**** BEGIN LOGGING AT Fri May 07 02:59:57 2021 May 07 03:11:15 There appears to be patches for this chip id (or what I presume the number to match the chip id is), https://github.com/openwrt/openwrt/search?p=1&q=63268. I'm currently trying to figure out if the chip id is just the cpu, or if it's the board + peripherals. If the latter, then potentially an existing image will have enough support for it to work. May 07 03:11:16 I just am trying to be very careful as I do not know how to use jtag and don't know if jtag or serial connection can fix a brick by flashing the wrong image (or one not fully supported). May 07 03:15:37 Okay, interesting, so my chip id is mentioned as being supported in https://openwrt.org/docs/techref/hardware/soc/soc.broadcom.bcm63xx?s[]=bcm63168. I can try and see if any premade images will get me to the point of tftp accepting the image and asking me to confirming the flash. May 07 03:33:31 Apparently I have a total of 123 Mb of RAM on my router. That's from `cat /proc/meminfo`. I'm not yet sure how to get info about the ram chip itself to confirm that. May 07 06:32:03 jow: after reading your netifd & UCI isssues comments from yesterday I start to understand the problem May 07 06:32:09 thanks for taking a moment and describing that May 07 07:22:19 jow: shit, i just checked netifd sources for the first time May 07 07:22:45 f*k me May 07 07:49:10 fuck, a single uci section (config interface 'lan'; option type 'bridge') is parsed by code in interface.c, bridge.c and (interface-ip.c XOR proto.c) May 07 07:49:22 jow: how is anyone supposed to follow that May 07 09:12:18 * enyc wonders if the ath10k CT firmware issues with 21.02 SNAPSHOT etc. are improved in 21.02.0-rc1 May 07 09:12:40 * enyc updates 2 ancillary and test lantiq xrx200 BT-HHv5a to 21.02.0-rc1 in any case May 07 09:21:52 Hi! I wanted to ask if anyone knows how to get the device tree directory when the kernel isn't compiled with the option to show it? Specifically, my `/sys/firmware` directory is empty and the `/proc/device-tree` symlink doesn't exist. May 07 09:23:26 Would it be somewhere in rootfs or the vmlinux file? I'm currently poking around the filesystem to find something that contains the hardware info for my modem. May 07 09:24:54 AlexisEvelyn: it depends on platform May 07 09:25:09 AlexisEvelyn: BCM4908 has .dtb on "bootfs" MTD JFFS2 partition May 07 09:25:27 AlexisEvelyn: usually bootloader should contain that somewhere May 07 09:25:35 AlexisEvelyn: but format is platform / bootloader specific May 07 09:26:24 I believe it's a brcm63xx (I'm still working out if this is correct). The CPU is Broadcom4350 v8.0. May 07 09:27:09 Is the bootfs also named rootfs? May 07 09:27:09 ``` May 07 09:27:10 cat /proc/mtd May 07 09:27:10 dev: size erasesize name May 07 09:27:11 mtd0: 03d60000 00020000 "rootfs" May 07 09:27:11 mtd1: 03d60000 00020000 "rootfs_update" May 07 09:27:12 mtd2: 00400000 00020000 "data" May 07 09:27:12 mtd3: 00020000 00020000 "nvram" May 07 09:27:13 ``` May 07 09:30:33 Is it possible that the kernel would be using DT but /sys/firmware not present? I doubt that. I'd say it means the kernel is loaded without DT, using hardcoded values from machine file or something like that. May 07 09:30:47 BTW, I've recently learnt that ACPI on ARM servers is a thing :/ May 07 09:31:52 It's possible that the kernel's not using a DT. I just have no idea how I would find the values and then be able to translate it to something that OpenWRT could use May 07 09:32:22 AlexisEvelyn: you're supposed to have the kernel sources and then you'd know which hardware is present and what addresses and options are used from the machine file then. May 07 09:32:48 AlexisEvelyn: a great deal of the data can be extracted by following symlinks in /sys/bus/i2c , spi etc. May 07 09:33:07 And of course looking at dmesg. May 07 09:33:57 AlexisEvelyn: it's quite often people are writing DT for OpenWrt basically from scratch, using existing ports for similar devices for inspiration. May 07 09:34:32 I hope I can get the data I need. Thankfully the `/sys/bus` directory and it's subdirectories exist (also dmesg). I know Sagem doesn't release the sources for anything except for the 1704N (as I already checked the website). May 07 09:35:27 AlexisEvelyn: probably you can ask them in an official letter. May 07 09:36:06 