<neggles>
stintel: is there a particular reason you went with BE for qoriq? (does it have to match u-boot?)
<Habbie>
(what is BE?)
<neggles>
big-endian
<Habbie>
oh right
<neggles>
the chip can run in either mode
<Habbie>
got it
<Namidairo>
he did it to infuriate you, neg /s
<neggles>
lol
<neggles>
i don't mind at all
<neggles>
i don't have an M300 (yet)
<neggles>
but pretty much everyone has given up on big-endian 64-bit POWER
<neggles>
debian have a userland but that's about it afaict
<neggles>
though FWIW i can confirm that the debian userspace works great as a chroot on OpenWrt
<Habbie>
nice
<Namidairo>
i think it might be more likely to find nxp nfc hardware in the openwrt toh than that
<neggles>
the firebox M300? stintel has one
<Habbie>
neggles, is that BE userspace on LE kernel? your debian chroot?
<stintel>
neggles: I don't think M300 CPU does LE
<neggles>
BE on BEA
<Habbie>
right
<stintel>
or maybe I misremember
<neggles>
ah no it does but altivec only operates in BE
<neggles>
so you would have to commit minor crimes to use it in LE mode
<stintel>
well altivec support in musl is shite anyway
<stintel>
it expects it to be there, if it's not (m200) -> booom
<neggles>
my friend's use case involves handwritten assembly
<neggles>
he's doing microbenchmarks on the core
tidalf has quit [Remote host closed the connection]
<stintel>
I tried having that fixed in musl but got I spent almost 2 weeks trying to figure out how PPC ASM worked and I got something working but the reviews were ... Chinese to me
<stintel>
on a 1000/600 fibre connection, SQM enabled, and multiple IPsec tunnels
<stintel>
the day I find something non-x86 with dual 10GbE that is capable of routing 10GbE that doesn't cost an arm and a leg, I'll replace the M300s
<stintel>
but for now they've proven really stable and I'm rather happy with them
danitool has quit [Ping timeout: 480 seconds]
<stintel>
right now my 10GbE switch is set up in L3 mode as gateway for my trusted VLANs but that doesn't have stateful filtering and building a stateful ruleset ... not worth my time :P
<stintel>
but it's a bit pricy still for just an experiment
<stintel>
and the sn2100 is even more expensive. although if capable of sqm+nftables at 10GbE, I'd consider 2 of those as new routers ;)
<stintel>
that ipng.ch link seems to suggest it can route at 50Gbps, but no idea how much impact nftables+sqm would have on that
<Habbie>
stintel, are you coming to FOSDEM?
<stintel>
I guess I could look at a friday-monday flight
<stintel>
but I've tried to visit fosdem so many times and I always end up arriving when it's almost over :P
<stintel>
or not going at all
<Habbie>
heh
minimal has quit [Quit: Leaving]
<neggles>
stintel: you will not get useful SQM performance out of an SN2100/SN2700, anything that's going via the mgmt CPU is going to be a waste of time
<neggles>
in theory on the SN2700 if you reimplemented it as a P4 program you could do it at line rate :P
<neggles>
but hairpinning traffic via the CPU is a big ew on a switch ASIC like that
<neggles>
ah wait, spectrum-1, so no P4 but maybe ebpf offload, unsure
<neggles>
nice that the price on the gen1 spectrums is finally coming down a bit
cmonroe has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
<schmars[m]>
mac mini and a thunderbolt-to-pcie-adapter would fit in 1U 19"
<schmars[m]>
10" 1U even, hehe
<schmars[m]>
very few sfp+ ports then, tho
tidalf has quit [Remote host closed the connection]
<SpectreDev_01>
Ansuel: could you please merge mx4200, it's fully reviewed by robimarko, and device is completely stable all that's left is to merge pull request
noltari has quit [Quit: Bye ~ Happy Hacking!]
nwf has quit [Read error: Connection reset by peer]
noltari has joined #openwrt-devel
ali3nx has joined #openwrt-devel
ali3nx has quit [Quit: Leaving]
hanetzer has joined #openwrt-devel
shibboleth has joined #openwrt-devel
SpectreDev_01 has quit [Quit: Connection closed for inactivity]