nitroshift has quit [Remote host closed the connection]
nitroshift has joined #openwrt-devel
<Grommish>
neggles: ping
<neggles>
Grommish: pong
<Grommish>
neggles: You're an industrious scamp.. I need to find some rust-lang code I can make a package out of for testing.. Any suggestions?
<Grommish>
Ideally something that tests float at some point
<Grommish>
I think I finally settled in on rust working and it being an suricata issue, so I need something else to test with :/
<neggles>
hmmm
<neggles>
Grommish: are we using rust stable or unstable?
<Grommish>
suricata decides it wants to use /usr/bin/ld no matter what else it finds in the setup (like the mips64-openwrt-linux-musl-ld.gold)
<Grommish>
neggles: Nightly
<neggles>
build `std` ?
<Grommish>
Yes
<Grommish>
I build all the tools
<Grommish>
even the broken ones like rls and miri
<neggles>
i mean if you can build the std crate I'd imagine most things ought to work, but, there was something I came across recently... hmmm
<Grommish>
I'm cross-compiling at the moment, because most of the archs don't have native support yet
<Grommish>
as in no rustc/cargo on device yet for native building
<neggles>
that's fair
<Grommish>
but I can generate a host version easy enough
<neggles>
are you looking for anything in particular? just something reasonably complex?
<Grommish>
neggles: The biggest issues are -msoft-float considerations, so *shrug* I'd rather have a rust Hello, World that does a float so it's easy to incorporate hehe but I can't be picky
<Grommish>
For simplicity sake, anything that has a standardized configure/make/make install would be great, but again, can't be picky
<Grommish>
I was looking at libpnet but it's x86_64 only
<Grommish>
Eh, don't think too hard on it.. I can just wait for suricata devs to answer :)
<neggles>
I don't use that much in the way of rust code heh and most of the things i'm diggin up don't use float
<neggles>
do I need to re-test my PR from september adding an ath79 device given that it's kernel 5.10 now and was 5.4 then?
rmilecki has joined #openwrt-devel
<slh>
always a good idea (especially if there are the DTS or patches that might differ between the different kernel versions), if the old PR doesn't apply anymore, I doubt anyone (who doesn't own the device) will sit down and rebase it
dedeckeh has joined #openwrt-devel
nitroshift has quit [Remote host closed the connection]
dedeckeh has quit [Remote host closed the connection]
<neggles>
slh: it's near-identical to 3 other already supported drivers
<neggles>
er, devices
<neggles>
i doubt i'll need to change much of anything, but, well, i've got eleven of the things so easy enough to test :P
<neggles>
have been meaning to try hacking up the ath eeprom caldata to enable the third radio chain
nitroshift has joined #openwrt-devel
<neggles>
stintel: nooooooo
<neggles>
surely not
<stintel>
?
<neggles>
oh, this is in bridge, my bad - thought it may be octeon-related :P
<neggles>
though, uh, I am getting a memory leak on this build
<stintel>
well, I currently have 2 SNIC10e running OpenWrt, identical config save for their IP address ... not seeing the leak
snh_ has joined #openwrt-devel
<neggles>
restarting dnsmasq now eats 2MiB(!) of RAM
<stintel>
but kernel got bumped from 5.10.92 to 5.10.96 ... so I was reading changelogs
<stintel>
ok, I guess we're going to have to add printk's in the fpa alloc functions
<stintel>
and check if restarting dnsmasq triggers that ..
srslypascal is now known as Guest1617
srslypascal has joined #openwrt-devel
<neggles>
it seems to have something to do with listening sockets; if i tell it to explicitly bind to <NIC IP> and loopback, instead of both v4 and v6 "all addresses", the leak gets smaller
<neggles>
actually
<neggles>
what if i tell it to just bind to loopback
<Grommish>
Meh.. They are calling for 25mm of ice buildup over the next 36 hours or so.. This is going to suuuuuck..
