<SwedeMike>
interesting that they didn't make it support 2.5GbE all the way, considering the specs it kind of screams for 2.5GbE support
<neggles>
suspect they were just planning to use the same PHY chip on U6-LR/U6-Pro/U6-Mesh initially
<neggles>
then the MTK-based U6-Pro/U6-Mesh got canned because, well, they sucked, and didn't have 2.5G either
<neggles>
AIUI they're being reworked with QCA SoCs
<SwedeMike>
I never did understand the differentiation between U6-LR, U6-Pro etc, it seems very confusing market segmentation
<SwedeMike>
and they're selling 2.5GbE switches but barely have any 2.5GbE client devices at all
<neggles>
U6-LR and U6-Lite's dirty little secret is the same as all the other MT7622 WiFi 6 devices' - no 802.11ax on 2.4Ghz
<SwedeMike>
I remember the U6-LR early access hw being different frmo the U6-LR actual shipping hw. Friend of mine bought the EA and it had a different radio, from what I remember
<neggles>
yeah, that was a more minor redesign
<SwedeMike>
does the MT7622 even have forwarding performance to do 2.5GbE even if it has such a phy?
<neggles>
yeah, dual-core A53, it'd be fine
<neggles>
U6-Pro was just U6-LR with 2x2 802.11ax on 2.4GHz, iirc
Andi_ has quit [Ping timeout: 480 seconds]
Andi_ has joined #openwrt-devel
<SwedeMike>
ok, I only have had experience with MT7621, and without accelerator it was anemic forwarding speed
<neggles>
they are worlds apart. mt7621 is an old, slow, dual-core quad-thread 800MHz MIPS
<SwedeMike>
ah
<neggles>
would be nice if mediatek hadn't made the model numbers so close.
<neggles>
from what i've heard, the respun U6-Pro will be a quad-core IPQsomethingoranother with 4x4:4 802.11ax on both bands and 2.5G - but take that with a *huge* grain of salt
<stintel>
even if it doesn't do full 2.5, I'm sure it can do more than 1GbE
<neggles>
throughtput-wise, yeah
robimarko has joined #openwrt-devel
<SwedeMike>
I'm looking at the MT7622A block diagram and it does seem to be quite powerful indeed
<robimarko>
stintel: Whats the Aquantia PHY model inside of the LR?
<SwedeMike>
nice assortment of ports
<neggles>
just a damn shame it's only 802.11n 2.4GHz
<stintel>
neggles: aye
<neggles>
SwedeMike: AQR112
<neggles>
er, robimarko:, rather
owrt-2102-builds has quit [Quit: buildmaster reconfigured: bot disconnecting]
owrt-2102-builds has joined #openwrt-devel
owrt-1907-builds has quit [Quit: buildmaster reconfigured: bot disconnecting]
owrt-1907-builds has joined #openwrt-devel
owrt-snap-builds has quit [Quit: buildmaster reconfigured: bot disconnecting]
owrt-snap-builds has joined #openwrt-devel
<stintel>
lol, I have some shortcuts for some sites, bugs.gentoo.org = bgo, that works funny with openwrt things ... bugs.openwrt.org = boo, forum = foo, git = goo :D
<robimarko>
neggles: Does it show up on the MDIO bus?
<neggles>
not a clue, I only know which PHY it is from the DTS in the tree :P
<stintel>
the first problem is that the mtk eth driver doesn't do c45 mdio
<neggles>
yeah i've been to the repo quiite a few times and misread it every time... woops
<Tapper>
nbd latest master. SNAPSHOT r18061
<robimarko>
stintel: Do you have any info on the U6-Pro? Cause its cheaper then the LR
<nbd>
Tapper: what platform are you using?
Borromini has joined #openwrt-devel
<stintel>
robimarko: I do not
<robimarko>
I looked at the FCC pictures and they are terrible
<robimarko>
There is 2 and from like a meter with terrible resolution
<robimarko>
I see that more and more vendors are doing it that way now so you cant know whats inside
<stintel>
robimarko: as I mentioned earlier, only got the U6-LR to do 2.5GbE testing
<robimarko>
Thats what I am also interested, whether it has the AQR PHY as well
Borromini has quit []
<userx>
Hello, how can I install package (add files) to staging_dir/target-<arch>_musl/usr? I need it for build dependencies since I see that this path is used in PYTHONPATH variable.
