**** BEGIN LOGGING AT Sat Jun 13 02:59:57 2020 Jun 13 04:04:46 nick[m], https://openwrt.org/docs/techref/targets/ipq40xx "show devices with this target" , also wikidevi mirrors should have SoC search / search for radio variants even Jun 13 04:10:53 jow: zstd is available under tools/ Jun 13 04:12:10 with https://downloads.openwrt.org/snapshots/faillogs/powerpc_8540/packages/zstd/compile.txt , it sounds like the SDK should be updated/fixed Jun 13 04:15:13 CentOS 7 compile succeeded Jun 13 04:35:06 build #451 of lantiq/ase is complete: Failure [failed kmodconfig] Build details are at http://buildbot.openwrt.org/master/images/builders/lantiq%2Fase/builds/451 blamelist: Adrian Schmutzler , J. Scott Heppler Jun 13 06:08:28 I have a bit of a weird problem. im trying to boot openwrt on a router that's identical to the one i'm already porting to (it just has a bigger flash chip) and now im getting "ath10k_pci: Refused to change power state, currently in D3" when trying to initialize the 2 wifi cards. the oem dts files for the two routers are identical to each other (just a different board name) and wifi works on the one I was originally working with, Jun 13 06:08:28 but it just doesn't on this other one Jun 13 06:43:33 forgot to mention that the router is ipq8065 based and the two wifi cards are QCA9984 cards Jun 13 07:20:40 Hello! The docs for opkg say that using --dest is "more a curiosity than actually useful" https://openwrt.org/docs/guide-user/additional-software/opkg#non-standard_installation_destinations I've implemented it and gotten quite a lot of the kinks worked out of it. I need to modify opkg now dynamically configure the PATH and LD_LIBRARY_PATH (with musl, it's actually /etc/ld-musl-.path, e.g. /etc/ld-musl-mipsel-sf.path). Jun 13 07:21:09 *now TO dynamically configure Jun 13 07:21:32 I dunno if anybody tries to use this or not Jun 13 07:23:47 lmore377_: oooh! I think I know this one! Jun 13 07:24:36 dasan: oh? Jun 13 07:24:52 lmore377_: Depending upon the flash chip, Linux driver and u-boot, the driver might need to do reset the spiflash unit before rebooting Jun 13 07:25:01 Wait, which cpu is this? Jun 13 07:25:35 ipq8065 Jun 13 07:25:54 oh crap, I'm sorry. Maybe I'm wrong Jun 13 07:26:11 We ran into something similar when we initially got 32M flash chips, but it was mt7620a Jun 13 07:26:12 and also it uses nand flash so it's problably not that Jun 13 07:27:02 thanks for trying to help though! Jun 13 07:27:39 sure! now I'm curious :) Jun 13 07:32:22 lmore377_, afaik ther is some advancedd debugging documented for ath10k-ct that probably also dumps internal firmware state so you could compare this - and load the driver with debugging flags in kernel cmdline / modprobe Jun 13 07:33:14 https://www.candelatech.com/ath10k-ug.php for example Jun 13 07:36:01 dansan: I'm often using "-d ram" and I just manually specify LD_LIBRARY_PATH when starting programs installed like that. Jun 13 07:38:15 plntyk: let me try rebuilding with that module because the standard ath10k module doesn't print anything extra Jun 13 07:48:23 lmore377_: (refused to change power state) wild guessing: somehow the cards are not fully powered and you need a GPIO action to turn the power on. I'd check /sys/kernel/debug/gpio on vendor's firmware (if it has debugfs enabled). Jun 13 07:52:32 PaulFertser: We've got a /mmc that we're installing stuff into. I think I just need a little tweaking to manage the paths. For example, my /etc/ld-musl-mipsel-sf.path contains the string "/mmc/usr/lib:/usr/lib:/mmc/lib:/lib" Jun 13 07:52:50 But it's helpful to know that people are using it for ram Jun 13 07:53:50 dansan: tweaking default library path plus PATH should be enough; is it proving to be problematic for you? Jun 13 07:56:05 PaulFertser: Well that is the proper way to do it with musl. By default, LD_LIBRARY_PATH is not defined. I suppose it would be defined in /etc/profile though Jun 13 07:56:33 dansan: it shouldn't be defined normally, I just add it manually to the running env for one-off actions. Jun 13 07:57:21 Yeah, but we're installing daemons on the mmc card. Some monsters made me port OpenJDK to MIPS and they run a java-based daemon with it. Jun 13 07:57:39 dansan: so isn't modifying /etc/ld-musl-mipsel-sf.path enough for you? Jun 13 07:57:41 So that's why I added that Jun 13 07:58:31 Yeah, it works, but I'm doing it statically. I want to post my patches for making opkg --dest work (as a normal installation location), so I think I need to do that dynamically in opkg when you call --add-dest Jun 13 07:59:18 I had to make a lot of changes to the build system to make it behave correctly with init scripts and