<stintel>
ahh, KCFLAGS - thanks Grommish for pointing me in the good direction
simsasss has left #openwrt-devel [#openwrt-devel]
<rsalvaterra>
Pepes: False alarm: turns out I'm just an idiot and didn't update my kconfig when CONFIG_PHY_MVEBU_A38X_COMPHY=y moved to the specific mvebu/cortexa9 kconfig, that's why I was missing eth2. :P
<f00b4r0>
any 802.11r expert around? I have a WPA2-PSK setup with 802.11r enabled. The wiki says that "nasid" is a required option however the Luci tooltip says it is "Not needed with normal WPA(2)-PSK." - what to believe?
<stintel>
I think it was required initially, but a later change removed that requirement
<f00b4r0>
ok. Trying to debug "hostapd: nl80211: kernel reports: key addition failed" which seems to be connected to 802.11r
<f00b4r0>
I have this very weird log where a STA auths to the AP normally, then several hours later it starts jerking around between both radios (2.4/5GHz) with hostapd apparently kicking it out with the above error message
<f00b4r0>
RSN is never retried either
dangole has joined #openwrt-devel
f00b4r0 has quit [Read error: Connection reset by peer]
f00b4r0 has joined #openwrt-devel
<stintel>
f00b4r0: interesting, NetworkManager on my XPS13 with ax200 is dropping from the network if I leave it open for a longer time, I was blaming this on usteer (or NetworkManager not handling this properly) but did not take the time yet to properly debug
<rsalvaterra>
stintel: You can always disable CONFIG_NETWORK… That disables IPv4. :P
c0sm1cSlug has joined #openwrt-devel
<karlp>
C4930002EBF4
<rsalvaterra>
robimarko: Not terrible. About 200 kiB in my images.
<karlp>
wrong window sorry
<stintel>
rsalvaterra: 😂
<robimarko>
That just makes my point I made via email that its not worth it
<robimarko>
People talk about fitting PHP, Squid etc because of space savings
<robimarko>
Time to drop it finally, would make life easier for packaging
<jow>
I'm having a rather pragmatic view here
<robimarko>
jow: I agree
<jow>
we simply have the manpower to deal with this proliferation of configuration choices
<jow>
*simply don't have
Tapper has quit [Ping timeout: 480 seconds]
<jow>
also at some point there has to be a certain baseline functionality software, packages, user etc. can rely on
<rsalvaterra>
Yeah. Even for custom builds.
<jow>
I'm fine to ship 22.x with a still functioning CONFIG_IPv6 but after the branch we should probably get rid of it for good
<rsalvaterra>
The IPV6 cleanup will already be a lot of work in itself. :)
<stintel>
I think dropping it before branching would be a bit late since we're in a freeze already
<rsalvaterra>
+1 for dropping after branching.
<jow>
right, it's not possible anyhow anymore
Tapper has joined #openwrt-devel
<stintel>
but I agree it can be dropped, I have yet to read a valid excuse for keeping it around
<jow>
I think stub IPv6 support should be a baseline requirement
<jow>
people can still omit IPv6 specific software (actually just odhcp6c and odphcd)
<stintel>
I have IPv6 connectivity for 20 years now, ISPs had enough time to implement
<jow>
but the rest is more or less intrinsically dual stack, like e.g. nft
<stintel>
does anyone know if starlink provides IPv6?
<jow>
and the disable_ipv6 sysctl knobs still work
<robimarko>
stintel: I think they do provide IPv6
<stintel>
ok that's it, I'm ordering a starlink today
<rsalvaterra>
They'd be crazy not to…
<stintel>
I was just reading a post from a year ago they're testing it and will roll out soon
<stintel>
I'll be paying 5 times what I pay now for my internet, but none of the ISPs here provide ISP
<stintel>
time to put my money where my mouth is and stop paying ISPs stuck in the stoneage
<robimarko>
I am not ready to give up gigabit fiber and pay 5-6 more
<robimarko>
Though I hate A1 for not having IPv6, I even have my ASN and BGP just to get proper IPv6 for some stuff
<stintel>
oh you have the same ISP?
