<rsalvaterra>
This gives me 2048 connections on a 128 MiB system, which is comfortable for my usage.
<aparcar[m]>
rsalvaterra: I think everything sub 128MiB should be set to 16k connections, once bigger it should be based on memory size
<aparcar[m]>
I don't want to shrink the current conntrack for small devices
<rsalvaterra>
In that case, the hack on top of the backport must be updated.
<rsalvaterra>
Instead of 1024, we make the lowest possible number 16384.
<aparcar[m]>
rsalvaterra: that's just my opinion, maybe jow want's to jump in since he added the sysctl stuff some 10 years ago
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
user_ has quit [Quit: leaving]
<stintel>
2048 sounds a bit low :)
<stintel>
peak here last 12h is 1.69k, imagine having a family with kids :)
<xback>
i've got ath79 fully booting on 5.15 :-)
<ynezz>
xback: ship it!
<xback>
ynezz: got to add an approval stamp first
<xback>
i'll borrow yours if that's ok :-)
<ynezz>
xback: if you find some time, there are 22.03 images ready for testing
<xback>
sure, I can test some
<aparcar[m]>
rsalvaterra: please adjust the lower limit and ship it
<f00b4r0>
ynezz: hi, did my last buildbot comment make sense yesterday?
<ynezz>
xback: Re: ath79: if you can give me link to initramfs image for tplink-c7v5 I can run test it
<f00b4r0>
as for 2048 connections on 128MB system, internet cafes hotspots may find that a little low :)
<ynezz>
internet cafes don't run stock firmware images anyway
<f00b4r0>
I beg to differ
<ynezz>
f00b4r0: your analysis makes sense to me, thanks, but fixing it properly is another matter :P
<f00b4r0>
heh :)
<f00b4r0>
eventually I'll take a shot at it. Just not in the immediate future.
<f00b4r0>
at least you know what's going on, now
srslypascal is now known as Guest543
srslypascal has joined #openwrt-devel
torv has quit [Remote host closed the connection]
torv has joined #openwrt-devel
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #openwrt-devel
Guest543 has quit [Ping timeout: 480 seconds]
<rsalvaterra>
aparcar[m]: Shouldn't we ask the author to respin with the changes instead?
<xback>
pushed ath79 5.15 stuff to my staging. if ayone would like to test it ..
<rsalvaterra>
xback: Will definitely do, but I'm still adapting my patches to 5.15. :)
<rsalvaterra>
Also, CONFIG_SCHED_CORE should be default disabled in the generic 5.15 kconfig.
<xback>
fyi, I run-tested it on Mikrotik rb922
<f00b4r0>
robimarko: ipq60xx devices should be avoided and we have no interest in them, correct?
<f00b4r0>
xback: out of curiosity, have you netbooted or sysupgraded?
<xback>
both
<robimarko>
f00b4r0: Not in general
<f00b4r0>
xback: nice to hear it still works. Seems some devices are no longer able to netboot even 5.10 kernels
<robimarko>
The 256MB RAM ones should be avoided, thats ewaste
<xback>
netbooted as in initramfs? (to be sure)
<f00b4r0>
yes
<xback>
f00b4r0: it works, but I made a minimal for this to keep the size small. otherwise it indeed does not boot
<f00b4r0>
ah
<f00b4r0>
so it seems we're reaching a point where first install will become problematic :(
<xback>
well, simply building default openwrt works for netboot. but don't add to any packages
<f00b4r0>
robimarko: thanks, just wanted to confirm. Someone asked for help porting OpenWRT to ipq6010, I think I'll pass :)
<xback>
so I have 2 files over here. one default for initramfs, and a 2nd build with all required packages inside for sysupgrade afterwards
<robimarko>
f00b4r0: Its doable, parts of IPQ807x can be reused cause IPQ60xx is a cut down kind of upgraded version
<robimarko>
I have some boards with IPQ6010 and IPQ6018
<robimarko>
But no time to work on that as well
<robimarko>
Dont fall for the port OpenWrt to XYZ
srslypascal is now known as Guest544
<f00b4r0>
xback: I'm surprised, I'd have expected the kernel size to be the issue. Or are we using lzma-loader already and I failed to pay attention?
