**** BEGIN LOGGING AT Tue May 26 02:59:57 2020 May 26 07:49:56 nbd: looks good, thanks a lot! :) May 26 07:50:25 I've just pushed that fuzzer seed input as well May 26 08:21:45 nbd: thanks for pushing patches May 26 08:39:47 i'll push one more tiny unrelevant libubox commit & update package in master May 26 08:47:55 thanks nbd for waiting :P May 26 08:48:30 sorry, didn't see your message May 26 08:49:31 ping me next time :) May 26 08:49:59 just laughing, thanks for bumping so I don't have to deal with it :) May 26 08:50:10 and thanks for your cleanup! May 26 08:50:25 that was very helpful and it seemt to actually work :D May 26 08:50:29 :) May 26 08:56:50 should we cherry pick libubox update to the 19.07 or wait few days? May 26 09:01:47 should probably cherry pick it soon, since the issues fixed were rather serious May 26 09:16:30 jow: do the buildbots support some simple webhooks on notifying a specific build is done? May 26 09:24:21 xback: pling May 26 09:29:16 I just tested a random squashfs image and it's rebuild and once unpacked they seem identical (via diffoscope), however they have different sha256sums. Is it possible that squashfs-tools is still not deterministic? I tested the original image https://downloads.openwrt.org/snapshots/targets/ramips/mt7620/openwrt-ramips-mt7620-alfa-network_ac1200rm-squashfs-sysupgrade.bin and the rebuild May 26 09:29:17 https://images.aparcar.org/rebuild/ramips/mt7620/openwrt-ramips-mt7620-alfa-network_ac1200rm-squashfs-sysupgrade.bin. If anyone comes to the same conclusion I'd finalize the usage of squashfs-tools-ng via https://github.com/openwrt/openwrt/pull/2916 May 26 09:29:55 lynxis: -^ May 26 09:40:52 sqfsdiff says the file timestamps differ May 26 09:43:10 mksquashfs 4.4 has command line options "-all-time", "-mkfs-time" and *should* support the SOURCE_DATE_EPOCH environment variable May 26 09:43:45 goliath: how do you compare them? I get an error message:openwrt-ramips-mt7620-alfa-network_ac1200rm-squashfs-sysupgrade_re.bin: reading super block: wrong magic value in super block. May 26 09:44:56 aren't those u-boot images? I ran binwalk on both of them and got 2 squashfs images out May 26 09:48:02 I tried the same but it didn't work neither... _openwrt-ramips-mt7620-alfa-network_ac1200rm-squashfs-sysupgrade.bin.extracted/1AE73B.squashfs: reading compressor options: data corrupted. May 26 09:48:50 yes, I was about to write that I also had to comment out the compressor option path in sqfsdiff, because the options in the images seem to be broken. May 26 09:48:50 I will add a command line flag to sqfsdiff to ignore those, since they seem to break for pretty much every $VENDOR patched SquashFS. May 26 09:49:59 goliath: time to switch to squashfs-tools-ng... May 26 09:54:09 The OpenWrt version of squashfs-tools (squashfs-kit) doesn't support the all-time option... May 26 10:36:27 aparcar[m]: squashfs-kit is reproducible, I think you run into other problems, which does not depend on the tool May 26 10:39:07 aparcar[m]: IMHO: how we build images and package together with our build system is the problem. we build multiple times with different timestamps. maybe you start documenting how the builder works, than you'll see the problem and can fix it in our build setup or build system. May 26 10:39:35 jow: is there a starting point for aparcar[m]. do we have our build setup documented? May 26 10:39:53 sure, its all in buildbot.git, ready to go May 26 10:40:20 f00b4r0: pong May 26 10:59:33 aparcar[m]: goliath: if squashfs-kit wouldn't be reproducible, the tests.reproducible-builds.org would detect it, but those are reproducible. May 26 10:59:55 e.g. https://tests.reproducible-builds.org/openwrt/openwrt_mediatek.html May 26 14:50:04 so, ath10k-ct can dump some useful info when the firmware crashes, including full firmware mem dumps and such. This can be detected by a udev event. Is there any existing daemon that is listening for udev events that could start listening for firmware crashes and copy the binary info somewhere permanent-ish? May 26 14:53:46 greearb: procd is handling udev events May 26 14:57:44 is it paying attention to FW crashes? I'm not super interested in hacking on procd, but if someone can make the crashes available somewhere, I'll work on debugging those. May 26 14:58:40 I have some crash-event handling