**** BEGIN LOGGING AT Mon Dec 28 03:00:17 2020 Dec 28 03:56:44 yikes just read the NAND patch. that's quite ugly. Dec 28 04:03:55 i wonder if there's an alternative bootloader for those devices Dec 28 04:24:55 Still working on the P2041 (related to the P2020 finished in 2019ish). https://github.com/openwrt/openwrt/pull/1688 ... Dec 28 04:25:09 I find the same issue that CHKDSK88 had about booting FIT images Dec 28 04:27:50 I wonder, could it be because CHKDSK88 never booted an initramfs? That's what I've been trying. I should try booting directly perhaps Dec 28 04:28:01 (setting up as they did) Dec 28 04:28:04 (... in NOR) Dec 28 04:42:56 When I bootm from the beginning of the initramfs, U-boot unpacks everything and then just stops like here: https://github.com/openwrt/openwrt/pull/1688#discussion_r244647072 Dec 28 04:44:01 However, imxtract'ing the fdt and kernel from it does not work -- bootm refuses to boot using the same command: Dec 28 04:44:05 => bootm 02400000 - c00000 Dec 28 04:44:08 WARNING: adjusting available memory to 30000000 Dec 28 04:44:10 Wrong Image Format for bootm command Dec 28 04:44:11 ERROR: can't get kernel image! Dec 28 05:26:38 Are there any good techniques for reflashing u-boot? (is it possible to chain-load u-boot to preserve the original?) Dec 28 05:52:12 what determines the order in which wan uplinks are displayed on the Overview page? it doesn't seem to be interface metric Dec 28 06:42:16 hurricos: chainloading u-boot is possible, but you need to link it to use a different address. Dec 28 06:42:54 hurricos: you can use "mkimage" to produce a uImage from your extracted kernel image, then bootm will work. Dec 28 07:19:17 Struggling to understand what is going on in these two situations: https://github.com/openwrt/openwrt/pull/1688#discussion_r244647072 and https://pastebin.com/nsYV7JeY Dec 28 07:20:40 Console appears to kick the bucket, but eth autonegotiation continues to work -- not sure if that's possible if Linux is not running. Dec 28 07:23:26 console is set up with bootcmd, but not forced by any chosen in a dts -- I use the DTS already in flash as fsl's DTS from mainline has ERRORs Dec 28 07:37:54 fwiw the board does have a copy of Linux on it which boots fine, which is why I can't figure out why OpenWrt does not Dec 28 08:26:44 hurricos: you might want to enable earlyconsole Dec 28 08:26:58 hurricos: so that the kernel continues to print everything via a bootloader-configured UART. Dec 28 08:27:21 hurricos: but since you're already getting some output, hm, probably your debuglevel is too low? Dec 28 08:27:35 hurricos: in your latest paste there's zero kernel output though. Dec 28 08:27:44 hurricos: so I'd say it means your earlyconsole is not enabled. Dec 28 08:29:51 hurricos: it was called earlyprintk before, and earlycon now. Dec 28 11:39:24 dedeckeh: ping Dec 28 12:04:27 mangix: i have modest hope that eventually mikrotik will release an updated bootloader that's a little less stupid. Dec 28 12:04:36 mangix: their ipq variants support ubifs, apparently Dec 28 13:09:32 how would I tell openwrt: all packets with source address 1.2.3.4 forward out interface abc? Dec 28 13:11:36 reiffert: ipv6 can just use source-specific routes, ipv4 need policyrouting. that means you need another routingtable that routes everything out to abc and a rule that uses that tables when the source address matches Dec 28 13:14:07 I fail to add a simple rule Dec 28 13:14:18 ip rule add add from 10.50.0.3 table telekom Dec 28 13:14:18 Error: argument "add" is wrong: Failed to parse rule type Dec 28 13:15:17 add add Dec 