nlowe has joined #openwrt-devel
nlowe has quit []
nlowe has joined #openwrt-devel
nlowe has quit []
nlowe has joined #openwrt-devel
<Grommish> x86_64 targets stil use musl, right?
<stintel> everything but arc
<Habbie> what does arc use?
<stintel> glibc
<Habbie> ah
<stintel> because musl doesn't support it
swalker has quit [Remote host closed the connection]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
spacewrench_ has quit [Quit: Leaving]
goliath has quit [Quit: SIGSEGV]
nlowe has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
slh_ has joined #openwrt-devel
slh has quit [Ping timeout: 480 seconds]
slh64 has quit [Ping timeout: 480 seconds]
danitool has quit [Ping timeout: 480 seconds]
slh has joined #openwrt-devel
slh_ is now known as slh64
Monkeh has quit [Ping timeout: 480 seconds]
Monkeh has joined #openwrt-devel
<Grommish> Thanks! :) I've got successful toolchain compiles now for aarch64, armv7, mipsel, mips64, and x86_64. I'm working my way thru the Archs
<stintel> nice work
<Grommish> ppc seems like a no go, but I'll revisit in the future when I get bored
<stintel> I'm still fighting ppc asm to fix musl
<stintel> it's a fucking shitshow
<Grommish> Any idea what the armv6 SOC is called? i know there is a target that uses it
<stintel> no ppc asm examples I can find still work on modern compilers etc
<Grommish> stintel: Well I hit the ppc asm wall myself with it insisting valid memory registeres arent
<Grommish> https://pastebin.com/cHNyVvpB if you didn't catch it earlier
<stintel> I've just created a github repo that shows what I'm trying to do step by step
<stintel> and trying to make it clear that I am trying to do my homework
<stintel> so that hopefully someone who understands this cancerous voodoo wants to help me
<Grommish> stintel: https://godbolt.org/
<Grommish> It'll take C++ and split out PPC ASM
<Grommish> along with a lot of other ones it looks like
<stintel> I don't need that
<stintel> I can do that with gcc -S anyway
<stintel> but it doesn't work at all
<stintel> and so I'm trying to replicate the relevant code step by step
<stintel> to see what's happening and try to understand it
<stintel> but that's apparently impossible
<stintel> I'm getting errors that yield 3 search results in google
<stintel> all 3 irellevant
<stintel> irrelevant?
<stintel> it's probably the most frustrating thing I ever worked on
<Grommish> That looks to just patch configure though
<stintel> yes, I can use #ifdef __ALTIVEC__
<stintel> but musl folks tell me to check __hwcap instead
<stintel> they've been somewhat helpful
<stintel> but they don't seem to care about fixing it themselves
<Grommish> *nod*.. Well, you can get it to work doing it wrong and then make it right.. it's how I end up doing it most of the time
<stintel> yes, but I want qoriq target in master
<stintel> so I can get on with firewall4
<stintel> right now I have to constantly merge several WIP branches together to get something working
<stintel> often forgetting a step
<stintel> it's horrendously slowing down almost anything I'm working on
<stintel> so I figured qoriq is almost ready
<Grommish> haha I'm living that dream between rust, suricata6, libhtp, etc.. I just cherry-pick to a working branch from the others
<stintel> then found out musl hardcodes fsqrt and e5500/e6500 do not support it
<stintel> and while the only software that SIGILLs because of that on my main router is chronyd
<Grommish> then I can always switch back , amend on the correct branch and re-roll a working if it's that far out of sync
<stintel> I can't push it like that
<Grommish> Oof
<stintel> fuck
<stintel> it's >7 AM again
<Grommish> time and fun and all that
<Grommish> Can anyone tell me a target that is ARM but not ARMv7?
<stintel> Grommish: bcm27xx/bcm271x
<Grommish> Thanks!
<stintel> Grommish: welcome
valku has quit [Remote host closed the connection]
rua has quit [Quit: Leaving.]
danitool has joined #openwrt-devel
Tapper1 has quit [Read error: Connection reset by peer]
Tapper has joined #openwrt-devel
Rondom has quit [Remote host closed the connection]
Rondom has joined #openwrt-devel
<f00b4r0> Grommish: that's because your "valid memory registers" really aren't :)
victhor has joined #openwrt-devel
<neggles> are we doing gpio-export or gpio-hog for USB power control pins these days
Borromini has joined #openwrt-devel
rua has joined #openwrt-devel
goliath has joined #openwrt-devel
<dvn> are there docs on writing services?
