c0sm1cSlug has quit [Remote host closed the connection]
c0sm1cSlug has joined #openwrt-devel
<Grommish>
neggles: ping.. Leak is still present, just very very slow.. grr
<Grommish>
neggles: I'm just going to let it sit for a week and see if it levels off or not
rua1 has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
f00b4r0 has quit [Ping timeout: 480 seconds]
dave has joined #openwrt-devel
dave has left #openwrt-devel [#openwrt-devel]
goliath has quit [Quit: SIGSEGV]
fblaese has joined #openwrt-devel
Tapper has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
philipp64 has quit [Quit: philipp64]
<hurricos>
one thought occurs to me RE: Octeon leaks
<hurricos>
CN6640 can have its memory read arbitrarily
<Slimey>
memory leaks?
<hurricos>
Yes, there's a leak in kernel code in (folks above seem to think) octeon-ethernet
<hurricos>
... I do wish we had a nice little prometheus exporter in Lua that could parse /proc/$pid/maps
<hurricos>
I don't imagine it'd be performant at all
philipp64 has joined #openwrt-devel
Tapper has quit [Read error: Connection reset by peer]
Tapper has joined #openwrt-devel
minimal has quit []
danitool has quit [Ping timeout: 480 seconds]
lmore377 has quit [Read error: Connection reset by peer]
lmore377 has joined #openwrt-devel
danitool has joined #openwrt-devel
<Grommish>
hurricos: So... You've spoken of odd and spooky magik.. Do tell :)
cmonroe has quit [Remote host closed the connection]
cmonroe has joined #openwrt-devel
GNUmoon has quit [Ping timeout: 480 seconds]
valku has quit [Quit: valku]
rose has joined #openwrt-devel
rose_ has joined #openwrt-devel
<rose_>
Do we need to get GPL License for versions above 15.05 in openwrt.
<rose_>
I see that most of the routers like TP-LINK and MI use 15.05 only despite having 21.02 in market is GPL stopping them from using versions above 15.05
rose has quit [Remote host closed the connection]
rose_ has quit []
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<stintel>
are people finally realizing most vendor SDKs are ancient and accidents waiting to happen?
<Grommish>
stintel: I think it's a matter of people trying to market low-cost consumer hardware to fill the "new" need of crap-tasic OEM firmware void and wonder what has spooked people away from beyond CC
<Grommish>
It's all been RaspPis or other low-end mass hardware
guidosarducci_ has joined #openwrt-devel
guidosarducci has quit [Ping timeout: 480 seconds]
<Grommish>
neggles: ping
<Grommish>
stintel: You don't have a platform for testing Octeon anymore, right?
<stintel>
I do
<stintel>
just not a lot of incentive
<Grommish>
stintel: I removed IPv6 which saw a slow-down, but steady leak
<Grommish>
stintel: I just removed procd ujail and I actually see RAM being free'd
<Grommish>
But I don't know how to verify if it's just an illusion or not, so I need someone who can actually understand what they see
<Grommish>
Yeah. Leak seems to be no more.. I actually had to check to see it was a 5.10 build
<Grommish>
Could ujail be seeing the fpa disappear with the ram and not freeing it when it gets back?
danitool has joined #openwrt-devel
<Grommish>
I mean, we know it does it because it causes the kmemleak log entries and we know the leak seems to be networking based, we pointed at dnsmasq.. ujail'd dnsmasq also came in between 5.4 and 5.10
<stintel>
my ERL was not running dnsmasq in backup state
<Grommish>
but what else uses ujail?
<stintel>
maybe ntp
<Grommish>
Rightnow, it's supposition on my part.. however, I am still at 29080
<Grommish>
and I dropped to 29000
<Grommish>
so might be worth a shot since you know what th eleak looks like
<Grommish>
I don't know if neggles is running ujail or not
<Grommish>
Not much is standard in his builds hehe
<stintel>
I don't know details really, all I know is my ERL rebooted every ~8h when running 5.10
<Grommish>
*nod* Well.. I'll keep testing with it
<Grommish>
We've had a couple of false leads, this is probably another one :D
<Grommish>
Ok.. I just pushed 64GB thru iperf and no increase
<Grommish>
RX bytes:4195237258 (3.9 GiB) TX bytes:102884268595 (95.8 GiB) - and at 33052 used on free.. it's procd ujail
Borromini has joined #openwrt-devel
<Grommish>
I'm going to turn IPv6 back on and leave ujail off and see what happens
<Borromini>
Grommish: doubling down on that octeon memleak? :)
<Grommish>
Borromini: I turned ujail off and the leak went away
<Borromini>
oh :)
<Borromini>
well my ER-4 is out the door :P
<Borromini>
i'm eyeing that sleek MikroTik RB5009UG
<Grommish>
Unfortuneatly, the 2 Itus boxes are the best I got hehe i use a Zyxel as a dumb AP and a Nighthawk R8000 that is currently broken
<Borromini>
:D
<Borromini>
i'm trying to resuscitate my XGS1250-12.