<userx>
In particular I'm interested in staging_dir/target-x86_64_musl/usr/lib/python3.9/site-packages. For example, I took python-twisted. I see this package installed to staging_dir/target-x86_64_musl/root-x86/usr/lib/python3.9/site-packages
* stintel
prints a label for the U6LR TTL pinout
<stintel>
gonna be less annoying next time :)
<karlp>
yah, I'm putting a label printer on the christmas wishing list I think :)
<stintel>
I already label all my power bricks and even most cables that come with devices
<karlp>
little concerned about thermal printer durability though,
<neggles>
robimarko: the U6-Pro that has been in the EA store / thru early FCC approval is not what will actually make it out to retail.
<robimarko>
So they are playing that game again
<neggles>
not playing any games really, they just realized that another MT7622-based device wasn't going to cut it
<neggles>
qualcomm stuff inbound instead
<neggles>
it got some *really bad* feedback from testers, so it has been yeeted out the window
<stintel>
booo qca
<stintel>
not buying anything qca based anymore
<neggles>
you're not wrong, but mediatek don't have a quad-core with dual 4x4 AX radios (let alone an 8x8 radio and 4x4 radio), soooo...
<Tapper>
nbd ipq806x
<neggles>
you can expect the U6-Pro to be essentially the same hardware as a TP-Link EAP660 HD
<stintel>
8x8 was coming I believe
<robimarko>
So its gonna be using IPQ807x then
<neggles>
most likely, yes
<robimarko>
Well, its the only SoC that has dual band AX
<neggles>
unless you use PCIe radios
<neggles>
but, $$$
<stintel>
I have this amazing ipq807x based device with 10GbE 802.3bt uplink
<robimarko>
Yeah, that costs way more
<stintel>
but try getting sources from Huawei
<robimarko>
Yeah, I just ignore anything made by Huawei
<stintel>
it has 8x8 5G and 4x4 2.4G
<neggles>
that'll be the IPQ8074
<robimarko>
So it uses the most beefy radios
<neggles>
which will probably show up in the U6-HD
<neggles>
assuming that happens
<stintel>
know of any other device with those specs? 10GbE 802.3.bt 8x8/4x4 rj45 console port?
<robimarko>
Not with BT for sure
<stintel>
802.3at is ok too
<neggles>
i mean you only need 802.3bt if you're pulling north of 25W, but
<robimarko>
QCA made the IPQ807x series really modular
<robimarko>
You just select the actual radios per the specs you want
<stintel>
but 8x8 + 4x4 only makes sense with 10GbE uplink imo
<robimarko>
Whats the Huawei model?
<neggles>
cisco C9130AXI-Z has all that and 5GbE
<stintel>
AP7060DN
<neggles>
I don't think the move from 5GbE to 10GbE serves any real purpose for any wifi AP
<robimarko>
Cause, I am messing with Xiaomi which shares the GPL in the same fashion as Huawei
<stintel>
are there u-boot sources for ipq807x ?
<robimarko>
But its dirt cheap, hence why I tolerate it
<robimarko>
Yeah
<stintel>
in that case I can probably just build u-boot and flash it via clamp on the SPI NOR
<neggles>
nah, there's a *shitload* of hardware in those ciscos
<Tapper>
nbd OK
<neggles>
it's a lot more than just an AP
<neggles>
but yes, no *person* should buy cisco/aruba wifi APs brand new, you buy them in 3 years once businesses start refreshing them for a tax break
<neggles>
with the notable exception of the aruba instant-ON stuff, which is actually reasonably priced and quite good.
<neggles>
no licenses either
<robimarko>
stintel: Unfortunatelly not as unlike every other vendor, they are not just putting the reference desing in the FIT config
<robimarko>
And since they disable kernel bootlog that is out of the question as well
<stintel>
got the ap7060dn for EUR540 new 1.5y ago
<stintel>
I'd say that beats the Cisco price :)
<neggles>
well yeah, but, huawei
<robimarko>
But if you can pull the DTB it shouldnt be hard
<neggles>
spi_flash_programmer_clip.jpg
<robimarko>
neggles: Its not like Cisco is nicer
<robimarko>
Just look at the MR33 trap
<Namidairo>
the blowing away of efuses
<Namidairo>
lol
<robimarko>
They will actually trip the secure boot fuse in the newer bootloader
<neggles>
hey don't think i'm defending cisco here
<robimarko>
If you try to load an unsigned FW
<neggles>
it forcibly turns on secure boot *after* you try to load an unsigned firmware?
<neggles>
WHY?!
<Namidairo>
spite
<neggles>
just turn it on from the get-go you assholes
<robimarko>
Because "security"
<robimarko>
And because Cisco
<robimarko>
So, its a d*ck move
<neggles>
why would they not just... turn on secure boot?
