Piraty has quit [Remote host closed the connection]
Piraty has joined #openwrt-devel
<stintel>
the usual marketing crap ;)
* stintel
remembers certain atheros draft N stuff
<stintel>
avoid!
<stintel>
hmmm my U6LR dropped offline after changing its autoneg to 100baseT Full
<stintel>
even though the other end shows it properly negotiated that speed/duplex mode
Namidairo has joined #openwrt-devel
<f00b4r0>
stintel: speaking of (after a good night's sleep :^), I noticed that after boot the M300 dmesg is flooded with fsl_dpaa_mac error messages: is this expected?
<f00b4r0>
for uboot env i recall following the commit instructions to the letter
<f00b4r0>
fsl_dpaa_fman.fsl_fm_max_frm=1530
<stintel>
yeah this doesn't seem related to the DSA problem due to too low MTU
<f00b4r0>
is this it?
<stintel>
yeah
<stintel>
kernel ?
<f00b4r0>
5.10.146 (22.03.2)
hanetzer has joined #openwrt-devel
<stintel>
bah
<f00b4r0>
?
<stintel>
maybe some kernel bump messed something up
<stintel>
I'm on master/5.156
<stintel>
5.15*
<stintel>
and I've not seen this on 5.10 before, so suspect some kernel bump
<f00b4r0>
hmm
<f00b4r0>
would you by any chance remember which was the last 5.10 kernel you ran?
<f00b4r0>
see if I can bisect this
<stintel>
you're not using the device in production yet right?
<f00b4r0>
no but that is my goal if I can run a stable release ;)
<stintel>
I don't see any recent changes in 5.10.146 in the fman driver
<f00b4r0>
incidentally these messages are a one-off occurence. They don't repeat
<stintel>
could you try master with TESTING_KERNEL ?
<stintel>
or wait
<stintel>
let me plug in serial console on my backup router and verify I'm not seeing them myself
<f00b4r0>
the device appears to be working normally (I'm connected over network)
<stintel>
well it appears to have something to do with enabling multicast or something
<stintel>
rebooting my backup router, hold on :)
<f00b4r0>
;)
<oliv3r[m]>
<robimarko> "So eth0 is the DSA master..." <- I belive that's what we mean :) e.g. switch lan1 - lan 28 (physical) == port 0 - port 27 (on the mac) and eth0 ('physical'-ish) is port 28 (on the mac), where we 'call' port 28 the 'cpu port'. DSA then translates this into the 'master interface' I would reccon indeed. we CAN control everything nicely of the mac on the switch chip, but the mac on the CPU port, I have not found any registers
<oliv3r[m]>
for this yet.
<oliv3r[m]>
E.g. afaik, traditionally oyu have MAC -MII -> PHY -> copper -> PHY -> MII -> MAC. Obviously they've done MAC -> MII -> MAC; but it could even be 'shorter' where they removed even more logic and just do MAC -> MAC or ... CPU -> MAC -> Switch? Would that even be possible? (or a shared MAC of sorts)
<stintel>
nope not seeing any of those, a bunch of messages from mv88e6085 though - this is 5.15.74
<stintel>
could you test master with testing kernel to see if it's not a config thing?
<f00b4r0>
sure
<stintel>
thanks
* f00b4r0
starts a build
<f00b4r0>
that reminds me I probably want to swap the sd card that came with the device for something new, just to be safe :)
<stintel>
f00b4r0: what's the specs of the card that came with yours?
<stintel>
brand/size/.. ?
<f00b4r0>
have to check. Transcend iirc
<stintel>
ok in that case I did replace mine ;)
<f00b4r0>
hahha
<stintel>
I've Sandisk 16GB
<f00b4r0>
yeah that's what I typically use everywhere. Very reliable.
clandmeter has quit [Quit: Alpine Linux, the security-oriented, lightweight Linux distribution]
danieli has quit [Quit: Alpine Linux, the security-oriented, lightweight Linux distribution]
clandmeter has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
<stintel>
Failed to exec upgraded.