<Grommish>
I'm profoundly greatful for itworking as well as it does.. My boxes are actually public facing hehe talk about doubling down
<neggles>
Grommish: wha
<Grommish>
neggles: No joke
stokito has joined #openwrt-devel
<Borromini>
=)
<stintel>
ugh, that mikrotik is almost useful
<Grommish>
neggles: I killed procd ujail
<neggles>
what uses ujail
<stintel>
if only the PoE-PD port would be 10GbE
<Grommish>
neggles: dnsmasq for one
<stintel>
sigh
<neggles>
should I be seeing a process?
<Borromini>
neggles: thanks for your help again btw. i'm going to try just write a hopefully functional openwrt image this time, before trying to replace the u-boot env
<Grommish>
neggles: and ujail went into as default between 5.4 and 5.10
<Borromini>
stintel: you're not easily satisfied are you :P
<Grommish>
neggles: Yes
<Grommish>
neggles: It'll show as a process
<stintel>
Borromini: am not :)
<neggles>
i see no ujail in `ps`
<Grommish>
neggles: I figured you didn't roll with ujail hehe
<neggles>
but
<neggles>
i stopped dnsmasq...
<stintel>
Borromini: where did you order the eap615 ?
<Borromini>
stintel: i think with Azerty
<Borromini>
Brother's getting it for his new office and i'm piggybacking onto his orders =) no VAT <3
<Grommish>
neggles: But, it's super easy to test.. just remove procd ujail from Base and build out don't even need a clean
<Borromini>
stintel: yeah, Azerty.
<stintel>
ETA?
<Grommish>
neggles: I'm actually seeing RAM get free'd
<Borromini>
but I assume that doesn't apply to you being in BG?
<stintel>
well my 2 came from dectdirect.nl :P
<stintel>
there was this dodgy polish webshop that required me to contact them with proof of company registration et
<stintel>
etc*
<Borromini>
:) yeah they don't have 'em anymore. Our RB5009UG's came from there
<Borromini>
from Dectdirect that is
<neggles>
Borromini: the RB5009UG is *really good* fwiw
<neggles>
can push 1gbps of wireguard at <35% cpu
<stokito>
Hi, Is there any process of deciding what to include into OpenWrt builds by default? I sent a letter to devlist "Pre-install MiniUPnPd on OpenWrt by default" but it seems not interesting in the context of devlist which is used mostly for patches.
<stintel>
can it route 10GbE on the SFP+ between different VLANs ?
<neggles>
yes'm
<neggles>
the 10G is direct-to-CPU
<Grommish>
stokito: UPNP of any kind as default is a huge security attack surface
<neggles>
(I think?) the rest of the ports are hanging off a PIPE
<stintel>
stokito: miniupnpd is in the packages feed, and afaik we do not install things from the package feed by default
<Grommish>
stokito: even OEM's have learned to turn it off or not include it
<Borromini>
neggles: sounds as awesome as it looks. Will be replacing an APU2 (at my brother's) and an EdgeRouter 4 here
<neggles>
10G full duplex to CPU, and the PIPE is not a dumb switch, it can do linerate L3
<Grommish>
Borromini: Oo.. but, that means the ER4 just will be ale to be tested on :D
<neggles>
it's just quite limited in capabilities
<stokito>
yes but still miniupnpd is supported and all security issues was resolved years ago
<neggles>
stokito: UPnP, in and of itself, is a *massive* security issue
<stintel>
stokito: there is no such thing as "all security issues resolved years ago"
<Borromini>
stintel: ETA end of this month, Azerty said
<neggles>
thousands of WD myclouds all around the globe hacked into and wiped, just because UPnP is enabled by default on most routers' OEM firmwares
<Grommish>
UPNP itself *is* the security issue.. the ability for any Application to open a port to the outside is ludicris
<stokito>
ok, what about to have just a NAT-PMP server which is much simpler and easier to audit?
