<FLD>
how come im getting "configure: error: fwknopd needs libnetfilter_queue ERROR: package/feeds/packages/fwknop failed to build." even after doing "apt install libnetfilter-queue-dev" ?
<FLD>
"Some packages may not be updated properly and built after they got stuck with old dependencies, resulting in warnings at the beginning of the compilation looking similar to: WARNING: Makefile 'package/feeds/packages/openssh/Makefile' has a dependency on 'libfido2', which does not exist". haha, you don't say: https://paste.debian.net/1223803/
<FLD>
there is no way automatically clearing these, instead i gotta manually uninstall/install each one?
<Namidairo>
i think you may have skipped a step regarding 'scripts/feeds'
<FLD>
Namidairo: ah, might be that i did that stuff before switching to the latest git branch
Grommish has joined #openwrt-devel
_lore_ has quit [Ping timeout: 480 seconds]
victhor has quit [Remote host closed the connection]
swalker has quit [Remote host closed the connection]
_lore_ has joined #openwrt-devel
fda- has joined #openwrt-devel
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
fda has quit [Ping timeout: 480 seconds]
swalker has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
Luke-Jr has quit []
Luke-Jr has joined #openwrt-devel
rmilecki has joined #openwrt-devel
rua has quit [Quit: Leaving.]
nitroshift has joined #openwrt-devel
valku has quit [Quit: valku]
Tapper has joined #openwrt-devel
srslypascal is now known as Guest8684
srslypascal has joined #openwrt-devel
gladiac is now known as Guest8685
gladiac has joined #openwrt-devel
Guest8684 has quit [Ping timeout: 480 seconds]
danitool has joined #openwrt-devel
danitool has quit []
srslypascal is now known as Guest8686
srslypascal has joined #openwrt-devel
rua has joined #openwrt-devel
Guest8685 has quit [Ping timeout: 480 seconds]
Guest8686 has quit [Ping timeout: 480 seconds]
xes has quit [Quit: bye..]
xes has joined #openwrt-devel
danitool has joined #openwrt-devel
<mangix>
i wrecked github actions. cool.
<rsalvaterra>
nbd: Am I safe to assume qosify is a superset of sqm-scripts (as far as the cake qdisc goes)?
<nbd>
yes
<rsalvaterra>
Thought so, thanks. :)
<rsalvaterra>
Thinking of playing with it, but I'm sure it will make my 8 MiB flash routers most unhappy.
<Namidairo>
you probably weren't putting those behind the largest connections to begin with
<dhewg>
it shouldn't take that much space though, looks like ~33kb for sched-bpf + qosify ipks here
<dhewg>
libbpf is already static, the other dependencies should already be there for sqm-scripts
<rsalvaterra>
dhewg: Ok, that's good to know!
<rsalvaterra>
I'm having trouble with the LLVM toolchain, though…
<rsalvaterra>
bpf.mk:70: *** ERROR: LLVM/clang version too old. Minimum required: 12, found: . Stop.
<rsalvaterra>
I mean, how too old? It should be downloading and compiling it. :P
<stintel>
is it not trying your host llvm?
<stintel>
CONFIG_USE_LLVM_HOST=y ?
<rsalvaterra>
stintel: Actually, I don't have that symbol in my .config. Let me see…
victhor has joined #openwrt-devel
<rsalvaterra>
I enabled CONFIG_BPF_TOOLCHAIN_BUILD_LLVM, though.
<rsalvaterra>
On the host side, I have libllvm13. Do I need anything else? Clang?
<rsalvaterra>
Right, looks like Clang needs to be installed too.
<rsalvaterra>
Dang, anoter 188 MiB…!
<rsalvaterra>
*another
<rsalvaterra>
Hmm… that's not it. It's complaining about clang missing from the staging_dir. But why? It should be building it, no?
