<opty>
i don't use dfs, i don't need ieee 802.11{d,h} then, right?
borek has joined #openwrt-devel
robimarko has joined #openwrt-devel
<mrkiko>
robimarko: hi!!
<robimarko>
mrkiko: Hi
<mrkiko>
robimarko: I am trying (as per my mail) to submit a mvebuDTS upstream. No hurry and didn't want to pressure, just to know the mail arrived to it's destination :D
<robimarko>
mrkiko: Yeah, it arrived yesterday in the evening
<robimarko>
I will reply today
<stintel>
vulpes2[m]: I believe with raspberry pi zero as client, setting pmf to required on the client, will allow them to connect to PMF optional network
<vulpes2[m]>
so far I haven't had much luck with any rpi (or any broadcom products for that matter) + PMF and/or WPA3
<stintel>
I think WPA3 is no go but PMF works at least on some
<vulpes2[m]>
mt7612e also refused to connect to a WPA3 ap as a station, didn't test ap mode
Borromini has joined #openwrt-devel
Tapper has joined #openwrt-devel
wvdakker has quit [Ping timeout: 480 seconds]
<vulpes2[m]>
ESP8266 is also quite problematic when it comes to WPA3 and PMF. Only the RTOS SDK has been updated to support PMF, while the NONOS SDK (now EoL) doesn't.
<vulpes2[m]>
The latest SDK for the ESP32 supports both which is pretty nice at least.
<stintel>
not sure I've tried 11w on my ESPs
<vulpes2[m]>
It doesn't work if you use Arduino based stuff like Tasmota, because the Arduino support for the ESP8266 is built on top of the NONOS SDK and they have no plans to migrate last time I checked.
<vulpes2[m]>
This is why I generally recommend the ESP32 instead of the ESP8266 to people who wants to play with iot stuff.
<stintel>
I'm running esphome
<stintel>
and I phased out all but one esp8266
<mrkiko>
robimarko: no hurry...
<stintel>
once EVE Energy firmware with matter is there, I'll replace the Shelly Plug S and I will have only esp32c3
<f00b4r0>
arduino sdk is an abomination anyway ;P
<robimarko>
But it can do digitalWrite
<\x>
robimarko: i found some other regdb.bin, is it cross compatible to other boards?
<\x>
the client oriented mediatek wifi 6 cards doesnt do that too haha
<\x>
cant blame them though
<\x>
client oriented.
<robimarko>
Well, cause clients dont have to do DFS
<\x>
yup. they can open it but needs to do s/DFS// and itll work.
<\x>
so, technically illega
<\x>
l
<robimarko>
Why would it be illegal if you dont allow AP mode?
<Borromini>
robimarko: yes, that's related. But somewhere someone said MT7613 just didn't do DFS (from what I gathered it wasn't just driver support that was lacking)
<Borromini>
either way, I ditched my MT7613 stuff, was just to make my point
<\x>
well, MT7921, MT7921K, MT7921AU and MT7922 all allows AP mode. but single AP only. no DFS support though.
<robimarko>
Well, then you cant do DFS only channels
<\x>
they are all client cards, once you put up AP with them theres no beamformer too, only beamformee
<Borromini>
I do recall getting kicked off DFS channels regularly on MT76x2 and MT7615, be it false positives or city noise idk.
<\x>
whats nice on them though is theyre the cheapest way to play with 6GHz currently, also outstanding headacheless monitor/injection
<robimarko>
Good monitor mode is alone a reason to get them
<f00b4r0>
\x: I have an mt7921e client card that appears to support dfs, is it lying?
<\x>
f00b4r0: it just prints that on iw phy, try opening an AP on a dfs channel, itll fail.
<f00b4r0>
i'll test that
<\x>
for 5GHz you can open like 36~48 and 149~165 most of the times
<Borromini>
f00b4r0: AFAIK my MT7613 APs would show DFS support too until you actually tried it
<robimarko>
Its probably mt76 advertising the features universaly
<robimarko>
But then handles the calls per chipset
<\x>
ehhh its like advertises that "radar detection" not "NO-IR" on DFS, but once hostapd starts a cac scan, nope.
