<mangix>
Ansuel: it says in the output With tx-internal-delay-ps of 2000. But then I see a division of 1000 in the patch. I take it 2 is the correct value?
<mangix>
ah nvm. I see in the kernel tree now.
<Ansuel>
anyway i think it's ready unless i made some mistake...
<Ansuel>
anyway setting to port5 can be ignored as it's set in theory.... anyway someone said there was a way to identify the switch type from the dts naming
<Ansuel>
so best place to post the script?
philipp64 has joined #openwrt-devel
philipp64 has quit []
<Ansuel>
what i didn't notice is pll disable
<Ansuel>
where is used?
<Ansuel>
ok plls needs to be addressed... the switch have that disabled by default and current qca8k code enables that by default (that goes against any logic)
<mangix>
Ansuel: as a git commit. Inside the scripts folder probably.
<Ansuel>
mhhh should we really post that in the scripts dir? isn't that overkill ?
<mangix>
one thing I'm curious about is the status of the LEDs. Will it be merged in master?
<mangix>
openwrt/master that is
Luke-Jr has quit [Ping timeout: 480 seconds]
<Ansuel>
imho yes... it's not that hacky... i will for sure update the patch and documentation but minor changes... (they poited out on upstream that leds should be placed in the phy node)
victhor has quit [Ping timeout: 480 seconds]
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schwicht has joined #openwrt-devel
schwicht has quit []
<mangix>
Ansuel: got it
<slh>
probably start putting the script with a public gist
<slh>
that gives you an editable/ stable location
<Ansuel>
i posted it on the qca8k pr with some other valueble info
danitool has quit [Ping timeout: 480 seconds]
Luke-Jr has joined #openwrt-devel
Luke-Jr has quit [Read error: Connection reset by peer]
Luke-Jr has joined #openwrt-devel
cp- has quit [Read error: Connection reset by peer]
Ansuel has quit [Read error: Connection reset by peer]
rmilecki has joined #openwrt-devel
guerby has quit [Read error: Connection reset by peer]
guerby_ has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
philipp64 has joined #openwrt-devel
<philipp64>
mangix: I put a USB thumb drive onto my Archer, but... what do I need to enable to use and mount an EXT2 (or 3) filesystem on that drive?
<stintel>
initialy I thought the m300 soc has 2 ethernet MACs connected to the mv88e6171, but attempts to bring up the 2nd one always result in: [ 2.250073] mv88e6085: probe of 0x0000000ffe4fc000:10 failed with error -22
<Tusker>
-22 means that you are trying to bring up two dsa roots on the same member. if you specify a different dsa member, it should allow it
<stintel>
are you suggesting there are 2 switches ?
<Tusker>
definitely on mine there are two switches, one at 0x10 and one at 0x11
<stintel>
OEM output seems to suggest the same for M300
<Tusker>
these two switches could probably be connected via a DSA interlink on one of the ports, meaning that they are trunked in some way (but hard to tell)
<stintel>
so the one I can access is on 0x10 - I suspect the 16 in the above paste
<stintel>
but I've tried all kinds of things to access the one on 4 but never works
<stintel>
so I guess I just don't properly understand how this works
<Tusker>
join the club, I am pretty confused :)
Kolusion has joined #openwrt-devel
<stintel>
also I hadn't properly it sent on my mail to you and wg answered the m200 I have is part of a cluster and cannot be used standalone and told me to contact sales :/
<stintel>
wankers
<stintel>
hmmm packet loss, is it local or is it my ISP
* stintel
goes to make coffee first
<Kolusion>
Hi all. I just checked out the OpenWrt website and wanted to provide some feedback. It is one of the worst open source projects I have seen. The website is so confusing. Things are written in developer terms instead of user terms. I tried to download it and every time I click something, I am presented with a directory listing of hundreds of cryptic names of which are all meaningless to me and there is
<mangix>
Y
<Kolusion>
... a major problem with FOSS projects is that they present everything in developer terms, when majority of people, like me, are users, making the developer jargon meaningless.
<Kolusion>
I would have liked to have been with a no frills firmware image to flash my router.
<Kolusion>
*been presented
<Habbie>
for free
<rsalvaterra>
Kolusion: What's your device?
<mangix>
Lol one of the worst.
