<kopijahe>
Hi, I was trying to port openwrt to this ipq4028-based router (all logs, stock and openwrt test dts, and u-boot outputs here: https://gist.github.com/kopijahe/1bed6d1d261a9e131f862aca2869af76 | pics here: https://imgur.com/a/YayazVJ ). I've found another router with almost exact dts as this one, compex wpj428, except that compex uses nand, and this uses spi-nor. So I use that for my base. The problem that the kernel is just too big (4.23 mb), I found
<kopijahe>
this thread that tells me to disable ubifs/zstd from kernel, but it's already disabled. What should I do now?
<xdarklight>
Ansuel: I've updated my ipq40xx Fritz Box 7520 (same marketing name as 7530) to the latest git master and noticed that Ethernet throughput is a lot lower with 6.1 (~15 Mbit/s) than 5.15 (as close to 1 Gbit/s between two computer as possible). Cabling etc. makes no difference, reverting e553deb4753c76498be5a9ed3bea80ff135fa13f and b5c55bb6b0acb87cae67b83e80c795a232fb50ae (kernel 6.1 switch - for testing only, this is not meant as a permanent
<xdarklight>
fix) solves the throughput issues. can you recall similar issues on other ipq40xx devices?
<xdarklight>
s/same marketing name/same hardware but different marketing name/
kopijahe has quit [Read error: Connection reset by peer]
kopijahe has joined #openwrt-devel
<kopijahe>
Can I overlap the kernel and rootfs partition to overcome the 4 mb limitation? I saw some discussion about it, but I don't know if it actually used in any device...
<dwfreed>
the issue is how the OEM flash process is going to behave
hanetzer has joined #openwrt-devel
hitech95 has joined #openwrt-devel
danitool has joined #openwrt-devel
danitool has quit [Remote host closed the connection]
<jow>
f00b4r0: i.e. should it be http centric, should it be part of a url parser, should it be part of some http client (uclient binding?)
<f00b4r0>
it's mainly the ability to (en/de)code www-urlencoded data
<f00b4r0>
string processing
<f00b4r0>
the more generic the better imho. I think you have everything laid out in lucihttp-ucode. It's just a matter of "renaming" it maybe.
<f00b4r0>
in fact I think I'm just going to depend on lucihttp-ucode for my own use case :)
<f00b4r0>
because I'm lazy
<f00b4r0>
jow: btw, how would you go about opening a PR for including a *native* package to the packages feed. i.e. a repo that is self-contained with it's own openwrt-ready makefile?
<f00b4r0>
i'm not completely clear how that would work
<jow>
f00b4r0: unfortunately that's only possible by introducing it as new feed
<f00b4r0>
arg
<f00b4r0>
i wanted to avoid splitting out the makefile so I don't have two repos to commit to whenever I e.g. move things around
<f00b4r0>
last but not least, we don't have the concept of "suggested" dependencies in our packaging system, right?
<jow>
correct. opkg has some rudimentary support for it but afair it's neitrher integrated into the build system nor hnadled on the cli level
<f00b4r0>
ok makes sense
<KanjiMonster>
the term you are looking for would be "recommends" btw. opkg does understand it, but I don't know whether opkg installs recommends by default or not, and how it displays them in the latter case
hanetzer1 has joined #openwrt-devel
<f00b4r0>
yeah it was a long shot. I also noticed that trying to install a package that "conflicts" and "provides" another one doesn't work, so I understand that opkg's dependency handling is rather limited :)
hanetzer has quit [Ping timeout: 480 seconds]
<jow>
KanjiMonster: it's also entirely possible that it's broken
<jow>
we badly need to rebase our opkg onto upstream again, but upstream opkg has grown very fat due to libarchivge and libsolv usage
<jow>
our move forward aparcar's apk work
<jow>
*or
* f00b4r0
notices that "require" and "import" aren't apparently documented in ucode's documentation pages :)
* russell--
dug an old linux computer out of the basement a couple days ago and powered it up: a pentium-s 133MHz, with 64MB of RAM, a 1GB harddisk, and a 2.4.20 kernel. most amazingly, despite being last shutdown in May 2003, the ds12b887 battery module had maintained the clock and bios settings for over 20 years, the clock was only off by about 25 minutes.