<stintel>
wtf
<olmari>
Old topic: sadly phone choices are 99.99% selecting least shit.. there is no best or mostly even rhat good for my tastes
Guest696 has quit [Quit: Page closed]
<f00b4r0>
stintel: sorry I'm slow with the test, I'm wrestling my config to get to a stage where it does build (further evidence that I don't want to run master ;P)
<stintel>
no worries
<stintel>
I'm fighting U6LR issues myself
<stintel>
traffic breaks at 100Mbps link speed (both half and full duplex), trying to sysupgrade from 22.03 to master resulted in "Failed to exec upgraded." and a hanging console
<f00b4r0>
ugh
Ansuel has joined #openwrt-devel
<stintel>
Ansuel: quassel on oracle cloud? :)
<stintel>
great, failed to exec upgraded again
<Ansuel>
stintel ah ah ah.... still trying to signup -.-
<stintel>
Ansuel: :/
<f00b4r0>
stintel: hmm master sysupgrade doesn't want to upgrade 22.03 (1.0 -> 1.1). I thought we had backported all changes to 22.03
<stintel>
f00b4r0: yeah but I never figured out how to add compat_version to /etc/config/system
<f00b4r0>
ah :)
<stintel>
so it has to be done manually
<Ansuel>
stintel it looks like they introduced some type of limited registration and if that limit is reached you just get "failed to process transaction" when they need to verify your card.... probably to handle the abuse of user selling free account
<Ansuel>
(the best thing is that if signup fail you have to retry and insert all the data all over again)
* stintel
is happy without cloud
castiel652 has joined #openwrt-devel
Ansuel_ has joined #openwrt-devel
Ansuel__ has joined #openwrt-devel
Ansuel_ has quit [Read error: No route to host]
Ansuel_ has joined #openwrt-devel
<Ansuel_>
just tried again and looks like i'm banned from the card verification lol.....
Ansuel has quit [Ping timeout: 480 seconds]
castiel652 has quit [Quit: Leaving]
<stintel>
heh, so we have both conloglevel and cronloglevel in /etc/config/system
<stintel>
I removed conloglevel thinking it was a typo
<stintel>
*facepalm*
Ansuel__ has quit [Ping timeout: 480 seconds]
robimarko has joined #openwrt-devel
<stintel>
probably better to name it consoleloglevel to avoid that confusion
<stintel>
or even better console_loglevel and cron_loglevel
<Ansuel_>
ahahahhaha
<Ansuel_>
golden stuff
castiel652 has joined #openwrt-devel
dangole_ has joined #openwrt-devel
dangole_ is now known as dangole
<stintel>
dangole: FYI unifi 6 LR traffic on eth0 breaks when setting to 100Mbps full or half duplex
<stintel>
on 22.03, trying to repro on master but sysupgrade appears broken
<stintel>
guess it's going to be one of those yak shaving days
<dangole>
stintel: I'm aware of that, unfortunately it has always been like that. The same is true for Puzzle M90x mvebu devices which also come with Aquantia AQR112 PHYs which got the same problem (AQR 10GE PHYs do NOT have this problem)
<robimarko>
dangole: what is the link type between the PHY and MAC?
<robimarko>
HSGMII/SGMII+ or?
<dangole>
stintel: as it was IEI-World themselves submitting support for their devices I have told them about this problem and keep pinging them about it every couple of months. They told me they asked Marvell/Aquantia what to do about it or if they would be allowed to share datasheets and/or reference code with me. They keep telling me Marvell doesn't reply to them.
<dangole>
robimarko: 2500Base-X
<robimarko>
So, its AQR112?
<stintel>
can we blacklist 100Mbps in the DTS somehow ?
<robimarko>
I have acess to AQR datasheet, but usually bugs like those are in the FW
<dangole>
robimarko: yes, it's AQR112, and it doesn't work for both 100M and 10M modes. The PHY correctly reports link speed and that the link came up, but no data passes (supposedly PHY-MAC link is configured wrong when operating in <1G speed, I suppose that's also your guess)
<robimarko>
I had something similar on IPQ8074 and AQR113
<dangole>
robimarko: the AQR firmware on the UniFi 6 LR is stored on a physical EEPROM next to the phy. aq-firmware-dl tools works, but doesn't make any different to load firmware via MDIO.
<robimarko>
10G would work, but 1G would fail
<robimarko>
In the end it was due to clocks as MAC needs to change its mode once you go to lower speed
<dangole>
robimarko: interesting. AQR113 in the Puzzle-M902 works fine for 100M/1G/2.5G/5G/10G modes without any problems also in lower speeds... just AQR112 (which anyway supports only 2.5G max) only works in 1G and 2.5G modes.
<dangole>
robimarko: we could try to reused @lynxis code for RealTek 2.5G PHYs on MT7622 SoC which also require switching the phy-mode when running at lower speeds
* stintel
looks
<stintel>
I have my U6LR open with serial console at the moment so I can do some testing
<robimarko>
Ahh, damn, I have the Gen4 AQR datasheets, not AQR Gen3
<robimarko>
Let me request those
Ansuel_ has quit [Read error: Connection reset by peer]
Ansuel has joined #openwrt-devel
<dangole>
stintel: same here, I got U6LR v1 without case and serial pins populated hanging on a wall next to me, connected to ZyXEL XS1930-10HP switch...
<Ansuel>
robimarko only uniphy2 doesn't work o.o?
<Ansuel>
uniphy0 and 1 are found correctly?