<Kolusion>
It is a TP-Link TL-WR841N.
<Habbie>
that's a -very- tight device in terms of RAM and flash i believe
<Kolusion>
rsalvaterra: Not likey definatly. I checked out the feature set of OpenWRT. It is way more advanced than what my stock firmware supports.
<Kolusion>
russell--: Thanks, but I am just presented with cryptic directories with no explanation which one I need.
<Kolusion>
I went to openwrt.org, clicked download, clicked stable, clicked any version, clicked 19.07.8, clicked packages, and then I get crypticness.
<Kolusion>
See the problem as a newcomer I am having?
<Kolusion>
Not here to stir the pot.
<russell-->
lol
<Kolusion>
I just wanted to provide feedback.
<russell-->
"The TL-WR841N is identical to the TL-WR841ND except that the antennas are not detachable. TP-Link naming scheme is, that devices ending in a D have detachable antennas."
<Kolusion>
russell--: Maybe. But I am out of time. Thanks and take care. :)
Kolusion has quit [Quit: Page closed]
<rsalvaterra>
*Sigh*
<stintel>
not here to stir the pot, comes in "one of the worst open source projects I have seen"
<stintel>
sigh indeed
<rsalvaterra>
Sometimes I hate being so patient. :P
<rsalvaterra>
Anyway, I don't see support for the v14 going all the way back to Chaos Calmer (which probably predates the v14).
<russell-->
also, *punches face of vendors* for model number confusions
<rsalvaterra>
russell--: Indeed. The same model name with completely difrerent hardware, jeez…
<rsalvaterra>
*different
<Tusker>
any idea where the UART is for this device Grandstream GWN7610 - https://fccid.io/YZZGWN7610/Internal-Photos/Internal-Photos-3098045 ? The 4 PIN square header you see is the most obvious candidate, but shows up nothing in 9600, 57600, 115200, so uart output is probably disabled. It shows 3.3v, ~0v, GND, ~2.5v. Which seems to me to indicate VCC, RX, GND, TX
<russell-->
Tusker: do you have any good pictures?
<Tusker>
let me take some
<Tusker>
i soldered on the header, so excuse the bad soldering
guerby_ has quit [Remote host closed the connection]
<russell-->
you will be ridiculed mercilessly
guerby_ has joined #openwrt-devel
<stintel>
lol
<Tusker>
waiting for google photos to do it's thing...
<Habbie>
Tusker, it always amazes me how long that takes!
<Tusker>
yeah, that 2x2 header with flux all over it is what I added, given that the voltages look normal for UART
<russell-->
the two traces that end near the NC silkscreen look vaguely like tx and rx
<russell-->
the seems to go through resistors to the two round test points
<russell-->
can you check continuity from the round test points to the pins?
<Tusker>
yeah, and those two test points continuity to the pins. the cloest test pad to the pins is connected to the pin closest to them ... let me draw
<russell-->
i'd find ground, then assume those two pins are rx and tx, try one way, then try them swapped
<Habbie>
'sigrok compatible' is key, the SPI trick is not (but somebody should fix that ;) )
chwba has quit [Remote host closed the connection]
decke has joined #openwrt-devel
pmelange has joined #openwrt-devel
hrw has left #openwrt-devel [#openwrt-devel]
guerby_ has joined #openwrt-devel
guerby__ has quit [Read error: Connection reset by peer]
goliath has joined #openwrt-devel
chwba has joined #openwrt-devel
<rsalvaterra>
mangix: No cigar. My C6 v2 is very unhappy. It boots, but I have no access to it, something's missing in the switch configuration, for sure.
Tapper has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
victhor has joined #openwrt-devel
Tusker has quit [Quit: Time wasted on IRC: 10 hours 43 minutes 44 seconds]
Ansuel has joined #openwrt-devel
Ansuel has quit []
Ansuel has joined #openwrt-devel
abiliomarques has quit [Read error: Connection reset by peer]
abiliomarques has joined #openwrt-devel
schwicht has joined #openwrt-devel
schwicht has quit []
<rsalvaterra>
hauke: Are we go for the kernel bumps merge? The user reported his RPi 3 is fixed. :)
Tapper has quit [Ping timeout: 480 seconds]
<rsalvaterra>
hauke: Hold on. 5.10.70 is out. xD
Tapper has joined #openwrt-devel
abiliomarques has quit [Quit: Leaving]
<stintel>
I see, I was wondering why kernel was so outdated in master :P
<rsalvaterra>
Ansuel: No joy with my C6 v2. I probably screwed up somewhere. :/
<stintel>
rsalvaterra: hey, just my 2 cents, if there are regressions in 5.10.65, I would squash 5.10.65 and 5.10.66 because bisect arriving on 5.10.65 would have known regressions so we should avoid that
<Ansuel>
can you link me the dts? nothing is reported in the kernel log?
