**** BEGIN LOGGING AT Fri Nov 08 02:59:57 2019 Nov 08 07:58:10 ynezz: Double checking my U4019 again Nov 08 08:14:09 thanks, it's quite interesting Nov 08 08:16:15 On my board, 2ms works fine. To make sure the pin had not already been reset or something like that, I also tested to disconnect and connect the power. Ethernet comes up and works fine after reboot Nov 08 08:16:57 I am not very familiar with all the details of ipq40xx, but maybe there is some internal difference between boards? Or can be, at least. And 5ms is a "safe" value since it is sufficiently high Nov 08 08:17:51 I think, that it's rather about the external switch Nov 08 08:18:42 there are more HW revisions of U4019? Nov 08 08:19:20 Yeah, I think you are right. Strange that two of the same board requires different delay Nov 08 08:19:31 anyway I think, that it might be as well good enough to remove that delay setting and use the 10ms default delay Nov 08 08:19:32 I dont think so, but I can ask my contact there Nov 08 08:19:48 Yeah, I guess we can survive 8ms extra :) Nov 08 08:20:41 I have asked Unielec, will let you know if or when they get back to me Nov 08 08:23:47 what is vendor doing, what is their reset delay? Nov 08 08:25:12 the vendor firmware uses the initial solution Nov 08 08:25:27 ah, the driver Nov 08 08:25:30 let me check Nov 08 08:25:45 usleep_range(1000, 10005); Nov 08 08:25:59 yup Nov 08 08:26:38 ok, so minimum 1ms max 10ms Nov 08 08:27:21 u64 delta = (u64)(max - min) * NSEC_PER_USEC; Nov 08 08:28:06 https://elixir.bootlin.com/linux/latest/source/kernel/time/timer.c#L1999 Nov 08 08:28:45 there is also Nov 08 08:28:47 ktime_t exp = ktime_add_us(ktime_get(), min); Nov 08 08:29:00 but then I guess 10 should be fine, since it will be higher than max of vendor driver Nov 08 08:29:44 so minimum sleep is 1ms, with a slack of 9 Nov 08 08:30:57 ah Nov 08 08:30:58 * The kernel give the normal best effort behavior for "@expires+@delta", Nov 08 08:30:59 * but may decide to fire the timer earlier, but no earlier than @expires. Nov 08 08:31:30 qca8075 80-Y9112-1 Rev. E (October 25, 2016) says "The RESETn signal must be asserted and kept low for at least 1ms after 3.3 V power and reference clock signals become stable. 3.3 V rising duration from 10% to 90% should be larger than 500 μs. The subsequent warm hardware reset needs at least 1 ms." Nov 08 08:34:21 so Nov 08 08:35:48 so that would be 1.5ms, or rather 1.501 Nov 08 08:36:04 seems vendor driver ignores the spec then and just pickes something "safe" Nov 08 08:40:24 so lets see what regression reporter finds out Nov 08 08:47:16 hi all i'm trying to compile avro (from libs/avro) using openwrt-sdk after ./scripts/feeds update -a i do ./scripts/feeds install avro without compliant Nov 08 08:47:55 but make package/avro/{clean,compile} fails Nov 08 08:48:29 i tried package/avro-c-utils/clean and it failed too Nov 08 08:50:45 On the "news"... https://www.reddit.com/r/HomeNetworking/comments/drfsr5/ubiquity_spying_feature_in_new_firmware_mandatory/ Nov 08 08:59:10 jow: ping Nov 08 09:00:12 aparcar[m]: pong Nov 08 09:00:33 I'm setting up the buildbots with phase1 on buildbot 2.5 Nov 08 09:01:16 it works locally but now I got it on an external server the worker (former slave) connects, but does not show up in the interface Nov 08 09:01:39 is that a typical thing when it is not running on localhost or something and I have to add something special? I feel like the buildbot example was somewhat incomplete Nov 08 09:02:25 https://phase1.aparcar.org/#/builders the targets show up, so the master "works" Nov 08 09:02:54 also the worker attaches itself https://paste.sr.ht/%7Eaparcar/9462d0591e448170e85ec587df6c147f75fe281a Nov 08 09:04:18 aparcar[m]: the interface looks totally unfamiliar to me Nov 08 09:04:29 I agree that from the logs everythzing looks fine Nov 08 09:04:39 can't offer any more insight tough Nov 08 09:04:54 maybe tail twisted.log on the master and see if it reports