<neggles>
that's just mean
<robimarko>
Because its not worth it
<robimarko>
Just complicates the production
<robimarko>
And they sign their FW anyway
<neggles>
if they sign their FW anyway it doesn't complicate production at all
<Namidairo>
i wonder what key is loaded in there
<neggles>
i am 100% sure cisco could be just using secure boot by default
<Namidairo>
or rather, if they are even doing it properly
<neggles>
heh
<neggles>
maybe it's all zeroes
<Namidairo>
^
<neggles>
probably just stuffing it with random bits if they're solely using it as a f&@$-you
<nbd>
and change the bpf toolchain option to "Use prebuilt LLVM toolchain"
<Tapper>
nbd let me see if i can get a newer build of it in my ubuntu
<nbd>
no need to update llvm on your system then
<Tapper>
nbd I will get back to you.
<Tapper>
nbd thanks
swalker has quit [Remote host closed the connection]
swalker has joined #openwrt-devel
minimal has joined #openwrt-devel
<stintel>
oooh
<stintel>
> I will be happy to provide you with the GPL files. We can provide you with the software version for which the appliances last held a valid License, if you can provide me with a company name or SN of the appliances I can check what that version was.
aleksander has joined #openwrt-devel
nitroshift has quit [Quit: Gone that way --->]
<Habbie>
stintel, remember two weeks ago when i asked you about saving license keys :D
<nbd>
i will add a minimum llvm version check soon
<nbd>
so that it bails out at build time
<stintel>
Habbie: :P
<nbd>
instead of producing weird runtime errors
pmelange has joined #openwrt-devel
<stintel>
robimarko: I'm lost :P
<stintel>
ahh
<stintel>
mdio mdio-bus mmd ... seems to return the same stuff as before
<stintel>
when not using mmd
<ldir>
nbd: Can you tell me what llvm toolchain you're using on macos and where you installed it from please?
castiel652 has quit [Quit: Page closed]
Tapper has quit [Ping timeout: 480 seconds]
<nbd>
ldir: i think i installed llvm 13 using brew
<nbd>
but the apple toolchain might work as well, not sure
<ldir>
ok - it's just that apple's clang can produce ebpf... I did it once... can't remember how?
Tapper has joined #openwrt-devel
<nbd>
ldir: took me a while to figure out how to compile kernel ebpf properly. the result is in include/bpf.mk
<ldir>
nbd: I'm sure it did! :-) I'll just brew install llvm
<nbd>
i think the apple toolchain should behave in the same way as upstream llvm
<nbd>
if it supports ebpf properly
<nbd>
so you probably don't have to brew install llvm
Tapper has quit [Read error: Connection reset by peer]
Tapper has joined #openwrt-devel
<Tapper>
nbd that's working now.
<nbd>
great
<Tapper>
nbd I think you should put this in the wiki. I would do it, but my spelling and gramer would not be a good look for openwrt.
<Tapper>
lol
<robimarko>
stintel: Well I am out of ideas
<stintel>
same :)
goliath has quit [Quit: SIGSEGV]
<ldir>
PaulFertser: not looked at dnsmasq stuff for a while, however reading the dnsmasq man page (owww my head hurtz) and looking at your change, all looks good to me.
<ldir>
if I can remember how to a apply a patch from patchwork I'll apply it :-)
<PaulFertser>
ldir: thank you! I'll add the documentation for the option to the wiki.
mangix has quit [Remote host closed the connection]
mangix has joined #openwrt-devel
stintel has quit [Ping timeout: 480 seconds]
stintel has joined #openwrt-devel
jlsalvador has quit [Remote host closed the connection]
jlsalvador has joined #openwrt-devel
minimal has quit []
Borromini has joined #openwrt-devel
jlsalvador has quit [Quit: jlsalvador]
jlsalvador has joined #openwrt-devel
<stintel>
can'
<stintel>
can't repro the issue with Pi Zero 2 on Raspberry Pi OS :(
<stintel>
checksums for the 3 firmware files are the same
niyawe has quit [Quit: niyawe]
<dwfreed>
stintel: what were you trying to repro? I have a 4 8 GB sitting next to me
<dwfreed>
It is still on Pi OS buster, but I could replace it with bullseye easily, as I haven't done anything with it yet
KGB-0 has quit []
KGB-0 has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
dedeckeh has quit [Remote host closed the connection]
arifre has quit [Remote host closed the connection]
userx has quit [Remote host closed the connection]
lmore377 has quit [Quit: No Ping reply in 180 seconds.]