<stintel>
yeah if you have CONFIG_BPF_TOOLCHAIN_BUILD_LLVM I'd expect the build system to build it
<stintel>
if you install it on the host you can use CONFIG_USE_LLVM_HOST=y and avoid long build times :P
<rsalvaterra>
From what I could gather, it fails when at make download, but just running make (-j 9, in my case) seems to build fine.
<nbd>
the easiest way is to download the llvm-bpf from snapshot builds and unpack it into your source directory
<nbd>
no need to install llvm on the system
f00b4r0 has joined #openwrt-devel
f00b4r0 has quit [Read error: Connection reset by peer]
f00b4r0 has joined #openwrt-devel
<rsalvaterra>
nbd: It didn't take that long to build, actually. And yeah, I uninstalled Clang from the host, since I don't need it (I only need libllvm because of Mesa).
fda has joined #openwrt-devel
<rsalvaterra>
That kmod-sched-core package, though…
<rsalvaterra>
I hate the fact it pulls in everything plus the kitchen sink.
<rsalvaterra>
Shouldn't we split it up in more fine-grained packages?
fda- has quit [Ping timeout: 480 seconds]
<nbd>
i guess we could do that
minimal has joined #openwrt-devel
<rsalvaterra>
I'll have a stab at it, probably this weekend.
<stintel>
just replace that 8MB flash hardware already ;)
<stintel>
just a warning: reworking kmod packages can be very frustrating
* rsalvaterra
groans and holds his WDR3600 tight
<stintel>
last time I looked at v4l related stuff
<stintel>
it's in need of a big cleanup
<rsalvaterra>
I wouldn't expect v4l to be a huge priority in OpenWrt… :)
<rsalvaterra>
Of course, people could use a USB webcam for surveillance…
<rsalvaterra>
Image built, the size difference is surprisingly small. I'm sure it's mostly due to tc vs tc-full.
fda- has joined #openwrt-devel
fda has quit [Ping timeout: 480 seconds]
<nitroshift>
rsalvaterra, 0/
<stintel>
man, I briefly looked at my v4l branch while my corp macbook was updating to 12.1, I want to throw computer out the window already
<jow>
IT is hard, let's go shopping
<stintel>
but I already ordered 2 pair of shoes last time I had this feeling :P
* jow
was impressed that nm-applet can import an OpenVPN configuration and properly handle both password and 2FA code queries
<stintel>
o_O
<stintel>
maybe I need to return the shoes for the complete experience
<stintel>
to be fair, nm+addons improved a lot
<stintel>
I'm doing Open vSwitch with bond + vlan via nm on my workstation
<rsalvaterra>
nitroshift: o/
<rsalvaterra>
Uhh… wan vs wandev in qosify.conf. Which one? Or both? :P
<rsalvaterra>
The possibility to disable ingress shaping is a nice touch in qosify.
<wvdakker>
perhaps I have missed something. For the new 5.10 kernel are there uboot changes necessary? For 5.4 I have https://paste.debian.net/1223840/. But booting with the latest trunk give me an overwrite error during kernel load. Uboot shows at boottime kerneldatasize 2.4M
<stintel>
rsalvaterra: either. config interface vs config device!
<stintel>
rsalvaterra: the former takes interface name as defined in /etc/config/network, the latter a real network device name like eth0
<stintel>
or at least that's how I understand it
<rsalvaterra>
stintel: That's how I understood it too, but I could be wrong.
<stintel>
I have to use the latter, otherwise it would configure cake on the macvlan device and that is completely performance-crippled for some reason
<rsalvaterra>
And in my device, both the interface and the device are named "wan". :P
<stintel>
same
<rsalvaterra>
tc -s qdisc show dev wan gives me what I expected, so it should be fine.
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
nitroshift has quit [Quit: Gone that way --->]
guerby has quit [Remote host closed the connection]
<rsalvaterra>
Hmm… I'm not seeing qosify binning my BitTorrent traffic correctly…
<rsalvaterra>
… I have the peer TOS set to "throughtput", which sets a DSCP value of 8 (dec).