<f00b4r0>
robimarko: hehe, amen to that :)
<robimarko>
f00b4r0
<robimarko>
f00b4r0: When I get those can OpenWrt be ported to XYZ, they always think of just using whater kernel works with OpenWrt
<robimarko>
Cause, the stock FW is already "OpenWrt"
Guest544 has quit [Ping timeout: 480 seconds]
<f00b4r0>
yeah, it looks exactly like that
<f00b4r0>
from the looks of it ipq6010 isn't even on its way upstream, so I assume the device runs some patched-to-death super old kernel
eduardo010174 has joined #openwrt-devel
<robimarko>
There is basic support upstream for it
<robimarko>
It was in better shape than IPQ807x actually
<f00b4r0>
interesting
<robimarko>
As they added regulator and the CPU clock drivers
<robimarko>
But spoiler alert, that driver doesnt work
<f00b4r0>
nice
<robimarko>
So, the usuall QCA peripherpals will work
<robimarko>
Cause its the same in most of the SoC-s
srslypascal has joined #openwrt-devel
hanetzer4 has joined #openwrt-devel
hanetzer3 has quit [Ping timeout: 480 seconds]
gleb has quit [Quit: Connection closed for inactivity]
eduardo010174 has quit [Remote host closed the connection]
<russell-->
xback: fwiw, ath79/wzr600dhp with CONFIG_TESTING_KERNEL=y with your staging tree, xtables-addons build fails
hexagonwin[m] has quit [Server closed connection]
hexagonwin[m] has joined #openwrt-devel
gch981213 has joined #openwrt-devel
<ynezz>
so when updating kernel config 5.10->5.15, I should first do `make kernel_oldconfig CONFIG_TARGET=target` and then `make kernel_oldconfig CONFIG_TARGET=subtarget` ?
<ynezz>
or vice versa? :p
JuniorJPDJ has quit [Server closed connection]
JuniorJPDJ has joined #openwrt-devel
<xback>
russell--: thanks. Still doing testing here with all modules etc
<russell-->
i'm doing a make dirclean world for good measure
mattytap has quit [Read error: Connection reset by peer]
mattytap has joined #openwrt-devel
<ynezz>
I've x86 5.15 bump in my staging tree, x86/64 works fine in QEMU
cbeznea has joined #openwrt-devel
robimarko_ has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
robimarko has quit [Ping timeout: 480 seconds]
csharper2005_ has quit [Remote host closed the connection]
Misanthropos has joined #openwrt-devel
<Grommish>
rsalvaterra: Any tips on setting up the config-5.15 stub? I get errors on scripts/kconfig.pl when I do it ;/
dlg has quit [Read error: Connection reset by peer]
dlg has joined #openwrt-devel
<xback>
russell--: I pulled a fresh clone from my staging, selected wzr-600dhp and testing kernel + selected "build all modules" It builds fine here atm. can you provide me a config or so?
<xback>
or steps to reproduce exactly
gch981213 has quit [Quit: Lost terminal]
ptudor_ has joined #openwrt-devel
ptudor has quit [Ping timeout: 480 seconds]
ekathva has quit [Remote host closed the connection]
ekathva has joined #openwrt-devel
<aparcar[m]>
anyone ideas for alternatives for grub on x86?
<stintel>
for OpenWrt? or laptop/workstation?
<mangix>
Systemd-boot?
<csharper2005>
Is it still 5.10 by default on ramips?
<f00b4r0>
mangix: oh dear, I thought this was a joke only to find it does exist ;P
Piraty_ has joined #openwrt-devel
Grommish_ has joined #openwrt-devel
aiyion_ has joined #openwrt-devel
aiyion has quit [Ping timeout: 480 seconds]
Piraty has quit [Ping timeout: 480 seconds]
gch981213 has joined #openwrt-devel
Grommish has quit [Ping timeout: 480 seconds]
aiyion_ is now known as aiyion
csharper2005 has quit [Read error: Connection reset by peer]
csharper2005 has joined #openwrt-devel
csharper2005 has quit [Remote host closed the connection]
csharper2005 has joined #openwrt-devel
T-Bone has joined #openwrt-devel
f00b4r0 has quit [Ping timeout: 480 seconds]
ecloud has quit [Ping timeout: 480 seconds]
ecloud has joined #openwrt-devel
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<neoraider>
systemd-boot is a fairly minimal EFI loader (I think the code is even based on a project with the name "efi-loader"). Working nicely on my notebook, but not quite appropriate for OpenWrt :)
<aparcar[m]>
neoraider: why is it not appropriate? to big?
