<csharper2005>
Znevna: Can we read mtd? Yes. Can we change the luci code? Yes. So, it's possible.
<Znevna>
out of my skills :P
<stintel>
skills can be learned ;)
danitool has joined #openwrt-devel
<f00b4r0>
Znevna: this looks similar to what mikrotik does (which we expose in /sys/firmware/mikrotik), only more straightforward since it's all fixed position. I'm guessing an ideal scenario would be to have an nvmem cell for model name
<f00b4r0>
this could then be exposed to userspace
<robimarko>
f00b4r0: NVMEM layouts got merged for 6.3
<f00b4r0>
robimarko: ah!
<f00b4r0>
i should take a look then
<f00b4r0>
would you happen to have the commit hash, so I can see which variant got in?
<robimarko>
Yeah, now its possible to parse whatever TLV you want by adding a layout driver
<robimarko>
It got pretty much replaced with NVMEM
<Borromini>
:P
<Borromini>
csharper2005: that's what that directory is for, for hacks. Literally
<csharper2005>
Ok. Is it worth using mac-address-ascii for the new devices If it can be helpful? Or "hack" is some kind of the temporary dir and these patches will be dropped soon?
goliath has joined #openwrt-devel
gladiac has quit [Quit: k thx bye]
floof58 is now known as Guest705
floof58 has joined #openwrt-devel
Guest705 has quit [Ping timeout: 480 seconds]
robimarko_ has joined #openwrt-devel
robimarko has quit [Read error: Connection reset by peer]
<Borromini>
they'll need to be migrated to NVMEM one way or the othery I'd say. So better implement that right away.
<robimarko_>
I dont know if NVMEM supports ASCII MAC-s currently
<rmilecki>
f00b4r0: robimarko_: yeah, i'm working on NVMEM layouts, will resume my work on Monday
<robimarko_>
rmilecki: awesome
<rmilecki>
csharper2005: i hope to replace mac-address-ascii with proper NVMEM based implementation next week, up to two
<robimarko_>
rmilecki: it would be awesome to see ASCII support
<rmilecki>
we will get that :)
<robimarko_>
Then we could remove the hack completely
<rmilecki>
that's my plan
<robimarko_>
I really like the NVMEM layouts, they up on everything that is dynamic
<oliv3r[m]>
What is NO_EXCEPT_FILL used for, and why would it cause bootfailure. Is it in any way linked to the ENTRY_POINT (which seems to be offset by 0x400 similar to what fill address is doing)
<oliv3r[m]>
thanks! Let me read that. Though it might not be the holy grail I was after :) (might still cause things to not work :S)
<oliv3r[m]>
but is it worth my time to try to get 'generic mips kernel' working for openwrt/realtek?
<oliv3r[m]>
crap, I had seen already that post :(
<oliv3r[m]>
so I'll poke hauke at some point :)
<robimarko_>
Poke him now if he is online
<hauke>
oliv3r[m]: I forgot about this topic
<hauke>
it was 10 years ago
<oliv3r[m]>
So I'm trying to use generic_mips_kernel, cause, it sounds like the right thing to do. However, it doesn't boot (like zero output nothing). I'm trying to pin-point what (combination of) option is causing this, and so far, I found that NO_EXCEPT_FILL seems to play a part of this. So I can remove it from generic mips kernel for now, but if I wanted to make it work, does for example 'kernel_entrypoint' have anything to do with it? As they
<oliv3r[m]>
are both doing offsetty things at 0x400
<oliv3r[m]>
This is dark water for me, so not quite sure what i'm looking at
<hauke>
I do not know
<hauke>
but I am currently having a similar problem with the pistachio target
<hauke>
It was chanegd to mips generic in upstream kernel and now it does not boot up any more
<Tapper>
Hi people any news on the rebooting bug for r7800 in master?
