clayface has joined #openwrt-devel
clayface_ has quit [Ping timeout: 480 seconds]
<Tusker> hmm... is there a way in the build scripts to add in the kernel and dtb into the rootfs filesystem itself ?
<Tusker> can I add files into $(STAGING_DIR_IMAGE) prior to sysupgrade-tar ?
kenny has joined #openwrt-devel
dangole has quit [Remote host closed the connection]
Ansuel has quit [Quit: Probably my PC crashed or time to sleep.]
danitool has quit [Ping timeout: 480 seconds]
victhor has quit [Ping timeout: 480 seconds]
kenny has quit [Ping timeout: 480 seconds]
Andi_ has quit [Remote host closed the connection]
Andi_ has joined #openwrt-devel
valku has quit [Quit: valku]
fda- has joined #openwrt-devel
<genuser1> I managed to compile this "openwrt-21.02.0-rc4-ipq40xx-generic-netgear_wac510-xxxx" and it boots with luci and opkg seems to be working so far. It's a good enough milestone for me for this weekend and I hope to test it further throughout the week. Thanks all here who helped me get through the compilation pangs. :)
hanetzer1 has joined #openwrt-devel
hanetzer is now known as Guest3250
hanetzer1 is now known as hanetzer
fda has quit [Ping timeout: 480 seconds]
Guest3250 has quit [Ping timeout: 480 seconds]
nitroshift has joined #openwrt-devel
rmilecki has joined #openwrt-devel
Tapper has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
fda has joined #openwrt-devel
fda- has quit [Ping timeout: 480 seconds]
goliath has joined #openwrt-devel
Tapper1 has joined #openwrt-devel
Tapper has quit [Read error: Connection reset by peer]
Acinonyx_ is now known as Acinonyx
goliath has quit [Quit: SIGSEGV]
aleksander has joined #openwrt-devel
fda has quit [Ping timeout: 480 seconds]
fda has joined #openwrt-devel
decke has joined #openwrt-devel
goliath has joined #openwrt-devel
floof58_ has quit []
floof58 has joined #openwrt-devel
pmelange has joined #openwrt-devel
dangole has joined #openwrt-devel
swegener has quit [Ping timeout: 480 seconds]
pmelange has left #openwrt-devel [#openwrt-devel]
ynezz has quit [Ping timeout: 480 seconds]
Ycarus has joined #openwrt-devel
danitool has joined #openwrt-devel
ynezz has joined #openwrt-devel
ynezz is now known as Guest3267
Guest3267 has quit []
ynezz_ has joined #openwrt-devel
ynezz_ is now known as ynezz
ynezz has quit []
ynezz has joined #openwrt-devel
ynezz is now known as Guest3268
Guest3268 is now known as ynezz
Andi_ has quit [Ping timeout: 480 seconds]
Andi_ has joined #openwrt-devel
genuser1 has quit [Quit: Page closed]
victhor has joined #openwrt-devel
aleksander has quit [Quit: Leaving]
megabro has joined #openwrt-devel
embargo has quit [Read error: Connection reset by peer]
rsalvaterra has quit [Quit: rsalvaterra]
rsalvaterra has joined #openwrt-devel
<karlp> jow: I got that resource loading working in the end, thanks for your help, but there's now a double class invoke that seems weird? https://paste.jvnv.net/view/bQHAi
<karlp> the line 39 there seems really odd, like something else isn't quite right.
<karlp> I think this is not layering properly for sharing a load() method either.
<karlp> like the parameters that go to a render call from a load call aren't obvious how to pass along
<karlp> huh, actually just dropping the last wat.render() works too. the "var wat = new m()" is enough to let the resource class do it's own load/render things...
pmelange has joined #openwrt-devel
<Tusker> I am trying to copy the kernel image and dtb into the ext4 filesystem prior to sysupgrade-tar being called, at what step can I intercept and copy those generated files to the staging directory, prior to the mkfs.ext4 etc ?
<hitech95> best way to boot an image from uboot without flashing?
<karlp> tftp an initramfs image and boot it?
pmelange has left #openwrt-devel [#openwrt-devel]
<hitech95> karlp: no idea on how, is there some info about the commands to use? I was able to do this on lantiq via serial, but on ath79 I have no idea on how to do it
<PaulFertser> hitech95: usually it's "tftpboot" and then "bootm".
<hitech95> PaulFertser: yea but how about the addresses?