errors Nov 08 09:04:58 do the same on the worker Nov 08 09:05:33 already open, nothing suspicious comes up Nov 08 09:05:50 the web ui is simply 2.5 instead of 0.8, the config files are nearly the same Nov 08 09:06:19 anyway I though you might have a simple answer like "auth xy is blocked" Nov 08 09:06:26 https://phase1.aparcar.org/#/builders is empty too Nov 08 09:06:32 something is odd there Nov 08 09:06:40 maybe you need to be logged in to see proper status info Nov 08 09:06:54 give it 1-2 seconds to load Nov 08 09:07:02 I gave it 20 Nov 08 09:07:47 firefox console said: Firefox can’t establish a connection to the server at wss://phase1.aparcar.org/ws. Nov 08 09:07:59 maybe you need to enable websockets support on your server externally Nov 08 09:08:02 or somehow disable it Nov 08 09:08:49 seriously... websockets Nov 08 09:09:28 I think websockets are for the live logs Nov 08 09:09:34 you don't like them? Nov 08 09:09:58 WEBSOCKETS! Nov 08 09:10:03 now it seem to work! Nov 08 09:10:17 I should have opened the debug console earlier! thanks jow Nov 08 09:10:42 I find the websockets protocol overly complex and I do not like the idea of browsers open arbitrary tcp connections Nov 08 09:10:59 https://phase1.aparcar.org/#/builders/33/builds/1 Nov 08 09:11:03 ooh, 2.5 looks nice Nov 08 09:11:07 anyhow, its probably needed for the ui to be fancy Nov 08 09:11:57 jow: the js luci stuff does not use websockets? Isn't that like, the new cool thing? Nov 08 09:12:05 no it does not Nov 08 09:12:17 you need like 100KB worth of server side libraries to support it Nov 08 09:12:21 jow: ack Nov 08 09:12:32 jow: is the connection between workers and master secured by ssl? Nov 08 09:12:44 you mean in the existing setup? Nov 08 09:12:44 ynezz: I got a reply, there is (or actually "should") only be one revision of the u4019 Nov 08 09:12:54 kristrev_: ok, good to know, thanks Nov 08 09:13:12 jow: yea, I'm basically try to rebuild the existing setup with docker + buildbot 2.5 Nov 08 09:13:17 ynezz: Will you take care of patching the delay or should I do it? Nov 08 09:13:23 aparcar[m]: I opted for ssh reverse tunnels in some cases Nov 08 09:13:37 kristrev_: I would wait for the reporter first Nov 08 09:13:51 okay, couldn't find that in the example configs Nov 08 09:14:57 oh buildbot seems to use angular, we should motivate ianchi to tweak the interface a bit Nov 08 09:15:24 russell--: it's the ioctl for doign non-standard baudrates. most proper drivers and most proper userspace tools know how to use it, even if stty isn't one of them... Nov 08 09:15:54 picocom/minicom/tio et al all do, and the ftdi/cp210x/pl23xx and friends all do too Nov 08 09:19:30 ynezz: Sounds good Nov 08 09:24:20 karlp: thanks for the pointer Nov 08 09:25:37 http://www.hce-engineering.com/setting-custom-baud-rate-on-serial-port/ and so on. Nov 08 09:25:41 what's the context of this again? Nov 08 09:25:47 who needs the custom bauds? Nov 08 09:26:27 even things like pyserial are "doing the right thing" in the background, so it's normally "only" quick privat thinge that have been hacked up that run into problems. Nov 08 09:27:27 karlp: some other badly designed device you might need to communicate with.. this would be my short answer =) Nov 08 09:29:58 hauke's arduino yun patch he referred to doesn't seem to have any patch, just mentions it Nov 08 09:31:38 not really sure what's going on with that ramips one though. Nov 08 09:49:59 problem resolved - make menuconfig and save was missing Nov 08 09:58:53 jow: i yesterday investigated the iface trigger for haproxy and i think it is completely unnecessary. I feel like dropping it. Nov 08 10:00:46 can somebody please push to openwrt.git so I see if my phase1 server is triggered :> Nov 08 10:01:38 ok Nov 08 10:02:38 so this is my latest haproxy.init : https://pastebin.com/NyjZvmFD Nov 08 10:03:08 oh a good commit Nov 08 10:05:44 thanks https://phase1.aparcar.org/#/changes/1 Nov 08 10:06:28 and here is