<rsalvaterra>
stintel: No regressions. It was a bad patch refresh by yours truly, sorry about that. ;)
<stintel>
oh
<stintel>
ok then nvm me
pmelange has quit [Read error: Connection reset by peer]
<rsalvaterra>
Ansuel: No access to it, it was at factory defaults, Wi-Fi radios disabled (and no serial console). Need to recover and retry, but probably not today.
pmelange has joined #openwrt-devel
<Ansuel>
rsalvaterra: bad :( can you give me the dts so i can check if something is wrong here?
<stintel>
ugh. that change to fail builds on missing kconfig symbols is useful, but now I'm forced to figure out why I'm being hit by some that make absolutely no sense
<stintel>
like the one I just pushed for KVM
<stintel>
I'm also bitten by SND_SOC_ALL_CODECS missing on rpi
<stintel>
no, wrong paste buffer, SND_SOC_AD193X_SPI
<stintel>
that's going to be another bald yak today
<stintel>
goddamnit where the hell does this come from
<stintel>
it's even set CONFIG_SND_SOC_AD193X_SPI=n in sound-soc-bcm2835-i2s
decke has quit [Quit: Leaving.]
<blocktrron>
mangix: regarding your "PHY-Swap does not work" issue - what behavior are you getting with your device?
<mangix>
No traffic.
<blocktrron>
link status works?
<stintel>
ok so I don't have sound-soc-bcm2835-i2s enabled, I'm guessing enabling that would magically solve the problem for me but that still leaves the possibility for it to go missing with certain config options
<mangix>
hmmm I don't remember. I want to say yes.
<blocktrron>
so you see that the interface goes up / down? And the lan ports work correctly?
<mangix>
Needs patching. Upstream only supports ar9331.
<mangix>
After I finish this qca8k stuff I'll look at ar9331
<mangix>
Ansuel: btw did the suspend/resume fixes get in the qca8k PR?
<Ansuel>
nope (i think).... Leo reported some problem with half duplex with 100m so i'm still investigating that...
en4rab_ has joined #openwrt-devel
<hauke>
rsalvaterra: thanks for integrating the kernel update fix, I was already sleeping at that time.
<stintel>
I'll have some Tested-by for the 5.10.70 hopefully later today if I can sort those bloody missing symbols
<stintel>
I think I know the proper fix, but the commit message for the KVM symbols was kind of sketchy and not properly explaining the actual problem, so I'd like to avoid doing that again
floof58 has joined #openwrt-devel
floof58__ has joined #openwrt-devel
floof58_ has quit [Ping timeout: 480 seconds]
floof58 has quit [Ping timeout: 480 seconds]
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
eck has joined #openwrt-devel
<mangix>
rsalvaterra: ping
eck has quit []
eck has joined #openwrt-devel
<mangix>
Ansuel: which devices need sgmii pll enabled?
<blocktrron>
aparcar[m]: regarding your kernel bumps - remove 5.4 in the same patchset
<blocktrron>
At this point, it makes no sense to track both kernels
<philipp64>
slh: thanks... building
<Ansuel>
mangix: i updated the script in the pr to tell you the devices
<mangix>
Ansuel: LOL they're both qca9563
pmelange has quit [Quit: Leaving.]
eck has quit [Remote host closed the connection]
eck has joined #openwrt-devel
<stintel>
how can bcm27xx even build if RPI_AXIPERF isn't in any kernel config
<stintel>
is this problem reflected in the buildbots ?