28 13:15:35 facepalm. thanks. Dec 28 13:16:43 how can I have openwrt re-read the /etc/network/rt_tables file? Dec 28 13:17:02 once it's there, it's there Dec 28 13:17:45 I was following https://openwrt.org/docs/guide-user/network/routing and was adding "100 telekom" to the /etc/iproute2/rt_tables file, however ip route show table 100 is empty Dec 28 13:18:12 because it's an empty routing table Dec 28 13:18:18 you need to add a route in there Dec 28 13:18:36 ip route add default dev abc table telekom Dec 28 13:25:19 default dev tun0 scope link metric 5 Dec 28 13:25:19 default dev pppoe-wan scope link metric 10 Dec 28 13:25:19 default via 10.7.99.2 dev tun0 metric 20 Dec 28 13:25:45 thats my table 100 and is working great. How would I make sure the 1st entry goes away when the tun0 device goes down? similar question for the pppoe-wan interface Dec 28 13:26:32 that should be removed automatically Dec 28 13:27:30 how is it going to be added again once tun0 comes up again? Dec 28 13:28:15 depends on what it is managing it Dec 28 13:29:20 "openwrt" is managing it.. well. the openvpn service and the pppoe openwrt setting Dec 28 13:29:46 I can have the up scripts do it I guess. Dec 28 13:30:43 last question: the 3rd line in the above route table has the wrong interface and should be eth1.99 but I fail to tell it while ip route add'ing Dec 28 13:31:49 ah managed to do it :) Dec 28 13:32:31 how would I make that static within openwrt, is that in luci network firewall custom rules? Dec 28 13:34:44 i usually edit /etc/config/network for this, but there might be a better option, don't know Dec 28 13:49:05 how would I have openwrt add routes when the ethernet interface abc comes up? Dec 28 13:49:17 there is no openvpn or pppoe daemon Dec 28 14:29:42 lemmi: hotplug.d/iface does the trick. Thanks for your help Dec 28 15:19:31 any known issues in 19.07, sysupgrade throwing "Failed to connect to ubus" and not doing anything ? Dec 28 16:09:07 * funman should work on his .dts files rather than put this off to later Dec 28 17:13:09 reiffert: cool :) Dec 28 17:21:44 PaulFertser: Thank you, I think this is what I was looking for. Sadly the way I wired the relay at our lab doesn't actually trigger the device to turn on. Have to actually hit the power button Dec 28 18:44:52 any IPsec users out there? I need testers for a PR… or reviewers. Dec 28 18:45:17 jow: ping Dec 28 18:45:39 got a question about debugging the startup order of bind and ISC dhcpd… Dec 28 18:47:00 I just pushed a staging branch to master Dec 28 18:47:08 gimme a sec to figure out how to get rid of it Dec 28 18:47:44 blogic: last time it happened to me, I asked jow ... Dec 28 18:48:05 I didn't have the right to delete it myself Dec 28 18:48:17 realtek (deletion prohibited) Dec 28 18:48:22 exactly Dec 28 18:48:23 yeah, I just pinged him Dec 28 18:48:47 i removed the # in my .git/config but forgot to add it to the following line ;) Dec 28 19:19:06 hrrm seems one of my BT-HHv5a's running wifi-APs and VLANs running on a development/master build is falling-over rebooting itself! OpenWrt SNAPSHOT r15324-920b692667 Dec 28 19:19:28 I shall try updated nightly build unless anybody wishes me to collect debugging etc. Dec 28 19:27:21 enyc: that would help other people (and ideally help fix it) Dec 28 19:27:32 if lantiq is also seeing breakage... Dec 28 19:30:16 Borromini: i've updated it and observe it doesn't have upstream dns configured correctly, wonder if some ntp/whatever is failing and then system crashing itself Dec 28 19:30:41 Borromini: FAQ/Wiki/instructions/etc on how/what to test/report