<PaulFertser> hitech95: I think on all ath79 SDRAM is mapped to same address so you can try instructions from some other board.
<karlp> "somewhere in ram" just experiment with ~any other ath79 directions.
<karlp> try 0x81000000 for instance...
<PaulFertser> Or 0x80060000
<karlp> if you want to run it, not just copy it to flash, you might want higher... at least my own stuff uses 80060000 for "tftp, then flash" but 8100000 when it's goign to run it.
<karlp> so it gets decompressed properly
<rsalvaterra> Hmm… I think I'm seeing a cute behaviour regarding ARP ageing on my MT7621 thingy. Digging…
<hitech95> karlp, PaulFertser , thanks I'll try. Do i need some special built images? I'm tring to develop a kernel driver and I dont want to have to flash 50+ time the memory IC with all my fails!
decke has quit [Quit: Leaving.]
<PaulFertser> hitech95: initramfs as built by OpenWrt should work
<karlp> enable ramdisk images in menuconfig
<jow> karlp: require 'Junk'; return Junk.Junk(...);
<jow> sorry, 'require Junk';
<karlp> as just another line after the require module.file as some_name?
<jow> no need to return view.extend() from the view again, after all Junk.Junk() already returns a view class
<jow> yes
<karlp> yeah, the view extend twice seemed quirky. but I wasn'
<karlp> t sure how ot strip it out.
* karlp pokes it some more...
<karlp> so the final view is just "return new Junk('blah')" ? instead of return view.extend() ?
<jow> no, reutrn Junk.Junk()
<jow> 'require' already implicitely calls new on the class being loaded, so you can't pass parameters to hit
<jow> hence the need for another "constructor" like function, which we named Junk in our current example
<jow> 'require remake.test1' implicitely calls new reusableLogView()
<jow> which is then passed as `logView` into your view
<jow> on that `logView` base class instance you can call the method `makeLogView`
<jow> which simply returns a class declaration
<jow> so it is a class factory function basically
kenny has joined #openwrt-devel
<jow> and you return that class directly without instantiating it
<jow> because the code requiring your view expects a class, not a class instance as return value
<jow> it will then instantiate the view class you returned and call various methods on it
<karlp> right, that name changing makes it a lot clearer.
rua has quit [Quit: Leaving.]
rua has joined #openwrt-devel
<karlp> I'mnot sure how well this fits into the wdigets.js though, that's all has renderWidget() functions, and is CBI focussed
<karlp> would this be ideally "widgetized" ? or just a new file in resources/tools?
pmelange has joined #openwrt-devel
dangole has quit [Remote host closed the connection]
dangole has joined #openwrt-devel
<tmn505> Tusker: check out how is the tegra, sunxi or mvebu creating sdcard image, which shoudld be easly bootable when written also to USB stick or other block device. For more sophisticated image creation check the mediatek target.
<f00b4r0> silly question maybe but is there a way to configure hostapd so that when it switches channels due to DFS, it stays in a regulatory band where the same txpower can be maintained?
<Tusker> tmn505: yeah, i've seen those, just want to re-use the sysupgrade-tar instead of just using an sdcard image
<Tusker> will keep digging :)
f00b4r0 has quit [Read error: No route to host]
f00b4r0 has joined #openwrt-devel
victhor has quit [Ping timeout: 480 seconds]
nitroshift has quit [Quit: Gone that way --->]
<Tusker> thanks
Tusker has quit [Quit: Time wasted on IRC: 15 hours 41 minutes 40 seconds]
dangole has quit [Remote host closed the connection]
dangole has joined #openwrt-devel
dangole has quit [Remote host closed the connection]
valku has joined #openwrt-devel
dangole has joined #openwrt-devel
enyc has quit [Ping timeout: 480 seconds]
goliath has quit [Quit: SIGSEGV]
<mrkiko> hello all!
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
Andi_ has quit [Read error: Connection reset by peer]
Andi_ has joined #openwrt-devel
KGB-0 has quit [Quit: KGB-0]
KGB-0 has joined #openwrt-devel
goliath has joined #openwrt-devel
<rsalvaterra> Ok, ARP ageing only works when I enable hardware flow offloading in MT7621.
<rsalvaterra> Is that expected?
<rsalvaterra> Without hardware flow offloading, ARP entries linger ad eternum in FAILED state.
minimal has joined #openwrt-devel
swegener has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
minimal has quit []
clayface_ has joined #openwrt-devel
clayface has quit [Ping timeout: 480 seconds]
danitool has joined #openwrt-devel
kenny has quit [Read error: No route to host]
kenny has joined #openwrt-devel
Ansuel has joined #openwrt-devel
<Ansuel> robimarko: ping ?
