<mangix>
can't find bootlogs for a bunch of these routers
<mangix>
Ansuel: alright. these routers are probably all 8337
<Ansuel>
good
<mangix>
I'll be making the change.
<mangix>
Ansuel: as for led-open-drain :)
<mangix>
ah right. need rsalvaterra to test
<mangix>
also need that qca8k backport PR to get updated as I updated the bindings to match the current WIP.
Tusker has quit [Quit: Time wasted on IRC: 1 hour 56 minutes 5 seconds]
<Ansuel>
time wasted on irc LOL?
<Ansuel>
mangix: tons of changes with v5 so....
<mangix>
80+ devices?
<mangix>
pretty sure it's more than that
<mangix>
what's the impact of falling edge again?
<mangix>
I ask since this router works fine without it.
<Ansuel>
port instability leds misbheaviour
<Ansuel>
is short the signal is wrongly received in some situation
<mangix>
hmm OK
<slh>
ipq806x are 20 devices alone
<Ansuel>
it's very present onbcm53xx
<mangix>
funny how upstream really likes private structs
<Ansuel>
example?
<mangix>
I was reading the comments on v2 or v3
<mangix>
what was the reasoning behind mac6-exchange being 8337 only again?
<Ansuel>
qca8327 the bit is flagged as reserved
Tusker has joined #openwrt-devel
<mangix>
oh OK
<mangix>
makes me really curious...
<Ansuel>
same reaction that's the reason i started to check the device actually using it
<mangix>
I just added mac6-exchange and made port 6 sgmii
<mangix>
let's see if this works, LOL
<Ansuel>
LOL port 6 is rgmii ?
<mangix>
yeah
<mangix>
port0 = sgmii, port 6 = rgmii-id
<Ansuel>
rip ?
<mangix>
yep
<mangix>
didn't work
<Slimey>
heh
<Ansuel>
that's good
<Ansuel>
i mean if at the end we chose to reintroduce mac06 exchange
<mangix>
oh you're removing the binding?
<Ansuel>
we are using the implementation with port6... we don't swap we support configuration with port 6 as unique cpu port declared
<mangix>
OK. as far as this router is concerned, it works with ether port0 or port6 defined.
<mangix>
What's interesting is that qca956x is all 8337 and all single port
<Ansuel>
what a strange config
<mangix>
?
<mangix>
QCA956x is similar to QCA955x except cheaped out everywhere.
<Ansuel>
ok only thing that i'm missing is the conversion of qca8k.txt to ymal...
<Ansuel>
i think all the other changes are ""clear""
<mangix>
I'd skip that
<mangix>
or do it in a separate patch
<Ansuel>
yhea... that's not how it works with reviewers AHAHHA
<mangix>
LOL
<mangix>
I typically say "that's beyond the scope of this PR" here. Mostly works.
<Ansuel>
considering andrew want this to be not stopped... i think will post v5 and will convert to ymal to v6
<Ansuel>
we gotta get that pricey reviewed-by LOL
<Tusker>
stintel: I'm taking a break on the xtm330 and working on the t30-w now... :) still no word from WatchGuard for the source code
<mangix>
Ansuel: meh. All my patches to the kernel have been small.
hubvu_ has joined #openwrt-devel
titanous has quit [Ping timeout: 480 seconds]
Vaughn has quit [Ping timeout: 480 seconds]
hubvu has quit [Ping timeout: 480 seconds]
zx2c4 has quit [Ping timeout: 480 seconds]
hubvu_ is now known as hubvu
zx2c4 has joined #openwrt-devel
titanous has joined #openwrt-devel
<Ansuel>
v5 post... time to sleep....
Ansuel has quit [Quit: Probably my PC crashed or time to sleep.]
zx2c4 has quit [Remote host closed the connection]
hubvu has quit [Read error: Connection reset by peer]
titanous has quit [Read error: Connection reset by peer]
zx2c4 has joined #openwrt-devel
hubvu has joined #openwrt-devel
titanous has joined #openwrt-devel
Vaughn has joined #openwrt-devel
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
dangole_ has quit [Ping timeout: 480 seconds]
<hanetzer>
anyone have any experience with https://openwrt.org/toh/ubiquiti/unifi_6_lr ? it looks appealing, but I'm a bit wary of ubiquiti stuff since the one's I've encountered before don't work on standard (eg, not ubiquiti branded) PoE switches
<mangix>
Passive PoE 48 VDC, 0.5 A
<mangix>
sounds about right
<hanetzer>
'right' as in, should work with stuff properly?
dangole_ has joined #openwrt-devel
victhor has quit [Ping timeout: 480 seconds]
<mangix>
hanetzer: no, as in, ubiquity runs weird PoE
<hanetzer>
ah, yeah.
<mangix>
could work well. I have an mt7915 router and am very happy with it.
<hanetzer>
mangix: looking to upgrade our 10-15 meraki mr24's :)
<digitalcircuit>
slh: If that is the same link as in my second message, I'm glad I found the right starting point!