<oliv3r[m]>
hauke: well it must work for someone, if they pushed that through :)
<oliv3r[m]>
hauke: but i'll just keep rambling and keep you in the loop of my findings
<hauke>
oliv3r[m]: yes I plan to ask on the mips mailing list for pistachio
<stintel>
sooo I can use matter's chip-tool to join my eve motion sensor with matter upgrade into my OpenThread network running on OpenWrt, but then on step 18 it fails because there is no response to the mdns query for the motion sensor. probably something missing in my otbr-agent
<stintel>
but good progress
<stintel>
I probably need SRP and or Advertising Proxy
<hauke>
stintel: nice
<hauke>
do you only have this eve motion ensor with matter or more?
<stintel>
eve motion and eve energy (the latter is not on the matter firmware yet)
<stintel>
I want to trick the eve app into believing there is an apple homepod or apple tv 4k thread border router in the network
<stintel>
and try updating the eve energy like that, without the apple tv 4k I bought yesterday
<hauke>
these eve devices are expensive
<stintel>
I tcpdumped everything while I was doing the upgrade of the eve motion, on an isolated network with only the ipad, apple tv and openwrt router
<hauke>
I think I paied less than 10€ for my power switches which work with Hue
<stintel>
hauke: they used to be homekit only and apple-only crap typically comes with a premium price tag
<stintel>
now that matter is out, we shall hopefully see much cheaper devices
<hauke>
yes I hope it will be adapted by many vendors
<hauke>
the fragmented smart home world is shitty
<stintel>
I'm like *really* happy with this sunxi a64 olinuxino devices
<stintel>
eMMC ftw
<hauke>
I hope it will be more successfully than AllJoyn and IoTivity/OCF
<stintel>
I've two of those olinuxino with a thread stick and otbr-agent (on OpenWrt of course!)
<csharper2005>
Tapper: do you have the issue number?
<Znevna>
tmp/.config-package.in:56157:symbol PACKAGE_luci-app-pbr depends on PACKAGE_luci-app-pbr
<Znevna>
x.x
<Znevna>
how do I get rid of this
<Borromini>
stintel: can I bug you about domotics? I'm looking around and I thought you had it to some extent
<Borromini>
it sounds like you took a deep dive though and that's not what I'm looking for =)
<stintel>
Borromini: I have not used domoticz in years
<Borromini>
i meant 'domotica' in general (no idea how that translates to English, thought it was 'domotics')
<stintel>
ah :)
<Borromini>
anything else you're using?
<stintel>
home-assistant
<Borromini>
and hardware/protocol wise?
<stintel>
zigbee, zwave, wifi
<Borromini>
ok ty
hanetzer1 is now known as hanetzer
Ryncewynd has joined #openwrt-devel
<stintel>
Borromini: but if I succeed in my current thread/matter experiment, and stuff appears stable, I'll most likely replace all the zwave and zigbee stuff
<stintel>
I'm actually hoping IKEA will come with thread/matter firmware upgrade for all the IKEA Zigbee lamps I currently own
<stintel>
but that might be wishful thinking ;)
<aiyion>
ucidef_set_led_netdev taking br_lan as trigger in DSA targets is an utter pain -.-'
<aiyion>
Would it be possible to refactor it to consume multiple interfaces instead?
<Borromini>
stintel: thanks, sounds cool.
<robimarko_>
aiyion: Its not possible
<robimarko_>
As you can only have one trigger per LED
<aiyion>
mhm. pre dsa the underlying switch was the trigger.
<aiyion>
Is there any way to access that?
<Znevna>
what is the source of the pain?
<aiyion>
We've got a Router, that has Lan-Ports,
<aiyion>
but not one LED per interface, but one for the whole switch.
<robimarko_>
aiyion: swconfig was providing that trigger
<aiyion>
ist there something similar, that can provide the same trigger again?
<robimarko_>
If there is a single LED for all traffic, just trigger of the DSA master
<robimarko_>
Aka eth0 probably
<damex>
mrnuke: did you manage to (re)check your sg2210p?
<aiyion>
for non-DSA the trigger was set to switch0 with a port mask iirc.