<stintel>
interesting :)
<robimarko>
Well, they are all over this part of europe
<robimarko>
Its either them or DT/their subsidiary
<robimarko>
But they dont have fiber at my home and also dont do IPv6
<stintel>
ugh
<stintel>
> "2.2 Residential Use. Services and the Starlink Kit are for use exclusively at the address you provided in your Order, and only for personal, family, household or residential use."
<stintel>
why can't we have nice things :(
robimarko_ has joined #openwrt-devel
<stintel>
starlink premium appears to be more enterprise oriented but not available here and would be ~138x more expensive
<stintel>
lol
<stintel>
wait
robimarko has quit [Ping timeout: 480 seconds]
<stintel>
~14x
* stintel
needs more coffee
<stintel>
it's still wrong. normal starlink is 99$ -> 199BGN, so $500 would be ~1000BGN, compared to ~36BGN now, so ~28x more expensive
<stintel>
that's a bit too much
shibboleth has joined #openwrt-devel
<Pepes>
rsalvaterra: That happens, nothing to worry about!
kenny has joined #openwrt-devel
valku has joined #openwrt-devel
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
<rsalvaterra>
jow: Speaking of IPv6… has anyone ever tried to use the BusyBox-native DHCPv6 client instead of odhcp6c?
<rsalvaterra>
(I know the BusyBox client wasn't usable at the time odhcp6c was written.)
<jow>
rsalvaterra: I don't think so. One reason we went with the odhcp6c route is that we want to do RA processing in userspace
<jow>
so its not just DHCPv6 but also SLAAC which needs to be covered and I think busybox dhcpc only does the stateful portion
<jow>
(we don't want the kernel to randomly pull and establish routes beneath netifd managed interfaces which would lead to a lot of races and complicated address/route table synchronization logic)
<jow>
rsalvaterra: hm I see. Maybe this is the way. But in order to use it you'd need to do all the scripting around it, extract prefix info, hand it over to netifd etc.
<rsalvaterra>
Yes, I know, that's the hard part (to me, at least :)).
<aiyion_>
Thanks, just wanted you to know I'm on irc as well.
Misanthropos has joined #openwrt-devel
srslypascal has joined #openwrt-devel
<stintel>
ok none of HOST_CFLAGS TARGET_CFLAGS KCFLAGS seems to respect -Wno-error=array-bounds, anyone any other suggestions besides falling back to GCC 8 ?
<stintel>
+ for the kernel build
Misanthropos has quit [Ping timeout: 480 seconds]
<tmn505>
force it in rules.mk?
<stintel>
that's where I forced it in TARGET_CFLAGS
<stintel>
doesn't work
<tmn505>
I don't know any other way, someone with more knowledge would need to answer that
<ynezz>
KCFLAGS?
<stintel>
doesn't work
<ynezz>
well, where/how are you setting that?
<stintel>
in KERNEL_MAKE_FLAGS in include/kernel.mk
<ynezz>
then it's forced in the kernel and you need to patch it?
<ynezz>
if it's properly set and passed it should work
<ynezz>
I don't know how it behaves if you have conflicting gcc args on the commandline, probably the last one takes effect?