<Borromini>
Grommish: if only... it sold in a jiffy :). And I reflashed the EdgeRouter Lite to Ubiquiti's stuff again as well, that's up for sale too
<neggles>
stokito: Any system which allows a device to request an external:internal port mapping without any end-user input is a HORRIBLE IDEA.
<neggles>
If you understand enough about networking etc. to use UPnP and/or NAT-PMP "safely" (insofar as that's possible), then you are also capable of building a custom image, or just installing it yourself.
<neggles>
it does not need to be, should not need to be, and probably will never be built in by default.
<Grommish>
stokito: But.. You can use Imagebuilder if you don't want to build from source and include miniupnp in your builds
<Grommish>
stokito: You'll better understand what issues your network might face and can decide for yourself..
<stintel>
stokito: it's unlikely to happen, due to miniupnpd being in the packages feed, unmaintained, considered a major security risk, we prefer security over convenience, etc
<neggles>
Borromini: i am very, very happy with mine - it's actually slightly too good, which is annoying, because now I don't want to use it for what I bought it for
<neggles>
my only complaint is the internal flash is a bit small, and it doesn't have a microSD slot, so i'm stuck using up the only USB port on a flash drive + using a hub if I want to connect something else
<stintel>
neggles: but can it do 10GbE inter-vlan routing?
<neggles>
yes
<neggles>
basic inter-vlan routing is handled in the PIPE
<Borromini>
neggles: the ER4?
<stintel>
sorry missed your previous answer
<Borromini>
neggles: oh you meant the RB5009UG I assume :)
<neggles>
i'm running rOS 7.1.1 on it atm because while most of the bits and pieces are mostly mainlined, that code's relatively fresh
<Grommish>
stokito: Why not just Port forward then
<neggles>
or just install it yourself?
<Grommish>
Ehich is what UPnP does.
<neggles>
if you want UPnP, just... install miniupnpd? if you really badly need it baked into the image, as stintel and myself and at least one other have already said in here, use sysupgrade.openwrt.org or the image builder to make a custom image
<neggles>
if you are capable of installing and configuring OpenWRT, you are capable of installing and configuring a UPnP daemon
<stintel>
OpenWrt*
goliath has joined #openwrt-devel
<neggles>
stintel: the name comes from the WRT54G, fight me >:D
<Grommish>
Wait.. What kind of Application are you making that you want to make UPnP a default in OpenWrt?
<Grommish>
What possible Venn Diagram from hell are you using
<stintel>
neggles: check the official logo
<stintel>
or the banner when logging in
<Grommish>
Malware? Wormable Hello World? I mean.. What?
<neggles>
stintel: I have read through the branding pack, I am aware, I just disagree :P
<neggles>
raise your hand if you've installed white russian on a WRT54G v1.1
* neggles
raises hand
<stintel>
o/
<Borromini>
does it count if i installed it on a shitty BCM 802.11bgn router? :P
<neggles>
probably :P
<neggles>
I still have that v1.1 somewhere
<neggles>
and a WRT54GL v1.0
<neggles>
anyway
<Borromini>
=)
<Borromini>
guys flashrom is balking at me because the image size doesn't match the flash size... Yet i'm only trying to write to a specified region
<neggles>
stintel: you're right and I won't even begin to pretend you aren't :P
<neggles>
want me to do some throughput tests on the 5009 with some assorted filter rules?
<stintel>
neggles: would be cool
<stintel>
right now I'm terminating some VLANs on my L3 switch but it has no stateful filtering so I consider it somewhat a security risk
<stintel>
and ACLs are a big nono
<neggles>
got a particular set of rules or shall i just mix some stuff up
<stintel>
I could wgetpaste you my /etc/config/firewall but ehm :P
<stintel>
you'd go "WTF?!"
<neggles>
heh.