<rsalvaterra>
With sqm-scripts, it was binned as bulk. Now it's… weird, I don't even see the traffic in any bin.
<rsalvaterra>
Hm…
f00b4r0 has quit [Quit: reboot]
SamantazFox has joined #openwrt-devel
decke has joined #openwrt-devel
<neggles>
stintel: you've been doing shenanigans on mpc85xx - any clue how i can get around the ole "kernel image too big, can't decompress without running out of ram" without getting my hands on an OEM GPL tarball to build a new u-boot with?
<neggles>
i have tried yeeting the debug info and it didn't seem to accomplish much
<stintel>
not mpc85xx really
<neggles>
oh were you not doing the ap3710i? who've i gotten you mixed up with then...
<stintel>
I am working on WatchGuard Firebox M200/M300
<neggles>
ah
<neggles>
well i got the right architecture at least.
<stintel>
for that I started as subtarget of mpc85xx
<stintel>
but afterwards created a new target qoriq
f00b4r0 has joined #openwrt-devel
<neggles>
ah i see
<neggles>
ah, right, T1042 as opposed to p1020
<neggles>
nowhere near as similar as the model number might lead you to believe
<neggles>
wvdakker: suspect you've hit the same issue i have, u-boot decompresses the kernel image and runs out of RAM in doing so - it seems 5.10 is a (relatively) chonky kernel
<stintel>
I believe rmilecki had some issues with things not booting due to too big kernel
<rmilecki>
i did
<neggles>
is the solution going to be chaining through that other loader I've forgotten the name of
<neggles>
the frustrating bit is it loads just fine if I tftp over the fdt and initramfs image, but I don't know enough about the difference between initramfs-kernel and regular flash boot images/generation/booting to know where to even start
<neggles>
time for more staring at source files / makefiles until they make sense :P
goliath has joined #openwrt-devel
<neggles>
ah, i see. u-boot compile time #define dictates how big the uncompressed image can be, and it seems like it's 8MB in these.
<neggles>
welp. time to strip more symbols
<stintel>
rmilecki: can we put that ATF stuff for Broadcom in a separate repo instead of "polluting" openwrt.git with it?
<rmilecki>
stintel: i guess we can
<rmilecki>
hauke: ?
<hauke>
rmilecki: I would also prefer to have this in a seperate reposetory
<hauke>
is this curently extracetd from a device vendor gpl tar?
guidosarducci has quit [Remote host closed the connection]
guidosarducci has joined #openwrt-devel
victhor has quit [Remote host closed the connection]
<rmilecki>
hauke: yes, check my commit message
dangole has joined #openwrt-devel
valku has joined #openwrt-devel
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
aleksander has quit [Quit: Leaving]
<rsalvaterra>
Guys, is there any situation where we shouldn't use $(AUTORELEASE) for a package?
<rsalvaterra>
I see lots of packages which migrated to PKG_RELEASE:=$(AUTORELEASE), but others still require manual bumps.
decke has quit [Quit: Leaving.]
<dwfreed>
Have to always remember to make your commit message appropriately? and do the same for outside contributions?
<hauke>
rsalvaterra: I think aparcar[m] tried to convert all packages, but the build took significant longer
<dwfreed>
yeah, the repeated calls to git log would use a lot of cpu time, especially for packages repo
<dwfreed>
perhaps memoization would help, or having something produce the whole git log once, and then commitcount greps through that
<rsalvaterra>
hauke: I see. So, no criteria? :P
fda has joined #openwrt-devel
fda- has quit [Ping timeout: 480 seconds]
<wvdakker>
neggles: hmm makes sense. The lantiq/xrx200 does not have the BZ_TEXT defines. I did find a changed LZMA_TEXT_START which was empty. After reinserting the value from the previous version the image did boot. However dont know if the change makes sense ....