<stintel>
ahh, still didn't get enough coffee
<stintel>
I added it on the KCFLAGS line but outside of the ""
<stintel>
thanks for making me recheck things ;)
guidosarducci has joined #openwrt-devel
guidosarducci_ has quit [Remote host closed the connection]
tizbac has quit [Remote host closed the connection]
tizbac has joined #openwrt-devel
Borromini has joined #openwrt-devel
Misanthropos has joined #openwrt-devel
Tapper has joined #openwrt-devel
cmonroe_ has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
goliath has joined #openwrt-devel
Guest1274 is now known as foxtrot
<foxtrot>
Hey all, I'm currently working on updating the MTR package (https://github.com/openwrt/packages/pull/18047) and just realised that there is an update available for it upstream. I'm wondering if it would be best to update the version of the package and change PKG_RELEASE $(AUTORELEASE) in the same PR, or whether it would be preferred to do that in a seperate PR
<foxtrot>
PKG_RELEASE to $(AUTORELEASE) *
<Habbie>
foxtrot, just do it in one go
<foxtrot>
Cool, will do then.
<foxtrot>
thanks :)
<Habbie>
i'm speaking from experience submitting PRs, not reviewing them for the project, but i'm pretty confident about this one :)
valku has joined #openwrt-devel
valku has quit []
dangole has quit [Remote host closed the connection]
gladiac has quit [Quit: k thx bye]
Ansuel has joined #openwrt-devel
<Ansuel>
Hi, quick question... does failsafe load kmods? Is having the switch driver as a module problematic for failsafe?
<Ansuel>
I didn't check it but i'm doing some test with qca8k as a module and i notice failsafe doesn't work... had no time to check via serial if the problem was something else
<tmn505>
yes, failsafe doesn't load kmods.
<tmn505>
because rootfs is not mounted yet
<Ansuel>
mhh why rootfs is needed for kmods ?
<Borromini>
that's where they're stored AFAIK
<Ansuel>
not if they are part of the image from the start
<Borromini>
tmn505: wait i thought it was just /overlay not being mounted in failsafe?
<Ansuel>
the kmod are present if they are compailed with the image
<tmn505>
hmm, I might have missremembered something, let me check
<Ansuel>
we are not talking about kmods installed with opkg
<Borromini>
Ansuel: no serial for boot errors?
<tmn505>
ok, now I see, only the packages which are in DEFAULT_PACKAGES end up in squashfs
<Ansuel>
sorry didn't check but if kmods are not loaded
<Ansuel>
then no ethernet traffic
<tmn505>
if You put it in that array then they should be loaded
<Borromini>
tmn505: anything compiled in should end up in squashfs AFAIK.
<Borromini>
not just stuff in DEFAULT_PACKAGES
<Ansuel>
i honestly think the problem is that they are not loaded
<tmn505>
Borromini: not in the case of images produced by buildbot
<tmn505>
which are the images served on openwrt.org
<Borromini>
tmn505: well Ansuel is not using the buildbot, I'm pretty sure.
<Borromini>
anything you produce with a buildroot has packages in squashfs.
<tmn505>
but the average users will
<Borromini>
Ansuel is not your average user :)
<Ansuel>
correct
<Borromini>
AFAIK he's instrumental in QCA DSA support and whatnot
<tmn505>
well that is fine until You want those changes in openwrt git tree
<jow>
Ansuel: afair there's an extra flag in kmod Makefiles defines which marks them as required for booting
<Borromini>
jow: for some kind of basic initrd?
<tmn505>
thanks learned something new
<Ansuel>
jow can you link an example?
<jow>
Ansuel: check for example package/kernel/linux/modules/netdevices.mk, KernelPackage/phy-realtek
<jow>
that extra ",1" in the AUTOLOAD:= macro call
<jow>
it causes the module file to get placed in /etc/modules-boot.d/ instead of /etc/modules.d/
<jow>
modules-boot.d is processed by procd init before launching preinit
<jow>
(and before the prompt stating "press f enter to enter failsafe now"
<jow>
it'll obviously not work for opkg-installed kmod packages, but for builtin ones it should do what you want
<Ansuel>
that's exactly what i need thx jow
<Ansuel>
i needed to make the ar8xxx swconfig driver working as a module
<Borromini>
so you can swap both in and out?
