
<Znevna>
f00b4r0: same thing sadly
<f00b4r0>
damn, thanks
<f00b4r0>
then i'm afraid if you want to debug further you'll have to get serial
<Znevna>
I have serial but no output since .. routerboot. how can I flip that hard_config bit?
<Znevna>
so I have to strip the 5.15 kernel somehow
floof58 is now known as Guest2452
floof58 has joined #openwrt-devel
<robimarko>
Is that with the LZMA loader?
<Znevna>
whatever is set in master?
<Znevna>
idk :D sorry.
<robimarko>
Then without
<robimarko>
I think f004br0 suggested enabling it
<Znevna>
I'll try a 5.15 build with his patch
<Znevna>
if that's what it does
<robimarko>
Yes, its enabling LZMA loader
nixuser has quit [Ping timeout: 480 seconds]
Guest2452 has quit [Ping timeout: 480 seconds]
<\x>
robimarko: remember theres this quad spi difference thingy
<\x>
i dont know the details hehe
<stintel>
quad spi is just reading/writing on 4 pins simultaneously for increased speed, aiui
<\x>
ah
<robimarko>
There is dual, triple and quad SPI
<robimarko>
But all of the support plain old single IO SPI as well
<stintel>
and datasheets might mention it's supported, doesn't mean it actually works :P
<\x>
ah
<\x>
robimarko: oh yeah you were asking about xiaomi MT7981 some time ago, it is known in two names, AX3000 NE and WR30U, but theyre expensive and doesnt make sense
<Znevna>
so when ramips will switch to 5.15 we'll have to use an older initramfs to flash it
<G10h4ck>
hi nbd how are you?
<slh>
dhewg: I would expect 700-800 MBit/s from mt7921au (albeit on a x86_64 host in STA mode), weirdly 6 GHz only gets half of that though (on linux 6.1~rc6; windows would get roughly the same figures)
<nbd>
G10h4ck: fine, thx
<nbd>
how about you?
<G10h4ck>
I had some stomachache but now seem going good, also have been reading hostapd code, I have now some undertanding, still I need your suggestion on what should be the flow to add an AP as a station in the station list, AFAIR right now when a station try to associato to our AP an event is bubbled from the kernel to the hostapd code and the hostapd code do a few stuff, authentication etc. then add it in it's station list and call a driver specific callback (
<G10h4ck>
driver->add_station or something similar) that in case of mac80211 is NULL so does nothing
<G10h4ck>
so in case there is another AP, I guess that event is not triggered because the APwill not attempt to associate
lmore377 has joined #openwrt-devel
<G10h4ck>
so there is some useful event bubbled to hostapd, for example when a beacon from another AP is received, where we can plug our code which create a station entry or should I "scan" for available AP in another way, and then trigger the station adding code?
<nbd>
it's been a while since i looked at that part of hostapd
<nbd>
so i don't have any answers yet
<nbd>
i don't think you should scan
<nbd>
beacons should be received already
<nbd>
for coexistence purposes
<nbd>
i just don't know in which part of the code
<dhewg>
slh: my setup isn't optimal :) the 200mbits/s figure is from iperf3 with a mt7921e card connected to a mt7921u ap with a thick wall in between
<G10h4ck>
nbd: so in some part of the code hostapd should receive the beacons from other AP ?
<nbd>
yes
<G10h4ck>
so I need to keep digging into hostapd code and then come back with more questions :)