<\x>
intel is the one who does it correctly, it puts NO-IR on them
<robimarko>
Yeah, that is what I said
<robimarko>
The driver just advertises the feature universally
<robimarko>
But then once you try calling the handler it probably returnes ENOTSUP
<robimarko>
As then its checking whats the chipset and the appropriate handler for it
<\x>
the thing though robimarko, i think they are technically correct since once you do s/DFS// on db.txt itll work lmao
<robimarko>
Not really
<\x>
atleast on MT7921
<robimarko>
Properly it should advertise DFS after checking per chipset if its even supported
<f00b4r0>
good thing I don't need it for outdoor setup ;)
<stintel>
with 6E you can avoid that DFS pain :)
<stintel>
indoor at least, outdoor is another level of meh
<robimarko>
As long as WiFI gets crumbs of RF spectrum its not gonna get better
<f00b4r0>
i don't see any dfs mention for 6GHz freqs indeed
<robimarko>
There shouldnt be any
<stintel>
hopefully soonish there will be something like EAP615-Wall or U6 In-Wall, 3-band, mtk based
<stintel>
outdoor requires AFS, which afaik requires Internet uplink :P
<stintel>
AFC* sorry
<stintel>
so now bluetooth will start using 6GHz too
<stintel>
fffffuuuuuuuuuu
<Borromini>
stintel: are you using the 615 on a daily basis?
<f00b4r0>
robimarko: the bottomline is, 17€ to bring AP capabilities (if subpar) to the Mochabin was a no brainer ;)
<stintel>
Borromini: they're the only APs I'm using
<Borromini>
on master i suppose?
<stintel>
best WiFi experience ever
<stintel>
yes
<robimarko>
Ugh, AFC looks terrible
* f00b4r0
doesn't care very much about wifi anyway
<stintel>
robimarko: you hadn't heard about it?
<Borromini>
stintel: ok. I've seen regular timeouts on master mt76 (backported to 22.03)... >_>
<robimarko>
I did, but never bothered to look at what it is
<stintel>
robimarko: ah :)
<Borromini>
well, also on 22.03 mt76 >_>
<stintel>
yeah I've had some semi-brief look into 6E and came across AFC somewhere 5-6 months ago
<f00b4r0>
wow, reading an extremenetworks article about AFC, what broken mind came up with this kludge?
<robimarko>
stintel: So far I just saw bits and pieces over various QCA docs
<robimarko>
As they are readying that sh*t to somehow roll it out
<stintel>
:)
<stintel>
if you need outdoor probably stick to 5GHz :P
<robimarko>
Looks like worse sh*t that Passpoint
<f00b4r0>
stintel: looks that way yeah :P
<robimarko>
Or whatever other horrible things WFA makes up in their "specification"
<f00b4r0>
heh
<robimarko>
And then we end up with black-box with 2 providers globally
<stintel>
the nice thing is that indoor won't require AFC at all because attenuation is high enough to not interfere with outdoor 6GHz stuff
<stintel>
which will also make 6E amazingly useful for crowded residential areas
<f00b4r0>
i plan to stick to 2.4 as much as I can. I need range ;)
<stintel>
yikes
<robimarko>
RIP, I hate 2.4G
<stintel>
just add more APs ;)
<robimarko>
On AX its usable, but still crap
<f00b4r0>
stintel: not an option in outdoor situation ;)
<stintel>
I want to phase out WiFi on 2.4G
<robimarko>
I just keep it for IoT
<stintel>
leave that band for BT and Zwave/Zigbee/Thread
<f00b4r0>
and the other setup where I (well, "family") need wifi has 1m-thick walls. 5G already doesn't work there, so I doubt 6 will make things better ;P
<stintel>
drill a hole, pull an UTP ;p
<\x>
halow where when
<robimarko>
stintel: Do you know the state of AFC in ETSI?
cbeznea has joined #openwrt-devel
<f00b4r0>
stintel: at that point I'd rather have everything on wire, as it should be ;)
<stintel>
robimarko: no, haven't looked into it deeply
<robimarko>
\x: You know of any halow device that is in production?
<\x>
none yet, theres some yourube video though where oems are shilling their shit
<\x>
theres one where it can be used as a wireless bridge connecting two distant networks via halow
<robimarko>
I ask cause I wanted to try it out on my previous student job, where it looked perfect
<robimarko>
But its been years and years of coming soon TM
<\x>
but cards and things using it? nah, not in market, theres some crowdfunding shiz atleast
<robimarko>
But ETSI has reduced 6E spectrum available anyway
<\x>
<stintel> with 6E you can avoid that DFS pain :)
<\x>
sucks that my country isnt allowing it. spatial reuse really cant be realized with legacy clients being mixed in :X
<\x>
with 6E/6GHz those can be fully realized so it wont hurt that much if theres a lot of channel reuse happening
borek has joined #openwrt-devel
<opty>
low-bandwidth 5 GHz (i don't need more): n@20 MHz or ac@80 MHz?