<mangix> rsalvaterra: upstream has a patch that can be backported for this AFAIK
kenny has quit [Ping timeout: 480 seconds]
<rsalvaterra> mangix: Can you point me to it?
<rsalvaterra> I can backport it, this has been annoying me for months. :P
<rsalvaterra> (I had never used hardware offloading, only in the last couple of days.)
arifre has joined #openwrt-devel
philipp64|work has quit [Remote host closed the connection]
<hitech95> I'm getting frustrated over a simple SPI bitbanged device driver that works with basic bash script but my kernel module does not!
pmelange has joined #openwrt-devel
genuser1 has joined #openwrt-devel
genuser1 has quit [Remote host closed the connection]
Tapper1 has quit [Ping timeout: 480 seconds]
<hitech95> if a target have both the 5.4 and the 5.10 kernel, I have to create two distinct commits for adding patche for both kernel versions?
<rsalvaterra> hitech95: Not really. You can add both patches in the same commit.
<rsalvaterra> It's the same feature/bugfix.
<Ansuel> anyway wps is currently broken....
<Ansuel> everytime is use it it does report wrong auth option
<Ansuel> a device registred with that option result in hostapd complaining for wrong auth mechanism
<Ansuel> any idea how to debug that?
<Ansuel> it's present from at least a year i tuink
<Ansuel> think*
<Andi_> would be possible to enable KERNEL_MIPS_FPU_EMULATOR?
<hitech95> rsalvaterra: so something like this might be good? https://github.com/hitech95/openwrt/commit/d351b4fc8baa811ff840015496492dbf56518eb0
<rsalvaterra> Andi_: Don't. Trapping and emulationg in kernel space is way slower than userspace emulation.
<rsalvaterra> *emulating
<Andi_> i would like to try it, i've just compiled alsa-equal + caps, but it is unusable due to cpu going to 100% usage on MT7628, maybe with fpu emulation would work better..
<rsalvaterra> hitech95: No C++-style comments (//), please. Otherwise, looks fine, if it's been tested, of course (this isn't a review, mind you).
<aparcar[m]> rsalvaterra: did you test ath79 with 5.10?
Tapper has joined #openwrt-devel
<rsalvaterra> aparcar[m]: Yes, I've been running 5.10 on ath79 for months, on my devices. And DSA too, since a couple of weeks ago.
<dwfreed> Andi_: kernel fpu emulation is going to be worse, 100%
<aparcar[m]> rsalvaterra: please report here https://github.com/openwrt/openwrt/pull/4619
<dwfreed> Andi_: if there is no fpu available in hardware, then the compiler will correctly generate instructions to emulate one
<aparcar[m]> also mangix here is an overview please report if you got experience with the kernels https://github.com/aparcar/openwrt/issues/15
<rsalvaterra> Andi_: If you're trapping/emulating in kernel space, you'll have a context switch for every emulated instruction.
<Andi_> ok i understand it
<Andi_> do you have any sugestion for a player that replaces mpg123 that supports equalizer?
<hitech95> rsalvaterra: yea I know that baout comments, I'm still testing the patch. I was just asking about the single commit for 5.4 and 5.10
<mangix> aparcar[m]: ath79 and ramips work pretty well under 5.10. All I can test.
<mrkiko> Andi_: mpd?
<mrkiko> Andi_: not exactly command-line player but you can use it as such somehow
<mangix> Andi_: the FPU emulator not only increases size, but it's considered to be a security risk.
<mangix> it's also slower than just compiling as softfloat
<mangix> yousong had a patch more thoroughly removing it. Need to find it...
<aparcar[m]> mangix: okay thanks
<mangix> actually I'm running rampips 5.10 right now. Need it for proper LEDs on my device
<mangix> sorry, LED control
<Ansuel> you earned an enemy mangix >:(
eduardo010174 has joined #openwrt-devel
<Habbie> somebody mentioned a project that builds openwrt images, 'as a service' but also locally hostable, in here, recently
<Habbie> can somebody remind me what it was
<Habbie> ?
<mangix> aparcar's ACU I thimk
<Habbie> asu
<Habbie> thanks!
