**** BEGIN LOGGING AT Mon Jul 15 03:01:15 2019 Jul 15 04:52:02 hey guys, anyone about Jul 15 05:32:51 any way I can verify a built bin file's package contents? Jul 15 06:35:25 Hauke: is puma7 still used ? Jul 15 06:41:01 seems so, according to wikidevi some device got FCCed in 2019 Jul 15 06:43:21 ok Jul 15 06:58:33 any idea what might be reader#1 in this context `SIGSEGV to reader#1 for invalid read access from 77a87d90` ? Jul 15 06:58:49 FS#2201 Jul 15 07:01:49 I write bug free software :) so I've never used core dumps on OpenWrt, makes me wonder if setting `ulimit -c unlimited` is enough to enable them, as I can see, that sysctl.d/10-default.conf sets coredumps to /tmp already Jul 15 08:04:48 ynezz: don't know if the description in 7765e442d04e4c19690f81084a9726776aea8b76 would help you ? Jul 15 08:17:03 ynezz: ulimit -c unlimited is usually enough Jul 15 08:17:13 ynezz: assuming "normal" software (no suid etc.) Jul 15 08:18:23 ynezz: gdb-server works well too (gdbserver :9000 crashing-command args... + ./scripts/remote-gdb 192.168.1.1:9000 ./build_dir/target-foo/crashing-command-1.2.3/src/crashing-command) Jul 15 08:20:02 i just received a M.2 wifi card Jul 15 08:20:07 and they are really tiny Jul 15 08:20:28 and the antenna plugs are even tinier Jul 15 08:20:38 yeah, always makes me sweat to connect them Jul 15 09:04:57 blogic: smaller than u.fl? Jul 15 09:20:31 Hi people. Jul 15 09:20:53 Why is openwrt wanting to switch from github to GitLab ? Jul 15 09:21:18 It makes no dif to me but... Jul 15 09:23:20 ldir: the git grep consensus seems to be Config.in syntax is "depends on BROKEN" Jul 15 09:45:58 ldir: nice, thanks! Jul 15 09:48:35 jow: yeah, gdb server would be nice, but the human reporting FS#2201 is barely able to produce usable bug report... Jul 15 09:49:10 Tapper: https://openwrt.org/meetings/hamburg2019/start#defragmentation_of_code_and_tools Jul 15 10:06:34 Tapper: the fragmentation of the tools is just one pain point, another paint point is missing CI, then some of us simply don't like promoting/using of this proprietary platform (which can change it's usage terms/UI/whatever tomorrow without any further notice), we can't bend it to our workflows, we bend our workflows to GitHub's, we can't for example disable that crappy web editing feature offered by Jul 15 10:06:40 GitHub (producing very low quality contributions), OpenWrt as FS/OSS project should promote/use/help develop other OSS/FS projects etc. Jul 15 10:07:16 that's what I remember from the discussion in Hamburg Jul 15 10:08:57 Chiming in from the side: indeed github's current ownership is already kind of bad point in itself, but ofcourse all the practical stuff needs to be.. well.. practical Jul 15 10:10:17 OK thanks. Jul 15 10:10:19 if the GitLab idea pans out in the end, we'll probably still keep GitHub as one of the sources for possible contributions, there's no plan to ditch it completly AFAIR Jul 15 10:10:29 * Tapper nods Jul 15 10:11:44 And indirectly there ought to still be hundreds of github repositories with "random" package contributors anyway Jul 15 10:12:06 we're not talking about packages feed BTW Jul 15 10:12:27 just the master tree and stuff around Jul 15 10:14:07 for the start I mean, in order to evaluate the GitLab we just plan to setup playground with the master tree and stuff around Jul 15 10:14:20 yep, but point was still that no need or desire to be hostile either towards github :) Jul 15 10:14:47 hm, what is hostile towards GitHub? Jul 15 10:18:25 ynezz I think they meen pushing people to use one over the other. Jul 15 10:19:02 well term in broad sense and no specifics... but example would be "oh you put code into github for your stuff, get lost" ;D Jul 15 10:24:11 nope, the overall idea is (we don't know yet