SpectreDev_01 has quit [Quit: Connection closed for inactivity]
dangole has quit [Remote host closed the connection]
philipp64 is now known as Guest153
philipp64 has joined #openwrt-devel
Guest153 has quit [Ping timeout: 480 seconds]
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
floof58 has quit [Ping timeout: 480 seconds]
floof58 has joined #openwrt-devel
<philipp64> I'm trying to backport some igc driver fixes to 5.15 but... they're not dropping in and I'm not sure why. the existing patches don't overlap drivers/net/ethernet/intel/igc/ ...
tSYS has quit [Quit: *squeak*]
tSYS has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
Danct12 is now known as Guest169
Danct12 has joined #openwrt-devel
<philipp64> anyone around with strong jshn-fu?
floof58 has quit [Ping timeout: 480 seconds]
floof58 has joined #openwrt-devel
sauce has quit [Ping timeout: 480 seconds]
dgcampea has quit [Remote host closed the connection]
dgcampea has joined #openwrt-devel
Danct12 is now known as Guest184
Danct12 has joined #openwrt-devel
Guest184 has quit [Ping timeout: 480 seconds]
gladiac has joined #openwrt-devel
xavifr has joined #openwrt-devel
Danct12 has quit [Quit: WeeChat 4.0.4]
hanetzer1 has joined #openwrt-devel
hanetzer has quit [Ping timeout: 480 seconds]
xavifr has quit [Ping timeout: 480 seconds]
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
Tapper has joined #openwrt-devel
xavifr has joined #openwrt-devel
robimarko has joined #openwrt-devel
tidalf has quit [Remote host closed the connection]
tidalf has joined #openwrt-devel
tSYS has quit [Ping timeout: 480 seconds]
hanetzer1 has quit [Ping timeout: 480 seconds]
Danct12 has joined #openwrt-devel
hanetzer has joined #openwrt-devel
Danct12 has quit [Ping timeout: 480 seconds]
tSYS has joined #openwrt-devel
Danct12 has joined #openwrt-devel
<ynezz> Hi, there seems to be some regression with wlans on ipq40xx/glinet-b1300 introduced between 09d13cd8d8 (good) 6cf27094e9 (bad), both wlans wlan0/wlan1 are missing https://gitlab.com/ynezz/openwrt-device-runtime-testing/-/jobs/5096165288#L83
gladiac has quit [Quit: k thx bye]
hanetzer has quit [Ping timeout: 480 seconds]
hanetzer has joined #openwrt-devel
Danct12 has quit [Ping timeout: 480 seconds]
rua has quit [Quit: Leaving.]
<KGB-2> https://tests.reproducible-builds.org/openwrt/openwrt_tegra.html has been updated. (100.0% images and 99.9% packages reproducible in our current test framework.)
tidalf has quit [Ping timeout: 480 seconds]
robimarko has quit [Remote host closed the connection]
robimarko has joined #openwrt-devel
<hanetzer> stintel: eyo. around? apparently the second tp-link eap615-wall I bought was v 1.8 and I'm not certain bout the serial port/uart on it; every time I hook it up it powers off the machine lmao.
tidalf has joined #openwrt-devel
goliath has joined #openwrt-devel
hanetzer has quit [Remote host closed the connection]
hanetzer has joined #openwrt-devel
hanetzer has quit [Read error: Connection reset by peer]
hanetzer has joined #openwrt-devel
hanetzer has quit [Ping timeout: 480 seconds]
tidalf has quit [Remote host closed the connection]
tidalf has joined #openwrt-devel
minimal has joined #openwrt-devel
hanetzer has joined #openwrt-devel
hanetzer has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<Ansuel> ndb ping ?
xavifr has quit [Ping timeout: 480 seconds]
xback has joined #openwrt-devel
philipp64 is now known as Guest239
philipp64 has joined #openwrt-devel
Guest239 has quit [Ping timeout: 480 seconds]
<xback> nbd: I just tested the mac80211 6.5 dump on a Mikrotik Wap-AC device.
<xback> while both radio's are discovered by ath10k-ct, there are no "wlan" devices available after boot
<xback> On this device, the radio's are probed as AHB device iso PCI devices
<xback> (it's a ipq40xx target)
<xback> "wlan devices" --> "wlan interfaces"
<EqUaTe> xback: sounds related to what ynezz was mentioned earlier.
<xback> EqUaTe: ah? what did he mention?