<\x>
i can attest that he is running a 100% stock, unrooted, completely clean android image! https://imgur.com/a/FNPFi6S
<\x>
nvm wrong channel sory
<rotanid>
russell--: wow!
<Habbie>
gh actions on /openwrt/packages/ appear to be stuck, or is another repo eating all the slots?
<Habbie>
ah, /openwrt/openwrt is plenty busy
swalker_ has joined #openwrt-devel
swalker has quit [Read error: No route to host]
valku has joined #openwrt-devel
SeriousD has joined #openwrt-devel
<SeriousD>
Hello together :) I'm trying to build a package(fluent-bit) with the openWRT sdk. For that I'm using a docker container( openwrt/sdk), I followed a tutorial and was able to build a hello world packet with own c code. Also I managed to compile fluent bit on my host machine. Now I want to combine that.
<PaulFertser>
SeriousD: usually it's best to use a full regular OpenWrt build system and write a Makefile for your package and then upstream it to packages feed.
<SeriousD>
Unfortunately *sometimes* I get errors in the compile step. I enter the docker container, feed and update packages and then build the fluent-bit package. I think I have some errors in my manifest but I cant figure it out. I attached all relevant files here: https://gist.github.com/SeriousD/821e63841bace89f377bcb1071be63da
<PaulFertser>
It takes care of properly setting all the flags etc.
<PaulFertser>
Your gist has no examples of errors you sometimes get.
<PaulFertser>
You shouldn't run cmake manually like that, you should instead use the common code that gets all the details right.
<PaulFertser>
Provide by include/cmake.mk
<PaulFertser>
Provided
<PaulFertser>
You shouldn't have to override Build/* sections.
<SeriousD>
commented that :) The strange part is that I got one run where it worked and I got the package but wasnt able to repoduce that.
<PaulFertser>
SeriousD: those errors are actually self-descripting. They say your Makefile is missing DEPENDS.
<PaulFertser>
SeriousD: during packages all generated ELF files are checked for dependencies on external .so libraries. And that is checked against the list of the packages to see if the needed .so files are provided.
<PaulFertser>
SeriousD: probably it gets built for your host? Because you override proper cmake things that are needed for cross-compilation.
<Ansuel>
stintel i use odhcpd + unbound from ages
<stintel>
Ansuel: I use Kea, it allows you to do a redundant setup
<Habbie>
i'm behind odhcpd+dnsdist here :)
<stintel>
with dnsmasq I abused static leases to also add DNS records for devices that don't actually use DHCP. wanted to use the Kea subnet_cmds hook but that's not open source
<stintel>
I guess I can just hardcode those in Lua for now
<SeriousD>
I tried to add the libaries on the DEPENDS but it didnt help. It gets build for an x86 system.
<SeriousD>
What do you mean by "you should instead use the common code that gets all the details right" ? Where do I find resources on that? Especally the cmake part. It felt a little wrong to just copy & paste the build instructions of the fluent-bit package in the build step and obviously that caused problems.
SeriousD has quit [Remote host closed the connection]
SeriousD has joined #openwrt-devel
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
SeriousD has quit [Remote host closed the connection]
dgcampea has quit [Remote host closed the connection]
rua has quit [Quit: Leaving.]
dgcampea has joined #openwrt-devel
rua has joined #openwrt-devel
rua has quit []
rua has joined #openwrt-devel
dgcampea has quit [Remote host closed the connection]
<hitech95>
hi, can someone give me any hint to troubleshoot why an FIT image is not correctly flashed into the ubi volumes? The process doesn't throw any error on the syslog.
minimal has quit [Remote host closed the connection]
ldir has quit [Quit: Out of Cheese Error]
Ansuel has quit [Remote host closed the connection]
dgcampea has quit [Remote host closed the connection]
djfe has quit [Remote host closed the connection]
Ansuel has joined #openwrt-devel
djfe has joined #openwrt-devel
minimal has joined #openwrt-devel
goliath has joined #openwrt-devel
<robimarko>
SeriousD: Why are you calling cmake manually in the Compile step?