<neggles>
I suspect i'd go "Ah, I see you are also a terrible person"
<stintel>
1213 /etc/config/firewall
<neggles>
i have 42 filter rules, 38 NAT rules, and 37 mangle rules (I run routeros because i like it and i use it at work)
rua has joined #openwrt-devel
<stintel>
also have like 20 network interfaces :P
<PaulFertser>
Borromini: run with more debug. Guess it doesn't like something about your layout file
<neggles>
stintel: 17 :P
<Borromini>
PaulFertser: ok, will do
<stintel>
neggles: ah but the rb is running routeros? not openwrt/
<neggles>
yes, I don't think we have a build for the 5009 yet? iirc it needs to wait for kernel 5.15
<stintel>
in t hat case nvm
<stintel>
I can't stand routeros
<PaulFertser>
Borromini: and you didn't show the contents of the layout file
<stintel>
that CLI makes my head explode
<neggles>
rOS 7's CLI is much, much improved
<neggles>
but you're a crazy person if you use the CLI for most things, really
<neggles>
winbox <3
<stintel>
*puke*
<stokito>
@Gromish I building a p2p program with VoIP capabilities intended for a not so expirienced users. And I would like it to make it simple to use. That's why I'm interseted to have a wider support of NAT-PMP. I wondered that OpenWrt doesn't have it out of the box and in fact this makes my current VoIP programs works slower.
<PaulFertser>
Borromini: I think you need to pad the sysupgrade image to match the region size exactly, you can do that with "truncate -s"
<neggles>
stokito: yeah, no, there's a reason why there are no p2p voip applications that anyone actually uses
<neggles>
UPnP will not solve your problem
<neggles>
you have forgotten about CG-NAT, for one
<Borromini>
PaulFertser: ok, how would i do that exactly?
<Borromini>
i assume i need dd for that then
<neggles>
Borromini: if it's too long, no
<PaulFertser>
Borromini: no, just 'truncate -s <region size in bytes> sysupgrade.bin'
<stokito>
Ok, thank you for answers. You refined more to me
<neggles>
stokito: VoIP is an absolute nightmare at the best of times; just having UPnP present on OpenWrt by default would not make a significant difference to this.
<neggles>
even if it was preinstalled, it would *never* be enabled out of the box
<Borromini>
neggles: PaulFertser: ok. so truncate to 13631487 bytes then
<stintel>
the other problem with have upnp by default is that at some point someone is going to hold us responsible for their network being hacked
<PaulFertser>
Borromini: probably it should end in 8
<stintel>
and while the license says "no warranty" ... having to deal with such people is very very very demotivating
<stintel>
so no, you're probably not going to convince many of us that including that by default is a good idea
<Grommish>
Well.. rsalvaterra would nix it simply for the 2k it adds or whatever
<PaulFertser>
neggles: +1 byte
<neggles>
PaulFertser: you are right
<Borromini>
ok so 13631488
<neggles>
zero-indexing
<neggles>
always gets ya
<Borromini>
:)
<svanheule>
Borromini: and if that doesn't work, try adding 0x300000 bytes of padding at the front too
<Borromini>
o_O
<svanheule>
Borromini: flashrom migth require a full flash image to extract the partition from
<PaulFertser>
Ah, yes
<svanheule>
at least, that's what the error message appears to indicate
<PaulFertser>
I think that's the right way to look at it, thank you svanheule
<Borromini>
truncated file (which is bigger?) gives 'Error: Image size (13631488 B) doesn't match the flash chip's size (16777216 B)!'
<neggles>
Borromini: I would be inclined to just dump the whole flash chip, copy the dump, `dd if=sysupgrade.bin of=copy-of-dump oflag=seek_bytes seek=$((0x3000000)) conv=notrunc`, then flash back the whole dump
<Borromini>
svanheule: so you're saying i should still write a full image same size as the flash?
<Borromini>
neggles: i will do that then, thank you.
<PaulFertser>
Borromini: you should have it but flashrom will write only part of it
<neggles>
i've never had much luck with flashing regions through flashrom
<Borromini>
bit weird flashrom refuses to write regions
<Borromini>
good to know neggles :P
<neggles>
I think PaulFertser and svanheule have the answer though
<Borromini>
PaulFertser: what do you mean by 'it'? I have a full dump.
<neggles>
flashrom still wants a file that's full-dump-length
<PaulFertser>
I checked my notes, the image needs to be the size of teh flash indeed, and flashrom can choose parts of it to be written.
<Borromini>
PaulFertser: ah!
<svanheule>
neggles: I remember having to fiddle with it to make it work, but I don't remember what I did in the end ...
<neggles>
but if you tell it to only flash a region, it will only flash that region from the file?
<Borromini>
thanks
<neggles>
wow okay
f00b4r0 has joined #openwrt-devel
<Borromini>
that's dense.