<stintel>
dwfreed: Zero 2 locks up on OpenWrt
lmore377 has joined #openwrt-devel
<stintel>
it's a new device, not supported yet
<dwfreed>
ah, bummer
<dwfreed>
I hate how much of raspberry pi is still closed source
<dwfreed>
it's better in bullseye, but still not great
goliath has joined #openwrt-devel
<stintel>
bummer indeed
<stintel>
I'd hoped to replace the zero w that I use as security cam with the zero 2
<stintel>
the zero w is at 100% CPU all the time (motion)
<Habbie>
the zero 2 really hits a lot of sweet spots
<rsalvaterra>
dwfreed: How much of the RPi is actually closed, these days? I thought the last remaining pieces were the GPU and hardware video decode drivers…?
<rsalvaterra>
Oh, I remember, that horrible VideoCore blob…
<Slimey>
heh
<rsalvaterra>
… which basically boots the whole system and kicks the "main" CPU…
<Habbie>
is that really different from the closed source BIOS on the average PC or server?
<Habbie>
other than it having a separate CPU
<rsalvaterra>
Habbie: Yes, completely different.
<rsalvaterra>
The BIOS run on the CPU.
<PaulFertser>
Current computers no longer have BIOS.
<PaulFertser>
Unfortunately, most UEFI implementations are just as shitty.
<PaulFertser>
If we consider current Intel CPUs then it's even worse than any videocore code because you can run unsigned code on VC while you can't run unsigned code on the ME (a dedicated CPU in PCH).
<rsalvaterra>
s/BIOS/UEFI. Only the smell differs. :P
<PaulFertser>
rsalvaterra: not really, BIOS was supposed to be running all the time (DOS syscalling it to do many jobs). While UEFI itself is no longer active once the kernel boots.
<PaulFertser>
And Linux was calling into BIOS too (remember APM for power management?) but not often, agreed.
<rsalvaterra>
PaulFertser: Uh… what about UEFI Runtime Services…?
<PaulFertser>
rsalvaterra: they're no longer active after GRUB transfers control to Linux. IIRC it's called "shutdown services" even.
<PaulFertser>
rsalvaterra: I think the only part that gets used is to read/write "nvram variables", probably via SMM handlers.
<rsalvaterra>
Hm… good to know. But there's still a bigger elephant in the room in x86…
<rsalvaterra>
… SMM.
<PaulFertser>
rsalvaterra: I think ME is the biggest.
<rsalvaterra>
Hard to choose… ME or SMM…? :)
<PaulFertser>
rsalvaterra: you control all of the SMM code and can change it.
<rsalvaterra>
Theoretically…
<rsalvaterra>
… becuse in practice, that code is part of the firmware.
<rsalvaterra>
So, unless you're engineering a computer from scratch…
<PaulFertser>
rsalvaterra: I have to admit, I had ExitBootServices() in mind. Runtime services is different thing and it's indeed used for nvram variables modification. But I do not think they normally run in the background.
<PaulFertser>
rsalvaterra: coreboot was ported to many COTS devices
<rsalvaterra>
Sure, but not yet any of the devices I own, unfortunately. :)
<PaulFertser>
And if the board is based on reference design it's probably supported by TianoCore edkII directly. E.g. our Tioga Pass server seems to be booting vanilla "Purley" edkII, and probably not much tweaking is needed to make everything work fully.
<PaulFertser>
I understand this all is unhelpful in most cases. Fuck Intel and the others. Only Pine seems to be really caring about the ordinary end users.
<rsalvaterra>
It's the reason I drool over the Talos POWER9 machines…
<rsalvaterra>
… but they're way out of my budget.
<stintel>
if only someone would build them in EU
<rsalvaterra>
stintel: Preach it, brother! :)
<stintel>
the shipping costs are what is killing it for me
<stintel>
and customs fee, import tax
<rsalvaterra>
stintel: Lucky you, I get thoroughly trounced by the price of the hardware already. :P
<stintel>
well becoming an economic refugee helped my finances :P
<rsalvaterra>
When you can get a dual socket EPYC for the same price, you start thinking twice about openness vs performance…
* stintel
reverts mac80211 to 5.10 and tries that on the Zero 2
<stintel>
I want that thing to work so I can replace the Zero security cam and have an excuse to order another Zero 2 😂
<rsalvaterra>
stintel: What's the Wi-Fi hardware on the thingy?
danitool has joined #openwrt-devel
<stintel>
another variation of bcm43430
Borromini has quit [Quit: leaving]
<stintel>
ooh
<stintel>
it's not hanging now
<stintel>
not working either, but not hanging is an improvement
<nbd>
there's really no need to install a package onto the system. you can simply download the tarball from openwrt.org and unpack it into your build directory
<nbd>
that way you get the same version that the snapshot builds use
<Tapper>
cant OpenWrt auto do that with out checking for a llvm on the host?