<aiyion>
Is that possible in dsa?
<robimarko_>
No
<aiyion>
ugh -.-'
<Znevna>
but what's wrong with using br-lan?
<robimarko_>
You are gonna get WLAN traffic as well
<aiyion>
and you might not even have that bridge depending on your setup.
<damex>
robimarko_: did you ever try to connect antennas on hap ac2 to ms156 port? i have seen reports that connecting disables onboard antennas and let you use external ones
<robimarko_>
damex: No, I just used the factory ones
<damex>
oh okay :(
<damex>
10x of ms156 -> rp-sma hopefully coming over next couple of weeks and i will be able to test ;)
<oliv3r[m]>
hauke: first mini-win (for me at least) that it _seems_ (running some additional tests) that I can keep NO_EXCEPT_FILL enabled, if I remove KERNEL_ENTRY from the realtek Makefile
<mrnuke>
damex: I did not yet. I have to pick it up from another location
<damex>
Znevna: there is reports online for mikrotik to use ms156 across their whole lineup. older routers/aps have ms156 inside and connectors look the same. also seen people mentioning hap ac2 having ms156 :)
<olmari>
Znevna: I'd suspect literally no hardware capable of doing so
robimarko has quit [Quit: Leaving]
csharper2005 has left #openwrt-devel [#openwrt-devel]
Borromini has quit [Quit: leaving]
Ryncewynd has quit [Read error: Connection reset by peer]
<dangole>
Znevna: it's the other way around: all devices with switch ICs (or SoC-integrated switches like MT7621) are capable of bridge offloading. it's those two devices which had LEDs defined in such a way that they would only indicate user-port<->cpu-port traffic and *not* indicate user-port<->user-port traffic. hence this is changed to have those LEDs controlled by the switch itself rather than having Linux drive them using GPIOs
<dangole>
Znevna: sadly this also removes to option to completely disable those LEDs from the user...
<Mangix>
Ansuel: I updated my macOS CI PR. Removed more tools now
<Mangix>
I'm curious what git-extras is used for
<Ansuel>
Mangix I'm having fun with creating an alpine image and i got a big idea
<Mangix>
do tell
<Ansuel>
considering everything got compiled with prebuilt tools and all and we are using external toolchain
<Mangix>
last I remember tools don't build with alpine
<Mangix>
elfutils specifically
<Ansuel>
in theory from the host we should use only extraction tool and some others.... I really want to create a dev env that will 100% fail as soon as we are ""leaking"" host tools
<Mangix>
uhhh
<Mangix>
are you thinking of something like nix-env --pure ?
<Ansuel>
yes something similar
<Ansuel>
btw i'm curious why tools doesn't compile on alpine
<Mangix>
Ansuel: elfutils relies on gettext, which alpine does not include by default. rather, elfutils assumes libc libintl.h
<Ansuel>
(the fact that we had gzip in prereq missing from ages and every makefile using the host tool makes me wonder if we have other similar case
<Mangix>
anyway if you want to test host tool leaks, nix might be a better way to test than alpine
<Ansuel>
i remember one crazy user compiling openwrt on a nix env
<f00b4r0>
Ansuel: care to elaborate what's the point?
<f00b4r0>
(i'm not seeing the obvious benefits, is why i'm asking)
<Habbie>
i've also noticed builds sometimes picking up host headers, which is bad
<Mangix>
Build dependency: Please install 'gzip'
<Mangix>
alright
<Ansuel>
point is that the idea is to create a slimmer container image for CI and as a side effect having something that will quickly catch package building with host tools/header
<f00b4r0>
Ansuel: not sure how it's going to get slimmer if it essentially has to be a full LFS? Or am i misunderstanding something? :)
<f00b4r0>
Habbie: yeah host headers leakage *is* a problem.