<dwfreed> Habbie: hai
<dwfreed> :)
<Habbie> haha
<Habbie> dwfreed, so re your question in the unrelated channel, did you see https://github.com/aparcar/asu ? :)
<dwfreed> I did not
<aparcar[m]> auc is the cli for the server (asu)
<Ansuel> asu stands for?
<aparcar[m]> Attended SysUpgrade
<Ansuel> owww ok that project
<aparcar[m]> actually you can have it unattended by using `auc -y` in a cron job...
<aparcar[m]> yea that project again
<aparcar[m]> anyone here using a "large" wireguard setup? The fact that it needs a `network` restart to reload the configuration seem terrible
<dwfreed> wireguard itself supports a config sync
<dwfreed> 'wg syncconf interface config-file'
<aparcar[m]> dwfreed: sure but network reload doesn't trigger that from what I can tell
<eduardo010174> somebody have time for review my pull request ? https://github.com/openwrt/openwrt/pull/4262
<dwfreed> so we should fix the proto handler
<aparcar[m]> dwfreed: that would be great. Ideas on how to do that?
<dwfreed> nope, I have no idea what netifd protocols are like
<dwfreed> does it implement a reload?
<aparcar[m]> from what I can tell, no
dedeckeh has quit [Remote host closed the connection]
eduardo010174 has quit [Quit: Page closed]
eduardo010174 has joined #openwrt-devel
eduardo010174 has quit []
<aparcar[m]> dwfreed: looking at netifd-proto.sh maybe proto_wireguard_renew could to do the trick...
genuser1 has joined #openwrt-devel
<mangix> huh
<mangix> i thought mt7628 was using DSA already
<Ansuel> btw mangix: looking at the qca8k driver i just notice how immature it is... we lacks many things that should be enabled... but i need to check the status of dsa in 5.10
<Ansuel> as i'm all checking stuff directly from net-next and adding feature
<mangix> Ansuel: lol. why would it be mature? it's a new driver.
<mangix> I don't think QCA has worked on it at all.
<mangix> IIRC blogic wrote and upstreamed it.
<Ansuel> actually it's present from 4.9
<Ansuel> it's one of the first driver... yes no work from qca it's all from community
<slh> Ansuel: btw., I've extended my DSA test (trunk port to a smart-managed switch) setup with bridge-vlan filtering setup on the g10, works fine with the single cpu-port support as of ~2 weeks ago (need to find a good time to check the newer multi cpu-port support again). iperf3 is limited to 260 MBit/s that way (real-world testing using an external speedtest more like ~230 MBit/s)
<Ansuel> slh: aside from shit perf did you notice any problem?
<Ansuel> packet dropping instability?
<slh> so far it works, but only without multi cpu-port patches (but again, r17677-f82c93b93c plus your PRs from 2021-10-05)
<Ansuel> mangix: the problem is that it seems in some time it was dropped and was leaved to a semi working state. basic function but this switch can support much more looking at the Documentation
<Ansuel> slh: god i should really update the multicpu pr.... but i have an exam this week so no time at all.... i hope e9hack can help with his super special rebased patch
<mangix> Ansuel: oh don't know if I showed you this: https://gist.github.com/neheb/c86e3898d765aa4d53442773e123c5e0
<slh> Ansuel: take your time, I'm not in a hurry
<Ansuel> anyway if you want you can use the one present in the pr... should be fine just you need to check if the script at init.d still work
<Ansuel> (the one that autoset the port to the correct cpu port)
<slh> the nbg6817 works with software bridging and multi-cpu patches at sufficient speeds
<Ansuel> mangix: wonder if the problem comes from the ignore speed strap?
<mangix> maybe
<Ansuel> there are some strap that are not controllable by regs
<mangix> the initvals did not have them btw
<Ansuel> wonder if that option just ignore ANY strap an not just eeprom and led
genuser1 has quit [Quit: Page closed]
<slh> I'm just happy that I can finally test a bit more with bridge-vlan filtering, getting closer to the (slightly complex) setup I'm looking for. I was already questioning my sanity, given that how I understood DSA didn't (appear to) work with the multi cpu-port patches
Tapper has quit [Ping timeout: 480 seconds]
<rsalvaterra> mangix: Hold up, I'm already carrying that patch. :/
<dwfreed> aparcar[m]: so netifd doesn't provide anyway for protocols to differentiate between reload and start inherently; one would have to check existing config in the init, and then do sync instead of recreating the interface if there's an existing config
<dwfreed> s/anyway/any way/
<hitech95> ralking about the QCA driver, blogic had created the offload patches for the NSS. Are they still not available/not mergable to mainline? LAst time i've seen his works was years ago...