True. I'll have to write a letter to ask for the sources. Thanks! May 07 09:37:04 In the meantime, I'm currently going to be looking through `/sys/bus` and I've already copied the dmesg log to my laptop. May 07 09:39:54 I definitely have to figure out how one is supposed to write a letter for a device sold to an ISP as I don't know the generic board's name. May 07 09:46:48 Would `/sys/devices` be able to be used to help make a device tree? May 07 09:59:43 If you are interested, this is where Sagem would normally release their source code. I don't remember how I originally found this site, but since I wrote a paper on the security of the 1704 modem a few years ago, I had the link in the paper. http://opensource.sagemcom.com/index.php?id=42&L=0 May 07 11:09:49 blogic: https://pastebin.com/raw/ZRDdK3yb May 07 11:10:11 bootlog from the RB912 patches which arrived yesterday on patchwork May 07 14:06:51 jow: please have a look at https://github.com/openwrt/openwrt/pull/4138 busybox/nslookup_lede.c might be obsolete May 07 14:54:58 nbd: netifd doesn't create bridge if there is no interface for it May 07 14:55:05 nbd: can you look into that? May 07 14:55:23 config device; option type 'bridge'; option name 'test'; list ifname 'lan1'; list ifname 'lan2' May 07 14:56:02 if there happens to be interface for that bridge, then bridge gets created thanks for the following code path: May 07 14:56:22 DEV_EVENT_ADD -> interface_main_dev_cb -> interface_set_available() -> interface_set_up() -> device_claim() -> ->set_state() -> bridge_set_state() -> bridge_set_up() -> bridge_enable_interface() -> system_bridge_addbr() -> May 07 14:56:37 but without interface DEV_EVENT_ADD doesn't occur of course May 07 14:56:40 jow: ^^ May 07 15:14:18 rmilecki: that's intentional. a device needs to be used by something in order to be brought up May 07 15:14:30 rmilecki: if you want to bring it up, create an interface with proto=none that uses it via ifname May 07 15:46:07 Hi I just flashed master to my r7800 and it will not boot. May 07 15:46:14 Still on kernel 5.4 May 07 15:46:56 I just switched it out for my wrt3200 for now I will look more in to it later to see if i can find out what it is. May 07 17:49:07 aparcar[m]: it is obsolete indeed, all relevant changes are upstream now May 07 17:49:28 aparcar[m]: we just need to enable the "fat" nslookup in busybox and drop the nslookup_lede patch entirely May 07 17:50:02 rmilecki: afair you can set option bridge_empty 1 to force creating a bridge even if there's no ports May 07 17:50:40 it's yet another of these marvellously named options :) May 07 18:09:53 jow: nbd: bridge_empty didn't work May 07 18:10:01 what a mess May 07 18:10:24 if i just want to bridge switch ports, why force me to create a dump interface?! May 07 18:10:30 *dumb May 07 18:10:51 nbd: what's the purpose of such magic behaviour? May 07 18:13:52 rmilecki: in order to not bring up devices on unused/disabled interfaces May 07 18:14:51 if something is explicitly put in /etc/config/network, just bring it up :| i could use "enabled 0" if I didn't want it May 07 19:02:11 rmilecki: May 07 19:02:18 there is no devup/devdown similar to ifup/ifdown May 07 19:02:33 netifd configuration is interface centric May 07 19:03:00 so unconditionally bringing up some devices could lead to quirky behavior May 07 19:03:17 having a config device and wanting it brought up by default is not a very common case, in my opinion May 07 19:03:42 so the way to make it work is to tie it to an interface May 07 19:03:51 which also makes it easier to bring it up and down May 07 19:04:01 nbd: i'm looking for a way to desesign that cleanly May 07 19:04:54 nbd: so does netifd's interface code first creates interface and then waits for interface up event? May 07 19:05:21 AKA it doesn't try to setup just-created interface right after creating it? May 07 19:07:10 so in netifd terminology, the interface is the l3 config and the device is the underlying linux netdev May 07 19:07:21 right May 07 19:07:30 it keeps track of which devices can be brought up via the available status May 07 19:08:03 so for device types like bridge, vlan, etc. they are created when an interface claims them (i.e. wants to bring them up) May 07 19:28:27 > having a config device and wanting it brought up by default is not a very common case, in my opinion