code that I could share too, but it is from a closed-source c++ app of mine, so it would require a bit of work to make it slot into procd I guess. May 26 14:59:38 It listens to udev events and one can configure it by editing /etc/hotplug.json and it can call any external application when receiving such an event. May 26 15:01:00 Is crash info not present in dmesg? I thought OpenWrt users are usually saving full logread and then attaching that to the tickets in your repo. Is it not working good enough in practice? May 26 15:01:44 for thinks like an NPE, I don't get much backtrace, so being able to look at the full memory (effectively a core dump) allows me to sometimes make more progress on tricky bugs. May 26 15:02:51 and I can debug hot-path bugs by writing to static structs in memory and then debug them after the crash to see where code was and what were values...like a printk to ram May 26 15:02:53 I see, so basically full "core dump". May 26 15:03:02 nod May 26 15:04:27 So first step would be to package a userspace utility which can fetch the crash from the driver, right? It's not like the full dump is encoded inside a udev event anyway. May 26 15:05:02 The driver already packages it in some sysfs file, let me check my code May 26 15:05:06 Or is it just exposed via a sysfs file/ May 26 15:05:15 Asked before seeing the answer :) May 26 15:06:42 sorry, uevent, not udev event May 26 15:07:38 build #145 of lantiq/ase is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/lantiq%2Fase/builds/145 blamelist: Rafa? Mi?ecki , Petr ?tetiar , Felix Fietkau , Jo-Philipp Wich May 26 15:08:29 Yes, procd uses NETLINK_KOBJECT_UEVENT May 26 15:10:05 https://pastebin.com/KnbDpGzv May 26 15:10:37 that is how I detect it....vast majority of that method can be ignored May 26 15:11:20 so, think I could use the json thing to cause it to call some script when the crash is seen? May 26 15:17:43 I would expect it to work, yes, it would be an "add" ACTION with "virtual" SUBSYSTEM and "devcoredump.*" DEVNAME I guess or something close to that. May 26 15:18:43 And then a script can be run to save the data exposed to /tmp May 26 15:19:26 I think in any case it's worth having some simple guide at collecting it manually too. May 26 15:20:06 it is a pain to find the coredump in sysfs, it moves around for each crash, so really you need something parsing the uevent May 26 15:20:50 I had a tool that would find it in the old location, I guess I could update it to handle some callback from procd May 26 15:21:17 parsing the uevent? Isn't it a variable with a path? May 26 15:21:40 it has some unique id hash in each crash, so you cannot just 'cat' some file in known location May 26 15:22:21 yeah sure, but isn't that iD just an uevent variable? May 26 15:22:30 yes, I suppose May 26 15:23:55 (strncmp(ln.c_str(), "add@/devices/virtual/devcoredump/", strlen("add@/devices/virtual/devcoredump/")) == 0) { May 26 15:24:17 String crashdir = ln.c_str() + strlen("add@/devices/virtual/devcoredump/"); May 26 15:25:19 in this case you can simply use that variable and forward it as parameter to the invoked script May 26 15:25:48 might be that the script even inherits it by defualt as envvar May 26 15:42:33 build #145 of at91/sam9x is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/at91%2Fsam9x/builds/145 blamelist: Rafa? Mi?ecki , Petr ?tetiar , Felix Fietkau , Jo-Philipp Wich May 26 15:50:05 build #144 of x86/64 is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2F64/builds/144 blamelist: Rafa? Mi?ecki , Petr ?tetiar , Felix Fietkau , Jo-Philipp Wich May 26 15:53:04 thanks for the procd suggestions. I'll see if I can make some progress on this when I get a chance to work on Openwrt some more. May 26 15:54:13 Probably the end user can save all the generated dumps and attach it May 26 15:55:10 build #144 of ipq806x/generic is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/ipq806x%2Fgeneric/builds/144 blamelist: Rafa? Mi?ecki , Petr ?tetiar , Felix Fietkau , Jo-Philipp Wich May 26 15:55:43 I'm working with an OpenWrt related project and building a big automated CI/CD test rig for it. I should be able to extract the files as part of my automated test reporting. I'll push anything I can back upstream. May 26 15:56:12 In this case it'd be handy to have automation indeed. May 26 15:57:10 users