<aparcar[m]> dwfreed: well that doesn't seem to hard to implement, does it?
<dwfreed> aparcar[m]: just needs someone to take the time to do it
<aparcar[m]> so "reload" actually calls the start function?
<dwfreed> it just runs 'config_init_all' again
<dwfreed> in netifd internals
<dwfreed> which is the same thing startup does
<hitech95> any info on what happened here? https://forum.openwrt.org/t/hardware-nat-for-lede/1094/49
<dwfreed> netifd internally knows reloads separately, so it could theoretically be reworked to do things differently
<slh> hitech95: what exactly are you referring to, as QCA driver is extremely ambiguous with qca8k (switch), ipq806x and ipq807x
<slh> or ipq60xx/ ipq50xx
<hitech95> slh: see my last link to the reply from blogic in the forum
<aparcar[m]> dwfreed: I'll try to make more sense out of it
<slh> hitech95: qca8337N hardware offloading has nothing to do with NSS
Andi_ has quit [Ping timeout: 480 seconds]
<hitech95> slh: Uhm... so my bad :)
<slh> sadly not too many device have the n-variant of qca8337 either
Andi_ has joined #openwrt-devel
<hitech95> NSS is the old offloading?
<slh> the other way round, more or less
<hitech95> oh ok!
<slh> but getting qca8337n hardware NAT working is probably easier than NSS
<hitech95> But ath79xx ICs what offloading "type" they have?
<slh> ath79, nothing - only ipq806x/ ipq807x have NSS. qca8337N is the switch, which is orthogonal to the other things (but ipq806x devices usually don't come with the n-variant and ipq807x is another topic altogether)
<slh> but you usually can't rely on wikidevi and friends being accurate about qca8337 vs qca8337n, no way around opening the device and inspecting it visually
rmilecki has quit [Ping timeout: 480 seconds]
<hitech95> got that! I'm on x86 due to MAP-T and GPON SFP from my provider. WAs looking about the MTK PPE but even that only IPv4. So the older ath79 socs does not have any HW offload in them?
<mangix> nevwer published AFAIK
Andi_ has quit [Ping timeout: 480 seconds]
<slh> hitech95: using qca8k (as in the ongoing PRs) instead of swconfig will bring a moderate speedup for ath79 (~+70 MBit/s), but apart from that, none
<Ansuel> i already have functional hwnat/routing offloading patches, that i hope to publish under GPL in the near future.
<Ansuel> rip
<hitech95> Ansuel: dumb question, you are italian or it was a just by chance that you worked on Technicolor HW?
<dwfreed> their IP address suggests they are Italian :)
<hitech95> slh: got that, unfortunatly by beeing in MAP-T I think that the only hardware that can accellerate it is MArvell and Broadcom. But they both lack support.
<Ansuel> italian
<hitech95> Good to se another one :)
<Ansuel> mangix: wonder if blogic still have that patches... it's really a pitty that they were just dropped...
<Ansuel> it's just sad
<hitech95> Ansuel: theese? https://git.openwrt.org/?p=openwrt/staging/blogic.git;a=commit;h=dd3bdac6d1dcd98d4d494052f7df31ca21558d6f
<Ansuel> also looking at the discussion everyone started using shitty qsdk stuff... instead of asking blogic for updates...
<Ansuel> O_O
<mangix> Ansuel: probably on some computer somewhere
<hitech95> yea, common people moved to fastpath again
<mangix> also, I remember using those patches way back in the day
<mangix> I had no idea what I was doing.
<Ansuel> but fastpath is just removing some stuff from kernel logic... the cpu still has load
<Ansuel> lol the amount of work blogic put....
<Ansuel> should i just take it and rebase it on top of 5.10 ?
<Ansuel> that seems to be a nice project
<mangix> Ansuel: oh yeah. He also had a slimmed down version of the NSS driver. Didn't post or upstream it.
<slh> at least for ipq806x it won't help much, afaik none of the devices have qca8337n
<mangix> remember him talking about it on here.
<hitech95> Ansuel: if I remember correcly that patche was using the initial netfilter idea/boilerplate for offloading. I think that you will need to support the new offload structure...
<Ansuel> hitech95 i know...
<Ansuel> but the fact that 99% of the code is done it's always a good thing...
<hitech95> yea for sure!