<Ansuel>
yep
<Borromini>
:)
rsalvaterra has quit []
<Ansuel>
the idea is that swconfig is totally not compatible with dsa
<Borromini>
yeah, you can't have both
rsalvaterra has joined #openwrt-devel
<Borromini>
i have no ipq806x stuff but I have tested qca8k a bit with DSA on my ipq40xx hardware :)
<Borromini>
thanks for all your work
<Ansuel>
i still have to work about offload...
<Borromini>
:) i'm already happy DSA works on QCA stuff :D
<slh>
Borromini: ipq806x and ipq40xx are easy, relatively few devices, all with very uniform hardware (all with the same switch), ath79 is the elephant in the room which can't be converted in one go
<slh>
there some way to migrate in stages, device by device, is necessary
<Ansuel>
there is a pr to improve the situation
<Ansuel>
i just have to mod it to set the magic value to 1 to fix failsafe
<Borromini>
:)
<Ansuel>
and ath79 can start to have dsa
<slh>
if only because of the truly ancient devices (AR7161) not supported by qca8k or ar9331(?) which also needs (a lot-) more time to work with dsa
<Ansuel>
but i assume we have to wait 22.x branch as we are in feature freze
<Borromini>
at some point ath79 will obsolete itself. at this point, i won't be able to build 22.xx or master with wpad anymore e.g. for my 4 MB ath79 stuff
<Borromini>
haven't tried wpad-mini yet but there's little point in keeping WPA2 only etc stuff around, AP wise
<slh>
Borromini: still plenty ath79 devices with 8 or 16 MB (and more) flash
<Borromini>
true :)
<Borromini>
still pretty popular and I assume QCA is still producing that SoC, if not just under their renamed QN5xxx whatever moniker
<Ansuel>
for some reason also ipq806x is still used by some oem there is a device produced late 2021 o.O
<slh>
it's a good SOC, but still surprising, as it vanished from the consumer market completely by now
<slh>
even the good sellers (r7800)
<slh>
(or archer c2600)
<Ansuel>
also wonder if we should introduce a tag in openwrt to check pr to merge after 22.x merge
<Borromini>
jow what do you think? ^^
<xdarklight>
ynezz: could you please also close https://github.com/openwrt/openwrt/issues/7018 ? the latest two comments confirm that the issue has been resolved a while ago. I am writing this message from a computer connected to a "BT Home Hub 5A" via wifi, so I an confirm that this is really solved ;-)
Borromini has quit [Quit: Lost terminal]
Larhzu has quit [Quit: leaving]
<robimarko_>
Borromini: As you are, here finally traced the SFP GPIO-s for the RB5009
<robimarko_>
Got the SFP breakout right in second attempt, first time I forgot to select the 1.0mm PCB thickness
<robimarko_>
jow: Thanks for explaining what the extra 1 in AutoLoad does
<stintel>
russell--: how's the m300 working for you?
<robimarko_>
I had that exact problem with kmod based ethernet drvier
GNUmoon has quit [Ping timeout: 480 seconds]
<russell-->
stintel: i haven't touched it yet
<aiyion_>
Can one of you kick me in the right direction regarding uqmi? I'm interested in how the field `description` is filled for current_plmn in commands.nas.
<Ansuel>
robimarko_ we should add that to the wiki honestly
<aiyion_>
felix implemented it in e5e70f32ea if I'm not mistaken; but I can not figure out yet, where the function `qmi_parse_nas_get_serving_system_response` is implemented.
