00:40 <JustFly> Hi! I'd like to add a heavily modified U-Boot package for a new target, but the compilation fails with MUSL, because it seems to not support thumb instructions. It says "sorry, unimplemented: Thumb-1 'hard-float' VFP ABI"... I've tried to build it without the thumb flags, but the startup (assembler) code offsets then do not fit anymore. Is there a convenient way in OpenWrt to use GLIBC instead for compiling a single packag

01:00 <Mangix> JustFly: glibc can be defaulted for certain targets. this is done with ARC

01:14 <slh> but the fact that this -only- target (arc) was different, was not appreciated at all - and a large part in the decision to retire it (there were more reasons, but the glibc dependency has always be a big point of contention)

01:57 <djfe> rmilecki: obviously OpenWrt could do different, but PSK has been cut from 6GHz like anything else legacy. since it's a new frequency spectrum and they are starting fresh: why keep old stuff? so most Clients won't accept PSK2 anyways even if you would find a way to use it on 6GHz which I doubt. SAE is the way to go now

01:59 <djfe> on 2.4GHz and 5GHz you need to keep providing PSK2 for legacy reasons but 6GHz is only supported by new clients, new enough to know wpa3

02:01 <slh> not quite, the common setup for most is to have one single ESSID for 2.4+5+6 GHz, clients don't like it that much if the access credentials/ encryption settings differ between BSSIDs with the same ESSID

02:03 <slh> that effectively means WPA3SAE everywhere, which few can do (IoT crap on 2.4 GHz, slightly older gear on 5 GHz) or WPA2PSK/ CCMP everywhere, mixed mode isn't accepted by many clients and doesn't give you much beyond WPA2PSK anyways

02:05 <slh> (or two concurrent ESSID, one with WPA2PSK on 2.4+5 GHz and one WPA3SAE ESSID on 2.4+5+6 GHz)

03:14 <KGB-2> https://tests.reproducible-builds.org/openwrt/openwrt_sunxi.html has been updated. (0% images and 100.0% packages reproducible in our current test framework.)

08:15 <PaulFertser> JustFly: but what do you need floating point in u-boot for?

08:28 <JustFly> As far as I understood this thumb instruction option decreases binary size by using 16 bit instructions. The target uses bootstrap code for starting U-Boot. If the size of that bootstrap code increases some hardcoded offsets do not fit anymore. I will try turning off hardwaer floating point instructions and see if that fixes the issue...

09:26 <PaulFertser> JustFly: Thumb instructions help to save space, right, but the error you get is because you're using hardfp ABI, but if you do not need FP at all, why not use softfp ABI instead?

09:26 <PaulFertser> -mfloat-abi=soft

09:27 <PaulFertser> Also, what core is that, is it really restricted to very old Thumb-1 instructions (for the Thumb mode)?

09:29 <PaulFertser> Cortex-A9 supports Thumb-2

09:36 <PaulFertser> JustFly: I have impression the problem are your compile flags, not MUSL or ABI.

10:06 <djfe> slh: I guess different encryption settings are going to become more common then. Hopefully 6GHz device don't suffer from this issue, then it should be fine.

10:07 <djfe> else: roll out 6GHz with a different ESSID. the spec leaves you no choice

10:08 <djfe> not sure if I would roll out the new ESSID for 2.4 and 5GHz as well but it could be of use :)

10:15 <PaulFertser> djfe: so is it established 6 GHz can't be used with mandatory 802.11w but RSN/WPA2?

10:15 <djfe> what is rsn?

10:15 <djfe> fairly certain

10:15 <djfe> why would you keep legacy "shit" for new clients?

10:16 <PaulFertser> djfe: WPA2 isn't that shitty

10:16 <PaulFertser> djfe: and with 802.11w required it's all encrypted so what's the problem?

10:16 <djfe> they dropped everything old and prohibited everyone from deciding against useful configs

10:17 <djfe> WPA3 is an improvement and WiFi alliance wanted to enforce it's usage on new spectrum

10:17 <djfe> there's no reason to stick with wpa2 on new frequencies in their mind

10:17 <djfe> I can honestly relate and see why.

10:18 <djfe> if this causes issues with shared ESSIDs in practice that would be a shame. But I haven't tested this. Have you?