<neggles>
that explains why I've never had that work
<neggles>
unintuitive
<svanheule>
stintel: what PoE chip is in the Unifi Switch Flex?
<Borromini>
while we're at it: is there a way to upgrade a PoE chip's firmware on OpenWrt? :P
<Borromini>
my GS1900-8HP balks when i try an actual PoE device on it since quite a while (both 21.02 and master)
<Borromini>
gonna try OEM firmware as well so see if it's an OpenWrt issue. Same PoE+ AP just works fine on my 'production' GS1900-10HP with 21.02, weirdly enough
<neggles>
Borromini: gonna depend on what the PD controller chip is :P
<Borromini>
:P
<neggles>
s/PD/PSE
<Borromini>
ST Micro ST32F100 or a sibling
<stintel>
booted an SNIC10e, configured an interface, enabled packet steeering, configured bird (ospf), and a cron job: * * * * * ( date; free; echo ) >> /tmp/mem.log
<neggles>
Borromini: so they've put custom firmware on a microcontroller
<Borromini>
i just checked on the GS1900-10HP, they're very similar, will boot the GS1900-8HP
<Borromini>
neggles: isn't that customary for PoE?
* Borromini
has no idea
<svanheule>
stintel: thanks, didn't manage to find the commit message
<stintel>
actually I added prometheus-node-exporter-lua, no need for the cron job
<neggles>
Borromini: nope! there are plenty of purpose-built PoE PSE controllers
<stintel>
svanheule: you need to try harder or get a coffee ;p
<svanheule>
stintel: already did the coffee thing, guess trying harder is up next
<Borromini>
:P
<neggles>
upgrading the firmware on that custom job is definitely something you can do from openwrt, assuming it has SWD/nRST hooked up to the SoC, but... good luck getting the firmware for it, unless it's been bundled into a later OEM fw
<stintel>
svanheule: but you're welcome ;)
<Borromini>
neggles: yeah, i figured so. not something they're gonna hand out to everyone i reckon
<neggles>
if they've issued an update for it ever, it'll be in the OEM rootfs
<neggles>
but if that's the case, upgrading to the latest oem fw would also upgrade it
<svanheule>
stintel: found it! a98396978 doesn't mention "unifi switch flex", so that's why I missed it first
<Borromini>
neggles: good point, i'm going to give the OEM firmware a shot first
<stintel>
svanheule: yeah, I searched for Flex
<stintel>
Grommish: my snic10e image does not include dnsmasq nor ujail and it showed the leak
<Borromini>
neggles: on the u-boot settings, since you said the tarball contains XGS1250-12 code as well, would it be feasible to recompile u-boot with the prompt setting enabled? then flash it?
<Grommish>
stintel: Interesting.. I just turned IPv6 back on and no leak
<Borromini>
since i'm messing with flash dumps anyway >_>
<neggles>
Borromini: thing is, the prompt *is* enabled
<neggles>
it's just set to an empty string with a 1s timeout
<neggles>
so you can't see it
<Borromini>
nor can i intercept it i suppose, since i've been mistreating my keyboard like a crazy person and u-boot doesn't budge...
<neggles>
tapping spacebar repeatedly should halt it, but they might have bypassed that
<Borromini>
didn't work, tried a lot of keys, spacebar chief among them
<neggles>
since you've got the flash dump you could just erase the kernel partition...
<neggles>
then it'll fail to boot and drop you to a prompt
<neggles>
and dnsmasq still causes mem usage to climb with it only bound to ipv4
<neggles>
it's just a *lot* slower, though it climbs faster if I do more dns lookups or restart it
<stintel>
yes IPv6 is on
<Grommish>
I'm going to run a multi-thread iperf for 5 minutes and see what happens
<Grommish>
Any suggestions on how to rapidly DNS query to my gatewy then?
<neggles>
I can reliably add 300kb to used ram just through `service dnsmasq restart`
<neggles>
actually over 1mib
<stintel>
neggles: does procd mem usage grow ?
<neggles>
stintel: good question, let me see
<stintel>
anyway, I've set up prometheus-node-exporter-lua on my snic10e and configured bird and conntrackd, which is what's always active regardless of router state (primary vs backup)
<stintel>
let's see if that shows anything useful in the graphs after I take a nap
<neggles>
looks like no
<neggles>
but i am not entirely sure how to check since htop is not being very helpful
<stintel>
cat /proc/1/status
<neggles>
stintel: completely unchanged
<neggles>
it gained 150 voluntary context switches, that's it
<dwmw2_gone>
Hm, is a mib 1/1048576 of a bit?
srslypascal is now known as Guest993
srslypascal has joined #openwrt-devel
<neggles>
a MiB is 1048576 bytes is it not
<dwmw2_gone>
MiB is, yes.