<dvn> can't find it from searching, but i could just be missing it
<PaulFertser> dvn: you mean procd services?
<dvn> yes
<neggles> it would be super duper nice if someone could take another look at https://github.com/openwrt/openwrt/pull/4517 (ath79, adding device support) now that i've found the USB power control GPIO
<dvn> PaulFertser: thanks
<neggles> it's been sitting around for a while now due to some disagreement over name of LED DTS nodes
<neggles> which does not seem to have come to a conclusion
xdarklight has quit [Quit: ZNC - https://znc.in]
oftc has joined #openwrt-devel
nlowe has joined #openwrt-devel
oftc is now known as Guest7028
<dvn> PaulFertser: what else can i reference for info on procd init scripts? i see things which are not mentioned on the linked page.
<dvn> i.e. in an init script i'm looking at it sets START=99 at the top
<dvn> what does that do?
<dvn> i don't see that referenced in here https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=package/system/procd/files/procd.sh
<PaulFertser> dvn: it just tells that when /etc/rc.d symlink is created it'll have number 99, that is the last one.
<dvn> PaulFertser: thanks. and STOP=10 ?
<PaulFertser> dvn: same about K10
<PaulFertser> dvn: the order of stopping when the system shuts down.
<aparcar[m]> Anyone compiling OpenWrt on a M1 CPU?
<hauke> aparcar[m]: OpeNWrt compiles for me on Ubuntu running on ARM64
swalker has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
dangole has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
dangole has quit [Remote host closed the connection]
minimal has joined #openwrt-devel
dangole has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
isak has quit [Read error: Connection reset by peer]
<aparcar[m]> hauke: I guess it’s the issue to use both Arm64 and macOS at the same time
<aparcar[m]> Currently it fails mentioning some undefined symbols for the architecture
valku has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
<hurricos> neggles: > Third device does not contain calibration data
<hurricos> should I buy one of these devices and try to find how it's being done in OEM?
<hurricos> I've been on the hunt for good OpenWrt 11ac devices for the last couple of months, so I'm interested
<hurricos> Sorry, I went on a tangent. To your original question: don't worry, you're not being ignored, PR review is just backed up right now :^)
<hurricos> Your PR looks very clean. If I can get my hands on some devices I will build and run-test and ACK, in my experience adrian usually focuses on the stuff that multiple people are looking at
<hurricos> or, well, the conversation looks clean.
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Guest7028 is now known as xdarklight
<owrt-snap-builds> Build [#377](https://buildbot.openwrt.org/master/images/#builders/65/builds/377) of `archs38/generic` failed.
<spacewrench> Is there a reason the native GCC version doesn't keep up with the cross-compile GCC version? On my target (Mediatek MT7688 / Onion Omega2+) the cross version is 11.2, native is 7.4.
<spacewrench> Hmm, maybe it's harder to cross-compile GCC...maybe it wants to compile some of itself during bootstrapping, and that's a weird dance in cross-compile land
<spacewrench> I guess I can always compile 11.2 on the OpenWrt board itself.
isak has joined #openwrt-devel
<hurricos> spacewrench: If you need a particular GCC version, pick the right OpenWrt version.
<hurricos> wait, why do you need 7.4?
<hurricos> (do you need 7.4?)
<hurricos> oh
<hurricos> wait I see what you mean; you want GCC to run on the Onion Omega2+?
<hurricos> > compile gcc ... that's not going to happen, genautomata takes 400-500MB.
<hurricos> It took 4 days for my MX60 to compile gcc under a gentoo chroot, of which 2 of those days were about digging deep into swap for genautomata to compile
<hurricos> Are you trying to compile something specific that isn't a package?
<hurricos> Or are you trying to compile a part of your own toolchain?
<hurricos> For the latter, I'd be curious what you're doing. For the former, just use the cross-compiler toolchain by exporting some shell variables and just using the existing toolchain; better, create a package
<hurricos> > 4 days ... MX60 < the device has 512MB of DDR2
isak has quit [Read error: Connection reset by peer]
isak has joined #openwrt-devel
philipp64 has quit [Quit: philipp64]
philipp64 has joined #openwrt-devel
<spacewrench> hurricos: I just wondered why it wasn't the same version between the cross & native environment. Anyway, you're right: I really need to learn to use the cross environment. I tried compiling a simple Boost app on the Omega2+, and it took FOREVER! (I just included one of the Boost header files, and it churned on and on.)