<EqUaTe> [10:23] <ynezz> Hi, there seems to be some regression with wlans on ipq40xx/glinet-b1300 introduced between 09d13cd8d8 (good) 6cf27094e9 (bad), both wlans wlan0/wlan1 are missing https://gitlab.com/ynezz/openwrt-device-runtime-testing/-/jobs/5096165288#L83
<EqUaTe> [10:23] * gladiac (~gladiac@77.116.207.120.wireless.dyn.drei.com) Quit (Quit: k thx bye)
<EqUaTe> [10:36] * hanetzer (~hanetzer@206.217.205.16) Quit (Ping timeout: 480 seconds)
<EqUaTe> [10:47] <ynezz> same situation with mt7622/belkin-rt3200 https://gitlab.com/ynezz/openwrt-device-runtime-testing/-/jobs/5096165290#L87
<xback> Thanks!
<EqUaTe> sorry, missed you joining, thought you'd already been in the channel at the time :)
<EqUaTe> I should hasten to add that when I say it sounds related, I mean that it sounds like the same symptom. Could be completely different though.
minimal has quit [Quit: Leaving]
<nbd> xback: i disabled the auto-created wlan* devices
<nbd> xback, ynezz: these auto-created devices are useless, and they get removed on a configured radio after wifi down anyway
<Ansuel> nbd do you have time to have a fun talk about stmmac?
<nbd> not sure how much time i have, but go for it :)
<Ansuel> i'm checking the problem and it really seems to be the timer thing is just useless... like i'm not following why napi shouldn't already handle it with the budget logic
<nbd> apparently some people are bothered by having too many interrupts
<nbd> which might potentially be an issue on battery operated devices
<Ansuel> still in the current state it really seems to be the timer conflicts with napi serulting in dma getting enabled and disabled a lot of times
<nbd> so the guy on netdev@ that we've been discussing this with posted a patch that at least allows disabling the timer via ethtool
<xback> nbd: so now that they are gone, how should I interface with wireless? :-)
<nbd> so my suggestion would be to backport that and add an openwrt specific patch to tune the defaults
<nbd> xback: configure it in /etc/config/wireless, as usual
<nbd> it's really the same state as on a configured radio after wifi down
<xback> aaah, I see
<Ansuel> nbd default would be 0, but looking at the code it really seems there is a chance to fix it... like why we arm the timer if napi will poll again? (case when poll function return > budget )
<nbd> xback: what were you even doing with the auto-created devices before?
<nbd> Ansuel: i think with the fix you can prevent the timer from being armed
<nbd> set tx-frames to 1, and usec to 0
<xback> nbd: they were fully reconfigured with an own application. We are not using openwrt wireless config file over here
<robimarko> xback: You can then just create them with iw
<nbd> iw phy1 interface add wlan1 type managed
<nbd> brb
<Ansuel> have to check if I manage to repro the regression case... well time to compile some code..
<Ansuel> why web is empty of device with serial + eth port on usbc ....
<nbd> xback: so what kind of application are you using to configure your wireless devices?
<mrkiko> Ansuel: yeah finding devices with readily accessible serial console isn't easy, tough they exist
<Ansuel> anything to suggest?
<mrkiko> Ansuel: what target do you need? ramips/???
<xback> nbd: i've build a binary which controls everything internally, wlan, userspace mesh, uart, gps, ais, radio monitoring, packet capture, qmi, etc, etc .. no dependencies, and using native netlink where possible
<nbd> interesting
<Ansuel> overkill o.o
<mrkiko> xback: and do you handle qmi reconnection as well ?
<xback> mrkiko: yes. I adapted the perl scripts from nbd to build c files out of the json data, so it also adds support for events and such
<mrkiko> xback: :D
<mrkiko> xback: well, it should be efficiente from a resoure perspective
<mrkiko> xback: that's a nice pro
<xback> mrkiko: that was the original intent years ago. our hardware was low-end at the time.
<xback> all these requirement were ment to work on a device with 8MB of flash nd 64MB of ram
<xback> so it was simply no possible to use external libs and tools
<xback> the binary is now roughly 720kB in size
<xback> it also contains a self written webserver, google earth KML server, etc etc
<mrkiko> xback: impressive :D
<xback> mesh was written from scrath, using rfc3626 as a base, and then extended with bidirectional rssi, added to the hello messages, and a few custom messages to exchange gps data and such
<Ansuel> nbd like why something like this wouldn't work? https://pastebin.com/t8MP9scF (notice the extra check right above the hrtimer call)
<Ansuel> should follow napi logic to reschedule if we have other stuff to do and save the timer if we have something to handle
<nbd> worth a try
<Ansuel> well already did some iperf run also with --bidir and seems to be working just fine
<Ansuel> (300 mbps at 200 mhz)
<Ansuel> (don't ask why i have an ipq806x running at 200mhz, just don't)
<nbd> :)
<nbd> regarding wifi, i'm happy that i finally got to the point where config reload is working much better
<nbd> next step is to allow ifdown of an interface to bring down wifi-ifaces belonging to that interface
<nbd> without affecting any unrelated interfaces
<Ansuel> did you discover some big bug in improving it?