<wvdakker>
The change I made was in Makefile and +LZMA_TEXT_START := 0x80a00000
srslypascal is now known as Guest8708
srslypascal has joined #openwrt-devel
Guest8708 has quit [Read error: Connection reset by peer]
srslypascal is now known as Guest8709
srslypascal has joined #openwrt-devel
Guest8709 has quit [Ping timeout: 480 seconds]
<hauke>
rsalvaterra: no not really
<rsalvaterra>
stintel: I think I jumped the gun on the eBPF classifier. I think it's working fine, I just don't have a lot of traffic at the moment. :)
felix has quit [Read error: Connection reset by peer]
valku has quit [Quit: valku]
felix has joined #openwrt-devel
<mkresin>
wvdakker: likely the uncompressed kernel is to big. which board do you have the issue with? can you pastebin the uboot env?
<wvdakker>
mkresin: I have already made it working by changing the LZMA_TEXT_START in the Makefile from the loader. It was empty and changed it back to 0x80a00000.
<wvdakker>
Now it boots ok. But not sure if it is a valid patch.
<wvdakker>
The board: ZyXEL P-2812HNU-Fx
valku has joined #openwrt-devel
<mkresin>
wvdakker: which Fx is it? The F1 or F3?
<wvdakker>
F1
<mkresin>
wvdakker: imho the correct fix is an update of the nboot command
<mkresin>
wvdakker: as soon as the uncompressed kernel is > 8 MByte, you are hitting the issue
<wvdakker>
mkresin: ok and what should the nboot command be?
<mkresin>
wvdakker: the compressed kernel is loaded to mem addr 0x80800000 and uncompressed to 0x80000000
<mkresin>
wvdakker: as soon as the kernel is > 8M the memory holding the compressed data gets overwritten during uncompression, you see the issue...
<wvdakker>
mkresin: If I understand you correctly then the 0x80800000 needs to be adjusted to a higher level
<mkresin>
wvdakker: compressed kernel is read to memory location 16M
<wvdakker>
mkresin: I shall test it and update the wiki if it works
<dwfreed>
notably that will also fail if uncompressed kernel is 16 M
<dwfreed>
why not put the compressed kernel before the decompression point?
<wvdakker>
Just a silly question, can 0x8100000 be any value as long as there is enough memory?
<dwfreed>
the nand read appears to be reading at most 3 MB, so 7FD00000 would put the compressed kernel right before the uncompressed, and then decompression can use however much space it wants
<dwfreed>
(up to the limit of memory, of course)
rmilecki has quit [Ping timeout: 480 seconds]
<mkresin>
dwfreed: mem is mapped to 0x8000000, default entry point for this arch is 0x8002000, hence you load the compressed to stuff to high mem
<mkresin>
wvdakker: more or less it can be any mem address with in the amount of ram provided by your board. as long as the is enough free ram after to hold the compressed data
<mkresin>
wvdakker: yes, 16M limit. but it's the current bootm limit for the lantiq u-boots anyway
<wvdakker>
mkresin: clear.
victhor has joined #openwrt-devel
Borromini has joined #openwrt-devel
<wvdakker>
mkresin: the nboot adjustment works. I will add it into the wiki page of the F1
danitool has joined #openwrt-devel
<rsalvaterra>
nbd: I also used sqm-scripts as a practical means to limit the bandwidth on two other interfaces/devices. Is it possible to do the same with qosify?
<rsalvaterra>
(Assuming it supports multiple config interface/device sections, I guess it should.)
<nbd>
it supports as many interface/device sections as you like
<blocktrron1>
They've changed their addressing scheme, and with mac80211 auto addressing of VAPs, a second VAP leads to conflicting mac addresses between radios.
<hanetzer>
glad stintel pointed me at that interesting tp-link 802.11ax family of devices
<blocktrron1>
well, it's not ubiquitis fault per se
<blocktrron1>
i just needed someone to blame which is not me :P
<blocktrron1>
I'll work around this by making the last octet not conflict, add a comment, get hate from people because MAC addresses do not match the vendor firmware regardless of the circumstance and live happy into the future.