<hurricos> yeah, haint made for compiling :^)
<spacewrench> I compiled linux natively...it took most of a day. And then I couldn't boot it.
<hurricos> I don't really know why OpenWrt provides gcc as a package tbh
<hurricos> it's certainly not made to self-host
<hurricos> if you REALLY need a specific version, I always recommend just getting a gentoo chroot
<hurricos> but like, why
<hurricos> lol
<hurricos> You'd do just as well to run the malta target in qemu
<hurricos> fwiw
<hurricos> though I've never had SMP working, likely because malta's not meant for that
philipp64 has quit [Ping timeout: 480 seconds]
<spacewrench> Actually, I benchmarked qemu...on a good x86 machine, compiling lua in an emulated MIPS was about as fast as compiling it on the real MIPS. A Mac M1 was only a little bit faster. (M1 QEMU for ARM is pretty zippy, though.)
<hurricos> Oh, cool!
<spacewrench> It's amazing that it works as well as it does, but it's too slow to use in a development cycle. I don't even know whether they make regular PCs with MIPS processors.
<spacewrench> I guess they do; and you can probably get time on one on the GCC compile farm. Maybe I should look into that.
<hurricos> They do, Longsoon :^)
<hurricos> but you still need cross compilation, different MIPS variant
nlowe has joined #openwrt-devel
<hurricos> OK, time to try to inject lorem ipsum into the kernel ring buffer
<hurricos> kernel *log* ring buffer
<hurricos> and by inject, I mean compile in. Wow, that's a lot less interesting now that I put it that way.
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<stintel> spacewrench: 7.4 just the version that is in OpenWrt-packages, it's been bothering me too
<hurricos> but why .... would you have GCC on openwrt?
<hurricos> I guess llvm bpf stuff is going the same way :\
<stintel> to play around - I used it on my m300 to do so asm poking
<hurricos> ah
<stintel> sometimes you really need that to figure out stuff about the device / cpu
<hurricos> OK, that's fair ...
<stintel> so I was wondering, since OpenWrt runs on multiple targets with plenty of storage
<stintel> should we make it possible to install toolchain & headers as ipk
<stintel> and I am leaning towards "yes"
<hurricos> imagine my shock when I find my lorem ipsum isn't in my kernel
<hurricos> oh
<hurricos> because I added it to patches-5.4
<hurricos> stintel: I think any extensibility like that *has* to be merged in with official ways to extend the userland
<hurricos> IMO the only reasonable way to replace the OpenWrt userland is in a chroot or, better, container
<hurricos> when I say "replace the userland" I mean all of the work that will be required to make build systems in general happy
<hurricos> I don't believe the toolchain is compiled with a musl-compiled GCC
<hurricos> e.g.
<stintel> just a thought, have not looked at specifics at all
<hurricos> gotcha for sure
pmelange has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
<Grommish> f00b4r0: Probably :D But, it was an easy solve by just ignore PPC for now ;)
<f00b4r0> Grommish: without looking at the code I can't say for sure what's going on, but my bet is that there's been a copy-pasta that mixed ppc instructions with x86 register monikers.
<f00b4r0> Grommish: on ppc registers don't have names like rax and rdx. That's an x86ism
<Grommish> f00b4r0: According to the rust devs, PPC under LLVM has been a pain and sore point for a whole
<Grommish> and it's certainly beyond my skills, so I'll ignore it until I have nothing else to do heeh
<Borromini> Grommish: you got 5.10 running fine on your Octeon device(s) don't you?
<Grommish> Borromini: Not anymore.. 5.10/5.15 had massive kernel memleaks I was trying to chase down
<Grommish> I've not been back in Octeon land to check the patches I've been shown to see if I might help or not
<f00b4r0> and does not match your pasted output
<Borromini> Grommish: ok. pesky thing is my ER4 is in production so a bit hard to swap out/use as a guinea pig
<Grommish> Right I've got multple Shields, both are running 5.4 without issue
<f00b4r0> (the ppc cache invalidation routine at the end, that is)
<Borromini> Grommish: yeah no issues there :)
<Grommish> but rust-lang came back around (apparently people are like.. wiating for it?)
<Grommish> and I was tired of watching RAM crawl by the second
<f00b4r0> Grommish: anyway, I'm not really interested in rust tbh, but the file I linked is correct and should build fine. What you have is probably garbage, judging by the error messages. I hope this helps :)
<Grommish> f00b4r0: From what I understand, rust-lang pulls LLVM from github as a sub-module, but I'll go look. I do appreciate the help!