csharper2005 has quit [Ping timeout: 480 seconds]
<neoraider>
Nah, it's tiny. But I assume that like udev, it's difficult to build without building the whole of systemd, or at least providing all its build dependencies. I might be wrong though, I have never built it myself
<neoraider>
Another thing is that it's EFI-only
<neoraider>
And as OpenWrt also supports a lot of legacy x86 hardware...
<neoraider>
Of couse, with EFI you don't strictly need another bootloader, you can just enable the EFI_STUB in your Kconfig and boot the kernel image directly from the ESP :D
ekathva has quit [Remote host closed the connection]
danitool has joined #openwrt-devel
hanetzer4 has quit [Quit: WeeChat 3.4.1]
minimal has joined #openwrt-devel
csharper2005 has joined #openwrt-devel
<T-Bone>
aparcar[m]: what's wrong with grub?
T-Bone is now known as f00b4r0
hanetzer has joined #openwrt-devel
hanetzer1 has joined #openwrt-devel
guidosarducci_ has joined #openwrt-devel
hanetzer has quit [Ping timeout: 480 seconds]
Borromini has joined #openwrt-devel
guidosarducci has quit [Ping timeout: 480 seconds]
hanetzer1 has quit [Ping timeout: 480 seconds]
hanetzer1 has joined #openwrt-devel
hanetzer1 is now known as hanetzer
<hanetzer>
stintel: hey, question. how do the ports lan[0-3] map to the physical labels, 1:1?
<hanetzer>
uh. on the tp-link-eap615-wall :)
csharper2005 has quit [Ping timeout: 480 seconds]
<Borromini>
hanetzer: same as on the EAP235, I believe they follow the outer numbering
<Borromini>
lan0 is the PoE-PD port, lan3 the PoE passthrough one
<hanetzer>
hrm. having issue setting stuff up. I want lan0/PoE-PD to be a trunk port
<Borromini>
i haven't dug into DSA and vlan tagging yet, unfortunately.
hanetzer has quit [Quit: WeeChat 3.4.1]
hanetzer has joined #openwrt-devel
srslypascal is now known as Guest577
srslypascal has joined #openwrt-devel
Guest577 has quit [Ping timeout: 480 seconds]
svanheule has joined #openwrt-devel
rua has quit [Remote host closed the connection]
rua has joined #openwrt-devel
Znullptr has joined #openwrt-devel
<rsalvaterra>
Oh, this will be fun…
<rsalvaterra>
… so, how should I prefix pending kernel patches for core mm functionality? ;)
<Borromini>
robimarko_: got a chance to test QCA8081 with 2,5G?
<robimarko_>
No, unfortunatelly not
<Borromini>
no worries, just curious.
<robimarko_>
The issue is getting the setup up
<robimarko_>
Cause, my switches that are always on dont have a 2.5G port
cbeznea has joined #openwrt-devel
<Borromini>
:)
<Slimey>
mine do ;)
<Slimey>
cat 9300's
csharper2005 has joined #openwrt-devel
cbeznea has quit [Quit: Leaving.]
csharper2005 has quit [Read error: Connection reset by peer]
csharper2005 has joined #openwrt-devel
<csharper2005>
I would like to upstream this patch to linux https://github.com/openwrt/openwrt/pull/1318#issuecomment-420607394. Should I keep the original license text from scpart.c untouched? Or it should be changed to SPDX-License-Identifier: GPL-2.0-or-later OR MIT?
<f00b4r0>
it should certainly not be "OR MIT" since the author did not include it in his license
Borromini has quit [Quit: Lost terminal]
<f00b4r0>
and since you're pushing someone else's work, personally I would leave it as is, at least for the initial submission. My 2c.