<robimarko_>
Ansuel: yeah
aiyion_ is now known as aiyion
<robimarko_>
Ansuel: BTW, I saw that you were doing some struct optimisations for qca8k to get rid of the holes reported by pahole
<stintel>
I dunno what the hell fsl/nxp people where smoking when they worked on the qoriq ppc dts files, what a horrible mess
<Ansuel>
robimarko_ yhea but i think i will postpone working on that as i have some idea on optimizing the mgmt eth stuff
<Ansuel>
i still have to understand if using skb_get is correct or not (to prevent release of the skb when the driver forward it
<robimarko_>
Ansuel: Nice, BTW the IPQESS driver is being prepared for upstreaming
<robimarko_>
Still gotta rebase the qca8k part after you recent changes so it can go upstream soon as well
<Ansuel>
we need to understand how to handle some write thing and what to drop
ekathva__ has quit [Remote host closed the connection]
<Ansuel>
feel free to drop an email to check how we should handle the code split
<aiyion>
Ah found the perl code generator, nevermind :S
ekathva__ has joined #openwrt-devel
<robimarko_>
Ansuel: I was thinking of sending the current version you have seen with the common code split soon
<robimarko_>
As RFC to see what would be best
<robimarko_>
Unfortunatelly, other work prevented me from working on IPQ40xx more
<robimarko_>
Wired networking is the last thing not upstream
<Ansuel>
oh nice
<Ansuel>
after that ipq40xx is fully supported?
<robimarko_>
Yes, there are some hacks like TCSR
<robimarko_>
But other than that we have upstreamed everything
<Ansuel>
i still have to work on tcsr
<robimarko_>
That thing is hopeless without a datasheet
<Ansuel>
i will have to probably drop tcsr for ipq806x and make it part of dwc3 qcom
<Ansuel>
and then try to propose something for tcsr for ipq40xx
<robimarko_>
Its a pain since it touches everything
<Ansuel>
there is a maintainer that have some info about it
<robimarko_>
So you cant just make it a part of other drivers
<Ansuel>
for example he suggested to drop multiple definition of tcsr and just make it one node with the 0x1000 reg size
<Ansuel>
my idea is that he has the datasheet...
<robimarko_>
If you need some info about USB TCSR regs, contact me
<robimarko_>
I have the USB regs datasheet
<robimarko_>
Thats if they left the TCSR stuff in the USB reg
<robimarko_>
Yeah, nope
<Ansuel>
... we have to hope they won't ask impossible thing upstream
<Ansuel>
99% they will...
<robimarko_>
Well, I dont blame them for not willing to merge stuff without any idea what it does
<robimarko_>
They got burned by broken downstream stuff before
<Ansuel>
like the entire krait stuff
<robimarko_>
Yeah, now they dont trust even QCA when they push stuff as they tend to push broken stuff more than guys who reverse engineer
GNUmoon has joined #openwrt-devel
<slh>
robimarko_: btw. ipq807x/ ax3600 (v5.15) seems to work pretty nicely, memory usage appears to be mostly stable (https://i.imgur.com/tiD5KUg.png); only tested with few devices connected so far
<robimarko_>
Thanks, its been way better than 5.10 now that its in decent
<robimarko_>
shape, redoing it from scratch really helped to find various stupid stuff I did
grid has quit [Quit: grid]
<Ansuel>
sorry for being late to the ipq807x party... someday i will officially close work on ipq806x and move on that...
<robimarko_>
Nothing to be sorry about
<robimarko_>
There is still plenty of work that can be done
<robimarko_>
The big thing is wired networking without QCA drivers
<robimarko_>
I dream about DSA
<robimarko_>
And there is lots to upstream yet
Tapper has quit [Ping timeout: 480 seconds]
<Ansuel>
yhea network is....
<Ansuel>
problem is dsa with nss... wonder if that can be integrated somehow
<robimarko_>
Well, not in the current scheme
<robimarko_>
Cause it using SSDK to configure the PPE on the fly
<robimarko_>
But I honestly dont care about offloading at all currently
<robimarko_>
Just want a proper DSA driver and not this POS that doesnt even handle roaming between ports properly
<Ansuel>
i have this crazy idea of testing qca8k + nss with the offload api... no idea how and if is doable but it would be intresting
<robimarko_>
BTW, do you have an idea on how could I monitor port changes
<robimarko_>
And then update the FDB
<robimarko_>
Adding netlink notifications in SSDK and then a kmod that would just listed to that comes to mind
<Ansuel>
mhh how port changes are done ?