<f00b4r0> yw
<hurricos> setenv ipaddr 10.0.7.2; setenv serverip 10.0.7.1
<hurricos> setenv ipaddr 10.0.7.2; setenv serverip 10.0.7.1
<hurricos> ugh.
<hurricos> Sorry.
<enyc> hurricos: hehe, easily done!
<hurricos> I had another thought RE failed P1020 init
<hurricos> There's a fixup for L2 cache that goes on in stock kernel but not in mine
<hurricos> I don't know if the fixup is actually necessary, since HiveAP 330 doesn't require it. But
<hurricos> the AP3825i runs at a different CPU frequency ...
<hurricos> and the L2 cache not syncing out to RAM is consistent with not seeing logs in RAM
<hurricos> the vendor's kernel dtb doesn't have this issue, so I might see if I can diff there
jschwart_ has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
jschwart has quit [Remote host closed the connection]
rua has joined #openwrt-devel
jschwart_ is now known as jschwart
<hurricos> OK, nice, finally found the device tree. Even better, it's very populated and confirms that by default CPU2 is actually disabled
<hurricos> enabled only later by the kernel directly, which isn't consistent with what my device-tree says. So let's compile this
<hurricos> tfw still no boot :^)
<hurricos> Huh ....
<hurricos> the new device tree ... doesn't ... reserve ...
<hurricos> memory for itself
<hurricos> so it ...
<hurricos> gets corrupted by the kernel ...
<hurricos> ?
<hurricos> No, it's valid
<stintel> hurricos: I see you have the same knack for going down rabbit holes as me :D
<hurricos> I mean, the problem isn't thta deep
<hurricos> the board is a known board using known hardware
<hurricos> all I have to do is get it to boot
<hurricos> there's something u-boot isn't doing here that it is on the other identical platform(s)
<hurricos> it even boots and is maintained by an existing vendor in the modern day
<stintel> meanwhile I ordered shoes to make me feel better 😂
<Habbie> haha
<f00b4r0> lol
<hurricos> :D
<hurricos> the one oddest thing ...
<hurricos> this board's original device tree uses fsl,p2020-guts instead of fsl,p1020-guts.
<hurricos> could just have been a 2014 tihng
<stintel> I have not driven a motorcycle in 5 years or so :(
<stintel> they're in a similar theme though :)
<stintel> that reminds me I have to get new winter tyres, have to look extra hard for the version with W speed index
<stintel> having to set a warning on 240kph is annoying
<hurricos> > 240kph < o_O
<stintel> car can do > 300kph in theory
<hurricos> oh
<hurricos> car, OK :D
<PaulFertser> You want to go in winter faster than 240 km/h?
<PaulFertser> :-O
<stintel> PaulFertser: winter tyres are kind of obligatory in several countries starting from November
<stintel> doesn't mean it's going to be freezing cold when I'm on the autobahn
<stintel> so yeah I've hit that 240 kph warning numerous times with the winter tyres
<PaulFertser> stintel: in Germany or somewhere closer to where you live? ;)
<stintel> autobahn == germany
<stintel> and I'll most likely be driving to Belgium for the holidays, so .. :)
<PaulFertser> Awesome, have a nice trip.
<stintel> now there's another problem :P the car is normally limited on 250, but mine isn't, unfortunately the speedo stops at 260, and I doubt I can set the warning any higher than that :P
<stintel> W = 270 and that's the highest I could find for the winter tyres
<stintel> so... I should write an Android app that warns me at 270 on GPS? :P
<stintel> PaulFertser: thanks
<stintel> hadn't done autobahn for > 1y due to covid
<stintel> had to get used to >240 again
<stintel> now the engine has new conrod bearings and improved oil sump and oil pump, it's time to get a 3rd set of wheels for (semi-)slicks and improve my track times
isak has quit [Quit: leaving]
Borromini has left #openwrt-devel [#openwrt-devel]
isak has joined #openwrt-devel
nlowe has joined #openwrt-devel
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nlowe has joined #openwrt-devel
Acinonyx has joined #openwrt-devel
<hurricos> to avoid further self-harm I am also going to buy a AP3710i and confirm that it still boots on openwrt master
<hurricos> unless @blocktrron can confirm as much, of course
Acinonyx_ has quit [Ping timeout: 480 seconds]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<jow> nbd: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=d0b33833aed0226a7fba679699ccae0f8af197a4
<jow> nbd: won't that break a lot of existing wireless configuration when upgrading from 21.02.1 to 21.02.2 ?