<digitalcircuit>
Thank you for continuing to follow my efforts, too :)
<slh>
in simple cases (no patch conflicts), you can 'just' dump https://github.com/torvalds/linux/commit/6351cac92a8c598115d74a700f1720861d57ef20.patch into target/linux/ipq806x/patches-5.10/999-6351cac92a8c598115d74a700f1720861d57ef20.patch (yes, it would actually belong into target/linux/generic/backport-5.10/, but ipq806x is applied after the generic parts, do either to get it in there for testing)
<slh>
the lazy person in me always tries this simple case first, before going all the way and rebasing the patch
<slh>
there's often a good chance for it to work (depends on how big the delta is and how much refactoring has happened in that area of the kernel)
<digitalcircuit>
slh: Ooh, good point. As I'm pretty tired right now and need to get dinner/etc (11 pm EST), I'll try this efficient way and let the build run, see what happens. I can diagnose it more tomorrow or such. Thanks! I also didn't realize that about generic vs ipq806x patching, good to know.
<slh>
it's always less work if your new patch gets applied last <-- no need to rebase patches that are applied behind it, less churn
<slh>
for a formal patch submission, you'd have to put it into the right place - but that's for later
<digitalcircuit>
slh: Makes sense! No point in formalizing everything only to find out it's not even the right patch to solve the boot issue.
<slh>
exactly, quick'n'dirty (not really dirty, but just not quite release-ready) can save a lot of time and (needless, for just testing something locally) efforts
<digitalcircuit>
slh: Hey, look at that - the build succeeded! Something I probably should've asked earlier - how can I be sure the patch was actually applied? As suggested, I saved it as "openwrt/target/linux/ipq806x/patches-5.10/999-6351cac92a8c598115d74a700f1720861d57ef20.patch"...
* digitalcircuit
is assuming it's applied, but isn't sure if there's anything he should check in case the new build does NOT work.
<digitalcircuit>
(I'm waiting to test because I should be eating and such, not fiddling with the router...)
philipp64|work_ has joined #openwrt-devel
philipp64|work has quit [Remote host closed the connection]
philipp64|work_ is now known as philipp64|work
nitroshift has joined #openwrt-devel
Tapper has joined #openwrt-devel
rmilecki has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
lmore377 has quit [Read error: Connection reset by peer]
lmore377 has joined #openwrt-devel
<nitroshift>
morning guys
lmore377 has quit [Read error: Connection reset by peer]
lmore377 has joined #openwrt-devel
<nitroshift>
anyone knows if mv88e6xxx dsa driver can balance the load between 2 cpu cores?
<mangix>
if you mean bridge the 2 CPU cores, no
lmore377 has quit [Read error: Connection reset by peer]
lmore377 has joined #openwrt-devel
goliath has joined #openwrt-devel
<nitroshift>
mangix, i mean use both cpu0 and cpu1, at the moment its irq sits only on cpu0
<mangix>
oh you're asking about using both CPU0 and 1
<mangix>
the answer is most likely no.
<nitroshift>
mangix, yeah, that's what i'm asking
<nitroshift>
the answer though is not what i expected :p
<mangix>
typically these are pegged to a single core
<mangix>
at best you can use cpu0 for ethernet and cpu1 for wifi
<nitroshift>
mangix, mwlwifi is already using both cores
<mangix>
maybe _lore_ knows more
<nitroshift>
mangix, thanks, will ping him later when he comes online
<mangix>
he's worked on the mvneta driver
<nitroshift>
by the way, is mvneta still used after the switch to dsa?
<mangix>
mvneta is an ethernet driver, so of course.
<Tapper>
Some one should start a blog naming and shaming venders for all there shitty code and buggy drivers. There would be a lot to post about and you could keep it going for years!
<Habbie>
mangix, i see #4622 depends on a few other PRs - to test, what branch or commit can I just check out and build? :)
<nitroshift>
_lore_, ping
goliath has quit [Quit: SIGSEGV]
<Tapper>
The COMMENTS section would be a blood bath. I see it all the time on blog posts about anything linux.
<jow>
hmm. Error relocating /usr/lib/lua/luci/template/parser.so: __stat_time64: symbol not found
<jow>
maybe we should consider adding abi version tracking to libc as well
<jow>
to prevent installing time64 executables on older libc's
<rsalvaterra>
Morns!
<jow>
but that ship has sailed already
<rsalvaterra>
mangix: What do you need me to test?
<jow>
would only help for future incompatible bumps
<mangix>
Habbie: just the qca8k backport one
<Habbie>
4528
<mangix>
rsalvaterra: Add qca8k,led-open-drain and see if it works
<mangix>
Although with the PR as is, it needs to be modified slightly as the wca8k backport PR needs to be updated to the latest version.
<rsalvaterra>
mangix: Hm… but weren't the LEDs already working before?
<rsalvaterra>
Or have I missed something?