<robimarko>
That aint really a fair comparison
<\x>
Whats this TPC TX power thing? is this a wifi 6 feature?
<robimarko>
AFAIR Its a 802.11h feature
<stintel>
no TPC is around for a long time
<robimarko>
Basically, it was meant for 2 devices to negotiate the minimum power they need
<\x>
is this the one lowering my tx power depending on the client/environment?
borek has quit [Read error: Connection reset by peer]
<robimarko>
But I am not sure how widespread it is
<\x>
seems neat
borek has joined #openwrt-devel
<stintel>
I often see this in iwlwifi
<stintel>
[1339999.132111] wlan0: Limiting TX power to 20 (20 - 0) dBm as advertised by b0:a7:b9:cb:ee:ba
<stintel>
[987139.197328] wlan0: Limiting TX power to 17 (20 - 3) dBm as advertised by b0:a7:b9:cb:ee:bb
<stintel>
b0:a7:b9:cb:ee:b{a,b,c,d} being my EAP615-Wall running OpenWrt
<robimarko>
Great to see it working
<\x>
seems ansuel hasnt been here in a while I want to bully him to bring up VHT/2.4 again, atleast reopen discussion on it either on a github issue or forum
<robimarko>
You can reach him on the forum or github
srslypascal is now known as Guest2113
srslypascal has joined #openwrt-devel
wvdakker has joined #openwrt-devel
<\x>
there seems to be a weird wired issue on the MR7350 on wired, but it only shows up after a long uptime robimarko. where it shows its connected at 1Gbit but on the other side its 100Mbit
<\x>
i dont think those cables ive ran are bad since those ran GbE fine with 40xx with pr-4721
<robimarko>
\x: Honestly no idea, I am not using that board nor platform currently
Guest2113 has quit [Ping timeout: 480 seconds]
<\x>
yeah
<\x>
seems 60xx is a box of chocolates with chili on them
<\x>
time to wait for a better specced 7981/7986 that has usb at least
<robimarko>
There is Redmi AX6000 that is supported, but no USB
<\x>
bpi r3 seems nice but after counting case and cables its kindaaaaaa
<robimarko>
BPi is nice, but its not an end product
<\x>
and its a proper board, I like it, it can have nvme
<robimarko>
BTW, for PCIe: Mini PCIe via USB
<\x>
maybe itll take some time for an R619AC replacement, I really liked it for having pcue
<\x>
ehh bpi r3 has a real nvme slot undernath
<robimarko>
Yeah, Key-M
<\x>
kind of a waste that they didnt split it though, it is set as 1x2 not 2x1
<\x>
fastest networking there is 2.5GbE so its not like a pcie x1 will be bottlenecked
<robimarko>
My biggest gripe with MT7986 is lack of 10G
<robimarko>
They have handicaped themselves from high-end again
<\x>
ehh, they might be doing what theyve done on mobile, slowly creep up to highend
<robimarko>
Well, they had low and mid tier before
<robimarko>
I really hoped they will try to high end end now
<stintel>
isn't this what they are? toys for men :D
wvdakker has quit [Quit: ZNC 1.8.2+deb2+b1 - https://znc.in]
<\x>
what sucks with ciaomi recently though is lack of uart access
<robimarko>
If thats the case then we are kind of done with OpenWrt on Xiaomi gear
<robimarko>
As they are being more and more annoying with locking everything down
<\x>
thats 67$ on xiaomi's page
<\x>
seems only available on mi.com
<\x>
so clearly the 360T7 is priced so much better
<\x>
both has the same number of ports anyway, but hay atleast you get a black solder mask on xiaomi's board haha
wvdakker has joined #openwrt-devel
<mrkiko>
robimarko: going via attempts, I obtained a arch/arm64/boot/Image: Linux kernel ARM64 boot executable Image, little-endian, 4K pages ; I used a .config I saved from make kernel_menuconfig, applied only some patches to mvneta
<robimarko>
mrkiko: to sum it up, best would be to spin-up a Buildroot or OpenWrt with external Linux-next
<robimarko>
And make/test the DTS for it
<robimarko>
Also, make sure to run DT schema checks on the DTS
<mrkiko>
robimarko: I am using linux.git based kernel right now; how do you run schema checks?