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
<rsalvaterra>
mangix: Not at home, now. What's up? :)
<Slimey>
this happens every time, even pulled new git clone. cc1: fatal error: ../dts/_adtran_bsap1920.dts: No such file or directory
<mangix>
rsalvaterra: the Archer C7 commit in my branch was updated with your router
<mangix>
c6 v2
<Slimey>
filename is actually ar9344_adtran_bsap1920.dts but it keeps putting _adtran_bsap1920.dts which doesnt exist
<PaulFertser>
Slimey: guess it means SOC is not assigned in your image Makefile definition.
<Ansuel>
mangix: coincidence? i don't think so....
<Slimey>
PaulFertser its basicly a clone of the ar9344_pcs_cap324
<Slimey>
what makefile you thinking about?
<stintel>
bloody kconfig voodoo and black magic
<mangix>
Ansuel: what do you mean?
<Slimey>
the one at /target/linux/ath79/image/Makefile??
<PaulFertser>
Slimey: yes, or some of the files included from it, the one that mentions adtran_bsap1920.
<Slimey>
its in generic
<rsalvaterra>
mangix: Oh? Ok will test later. I bricked mine today, need to recover it.
<PaulFertser>
Slimey: so you're missing SOC assignment there.
<Slimey>
ah i see what you mean
<rsalvaterra>
hauke: Don't mention it, I thank you for finding the bug so quickly. :)
<Slimey>
PaulFertser its from a old snapshot but it was saying ATH_SOC instead of just SOC
<Slimey>
ath79 with kernel 4.14 i think
<Ansuel>
mangix: it's probably a problem of the soc and qcom advice that
<Slimey>
yay now to see it if boots :P
<mangix>
Ansuel: qca9563_xiaomi_aiot-ac2350.dts is strange. It says cpu port5, Mode rgmii-rxid
<mangix>
I didn't know port 5 could be a CPU port
<Ansuel>
mangix: yes that is normal with qca8337. with every qca8337 if rgmii-id is set for cpu port0 also port5 require to set delay for rx... i should hide it...
<Slimey>
:/
<Slimey>
[ 1.138020] Run /sbin/init as init process
<mangix>
aparcar[m]: no. ramips is mt7621 only. ath79 is tiny target only.
<mangix>
forgot what else there is
<aparcar[m]>
mangix: how can I tell?
<mangix>
tell what?
<mangix>
Ansuel: can you check qca9563_xiaomi_aiot-ac2350.dts ? Everything is weird about it. 02_network says "0@eth0" "2:lan" "3:lan" "4:lan" "1:wan" . I don't see any eth1.
<aparcar[m]>
mangix: well I want to track status of all target/subtargets. I'm guessing I'll just look whereever SW is disabled or something?
pmelange has joined #openwrt-devel
<mangix>
aparcar[m]: just checked. All targets with CONFIG_NET_DSA=y use DSA. Only exceptions are ramips and ath79 which only do so partially.
<Ansuel>
mangix: yep it's correct qca8337 with sgmii (so the port5 fixup won't be applied and they set to 0 any reg to port6 (so no secondary cpu) also we lack some undocumented stuff for that... do we have a user to test some stuff? i'm curious
<mangix>
nope
<Ansuel>
what a strange config..... the fact that they set rgmii delay for port5 is interesting... in the current qca8k code we set that only with rgmii-id
<G10h4ck>
anyone know if it is possible to reasonably package programs written in rust-lang for openwrt?
<mangix>
nope
<G10h4ck>
last time I had some news someone had success crosscompiling for MIPS but the binry where minimum 2MB each
<Habbie>
at a quick grep i spot zero rust things in the openwrt packages repo
<stintel>
there's a PR for Rust in the packages feed, I suggest you test it
<mangix>
Ansuel: anyway, ar9344_atheros_db120.dts has a dual switch setup. Need blockttron's stuff to get that working (not that I have the hardware). I have no idea what ar9344_netgear_wndr.dtsi is doing
<PaulFertser>
aparcar[m]: is anywhere aware of the cert error that makes 21.02 opkg not work atm?
<mangix>
maybe dual switch setup as well?
<Ansuel>
ar9344_netgear_wndr looks normal... aside from the undocumented bit in pws reg
<aparcar[m]>
PaulFertser: no clue
<mangix>
Ansuel: right but it has a builtin_switch node
<Ansuel>
give me a link pls
<Ansuel>
(sorry i'm lazy)
<PaulFertser>
aparcar[m]: https://bugs.openwrt.org/index.php?do=details&task_id=4059 and the users on #openwrt complain about related issues now. I thought you might be involved with the openwrt.org infra enough to know whom to ask about it.