on devel builds? Dec 28 19:31:45 Hello, could someone please help me with these 2 questions? https://forum.openwrt.org/t/imagebuilder-question/83044 Dec 28 19:36:18 host 0.openwrt.pool.ntp.org Dec 28 19:36:48 hrrrrrrrrrrrrm seems 2.openwrt.pool.ntp.org is the only part with AAAA records ;-( Dec 28 19:38:06 mangix: ping Dec 28 19:45:23 enyc: you mean ntp is broken, not your device? Dec 28 19:46:45 Borromini: not entirely sure, but not sure openwrt supposed to self-reboot just because ntp not working Dec 28 19:47:38 Borromini: I am observing through further debugging/experimentation that openwrt + busybox-ntpd is doing bad job of just syncing the time without ipv4-access (bearing in mind this is a secondary AP type thing, not supposed to be taking up public v4 space etc) Dec 28 19:49:51 Borromini: that said it has access to a vlan that has natted v4 it could be on, but I still think OpenWRT should play nicely with ipv6-only nat64-provisioned upstream internet Dec 28 19:51:07 much -ddddddddddddd and so on, explicitly using 2.openwrt.pool.ntp.org , ... it doesn't seem to be trying any IPv6 ntp query addresses Dec 28 19:52:18 busybox ntpd has much # if ENABLE_FEATURE_IPV6 -- I wonder if this needs enabling Dec 28 19:54:03 give it temporary ipv4 access with manual ip -4 commands and ... voila,,, ntpd works Dec 28 20:08:42 anyone using arc here? Dec 28 20:11:17 aparcar[m]: Only the peole from Synopsys and their customers have access to the hardware Dec 28 20:11:26 which is supported in OpenWrt Dec 28 20:14:13 so can they pleaes send me a dev board Dec 28 20:14:24 or how else are we suppose to support that? Dec 28 20:15:42 aparcar[m]: you can ask Alexey Dec 28 20:15:59 he normally answers when you write him a mail Dec 28 20:16:32 I assume that the dev boards are expensive Dec 28 20:31:58 mhh then I'll not review these patches :P Dec 28 20:41:57 seems basically that busybox ntp client somewhat limited - does not cycle thorugh all addresses properly, does not fallback quickly with unreachable addrseses ,, and nothing in openwrt resolv library config, nor anything in ntpd etc is doing anything to help this... Dec 28 20:42:35 that's the reason why busybox is small, limited functionality Dec 28 20:44:51 jow: =) yes though it seems there is a small, pragmatic patch available, https://dev.archive.openwrt.org/attachment/ticket/12167/0001-busybox-make-ntpd-prefer-IPv6-addresses.patch Dec 28 20:45:03 jow: but I'd need to test it on IPv4-only to make sure it solves both situations right Dec 28 20:46:17 jow: I think its' not sensible in this day-and-age for OpenWRT to not work on IPv6-only network just because ntp client ... I really don't think it should need full ntpd to behave... Dec 28 20:47:30 https://bugs.openwrt.org/index.php?do=details&task_id=2608 hrrrm Dec 28 20:58:49 Have left lantiq system with 'working' ntpd and see if it now does/doesn't fall over Dec 28 20:59:31 NB: (Suggests) may be something in openwrt master that likes failing/resetting if time-wrong, ntp... or wpad processing wrong dates, etc etc.... would need to check further to confirm! Dec 28 21:02:23 aparcar[m]: pong Dec 28 21:02:35 can you test the arc thing? Dec 28 21:03:07 only compile Dec 28 21:03:31 how much are those dev boards? Dec 28 21:03:34 the patch should be fine as Synopsis carries a similar one in their glibc fork. Dec 28 21:04:00 I have no idea. Dec 28 21:06:40 I forwarded your email, let's see if they can test it for us Dec 28 21:06:58 could you test my squashfs-tools-ng patch? I don't want to get it into 20.x Dec 28 21:10:10 enyc: that patch looks sane