<nbd>
well, it only works on linux/x86_64 hosts
<Grommish>
stintel: Not ready to tackle this memleak yet? :D
<nbd>
but i guess at some point i'll add a script to scripts/
<stintel>
Grommish: that's over my head
<Grommish>
stintel: *sigh* Mine too, which is why i was pinning my hopes on you :D
<stintel>
Grommish: but you should probably have added all maintainers returned by the get_maintainers.pl script and also the ML
<Grommish>
stintel: Also beyond me.. I don't expect to hear back from the single, and I'm not on the mailing list.. Meh
Tusker has joined #openwrt-devel
<Grommish>
Well, does that mean Octeon targets are going away?
<Grommish>
Since none of them can seem to handle 5.10
<stintel>
it's just a bug that needs to be fixed ..
<Grommish>
Well, yeah :) But aren't they all? You know how slim the dev pool is for the target, and I have nothing practical to offer except obstinance
<Grommish>
Anyway, if you think there is anything I can help with, hit me up. I don't mind putting in an OWRT bug report (not that it's needed) but upstream I wouldn't have any idea how, or the knowledge to answer them if they answered
<stintel>
try the ML returned by the get_maintainer script
<Grommish>
Ok.. CC you again?
<slh>
Grommish: Ansuel has been working on kernel v5.15 for ipq806x (WIP PR on github), might be worth looking at for octeon, maybe it's fixed there already
<stintel>
Grommish: yes, I'd like to get a response if anyone sends one
<Grommish>
slh: Ok.. I'll look, thanks
<stintel>
I have not revived the box yet where I have my octeon cards in so can't even test anything myself
<hauke>
stintel: this diff from the kernel should apply more or less: git diff v5.12.19 v5.11.22 -- drivers/net/wireless/broadcom/brcm80211/
lmore377 has quit [Quit: No Ping reply in 180 seconds.]
lmore377 has joined #openwrt-devel
<fda->
dangole: after update avm7320 to uorrent version i noticed some "random: ubusd: uninitialized urandom read (4 bytes read)" and "urandom-seed: Seed file not found (/etc/urandom.seed)". are the messages okay and should be ignored?
dansan has quit [Quit: The C preprocessor is a pathway to many abilities some consider to be unnatural.]
dansan has joined #openwrt-devel
aiyion_ has joined #openwrt-devel
aiyion has quit [Ping timeout: 480 seconds]
<neggles>
Grommish: definitely does not have USB3 no
<stintel>
forums suck
<stintel>
impossible to find anything
<neggles>
bad search implementations suck
<neggles>
and most forum search implementations are bad
bookworm_ has quit []
bookworm has joined #openwrt-devel
<Tapper>
I go back to google and sirch for what I want with the words openwrt forums after it lol
rmilecki has quit [Ping timeout: 480 seconds]
goliath has joined #openwrt-devel
<Grommish>
neggles: I didn't think so.. Wonder if it'll have any impact on the ERL if it's enabled, other than just taking up space
<neggles>
Grommish: doubt it, there's no XHCI controller in the octeon chip they used, way way way too old
<neggles>
it predates USB3 by at least 3 years
<Grommish>
neggles: Yeah. there is a PR that wants to enable it by default.. Any good with memleaks? :D
<stintel>
that PR is for a way newer Edgerouter than the ELR ?
<stintel>
the supported octeon devices have plenty of storage, so enabling things in the kernel config rather than adding a kmod-package to default_packages is the usual way things are done
<neggles>
*distant sad SNIC10e noises*
slh has quit [Ping timeout: 480 seconds]
<neggles>
my kingdom for a 3mb (compressed, LZMA is fine even iirc) octeon kernel+initrd
<stintel>
:D
<stintel>
you live in a fantasy world :P
slh64 has quit [Ping timeout: 480 seconds]
<neggles>
I will settle for a 3mb kernel image that has *just barely enough features* to mount a ubifs volume on NAND and kexec() into it
<neggles>
which... is pretty much the cavium stage1+stage2 setup...
<stintel>
gahhh bloody brcmfmac
<neggles>
ah yeah, that PR *is* for a much newer ER, the ER-6P