<\x>
so in theory it can turn those three slots on the omnia to six slots hehe
<Mangix>
LOL why?
<Mangix>
in any case, the interface would get band limited
<Mangix>
assuming a pcie 2.0 slot, pcie 1.0 speeds per device
<Mangix>
2.0 == 2gbit/s
<Mangix>
no wait, 1.0 == 2gbps
<Mangix>
2.0 == 4gbit/s
schwicht has joined #openwrt-devel
schwicht has quit []
schwicht has joined #openwrt-devel
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nixuser has quit [Remote host closed the connection]
nixuser has joined #openwrt-devel
<rmilecki>
bluecmd[m]: thanks a lot!
<rmilecki>
bluecmd[m]: i'm not familiar with SFP or RX_LOS
<slh>
\x: wifi6 cards are getting quite hot, such an adapter would concern me in terms of heat dissipation
<slh>
at least bulky passive coolers are the norm, active cooling in quite a few cases
xavifr has joined #openwrt-devel
goliath has joined #openwrt-devel
<rmilecki>
nbd: it's great I can add another wifi-iface (and reload_config) without bringing 1st BSS down
<rmilecki>
nbd: i found another hiccup, i'm verifying now and i'll provide info shortly
<rmilecki>
(also comparing brcmfmac and mt76)
<Mangix>
i have never used brcmfmac. Am I missing out?
philipp64 has quit [Ping timeout: 480 seconds]
Misanthropos has quit [Ping timeout: 480 seconds]
philipp64 has joined #openwrt-devel
<KanjiMonster>
Mangix: going by the amount of headaches it is causing, I'd say only if you like debugging things ;)
madwoota has quit [Quit: .]
<PaulFertser>
Mangix: you kinda have to if you're using one of the popular SBCs. Or Pinebook Pro. Or Apple Macbook.
<KanjiMonster>
rmilecki: RX_LOS is a well defined pin for SFP modules to report no link, which is usually exposed as a GPIO or bit in a CPLD register
madwoota has joined #openwrt-devel
<PaulFertser>
Mangix: (re DBDC is great) but that card is only 2x2 on 5 GHz, so you'd miss an opportunity to serve many MU-MIMO clients faster as they would all be sharing 2x2 bandwidth.
Misanthropos has joined #openwrt-devel
<rmilecki>
Mangix: brcmfmac is driver for FullMAC devices, so you have limited control over them, similarly to ath10k/ath11k
<rmilecki>
Mangix: i wouldn't call Broadcom's FullMAC devices or firmware or devices terrible
<rmilecki>
it requires some quirks but nothing really outstanding
<rmilecki>
my smartphone seems to connect but immediately drops the connection
<PaulFertser>
Also, FullMAC kind of makes sense when you're tight on power budget (mobile devices).
<KanjiMonster>
Would there be interest in actual, working* MSTP in OpenWrt? (dependend on switch chip and driver support)
<nbd>
i think that would be a good idea
<KanjiMonster>
I have hacked together a proof of concept for mstpd (using the in-kernel MSTI support introduced in 5.18)
<KanjiMonster>
the mv86-something driver supports msti in its driver, and I have a local patch for b53. No idea about other switch chips if they could support it
<KanjiMonster>
the kernel interface is annoying to work with though
<KanjiMonster>
nbd: seeing that ustpd is just a fork of mstpd, it should be fairly easy to port to it, but I would need some help in adapting netifd and the ubus parts for it
<KanjiMonster>
e.g. the bridge needs mst mode explicitly enabled, and you can only do that as long there are no vlans defined on the bridge (so ideally you do that on creation)
<KanjiMonster>
you need some way to assign vids to mstis, and set a config id for mst (used by different MSTP instances to indentify that they are supposed one i-forgot-the-term-for-that)
<KanjiMonster>
since it will require a kernel >= 5.18, this will be main/openwrt-24.x feature, so hopefully enough time to implement it and fix most bugs ;D
xes has quit [Ping timeout: 480 seconds]
dangole has joined #openwrt-devel
<nbd>
KanjiMonster: i can take care of the netifd part
<nbd>
KanjiMonster: the vid->msti assignment and the config id part is configured in the bridge layer via rtnl, right?