<mangix>
Sure but the question is if they work with led-open-drain
goliath has joined #openwrt-devel
rua1 has joined #openwrt-devel
<rsalvaterra>
I see, thanks.
rua has quit [Remote host closed the connection]
indy_ is now known as indy
mrkiko has joined #openwrt-devel
pmelange has joined #openwrt-devel
<mrkiko>
Why aren't nathelpers shipped by default in openwrt? Are they useless nowadays or what??
pmelange1 has joined #openwrt-devel
<dwfreed>
most people don't use the protocols that there are helpers for
<dwfreed>
irc isn't plaintext anymore, so the irc helper is useless (and nobody uses dcc anymore either, which was the irc helper's only purpose)
<dwfreed>
And some things that previously needed a helper now use their own nat traversal mechanisms, like STUN and TURN
<mrkiko>
dwfreed: thanks! The only protocol that comes to mind is FTP
pmelange has quit [Read error: Connection reset by peer]
pmelange has joined #openwrt-devel
<dwfreed>
mrkiko: and that's all passive mode these days, and thus doesn't need a helper
<dwfreed>
(active mode was really not a forward-thinking idea)
pmelange1 has quit [Ping timeout: 480 seconds]
pmelange has left #openwrt-devel [#openwrt-devel]
nitroshift has quit [Remote host closed the connection]
nitroshift has joined #openwrt-devel
clayface_ has joined #openwrt-devel
clayface has quit [Ping timeout: 480 seconds]
rsalvaterra has quit [Quit: rsalvaterra]
rsalvaterra has joined #openwrt-devel
<rsalvaterra>
mrkiko: The only NAT helper I ever used was for PPTP, which nobody sane should ever use, nowadays.
Tapper has quit [Ping timeout: 480 seconds]
victhor has joined #openwrt-devel
Tusker has quit [Remote host closed the connection]
Acinonyx_ has quit [Quit: Peer reset by connector]
Tapper has joined #openwrt-devel
Acinonyx has joined #openwrt-devel
Ansuel has joined #openwrt-devel
dangole_ has joined #openwrt-devel
Dgrey has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
pmelange has joined #openwrt-devel
<fda^>
after 4 weeks uptime i can say something eats the CPU of the E8450: https://ibb.co/gTCB0f5
<fda^>
...and the firewall works not well as new ipv6 adresses are not recognozed correctly as i have after a internet-reconnect run manually: "service firewall restart" to allow opened ports again
<fda^>
does wsdd2 no longer segfaults with updated gcc11? this prevents me from updating the image
Tapper has joined #openwrt-devel
minimal has joined #openwrt-devel
dangole_ has quit [Remote host closed the connection]
dangole has joined #openwrt-devel
<Habbie>
what does "Wireless is not associated" mean?
<Habbie>
i updated from 21.02 to master without resetting, i'll try a reset now
<Ansuel>
no device connected but wifi on ?
<Habbie>
the wifi LED was off
<Habbie>
after reset, it is 'disabled' .. and also 5ghz is gone
<Habbie>
archer c7 v5
<Habbie>
enabled 2.4ghz and now that works
<Ansuel>
in that case then it's correct after a reset wifi is always off by default
<Habbie>
yes, that part is not surprising
<Habbie>
5ghz being gone is, though
<Habbie>
unless i'm missing something, that apparently broke somewhere between 21.02 and master
<Habbie>
(or i made a mistake in building my image)
<Habbie>
mm, the image without 5ghz is for the wrong target entirely :D
<Habbie>
it's a miracle anything works then!
<Habbie>
oh, no, wait, -that- image is right
Dgrey_ has joined #openwrt-devel
Dgrey has quit [Ping timeout: 480 seconds]
Dgrey_ is now known as Dgrey
nitroshift has quit [Quit: Gone that way --->]
rsalvaterra has quit [Quit: rsalvaterra]
<jow>
fda^: usually it is better to let ipv6 port opening rules target the host part of your ipv6 addresses only, this way rotating prefixes do not matter
<fda^>
jow: it seems firewall is just not reloaded if i get a new ipv6
<fda^>
or to early and the ipv6 is not assigned yet
<jow>
e.g. dest_ip ::1234:5678:abcd:0001/::ffff:ffff:ffff:ffff instead of dest_ip 2001:dead:beef:1234:1234:5678:abcd:0001
<jow>
there should be a syslog note
<jow>
reloading firewall due to if{up,update} on ...
<jow>
considered using ndp proxy mode to mirror your wan-side /64 to lan?
<fda^>
im just using ULA in my lan
<fda^>
and i have some vlans/APs so a /64 is to small
<fda^>
so this does not work after i hit on the webinterfce "restart pppoe-dsl"
<jow>
next time you observe the problem where you would need to restart the firewall manually, check cat /proc/net/nf_conntrack
<jow>
maybe there's stale conntrack entries preventing DNAT to the new addr
<jow>
or from the new addr rather
<fda^>
icat /proc/net/nf_conntrack|wc -l is currently 194
<fda^>
but most ipv4
<jow>
I mean specifically look for ESTABLISHED .+ dst=fda0...
<jow>
and see if an "echo f > /proc/net/nf_conntrack" (equivalent to conntrack -F) make things work again without a firewall restart
<jow>
if yes, then it is a stale conntrack issue
<fda^>
cat /proc/net/nf_conntrack|grep ESTA | grep ipv6 shows me 3: 1 of irc and 2 of fetchmail
<fda^>
in /etc/hotplug.d/iface/20-firewall i notced $IFUPDATE_DATA but this seems to be empty always
<jow>
that is only set on IPv6 delegated prefgix changes (deprecation, renewal, ...)
jbowen has joined #openwrt-devel
<fda^>
ah, okay...