<stintel>
just a regular winter Thursday :P
<Grommish>
I didn't see a leak, aside from the dnsmasq bump on restart, after I had turned off Ipv6, but it could have been coincidence with the last 5.10 bump
<Grommish>
stintel: All the utilities here are above ground :D
snh has quit [Ping timeout: 480 seconds]
Guest1617 has quit [Ping timeout: 480 seconds]
<neggles>
so if dnsmasq still leaks 1MB of memory even though it's set to `listen-address=127.0.0.1`
<neggles>
that implies it's a userspace bug right
<neggles>
or at least, not octeon-ethernet-specific (though maybe mips/octeon-specific)
srslypascal has quit [Ping timeout: 480 seconds]
srslypascal has joined #openwrt-devel
goliath has joined #openwrt-devel
Grommish_ has joined #openwrt-devel
t4h4[m] has joined #openwrt-devel
Grommish has quit [Ping timeout: 480 seconds]
Tapper has quit [Ping timeout: 480 seconds]
cmonroe has quit [Ping timeout: 480 seconds]
Tapper has joined #openwrt-devel
Grommish has joined #openwrt-devel
mrkiko has joined #openwrt-devel
<mrkiko>
Good morning!!
Grommish_ has quit [Ping timeout: 480 seconds]
danitool has joined #openwrt-devel
GNUmoon2 has quit []
GNUmoon2 has joined #openwrt-devel
GNUmoon2 has quit [Remote host closed the connection]
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
robimarko has joined #openwrt-devel
<robimarko>
f00b4r0: I finally replied, I dont know what else to do
<robimarko>
I just dont understand whats the issue
<f00b4r0>
robimarko: ack. Let's hope Kalle will reply. I don't understand either what the issue is. Maybe hauke has an opinion?
Tapper has quit [Ping timeout: 480 seconds]
<robimarko>
f00b4r0: Yeah, hopefully Kalle finally replies
<robimarko>
Last time it was like, can somebody look at this
<f00b4r0>
I'm not sure why there's so much reticence, even within OpenWRT devs. This is a boot-only code path, it doesn't disturb anything and is a NOP as far as every other platform is concerned
<robimarko>
Yeah, it just does basically what pre-cal does
<robimarko>
First lookup the full name file
<f00b4r0>
my takeaway here is to no longer purchase mikrotik hw, and to stop caring about porting it to OpenWRT altogether.
<robimarko>
Cause just requesting board.bin makes no sense
<neggles>
FWIW, this same loading approach would be very useful for at least one other device (that i still need to put together a PR for)
<robimarko>
neggles: Can you comment on the PR?
<f00b4r0>
^
<neggles>
sure
<robimarko>
And potentially on the linux-wireless patch as well
<robimarko>
Cause, currently buying a new IPQ40xx Mikrotik board is fify
<neggles>
the sophos APX320/530/740 have *six* different pairs of board calibration data depending on the region you're in, and they all have the same bmi_id
<robimarko>
*iffy
<karlp>
anyone seen "*** pylibfdt does not seem to be available with python3" with a sunxi build?
<f00b4r0>
neggles: it was only a matter of time before other vendors started doing the same
<neggles>
f00b4r0: you should see the absolute nightmare of a script they run to determine which files to load
<f00b4r0>
neggles: I'm sure I'd rather not ;-)
<robimarko>
neggles: So, they are one of the rare ones that actually care about the region in calibration?
<robimarko>
Usually everybody including QCA just ignore that compeltely
<neggles>
robimarko: yep - you've got FCC, ETSI, Israel, and one other I think
<neggles>
they actually have seven or eight region folders but a few of them have files with the same sha256sum
<robimarko>
Probably Japan
<neggles>
e.g. there's a dedicated folder for australia, but the files are the same as ETSI
<robimarko>
Most vendors kind of have ETSI/FCC and JP
<robimarko>
BDF-s
<robimarko>
And then we pick ETSI usuallly
<robimarko>
Cause, we really have no way or picking the right one on runtime in OpenWrt
<neggles>
ahhh that's right, the reason the script is such a nightmare is that it supports five different chipsets as well
<karlp>
I can build uboot locally, ok, same 2020.04 version that we have.