<dwmw2_gone>
Mib is 1048576 bits, so 262144 bytes
<dwmw2_gone>
Mb is 1000000 bits
<dwmw2_gone>
mb is 1/1000 of a bit.
<dwmw2_gone>
So I guest mib would be 1/1024 of a bit?
<dwmw2_gone>
512mib would be half a bit?
srslypascal is now known as Guest994
srslypascal has joined #openwrt-devel
<neggles>
how does one have a fractional bit
<dwmw2_gone>
Dunno, but I see people saying mb so it must be possible.
<dwmw2_gone>
Must be something to do with quantum computing
<russell-->
ask a statistician
<f00b4r0>
:)
<dwmw2_gone>
oh... dangole_?
<dwmw2_gone>
Are you using i7.infradead.org right now? Can I reboot it a few times to set the Linux parallel CPU bringup stuff I've been playing with ?
<dwmw2_gone>
looks like you aren't logged in atm
stokito has quit [Quit: Page closed]
<Grommish>
neggles: I'm going to run iperf for the next hour or so and see what happens. I'll check it when I get up. Whee!
<neggles>
Grommish: whee indeed :P
Guest993 has quit [Ping timeout: 480 seconds]
<neggles>
I can very very reliably add >1MiB to used RAM just by cycling dnsmasq on and off though...
<neggles>
not looking forward to attempting to get the eMMC off the board...
<neggles>
Grommish: it does the same on yours?
<Grommish>
neggles: Yep
<neggles>
gotta be something to do with listening sockets
<dwmw2_gone>
neggles: what's that board?
<neggles>
dwmw2_gone: it is a T-Mobile Personal Cellspot V2 aka Nokia SS2FII aka Askey SFE3160S
<neggles>
LTE femto eNodeB
<dwmw2_gone>
nice
<neggles>
I have five of them, and I am almost certainly going to murder at least one of them
<stintel>
neggles: what kernel are you running on your SNIC .
<neggles>
stintel: your 5.10.35
<dwmw2_gone>
going to play with opengsm ?
<neggles>
with some bits and pieces added in for the pcie console driver, which does not work
<stintel>
neggles: ok, I'm on 5.10.92 and not seeing any leak atm
<Grommish>
I'm on .92 and I didn't see it
<Grommish>
Until I cycled dnsmasq
<stintel>
I think we can safely say we are looking at 2 different leaks
<Grommish>
then, as neggles reported, it's a steady climb each restart
<neggles>
dwmw2_gone: LTE, so Open5GS - assuming I can get into it
<stintel>
1 related to userspace / restarting dnsmasq
<stintel>
1 in kernel/network
srslypascal has joined #openwrt-devel
<neggles>
i have a few other LTE femtos inbound but it turns out they all use the same damn qca chip
<stintel>
kill it with fire!
<neggles>
it's LTE, there *are* other games in town but they're big angry TI FPGDSPGASoCs
<neggles>
yeah that acronym mashup didn't work, *TI FPGA/DSP/ARM SoCs
<Grommish>
neggles: What Cat is it?
srslypascal is now known as Guest997
srslypascal has joined #openwrt-devel
<Grommish>
I've got a older Cat4 Pepwave I use on vehicle
<neggles>
Grommish: categories don't quite apply to eNBs the same way as UEs, but, it can do 2x2 MIMO with QAM64 @ 20MHz, which is theoretically good for 150/50, and it can do 2CCA so twice that
<neggles>
oh nono, this is a *base station*
<neggles>
not a client, but a tiny cell tower
<Grommish>
Ah, gotcha I had one of those for Sprint at one point
<neggles>
MagicBox?
