<stintel>
someone at work asked me to document the code base structure, figured it's best to improve the public documentation and link to that
<karlp>
yeah, ok, htere's more stuff in tools than I thought there was :)
<stintel>
updated the gist a bit
<karlp>
we have "lzma" and "lzma-old" in tools?!
<karlp>
squashfs uses lzma-old?
* karlp
closes his eyes, other things to be doing.
<stintel>
=)
robimarko has joined #openwrt-devel
<stintel>
I could just edit the wiki and add what I drafted, but would like it if a few people could review it first :)
ptudor has quit [Ping timeout: 480 seconds]
ptudor has joined #openwrt-devel
<oliv3r[m]>
is it possible to get edit access to the wiki with my openwrt forum user account?
Acinonyx has quit [Ping timeout: 480 seconds]
<ukleinek>
oliv3r[m]: that's different accounts AFAIK, PaulFertser is the one to create you a wiki account
<oliv3r[m]>
ah, no sso. ironically, I used to have a few openwrt wiki and what not accounts 5 years ago :p
<oliv3r[m]>
PaulFertser: can I get a wiki account please? :)
<oliv3r[m]>
funny is that there's a button to use github auth, but then that also doesn't work :(
gromero has joined #openwrt-devel
<ukleinek>
oliv3r[m]: I think you need to /msg PaulFertser your desired username and your email address
<oliv3r[m]>
done
<oliv3r[m]>
the wiki states that alpine can be used to build openwrt; however 7z support is broken due to sched.h support being borked. Who was the last to use openwrt on alipine?
<robimarko>
I dont see how anybody can know that
<oliv3r[m]>
well someone might be in this channel that is using openwrt on alpine :)
<robimarko>
I know that Ansuel was trying to use Alpine for CI builds
<robimarko>
No idea how far he got
<stintel>
at some point I build OpenWrt on gentoo hardened musl
<stintel>
built*
<stintel>
some things worked. but iirc targets that built their own u-boot didn't
xes has quit [Ping timeout: 480 seconds]
<oliv3r[m]>
well you can't build 7z 'off the shelf' right now, alpine has a few patches for it too; so I doubt it still works; maybe on targets that don't need 7z of course it might
<oliv3r[m]>
Ansuel: how far DID you get? :)
torv__ has joined #openwrt-devel
torv_ has quit [Remote host closed the connection]
<oliv3r[m]>
i'm re-building a debian container from scratch to see if I can get rid of this faked bs; as for some reason, that also causes (i have no idea how or why) to cause kernel panic attempt to kill init. I did a dockerized build on my server, and it seems to work there (gotta try the image still though)
<oliv3r[m]>
(i realize i'm shooting at straws here :;)
<robimarko>
Are you still building as root?
<oliv3r[m]>
never did :) i thought I did, but the openwrt-sdk container wasn't; my new alpine/debian contianers don't either
<oliv3r[m]>
and I really want to keep building in docker; as I have no desire to polute my host :)
<robimarko>
Well, it seems that building in Alpine container works
<robimarko>
Just built ipq807x in it
<oliv3r[m]>
then you don't need/use 7z probably
<oliv3r[m]>
the realtek target needs 7z :(
<robimarko>
Indeed
<oliv3r[m]>
but even with a clean debian container, with a cleaned git repo (i'm sure something dirty that doesn't show in git status) still causes 'faked' to run forever
<oliv3r[m]>
and (while not finished yet) breaks stuff
<Piraty>
is anyone aware if user facing parts of openwrt have been adopted by other distros/projects? i think of uci/luci , maybe even libuci
cbeznea has quit [Quit: Leaving.]
cbeznea has joined #openwrt-devel
<dhewg>
oliv3r[m]: it just has to be the setup on your end. I build a few targets in a debootstrapped ubuntu lxc container last week
winternull has joined #openwrt-devel
nixuser has quit [Ping timeout: 480 seconds]
<nbd>
rmilecki: ping
<nbd>
rmilecki: i managed to come up with a simple hack for making the network stack more threaded: https://nbd.name/p/8fc302e1
<nbd>
please apply that on top of the other patches that i made
<rmilecki>
nbd: will do during next days
<nbd>
and run it with threaded napi enabled and with the default netifd rps/xps script
<nbd>
and you also need to do this:
<nbd>
echo 1 > /proc/sys/net/core/rps_threaded
<rmilecki>
ok
<nbd>
this changes the code to create one core napi thread per cpu
<nbd>
and instead of steering packets to specific cpus using rps, it steers them to specific threads
<nbd>
allowing the scheduler to bounce things around based on cpu load
<nbd>
the patch turned out much simpler than i thought it would
<oliv3r[m]>
<dhewg> "oliv3r: it just has to be the..." <- i figure; but what? I've nuked all directories that are basically in .gitignore; so nothing should be left anywhere?
