<mangix>
Guest9490: I just realized. -fpic could be the problem. replacing with -fPIC probably solves it.
<mangix>
stintel: so flow offloading was never done on 6in4 tunnels?
<stintel>
mangix: firewall4 doesn't support flow offloading yet, so your question doesn't make sense?
<mangix>
oh OK
<mangix>
I ask since flow offloading is broken on kernel 5.4 with IPv6, but works with 6in4
<mangix>
if it never got enabled, that would make sense
<stintel>
in my wip branch, initial version did not work at all due to not setting devices in the flow table, which is required
<stintel>
in later iteration, it skipped interfaces with names starting with a digit
<stintel>
and eventually I fixed it with the quoted strings
<stintel>
but I don't think this is related at all with what you're talking about
<stintel>
I'm working on userspace, the problem you mention is in the kernel
<stintel>
anyway, time for bed, being discharged from the hospital tomorrow
SamantazFox has joined #openwrt-devel
SamantazFox has quit [Read error: Connection reset by peer]
SamantazFox has joined #openwrt-devel
rmilecki has quit [Ping timeout: 480 seconds]
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<mangix>
Guest9490: add HOST_CFLAGS += -DLUA_USE_LINUX $(HOST_FPIC) -std=gnu99 to lua/Makefile.
SamantazFox has quit [Ping timeout: 480 seconds]
<mangix>
or just HOST_FPIC
<mangix>
std=gnu99 is pointless since openwrt requires recent compilers
danitool has quit [Ping timeout: 480 seconds]
victhor has quit [Remote host closed the connection]
minimal has quit []
<mangix>
so ath10k is pretty bad
<stintel>
tell us something we don't know :P
<neggles>
ath10k-ct to the rescue?
<neggles>
:P
<stintel>
unfortunately at least for wave 1 that is also unreliable and will not be fixed
<stintel>
changed jobs somewhere mid this year, corp galaxy s9 caused the ath10k-ct firmware on my dap-2695 to start crapping itself once every few seconds or so, rendering the 5ghz radio completely unusable
<stintel>
as a result I now no longer have any QCA based APs in use
<stintel>
and I bought a Galaxy S21 mostly to not buy a QCA based smartphone anymore
valku has quit [Quit: valku]
<Namidairo>
did samsung build an integrated wifi chipset yet or are they still using broadcom on their exynos platforms
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
nitroshift has joined #openwrt-devel
nitroshift has quit [Remote host closed the connection]
nitroshift has joined #openwrt-devel
<\x>
hmmm, so theres that MT7922? aka AMD RZ616. and yeah I guess no AP mode.
<mangix>
neggles: ath10k-ct is unreliable too
<neggles>
yaaaaaaaaay!
<mangix>
what happened was I broke my ax router and am not stuck with ath10k garbage until the new ax one comes
<mangix>
/s/not/now
<neggles>
qualcomm are really determined to turn themselves into late-2000s Apple
<neggles>
their latest ARM windows SoC, they won't even admit how many cores it has
<mangix>
well, QCA is not the same as qualcomm
<neggles>
they're not better though :/
<mangix>
I mean can anyone really compete with M1?
<mangix>
I don't think so
<neggles>
well yes, mediatek are trying, the problem is that qualcomm haven't been trying
<neggles>
and they have an exclusivity deal on ARM windows SoCs
<neggles>
the dimensity 2000 (MT6983) looks like an absolute beast
<neggles>
and there have been rumors of a PC-focused mediatek chip with four cortex-X2 and four cortex-A710 in the latter half of 2022 - just as qualcomm's exclusivity on WoA expires
Tapper has joined #openwrt-devel
<neggles>
they're not going to catch up with M1 and future SoCs anytime soon, but they don't really need to
<neggles>
in any case, it'd just be nice if QCA would give us some open-source or even quasi-open-source drivers that didn't suck.
<mangix>
I mean it's not like they don't upstream anything
<Namidairo>
it's just that what they do is sometimes very broken
<mangix>
that's true
<neggles>
upstreaming broken crap is... arguably worse than not upstreaming anything
<neggles>
:/
castiel652 has joined #openwrt-devel
<neggles>
my wife just walked over and dropped a slightly wet 2MB SPI flash chip on my desk, apparently it came out of her hair in the shower
<mangix>
lolllll
castiel652_ has joined #openwrt-devel
castiel652 has quit [Ping timeout: 480 seconds]
<neggles>
i thought i threw this one out weeks ago
<neggles>
genuinely no clue how that happened... did i marry a cyborg? :O
<Grommish>
Magnetism due to the evil vax?
<Grommish>
(I live in a place a RN tried to tell the legislative body CoVID vax made you magnetic.. including putting spoons on herself.. live.. on TV.. it was great)
\x has quit [Ping timeout: 480 seconds]
ecloud has quit [Ping timeout: 480 seconds]
castiel652_ has quit [Remote host closed the connection]
rua has quit [Ping timeout: 480 seconds]
<yrrips>
lmao
rua has joined #openwrt-devel
\x has joined #openwrt-devel
nitroshift has quit [Quit: Gone that way --->]
ecloud has joined #openwrt-devel
hitech95 has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
onemarcfifty has joined #openwrt-devel
Borromini has joined #openwrt-devel
onemarcfifty has quit [Ping timeout: 480 seconds]
<hitech95>
To end the year I've sent my first mail to the mail list! Lets discuss some improvements!