<Grommish>
because they had shite coverage here
<neggles>
the MagicBox devices are insane. They have one of these same setups for the eNB part, then *two* other full SoC setups
srslypascal is now known as Guest998
srslypascal has joined #openwrt-devel
<neggles>
made by AirSpan. the other Sprint ones are Casa Systems Pebbles, which are also FSM9955
<Grommish>
This wasn't recent
<Grommish>
But yeah I connected it to the the LAN so it could act as the base station
<neggles>
neither are these - the T-Mo ones were all shut off in 2020 iirc, and the Sprint ones were shut off on or before dec 1st 2021
<Grommish>
I had ot provide the backhaul but it was worth the service
<dwmw2_gone>
all a bit pointless these days now the phones themselves will just use your wifi for the backhaul
<neggles>
ah okay, not a magicbox, so either a Pebble or one of the older ones - little black rounded thing with a bunch of vent holes on top?
<stintel>
ahh, looks like conntrackd was not running on my main router
<stintel>
which greatly reduces the amount of traffic the backup router receives
<stintel>
fixed that, let's see if I get more leaky behaviour now
<neggles>
i have a wide assortment of functional GSM BTSes/LTE eNBs of various sizes, and a couple UMTS hNBs
<neggles>
AT&T ones
<Grommish>
stintel: I actually saw no issues pushing data, but the minute I flip dnsmasq it goes nuts
Guest997 has quit [Ping timeout: 480 seconds]
<Grommish>
Now, question.. I started this hunt looking at adblock and dnsmasq
<Grommish>
could adblock updating be causing dnsmasq to restart?
<f00b4r0>
neggles: you're lucky; here the operators will no longer provide cells to customers, relying on wifi calling instead, which as you pointed out, really sucks.
<neggles>
f00b4r0: oh make no mistake, none of these are attached to any carrier network. i have my own network.
<f00b4r0>
ah
Guest998 has quit [Ping timeout: 480 seconds]
<f00b4r0>
neat
srslypascal is now known as Guest999
srslypascal has joined #openwrt-devel
<neggles>
most of them are not even certified for use in australia, but I have a low power testing license
<neggles>
the ericssons are certified though, and quite easy to set up once you know the trick (and have the software package)
<f00b4r0>
nice
<neggles>
you can often pick up BNIB ericsson RBS 6402s on ebay for about US$30-40 because nobody knows how to make them do anything useful
<neggles>
(except for myself, some people who are under NDA, and about a half-dozen others that i know of)
<f00b4r0>
heh. Small audience :)
<neggles>
there are a lot of people who would like to know, but there is a lot of potential to get into a lot of trouble with your local regulatory authority if you are not careful with this gear
srslypascal is now known as Guest1000
srslypascal has joined #openwrt-devel
<f00b4r0>
makes sense
<neggles>
so I don't hand out the details to people who can't demonstrate some level of competency and understanding
<Grommish>
Small Private Cell Network? I saw a Defcon talk about SIM hacking that setup their own private netwokr for the event
<neggles>
that would've been Harald from Osmocom, and yep, that network was built using 6402s :)
<Grommish>
Fell outta the chair when they mentioned the Hayes baseband set still being used
Tapper has quit [Ping timeout: 480 seconds]
<neggles>
n.b. don't go asking the osmocom guys how to set up a 6402, they can't tell you
Guest999 has quit [Ping timeout: 480 seconds]
<Grommish>
Eh, beyond my knowledge, but it was a great talk
<dwmw2_gone>
What's so bad about wifi calling?
<f00b4r0>
i would be happy with my carrier being more cooperative. I don't get any signal here, and so if I don't actively try to catch some to re-register to the network, after a while wifi calling stops working (SMS and MMS stop even earlier), resulting in missed calls
<Grommish>
dwmw2_gone: If your Internet connection is good, nothing
<f00b4r0>
Grommish: not exactly, see above
<neggles>
dwmw2_gone: relies on your wifi signal being solid/stable, handoff doesn't work half the time
<dwmw2_gone>
Surely that's just as true of a femtocell?
<Grommish>
If your High Speed is 5Mbs, not so much
<neggles>
depends on the kind of femtocell, and because they use LTE to talk to your handset, there's no wifi interference issues
<f00b4r0>
many things don't work reliably with wifi calling. MMS being prime issue.
<neggles>
range is often much better than wifi range, as well
<f00b4r0>
^
<neggles>
and handoff is a lot more reliable since it works the same way as regular cell handoff
<f00b4r0>
^
<Grommish>
neggles: Less interference
<f00b4r0>
but as I said, it gets really ugly when you have zero coverage
<neggles>
Sprint's MagicBox devices took another approach that solves the local backhaul bandwidth problem, too - the MagicBox contains a very high-power 4x4 LTE client radio with high-gain antennas which it uses for backhaul
Guest1000 has quit [Ping timeout: 480 seconds]
<neggles>
downside, it does require you to have *some* coverage - but the Pebble and the t-mo cellspot both work even if there is no coverage, all you need is internet and GPS lock
<Grommish>
neggles: So the MagicBox was basically a customer-based amplifier?