<stintel>
dangole: ah cool
* stintel
looks into the sysupgrade problem
<dangole>
stintel: which sysupgrade problem?
<robimarko>
Ansuel: For me yeah, but somebody else pointed out AX3600 having an issue as well
<dangole>
robimarko: I found this in Ubnt stock U-Boot sources: // 5.1.77 Autonegotiation Receive Reserved Vendor Provisioning 1: Address 7.E410
<dangole>
// enable aqr112c 100Mbps link detection
<stintel>
dangole: on the U6LR running image from openwrt-22.03
<dangole>
stintel: with stock loader or ubootmod?
<stintel>
dangole: stock
<stintel>
traffic breaks at 100Mbps link speed (both half and full duplex), trying to sysupgrade from 22.03 to master resulted in "Failed to exec upgraded." and a hanging console
<stintel>
(just repeated what I wrote before you joined)
<dangole>
stintel: cheers, will check IRC logs to follow up
<robimarko>
dangole: If that comment about the register name is correct
<stintel>
reading `man execvp`
<robimarko>
Then 0x0 is default for that register anyway
<dangole>
robimarko: it also simply doesn't change anything (I've tried doing this in U-Boot as well as aqr phy driver in Linux)
<robimarko>
dangole: Yeah, cause thats the default value
<dangole>
robimarko: makes sense then ;)
<robimarko>
But also note that if you change BIT(0) which is Disable
<robimarko>
100BASE-TX
<robimarko>
Parallel
<robimarko>
Detection
<robimarko>
Then autonegotiation must be restarted
<robimarko>
BTW, what is the AQR silicon revision?
<robimarko>
A0 or BO?
<dangole>
robimarko: I initially thought that 100M links would work with stock U-Boot, then realized a few months back that 100M also doesn't work on a bunch of new v2 devices I used in a deployment running stock loader and OpenWrt...
<stintel>
ah btw what's the proper way to distinguish v1 from v2 ?
<dangole>
robimarko: it's printed on the chip (extermely tiny, I will need optical tools to decipher) or there is a way to read the revision via MDIO?
<robimarko>
It can be read from the PHY
<robimarko>
It says MMD1.3
<dangole>
stintel: ubnt doesn't mark revisions on U6LR. v1 was only pre-orders and very few people actually got it in early 2021. v2 was mass-produced and is what most people received. now there seems to be v3 with 1G RealTek PHY instead of AQR112...
<\x>
same opps were set on the 5.15 one it just doesnt scale though
<dangole>
mrkiko: the nand controller built into mt7621 is for parallel SLC NAND. as the flash node of the r6220 is using that nand controller, it got to be parallel nand. seeing spi-nand bring used on mt7621 is quite rare, as the SoCs bootrom doesn't really support SPI-NAND -- however, some SPI-NAND chips bring compatibility with the SPI-NOR read command, so the bootrom can work there and you need a special preloader to load U-Boot...
<\x>
its all usable now minus that cpu scaling and usb, i tried enabling the usb2 node last time but that didnt work, seems it is enabling a different port, as for the usb3 one that one is undefined on the dts, defining it still didnt make it work, the system does know about it though https://i.imgur.com/3CpcBxM.png
<stintel>
dangole: yeah I just spotted the dma-requests warning in 22.03 on U6LR that's why I was asking, the less of those we have the lower the chances people might report an issue for it was my reasoning
<\x>
so i guess a lot of things needed to be backported or just go straight 6.1
<stintel>
checking lex output file root... configure: error: cannot find output from flex; giving up
<stintel>
pffff
<stintel>
I'm also giving up
<stintel>
it's one of those days
<castiel652>
nbd: Is it possible to do RX reordering with identical seqno? Or can just rely on the AMSDU_MORE flag like in mac80211? ath10k does deaggregation but the frames have identical seqno.
<robimarko>
\x: IPQ60xx is still a long from being usable without the hacks that are carried locally
<robimarko>
At this moment I dont have time for it as well
<hgl>
Is it a good idea to upgrade an openwrt device (without keeping the configs) by directly executing dd, which writes bin to the disk, on the running system and then reboot? I tried it and it seems to be fine, but I wonder if this method has any flaws?
shibboleth has quit [Quit: shibboleth]
<dwfreed>
hgl: sysupgrade can be told to not keep configs
<dwfreed>
generally better to do that, because sysupgrade can validate metadata
<hgl>
dwfreed, validating is nice. I guess I'll just stick to sysupgrade
<hgl>
thanks
MAbeeTT4 has joined #openwrt-devel
<f00b4r0>
hmm so 'bonding' proto always requires an ipv4 address.