<robimarko>
And ETSI is usually the most restrictive one
<neggles>
AU, ETSI, FCC, IC, IL, ROW
<neggles>
ah sorry ignore AU, that's an artifact of me being in australia - ROW has 4 subfolders, AU BR CA CR
<neggles>
t reads out the country code the device was originally manufactured with, and if that matches the country code in the config, it loads from yet another copy stored in the boot NOR flash
<neggles>
if it doesn't it does a bunch of gymnastics and copying
<f00b4r0>
that must take forever
<neggles>
f00b4r0: they are running four separate access point models out of a single 64MiB FIT image with no rootfs
<f00b4r0>
yum
<neggles>
some ipq4019, some ipq8064
<neggles>
oh and two of them have the ability to run in 5GHz+5GHz mode instead of 2.4GHz+5GHz
<neggles>
which ALSO requires the BDF be swapped out
<neggles>
AFAIK, at the moment there is no practical way to allow a single openwrt image to support both modes
<neggles>
I believe this would allow that with a little bit of effort (and yes, all the wifi chips in the devices in question are ath10k)
GNUmoon has quit [Ping timeout: 480 seconds]
<neggles>
robimarko / f00b4r0: commented :) probably longer than it needed to be, but.
<rmilecki>
stintel: wow, nice someone found & fixed that leak
FireBurn has quit [Read error: No route to host]
FireBurn has joined #openwrt-devel
Tapper has joined #openwrt-devel
wulfy23 has joined #openwrt-devel
amaumene has joined #openwrt-devel
eduardo010174 has joined #openwrt-devel
<ynezz>
FireBurn: exactly what you're doing now, I'll merge it if nobody objects
<reiffert>
Hey guys - I'm migrating my working strongswan/swanctl setup from Debian to Openwrt but it appears the updown script listed in the config never gets called. swanctl --stats lists the updown plugin as being loaded.
jow_laptop has joined #openwrt-devel
jow_laptop is now known as jow
dangole has joined #openwrt-devel
GNUmoon has joined #openwrt-devel
<reiffert>
stintel: I believe it was you mentioning xfrm to me yesterday when I found that I cannot pass traffic while the openwrt firewall is started. Are you by chance using an updown script in your swanctl.conf in openwrt?
minimal has joined #openwrt-devel
<stintel>
reiffert: not on OpenWrt
<FireBurn>
ynezz, fantastic
<stintel>
I just use option zone 'vpn' on the interface with xfrm proto
<stintel>
and then it integrates nicely with firewall3/4
<FireBurn>
I've been having issues building openwrt 21.02 since I upgraded to glibc 2.34
<FireBurn>
So I've not rolled my own 21.02.1 yet
<ynezz>
it's unlikely to be available in 21.02
<ynezz>
it would be available in next release
<FireBurn>
23.08?
<FireBurn>
Also are there any recommend wifi6 routers for OpenWRT yet?
<PaulFertser>
Filter ToH by 802.11ax , you'll see FireBurn
<robimarko>
Officially there are only 3 or so
<robimarko>
U6 Lite/LR
<FireBurn>
It was more, which one would you recommend folk go out and buy
<robimarko>
And there is Belkin/Linksys rebranded one
<PaulFertser>
FireBurn: rt3200 is the current favourite for "folks".
<FireBurn>
"thanks"
<reiffert>
strongswan on openwrt seems to ignore the if_id_in if_id_out setting ...