the list of to be build targets https://phase1.aparcar.org/#/pendingbuildrequests Nov 08 10:07:53 so no difference to the current setup, right? Nov 08 10:08:19 ynezz: only the design is more spicy - sorry, no new features yet Nov 08 10:08:43 well, just asking :) Nov 08 10:52:56 Hmm, does anyone know why the sitecom_wlr-7100 (ath79) uses the mainline ath10k instead of -ct? It's the only device in master that does Nov 08 10:53:01 Was this just an oversight? Nov 08 10:54:44 I'm working on making the ath10k kmod and firmware variants configurable through a global .config symbol Nov 08 11:02:35 gladiac1337: looks heaps nicer now :) Nov 08 11:46:32 neoraider: this router has only 64MB ram, and with -ct variant oom killer started it's work, that was with single client on DFS channel, opened LuCI interface and ssh connection. While on upstream ath10k it worked without problems. Nov 08 11:56:06 procd - does anyone have objections against this one ? https://patchwork.ozlabs.org/patch/1189235/ Nov 08 11:56:24 tmn505: I see. The same issue exists on all 64MB devices though, and the workaround we applied for the mainline ath10k was rejected by the -ct maintainer greearb__: https://github.com/openwrt/openwrt/pull/1077 Nov 08 11:57:38 karlp: thanks :) Nov 08 12:02:25 wigyori: hmm, would this allow to get a console on an otherwise closed device via a USB gadget? Sounds dangerous if the admin doesn't expect this behaviour Nov 08 12:07:54 neoraider: that's a pity, well, ootb experience needs to be stable. Until we have fwcfg configuration interface, I would like to keep the current variant, but if someone will change it I won't complain, someone other probably will. Nov 08 12:08:06 wigyori: I'm all for it please. Nov 08 12:08:14 neoraider: only if the gadget is configured on the device already Nov 08 12:08:17 it doesn't ahppen by itself. Nov 08 12:08:48 the device has to have a usb otg/device port, and have configured a gadget serial port there. Nov 08 12:09:26 I've not tested mhei's version of it, my gadget is powered via the gadget port, so it's always there. Nov 08 12:09:42 I've not needed the hotplug stuff, just adding ttyGS0 to the inittab was enough for me. Nov 08 12:27:40 should I not be able to use "md" in uboot to drump the uboot-env and art partitions? I just seem to get "abbaaba" variants, instead of any of the uboot env strings, or anything that would match the mac adresses Nov 08 12:28:12 (just trying to get the actual mac address out, instead of the ethaddr = 00:AA:BB:CC:DD:EE default it shows Nov 08 12:33:02 nvm, was using logical addresses, instead of the ranged into the memory map properly at 0x9fxxxxxxx Nov 08 12:50:44 More defaults weirdness: nec_wg800hp (also ath79) defaults to CT-HTT ath10k firmware Nov 08 12:59:18 wigyori: I've noticed this one, but didn't have time to comment on it, but shouldn't it go through some wrapper script? Nov 08 13:00:14 I mean package/base-files/files/etc/inittab:::askconsole:/usr/libexec/login.sh Nov 08 13:00:24 or this is being handled by the procd itself? Nov 08 13:00:43 (don't know about the internals) Nov 08 13:03:48 ok, answering to myself, start-console runs procd_inittab_run("askfirst"); so it should be ok Nov 08 13:04:47 erm no, start-console calls askfirst, but /usr/libexec/login.sh wrapper uses askconsole Nov 08 13:08:50 ah, askfirst is being defined in target's inittab, so looks good to me Nov 08 13:32:07 ynezz: https://github.com/openwrt/openwrt/pull/2533 (one last question for you) Nov 08 13:36:09 champtar: well, you're the author :) Nov 08 13:53:55 alright, that's 3 acks on the procd stuff - thanks :) Nov 08 14:12:41 wigyori: its good to go Nov 08 14:13:02 I had a look at that one a couple of days but wanted to give it a spin on HW to be sure Nov 08 14:13:21 any issues in that code breaks stuff heavily for everyone :-) Nov 08 14:14:43 * ynezz wonders if it would be possible to somehow test the working console in CI under QEMU or such Nov 08 