<neggles>
Grommish: it is a fully-fledged fully-functional femto eNodeB, connected via internal ethernet to what's essentially an insanely overpowered 4G CPE device
<f00b4r0>
hmm. "Load key ".ssh/id_dropbear": invalid format" - for a freshly generated key, that's rather strange
<neggles>
you put it at the one spot in your house where you get vaguely usable signal, and it takes advantage of its large array of high-gain antennas to provide itself with backhaul
<neggles>
it's a lot bigger than a phone - they're not small at all, really - so it's got a lot of room for antenna
<dwmw2_gone>
That's useful information; thanks. I had never really had many issues. I mostly have crap reception at home so it just goes onto wifi calling and stays there; handoff isn't that much of an issue
<dwmw2_gone>
and I don't use MMS much :)
<dwmw2_gone>
And I also have decent wifi coverage around the house because there are lots of OpenWrt boxes :)
<f00b4r0>
heh :)
<dwmw2_gone>
and I made sure 802.11r fast roaming works properly too.
<neggles>
I have never had wifi calling work properly... maybe it works better these days, but i've had it disabled for years
<dwmw2_gone>
I once broke the GSM SoC firmware while flashing my phone, while I was in the US
<dwmw2_gone>
Which was nice because then it didn't *know* I was in the US, and didn't refuse to use wifi calling :)
<Grommish>
Ok.. I'm going to bed before I get sucked into something else.. Night all
<dwmw2_gone>
So I got "not-roaming" rates for calls etc.
<dwmw2_gone>
night Grommish
<neggles>
night :)
<f00b4r0>
nn
<Borromini>
nn
<neggles>
dwmw2_gone: I would like to have functional 802.11r but, well, i'm still using unifi firmware on my unifis because of reasosn
<neggles>
and yeaaaaaaaaah doesn't work. (well it works on the wpa2-enterprise SSID)
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
GNUmoon has joined #openwrt-devel
neggles has quit [Quit: bye friends - ZNC - https://znc.in]
Misanthropos has quit [Read error: Connection reset by peer]
<PaulFertser>
Now you have skills and tools to repair many more devices :)
<PaulFertser>
And to experiment more freely with this one.
<Borromini>
:)
<Borromini>
i still have a bricked TL-WR1043ND v2 here, might try my hand at it as well
Misanthropos has joined #openwrt-devel
srslypascal is now known as Guest1018
srslypascal has joined #openwrt-devel
Guest1018 has quit [Ping timeout: 480 seconds]
gladiac is now known as Guest1023
gladiac has joined #openwrt-devel
Tapper has joined #openwrt-devel
Guest1023 has quit [Ping timeout: 480 seconds]
srslypascal has quit [Ping timeout: 480 seconds]
<Slimey>
gm
srslypascal has joined #openwrt-devel
clayface has quit [Ping timeout: 480 seconds]
clayface has joined #openwrt-devel
minimal has joined #openwrt-devel
bkallus has joined #openwrt-devel
dangole_ has quit [Ping timeout: 480 seconds]
<bkallus>
Can anyone here explain to me what the cferam.000 files are in the bcm63xx-cfe repo? Are they extracted from official firmware images? Thanks.
goliath has quit [Quit: SIGSEGV]
<Slimey>
ugh java
goliath has joined #openwrt-devel
aiyion_ is now known as aiyion
bkallus has quit [Remote host closed the connection]
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<aparcar>
nbd: the build system passes ipk package paths to opkg to install them (during packages/install and target/install). why the full path and not just hand it the repositories and package names?
<aparcar>
instead of opkg /foo/bar/package.ipk we could do opkg --repository /foo/bar package
Gaspare has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
Gaspare has quit [Ping timeout: 480 seconds]
GNUmoon has quit [Ping timeout: 480 seconds]
<jow>
aparcar: I can think of two reasons... 1) it is easier (no need to prepare repository metadata beforehand) and 2) full control
<jow>
if you let opkg operate on an actual repo it might pull in additional dependencies which for whatever reason aren't selected or expected by the build system