<Ansuel> slh: no qca8337n on ipq806x ? really ?
<hitech95> I finally managed to have my dumb LED driver to work.
<slh> Ansuel: I /think/ all ipq806x have only qca8337, not qca8337n - no guarantees though
<slh> newer ath79 devices however...
Tusker has joined #openwrt-devel
<Ansuel> ath79 device have qca8337n but not ipq806x ?
<slh> yes
<slh> some
<Ansuel> i assume it doesn't make sense to have qca8337 n on ipq806x as we have nss
<slh> probably, yes - 10ct saved on qca8337
<Ansuel> blog work was incredible and honestly he had done the same thing i wanted to do....
<mangix> Ansuel: those IRC logs are probably 3 years old
<Ansuel> aka mod the gmac driver and add support for offload using nss
<Ansuel> mangix: sad... wonder what blocked him... the amount of work needed to propose that?
<Ansuel> the fact that is was litterally impossible to do?
<mangix> probably busy with IRL stuff
<hitech95> is he retired?
<mangix> no
<mangix> blogic is quite young IIRC
<hitech95> blogic made all the hardware I own/owned to work.
<Ansuel> mangix: probably same situation like me... he got documentation and source at times and he notice that it was a nice project to understand the entire kernel net part
<hitech95> He really did a great job!
<Ansuel> and then he started to work seriusly LOL
<Ansuel> what i would be interested would be if he even worked on the driver loading part...
<mangix> anyway. I haven't seen him recently
<slh> late august
<Ansuel> anyway it looks like r7800 have 8337n
<Tusker> looks like blogic is working on the ipq40xx devices still - https://github.com/Telecominfraproject/wlan-ap
<Ansuel> he probably just started to work
<slh> san't see it clearly on the r7800 FCC images, but I wouldn't trust wikidevi or the OpenWrt wiki (without visually confirming it) in that regard
<slh> can't
<Ansuel> the fcc internal photo are shit....
<slh> sadly, yes
<slh> tl-wdr4300 has qca8327n though
<Tusker> r7800 bootlog says 'switch0: Atheros AR8337 rev. 2 switch registered on gpio-0'
<Ansuel> they have same id so....
<Ansuel> they can only be identified with the id on the chip i think
Tapper has joined #openwrt-devel
<Ansuel> what about g10 nbg ?
<slh> nbg6817 has qca8337n, clearly readable on the FCC images
<slh> err
<slh> nbg6817 has qca8337, clearly readable on the FCC images
<slh> no-n
<Ansuel> link image? ahhaha
<slh> can't read the writing on the g10 FCC images
pmelange1 has joined #openwrt-devel
<Tusker> r7800 according to various people is QCA8337N/IPQ8065 - ie, https://www.snbforums.com/threads/r9000-dd-wrt.40032/
<slh> can't quite read https://hwready.it/wp-content/uploads/2015/11/IMG_1071.jpg (g10), the FCC images are blurry
<mangix> let's see, 4900
<mangix> dual CPU switch. nothing special
<slh> so both nbg6817 and g10 are no-n qca8337
<slh> need to bookmark https://hwready.it/recensioni/asrock-g10-router-gaming-recensione/5/ those internal images are good
<mangix> looks italian
<mangix> ah just noticed the .it
<Ansuel> 2nd italian tech news website LOL
pmelange has quit [Read error: Connection reset by peer]
<Ansuel> andyway it's incredible how blogic patch are very similar to min except some difference that i will ""steal""
<Ansuel> wonder if he had access to some special documentation...
<mangix> hahaha do tell
<Ansuel> the number of work on the offload patch tho... it's just incredible
pmelange has joined #openwrt-devel
<slh> https://www.expreview.com/47447-all.html <- r7800, no-n qca8337
<Ansuel> well rip offload patch 1?!?!?!
pmelange1 has quit [Ping timeout: 480 seconds]
<slh> the archer c7 has qca8337n (and there it will presumably make a real difference), and the tl-wdr4300 has qca8327n (iirc you had a tl-wdr4300 or tl-wdr3600?)
<Ansuel> anyway the real problem is understand how the offload structure now works...
<Tusker> i've got a tl-wdr4320 here if you want me to check the chip itself ?