10:18 <djfe> If this causes issues, then it's the fault of the new clients that support 6GHz and still receive updates. this decision doesn't affect legacy clients at all

10:20 <djfe> also it was a good decision to drop WEP and WPA (TKIP) once and for all

10:20 <djfe> or rather: not to keep them

10:21 <PaulFertser> Here's the list of specific requirements from hostapd maintainer https://w1.fi/cgit/hostap/tree/src/ap/ap_config.c#n1202

10:23 <djfe> also: WPA3 (SAE) has the advantage of perfect forward secrecy, which is already common on the web https://www.thesslstore.com/blog/perfect-forward-secrecy-explained/

10:47 <JustFly> PaulFertser: It's a Cortex A7. The flags of the two commands are identical.

10:47 <JustFly> Original command: armv5-eabi--glibc--stable/bin/armv5-glibc-linux-gcc -Os -Wall -g -nostdlib -fno-builtin -Iinclude -march=armv5te -mthumb -mthumb-interwork -fno-pie -fno-PIE -fno-pic -fno-PIC -fno-partial-inlining -fno-jump-tables -fdata-sections -ffunction-sections -DCSIM_ASIC -DNO_QUICK_SIM -DSDRAM0_SIZE_ -DSDRAM1_SIZE_ -DDRAM_INIT=1 -c -o arch/arm/start.o arch/arm/start.S

10:48 <JustFly> My command: arm-openwrt-linux-muslgnueabi-gcc -Os -Wall -g -nostdlib -fno-builtin -Iinclude -march=armv5te -mthumb -mthumb-interwork -fno-pie -fno-PIE -fno-pic -fno-PIC -fno-partial-inlining -fno-jump-tables -fdata-sections -ffunction-sections -DCSIM_ASIC -DNO_QUICK_SIM -DSDRAM0_SIZE_ -DSDRAM1_SIZE_ -DDRAM_INIT=1 -c -o arch/arm/start.o arch/arm/start.S

10:48 <JustFly> When adding -mfloat-abi=soft the build succeeds, but I get a whole lot of these warnings: warning: array subscript 'const struct romvsr[0]' is partly outside array bounds of 'void[4]'

10:48 <JustFly> I'll continue with the other missing stuff for board support and then check if the device boots. Thanks for the help so far.

10:59 <PaulFertser> JustFly: why are you using ARMv5 instruction set for Cortex-A7?!

11:00 <PaulFertser> It supports ARMv7-A and of course Thumb-2 as well.

11:01 <robimarko> nbd: Whats the deal with: https://github.com/robimarko/openwrt/blob/master/target/linux/generic/pending-6.1/203-kallsyms_uncompressed.patch

11:02 <robimarko> I was testing 6.6 and its completely breaking booting, causing kallsyms to throw a BUG() and then hang

11:03 <nbd> robimarko: the point of that patch is to reduce lzma/xz compressed kernel size

11:05 <nbd> i think we should measure how much of a difference it makes these days

11:05 <nbd> and if the size saving is small, we can just drop it

11:06 <robimarko> Well, my issue with it is that its being unconditionally enabled via: https://github.com/robimarko/openwrt/blob/master/include/kernel-defaults.mk#L109

11:06 <robimarko> So, seeing that its disabled in config-x.y is misleading

12:05 <robimarko> nbd: Well, it seems that its saving around 14k on ath79

12:06 <nbd> if you can give me your current tree with 6.6 and the .config you're using, i can help with debugging that patch

12:07 <nbd> please also send me the location of the BUG() that triggers and i'll take a look

13:25 <Piraty> i remember readin an article about a router that required very little modification to make it capable of being operated with via poe , does anyone have the reference?

13:29 <djfe> Piraty: what makes you think there's just one?

13:30 <Piraty> i don't ;) i just remember i read a write-up some weeks ago and i didn't bookmark it

13:31 <djfe> here?

13:38 <Piraty> no

14:12 <robimarko> Oh, musl finally has a new release

14:16 <stintel> cool

23:30 <KGB-2> https://tests.reproducible-builds.org/openwrt/openwrt_omap.html has been updated. (11.1% images and 100.0% packages reproducible in our current test framework.)