<Ansuel>
i mean it's the normal phylink stuff or something else?
<Ansuel>
can't you just attach something there?
<robimarko_>
Well, that would be easy if it was a normal switch driver
<robimarko_>
No PHYLINK here
<Ansuel>
but ssdk have something like that if i'm not wrong
<robimarko_>
It registers each switch port as a separate netdev
ekathva__ has quit [Quit: Leaving]
<Ansuel>
i mean at worst you can think of implementing a notifier
<robimarko_>
Yes, SSDK has up/down notifications
<Ansuel>
and export it as a symbol
<Ansuel>
and attach what you need to it
<robimarko_>
Yeah
<Ansuel>
that seems to be the easiest and less hacky way
<robimarko_>
And then you basically have to manage the FDB in whatever kmod
<robimarko_>
And call SSDK to update it on the switch
<robimarko_>
Anyway, off to bed
<Ansuel>
that seems to be a correct way but why you need that ? i'm curious
<robimarko_>
Cause they hacked this thing together
<robimarko_>
As its registering ports as netdevs then there is nothing managing the FDB of the switch
<robimarko_>
So if you move a client from port 1 to port 2
<robimarko_>
Then you gotta wait for 30s aka the default timeout
<Ansuel>
o.o
<Ansuel>
OMFG
<robimarko_>
So that the duplicate FDB entry would be dropped
<Ansuel>
so your solution is to force remove the entry from fdb
<robimarko_>
They solved it with a NSS bridge kmod
<Ansuel>
....
<Ansuel>
hack on top of an hack on top of another hack
<robimarko_>
But that requires kernel changes and a running NSS-DRV with FW
<robimarko_>
Yeah, so I want to avoid having to have NSS-DRV at all for basic networking
<robimarko_>
Some users have been working around this with scripts, waiting for port change
<robimarko_>
And then flushing the FDB with SSDK shell
<Ansuel>
so your current idea is to skip any use of the nss firmware
<robimarko_>
Yes, why would I need it for networking without offloading
<Ansuel>
yes the quick way is just fast age everything...
<robimarko_>
Yes, they are using ssdk shell cause it doesnt require any coding
<Ansuel>
yep they flush the fdb but that cause downtime am i wrong ?
<robimarko_>
Just a shell script
<robimarko_>
Yes, techically you can have dropped packets until the switch relearns
<robimarko_>
Its fast, but still not a correct way to solve it
<robimarko_>
I posted the workaround link
<Ansuel>
anyway i think you can use a mix of notifier + complation to wait for the fdb removal
<robimarko_>
Yes, basically implent a SW FDB
<robimarko_>
Anyway, off to sleep now really
<robimarko_>
Long day tommorow
<Ansuel>
good night
robimarko_ has quit [Quit: Leaving]
Ansuel has quit [Quit: Probably my PC crashed or time to sleep.]
danitool has joined #openwrt-devel
Ansuel has joined #openwrt-devel
shibboleth has quit [Quit: shibboleth]
AtomiclyCursed has quit [Ping timeout: 480 seconds]
<Ansuel>
jow i added the refrence of the AUTOLOAD in the wiki
<Ansuel>
(i assume it wasn't present)
Ansuel has quit [Quit: Probably my PC crashed or time to sleep.]
<philipp64>
anyone have an easy/portable way to parse output of a command for a multiline match? I need to parse the output of "named-checkconf -p" for /^options {\n(\t.*\n)*\tlisten-on-v6 {\n\t\t"none";\n\t};/m without introducing excessive dependencies (like Perl).
lmore377 has quit [Ping timeout: 480 seconds]
<dwfreed>
philipp64: sed
<philipp64>
yeah, i was hoping for something simpler...
<dwfreed>
sed is simple
<dwfreed>
just load the whole file into the pattern space and then apply that regex