<Ansuel> i have a wdr3600 but in another home so no luck in testing on that
<Ansuel> think things will go with understand how to port everything and you guys will test
<dwfreed> datasheet for 8337n
<Ansuel> old version there are newer online
<slh> even the RB3011UiAS-RM appears to have only no-n qca8337
<Ansuel> it makes sense as we have nss
<slh> does anyone know of a "cheap'ish" (used?) PCIe ethernet card that supports copper 2.5GBASE-T, 5GBASE-T and 10GBASE-T?
<stintel> intel x540-t1 on ebay ?
<hauke> mac80211 from 5.15 compiles for me, but it crashes when a station connects to my ath9k AP, see here: https://pastebin.com/GENvSU2j
<slh> thanks, intel x540-t1 looks indeed pretty reasonable :)
<hauke> if someone else wants to have a look see here: https://git.openwrt.org/?p=openwrt/staging/hauke.git;a=shortlog;h=refs/heads/mac80211-5.15
robimarko has joined #openwrt-devel
<robimarko> hauke: I will give it a go tommorow, still fighting with the stupid ath11k embedded regulatory anyway
<Ansuel> still can't understand how roby join and answer stuff
<Ansuel> irc blackmagic?
pmelange has left #openwrt-devel [#openwrt-devel]
<robimarko> Ansuel: Its the IRC log
<robimarko> I usually keep an eye on it when not doing anything
<Ansuel> ow ok so you check the log without joining ok makes sense
<Ansuel> is it real time?
<Ansuel> oh wow
<robimarko> Yeah, if you leave the live updates enabled
<Ansuel> good to know
eduardo010174 has joined #openwrt-devel
<robimarko> Its also easier to follow as you can disable the leave/join messages
<robimarko> BTW those patches from blogic look insane, he really did strip down NSS to do what he needed
<Ansuel> mhh they are not about nss tho
<Ansuel> or did you also find the nss patch
<robimarko> Yeah I know
<robimarko> Its just HNAT in the switch
<Ansuel> yhea... it's incredible how he fixed ssdk
<Ansuel> massive work
<robimarko> I am sure he went insane considering that SSDK is so terribly written
<Ansuel> pity there is no patch about nss wonder if he tweaked also the load driver
<Ansuel> he posted some irc chat about that and well... same reaction we had at times when we checked ssdk
eduardo010174 has quit []
<robimarko> I am following all of the qca8k stuff as IPQ40xx technically uses QCA8337 inside of it
<robimarko> Should even be th N variant
<Ansuel> also same reg?
<robimarko> But obviously MMIO based instead
<robimarko> Yeah, the switch part uses the same register layout
<Ansuel> good but is it used somehow?
<Ansuel> i mean the driver for ipq40xx
<robimarko> I am trying to get approval at work to publish our WIP ethernet + qca8k based driver for the IPQ40xx
<robimarko> Its based of old qca8k so it doesnt really have fancy features like the current one but still works remarkably fine
<Ansuel> so happy that i'm indirectly helping other stuff
<robimarko> The biggest issue is that whole mess of PSGMII
<Ansuel> that is also used for ipq807x right?
<robimarko> Its full of bugs and it requires a lot of hacks to get it working
<robimarko> Yeah, IPQ807x and IPQ60xx use the same switch for the 1G ports
<robimarko> But the ethernet controler is much newer and has 3 SERDES-es in IPQ807x and 2 in IPQ60xx
<Ansuel> you specify "for 1g" 2.5 use different stuff ?
PaulFertser has quit [Ping timeout: 480 seconds]
<slh> aquantia for those
<Slimey> yup
<robimarko> 2.5G just uses QCA8081 PHY-s
<robimarko> 10G is Aquantia AQR107
<Ansuel> so the blocking stuff currently is the psgmii hack stuff and the "to be approved" driver
<robimarko> Well, the PSGMII hacks are the blocker
<robimarko> Otherwise it would be a simple adaptation of qca8k to use MMIO regmap and disable the internal MDIO
<Slimey> aqrate aqr106
<slh> and ipq807x is probably another level, compared to 'just' ipq40xx for which this driver is being developed
<robimarko> Yeah, IPQ807x is a whole new beast
<Ansuel> sacred of the thing behind ipq807x...
<Ansuel> scared*
<robimarko> It just shares the same switch setup for 1G ports
<robimarko> Good luck getting that to work semi usefully without an datasheet
<Ansuel> so you are telling me they just recycled qca8k regmap for everything lol ?
goliath has quit [Quit: SIGSEGV]
<robimarko> They just reused the same switch setup as its simple and good enough for 1G
<robimarko> Despite it actually not being so good due to a single 1G uplink to the CPU
<Ansuel> 2g if you use the 2 cpu with lags
<robimarko> Ansuel: But you cant
<Ansuel> ?