<karlp>
linux shooting people in the foot hard again.
torv has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
<stintel>
pfff libgpiod
<f00b4r0>
heh
<stintel>
not a fan of this lib after it dropping contextless interfaces
<stintel>
it required non-trivial rewrite of stuff using it
<stintel>
so I ditched it completely and replaced raspberry pi motion sensor with esp32c3 running esphome
<Ansuel_>
karlp more like devs using deprecated stuff and getting surprised things gets droped after years
shibboleth has joined #openwrt-devel
<shibboleth>
so, to build "current master" for the 22.03 branch, my assumption was that pulling master and then "git checkout openwrt-22.03 & git pull" should be sufficient?
<dwfreed>
yes
<shibboleth>
for some reason git log then shows last commit 9dc46d65494c4956677694ea050512bf8971e03e
<shibboleth>
so, git pull what exactly? i've no black belt in git
<shibboleth>
remote=what?
<dwfreed>
origin...
<dwfreed>
(without the dots)
<aiyion>
I've got trouble finding gpio numbers for this mt7621 device. Usually I go with the blink script and search by cutting the space in half once I found.
<aiyion>
Last thing I did was adding uart2 to state default, I think to free it from its assumed purpose.
<aiyion>
Ist there a list of recommended strings I can add for nor devices, to find which gpios I curently can not control?
<aiyion>
This feels a little too much like blindly poking around for my taste
rua has joined #openwrt-devel
_zer0def has joined #openwrt-devel
_zer0def has quit []
zer0def has quit [Ping timeout: 480 seconds]
zer0def has joined #openwrt-devel
<f00b4r0>
robimarko: out of curiosity, do you have a recommendation for an openwrt-supported pcie wlan module for the mochabin?
Piraty has quit [Remote host closed the connection]
Piraty has joined #openwrt-devel
<shibboleth>
there's something wrong with tl-wpa8630(p) sysupgrade. it simply won't preserve config
<f00b4r0>
flash full?
<shibboleth>
736K free
<shibboleth>
of 834k total
<shibboleth>
isn't the jffs2 backup stored in ram during upgrade anyway?
<f00b4r0>
no
<shibboleth>
well, in any case 736k free of 824k total should... suffice, no?
<f00b4r0>
should, yes
<shibboleth>
ssh sysupgrade doesn't throw any warnings/errors, serial console is kinda risky with these devices (internal mains voltage due to powerline)
<shibboleth>
is the dev/maintainer (jwmullwalley something) on irc?
<shibboleth>
jwmullally
shibboleth has quit [Quit: shibboleth]
<robimarko>
f00b4r0: Not really, I just tried QCA6390
<robimarko>
But thats not yet supported
<robimarko>
I dont really have any MTK cards to try
<aiyion>
I found the two missing LED on my 7621 board to be 67 and 68. How do I tell, whether they were shadowed by the uart2 or the i2c node?
<f00b4r0>
robimarko: ok thx still. For mtk it seems the only alternative are AsiaRF's designs. I suppose the 7915 should be a fairly safe bet
<f00b4r0>
oh i see mt76 seems to support the 7916 even
<f00b4r0>
nbd: is mt7916 support stable?
<karlp>
Ansuel_: no, libgpiod is mainted on git.kernel.org and doesn't have a released version that uses the v2 interface. and kernel defaults the v1 interface to off. this is kernel people hating users IMO.
<karlp>
and then there's the "gpio-line-names" joke...
<karlp>
gpiochip has ports a..g, 32entries each, 200 somethign gpios.... I'm meant to make a complete array in the dt for the entire set if I want names...
<karlp>
but it's ok, I can use "" for the ones I don't need...
<karlp>
left hand not talking to the right hand.
<robimarko>
f00b4r0: Yeah, AsiaRF is good but expensive to import in EU
<robimarko>
As they charge hefty shipping and then there is VAT on top
<f00b4r0>
ah
<f00b4r0>
damn
<f00b4r0>
oh wow indeed
<f00b4r0>
shipping is as expensive as the card itself :P
<robimarko>
No idea how good those 7921 cards are, I know that 7915 ones are good
<f00b4r0>
cheap enough to give it a try. And it's free shipping :)
<dangole>
robimarko: mt7921 are meant for use in laptops, so supports only few virtual interfaces, AP mode has only been added recently to mt76 driver, but works fine. I got AMD RZ508, and in terms of range and power consumption I quite like those modules.
<schmars[m]>
ynezz i find your testbed work super interesting and might try and build a testbed myself - can i see somewhere how the uart-relay thing is set up? a usb switch and a couple of uart/jtag adapters seem straightforward, but i'm wondering about power on/off. was thinking to solve power control with just a poe switch and a matching poe splitter per device
hauke has quit []
hauke has joined #openwrt-devel
SlimeyX has quit [Remote host closed the connection]