**** BEGIN LOGGING AT Sun Jul 05 03:01:09 2020 Jul 05 03:18:57 build #349 of layerscape/armv7 is complete: Failure [failed kmods] Build details are at http://buildbot.openwrt.org/master/images/builders/layerscape%2Farmv7/builds/349 blamelist: John Audia , DENG Qingfang , Daniel Gonz?lez Cabanelas , Scott Roberts , Hauke Mehrtens Jul 05 03:18:57

Jul 05 03:47:20 build #149 of mpc85xx/p1020 is complete: Failure [failed updatefeeds] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/mpc85xx%2Fp1020/builds/149 blamelist: Hans Dedecker , Stijn Segers , Hauke Mehrtens Jul 05 05:25:36 kernel wants to get rid of "non-inclusive" words like blacklist. Jul 05 05:25:41 * mangix grabs popcorn Jul 05 05:30:15 mangix: where did you read this? Jul 05 05:38:40 build #144 of armvirt/64 is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/armvirt%2F64/builds/144 blamelist: Hans Dedecker , Stijn Segers , Hauke Mehrtens Jul 05 06:21:12 aparcar[m]: a website that can't stay up Jul 05 06:22:03 https://lkml.org/lkml/2020/7/4/229 Jul 05 06:27:29 wow what a garbage conversation Jul 05 06:58:08 karlp: ? Jul 05 06:58:21 charming as always Jul 05 06:59:20 karlp: ah, i get it. I saw hauke talk to himself, or so it seemed Jul 05 09:48:01 mangix: probably I2C master and slave will be called differently Jul 05 10:51:57 build #136 of x86/generic is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2Fgeneric/builds/136 blamelist: Hans Dedecker , Stijn Segers , Hauke Mehrtens Jul 05 11:47:18 Hauke: great, thanks! also kudos to you for working out that 74Kc patch (C0 hazard) - I'm amazed by these type of fixes, especially when seeing that it takes two different kernel patches and a GCC version update to trigger this... :) Jul 05 12:33:33 any idea how I can give an unatuhenticated user access to some status pages in luci? Jul 05 14:50:38 rmilecki: I just tried openwrt with kernel 5.4 on a WRT54GS (brcm47xx with SSB) and saw no problems Jul 05 15:04:04 Someone is still using Trillian? Holy hell Jul 05 15:04:17 "(Quit: ~ Trillian - www.trillian.im ~)" Jul 05 15:59:14 build #177 of x86/geode is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2Fgeode/builds/177 Jul 05 17:04:03 xdarklight: what is the fastes way to reproduce the traffic problem with DSA? Jul 05 17:06:36 so I am about to unbox a 3200acm to play with dsa Jul 05 17:06:55 the design of this router is awful Jul 05 17:06:59 jow: you are now in the mvebu group ;-) Jul 05 17:07:18 Did support for the dual CPU ports ever get added to DSA for that device? Jul 05 17:07:33 likely not Jul 05 17:08:01 I want to look into uci support for bridge vlan filtering Jul 05 17:11:05 Hauke: yeah... two weeks from now I'll send patches that micro-optimize target cflags and that drop any targets with less than 64MB flash Jul 05 17:11:41 because I spent all that money for this... so I deserve all the best Jul 05 17:12:07 I have a 1200acm the cheapest version Jul 05 17:13:18 will those wrt ac devices even get wpa3 support? Jul 05 17:15:31 the wrt54g will get it Jul 05 17:15:35 ;-) Jul 05 17:15:59 but slow because you wan not do HW crypto with ieee80211w Jul 05 17:16:18 there is not development going on for the marvell ac wifi Jul 05 17:19:32 jow: lol. Jul 05 17:27:34 hah Jul 05 17:28:00 funny that wrt54g will get wpa3 yet the newer wrt ac series wont Jul 05 17:28:05 mwlwifi btfo Jul 05 17:29:44 pardon my naiveness but I can't fathom why anyone would still be using a wrt54g in this day and age. Am I missing something obvious? ;) Jul 05 17:30:46 he meant it as a funny thing Jul 05 17:30:49 but its real Jul 05 17:31:04 wifi drivers on older wrt54g series are better supported Jul 05 17:31:46 jow: would you have an opinion on the correct way to address tftp boot limitation on some devices (where a 'big' kernel can be booted from flash but not from network - specifically thinking about RouterBoards here): IMO we'd need an install-only initramfs image, and it would have to be configured differently from the sysupgrade image (kernel AND openwrt configs) Jul 05 17:31:59 is there a "correct" way to do that? Jul 05 17:32:47 (lzma-loader is just delaying the inevitable and causes other issues, so I'd like to do away with it) Jul 05 17:34:46 f00b4r0: I'd probably provide some sort of bootstrap tftp firmware Jul 05 17:35:01 which offers some web or scp flash emchanism Jul 05 17:35:21 yes, that's what I have in mind. Basically stripped down luci, kernel with network+flash support, done. Jul 05 17:35:43 but what's unclear to me is how to implement that in the current image build system :) Jul 05 17:37:35 I'd typically like to turn off the automatic setting 'm' (or 'y') of the kernel config. First step to slim it down Jul 05 17:38:15 so that only the target config is used, without all the extra added stuff that's unnecessary in this context. Jul 05 17:39:11 Hauke: run: iperf3 -c Jul 05 17:39:13 then there's the issue of actually setting a tiny subset of packages for the image, as well. AIUI it's impossible to do anything like this currently. Jul 05 17:39:19 Hauke: (on the HH5A that is) Jul 05 17:45:58 nickname_: this adds support for MFP to b43: https://git.kernel.org/linus/75d057bda1fbca6ade21378aa45db712e5f7d962 Jul 05 17:46:43 xdarklight: so your HH5A is the client and your PC is the server? Jul 05 17:47:21 xdarklight: thanks I see the problem Jul 05 17:51:30 jow: thanks for looking into the DSA VLAN configuration Jul 05 18:03:06 xdarklight: ifconfig eth0 down && ifconfig eth0 up makes it recover Jul 05 18:06:54 Hauke: on the 3200acm, stuff seems to work by just sticking vlan devices together Jul 05 18:07:19 e.g. I changed lan from ifname lan1 lan2 lan3 lan4 to lan1 lan2 lan3.5 lan4 and it just worked Jul 05 18:07:34 however, the mtu on the cpu port appears to be too low Jul 05 18:09:16 eth0 needs to be bumped to mtu 1504, otherwise tagged traffic gets stuck Jul 05 18:19:56 Hauke: so probably a PMAC issue (compared to a switch/GSW issue)? Jul 05 18:24:31 nvm, slightly more complicated setup and traffic ceased to function Jul 05 18:24:46 looks like a mtu issue... ping and ssh works https gets stuck Jul 05 18:47:24 urgh the mtu fun Jul 05 18:47:46 needs redesign/whatever Jul 05 18:55:41 build #161 of x86/64 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2F64/builds/161 **** BEGIN LOGGING AT Sun Jul 05 19:03:35 2020 Jul 05 19:15:07 jow: on the eth0 interface you have the VLAN and also a special tag Jul 05 19:16:14 the tag is 4 or 8 bytes on marvell Jul 05 19:18:08 xdarklight: this looks like a software problem in the driver to me Jul 05 19:22:42 jow: does it work when you put lan3.5 into a bridge? Jul 05 19:23:14 when you directly reveice on lan3.5 there is no problem because Linux will do the filtering in software Jul 05 19:23:55 jow: this helped me to understand the vlan filtering with the bridge util: https://developers.redhat.com/blog/2017/09/14/vlan-filter-support-on-bridge/ Jul 05 19:53:27 Hauke: I came up with this for now: https://pastebin.com/iqVsiBkP Jul 05 19:53:41 this results in switch0.1, switch0.2, switch0.5, switch0.8, switch0.11 Jul 05 19:53:56 with the salve ports tagged/untagged according to the vlanXports spec at the top Jul 05 19:54:36 however I have no idea if this setup is accellerated or not Jul 05 19:54:42 requires some further testing Jul 05 19:55:10 if this is working though, it should be trivial to implement in C and replace the vlanXports variables with "config switch_vlan" sections Jul 05 19:56:02 I made a standalone wifi interface to connect to and deleted all other existing ifaces (lan, wan, wan6) Jul 05 19:57:47 jow: ifconfig shows the SW counters and ethtool -S shows the HW counters Jul 05 19:57:55 this helps to see if it is offloadded Jul 05 19:58:43 thanks for the hint, will check Jul 05 20:04:57 mh, need to build some test setup for iperf first. Going to do that tomorrow Jul 05 20:05:16 had a little branfart here. I guess traffic (ping) to the router itself is always handled in SW Jul 05 20:05:39 need to check traffic flowing from one port to another, I guess that should solely happen in hardware Jul 05 20:05:50 but I am fairly confident that it'll work as expected Jul 05 20:12:21 jow: yes Jul 05 20:12:35 jow the fabric has the flow db (fdb Jul 05 20:13:02 and in that there are the macs, ports, vids, ... Jul 05 20:13:21 .win 20 Jul 05 20:13:24 grrr Jul 05 20:15:32 jow: you have to test it with a second device on the switch Jul 05 21:00:03 build #151 of x86/legacy is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2Flegacy/builds/151 Jul 05 21:21:54 updated openwrt/upstream, https://sdwalker.github.io/uscan/index.html Jul 05 21:46:49 Does anyone know why the package e2guardian was abandoned? Just lack of a maintainer or was it replaced by something better? Jul 05 21:55:29 xdarklight: it is hitting this netif_stop_queue() and the queue is not started again: https://elixir.bootlin.com/linux/v5.4.50/source/drivers/net/ethernet/lantiq_xrx200.c#L319 Jul 05 22:03:00 xdarklight: I will have a closer look at this in the next days, but there is a bug in the MAC driver Jul 05 22:08:49 Hauke: thank you! in case it's of any help (otherwise sorry for the noise), the etop driver does this: https://elixir.bootlin.com/linux/v5.4.50/source/drivers/net/ethernet/lantiq_etop.c#L190 Jul 05 22:16:34 xdarklight: yes I tried something similar Jul 05 22:16:40 have to experiment some more Jul 05 23:54:29 build #150 of mpc85xx/p1020 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/mpc85xx%2Fp1020/builds/150 Jul 06 01:25:03 build #145 of armvirt/64 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/armvirt%2F64/builds/145 **** BEGIN LOGGING AT Mon Jul 06 02:59:03 2020 **** ENDING LOGGING AT Mon Jul 06 03:00:56 2020