indeed Dec 28 21:10:34 aparcar[m]: Synopsis added their approval: http://lists.openwrt.org/pipermail/openwrt-devel/2020-December/032936.html Dec 28 21:13:41 mangix: can you please elaborate on this situation? you did recently remove uclibc-ng from openwrt and now there is only musl and glibc left, right? Dec 28 21:14:22 yes but musl has no support for ARC Dec 28 21:15:01 and glibc only with your patch I guess Dec 28 21:15:28 for ARC700 yes Dec 28 21:15:42 the otherone they upstreamed support Dec 28 21:16:58 I don"t think they care much for ARC700 since it's older Dec 28 21:17:12 jow: ok, in which case can you apply it so it goes into upcoming master builds, ping me and I'll test it in various v4-only v6-only NAT64 configs which I'm well set-up to do. Dec 28 21:17:49 guys, i've been seeing this with blogic's tree for a while... i wonder if there's some broken failover thing or sth at work again? https://paste.debian.net/1178657/ Dec 28 21:20:31 Borromini: repo to large apparently, try cloning via git://git.openwrt.org/openwrt/staging/blogic.git instead Dec 28 21:20:43 that avoids the http reverse proxy timeouts Dec 28 21:20:50 jow: oh, ok. thanks. Dec 28 21:21:00 $ git clone git://git.openwrt.org/openwrt/staging/blogic.git Dec 28 21:21:00 Cloning into 'blogic'... Dec 28 21:21:01 remote: Enumerating objects: 2500612 Dec 28 21:21:08 and still counting... Dec 28 21:21:23 taking too long for nginx Dec 28 21:21:24 yep git:// just works Dec 28 21:21:31 mangix: nevermind dedeckeh takes care of your patch Dec 28 21:21:36 ah :) Dec 28 21:29:42 aparcar[m]: I talked with dedeckeh and we want to wait till after the vacation season Dec 28 21:30:10 removing the arc770 target was our preferred solution Dec 28 21:32:51 ok sounds good Dec 28 21:32:52 thanks Dec 28 21:40:01 jow: [or, give me some kind of clone/fork package or similar??] I'm re-learning my languagega and coding ando ther eexrecises and not yet into git-voodoo and opnewrt-dev voodoo yet! Dec 28 21:59:45 Hello, could someone please help me with these 2 questions? https://forum.openwrt.org/t/imagebuilder-question/83044 Dec 28 22:02:50 Hauke: can't reply to the email on mobile. but yes, itshould fix the error on ARC. Dec 28 22:03:39 ldd no longer shows anything special. Dec 28 22:06:11 mangix: were you able to reproduce the problem? Dec 28 22:16:07 no Dec 28 22:16:27 but this will fix it no matter what Dec 28 22:18:25 Hauke: try running ldd in the SDK's tools Dec 28 22:23:22 OK my bt-hhv5a (lantiq) doing wifi-APs and VLANs, that did work fine with 19.07.5 is now crashing/resetting under Master. Should I set up some network-logging / debugging somehow?? Dec 28 22:24:21 not todo with the ntp limitations or so, seemingly Dec 28 22:25:07 Hauke: yeah just downloaded the SDK. Seems the problem is bigger: https://gist.github.com/neheb/81c717df125247a142d7d37c4ef58869 Dec 28 22:27:35 syslog https://forum.archive.openwrt.org/viewtopic.php?id=11912 and serial log maybe... Dec 28 22:38:08 mangix: pkg-config.real is a shell script for musl and a binary for glibc, strange Dec 28 22:47:59 GNU's Embed-Friendly Web Server Updated With Better OS Portability, Performance https://www.phoronix.com/scan.php?page=news_item&px=GNU-libmicrohttpd-0.9.72 Dec 28 22:48:49 Binary is only about 32k (without TLS support and other optional features) Dec 28 22:50:01 it's already in packages Dec 28 22:54:43 * Tapper nods Dec 28 22:56:46 Hauke: uhhh what? It should be the same for both Dec 28 22:57:41 yeah pkg-config.real succeeds with ldd on both the SDK and locally. Dec 28 22:58:00 except the SDK has not found on the