<fda^>
btw, i have currently dual-stack and will lose seen ipv4 due to switch to CGNAT. So im using a vserver to route incomming ipv4 to my ipv6. in /etc/hotplug.d/iface/ i have a wget so the server updates the destination-ipv6
<fda>
/proc/net/nf_conntrack show ESTABLISHED only 2 unrelated
fda^ has quit [Ping timeout: 480 seconds]
fda- has joined #openwrt-devel
fda| has joined #openwrt-devel
fda^ has joined #openwrt-devel
Ansuel has quit [Read error: Connection reset by peer]
Ansuel has joined #openwrt-devel
<fda^>
firwall log shows "Flushing conntrack: 2003:db:cf ....." but only with 1 ip. maybe the whole subnet should be flushed (as im use ::face and not the assigned?)
<fda^>
or netifs is the problem? "netifd: You have delegated IPv6-prefixes but haven't assigned them to any interface. Did you forget to set option ip6assign on your lan-interfaces?"
fda has quit [Ping timeout: 480 seconds]
fda- has quit [Ping timeout: 480 seconds]
fda| has quit [Ping timeout: 480 seconds]
<mangix>
Habbie: ath10k is not selected for some reason
<Habbie>
ah, that's a good hint
<Habbie>
hopefully unrelated, is mac80211 building known to be broken on master?
<fda^>
Habbie: for me to @avm1200
<Habbie>
fda^, mac80211? ok
<fda^>
no
<fda^>
does not build
<fda^>
boolean symbol CRYPTO_LIB_ARC4 tested for 'm'? test forced to 'n'
<Habbie>
build_dir/target-mips_24kc_musl/linux-ath79_generic/backports-5.10.68-1/net/mac80211/iface.c:1609:17: error: implicit declaration of function 'kcov_remote_start_common' [-Werror=implicit-function-declaration]
<Habbie>
can get a fuller log soon
<Habbie>
currently trying some bisecting
<Habbie>
indeed, what fda^ pasted
<fda^>
maybe just "cc1: some warnings being treated as errors" ?
<Habbie>
nbd, so 978e822db354daf974811f2717c6013fa3eb8079 is suspect
<nbd>
i think it's not about kcov being enabled or disabled,
<nbd>
but somehow kcov.h gets implicitly included for some reason on my build, whereas it doesn't in your build
<nbd>
i will remove the kcov lines from the patch
<Habbie>
ack
<Habbie>
thanks :)
<hurricos>
chunkeey: I just had chunkeey's staging tree @ a704ca791 have a build failure when run -j24, but success with -j 1
<Habbie>
nbd, i'll stop my bisect then :)
<Habbie>
nbd, happy to test your PR when it's in, let me know
<fda^>
hurricos: this happend to me also from time to time
<hurricos>
I'm sure this happens often and the only reason it wasn't caught elsewhere was because this tree is not built by nearly as many people. But I will test the MX60 with the result later today
<hurricos>
fda^: Yeah, it's weird but I realize after some thought that it must just happen sometimes.
<fda^>
somewhere a dependency is missing and the build order is then not okay. with 1 paralled this is not a problem
<Habbie>
nbd, patch applies
fda has joined #openwrt-devel
<fda>
oh great wsdd2 does not crash on E8450 as it did on avm7320
<nbd>
Habbie: does it build?
<fda>
but after update dnsmasq fails as it can not read anymore files from /etc/dnsmasq/
<Habbie>
nbd, finding out :)
<jow>
fda: yeah, incompativle musl bump
<jow>
fda: stuff uses time64 syscalls now (e.g. stat()) and the installed older libc.so does not know that
<fda>
ah, the thing with rrdtool...
<Habbie>
nbd, i changed kernel settings meanwhile so it'll be a few more minutes
<Habbie>
nbd, (not kcov settings)
<jow>
rrdtool is just another victim
<Habbie>
nbd, after that i can undo your patch to quickly see if it breaks again
fda^ has quit [Ping timeout: 480 seconds]
<jow>
its an unclean library update without dealting with the ABI version fallout
<fda>
as the E8450 has aarch64 im wondering if my rrd files still working...
<fda>
seems so
<fda>
how can i allow dnsmasq to acces a directory?
<jow>
ah, the newly enabled jhail crap might be an issue too
<fda>
sure...
<fda>
but i forgot the variable name
<fda>
ah, no variable but commands: procd_add_jail_*
<tmn505>
mangix: pong
<jow>
so whenever you reference a new directory in dnsmasq.conf or elsewhere you also have to amend the init script?
<jow>
a truly good implementation would gather the required paths directly from the config
<fda>
i just added "procd_add_jail_mount /etc/dnsmasq/" to the init script, and now it works again. the procd my be good for security but annoying for usage...
<jow>
s/procd/ujail/
<fda>
i dont know exactily the differenced, but it seems to be ujail is a part of procd
<jow>
yeah but until very recently it was disabled by default
<fda>
i noticed the some very basic bugs ... eg when you close a browser tab with luci, and then just open the main url of the same luci device again, it asks again for a password. but the luci menu is visible and you could just use it to navigate
goliath has quit [Quit: SIGSEGV]
<Ansuel>
mangix: can't remmeber if we have more testing to do with 1ca8k
<mangix>
Ansuel can you ask the qca8327 guy to test with mac6-exchange disabled? I don't know how the initials would need to be modified.