May 07 19:29:09 I disagree May 07 19:30:37 if I put a config device in my config, I expect it to be created at least May 07 19:31:14 but I know that we disagree on the dummy "config interface" requirement for bridges since we already discussed that here May 07 19:34:43 rmilecki: https://pastebin.com/6B4D4S92 May 07 19:35:01 dunno if this is what you guys are discussing May 07 19:35:08 but been using that for years May 07 19:36:16 that will make a config interface / type bridge appear even if there is no ifname and allows me to then just put a config wifi-iface / network foo into the bridge May 07 19:44:52 i have a netgear gs108t-v3 (realtek) that i flashed the current master last night, it seems to stop talking to its WAN interface after a few hundred seconds (300 in the last test). /etc/init.d/network restart makes it work for another few hundred seconds. nothing in dmesg, nothing obvious in syslog. May 07 19:52:39 russell--: does the LAN keep working? May 07 19:52:56 Borromini: i don't know, nothing is connected to LAN May 07 19:53:47 last test ping stopped at 433, started very shortly after the interfaces were up May 07 19:54:54 yeah, nothing in syslog May 07 19:56:18 ok. i'm seeing weird stuff on my gs108t-v3 as well since like a week, but i'm on 21.02. May 07 19:56:23 tcpdump shows the wan interface (switch.1 in my case) arp'ing for the gateway address, not seeing a reply May 07 19:58:51 built from cde31976e3 (because i saw the series of realtek commits) May 07 19:59:18 which hash did you use before? May 07 20:01:08 21.02 doesn't have those May 07 20:01:11 blogic: at least I can count you as another person struggling with netifd May 07 20:01:19 an older one, ec6293febc from about a month ago May 07 20:02:11 ok. i'm going to try rolling back the latest kernel bump, see if that has any effect. May 07 20:02:18 ec6293febc is .109 May 07 20:04:48 * russell-- is trying just reverting the last batch of realtek commits May 07 20:13:39 russell--: ok keep me posted May 07 20:23:25 Borromini: 3bae740311e seems okay, ping still going after 500+ seconds May 07 20:24:22 trying 515d9c85f2 May 07 20:29:49 timeouts here don't start immediately, FWIW May 07 20:29:55 don't know how it is for you May 07 20:33:03 the failures in a sample of about 4 or 5 runs is in the 300-500 seconds range May 07 20:40:22 that seems okay too, trying d4649942ad May 07 20:40:33 ok May 07 20:59:24 d4649942ad is okay too May 07 21:03:55 rsalvate_: your UBIFS patch made the news: https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.13-UBIFS-Zstd May 07 21:44:07 So people I looked at my soft bricked r7800 and it is the lan ports that don't work. Wifi comes up. May 07 21:48:49 O I see the bug is known about on the ML. May 07 22:01:15 Tapper: I assume you're being bitten by https://github.com/openwrt/openwrt/pull/3954 (the last ~10 messages in that thread) May 07 22:02:49 not really an issue in that PR, but in the patches just merged before it to support Ubiquiti UniFi AC HD May 07 22:04:17 I've just successfully built a kernel 5.10/ dsa based image for my nbg6817 last night (but not including the apparently problematic changes adding Ubiquiti UniFi AC HD support) May 07 22:25:23 pkgadd yes mate. That's the one! I will just wate for a fix and flash using my laptop or phone. May 07 22:25:50 I am still on kernel 5.4 for now. May 07 22:26:34 I will switch to Kernel 5.10 after it's being tested a bit. Hows it going on your box? May 07 22:27:34 fine, I was using the non-dsa one for about a week, the dsa PR for about 3 weeks now May 07 22:27:59 but switching to dsa requires reconfiguring from scratch May 07 22:31:11 I still haven't quite managed to get my VLAN trunk working yet, but I didn't have a lot of time to look into the different config syntax in detail either May 07 22:43:47 Tapper: https://forum.openwrt.org/t/pr-ipq806x-kernel-5-10-bump-code-propose/90022/125 should get you going again, "Reverting the unrelated commit fa73183 11 fixes things with 5.4 for me with R7800." May 07 22:44:22 pkgadd Thanks. May 07 22:58:54 hmm, the belkin rt3200 is (back-) down to 79.99 GBP today, damn - if it wouldn't be for shipping and double taxation/ customs driving that up again May 07 23:02:52 Hauke: I don't know why, it's a mere two-liner… :) May 07 23:47:07 pkgadd: I think you do not have to pay UK taxes, but your