library Dec 28 22:58:30 is it encoding an absolute rpath to libpkgconfig? Dec 28 22:59:21 jow: looks like it Dec 28 22:59:40 libpkgconf.so.3 => /home/mangix/devstuff/openwrt/staging_dir/host/lib/libpkgconf.so.3 (0x00007f55663a4000) Dec 28 22:59:52 yeah, that's not going to fly with the sdk Dec 28 23:00:40 jow: it's the same problem with all of these though: https://gist.github.com/neheb/81c717df125247a142d7d37c4ef58869 Dec 28 23:01:36 mangix: that's the SDK ? Dec 28 23:01:39 yes Dec 28 23:01:44 which one? Dec 28 23:01:51 archs-glibc Dec 28 23:02:15 downloaded just now from https://downloads.openwrt.org Dec 28 23:02:23 well, an hour ago Dec 28 23:03:11 looking at the other tools, they're all linking to the OS libraries Dec 28 23:03:53 they're actually supposed to be shell script wrappers Dec 28 23:04:01 for example, libz.so.1 => /usr/lib/libz.so.1 (0x00007f4e7bece000) when there's tools/zlib Dec 28 23:04:04 I suppose bundle-libraries.sh failed, likely due to some glibc specifics Dec 28 23:05:36 makes sense. I don't think glibc SDKs have ever been tested. Dec 28 23:06:14 http://builds.openwrt.org/master/images/builders/archs38%2Fgeneric/builds/687/steps/images/logs/stdio Dec 28 23:06:55 ctrl-f "bundle-libraries.sh" Dec 28 23:06:58 2nd match Dec 28 23:07:19 see the "find: 'staging_dir/host/usr/bin': No such file or directory" ? I guess this is the culprit which makes the entire bundling process fail Dec 28 23:07:40 musl / uclibc builds probably staged at least something in usr/bin, avoiding the issue Dec 28 23:08:45 no nevermind, that is not the issue Dec 28 23:08:52 anyhow, compare with http://builds.openwrt.org/master/images/builders/ath79%2Fgeneric/builds/590/steps/images/logs/stdio Dec 28 23:09:08 ctrl-f "bundle-libraries.sh", see a lot of "Bundling xyz" messages Dec 28 23:09:32 these are missing for archs38 Dec 28 23:11:08 hmmm for host/bin/xmlwf I'm getting libexpat.so.1 => /home/mangix/devstuff/openwrt/staging_dir/host/lib/libexpat.so.1 (0x00007f46435ca000) Dec 28 23:11:19 that's absolute, no? Dec 28 23:12:08 libexpat.so.1 => /usr/lib/libexpat.so.1 (0x00007f8d3912b000) with the SDK Dec 28 23:12:22 does not matter Dec 28 23:12:55 xmlwf actually should be just a shell script containing something similar to this: Dec 28 23:13:00 #!/usr/bin/env bash Dec 28 23:13:00 dir="$(dirname "$0")" Dec 28 23:13:00 export RUNAS_ARG0="$0" Dec 28 23:13:00 export LD_PRELOAD="${LD_PRELOAD:+$LD_PRELOAD:}$dir/../lib/runas.so" Dec 28 23:13:00 exec "$dir/../lib/ld-linux-x86-64.so.2" --library-path "$dir/../lib/" "$dir/.xmlwf.bin" "$@" Dec 28 23:13:32 due to that it won't matter the slightest what kind of rpath is encoded or what ldd thinks where it should locate the libs Dec 28 23:13:48 hmm? xmlwf is a binary Dec 28 23:14:01 we move the actual elf executable aside and replace it with a wrapper script + preload library which invokes the correct ld.so and library path Dec 28 23:14:10 this is what bundle-libraries.sh does when packing the SDK Dec 28 23:14:17 that step failed for archs38 Dec 28 23:15:29 yeah it shouldn't be a binary Dec 28 23:16:10 within the SDK, the actual xmlwf should be .xmlwf.bin and xmlwf should be a shell script Dec 28 23:16:45 compare with https://downloads.openwrt.org/snapshots/targets/ath79/generic/openwrt-sdk-ath79-generic_gcc-8.4.0_musl.Linux-x86_64.tar.xz Dec 28 23:16:52 unpack and check staging_dir/host/bin/xmlwf Dec 28 23:17:02 you'll see that it is a shell script Dec 28 23:17:33 I have the mvebu SDK and that's how it is there. Dec 28 23:17:50 which subtarget? Dec 28 23:18:10 cortex a9. it's a shell script I meant to