<robimarko>
If its using CMake then just include cmake.mk and it should do everything that is required
cmonroe has quit [Ping timeout: 480 seconds]
<robimarko>
This makes no sense to me: PKG_BUILD_DEPENDS:=libc yaml
<micw>
this seems to affect several platforms and has a heavy impact in user experience
<micw>
from reading through the posts, I got the impression that this might be beneth the radar of core devs (all comments i have read seems to be from experienced users who tried to narrow the issue and figured out a workaround)
<micw>
is there a way (and does it make sense) to flag this issue as "important"?
<PaulFertser>
micw: this github issue mixes many things into one, it's impossible to investigate
<PaulFertser>
As many different devices are mentioned with different workarounds.
<micw>
I don't think so. The issue seems to affect multiple devices in the same way and there is currently only one working workaorund (multicast_to_anycast_all)
<micw>
There is the assumption that a wrong groupkey might be the cause - but I have no idea what that means ^^
<micw>
honestly it's a bit verbose since it documents all kind of try-and-error to narrow the issue
<micw>
Thing is, wifi seems to be seriously broken on 21.x and newer on several devices. And it is broken in a way that is almost impossible to debug (since nothing is in the logs). And it is broken in a way that can be temporarily fixed by users (by toggling wifi on/off on their device) but user experience is terrible.
<micw>
@PaulFertser, what would be required for further investigation?
<PaulFertser>
micw: some way a developer could reliably reproduce it. E.g. some popular mt76 device (and probably not low end like mt7628a) and exact settings and steps that would lead to reproduction.
<mrkiko>
micw: on what device are you experiencing issues?
<micw>
i have the issue on all wifi settings i tried (default, fast roaming on/off, almost all options in "advanced")
<PaulFertser>
micw: is it reproducible with wpa3-only?
<micw>
I just figured out the the "Multi To Unicast" on luci has no effect. it sets multicast_to_unicast = 1 but the setting had be renamed to multicast_to_unicast_all
<PaulFertser>
micw: is it the same for both mt7603 and mt7613?
<micw>
wpa3-only not tested yet. I'll do it.
<micw>
I do not understand the last question ^^
<micw>
Ah, figured it out. you mean is it on both 2g and 5g
<micw>
The problem is that the issue occurs randomly. I had it the whole day on my workstation so i got some opportunity to debug it.
<micw>
but I can confirm it for 5g
<micw>
i check with wpa3 only
<micw>
(this will interrupt my connection for a moment...)
micw has quit [Remote host closed the connection]
torv has quit [Remote host closed the connection]
micw has joined #openwrt-devel
torv has joined #openwrt-devel
<micw>
back. if you wrote something after my (this will interrupt my connection for a moment...), i have not seen it
<micw>
with wpa3-only my linux box does not connect to the AP at all. It re-asked fo wifi password but got no connection
<robimarko>
Nope, there was nothing after you lost connection
<micw>
thx
<micw>
with 2g only, i need to check. I'm now connected to 2g. need to run it for a few hours
<micw>
on 5g the issue occurs after a few minutes
<micw>
but not always. yesterday i had only a few outages on this machine
<micw>
From what I know, I think it's a 5ghz issue only. I have some 5ghz devices (laptop, phones, raspberry pi) where I had a lot of outages with this schema (connected but no traffic).
<micw>
On the other hand I have a lot 2ghz devices (mainly esp8266 IoT stuff) but I had only a few outages since the update
<micw>
that's no strong evidence, I'm sorry
<micw>
different question. When I did the first steps with fast roaming (802.11r Fast Transition) it worked for the 1st ssid on a radio only. does anyone know if that has changed meanwhile?
robimarko has quit [Remote host closed the connection]
<micw>
@PaulFertser, I got the same issue with 2.4 ghz wifi (mt7603) when I use my wife's phone. It has an ip nut shows "no internet connection". with tcpdump i see the same arp requests all the time
<micw>
after ~1-2 minutes arp seems to pass through and connectivity is there
<micw>
could also be a "similar" issue
<PaulFertser>
micw: nasty :/
<PaulFertser>
micw: why your "linux box" can't connect to wpa3-only, how come, what card is it using?