<Ansuel>
mangix: from what i can see it seems they attached wan to phy port and lan port are handled with the switch
<mangix>
wouldn't that make the wan port part of the switch?
<G10h4ck>
thanks Habbie and mangix, i was considering rewriting from scratch a program we started as a prototype in Lua, rust seemed the perfect match but if it doesn't supports well OWRT is no way...
<Ansuel>
mh no it's physicall attached to the soc phy... direct connection cpu port to phy
<mangix>
Ansuel: so if the builtin-switch is used, why is there even an ar8327 node?
<Ansuel>
cause there are 2 switch one with just one port attached (wan) and the separate dedicated for lan eth0 connected to switch eth1 for wan
<Ansuel>
you should port the qca8327 part with the wan port removed
<Ansuel>
also lan port are swapped phy0 is lan4 phy1 is lan3 ....
<aparcar[m]>
PaulFertser: okay I'm able to reproduce this, not great
<Ansuel>
o.O
<Ansuel>
dsa conversion is FUN
<mangix>
oh wait that's v2
<mangix>
v4 is using ar8327
<PaulFertser>
aparcar[m]: can you pull the right strings please?
<mangix>
I see the builtin-switch used nowhere
G10h4ck has quit [Quit: Konversation terminated!]
<mangix>
probably a mistake when porting from ag71xx
<aparcar[m]>
PaulFertser: do you have the strings at hand?
<Ansuel>
could be... but what is the compatible of builtin switch ?
<mangix>
it's ar9331
<PaulFertser>
aparcar[m]: I do not as I'm just an OpenWrt user, I have no idea what the infra consists of and who's taking care of it.
<mangix>
it's a fast ethernet switch, not gigabit
<mangix>
wikidevi says it's all gigabit
<mangix>
dts file is wrong
<Ansuel>
oh wow....
<Ansuel>
nice that we can now fix that
<aparcar[m]>
PaulFertser: I already pinged some people, will try to fix things
<mangix>
so are these initvals
<mangix>
there's no eth1
<PaulFertser>
aparcar[m]: thank you!
<Ansuel>
are we sure the nwdr is 100% correct? the rgmii-rxid on port6 looks wrong...
<mangix>
I mean...02_network only references eth0
<mangix>
the dts file only references eth0
<mangix>
only eth0 being used makes sense
<jow>
PaulFertser: yes, we're aware of the cert error
<jow>
and I curse whoever came up with the braindead idea to enable https for opkg repos by default
<jow>
crap like that was bound to happen
<stintel>
oh I had my fair share of cursing today due to that LE crap
<jow>
solution: sed -i -e 's#https:#http:#g' /etc/opkg/*.conf
<PaulFertser>
Not everybody knows packages are properly signed.
<stintel>
jow: but https good, http bad!
<mangix>
I would hope openwrt servers do not to HSTS
<mangix>
*do
<jow>
mangix: of course they do, because mar crypto much good
<jow>
there's various work around and solutions
<mangix>
even for packages?
<digitalcircuit>
Ansuel: Belatedly, thank you for your advice for the IPQ8065 crash/reset issue on the mailing list! After a trip, I've recently gotten back to tinkering with OpenWRT, and I hope to have an update in the upcoming weeks.
<jow>
client side (at least for openssl 1.0.x users), people need to throw out DSG Root CA X3 out of ther system cert store
<jow>
that should allow openssl 1.0.x to verify normal LE certs again
<jow>
server side we need to ensure that we bump our acme client infra to latest and request alternative chain certs for LE that do not use cross signing
<jow>
*X3 cross signing, to accomodate for the android zombies
<jow>
stintel: personally I've spend more time debugging (not working) ACME auto renewals than I ever did requesting and installing certs manually before LE existed
<aparcar[m]>
jow use caddy 😜
<stintel>
jow: to be honest the acme setup on our infra makes my head explode. it's overcomplicated
<jow>
stintel: its needed to avoid having an http port for every domain
<jow>
and to request wildcards
<stintel>
jow: there are better solutions for that (dns-01)
<jow>
stintel: which we use
<Ansuel>
digitalcircuit: hope we can find the real reason of ipq806x instability
<stintel>
jow: yeah, but less complicated :P
<jow>
how so?