<nbd> i spent 90% of my work time this week on debugging hostapd stuff
<nbd> :)
<Ansuel> :D
<mrkiko> nbd: thanks a lot for the great work
<nbd> you're welcome
<nbd> i'm really happy with how it turned out
<nbd> and that it can even deal with things like ifname changes because of section reordering
<nbd> and generated mac address changes
<robimarko> I have seen that you have just been cranking out changes
<nbd> so hostapd/wpa_supplicant/wdev.uc will simply switch around mac addresses and ignore the ones from the .conf files if necessary
<nbd> but only if the user didn't explicitly configure an address
<nbd> so when i'm done with the netifd changes, you will be able to do things like assign wifi-ifaces to extra interfaces with proto=none pointing to the same bridge as other interfaces
<nbd> and then control the up/down state of that wifi-ifaces through the extra interface
goliath has quit [Quit: SIGSEGV]
hanetzer has joined #openwrt-devel
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
<KGB-0> https://tests.reproducible-builds.org/openwrt/openwrt_ath79.html has been updated. (98.7% images and 99.9% packages reproducible in our current test framework.)
SpectreDev_01 has joined #openwrt-devel
danitool has joined #openwrt-devel
goliath has joined #openwrt-devel
danitool has quit [Ping timeout: 480 seconds]
<philipp64> nbd: what would be involved in fixing _jshn_append to not add the leading space before the first entry? would it break anything?
<nbd> i don't think it would break anything. this might work: https://nbd.name/p/9c9b6a8e
<nbd> but needs testing
<philipp64> Needs a backslash before the first added $
<nbd> right
<philipp64> otherwise works at first blush
<Ansuel> curious if this actually works correctly
<nbd> Ansuel: another idea might be to only arm the timer if !test_bit(NAPI_STATE_SCHED, &napi->state)
<Ansuel> stmmac_tx_timer already have if (likely(napi_schedule_prep(napi))) {
<Ansuel> that does just that check
<Ansuel> so myabe an idea would be to drop the check there and move it to the arm function
<nbd> yes, but timer rescheduling is expensive
<nbd> so it might be good to avoid unnecessary timer arming
<Ansuel> it's actually very clever !
<Ansuel> let me create a patch
<Ansuel> nbd mhhhh testing napi state is not easy... idea is to use napi_schedule_prep but no idea if correct
<Ansuel> there is napi_if_scheduled_mark_missed but it does set it as missed... and doesn't look correct
<Ansuel> (can't find an api that just check the value and nothing else)
<nbd> just use the test_bit directly
<Ansuel> is it ok without any locking ?
<nbd> test_bit is atomic
<Ansuel> ok
<nbd> i'm not 100% sure if it's race free
<nbd> but if it races, the race should be harmless
<Ansuel> napi_synchronize use test_bit and then msleep soooo maybe it's ok
<Ansuel> found meme driver that just does what we need
<Ansuel> wonder if the change is that simple
<Ansuel> in any case it's clear that the timer is armed most of the time for no reason at all
SpectreDev_01 has quit [Quit: Connection closed for inactivity]
<Ansuel> nbd i'm having mixed results with it but iperf tests looks to be working just right
<Ansuel> mixed results in the sense that the thing looks to work very well
<nbd> would be nice to compare it against having timers disabled entirely and tx-frames set to 1
<Ansuel> just doing that
<Ansuel> also the pr have a few guy that made good stats
<nbd> btw. the netifd feature to toggle individual wifi-ifaces via their interfaces is working now
<nbd> once things have settled down a bit, i'm planning on backporting the whole shebang to 23.05
<Ansuel> we should really compose a list of what is missing for 23.05
<Ansuel> like a pinned issue to track things
<philipp64> nbd: we don't store the name of an object into itself in jshn.sh, do we? K_{foo} points to the indices inside an array if it's an array, and the names of the members if it's a hash, right? But there's no way to retrieve the name of a node from itself, just from its parent... but its parent can have multiple children... and even then, they might be unnamed objects...
goliath has quit [Quit: SIGSEGV]
philipp64 is now known as Guest263
philipp64 has joined #openwrt-devel
Guest263 has quit [Ping timeout: 480 seconds]
robimarko has quit [Quit: Leaving]
goliath has joined #openwrt-devel
<philipp64> nbd: opened a PR with some enhancements to jshn.sh but it requires that patch above be applied first...
<dwfreed> why not include that patch in your PR
<philipp64> dwfreed: okay, done... didn't want to take credit away from Felix...
rmilecki has quit [Quit: Konversation terminated!]
rmilecki has joined #openwrt-devel
rmilecki has quit []
rmilecki has joined #openwrt-devel
<dwfreed> philipp64: I mean, you can make him the author of the git commit
<Slimey> mother nature sent up the ether/switch killer :\