<Slimey>
tmn505 whats the little board in the front all by itself
<fda>
or if you lougout + login again and have password saved in browser so you could login very quick there is a rpc or xml failure, but login succeeds
<jow>
fda: the menu is cached in session storage which is supposed to get flushed when the tab is closed
<jow>
fda: the rpc failure, yeah, needs investigation
<Ansuel>
mangix: mhhh so testing if it does work with a configuration of port6 enabled? can't we just ask him to test qca8k with port6?
<jow>
fda: can't reproduce your first issue
<jow>
closing tab and reopening base url does not ask for login
<mangix>
Ansuel: That too :). But I mean mac6-exchange being set is probably a copy/paste error.
<tmn505>
Slimey: that's antenna
<Slimey>
for BLE or dedicated scanning radio?
<Slimey>
or just normal wifi
<fda>
this is since i started to use openwrt currently with firefox93
<mangix>
plus my DSA PR needs a newer version of the qca8k backport PR.
<jow>
fda: neither on firefox, nor chrome
<mangix>
I did the v5 changes already.
<Ansuel>
mangix: problem is that i can create custom intval to test the configuration but at this point is just quicker to test qca8k with the same config...
<jow>
fda: maybe you once open https and then http or vice versa
<jow>
which might cause brpwsers to not send sysauth cookies
<tmn505>
Slimey: normal wifi 2.4, connected to ar1022
<fda>
jow: i have no https configured
<jow>
or something, in any case I cannot reproduce it with some simple close/reopen tab here
<Ansuel>
mangix :( wonder if github still have old version based on the current backport pr ?
<fda>
is use the hostname which resolves to its ULA ipv6
<Ansuel>
mhh nope it does require v3-4-5 for the port6 configuration...
<Slimey>
cool
<jow>
fda: I do as well
<fda>
jow: maybe you browser deleted cookies (or whatever) if you close a tab
<jow>
in that case I would be prompted for login
<fda>
i noticed this behaviour on all my 3 devices
<Ansuel>
same browser?
<fda>
yes, its shows the prompt. but im still authenticated and the menu shows and could be used
<jow>
maybe you're simply opening the logout url?
<fda>
hm, with EDGE it does not show login prompt, but a complete empty overview
<ldir>
nbd: ping - on your macos was does python -V vs python3 -V return ? And do you have anything that could be python related in your path, eg. /usr/local/bin containing a 'python' of some sort ?
<ldir>
nbd: s/was/what/
goliath has joined #openwrt-devel
<jow>
fda: what version of openwrt and luci do you run exactly?
<jow>
nvm, found it in your screenshot
<fda>
current of today
<fda>
these things are very minor, but i notice it all the time
<fda>
@nbd thy, build works now
<fda>
and the avm1200 booted :)
<fda>
sadly wsdd2 is also broken for this (relative new) device: "procd: Instance wsdd2::instance1 s in a crash loop 6 crashes, 0 seconds since last crash"
<fda>
i wun it from rc.local as the normal init script always fails: "samba36/4 or ksmbd is not running, can't start wsdd2!"
<fda>
i have no other logs as on ssh and syslog nothing is visble (except procd message on syslog). segfault seems only on serial console
fda- has joined #openwrt-devel
<fda->
on e8450 syslog is also like fopr avm7320 this on syslog: "procd: /etc/rc.d/S95done: Command failed: Not found"
<jow>
fda-: will push a fix for that later. It may happen is /etc/config/firewall contains syntax errors or is not present at all
<fda->
@jow my /etc/config/firewall has 494 lines
<dwfreed>
2,575 on one system here, but that could probably be improved...
<jow>
fda-: uci show firewall >/dev/null
<jow>
will display the syntax error on stderr if there's any
<jow>
also if at some point there was an error and it is fixed in the menawhilew, the menu might still be cached without firewall entries, in such a case rm -r /tmp/luci-* should restore it
<fda->
jow: no errors! on my 2nd device there are only 46 lines and shows also this error on syslog
<fda->
could be contents of /etc/firewall.user be the source of the error?
<Habbie>
mangix, thanks, ath10k hint plus some tinkering has given me both bands back :)
<fda->
jow: and it seem no be a problem to edit custom rules. i have there things like "ipXtables () { iptables $* ; ip6tables $* ; }"
<jow>
fda-: no.
<jow>
fda-: also which error on syslog do you mean specifically? Since you appaer to talk about multiple (related or unrelated) issues at once, I lost track somewhat
<fda->
open the luci url /cgi-bin/luci/admin/network/firewall/custom
<jow>
then this error is caused by some related request
<jow>
some XHR call likely
<jow>
for that it would be useful to open the browser debug console, network tab and see which call yields a 502 if you open/reload that url
<jow>
I guess it is an "expected" 404, like a probe for /ubus or translation feeds
<fda->
there is a 500, but for all urls in luci for /cgi-bin/luci/admin/network/firewall/logo.svg
<jow>
aha
<fda->
if you open /cgi-bin/luci/admin/network/firewall/logo.svg direct it will try a favicon which gives 404
<jow>
sounds like a bug in your theme's css
<jow>
it attempts to include an svg image relative to the cgi bin path
<jow>
which triggers a 404 which yields the error you see in syslog
<fda->
that would be really bad....