taxes only May 07 23:47:25 I think it was this way for me when I ordered this device to Germany May 07 23:54:24 psa: i'd advice against building+sysupgrading master onto ipq806x (tested ad7200+c2600) May 07 23:56:44 psa: i'd advice against building+sysupgrading master onto ipq806x (tested ad7200+c2600) May 07 23:57:26 sorry for the repeat, was meant to be: after sysupgrade, eth/lan-ports were entirely defunct, failsafe mode failed, had to rig a ttl console to get it working again May 07 23:58:32 commit, cde31976e37513cf2977e6d4b691f561344e61ed i noticed several ipq806x patches in the last two days. were these tested, by whom? May 07 23:59:47 jow: is there still a need for maintaing a seperate busybox musl version of nslookup (nslookup_lede)? apline linux also just use normal nslookup. can we just remove the patch and the config options? May 08 00:00:14 also, you should pull the current snapshot builds to avoid ppl semi-bricking their devices. tftp recovery to previous build works, ttl console works, also if you can manage ssh access through wan May 08 00:01:25 not my intention to be spamming, but i just spent over an hour recovering these devices. achtung, achtung May 08 00:02:21 pkgadd, ping May 08 00:02:24 jacekowski, ping May 08 00:02:41 you both have ipq806x devices, mind verifying? May 08 00:25:43 Anyone know where /etc/config/system gets synthesized into /etc/syslog.conf? Doesn't seem to be in /etc/init.d/system ... May 08 00:33:41 Hauke: what did you end up paying in total (I'm in .de as well)? May 08 00:33:57 shibboleth: ? May 08 00:35:44 I payed £65.80 , but it was used warehouse May 08 00:36:00 4342d27ec9 is the commit that kills gs108t-v3 May 08 00:36:20 thanks May 08 00:36:34 no VAT, but £4.71 shipping, and £10.51 import fees May 08 00:38:22 so just under 95 EUR, hmm May 08 00:39:25 shibboleth: I guess you're affected by >https://forum.openwrt.org/t/pr-ipq806x-kernel-5-10-bump-code-propose/90022/125 should get you going again, "Reverting the unrelated commit fa73183 11 fixes things with 5.4 for me with R7800."< as well May 08 00:39:50 pretty sure it was built with kernel 5.4 May 08 00:40:37 shibboleth: the kernel bump isn't the problem - the unrelated changes for the Ubiquiti UniFi AC HD merged just before were May 08 00:41:19 https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=fa731838c5247f29f68098d7c78fc444f068bda9 May 08 00:43:01 looks about right May 08 00:43:05 Hauke: you're in the UK now? May 08 00:43:12 i'll set aside some time for testing May 08 00:43:48 philipp64: the (very interesting) Belkin rt3200 is only available from there (or, as the Linksys e8450 from the US) right now May 08 00:45:16 what's interesting about it? May 08 00:46:25 philipp64: mt7622b (ARMv8, 2*1.35 GHz), mt7915e wifi6 - and fully supported by OpenWrt May 08 00:47:38 yeah, that would do it... so 802.11ax? May 08 00:47:45 yes May 08 00:47:53 how much memory and storage? May 08 00:48:28 128/512 --> https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=aa94e34c1d7ce40684ec01b8ddc056548ecd34f0 May 08 00:49:13 technically ipq807x should be faster, but that seems to be quite hard to get working May 08 00:51:05 shibboleth: I'm running last night's master (r16642-ed4641e9f1) just before those changes were merged on my nbg6817 (but with the kernel 5.10+dsa PR merged in) May 08 00:51:52 dsa, that's the "new" switchport/vlan style/config, right? May 08 00:52:04 yes May 08 00:52:38 i'm assuming the old style will still be supported? May 08 00:52:44 no May 08 00:52:48 and that current configs will be honored? May 08 00:53:13 well, right now - yes, once the dsa [WIP] PR is in, no May 08 00:54:10 the networking config is completely different - and you must nuke your config before flashing a dsa build May 08 00:57:28 you can do/ achieve (roughly) the same functionality, but the way it's configured is totally different May 08 01:12:22 pkgadd: at the same time as flashing is fine, no need to do it before ;-) May 08 01:13:21 Build [#81](https://buildbot.openwrt.org/master/images/#builders/64/builds/81) of `realtek/generic` failed. May 08 01:21:30 russell--: ack, that effectively what I meant May 08 01:51:34 Build [#40](https://buildbot.openwrt.org/openwrt-21.02/images/#builders/19/builds/40) of `layerscape/armv7` completed successfully. **** ENDING LOGGING AT Sat May 08 02:59:57 2021