14:15:24 ynezz: it should Nov 08 14:15:49 i am talking to aparcar[m] that we make buildbot generate images that we can automatically test using udron inside the build pipeline Nov 08 14:16:14 it'll be a couple of months till the basic infra is ready and we can start building udrone test cases but its wip Nov 08 14:16:54 build pipeline is long feedback loop Nov 08 14:17:09 s/build/buildbot/ Nov 08 14:17:45 the magic of CI nowadays is that you fork the repo on GH/GL, commit, get same feedback as you would on buildbot, but much faster Nov 08 14:18:40 so this could be actually added under the procd project repo CI Nov 08 14:19:07 you can still build procd out of tree, get it in the image, run test it Nov 08 14:20:03 aparcar[m] has some basic proof-of-concept somewhere already with Docker Nov 08 14:21:02 make ci-runtest-foo -f Makefile.ci (so one could run it locally without any CI as well) Nov 08 14:21:35 https://gitlab.com/ynezz/openwrt-ci/#usage-example Nov 08 14:22:39 GL offers really generous build resources, my current CI pipeline has ~5 jobs which usually run at once/parallel Nov 08 14:22:48 (depends on the daytime) Nov 08 14:23:32 5 jobs for upstream/fixes-improvements-refactoring-ci in 2 minutes and 51 seconds (queued for 7 seconds) Nov 08 14:24:26 (so... can GL just build releases too then..... Nov 08 14:24:44 snapshots, yes Nov 08 14:24:57 I wouldn't trust it with releases until it's 100% reproducible :) Nov 08 14:25:12 even then I do not feel like handing the private keys to GL Nov 08 14:25:20 but you can still add private runners Nov 08 14:25:23 and full write access to the repo servers Nov 08 14:27:09 that 2m51s is for this one https://gitlab.com/ynezz/openwrt-uledd/pipelines/94640958 SDK build under 4 different platforms mips/mips-be/arm32/arm64, x86 with 4 different compilers gcc-6,7,8,9 and release/debug combinations (8 compile steps), 2 static code analyzers (cppcheck, clang's scan-build) Nov 08 14:27:48 actually there is clang-9 as well, so 5 different compilers, 10 compile steps Nov 08 14:28:26 jow: but it could be still usefull to build it two times for reference before the release Nov 08 14:28:40 just to make sure it's 100% reproducible in other environment Nov 08 14:28:56 blogic: yeah, not being a procd expert i wanted to collect acks/naks on this one - thanks for the 4th Nov 08 14:31:42 ynezz: indeed, its fine for that. I'm just very careful when heavily relying on "free" and "generous" offerings by agile startup companies fueled by venture capital, preparing to go public soon Nov 08 14:32:34 eventually all these investments have to pay back, and thats not going to happen by giving rented compute resources away for free Nov 08 14:33:32 it's like max. 100 concurrent GCE instances, peanuts for such companies Nov 08 14:34:00 and actually you've time limits for private repos, unlimited for opensource Nov 08 14:34:12 s/time/build time/ Nov 08 14:34:37 anyway, I've the same oppinion on it, but with GL I've at least self-hosting option exit strategy if it goes in the wrong direction Nov 08 14:35:44 like Oracle buys them :p Nov 08 14:35:44 if we still know how to setup CI by then ;D Nov 08 14:36:58 sr.ht knows Nov 08 14:38:57 :D Nov 08 15:02:12 do we accept a signed-off-by with chinese characters? Nov 08 15:02:20 https://github.com/openwrt/openwrt/pull/2427/commits/6967cf9c72addd605b873f45ec9bf3b5ca3fb10d Nov 08 15:02:21 ynezz: sed 's/2/3/' I didn't feel as the author, should I keep your SOB ? Nov 08 15:07:33 lantiq Nov 08 15:23:40 adrianschmutzler: of course Nov 08 15:24:22 Petr Štetiar and Rafał Miłecki are allowed special chars aswell :-D Nov 08 15:57:24 neoraider, you could easily apply a patch to the ath10k-ct upstream driver, just like you do for stock ath10k driver? I just don't want the patch in the upstream driver because I think it will decrease performance on higher-end systems. Nov 08 15:57:52 and you could submit the patch to openwrt, I don't control that and wouldn't