<reiffert>
inside updown scripts on Debian the variable $PLUTO_IF_ID_IN is set to the value specified in the swanctl.conf file. on openwrt the variable is empty / not existing
<stintel>
reiffert: it definitely does not ignore that setting or none of my xfrm tunnels would work
<stintel>
stay away from Unifi 6 Lite/LR, it's wifi4 on 2.4GHz
<PaulFertser>
stintel: what do you think is so bad about using 802.11n on 2.4 GHz? Lower QAM? But are those higher QAM modes actually usable?
<robimarko>
UBNT release U6 Pro with dual band AX
<robimarko>
But I have no idea whats inside
<robimarko>
Its actually cheaper then LR
<stintel>
PaulFertser: no TWT
<stintel>
if anything needs improved spectrum efficiency it's the 2.4GHz band, and exactly that is missing from those Unifi APs
<PaulFertser>
stintel: does it help if plenty of 802.11n are still around?
<stintel>
I can recommend TP-Link EAP615-Wall :P I'll try to get it in master this month
<reiffert>
stintel: swanctl --load-all with default = 3 "debug" shows the if_id_in and _out value I chose in swanctl.conf - however the _updown script doesn't know that variable.
<stintel>
PaulFertser: I'm not an expert on all that wireless magic, but it doesn't make sense to buy something new that contains such outdated technology
rua has quit [Quit: Leaving.]
<reiffert>
on Debian it's in ${PLUTO_IF_ID_IN}
<stintel>
PaulFertser: if you settle for a 5ghz ax 2.4ghz n now, introducing 2.4ghz ax clients later will not benefit from the fancy stuff
<PaulFertser>
stintel: I see the reasoning.
<stintel>
and most people don't buy new APs that often, so we really shouldn't recommend anyone to buy these products
<FireBurn>
Does that mean there aren't any 6Ghz triband ones yet?
<stintel>
PaulFertser: you can also get TP-Link EAP615-Wall for example for less than Unifi 6 Lite
<stintel>
and it's dual-band ax
<stintel>
JCG Q20, Tenbay T-MB5EU-V01, TOTOLINK X5000R are currently supported in master
<PaulFertser>
Those are 2x2 on both bands.
<stintel>
like the U6Lite
<PaulFertser>
And EAP615 is 2x2 as well if it's mt7975dn radio based
<stintel>
it is
<stintel>
I'm running my U6LR with the 2.4GHz radio disabled
rua has joined #openwrt-devel
<PaulFertser>
I have an impression wanting 802.11ax on 2.4 GHz currently isn't worth trading for 4x4 capability on 5 GHz.
<stintel>
I'll gladly install a few dual-band 2x2 ax APs and have better coverage in the process than spend more on a single 4x4 5ghz ax 2x2 2.4ghz n
<reiffert>
stintel: something else is odd on openwrt: tcpdump -n -i tunnelif doesn't show any traffic and RX/TX counters on ifconfig tunnelif don't increment. But traffic is being passed. Hmmmm.
<stintel>
reiffert: I would revise my configs if I were you. sounds like something is misconfigured
<stintel>
you can't compare apples with bulldozers
eduardo010174 has quit [Quit: Leaving]
<ynezz>
are there any known issues in 21.02/master regarding dhcpv6 on wan interface? odhcp6c[2705]: Failed to send SOLICIT message to ff02::1:2 (Address not available)
<reiffert>
stintel: thank you. I'll dig into it.
<stintel>
ynezz: I've spotted a load of odhcp6c errors in my log archive but not that one
<stintel>
ynezz: although recently a similar one: odhcp6c[7152]: Failed to send RS (Address not available)
<stintel>
that apparently also appeared in June last year
wulfy23 has quit [Remote host closed the connection]
<reiffert>
ip xfrm policy is showing if_id 0x3 on debian but is missing on openwrt.
<stintel>
shows if_id 0x32 for me
<stintel>
maybe you need ip-full
<stintel>
you do have kmod-xfrm-interface and xfrm packages installed, right ?
<mangix>
Hrm missed ax discussion.