<jow>
(which is a bug, that 404 should not trigger a 500 that triggers a backtrace, it's an error introduced by recent refactoring)
<jow>
but that bad logo.svg request is the cause
<jow>
from your screenshots I assume you use some kind of modified openwrt-2020
<jow>
(it's also the only theme using a file named "logo.svg")
<fda->
im not sure if its the theme
<fda->
i just copied openwrt-2020 and replaced colors (also in the svg)
<fda->
with bootstrap therem there is no error
<fda->
have you openwrt-2020 installed?
<fda->
thanks god... openwrt-2020 has the same error in syslog :D
<fda->
with bootstrap there is no error
<jow>
using which browser?
<fda->
firefox93
<fda->
but the error is only in syslog
<fda->
firefox told me: "Die Zeichenkodierung des Reintext-Dokuments wurde nicht deklariert. Das Dokument wird in manchen Browser-Konfigurationen mit verstümmeltem Text dargestellt, wenn das Dokument Zeichen außerhalb des US-ASCII-Bereichs enthält. Die Zeichenkodierung der Seite muss im Transferprotokoll deklariert werden oder die Datei muss eine Byte-Order-Markierung als Kodierungssignatur verwenden."
<jow>
yeah, unrelated
<fda->
some character encoding error
<jow>
something is somehow requesting logo.svg relative to the current document url instead of relative to the CSS file
<jow>
firefox 93 is not doing that here
<jow>
it requests //$router/luci-static/openwrt2020/logo.svg as expected
<jow>
not //$router/cgi-bin/luci/admin/network/firewall/custom/logo.svg
<jow>
or //$router/cgi-bin/luci/admin/network/firewall/logo.svg
<jow>
mind to pastebin your http://.../luci-static/openwrt2020/cascade.css ?
<fda->
i did just opkg install luci-theme-openwrt-2020
<fda->
and switched by webif
<fda->
btw i have this http-500 on all luci urls! by syslog error is only on the 1 url
<jow>
well, I did too, and it is not doing it here
<jow>
I only tested with one open tab which I closed and then reopned
<jow>
having two tabs open, then closing one and opening another shows it for me as well
rmilecki has quit [Quit: Konversation terminated!]
<fda->
what is expected by openwrt wenn i open the device again? should the be a login again or should it behave a yet logged in?
<fda->
imho: dont show login again
<jow>
it should behave like logged in, it's cookie based after all
<jow>
and for me it only flashes the login view briefly and immediately redirects to the status page
<jow>
likely some timing thing
<jow>
ah okay, the cookie is https based, the newly opened tab initially opens the http:// url
<jow>
for that the browser does not send the cookie, since it is restricted to https://
<jow>
but then some browser auto-try https heuristic kicks in here, making the page reload as https://, then the cookie is sent and everything works
<fda->
i have no https configured for luci, so i use only http
<jow>
let me try that
<fda->
to the wrong svg path: what about to return 404 instead of 500 for things like /cgi-bin/luci/admin/services/unbound/logo.svg ? when i open /cgi-bin/luci/null there is a 404
<jow>
can't reproduce the login-on-new-tab issue with only http
<jow>
only when being logged in via https, then re-open a tab as http
<fda->
maybe its also something with wrong paths because of a addon
<fda->
im trying a new image and flash to clear all caches
<fda->
but now strace fails to build
<fda->
"Makefile:9303: *** missing separator. Stop". there is "@CODE_COVERAGE_RULES@"
<fda->
but there was no reason to rebuild this package as is worked to make in the same directory soem houts ago
Ansuel has quit [Quit: Probably my PC crashed or time to sleep.]
Ansuel has joined #openwrt-devel
Tapper has quit [Quit: Tapper]
Tapper has joined #openwrt-devel
<mangix>
fda-: no idea why that breaks. Can be patched out.
<Habbie>
mangix, so, just to be sure, if i want to test 'the DSA PR' on archer c7 v5, i merge 4528 and 4622 and build that?
<Ansuel>
Habbie i think it won't work as the ath79 pr now point to a new version that is still not present in the backports pr
<Habbie>
right
<Ansuel>
we are currently pushing things upstream so the patch changed a bit and they are not synced
<Habbie>
ok
<Habbie>
well, i'm happy to test, given a clear branch or commit to point my compiler at :)
<Habbie>
whenever that is convenient
<Ansuel>
magix can still give you the correct patches if you want to test
<fda->
@jow i started a complete rebuild, should be done soon. the wsdd2 segfault is in libc as you said: do_page_fault(): sending SIGSEGV to wsdd2 for invalid read access from 00000001" - " "epc = 77eab76c in libc.so[77e2f000+a5000]"
<fda->
rebuild sucessfully inclusing strace ^^
fda has joined #openwrt-devel
rmilecki has joined #openwrt-devel
fda- has quit [Ping timeout: 480 seconds]
<mangix>
Habbie: so for that router everything should be good to go. Except the sgmii-rx-falling-edge binding. That just has to get moved below led-open-drain
<mangix>
Ansuel says there's instability without it. I don't remember such a thing. But w/e.
<Habbie>
mangix, ok, so, did you want me to test, or are you confident? and if you want me to test, test which branch/commit?
<Habbie>
and 'qca8k-backport' is the local branch that was created by the first gh pr checkout
<Ansuel>
oww interesting...