<PaulFertser>
Even brcmfmac can connect to wpa3 if you use the right firmware and iwd instead of wpa_supplicant.
<micw>
I cannot easily change this. It's the machine I earn my money with, so i cannot risk to render the wifi unusable (there's another linux issue with the ethernet in the thinkpad dock, so I cannot use ethernet reliable and depend on working wifi)
danitool has joined #openwrt-devel
<micw>
If this observation is very important, I can try to setup another device this weekend and try to reproduce with that one.
<PaulFertser>
ax200 should work with wpa3 and wpa_supplicant no problem...
<PaulFertser>
No, it's not super important.
<PaulFertser>
So is there a correlation between group key rotation and loss of connectivity?
<micw>
That's what the long issue thread assumes at one point. I have no idea what that means and how I could provide this information
<micw>
which settings in wifi are related to group key rotation?
<micw>
and can i see in the log when this happens?
<PaulFertser>
I think there was some message in the log emitted by hostapd when GTK is changed.
<PaulFertser>
But do you need that rotation at all? What happens if you disable it?
<PaulFertser>
That's another problem with that ticket, some people were mentioning that option but then some others are saying it doesn't help, and some say it does. It's a real mess.
<micw>
how do i disable the rotation?
<Mangix>
micw: try iwd
<micw>
@Mangix, context?
<PaulFertser>
Mangix: I mention iwd only because that's the only userspace that can do wpa3 with brcmfmac.
<Mangix>
intel ax200
<PaulFertser>
In client mode.
<PaulFertser>
ax200 is iwlwifi so it works just fine with regular wpa_supplicant
<Mangix>
iwd performs faster connections for me
<PaulFertser>
Might be faster, I can imagine
<micw>
option wpa_group_rekey '0', right?
<micw>
i check (will interrupt my connection again...)
<PaulFertser>
Yes
<PaulFertser>
Hm, this might be relevant (from hostapd config) #wpa_strict_rekey=1
<PaulFertser>
OpenWrt supports option wpa_strict_rekey, set it to 0 too.
micw has quit [Remote host closed the connection]
micw has joined #openwrt-devel
<micw>
with rekey set to 0 I got no ip with my client. sure this is valid?
<micw>
one sec. i switch IRC to my windows laptop which is on ethernet
micw has quit []
micw has joined #openwrt-devel
<micw>
back
<micw>
ok, i have rekey interval 0 and a connection
<micw>
I let it run to check if the issue re-occurs
<micw>
I can also try with a short rekey intervall to see if it makes things worse
<micw>
@PaulFertser, i can reconfigure all my networks with rekey intervall 0 and check if the issue re-occurs. that should give at least some evidence
soxrok2212 has quit [Read error: Connection reset by peer]
<Ansuel>
hauke thinking about it
<PaulFertser>
micw: not only that, also disable wpa strict rekey.
<hauke>
Ansuel: you posted this in the forum: "ipq40xx EA6350v3, EA8300, MR8300 and WHW01 require tweak to the uboot env on update from 22.xx to 23.xx. Refer to the Device wiki or the instruction on sysupgrade on how to do this change. Config needs to be reset on sysupgrade."
<micw>
which setting in luci is this?
<Ansuel>
i think we should create another section in the release notes
<Ansuel>
or should we put that in known issue?
<PaulFertser>
micw: no idea about LuCI but current OpenWrt has "option wpa_strict_rekey"
<micw>
@PaulFertser, alright, i set both now and check how it behaves
<hauke>
Ansuel: I added it under upgrade now
<Ansuel>
ok let me check
micw has quit [Remote host closed the connection]
micw has joined #openwrt-devel
<micw>
has anyone a working wifi roaming? i have set up dawn, and the same ssid on several APs. I have tried several settings including the recommended in DAWN setup.