right now could grab the binary file with some coaching, but usually the dmesg dump is enough to go on I think. May 26 16:03:38 build #144 of brcm47xx/mips74k is complete: Failure [failed pkgbuild] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/brcm47xx%2Fmips74k/builds/144 blamelist: Rafa? Mi?ecki , Petr ?tetiar , Felix Fietkau , Jo-Philipp Wich May 26 16:50:10 jow: did sysauth_template disappear in luci 184ea6230076 ? My template just gets ignored now, and the new themes all just have it built into the header? I want to keep requiring people to set a password, not just have the warning and ignore it. May 26 16:52:50 I have to replace the /usr/lib/lua/luci/view/sysauth.htm with my own now, which runs into the "file blah is alreadyd provided by package wop" May 26 18:26:47 *yawn* May 26 18:27:07 nbd: have a glance at the multi psk patches May 26 18:27:18 i'll commit in 24h if i get no feedback May 26 18:27:44 ynezz: is upwrd on hold due to cdr ?! May 26 18:28:01 or whats the status on that ? May 26 18:28:40 blogic: is there documentation so the ones that are not so able to read the code are able to test it to give feedback? May 26 18:28:58 stintel: plan to add wiki & luci next week May 26 18:29:17 stintel: the commit on patch #3 is self explanatory though May 26 18:29:31 had 3 randoms reply and test already May 26 18:29:38 still a bug in hostapd that i need to fix May 26 18:29:57 if wlan0 already exists when hapd starts the vlan netdevs dont spawn May 26 18:30:06 will investigate the coming days May 26 18:30:28 and we need a wlanzone blaock target in fw3 May 26 18:30:30 so May 26 18:30:48 reminds me I still need to provide feedback on the initial breakage caused by wifi reconfig May 26 18:30:48 net1 -> ap -> net2 -> uplink May 26 18:30:52 ufff May 26 18:30:54 net2 is the wan zone May 26 18:31:10 so a guest network on net1 will ant to block net1 access May 26 18:31:37 but right now fw3 require you to define the proto ipv4 ipv6 ipv46 icmp .... May 26 18:31:47 we need a * proto match May 26 18:32:07 will fix that the coming days May 26 18:32:11 can you not just skip proto entirely? May 26 18:32:17 wrote piles of code for udevmand today May 26 18:32:26 stintel: infact you cant May 26 18:32:34 which is what i wanna fix May 26 18:32:47 so udevmand does topology mapping May 26 18:33:06 and in the next iteration i will add that you can add rules for the macs it finds May 26 18:33:28 sounds all very cool May 26 18:33:34 so we will have /etc/clients/${mac}.json May 26 18:33:43 wish I would be more active and able to test May 26 18:33:55 within we have unique psk/vid, parental control, fw3, qos .... May 26 18:34:13 stintel: still love ya even if you are slagging :P May 26 18:34:45 hahah May 26 18:35:02 i just recalled sitting in a taxi drunk as fuck in praque with you May 26 18:35:05 :D May 26 18:35:23 well let's plan this once the covid shite is behind us May 26 18:35:31 oh dear, once the zombie apocalypse is over we need to #replay that one May 26 18:35:38 *snap May 26 18:36:02 for now it is single dad, home schooling, hacking for me :) May 26 18:36:18 sounds like an interesting combination May 26 18:36:42 for me it's just "single" again since recently May 26 18:36:47 my parents, their best friends and my neihbours are high risk May 26 18:37:00 high risk for what? covid? May 26 18:37:14 so i have 6 70+ folks and a 7 year old that i do all the supply runs for right now May 26 18:37:25 cancer patients, lung whatever May 26 18:37:44 i am currently doing all the shite for 6 oaps and a kid :-) May 26 18:37:48 geez May 26 18:37:52 its fun May 26 18:37:59 I barely manage for myself May 26 18:38:02 day 78 if you wanna know May 26 18:38:14 good thing there's a 24/7 shop around the corner that always has alcohol May 26 18:38:14 dude not highly functional ? May 26 18:38:26 not for me May 26 18:38:32 clean 6 years, dry 2 May 26 18:38:47 proteins and my gym room ... May 26 18:39:04 admirable, keep it up May 26 18:39:12 f'yah May 26 18:39:19 its easy May 26 18:39:26 you just need to want it May 26 18:39:30 or have a kid ;) May 26 18:39:43 yeah I bet that helps May 26 18:39:48 it does May 26 18:39:56 is there any source for udevmand available? I got interested just by seeing the name, trying to find a replacement for some hacked together udev-replacement I am currently using (for managing modems) May 26 18:40:18 shit hits the fan real quick if you get into neurol network dev May 26 18:40:32 kristrev: hey dude May 26 18:40:39 kristrev: gimme 2-3 weeks May 26 18:40:45 it will solve plenty issues May 26 18:40:51 blogic: Thanks, looking forward May 26 18:41:05 been spending my spare cycles prio 1 on this one May 26 18:41:55 spare cycles sounds like a limite resource ;) May 26 18:42:12 kristrev: amazingly i have been pretty good on that May 26 18:42:38 thx to my 5hr sleep requirement May 26 18:42:50 its all about structure and plannign :-D May 26 18:42:51 blogic: I am impressed, my impression is that kids eat into the spare cycle budge. Though, I guess they go to sleep ... sometime :) May 26 18:42:54 And that May 26 18:43:15 kristrev: its a challenge but here is skype and the grannies May 26 18:43:19 ;) May 26 18:43:34 one being a retired teacher handling the home schooling May 26 18:44:12 Thanks to my building suddenly being moved to the front of queue for the garbage truck, I am also on forced five hours of sleep. Not sure about the productivity in the first three-four hours of the day though May 26 18:44:14 stintel: i also started tinkering a conf file tomcat style template parser with uci backend May 26 18:44:37 tomcat style? I read this as xml and shiver May 26 18:44:48 stintel: that would obselete 90% of our init.d crap May 26 18:44:49 Lucky you. I dont have kids, but I could hear the relief from all my friends with kids when schools opened here May 26 18:44:59 stintel: witout the xml May 26 18:45:17 yaml? toml? May 26 18:45:18 kristrev: its actually a blessing to spend that much time with her May 26 18:45:28 stintel: own syntax May 26 18:45:40 ужас May 26 18:45:56 I'm not a big fan of "own syntax" honestly May 26 18:45:59 <% uci_for each wireless wifi-device%> May 26 18:46:46 kristrev: i got a mail 4 days ago from MTK May 26 18:47:01 they reused your unilec emmc sysupgrade stuff May 26 18:47:12 and sent patches to baseline it for all mtk targets May 26 18:47:17 need to test/merge it May 26 18:47:37 block2mtd is just not that great May 26 18:47:54 blogic: Sounds good with your kid. It is what I hear here too. That it might be frustrating and hard to focus on work meetings, etc., etc., but in the end they have never spent so much time together and it is great May 26 18:48:19 And you are kind of force to find nice, local activities to do together. The sale of tents, canoos and all sorts of outdoor equipment has exploded May 26 18:48:49 i bought a wood horse for the garden, does that count ? May 26 18:49:00 she got named Luna May 26 18:49:06 why ever May 26 18:49:12 Hehe, cool, thanks for letting me know. I have been meaning to test myt mt7623 board with 5.4 for a while, but have not found the time yet. So much stuff going on May 26 18:49:21 Luna? Not Lua? ;) May 26 18:49:27 lulz May 26 18:49:57 right, gonna carry her from y to her bed now and hit the sack myslef May 26 18:50:18 I believe anything outdoor counts, or indoor for that matter May 26 18:50:22 good night May 26 18:50:24 :) May 26 18:50:41 kristrev: the whole house turned into a playground ... its insane May 26 18:51:00 hahaha May 26 18:51:06 nice chat ... May 26 18:51:31 well, I can think of worse things to have your house turned into than a playground :) May 26 19:56:29 hm May 26 19:56:31 opmini.o perlmini.o gv.o toke.o perly.o pad.o regcomp.o dump.o util.o mg.o reentr.o mro_core.o keywords.o hv.o av.o run.o pp_hot.o sv.o pp.o scope.o pp_ctl.o pp_sys.o doop.o doio.o regexec.o utf8.o taint.o deb.o universal.o globals.o perlio.o perlapi.o numeric.o mathoms.o locale.o pp_pack.o pp_sort.o caretx.o dquote.o time64.o miniperlmain.o -lpthread -lgdbm -ldb -ldl -lm -lcrypt -lutil -lc -lgdbm_compat May 26 19:56:31 ./miniperl -w -Ilib -I. -Idist/Exporter/lib -MExporter -e '' || sh -c 'echo >&2 Failed to build miniperl. Please run make minitest; exit 1' May 26 19:56:31 Attempt to free unreferenced scalar: SV 0x18adec0. May 26 19:56:35 Failed to build miniperl. Please run make minitest May 26 19:56:37 make[4]: *** [makefile:365: lib/buildcustomize.pl] Error 1 May 26 19:56:39 make[4]: Leaving directory '/home/dwmw2/git/lede/master/build_dir/hostpkg/perl/perl-5.28.1' May 26 20:03:34 gcc-10? :p May 26 20:04:29 https://github.com/openwrt/packages/issues/12262 May 26 20:04:38 Arch