<mangix>
Clever.
<Habbie>
you don't -need- 'gh' but it makes things easier
<mangix>
I used to use a git-pr python thing. Wonder what happened to it.
<Ansuel>
then how you get the pr in a normal way without downloading the .patch from the link?
<Habbie>
Ansuel, add fetch = +refs/heads/*:refs/remotes/origin/* to the [remote] in .git/config
<Habbie>
then all PRs appear
<Habbie>
(on git fetch --all, or something like that)
<Ansuel>
right
<Habbie>
before i knew that, i added the submitter's fork as a remote and checked out their branch
<Habbie>
which also works but now i have 20 remotes
<Ansuel>
yhea not that good
danitool has joined #openwrt-devel
minimal has quit []
<Habbie>
sysupgrading
<fda>
@jow login+menu seems to be some cross-site thing. i have a "startpage" on all devices/browsers with links. so no favorite sync by only 1 browser. login+menu shows only (always..) when i use a link on this site
<Habbie>
mangix, i appear to have lost WAN
<Habbie>
mangix, and LAN :)
<Habbie>
so whoever said 'configure wifi before you do this' was spot on
<mangix>
Habbie: that would be me
<mangix>
want to fix it?
<Habbie>
yep, sounds good, for me and other users :)
<Habbie>
(you were in my top 3 of people who might have been the one saying 'enable wifi before ..' indeed :D )
<mangix>
that most definitely needs ag71xx support though
<Ansuel>
i don't know if we support that
<Ansuel>
i mean it's something that the attached mac needs to support
<Ansuel>
so very SoC specific
<mangix>
it's in russell king's tree. so not upstream either :)
<Habbie>
folks, thanks for not making me feel like a noob here, i'm definitely out my depth in this :)
<Ansuel>
i need to check the in-band stuff...
<Ansuel>
i have this crazy idea of adding a special mode.... qca8k support special header where you can write read mdio with sending special packet to the switch
<Ansuel>
that would improve mdio access (in theory)
<mangix>
Habbie: so one of the PRs I linked to in the ath79 dsa PR is one that changes the default kernel to 5.10. I listed it as optional. I completely forgot 5.4 is the current default.
<Habbie>
right
<Habbie>
i figured it was something like that
<mangix>
Ansuel: that sounds like witchcraft
<Ansuel>
it is but i'm too curious to implement that... problem is that i found some similar implementation but nothing really similar...
<Ansuel>
it's all documented but i lack some knowled on how to correctly implement that
<mangix>
which I assume means you'd get to have more fun with upstream
<Habbie>
who is upstream? linux?
<Ansuel>
since we send packet to the switch i really lack the logic/idea on how to handle that... the transaction part of sending and wait for a response part
<mangix>
Habbie: yeah
<Ansuel>
yes upstream = linux
<Habbie>
ack
<Ansuel>
i should ask robi if he has any idea on how to implement that
<Ansuel>
from documentation te logic is not that hard and we also already have the tagger that use the special header
<Ansuel>
the main idea is that you send this specially crafted packet and the switch answer you with ack if it's an mdio write or the mdio regs asked to read...
<Ansuel>
this in theory should be quiker??? than using mdio read/write logic as we send one packet instead of writing 3 times to the mdio lin e
<Habbie>
but who handles the packet? the switch chip?
<mangix>
Ansuel: what's the benefit to faster mdio?
<Ansuel>
less cpu load
<mangix>
cpu load for what?
<hauke>
Ansuel: does this also work from normal LAN ports?
<Ansuel>
no soc have the switch interrupt pin attached (we never had a chance to verify this)
dedeckeh has quit [Remote host closed the connection]
<Ansuel>
so currently the system pool the switch to get port status
<hauke>
what happens when you send such packets around in some network like a local hotspot ;-)
<Ansuel>
hauke the packet needs to be sent to the cpu port. it does use atheros special header... currently in qca8k is used to tag packet and forward them to the correct port
<Habbie>
a lot of tools need some time to rebuild with a kernel version change it appears
<Ansuel>
you can set a priority to every packet so overload problem can be solved by using priority i think
<mangix>
Habbie: enabling ccache helps quite a bit
<Habbie>
my system has ccache but openwrt might not be benefiting
<Habbie>
probably because i use the $PATH hack
<mangix>
no it needs to be enabled for OpenWrt
<Habbie>
and this compiler is not in there
<Habbie>
ack
<Habbie>
i see it it in .config
<mangix>
make menuconfig > advanced > ccache
<Habbie>
why is it not default? :)
<mangix>
I have no idea
<mangix>
I think historically it caused issues but all of them have been fixed.
<Habbie>
running sysupgrade
<Habbie>
i'll be off to bed soon
<mangix>
ah right, Europe
<Habbie>
yep
<Habbie>
you're in US?
<mangix>
yeah
<mangix>
13:36 here
<Habbie>
nice
<Habbie>
22:36 here
<mangix>
Habbie: btw since you maintain dnsdist, are there any plans upstream to replace the build system with either cmake or meson?
<Habbie>
there are ambitions but no plans, if you understand what i mean :)
<Habbie>
it's a lot of work
<Habbie>
and at least some of those alternatives deal very badly with our monorepo, i understand
<Ansuel>
also eu here it's fun doing testing with magix
<Habbie>
ok wifi is up
<mangix>
I'm sure it is. Last I looked at dnsdist, it was difficult to run clang tools on it.