<Mangix>
Off the top of my head, I think libzstd is one thing that leaks from the host
<Mangix>
intentionally
<Ansuel>
using an alpine image with only the required tool instead of a big ubuntu image would reduce container size
<f00b4r0>
ah ok, yeah from that perspective it makes sense :)
<Mangix>
hmmm author of libdeflate got rid of the makefiles
<aiyion>
quick question about adding support for MT7981B: all I find in upstream linux for this platform is a line for a thermal driver added. Is Mediatek going to bring their patches upstream themselves?
<Mangix>
Ansuel: any insight into the cmake error?
<Ansuel>
any idea what to check?
<Mangix>
no i mean do you have logs?
<Mangix>
ERROR: tools/elfutils failed to build. <-- yep, still fails
<Ansuel>
oh god libdeflate needs cmake... and cmake needs libdeflate
<Ansuel>
LOL
<f00b4r0>
*sigh*
<Mangix>
unless tools/cmake gets switched to using git
<Mangix>
same with ninja
<Ansuel>
can't we just enforce a custom command for them?
<f00b4r0>
lovely tools that can't bootstrap themselves...
<Ansuel>
should work
<Mangix>
Ansuel: you mean falling back to gzip for them specifically?
<Ansuel>
yes
<Ansuel>
in theory we should be able to declare the variable in the makefile
<Mangix>
btw, is GNU gzip needed?
<aiyion>
I'll come back when I've got more crayons -.-' These mediatek repos do not look like fun to me, yet :/ gn
<Ansuel>
and that should overwrite the default value
<Ansuel>
Mangix you mean since we will use libdeflate-gzip...
<Mangix>
that's PKG_CAT I think
<Mangix>
Ansuel: no
<Mangix>
alpine linux has busybox gzip, which prereq-build.mk doesn't like
<Mangix>
it wants gnu gzip
<Mangix>
because busybox gzip doesn't implement --version, of course
<Ansuel>
think busybox gzip have some command missing
<Ansuel>
btw where the hell is a log for cmake to understand the error o.O
<f00b4r0>
Mangix: i'd guess if it were only that it'd be handled in prereq-build.mk, no? Seems simple enough otherwise.
<Ansuel>
/home/openwrt/openwrt/build_dir/host/cmake-3.25.1/Source/kwsys/kwsysPlatformTestsCXX.cxx:53:13: error: 'environ' was not declared in this scope; did you mean 'union'?
<Ansuel>
/home/openwrt/openwrt/build_dir/host/cmake-3.25.1/Source/kwsys/kwsysPlatformTestsCXX.cxx: In function 'int main()':
<Ansuel>
53 | char* e = environ[0];
<Ansuel>
| ^~~~~~~
<Ansuel>
| union
<Ansuel>
Test failed to compile
<Ansuel>
ok wtf LOL?
<Ansuel>
probably the problem is with
<Ansuel>
g++ has setenv
<Ansuel>
g++ has unsetenv
<Ansuel>
g++ does not have environ in stdlib.h
<Ansuel>
g++ has stl wstring
<Ansuel>
g++ has <ext/stdio_filebuf.h>
floof58 is now known as Guest755
floof58 has joined #openwrt-devel
* Mangix
looks
<dwfreed>
Ansuel: that's "normal"
<Mangix>
f00b4r0: most likely the gzip check was modeled after the others
<Ansuel>
dwfreed cmake test fails exactly with g++ and environ
<dwfreed>
that test is supposed to fail
<Mangix>
Ansuel: environ needs _GNU_SOURCE maybe?
<dwfreed>
no
<hauke>
aiyion: which device with MT7981B do you have?
<dwfreed>
that test is designed to detect whether stdlib.h defines environ; g++ does not, so it *correctly* fails; it does not fail the build
Guest755 has quit [Ping timeout: 480 seconds]
<Ansuel>
dwfreed any idea how to bisect it?
<Ansuel>
also mangix with my dockerfile you didn't repro the error?
<dwfreed>
that's not the failure
<Mangix>
Ansuel: that's not an error
<dwfreed>
the failure is elsewhere
<Mangix>
Ansuel: I don't use docker. Anyway, what dockerfile?