<csharper2005>
f00b4r0: Got it. I just changed the formatting and deleted 1 line... How to specify correctly the author of the patch - NOGUCHI Hiroshi? Should I place their signed-off or other reference?
<f00b4r0>
well ideally you would get them involved if possible. I wouldn't add their SoB (again, my take on this), but mention they're the author and CC them at minimum
<f00b4r0>
as for the spelling, yes, copy paste what's in MODULE_AUTHOR
<PaulFertser>
You can't add someone else's, please read the definition carefully.
<f00b4r0>
^
dedeckeh has quit [Remote host closed the connection]
<csharper2005>
f00b4r0: Thanks for the helpful advice!
<csharper2005>
PaulFertser: Looks like the letter b is my case
<PaulFertser>
csharper2005: indeed
csharper2005 has quit [Remote host closed the connection]
<rsalvaterra>
aparcar[m]: What's wrong with grub?
<mirko>
ynezz: ping
<mangix>
rsalvaterra: besides being big?
<rsalvaterra>
It's that an issue on x86?
<mangix>
I doubt it. I do prefer systemd-boot though
mangix has quit [Remote host closed the connection]
mangix has joined #openwrt-devel
<rsalvaterra>
I'm worried about the sheer amount of responsibility systemd is assuming…
<mangix>
I mean the main point of integrating systemd-boot is mostly for secureboot
robimarko_ has quit [Quit: Leaving]
<mirko>
what's the current way to include a cron job as part of a package? there's sooo many hack'ish ways within the base as well as the packages repo and none looks neat and clean..
<aparcar[m]>
rsalvaterra: we suspected it to be not reproducible
<rsalvaterra>
Ok, that's a different issue.
<f00b4r0>
aparcar[m]: I thought this was no longer an issue. IIRC both coreboot and debian worked toward ensuring that
<aparcar[m]>
f00b4r0: yea we found other issues, will check again
<f00b4r0>
if anything I'd suggest fixing the underlying problem instead of scraping the (trusted) bootloader entirely, if possible
<rsalvaterra>
Oh… it should improve the kernel's memory management.
<rsalvaterra>
Of course, I rebooted my router just 30 minutes ago, so no definitive conclusions for now.
<mangix>
seems neat. I don't see the point in backporting though
<\x>
rsalvaterra: 5.15 all good for ramips mt7621?
<rsalvaterra>
It's not even a backport. The final version should hit Linux 5.19. Hence the bravery. :P
<rsalvaterra>
\x: Generalizations are dangerous, but it seems to be the case. My n = 1, though. :P
<slh>
aparcar[m]: the only almost full featured alternative to grub would probably be syslinux (https://wiki.syslinux.org/), can do BIOS booting, can do UEFI and/ or multiple flavours of PXE, but the nice part about grub is it abstracting away all that stuff and providing a generic boot environment (almost) regardless of the underlying details <-- grub can also fill that role in an SBSA (Server Base System
<slh>
Architecture for ARMv8 servers), grub can be used on the RPi using a tianocore/ UEFI intermediary loader, etc.
<slh>
neoraider: systemd-boot is the old gummiboot, I've played with gummiboot a few years before it was merged into systemd-boot (mostly as a hack to provide a fallback efi loader in combination with grub (retaining upgrade support for distro packages) on a software RAID1), back then it worked quite nicely - but grub has the advantange of having seen much more real world testing
<Grommish_>
stintel: I'm running 5.10.108 on the Octeon, and with the network off (service network stop) I still see a very small leak. It was greater with networkibng on, but not like it was before (I pushed nearly 400Gb thru the WAN inside of an hour and it did rise, but not excessively).. Suggestions on what to try next or any idea when you might throw up a 5.15 config?
Grommish_ is now known as Grommish
<Grommish>
stintel: I've ended up reverting 3a14580411 locally so I still have a usable device :(
<Grommish>
stintel: I had posted about creating a config-5.15, but then realized only the includes were in place and not the linux/generic stuff yet, so I bailed on it. I'm happy to play locally if someone wants to give me a hint on how to do it, but I certainly can wait