<oliv3r[m]>
the only remaining step, is to re-clone and see if that makes a difference
nixuser has joined #openwrt-devel
minimal has joined #openwrt-devel
<oliv3r[m]>
so, i've patched 7z and building a whole new setup; lets see if anything else breaks :)
<dhewg>
no clue, I don't use docker
sauce has quit [Quit: sauce]
sauce has joined #openwrt-devel
valku has joined #openwrt-devel
<paper_>
oliv3r[m]: I use a docker container with debian 11 because I got tired ofconstantly fixing musl and other issues before I could build anything, I can send you a link to the dockerfile I use for building Turris OS if you want to
<mirko>
i know there's differences in the kernel versions (e.g. some ipv4/ipv6 got merged into the same modules - e.g. for nft-nat-chains) - however i tried to map everything back in include/netfilter.mk. would be just happy about any pointer.
romany0 has quit [Ping timeout: 480 seconds]
hanetzer3 has joined #openwrt-devel
hanetzer2 has quit [Ping timeout: 480 seconds]
fakuivan has quit [Remote host closed the connection]
fakuivan has joined #openwrt-devel
<oliv3r[m]>
hah! building with alpine fixed all that shit :)
<oliv3r[m]>
paper_: i was using the opewrtorg/sdk container for months; but something broke over the last few days. I now created an alpine container, and after fixing 7z it actually works; and my builds actually boot again!
hanetzer4 has joined #openwrt-devel
<nbd>
rmilecki: fyi, in my tests non-offload throughput on mt7621 is higher and more stable with rps_threaded
hanetzer3 has quit [Ping timeout: 480 seconds]
<rmilecki>
nbd: let's hope it improves bcm53xx as well however that hw turned out to be tricky way too many times
<oliv3r[m]>
something is still weird; with alpine as base, I now have about 20 'faked' things running; which from a throughput point is still far better then the single one :) but I'd have to admit, I never noticed or checked before :p
<Znevna>
o0
<oliv3r[m]>
actually it was close to 50 faked processes; that kept lingering after the make command was finished even, so, a clean worktree to nuke all the crap
<oliv3r[m]>
whadayaknow, clean worktree nuked those faked; something that even rm -rf and make distclean didn't manage. so 'something somewhere' was left being stale
lmahmutov has joined #openwrt-devel
lmahmutov has quit [Remote host closed the connection]
Slimey has quit [Remote host closed the connection]
cbeznea has quit [Quit: Leaving.]
cbeznea has joined #openwrt-devel
<oliv3r[m]>
hmm, the fakes are still there actually; but at least multiple of them instead of a single one taking forever :S
<oliv3r[m]>
what's more funny, they keep on doing their thing, long after the build has finished (or failed) ... killing them with killall doesn't seem to do anything bad (they are gone afterwards thougH)
<oliv3r[m]>
btw, I thought 'OF: Bad cell count for /soc/spi@1200/flash@0/partitions' was a false positive and fixed already on master?
Slimey has joined #openwrt-devel
gromero has quit [Remote host closed the connection]
gromero has joined #openwrt-devel
tmn505 has quit [Remote host closed the connection]
tmn505 has joined #openwrt-devel
<robimarko>
It is false positive but its not fixed
fakuivan has quit [Quit: No Ping reply in 180 seconds.]
Borromini has joined #openwrt-devel
fakuivan has joined #openwrt-devel
Acinonyx has joined #openwrt-devel
Borromini has quit [Quit: Lost terminal]
<oliv3r[m]>
ah ok; fair nuff
<robimarko>
Issue is that upstream isnt really sure how to fix it, at least that was the last discussion I found
cbeznea has quit [Quit: Leaving.]
<mrkiko>
nbd: ping
<mrkiko>
robimarko: hi!!
<mrkiko>
robimarko: I looked at the hwmon situation a little bit more. The other drivers use a macro called CLAMP that allows you to restrict the value n certain range
<mrkiko>
example of that behaviour are marvell.c and bcm54140.c
minimal has quit [Quit: Leaving]
<mrkiko>
robimarko: still, I'm little bit dubious on how I can proceed - I see they use sometimes macros to have the value rounded to one divisible for e.g.: 5 or something like that. But I don't know what the requirements for the chip are here
madwoota has quit [Ping timeout: 480 seconds]
<mrkiko>
robimarko: sorry, not a macro - actually clamp_val is a function in include/linux/minmax.h
<mrkiko>
yess - it is but written in lowercase :d :D
floof58 has quit [Remote host closed the connection]
floof58 has joined #openwrt-devel
<robimarko>
mrkiko: My point was that the issue is that you are getting impossible values from the hwmon core
<mrkiko>
robimarko: I agree, but I am in the impression it's not something "the core" threats as a bug
<robimarko>
mrkiko: I mean, providing you with the max of long int is kind of crazy
<robimarko>
I have a feeling that its not initialized
<mrkiko>
robimarko: I will look at that, but in the meantime I should understand why after definiting this thermal zone I deleted all the others :D
<mrkiko>
robimarko: do you think it can help f I movemy aqr_thermal outside of the tree, nearer thecpus and cluster ones?