such. Jun 13 08:01:36 And then there are also rogue packages that presume that the install root is /, but that's on a case-by-case basis. Jun 13 08:33:03 PaulFertser: a lot of the gpios between stock and openwrt are exact opposites so thats ineresting Jun 13 08:33:43 lmore377_: you can try changing them manually one by one with /sys/class/gpio Jun 13 08:34:24 yeah im about to try that. ill also check if anyhings different on the working router Jun 13 08:34:32 we still don't support ext4fs overlay, do we? Jun 13 08:37:55 dwmw2_gone: on x86 and raspberry pi and problably others you can choose between ext4 and squashfs Jun 13 08:38:09 looking at Banana Pi R2 Jun 13 08:38:23 it seems to have a tmpfs overlay over its ext4 root, which is obviously ephemeral Jun 13 08:38:31 trying to work out how to set up the image so that it has a rootfs_data partition Jun 13 08:38:38 do I just make another ext4 with that label ? Jun 13 08:46:54 Sat Jun 13 01:22:25 2020 daemon.info mount_root: rootdisk overlay filesystem has not been formatted yet Jun 13 08:46:54 Sat Jun 13 01:22:25 2020 daemon.notice procd: /etc/rc.d/S95done: sh: mkfs.f2fs: not found Jun 13 08:46:58 oh that would do it :)P Jun 13 08:49:32 * ldir keeps wondering where dwmw2 has gone Jun 13 08:49:59 astral Jun 13 08:50:03 good question Jun 13 08:52:30 hmm. How is it that I can never make sense of the memory numbers on Luci's overview page? No sum seems to ever amount to 100% Jun 13 08:53:32 total available: 41%; used: 49%; buffered: 4%; cached: 16%. What gives? Jun 13 08:56:11 PaulFertser: so uh it seems like the two boards have different gpio numberings. i tried to turn on what's supposed to be the red led and nothing happened Jun 13 08:56:35 also the different modes for the blue led were wrong Jun 13 08:56:37 lmore377_: probably different SoC even? :) Jun 13 08:58:26 i swear if its a different soc im going to hunt down whoever designed these things. 40 device tree files came out of the original firmware. ill check the soc on the other router when everyone's asleep Jun 13 09:05:45 actually 20, just accidentaly counted the dts files i generated from the dtbs Jun 13 09:42:17 PaulFertser: same soc, slightly different board layout. Jun 13 09:45:11 lmore377: time to re-RE GPIOs then, it seems Jun 13 09:45:43 plntyk: thank u! Jun 13 09:49:19 before i go down that rabbit hole, is it possible that a script somewhere in the oem firmware is just ignoring whatever the device tree says and is activating it with /sys/class/gpio? the only difference between the device trees for the two revisions is the model and a line that has "usb3_dev_reset_quirk" in it Jun 13 09:53:25 lmore377: of course that's possible Jun 13 09:54:00 lmore377: oem firmwares are often a huge pile of BS. Jun 13 09:55:07 PaulFertser: the fact that it's basically stock openwrt 15.05.1 with some extra packages made me think this would be easier Jun 13 10:18:19 no luck on the wifi but i think i discovered an unreleased model apparently? there's a script to check the board model based off a gpio pin and if its high that means the model is a rt4240w which i cant find any info about. also if i set that gpio pin to out and turn it on it turns on a hidden led Jun 13 11:09:14 zorun: can i ask you a babeld question? Jun 13 11:47:33 russell--: interested in the babeld question, say it Jun 13 11:48:56 guifipedro: i'm just starting to play with it ... like as of 10 minutes ago. have been running olsrd for a long time, just looking for some pointers on how to get started Jun 13 11:49:59 like how to export ipv4 prefixes for local networks and not advertise private networks Jun 13 11:53:59 we would like to switch from bmx6 to babel (from bird v2) Jun 13 11:53:59 which would need development/integration, we will see Jun 13 12:15:13 russell--: sure! not much time today, but you can start with https://openwrt.org/docs/guide-user/services/babeld Jun 13 12:15:42 over the years I added many examples Jun 13 12:17:23 I'm interested in feedback on this doc (for instance, I don't remember if using logical interface names still works) Jun 13 12:20:59 zorun: for us they still work in 19.07, using a wireguard interface and type tunnel Jun 13 13:48:12 build #164 of bcm27xx/bcm2711 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/bcm27xx%2Fbcm2711/builds/164 Jun 13 13:58:28 adrianschmutzler: ok, thanks, good to know! Jun 13 15:30:40 build #452 of lantiq/ase is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/lantiq%2Fase/builds/452 Jun 13 15:37:32 * zorun has been chasing the rpcd heisenbug today Jun 13 15:37:37 https://bugs.openwrt.org/index.php?do=details&task_id=3177 Jun 13 15:38:15 libubox is currently