<stintel>
jow: the fact that there are 2 different plays just for acme is ... ehm, wtf?
<jow>
stintel: ah you mean the "let's grant random tools access to the cloud infra so that they can autoprovision their stuff at will"-kind of simpler
<jow>
yeah sure, that would work
<jow>
could also simply do a managed k8s on aws and do cert-amanger in there
<Ansuel>
mangix: are we sure we didn't swap port5 with port6 and the switch is actual qca8337 instead of qca8327 ?
<stintel>
jow: random tools?
<jow>
"lego"
<Ansuel>
mangix: sorry i mean port6 with port5
<aparcar[m]>
can't you use certbot and let it do everything?
<Ansuel>
acme.sh ?
* jow
sighs
<Ansuel>
jow is angry
<jow>
well guys, just install caddy and run certbot to do everything
<jow>
and sprinkle a bit of https powder on top to make the crypto nerds happy
<jow>
I'm out
en4rab_ has quit [Remote host closed the connection]
<stintel>
jow: to me lego is as random as acme.sh, maybe less because I've actually used it and read some of the source
<jow>
by all means, feel free to convert it to lego
<jow>
or whatever client you guys are happy with
<stintel>
no experience with lego and DO, but in Azure you can create a service principal that can only change DNS records for zone X
<stintel>
and if those creds are used from our VMs I don't see the problem
<stintel>
if someone manages to access our infra we have a bigger problem
<jow>
acme-dns was the next best provider independant solution that didn't required elaborated name server setups or third party providers to have a programmatically manageable NS zone
<jow>
of course it is "simpler" to just define an API token and use a random client that speaks DO/AWS/Azure natively
<jow>
the same way it is simpler to just use Github compared to setup getweb + gitolite
<jow>
if that's what you're comfortable with then fine with me
<stintel>
well that's why we need an infra committee in the first place
<stintel>
so we can talk about things like this without having to involve 20 other people who know nothing about maintaining infra
<aparcar[m]>
certbot works with nginx to respond a temporary challenge which is consumed by the LE backend, I don't see how this is a lot of magic or converts HTTPS to magic powder. Am I missing something?
<stintel>
and while the README in ansible documents briefly how to use it, it doesn't explain how it works behind the scene, so issuing a new cert is one thing, fixing a problem in the setup it a different thing entirely
<jow>
aparcar[m]: yes, you miss the fact that wildcard certs require dns-01 verification, not http based one
<jow>
aparcar[m]: and the fact that not all services we might want to request certs for actually do have any http server
danitool has joined #openwrt-devel
<jow>
so you want dns based verification, which requires a DNS setup which you can programmatically manage (adding, removing TXT records and potentially CNAMEs)
* Habbie
looks up
<aparcar[m]>
I wasn't aware we use (and need) wildcards nor we have any non-http services using https
<Habbie>
CNAMEs are a great trick for LE, yes
<stintel>
aparcar[m]: SSL certs are not only used for https
<stintel>
SMTP is one example
<aparcar[m]>
jow: thanks for elaborating
<stintel>
IMAP/POP3 the next
<stintel>
can go one like this for a while
<stintel>
since Habbie is watching
* stintel
is running a custom PowerDNS-based dyndns setup for his LE :P
<Habbie>
:)
<stintel>
rsalvaterra: my small army of RPis is running 5.10.70, thanks for taking care of these bumps!
<Ansuel>
small army ?
<stintel>
I'm exaggerating, only 9
<stintel>
possibly 2 more coming up; we're doing automated testing of a captive portal solution with rpis but that's on raspbian, much bloat, I'm considering moving that to OpenWrt with automated builds
<stintel>
actually started working on that today but ended up yak shaving for a change
<rsalvaterra>
stintel: "only 9". That's an effin' battalion in my book. :P
<stintel>
I just lack one more pi3 for bcm2710 (both rpi3 are capturing adsb and for some reason these dongles don't work on 64bit kernel)
* rsalvaterra
is old enough for Beowulf cluster jokes.
<stintel>
ooh, migrated my power8 from gentoo hardened to gentoo hardened musl, and my ansible play finished o_O
<stintel>
surprisingly little breakage
<rsalvaterra>
Oh? There's a musl-flavoured Gentoo?