slh64 has joined #openwrt-devel
<KanjiMonster>
nbd: yes, but you can only assign existing vlans (i.e. with at least one member) to mstis
<KanjiMonster>
if there is no member, then you'll get an error
<KanjiMonster>
and if you remove the vlan and readd it, then its msti resets to 0
dangole has quit [Ping timeout: 480 seconds]
<KanjiMonster>
msti stp states can only bet set on bridge members that have at least one vlan of that particular msti, else you get an error
<KanjiMonster>
and if you try to assign a vlan range to a msti and one of the vlans doesn't exist, then the whole assignment fails (AFAICT)
<nbd>
so i'd take care of those msti config aspects within netifd
<nbd>
and leave it to ustpd to get the configuration state from rtnl
<nbd>
only passing the stp protocol type via ubus from netifd
<KanjiMonster>
there's also the question of where the source of truth is supposed to be; the in kernel vid -> msti assignment, or the mstpd configuration
<nbd>
i think it should be the in kernel assignment
<KanjiMonster>
the vid -> msti assignment is part of the configuration used by mstp instances to identify identically configured bridges to form one ugh, lets call it domain
<KanjiMonster>
so it might make sense to have it not depend on the currently configured vlans on the bridge but be static
<nbd>
at least in openwrt, i don't want ustpd do have its own configuration
<nbd>
netifd is responsible for managing the bridge configuration
<nbd>
and ustpd should follow whatever netifd configured on the kernel side
<nbd>
unless there is a practical reason why that doesn't work well
<KanjiMonster>
you would probably still need some parts to directly configure ustpd, like (per tree) priorities or path costs; there is no place in the kernel for that
<nbd>
is this global, per-vlan or per-mst?
<KanjiMonster>
per-mst
<nbd>
ok
<nbd>
i guess that can be passed via ubus as well
<KanjiMonster>
technically MSTP also works by assigning VIDs to FIDs, and FIDs to MSTIDs, but the kernel doesn't know about FIDs and has essentially a static VID to FID mapping
<KanjiMonster>
also I'm not sure if the exchanged configuration between MSTP instances uses the FID stuff, I think it works on VID -> MSTID
dangole has joined #openwrt-devel
<KanjiMonster>
nbd: for some reason the man page doesn't have all commands, see https://pastebin.com/DEvPtM1d and the commands that take a mstid (and the setmstconfid command) what you might want to configure for mst and per msti
<KanjiMonster>
the standard says maxum of 64 trees are supported, but switch chips may support less than that (e.g. the most b53 supported switches that do support mst support 8)
<KanjiMonster>
there is no way of getting that information from the kernel though, so you just assign vlans to mstis until you get an error from the kernel ;P
schwicht has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<mrkiko>
I am grepping with: grep -rn --group-separator='~~~~~~~~~~~~~~~~~' -B 7 -A 7 printk *
<mrkiko>
I would need to show the filename onlyh once, then only line output and the code, or even not line content. Anyone has any idea or tool to suggest?
schwicht has joined #openwrt-devel
dangole_ has joined #openwrt-devel
<mrkiko>
like "file.c:line" then output lines only in the next output lines, then group separator, then file.c:line,. then other lines of output without both fiulename and line
dangole has quit [Ping timeout: 480 seconds]
<KanjiMonster>
mrkiko: grep -rl printk * | xargs -n1 sh -c 'echo "$1:" && grep -n --group-separator='~~~~~~~~~~~~~~~~~' -B 7 -A 7 printk $1' sh
<neggles>
\x: yeah you'd be insane to put two MT7915AN or MT7916DAN cards on one of those double-stacks
<neggles>
they can pull 10W each
<neggles>
its rare but asiarf are very explicit about the "make sure you can supply 3.3v 3A!"
<mrkiko>
KanjiMonster: thanks a lot!!
<Christophe[m]1>
hauke: I'm beginning to think that this might be caused by the sd card. I managed to get it booting once after running wipefs before dd on the sdcard, now it's saying "Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000005" :'-(
<PaulFertser>
SD cards can't be trusted, they fail in many different strange ways.