<nbd> no, it's backwards compatible
<jow> ok
<nbd> compatibility is implemented in netifd-wireless.sh
<nbd> it translates the existing option into the band option
<jow> allright
nlowe has joined #openwrt-devel
nlowe has quit [Read error: Connection reset by peer]
<neggles> hurricos: I suspect generating the 3rd chain caldata would require some rather fancy equipment - at least a decent 6GHz spectrum analyser? - but if there *is* a relatively simple way to unlock chain 3 that would be pretty great; I’d offer to send you one but shipping from here is probably more expensive than buying one locally now they’re effectively EOL/EOS with sophos
<hurricos> neggles: I'm figuring that if the oem device has caldata, we have caldata.
<hurricos> that said
<hurricos> it's QCA988X, yes?
<neggles> as best I can tell it’s got two chains worth of caldata despite being 3 chains of device
<hurricos> so what does the vendor fw do to provide caldata to its driver/
<hurricos> I'm guessing copies one to the other and calls it a day?
<hurricos> Figuring that bit out is what I'd like to do
<neggles> it’s a QCA9880 for 5GHz yeah, the caldata bin is in the ART partition
<hurricos> can you dump me the bin and shoot me a board photo?
<neggles> sure
<hurricos> paste.c-net.org is fine
<hurricos> or wherever else
<neggles> both ART entries only have caldata for 2x2, it seems - I couldn’t find anywhere in the OEM firmware that did anything different with an AP100 vs AP55
<hurricos> you would also want to check the rootfs
<hurricos> I see what you mean by third chain
<hurricos> vendor might not use the third chain
<hurricos> :P
<hurricos> board photo
<hurricos> would tell us for sure whether they intended not to*
fda has quit [Ping timeout: 480 seconds]
fda has joined #openwrt-devel
Tapper has quit [Read error: Connection reset by peer]
<neggles> https://i.imgur.com/Z3jmlwy.jpg I can get a better pic but this is an AP55C, the “2x2” model
Tapper has joined #openwrt-devel
<neggles> they also sold an AP100C, which is 3x3; both use the same kernel and rootfs
<neggles> there’s an unpopulated micro-USB OTG on the C models too; the Edimax they’re based off has it present https://www.edimax.com/edimax/merchandise/merchandise_detail/data/edimax/global/edimax_pro_indoor_access_points_ac1750/cap1750/
<neggles> then there’s the non-C models which are the same hardware, but with u.fl connectors for the 5GHz antennas, designed for wall mounting
<neggles> (so a different PCB layout)
<neggles> there’s a lot of names for this hardware 😝
valku has quit [Quit: valku]
nlowe has joined #openwrt-devel
dangole has quit [Ping timeout: 480 seconds]
nlowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<hurricos> I have more! I have more!
nlowe has joined #openwrt-devel
<hurricos> Uh oh! It points to fdt_offset_ptr
<hurricos> and fdt_next_node ._.
<hurricos> Let's go back to the cleaner fdt and try again
Luke-Jr has quit [Ping timeout: 480 seconds]
hurricos has quit []
hurricos has joined #openwrt-devel
Luke-Jr has joined #openwrt-devel
spacewrench_ has joined #openwrt-devel
spacewrench has quit [Read error: Connection reset by peer]
<neggles> hurricos: here's a full OEM flash dump split into the various partitions & named appropriately: https://paste.c-net.org/CheckingPelvis - this is an AP55
<hurricos> thank you neggles; I'm going to keep hacking on AP3825i, so I probably won't have time to look tonight, but ...
<hurricos> definitely push the ART from that fw through apeethmgr
<neggles> hurricos: no problem :) yeah i'm about to heh
<hurricos> atheepmgr*
<neggles> it has the 2.4GHz fw @ 0x1000 and the 5GHz fw @ 0x5000
<neggles> which is fairly standard it seems
<hurricos> very, that's identical to how I've always seen 'em
kenny has joined #openwrt-devel
spacewrench_ has quit [Ping timeout: 480 seconds]
spacewrench has joined #openwrt-devel
<neggles> hurricos: oooooooo.
<hurricos> ?
<neggles> heh, sorry, jumped the gun a bit, should've made the gist first
<neggles> hurricos: here https://gist.github.com/neg2led/34827bea2fe57b6a2633f6f7ddac78dc - im not sure how to interpret this exactly, but it looks like there might be caldata for all 3 chains, but the Tx/Rx masks are set to 0b011?