<stintel>
for many years
<stintel>
my army of server VMs is running that for ~4 years
<rsalvaterra>
…
<rsalvaterra>
I…
<rsalvaterra>
Can't. Must… resist…
<stintel>
ppc64le not that long though
<stintel>
yeah, I've started it as an experiment and now I can't do anything else anymore
<Ansuel>
why musl is so much loved?
<stintel>
because glibc is bloat
<Ansuel>
even for normal system?
blogic has joined #openwrt-devel
swegener has quit [Remote host closed the connection]
swegener has joined #openwrt-devel
blogic_ has quit [Remote host closed the connection]
<stintel>
total used free shared buff/cache available
<stintel>
Mem: 358 53 30 0 274 294
<stintel>
try that with a modern glibc-based system
<stintel>
it's not because we have plenty of RAM and CPU that we need to waste it
<paper_>
void has 2 editions - glibc and musl, so we can compare and the difference exists, but not too big
<paper_>
the musl version is fun for finding bugs in software
<stintel>
-rw-r--r-- 1 root root 4.5M Aug 22 23:03 /var/cache/binpkgs/sys-libs/musl-1.2.2-r3.tbz2
<slh>
mangix: does the swconfig package need to be removed from DEFAULT_PACKAGES at this point (isn't it just dead-weight on qca8k supported devices, without any ill effects)? (retaining support for the not-yet-ported devcies)
<mangix>
yes it does. I explained why.
<slh>
thinking about e.g. tl-wr1043nd v1 (rtl8366rb, which would need porting to realtek_smi)
<mangix>
If swconfig can be patched to not interfere with qca8k, it can stay.
<mangix>
or ar8216, w/e
Ansuel_ has joined #openwrt-devel
<mangix>
Ansuel_: welcome back
<slh>
mangix: I'm talking about swconfig the package and CONFIG_SWCONFIG=y && CONFIG_SWCONFIG_LEDS=y, not CONFIG_AR8216_PHY=y
<Ansuel_>
mangix: hello for some reason ath10k decied to die... did i miss something?
<mangix>
if it's not supported, why is there a compatible string, LOL
<rsalvaterra>
I was looking for that line. :)
<rsalvaterra>
I hope that SMI companion driver isn't needed for DSA support.
<rsalvaterra>
mangix: Because FIXME? Should probably be TODO instead.
<mangix>
sounds like it should Ansuel should have wrote FIXME when submitting his LED patch, LOL.
<Ansuel>
FIXME: generilize leds for the entire phy framework
<Ansuel>
how did i miss this magic word
<mangix>
git grep FIXME | wc -l
<mangix>
4051
<mangix>
wow
<rsalvaterra>
Ansuel: ath10k dies periodically for no obvious reasons. Nobody even cares anymore, it's just what it does. :P
<Ansuel>
i'm curios to check the debug and submit it to ath10k-ct
<Ansuel>
unless the entire router is dead
<rsalvaterra>
You should ask stintel what he thinks of ath10k. ;)
<slh>
rsalvaterra: I've been successfully using ath10k-ct with and ath10k qca9980 and qca9984 for quite a while, don't even remember when I last had problems with those
<mangix>
i wonder if ben moved on honestly. I see him patching mt76 nowadays.
<Ansuel>
currently porting qca8k to FILED_PREP for no exact reason so that has priority than ath10k shitty driver
<rsalvaterra>
slh: Those are wave 2. Try a bit of QCA9880 wave 1, for the lulz. ;)
<mangix>
Ansuel: so this archer I have has a minipcie slot. I can replace shitty ath10k card. But there are no good replacements :(.
<Ansuel>
oww and you don't have any clue about the state of ath11k.... we are waiting a miracle for backports package to switch to 5.15 cause it's just a waste of time port 40 patch
<rsalvaterra>
mangix: Because the diplexers are in the card itself, right?
<slh>
rsalvaterra: I only have qca9880-br4a in my bthub5, haven't used that one in anger for quite a while (need to experiment with a dsa on that one soon, though)
<Ansuel>
don't know i see a good community with the wifi6 mediatek card...
<rsalvaterra>
Well, you can put a MT7615 there. Or MT7915, if going for ax.