<robimarko>
mrkiko: Stable or RC?
<robimarko>
You want to be targeting linux-next
<mrkiko>
robimarko: for the DTS, I compared those in upstream and they are very close to openwrt atm, and so I simply copied the dts from there with my partition changes applied
<robimarko>
For schemas, just add dtbs_check as the make target
<mrkiko>
robimarko: thanks! linux.git HEAD, so RC
<robimarko>
But you have to have dtschema and linter installed
<mrkiko>
robimarko: ok
<robimarko>
Oh, and yeah, you are gonna need to document that board in bindings
<\x>
block tron posted something 7981 some days ago
<\x>
on my network here i gave up with enabling any types of ft so I only offer 2, 6, 8. no 11r enabled at all. I do have 11k I push neighbors atleast to them
<mrkiko>
robimarko: 37xx
<\x>
try disabling 11r then have wpa3/2 mixed with 11w optional stintel and see if that ipad connects properly
<\x>
theres a bit of an issue with 11w and 11r since some clients will try to connect as ac-4 yet hostapd expected them to connect with ac-6 when the client flagged for the capability
<stintel>
sigh, do I really need to bring up another AP to figure out how crapple wants me to configure my AP
<stintel>
fuck sake
<\x>
tbh just disable any 11r hehe
<\x>
if you want a controller, dawn does work without 11r enabled, itll still ask clients to move via 11v
<\x>
but ofcourse clients will still decide so its a bit of a meme for me, I just let them do what they do haha, I do push neighbors to help them at least
<\x>
this will offer 2,6,8 (wpa2/psk, wpa2/sha256, wpa3/sae)
<mrkiko>
robimarko: but for my evice, what would I need to document exactly? I am trying to search for examples, but I am not using any new binding as far as I can tell
<stintel>
\x: I don't want to disable 11r
<\x>
stintel: then thats an issue
<\x>
since wifi alliance and hostapd doesnt match on what they want to do
<stintel>
\x: apple doesn't care about wifi alliance anyway
<\x>
wifi alliance is like 2, 6, 4, 8, 9
<stintel>
overcharge people for their crap, yet refuse to do any wfa certification
<\x>
hostapd is like 2, 4, 6, 8, 9
<\x>
see the mismatch on 4 and 6?
<\x>
thats why most manufacturers just ship with disabled 11r nowadays
<\x>
tbh wifi alliance is in the wrong here, wpa2/ft-psk isnt more secure than wpa2/sha256
<\x>
what you can do stintel is to either remove 6 from being offered, so you disable 11w. dont worry, it only toggles off wpa2/sha256, wpa3 will still be wpa3
<stintel>
wpa3 client support is limited, it would result in most clients not having MFP
<stintel>
which is unacceptable
<stintel>
any kiddie can render your wifi network broken if you don't have MFP
<\x>
then dont offer any ft. only offer 2, 6 and 8
<\x>
you can even go more with like 6 and 8 only, just move 11w to required
<stintel>
but this is ridiculous. _all_ my devices connect to psk3-mixed+11w-optional+11r, except the piece of shit crapple
<\x>
are you even sure that thats worth it?
<\x>
very few devices do wpa3/ft-sae currently
<\x>
check the akm suite of your clients
<stintel>
yes, I'm trying to find a regex to make that hostapd_cli output readable
<\x>
tbh this 11w + 11r should have its own wiki entry
<stintel>
instructing apple users to complain to apple about the combination not working :P
<stintel>
maybe if 1000 people complain about it they might do something
<stintel>
I have 30+ wireless clients, the only one that's bitching is the ipad
<stintel>
well not entirely, 1st/2nd gen oneplus is also not happy, but these devices are ancient and insecure anyway, they're just testing devices
<\x>
most androids and apple follow that wifi alliance shizzle
<\x>
i had a small rant on #openwrt about it back then
<\x>
tbh the goal is to only run 9|8|6, but im sure most of us has wpa2 without 11w clients hehe, those needs 4|2
snh has joined #openwrt-devel
<\x>
dont offer 6 or dont offer 4. you choose.