<mangix>
E7350 is a nice alternative to E7450 for those that want ax on 2.4ghz
<mangix>
PR needs review again
<mangix>
Hopefully before branching happens.
noltari_ has joined #openwrt-devel
<reiffert>
stintel: it is ip full since ip-tiny didnt know xfrm
<nitroshift>
stintel, o/
<stintel>
hey nitroshift
noltari has quit [Ping timeout: 480 seconds]
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<hurricos>
Slimey: Stop pointing your nvmem-cells section at partition@1ff0000 >:o(
<hurricos>
alternatively, if you're not doing so, post your dts and challenge someone else here to explain why the dtc is taking your dts instructions and blending them into a fine paste
<reiffert>
stintel: after emptying /etc/ipsec.conf where the old and also working config was in ...
<Slimey>
lol
<hurricos>
Pretty sure you're doing the latter correctly, so if make distclean / rebuilding everything isn't working, it'd floor anyone else here
<hurricos>
s/latter/former/
<hurricos>
wow, my brain is not awake today
<Slimey>
this is from the one that works but is kernel 4.x
<reiffert>
stintel: all I have to do to make it work after a restart is issuing "swanctl --load-all". What's the preferred way here, have a script run that or go for /etc/config/ipsec ?
<stintel>
reiffert: the strongswan package should have a file included that does that already
rmilecki has quit [Quit: Konversation terminated!]
dangole has quit [Remote host closed the connection]
dangole has joined #openwrt-devel
<reiffert>
stintel: hm.
<stintel>
I don't remember the details
<rsalvaterra>
Anyone around here using git-pw?
<stintel>
o/
<rsalvaterra>
Nice! I'm gonna nag you to help me configure it. xD
<stintel>
RTFM you noob :D
danitool has joined #openwrt-devel
<stintel>
jk
<rsalvaterra>
Of course I'm reading it. ;(
<rsalvaterra>
;)
<rsalvaterra>
I wonder if I have to authenticate just to get bundles and applying them locally…
* rsalvaterra
was today-years-old when he learned about git-pw.
<stintel>
I've had it in my gentoo overlay for years
<rsalvaterra>
stintel: Up and running. Pretty easy, actually. :)
<stintel>
yep
<reiffert>
ah .. so etc/config/ipsec is missing -> no config_foreach config_ipsec ipsec, etc
<rsalvaterra>
Now I need to read the docs, since I want to apply a specific series.
<rsalvaterra>
git-pw series list
<rsalvaterra>
Ooooh, this is sweet.
<stintel>
reiffert: strongswan is complex, ipsec.conf -> swanctl was not an improvement imho
<stintel>
but required as ipsec.conf wasn't getting new features like xfrm interface support
<dwmw2>
I'd be interested in actual packet performance. I spent a while playing with openconnect last year and found I could make it go faster than the kernel's ESP, on x86 :)
<dwmw2>
It's quite tempting to add IKEv2 support and call it an IPSec client :)
<Slimey>
how do i push my current branch to a new one?
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<Slimey>
nm got it
Borromini has joined #openwrt-devel
<mrkiko>
dwmw2: interesting! What's the approach used to make it go faster?
<dwmw2>
I abuse vhost_net to get packets to/from the tun device, like a VMM would.
<dwmw2>
On x86 I also lifted the stitched aes/sha asm routines from OpenSSL to really speed things up.
<dwmw2>
Basically trying to achieve as close to zero-copy as I cna.
<dwmw2>
Batching send/receive using recvmmsg/sendmmsg.
<mangix>
more like, get something actively maintained in the tree. Currently I'm looking at libarchive as it looks like it can replace a bunch of tools. Although it's maybe too heavy.
<aparcar>
mangix: this 1k line unzip thing looks fine
<aparcar>
hauke: thanks for pushing for 5.10 :)
rua has joined #openwrt-devel
<mangix>
aparcar: that 1k thing builds on freebsd. No idea about elsewhere.