<robimarko> It only has the port0 connected
<robimarko> The switch is built into the SoC
<Ansuel> ow....
<robimarko> It only has external PHY-s
<Ansuel> so 5 port even the wan port all connected to 1g ?
<robimarko> Yeah
<robimarko> Its all connected to a single 1G uplink
<robimarko> I am sure they inteded to use the HNAT features back when it was being designed
<robimarko> But I havent seen that actually being used
<slh> the bigger devices just combine that with (usually one) Aquantia 2.5/ 5 rarely 10 port for WAN or LAN purposes
<Ansuel> well it's all in ssdk and i think it's enabled by default
<robimarko> slh: For the IPQ807x/60xx they always use QCA8081 for the 2.5 port
<robimarko> As its cheaper and has no FW
<robimarko> Its actually a decent Clause 45 2.5G PHY
<robimarko> Ansuel: Only if you selected the HNAT SSDK
Tapper has quit [Ping timeout: 480 seconds]
<Ansuel> anyway the idea of using hnat is still strange... example on ipq806x we have 2 cpu and they allocated 1g to wan and 1g to lan but we all have qca8337 NOT N... so hnat is not supported at all... think the idea is that you have nss but still you are limited by 1g uplink
<robimarko> slh: Thats the first one I have seen using AQR111
<robimarko> But thats 5G
<Ansuel> anyway i still have to understand how to dected a hnat supported switch from sw
<robimarko> Ansuel: I dont think you can
<Ansuel> FUN.
<robimarko> Hence the build time selection
<Ansuel> in our case we can use the compatible
<robimarko> Yeah, that is fine
<robimarko> As the ID is 0x13
<robimarko> And the non N has the same one
<Ansuel> but wow.... that is really stupid considering they could just set a ro reg to declare support for it...
<Ansuel> question does ipq40xx have an interrupt for the switch ?
<hauke> mac80211 from kernel 5.15 does not crash for me in a simple test
<robimarko> Ansuel: I dont think so
<robimarko> At least not GIC based ones
<robimarko> Only GMAC has them
<robimarko> And since its built in into the silicon its not a GPIO based as well
<Ansuel> sad stuff... think they only fixed that with new soc
<Ansuel> i remember ipq807x having interrupt for each port
<Tusker> a question on the build process, is there a way to use sysupgrade-tar, but inject files into the rootfs, prior to the mkfs ? I know there is the possibility of putting files in ./files/ for static files, but what about the kernel and dtb themselves ?
<robimarko> Ansuel: Well, the whole IPQ40xx wired networking is a real mess
<robimarko> It all seems like half baked
<robimarko> They essentially made it out of existing stuff they had
<Ansuel> that soc looks like a transition soc... new stuff + old stuff - expesive one to make it cheap
<robimarko> Yeah
<robimarko> It was cheap and had everything built int
<slh> it's basically an ath79 replacement
<robimarko> Yeah
<robimarko> I mean its way better than ath79
<slh> yep
<robimarko> And they are still cranking them out
<aparcar[m]> zx2c4: Hi, unrelated to the recent GitHub merge: Do you know a way to reload the wireguard configuration without restarting the network? I'm currently looking into solutions for that. If you already have an idea that would be very helpful
<Ansuel> robimarko they are still producing device with ipq40xx ?
<mangix> Ansuel: archer c7v2 supports it AFAIK
<mangix> patch was denied way bakc when
<slh> Ansuel: yes, ipq40xx is still on the shelves - probably another year or two until ipq60xx and ipq50xx steal its thunder
<Ansuel> lol i hope not... ipq60xx and 50xx is just a meme now
<Ansuel> don't even know if they finish sending core patch upstream
<slh> bbbbutt, wifi6 ;)
<slh> (imagine a number of exclamation marks in the above) ;)
<robimarko> I wouldnt count on QCA to do the upstreaming
<robimarko> They did some initially like always
<robimarko> But trust me, not even the clock drivers are complete
<robimarko> But that the usual with QCA
<slh> RAM will be a(n even) big(ger) issue for ipq50xx/ ipq60xx
<robimarko> Oh yeah, that is why only the cheapest junk will have 256MB
<robimarko> Would not recomend that
<robimarko> As they barely made it run with that
<robimarko> Had to cut a lot of ath11k features and cap the number of clients