<Habbie>
running 5.10.70
<Habbie>
no WAN
<mangix>
Ansuel: you make it sound like I should move to Europe
<mangix>
I listed it as optional since I turn off all my LEDs anyway
<Habbie>
so, things to do: try without mac6-exchange (should be terrible), try that leds PR
<mangix>
I mean if you want LEDs, sure.
<Habbie>
well, many users would, i think
<Habbie>
and i can do builds and test a few things in between work
<mangix>
OK
<Habbie>
4562 merged cleanly into this
<Habbie>
so i'm off to bed but with some luck i'll have more news when -you- wake up :)
Ansuel has quit [Ping timeout: 480 seconds]
Ansuel has joined #openwrt-devel
<Ansuel>
mangix: news?
<Habbie>
Ansuel, everything works on my c7 v5 except for LEDs so i'll test 4562 tomorrow - if that is the news you were looking for
<Ansuel>
good to know
<Ansuel>
leds doesn't work at all ? or the uncorrectly blink ?
<Habbie>
power+wifi work correctly
<Habbie>
rest is off
<mangix>
Ansurl: LED definitions are missing in DTS
<Habbie>
'works' is 4622+4528+5.10 plus mangix' small patch
<Habbie>
ok, zz, good day all
<Ansuel>
mangix: strange tho so they set a completely not standard configuration for leds
<Ansuel>
nice!
<mangix>
Ansuel: ummm, isn't the qca8k LED patch needed?
<mangix>
or is that just for controlling them?
<Ansuel>
it depends if the switch implement custome rule or not
<Ansuel>
the switch have default rules so unless we need to edit something.... it should work (if the oem didn't do something strange)
<mangix>
for this one, looks like no
rmilecki has quit [Quit: Konversation terminated!]
<mangix>
or it's a bug in the output
<mangix>
who knows
<Ansuel>
it's strange that they doesn't work honestly
jbowen has quit [Quit: leaving]
<mangix>
Ansuel: rsalvaterra had a similar issue. He fixed it by adding led-open-drain. But this config doesn't see to have that.
<Ansuel>
let me check something
<Ansuel>
ok we actually disable power-on-sel and led open drain if they are not defined in dts
<Ansuel>
you could be right
<Ansuel>
so much bad configuration that probably only works because of uboot
<mangix>
Ah I see.
<mangix>
That's quite annoying then.
<mangix>
plot-twist: all qca956x devices use open-drain
<Ansuel>
we can consider ignoring the value but i don't know... it seems something badly defined from the vendor
<mangix>
Most likely. Only one I've really looked at is a CyberTAN SDK. Pretty bad.
<Ansuel>
what did you notice that was bad?
<mangix>
Switch config being done in userspace.
<Ansuel>
......
<Ansuel>
with a good sysfs entry LOL
<Ansuel>
ahahahah
<mangix>
With magic undocumented values
<mangix>
No
<mangix>
Not even sysfs
<Ansuel>
well if you check the current ar8216 code it's full of magic values
<mangix>
stuff like switch reg write 0x1f
<mangix>
or switch reg write 0x5d; usleep 500: switch reg write 0x7f in a loop
<mangix>
To get blinking behavior or w/e
<Ansuel>
majestic
<mangix>
Actually this switch supports driving the LEDs in GPIO mode. It was not used in the stock
<mangix>
Erm vendor SDK.
<mangix>
Had to figure out everything by trial and error.
<mangix>
One of the GPIOs is actually inverted for some reason.
<Ansuel>
ok that is terrible
<digitalcircuit>
Thanks for the help and patience! I'll follow up once I've either found a fix, or tried everything that's been suggested so far.
<digitalcircuit>
slh (and Ansuel): Low priority update for NBG6817 on kernel 5.10 boot failure with Kingston MMC - backporting Linux 5.12's commit for MMC voltage bus probing did NOT seem to fix the issue (test branch: https://github.com/openwrt/openwrt/compare/master...digitalcircuit:test-fix-5.10-nbg6817-new-mmc ). I'll now remove this change and try reverting the earlier commits suggested (7577316, 3e09a81).
<clayface>
Ansuel: I overlooked the conditions for those bindings. Did you start on this? I can try and figure it out now if you want
<Ansuel>
if you want we can search and devide the logic for the single binding
<Ansuel>
device.... does this word exist?
<Ansuel>
anyway and split the work
<Ansuel>
we need to limit the binding to port 0 and 6 and if we really want also limit it to mode sgmii
<Ansuel>
we can hove something like that is present in some documentation for the other dsa switch
Tapper has quit [Ping timeout: 480 seconds]
<mangix>
divide
<Ansuel>
wow it's actually the correct meaning just NEVER USED lol
<Ansuel>
devide = split
<mangix>
divide
<mangix>
Could also say bifurcate to confuse upstream.
<clayface>
I think we need to define mdio, ethernet-switch, ethernet-port containers, then put in our properties and conditions for port 0/6. Is it correct?
JiiPee has joined #openwrt-devel
<Ansuel>
right we lack also the mdio binding
<Ansuel>
claface: did you do some progress? i'm starting now