<micw>
when i move my device from one room into another, i gets associated with multiple APs and has no internet for several minutes
micw has quit [Remote host closed the connection]
micw has joined #openwrt-devel
<micw>
@PaulFertser, still the issue, with the rekey settings
<PaulFertser>
micw: that's weird but an essential datapoint.
<micw>
one more finding. without multicast_to_unicast it sometimes takes ages to get a dhcp response. is that also multicast?
<micw>
with multicast_to_unicast enabled, i instantly get the dhcp response
<micw>
i take this back. the dhcp issue seems to be a different one i need to dig into
<micw>
that's strange. the router gets the dhcp request (wifi client -> ap -> router), is see it in tcpdump on ap and on router. the router answers but the answer never reaches the ap
<micw>
on ap:
<micw>
22:48:42.283622 IP 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request from 2e:ed:ee:6f:09:f7, length 304
<dwfreed>
add -e to your tcpdump's please
<dwfreed>
this is actually important
goliath has quit [Quit: SIGSEGV]
<micw>
on router:
<micw>
22:44:59.162289 IP 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request from 2e:ed:ee:6f:09:f7, length 304
<micw>
22:44:59.172788 IP 192.168.1.210.67 > 192.168.1.195.68: BOOTP/DHCP, Reply, length 302
<micw>
then after a while, it suddenly works. i have the situation quite often. a wifi client waits ages for a dhcp answer. then suddenly it works
<dwfreed>
stops working shortly after roaming?
<micw>
no, that's unrelated
<micw>
i think it's a completely separate issue
<micw>
but it seems to happen on all aps in m home and i'm quite sure it happens at the same time
<dwfreed>
is this a dsa device?
<micw>
what does it mean?
<dwfreed>
if you look at ip link, do you see lan1, lan2, lan3, lan4
<dwfreed>
(or however many lan ports your router has)
<micw>
ah, yes. all devices have dsa
<robimarko>
BTW, does it happen that after 5 minutes it starts working?
<micw>
could be 5 minutes. i wrote it 22:37 and it worked around :51
<robimarko>
Well, if it is what I suspect then it has to be exactly 300 seconds
<robimarko>
It would be great if you could dump the switch FDB and look if there is a duplicate entry
<micw>
on the router or ap?
<micw>
and how do i dump this?
<dwfreed>
'bridge fdb' should work, I think
<micw>
root@lede:~# bridge fdb
<micw>
-ash: bridge: not found
<dwfreed>
probably have to install ip-full
<micw>
on the router or ap?
<dwfreed>
I would assume the router
<micw>
installed, still not found
<micw>
"brctl showmacs br-lan" maybe?
<dwfreed>
would work
<micw>
got the effect again after i roamed with my phone
<micw>
actually there are dupplicates. but not for the mac of my phone
<micw>
wokred again after 3 minutes
<micw>
after i roam, the mac is pressent on both aps and the router. is that ok?
micw is now known as Guest3166
micw has joined #openwrt-devel
<micw>
boah. i have the impression that with 21.x more things get worse then better
<dwfreed>
21.x is EOL
<micw>
oh. meant 21.x+
<micw>
i have latest 22.x
<dwfreed>
I mean, shit has bugs
Guest3166 has quit [Remote host closed the connection]
<micw>
sure. but atm it breaks at so many points. it's quite frustrating
<micw>
@dwfreed, what was you idea what the dhcp issue could have been?
<dwfreed>
mostly that the fdb may have had the wrong port
<dwfreed>
but it also sounds like it could be that the fdb has duplicates
<micw>
where is no port shown on the list
<micw>
but i have only one port used on my router with a managed switch on it
micw has quit [Remote host closed the connection]
micw has joined #openwrt-devel
<micw>
i changed the bridge with "brctl setageing br-lan 5" on the router and the 2 aps where i'm roaming. results are simlar, it takes several minutes until i get in ip address
<micw>
but not always ^^
<micw>
oh, it seems that with low ageing value roaming suddenly works