Tapper has quit [Ping timeout: 480 seconds]
<mangix>
Ansuel: there are no real mt76 minipcie cards. they're mostly talking about builtin ones.
<Ansuel>
well.. then you are stuck
<mangix>
it's connected through pcie, but it's all soldered to the board.
<mangix>
the turris people are testing an mt76 card courtesy of asiarf but they've released nothing.
<stintel>
ath10k-ct worked quite well for me also, until new client sent me a macbook pro and a galaxy s9, as soon as these devices connected the firmware started shitting itself 20 times per minute
<stintel>
same or similar issue as one I had seen sporadically, and one that Ben can't fix, for whatever reason (I don't blame him)
<Ansuel>
qcom proprietary shit having fun with qcom proprietary firmware... ath10k-ct have many part internally reworked so could be that they trigger some proprietary stuff and fails
<stintel>
so ath10k and by extension qca is dead to me
<rsalvaterra>
stintel: God rest its soul.
<stintel>
the reason why I pushed for ath10k-ct as default is because vanilla ath10k was way worse, it used to crash without recovering
<rsalvaterra>
I miss Atheros. Without the Q.
<stintel>
and nobody cares, bug reports are ignored, etc
<stintel>
so yeah, fuck qca, in the arse, with a cactus
<rsalvaterra>
No. Gently. With a chainsaw. :)
<stintel>
heh
<mangix>
anyway, look at the linux-wireless patchwork. it's all mt76.
<rsalvaterra>
Anyway, I too have experienced first hand what stintel just described.
<mangix>
everyone's migrating to mt79xx
<stintel>
well let's hope a decent mtXXXX AP comes out soonish
<stintel>
by decent I mean at least 2.5GbE PoE+ uplink, with RJ45 console port
<rsalvaterra>
With wave 1 hardware, mind you. Wave 2 seems much more stable, granted. I've never seen a firmware crash on my C6 v2.
<stintel>
my qca wave2 AP is twice as slow as the wave1 AP or so
<Ansuel>
for me r7800 was more stable than wrt3200acs for example... i solved all my problem with ath10k-ct
<stintel>
so yeah nope. not a good experience
<rsalvaterra>
Ansuel: Ah, Marvell Avastar. Well, when you set the bar so low… :P
<stintel>
yay, another 802.11g client replaced by something that does at least 802.11n
<Ansuel>
about ath10k-ct being slow... there is a big offload chunk of code that is disabled on ath10k-ct... i tried to spam ben and ask to fix it... we only manage to fix a firmware panic when the mode was enabled...
<Ansuel>
he said he never tested that part of the code... sad
<Ansuel>
(talking on firmware level... nothing related to the driver source)
<stintel>
Habbie: as I said, I'll be close to NL soon so could even drop them :)
<Habbie>
stintel, and if the price is 'come to belgium and have dinner with you' (not even talking about who pays), why not
<Habbie>
or that
<Habbie>
i'll cook
<stintel>
ooh
<stintel>
now I'm interested :D
<Habbie>
haha
<rsalvaterra>
stintel: Weird. I like it.
* stintel
foodie
<Habbie>
stintel, i don't know how much of a detour Hilversum is
<Habbie>
stintel, but i also have a car
<Habbie>
and i love meeting people
<Ansuel>
this chat LOL
<rsalvaterra>
stintel: Well, if you want to come to Portugal… ;)
<Habbie>
Ansuel, nice huh :)
<Habbie>
rsalvaterra, oh i do want that :D
<stintel>
Habbie: I'll be close to Brussels
<stintel>
rsalvaterra: that's quite the detour :P
<Habbie>
stintel, yes, i believe it's called Belgium
<Habbie>
stintel, :D
<stintel>
hahaha
<Habbie>
or as we like to say
<Habbie>
'drive south, and then when the roads really start to suck, it is only another 90 minutes'
<rsalvaterra>
stintel: Can't to anything about it, it's literally the westernmost country in continental Europe. :P
<stintel>
Habbie: it's less than 200km according to maps, and my brother is living just across the border somewhere north of antwerp, so could combine these things :)
<Habbie>
yeah
<Habbie>
i've done conferences in Antwerp while sleeping at home :)
<rsalvaterra>
Speaking of sleeping, I'm off to bed. Cheers, guys! :)