victhor has joined #openwrt-devel
mangix has quit [Read error: Connection reset by peer]
rua has joined #openwrt-devel
mangix has joined #openwrt-devel
<Guest9490>
mangix: seems like you fixed the lua bug
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
SamantazFox has joined #openwrt-devel
<hauke>
mangix: Apple Socs are pretty big and expensive, that is one of the reasosn they are so fast
<hauke>
chips from Qualcomm or other semiconductor companies are normally smaller and cheaper
<hauke>
a semiconductor company has to look after their gross margin otherwise the stock will go down, for semiconductor companies this is something between 30% and 60%
<hauke>
Qualcomm has a gross margin of 57% to 60% this year, if you build commodities it would be less
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<hauke>
if Qualcomm would invest 10$ more into producting their high end chip they probably would have to charge 30$ more for it to not make the stock market angry
<hauke>
apple has a gross margin of only ~40%
nlowe has joined #openwrt-devel
rua has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<neggles>
it ain't small... the Graviton2 is even bigger
<neggles>
it's quite possible to make significantly larger SoCs for non-mobile platforms - Qualcomm just haven't bothered to try
nlowe has joined #openwrt-devel
dangole has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
lmore377 has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
nlowe has joined #openwrt-devel
<stintel>
not even *that* expensive
nlowe has quit [Read error: Connection reset by peer]
<stintel>
damn you neggles why do you do this :D
<neggles>
stintel: what, link the CCR2116? :P
<neggles>
I have two on preorder with our local distributor already...
<stintel>
:D
<neggles>
s'even got an M.2 slot
<neggles>
and most of the annapurna alpine stuff has been mainlined or is on its way in, they're just Cortex-A76-i-think cores
rua has joined #openwrt-devel
<neggles>
CCR2116 = AL73400, CCR2004 = AL32400, which is essentially a quad-core variant of the same thing
goliath has joined #openwrt-devel
srslypascal is now known as Guest9734
srslypascal has joined #openwrt-devel
<hitech95>
neggles: my issue with all those soc companies are the cluttered BPS, no manline drivers etc. There are a tons of IPs that are still lacking proper support. And each year new Socs are out.
<hitech95>
none of such company really invest on porting the drivers into mainline. Especially the hw offloading part of them. Chips get old and documentation of 20 year old socs is still under NDA.
<neggles>
mediatek do better than most, and annapurna have a vested interest in mainlining things because they're owned by (ugh) Amazon, who would like you to be able to run whatever you want on the AWS graviton instances
<neggles>
but yes. Qualcomm/QCA/Broadcom/Marvell-sort-of/etc are Not Great at mainlining things an it's a problem.
Tapper has joined #openwrt-devel
Guest9734 has quit [Ping timeout: 480 seconds]
ecloud has quit [Ping timeout: 480 seconds]
<Namidairo>
did they even have any annapurna hardware in their eero stuff yet
ecloud has joined #openwrt-devel
<Namidairo>
looks like their ax gear is qca
<hitech95>
neggles: mediatek is doing good but their offload IP is old. It was designed by ralink and partially updated. No consumer Router are on Amazon Socs. so no support there. BTW I'm also a special case whit ipv4aas. but I think that according to this: https://www.ripe.net/manage-ips-and-asns/ipv4/ipv4-waiting-list this will be common in the near future...
<hitech95>
I'm waiting for the rb5009ug to support openwrt to consider some proprietary HW, for now I'll stick with x86.
<hauke>
neggles: Qualcomm currently does not have any real competitor in the high end mobile SoC market, you can not buy a chip from apple just the full phone. They do not have to make they SoC faster, they can just raise prices
<hauke>
neggles: the sillicon vendors are doing what the big customers want and the big customers do not care about mainline support
<hauke>
they get support for the sillicon vendor
<hauke>
mainline support takes time and then you could miss your market window
<hitech95>
hauke: I got that, but I'm angry about old stuff
<hitech95>
hauke: I got that, but I'm angry about old stuff
<hitech95>
*sorry double post :(
<hauke>
nobody cares about old stuff
<hauke>
*no vendor cares
<hitech95>
like first gen ADSL modem on broadcom
<hitech95>
just release the sources as is.
<hitech95>
then let the community develop drivers by its time
onemarcfifty has joined #openwrt-devel
<hauke>
they would not make any additional revenu by releasing the code and would loose some IP
<hitech95>
anyway I'll hope that the soc in rb5009ug will get some sort of HW acceleration in future. (No idea if now it is present). Or at least ipv6 implementation in the mtk ppe.
<hitech95>
hauke: then landfills will be always filled with junk. I just avoid such companies. Lantiqu before intel was doing "goodish".
<hauke>
it was not Intel that changed the attitude
<hitech95>
hauke: no idea on what happende GRX300 was released by them and it was the first not supported generation in a long time. now it is MaxLinear that it might be even worse.
<hauke>
GRX300 and GRX350/GRX550 were announced before the Intel stuff was announced
<neggles>
the SoC in the 5009UG is a recent marvell armada, it already has significant amounts of hw accel support
<hitech95>
Well I have to go, I wish you guys a happy new year! Hope next year I'll have lot of new patches about MAP-T. Some have to be discussed in the mailing list. Like implementing an ubus pathc for miniupnp :)
<hitech95>
neggles: yea but with nat46 is a bit hard to implement shuch flow. I think it will require some magic logic to properly route the traffic with DSA drivers.
<neggles>
FWIW, it also has more than enough CPU power that hw accel might not actually be necessary for most use cases ;)
<hitech95>
neggles: IDK, with 4xa72 we cant saturete the 1gbps uplink in ipv4, in ipv6 all is fine so something in SNAT or the translation is slowing down the connection. With a MT7622 we get 400mbps in ipv4 vs full 1Gbps with oem patches on broadcom hw (ISP router)
SamantazFox has quit [Read error: No route to host]