say. Dec 28 23:19:06 https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=scripts/bundle-libraries.sh Dec 28 23:19:34 if the archs38 build lacks the "Bundling xxx" messages, it means the script could not find $LDSO Dec 28 23:20:48 which could be either due "ldd" having unexpected output (maybe we called the cross-glibc's ldd instead of the host one?) or the system ldso not matching the "*/ld-*.so*" pattern Dec 28 23:20:55 I guess the culprit is with the former Dec 28 23:23:24 I do wonder what the issue is with libubox and friends though. I assume there's a native CMake way to fix this instead of hacking around it. Dec 28 23:25:38 to fix what? Dec 28 23:25:56 the gist I posted above. Dec 28 23:26:26 unless you mange to use rpath with $ORIGIN and relative search paths I don't see any fix for it Dec 28 23:29:05 I remember some hackery was needed with some hostpkg packages in the packages feed. eg. https://github.com/openwrt/packages/commit/e8da6d929451acbdcd94ae1fc747c876a84df8bc#diff-99de1cebf3f01cf590af488498732b2a46f9c8d0e6ae3fc506b90df708f06a13 Dec 28 23:29:38 -DCMAKE_SKIP_RPATH=TRUE in cmake.mk Dec 28 23:31:02 I found the reason for the archs38 build issue Dec 28 23:32:17 it is picking up staging_dir/toolchain-arc_archs_gcc-8.4.0_glibc/bin/ldd which fails since its obviously a cross ldd Dec 28 23:34:57 https://hastebin.com/yepesatule.rb Dec 28 23:35:55 ironically it does work when performed on my desktop since incorrectly hardcodes the builder path in its shebang so it fails when executed elsewhere, allowing bundle-libraries.sh to fallback to /usr/bin/ldd Dec 28 23:37:32 funny. I will test my cmake stuff to see if it works. Dec 28 23:40:06 jow: i asked a question a few days ago about iproute2 build failing under some all =m conditions, when the post build discovers what it thinks is a missing dependency (specified as a conditional dependency). do you have any idea what's wrong? Dec 28 23:42:23 2020-12-24 06:09 russell--: Package tc is missing dependencies for the following libraries: Dec 28 23:42:26 2020-12-24 06:09 russell--: libelf.so.1 Dec 28 23:42:29 2020-12-24 06:09 russell--: libmnl.so.0 Dec 28 23:42:52 russell--: not offhand Dec 28 23:46:00 russell--: where in menuconfig was tc? Dec 28 23:59:44 >KGB-0< https://tests.reproducible-builds.org/openwrt/openwrt_kirkwood.html has been updated. (100.0% images and 96.9% packages reproducible in our current test framework.) Dec 29 00:08:12 mangix: fixes pushed, the archs38 SDK should start working soon Dec 29 00:10:33 great. that should cut the build failures in half :) Dec 29 00:21:17 =) Dec 29 00:21:46 Sure I saw (somewhere) something about OpenWRT UI and MTU configuration or better display/control of MTU, at least... Dec 29 00:41:57 mangix: network / routing and redirection Dec 29 00:44:49 you can reproduce with a .config stub something like this: http://sprunge.us/o3v4PD Dec 29 00:45:22 those configs are nasty :) Dec 29 00:45:43 the way to reproduce is to select and then deselect Dec 29 00:46:04 This needs to be controlled in some way Dec 29 00:56:47 jow: do you have an opinion on https://patchwork.ozlabs.org/project/openwrt/patch/20201227213124.1638854-1-mail@aparcar.org/ ? Dec 29 02:40:02 mangix: still around? Dec 29 02:42:26 dangole: thinking about Perl… is there a way to automatically handle the $(eval $(call HostBuild)) through a hook? Dec 29 02:43:30 wondering if the package/*/compile template has a hook I could chain onto, but I’m not seeing anything in packages/Makefile that looks promising. Dec 29 02:45:20 oh, well. **** ENDING LOGGING AT Tue Dec 29 02:59:57 2020