the prime suspect... Jun 13 16:00:15 hmm.. are you aware of any problems setting bitrates in ath10k / gl.inet devices (using the vendor fork firmware). Setting bitrate on a device with a station-mac (4-address modes) for 5 GHz results in crashes / restarts Jun 13 16:11:03 * enyc meows Jun 13 16:11:18 irc chanel here also outdated r.e. release Jun 13 16:59:02 yanosz: gl.inet uses an ancient OpenWrt fork, setting bitrates works fine in 19.07 (apart on ipq40xx on 2.4GHz) Jun 13 17:07:19 I am too stupid to set a static tty name using udev rules Jun 13 17:07:32 creating a symlink works but i want to rename it instead Jun 13 17:30:23 fuck linksys, this EQ8300 does not have torx but the security bit torx Jun 13 17:30:30 the one with the small pin in the center Jun 13 17:30:38 luckily I am well equiped Jun 13 17:38:58 ffs now there are no serial pin headers and I need to whip out the solder iron Jun 13 17:40:11 blogic: there are 5.edges security torx, too Jun 13 17:40:27 Normal have 6 Jun 13 17:40:32 rubberduck: they will burn in hell Jun 13 17:41:06 Do you know the dbox2? This one had those 5-security-torx Jun 13 17:41:21 I dont watch TV so no Jun 13 17:41:27 ;) Jun 13 17:41:31 Nintendo uses some special triwing Jun 13 17:41:53 Or inverted torx for game modules Jun 13 17:41:55 there is that weird triple strip screw where the slots are angled Jun 13 17:42:19 so you can screw it shut but when you turn ccw the driver slips out the slope Jun 13 17:53:27 zorun: what is the babeld equivalent of olsrd's HNAs? i assume it's a filter, but i haven't grok'd all the filter options Jun 13 17:55:19 blogic: dremel can turn anything to a flathead slot screw ;-) Jun 13 17:57:17 heh Jun 13 17:57:28 that or make it no longer a issie Jun 13 17:57:38 issue* Jun 13 18:06:25 zorun, this seems to do what i think i want: uci add babeld filter ; uci set babeld.@filter[-1].type=redistribute ; uci set babeld.@filter[-1].ip=10.11.27.0/27 ; uci commit babeld Jun 13 18:13:08 russell--: I have no idea what a HNA is Jun 13 18:20:58 zorun: wikipedia says it stands for "Host and network association", basically advertising local route Jun 13 18:21:22 https://en.wikipedia.org/wiki/Optimized_Link_State_Routing_Protocol Jun 13 18:46:12 russell--: ok, I still don't understand what this is. OLSR seems to like custom acronyms Jun 13 18:46:21 russell--: maybe explain what you want to achieve in English? ;) Jun 13 18:46:35 if it's just redistributing local IP addresses, babeld does that by default Jun 13 18:47:59 if it's redistributing local routes (e.g. static routes), then yes, your approach is good Jun 13 18:48:58 see https://openwrt.org/docs/guide-user/services/babeld#example_filtering_configuration Jun 13 18:52:09 blocktrron: acient .. .it's 18.06 ... that's not corrent but not chaos calmer ;-). In the meantime, I noticed, that potential problems in 19.03 were actually cased by a broken powerbank and a broken network calble causing some hassle. Jun 13 18:52:09 the missing hostname functionality causing errors in openvswitch can 1st: be fixed by having a small bash script, and 2nd: won't cause any acutal harm... it's more like openvswitch init script is broken, requiring some manual ovs-ctl start --system-id=42 on 19.3 Jun 13 18:52:29 aka: I'm making progress, but things are unstable no matter what ;) Jun 13 18:54:09 the built up is not finished yet, but I'm curious, if there's a problem with bitrates in 18.06, in case I need to downgrade due to other protblems Jun 13 19:34:59 yanosz: okay, then this might have changed. My VIXMINI (which is a quite recent device) came with 15.05 pre-installed Jun 14 00:59:41 PaulFertser: if i have the original wifi card in, lspci -v shows "!!! Unknown header type 7f" on everything, but if I put in a random laptop wifi card lspci -v has a lot more information and everything looks populated correctly. I tried swiching the cards between the two routers and it's the same result Jun 14 01:16:14 Hello all! I guess I haven't dug too deeply into how the toolchain is built yet, but I've done gcc dev work in the past. There's a gcc-8.3.0, gcc-8.3.0-final and gcc-8.3.0-initial. I presume the "initial" is just the bootstrap compiler (or stage1), but I don't understand "final" and non-suffixed one. What is the difference between the two? And does buildroot still let gcc build two final versions to compare them? Jun 14 01:47:13 dansan: look in toolchain/Makefile Jun 14 02:38:39 russell-- thankee!! Jun 14 02:48:28 russell-- Well this still leaves a few things unanswered. It looks like I'll need to digest target/toolchain/Config.in and target/toolchain/Makefile **** ENDING LOGGING AT Sun Jun 14 02:59:57 2020