complain either way. Nov 08 16:47:01 greearb__: for Gluon, we don't really care about -ct, because we need 11s and it's easier to just use mainline/QCA ath10k everywhere instead of just on devices where -ct doesn't support 11s Nov 08 16:47:22 greearb__: but I'd like to find a nice solution for OpenWrt default builds as well Nov 08 16:53:36 ok, fwcfg is probably the best way to do it with -ct, and I'll certainly take a patch that adds that as an option to fwcfg. Nov 08 17:55:08 karlp: there was a arduino yun patch somewhere ;-) Nov 08 19:20:34 hey guys! I spoofed my mac address (by mistake) for WAN, and was wondering - how can I easily find what the original hardware mac was? Nov 08 19:20:42 using ssh preferably Nov 08 19:21:28 NobodyYet: cat /sys/class/net/ethX/address Nov 08 19:21:51 that shows my spoofed address Nov 08 19:21:53 or /etc/board.json if it was in a switch Nov 08 19:22:14 ah right Nov 08 19:22:18 yeah both show me the spoofed address unfortunately Nov 08 19:23:17 "/etc/board.json" will only show the changed address if you changed it before build Nov 08 19:23:46 I did, yes. I added /etc/config/networking before I compiled, and that overwrote the mac Nov 08 19:24:24 added that in the files/ folder Nov 08 19:25:26 then you have to look in the code repo where the original address comes from Nov 08 19:25:38 dts or 02_network depending on your device setup Nov 08 19:25:48 build #36 of ar7/generic is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/ar7%2Fgeneric/builds/36 Nov 08 19:26:08 build #42 of armvirt/32 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/armvirt%2F32/builds/42 Nov 08 19:27:15 build #40 of cns3xxx/generic is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/cns3xxx%2Fgeneric/builds/40 Nov 08 19:27:30 build #40 of ixp4xx/harddisk is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/ixp4xx%2Fharddisk/builds/40 Nov 08 19:28:11 build #38 of layerscape/armv8_64b is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/layerscape%2Farmv8_64b/builds/38 Nov 08 19:28:16 build #42 of mediatek/mt7622 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/mediatek%2Fmt7622/builds/42 Nov 08 19:28:25 build #40 of mvebu/cortexa9 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/mvebu%2Fcortexa9/builds/40 Nov 08 19:28:38 build #36 of octeontx/generic is complete: Exception [exception ccachecc] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/octeontx%2Fgeneric/builds/36 Nov 08 19:29:07 build #40 of mvebu/cortexa53 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/mvebu%2Fcortexa53/builds/40 Nov 08 19:29:12 build #36 of oxnas/ox820 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/oxnas%2Fox820/builds/36 Nov 08 19:29:25 build #37 of ramips/rt288x is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/ramips%2Frt288x/builds/37 Nov 08 19:29:37 build #36 of sunxi/cortexa53 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/sunxi%2Fcortexa53/builds/36 Nov 08 19:29:44 build #40 of x86/generic is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/x86%2Fgeneric/builds/40 Nov 08 19:30:00 build #36 of x86/64 is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/x86%2F64/builds/36 Nov 08 19:30:03 build #38 of x86/geode is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/x86%2Fgeode/builds/38 Nov 08 19:30:06 build #34 of x86/legacy is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org/openwrt-18.06/images/builders/x86%2Flegacy/builds/34 Nov 08 19:30:33 that was me, sorry for that Nov 08 19:37:05 O_o Nov 08 19:42:22 jow: already upgrading the buildbots? ;) Nov 08 19:46:02 nope, fat fingering Nov 08 19:46:51 :) Nov 08 20:08:33 jow: what do you think of changing the builder to a branch-target basis so there is no split between release builders and snapshot builders? Nov 08 20:40:26 worth a try **** BEGIN LOGGING AT Fri Nov 08 21:46:43 2019 Nov 08 23:16:23 https://news.ycombinator.com/item?id=21487854 **** ENDING LOGGING AT Sat Nov 09 02:59:57 2019