<\x>
¯\_(ツ)_/¯
<mrkiko>
robimarko: ok, so I understand I should play with quilt :D
<mrkiko>
robimarko: so I would try to set up the openwrt buildroot to use the external kernel and point it to the linux-next git tree with my changes, then use the /prepare recipe?
floof58 has quit [Remote host closed the connection]
floof58 has joined #openwrt-devel
<f00b4r0>
stintel: catching up I was away; 11r works perfectly well with idevices, provided you use over the air, not over the DS
<f00b4r0>
dunno about wpa3 which I don't use and don't know what MFP is. I don't use 11w either
<stintel>
MFP == 11w
<f00b4r0>
ah ok
<stintel>
if you don't use it, any kiddie can deauth your clients with a single aireplay-ng command
<f00b4r0>
that's cool
<stintel>
in an endless loop, rendering your network unusable
<f00b4r0>
any such kiddie would have to be within firing range from my house anyway, so not exactly an issue ;)
<olmari>
Suprisingly many things have been b0rked emprically with me either with 11w, or WPA3 or whatever stoopid combo... and never seems to fail same way
<f00b4r0>
I also don't enable 11w back "home" (for a little while longer) in Paris and never had issues either
<mrkiko>
robimarko: so in the end I can't leverage the external git thing, so I miss how toconnect my buildroot to linux-next in a sense...
<stintel>
and if you want to use e.g. MBO with WPA >=2, MFP is required (can be optional, but not off)
<mrkiko>
robimarko: I understand I probably should do something similar to what is done when porting to a new kernel, but looking at git history isn't probably enough to find out. I can probably do it with a 6.x stable kernel, maybe leveraging a pending PR... but I'm little bit lost
<f00b4r0>
i typically have a single band up at most APs, and when I have more than one, it turns out the idevices are pretty good at figuring themselves where to be
<f00b4r0>
so all is well in my fruit-based world :D
* f00b4r0
imagines stintel picturing him as a wireless caveman :-D
<stintel>
😂
<mrkiko>
In my case (apartment) this can be an issue probably. but I am under the impression that securing wi-fi is kinda lost cause right now. Even assuming all the software running in the CPU is perfect (no bugs), I can't say for sure a person out there won't be able to send a packet that simply crashes the firmware of the wi-fi card and/or takes control of the device somehow, rendering my network unusable
<robimarko>
I was away, so I am just now reading through
<mrkiko>
robimarko: yeah, sure...
<f00b4r0>
meanwhile i succeeded in cross-grading a raspbian buster rpi to debian bullseye, I should probably document that somewhere
<mrkiko>
robimarko: don't worry, sorry for me disturbing ... I am willing to learn and understand this is hard-core stuff, I just miss where to start, even tough I am reading about quilt and I know it will be my friend
<robimarko>
mrkiko: A3720 should work rather well without forward porting anything
<Borromini>
mrkiko: whatare you porting?
<robimarko>
Most of the patches are backports
<robimarko>
Honestly, the way I upstream stuff is with a buildroot using initramfs images and linux-next kernels
<mrkiko>
Borromini: trying to submit gl-mv1000 support upstream
<robimarko>
That is the most flexible way
<mrkiko>
robimarko: infact i was trying to achieve that, because I know that even if it will be overkill this time, it may well be useful for me in future
<mrkiko>
robimarko: even tough I have no serial access to the device
<robimarko>
Ugh, its not ideal without UART
<mrkiko>
robimarko: I bitterly know :D
<robimarko>
Since we are discussing here, I wont reply to the email
<robimarko>
Is it that hard to get UART on the device?
<mrkiko>
robimarko: I was thinking I could point buildroot to my external linux-next tree via the related config option, but when I try to make package/kernel/linux/prepare, it tries to use the testing kernel 5.15.79 regardless
<mrkiko>
robimarko: if you're blind and nobody around the help soldering, yes :D
<robimarko>
Sorry, I keep forgetting that you are blind
<robimarko>
The issue is that if it doesnt just boot its impossible to know where is it stuck
<mrkiko>
robimarko: no problem... I could get the UART on the device but should probably send it away and wait a good amount of time
<mrkiko>
robimarko: sure. Infact I am trying to work "methodically" to get to the point where at least I have a more precise idea of the situation