user? :P May 26 20:04:58 yeah, just updatred to Fedora 32 May 26 20:15:23 that fixes it; thanks May 26 20:25:57 hm, "etc/config/luci seems to be corrupt, unable to find section "main"" May 26 20:26:01 where's that supposed to come from in a new image? May 26 20:26:37 that can happen if rpcd is not running or crashing May 26 20:27:24 on this occasion it really has no 'main' section May 26 20:27:30 root@OpenWrt:/etc/config# cat luci May 26 20:27:31 config internal 'diag' May 26 20:27:31 option dns 'openwrt.org' May 26 20:27:31 option ping 'openwrt.org' May 26 20:27:31 option route 'openwrt.org' May 26 20:27:50 that looks broken indeed, then May 26 20:28:17 I did a minimal install, which lacked luci. Then built somethimg more useful and sysupgraded to it May 26 20:28:27 It's funny how both old and new platforms break May 26 20:28:58 dwmw2_gone: https://github.com/openwrt/luci/blob/master/modules/luci-base/root/etc/config/luci May 26 20:29:11 try that one May 26 20:30:37 that works better; thanks. May 26 21:00:37 dwmw2_gone: can you chime in on my patch to harden string functions in firewall3? thanks May 26 21:13:11 jow: do you keep track of our CDN "credits"? I would write them a mail if we can get more, or the CDN is dead within 10 days... May 26 21:13:21 aparcar[m]: no May 26 21:13:27 I didn't know there are credits May 26 21:14:31 philipp64|laptop: some of your fw3 string hardening changes just look like ocd May 26 21:14:46 e.g. sprintf(lodev.name, "lo"); -> strlcpy(lodev.name, "lo", sizeof(lodev.name)); May 26 21:15:04 we got 2k credits for free (equal to $2k), since download.pl uses sources.cdn.openwrt.org the credits vanish May 26 21:15:27 if you insist on ensuring that a three byte literal fits into a much larger buffer, I'd prefer snprintf() over strlcpy() May 26 21:15:28 Worst case the CDN dies and download.dl falls back to sources.openwrt.org, so no big deal May 26 21:17:19 philipp64|laptop: another thing, please don't move variable declarations into the function, I prefer to have them at the top May 26 23:03:36 New speed test. https://speed.cloudflare.com/ May 26 23:25:55 mmh, download only May 26 23:32:05 PtitGNU yeah I hope thay ad a uploade test soon. May 26 23:40:56 I use GoogleFiber's Speedtest for most things May 26 23:41:24 http://speedtest.googlefiber.net May 26 23:41:53 Or more specifically, atlanta.speedtest.googlefiber.net because I want consistant readings May 27 00:07:07 not accurate for me.. and the website says "Speed to: Raleigh-Durham".. I don't even know in which country it's located May 27 00:07:31 The US May 27 00:07:34 North Carolina May 27 00:08:04 Speedtest.googlefiber.net will send you to one of multiple locations.. I've seen Kansas City, Atlanta and now I guess RDU May 27 00:08:34 But, close or not, they are a Tier 1 backbone provider, so it should be good regardless May 27 00:08:46 Certainly better than speedtest.net May 27 00:08:52 ahah, okay :) (I thought it was in India ^^) May 27 00:09:10 and it does upload speed ;) May 27 00:10:06 but if it's the closest according to google, I think they don't have any servers in europe for this speedtest.. May 27 00:10:31 Doubtful, true.. May 27 00:10:45 I don't think they are dropping Fiber service overseas yet May 27 00:11:10 Or even in the US if their fail in Louisville is any indication May 27 00:11:26 The google one givs my ping as 109 ms in the uk May 27 00:11:27 Dug up all the streets and then decided to pull service May 27 00:11:38 I use speedtest in his CLI version here (speedtest-cli) and it's rather accurate here May 27 00:11:47 That goes to speedtest.net May 27 00:11:54 The speed one gives me 11 ms May 27 00:11:54 the python script May 27 00:12:11 yep but they have a lot of servers in .eu May 27 00:12:14 *nod* May 27 00:13:26 Tapper: 106ms here in .be, I win :p May 27 00:13:27 Does anyone have a way to extract a dtb from an ELF boot bin? May 27 00:14:20 no idea sorry May 27 00:14:54 This device is.. frustrating.. but if I can get it to work, it'll be worth it May 27 00:15:06 But they only made a few hundred :( May 27 00:21:21 has a dual-core 1Ghz 64-bit processor, 1GB ram and 4GB onboard storage.. no wifi and 3 1000GbE ports May 27 00:21:47 but absolutely no support anywhere.. and Marvell/Cavium is useless to try and contact **** ENDING LOGGING AT Wed May 27 03:04:13 2020