if it's going to work) to have GH<->GL bridge, so PR created on GH would get mirrored to GL, CI would run on GL, propagate the results back to GH PR as comments (or whatever else is available there) and the PR would be handled by devs on the GL side only Jul 15 10:26:09 russell--: its like a u.fl but about 25% of the size Jul 15 10:27:06 mhf4 ? Jul 15 10:29:53 blogic: wow Jul 15 10:30:06 fiddly stuff - nothing for me Jul 15 10:30:53 yeah, connect once, glue and pray Jul 15 10:31:28 blogic: i assume it is called pico.fl ;) Jul 15 10:32:24 does it produce a satisfying crunch as you break it? Jul 15 10:32:33 it's called mhf4 (0.81mm), u.fl is 2.0mm Jul 15 10:33:34 stereo dissecting microscopes, ftw Jul 15 10:33:39 so with my age/eyesight combination we're talking about something I can't see vs something I can barely see Jul 15 10:38:51 u.fl it is Jul 15 11:11:40 hi all, trying to solve my cifsd/restart issue, what procd sequence is luci/service/restart triggering? Jul 15 11:15:25 For some reason manually doing "stop" + "start" via luci works, while "restart" breaks the service? Is there a extra step involved in restart? Jul 15 11:16:39 Ah need to quicker hit start/stop than i also breaks... so seems tobe a timing issue? Jul 15 11:49:44 Good morning. I have built an image of the packages but the file is very small compared to the ones that I flash off the OpenWRT website. How may I check the contents of the package? Jul 15 11:50:13 The image manifest doesn't show all the packages I have selected to build, even though "make" show that all were indexed and installed Jul 15 11:54:59 Did you maybe select them as (m) package, instead of (*) ? Jul 15 11:55:53 Yes Jul 15 11:56:03 I chose (m) so that the dependencies are selected as well Jul 15 11:56:08 If you want the package included into the image file, you need to select them as (*) otherwise they go into the /packages folder for later installation Jul 15 11:56:32 Oh my goodness. No wonder I've bricked my router twice already xD Jul 15 11:59:08 should not brick the router, all base packages are (*) by default, just hit space twice so you see (*) than the extra packages are included in the image/sysupgrade file Jul 15 11:59:36 Do I need to go to dev-special and select gcc? Jul 15 11:59:54 for what? Jul 15 12:00:31 By the way, with the current set up I am not able to get over 300Mbps, out of a 400 Mbps connection. I read there's some sort of Number Generator or a package that allows me to get my full thoroughput. PS: I'm able to get it on stock f/w Netgear Jul 15 12:01:03 no clue i use wired mostly :p Jul 15 12:02:03 I am wired too Jul 15 12:02:05 Not wifi Jul 15 12:06:35 Hi! Could someone grant me permission on patchwork? Jul 15 12:12:56 blogic: ^ Jul 15 12:15:05 gch981213: ath79: add support for Qihoo C301: package/boot/uboot-envtools/files/ath79: qih... < yunco... Jul 15 12:17:12 pepe2k: er...what? Jul 15 12:18:00 gch981213: alphabetical order, q before y Jul 15 12:18:28 gch981213: ath79: add support for Qihoo C301: ar9344_qihoo_c301.dts: whitespace vs. tabs in &builtin_switch node Jul 15 12:20:23 pepe2k: package/boot/uboot-envtools/files/ath79: That's a case entry starting with "buffalo,bhr-4grv2" Jul 15 12:21:19 pepe2k: Thanks for spotting that whitspace problem :) Jul 15 12:22:22 gch981213: ah, sorry for the noise then, not enough lines in a diff :) Jul 15 12:24:31 gch981213: whats your username in patchwork ? Jul 15 12:25:01 blogic: it's "981213" Jul 15 12:27:53 blogic: while you are at it, you could do the same for me Jul 15 12:27:59 username should match IRC handle Jul 15 12:28:07 Hi, after a successful compilation, how can I close again an image after a change to the rootfs? Jul 15 12:50:41 I'm trying to build just an openconnect package - why would I get "*** No rule to make target 'package/openconnect/install'. Stop." Jul 15 12:51:20 it builds fine with 'make package/openconnect/compile', but I just can't do the next stage to build the ipk (and yes, in 'make menuconfig' openconnect is selected 'M') Jul 15 12:59:05 what is your command Jul 15 13:00:12 have you selected it with y to make a * Jul 15 13:00:24 I recently found out that M won't build the package in to an image Jul 15 13:00:32 Just compile and index it for ready/available Jul 15 13:05:18 gch981213: and what email addr ? Jul 15 13:05:23 blocktrron: ^^ same for you Jul 15 13:05:37 blogic: gch981213@gmail.com Jul 15 13:06:20 i actually knew that :-) Jul 15 13:07:05 blogic: mail@david-bauer.net Jul 15 13:12:20 mail sent Jul 15 13:16:10 ynezz: do you know how to add the Makefile variable PYTHON3? PYTHON exists but points to 2.x Jul 15 13:16:45 neonknight77, I'm not trying to build it into the image, and so it seems I don't need the 'install' step as it's already built the ipk within the 'compile' step yay :) Jul 15 13:18:53 ynezz: never mind I jus tadded it via "#!/usr/bin/env python3" Jul 15 13:19:59 blogic: Thanks! Jul 15 13:32:48 aparcar[m]: hm, there shouldn't be any PYTHON3 make variable, just PYTHON:=python, which translates into staging_dir/host/bin/python symlink pointing at the correct Python version on the host Jul 15 13:38:09 ynezz: thanks! Jul 15 14:19:09 jow: do you mind creating a releases.json files as (newly) explained here https://github.com/openwrt/openwrt/pull/2192 Jul 15 14:25:38 aparcar[m]: hey, wouldyou find a moment to test my sysupgrade patch, please? Jul 15 14:30:21 rmilecki: hey, sorry for the latency. I'm now in Berlin to work on other stuff but will try to test it tonight when back home Jul 15 14:30:33 aparcar[m]: ok, thanks Jul 15 14:32:50 ynezz: can you please have a look again at https://github.com/openwrt/openwrt/pull/1992 so we can move things forward? would be a shame if 19.07 is again without firmware signatures Jul 15 15:12:18 hmm... is there a strongswan-mod .. etc. package bundling chacha20poly1305 ? I tried a few, but chacha20 is not in the list of supported encryption algorithms. I'm using 18.06 Jul 15 15:18:53 aparcar[m]: i didn't know about that signature check pull request Jul 15 15:19:12 aparcar[m]: i think it could be integrated nicely with my plan for refactoring sysupgrade process Jul 15 15:41:37 aparcar[m]: I'd merge the shell script with the python script Jul 15 15:41:48 the menuconfig symbol name is too generic (CREATE_JSON) Jul 15 15:42:32 the make target name is not ideal either Jul 15 15:42:43 what about "make imageinfo" as make targert Jul 15 15:45:15 and the symbol config IMAGE_INFO_FILES Jul 15 15:50:43 yanosz, according to https://github.com/openwrt/packages/blob/master/net/strongswan/Makefile "chapoly" isn't included :< Jul 15 15:52:02 yanosz, here is the list of them all: https://wiki.strongswan.org/projects/strongswan/wiki/PluginList Jul 15 16:57:36 aparcar[m]: I've merged PR#1992 like a week ago into my staging tree, but didn't managed to test it yet Jul 15 17:01:29 aparcar[m]: it doesn't look like 19.07 material, it would be a nice goal if we could manage to prepare and test signed firmware images for next release, and enable image signature verification on a few targets Jul 15 17:03:13 (by default I mean) Jul 15 17:27:59 yanosz: it's not that hard to include them in strongswan, but you'll need to create a kmod for it as well (see https://github.com/cotequeiroz/openwrt/commit/a16a8d47d335d3d00dee3aa54cd5bb95d292e415) Jul 15 17:28:42 Here's an old attempt of mine at the strongswan package https://github.com/cotequeiroz/packages/commit/002bc123077e8f00ee6c2935f88624dcb120a448. I haven't updated it, so it may need adjustments. Jul 15 17:31:41 jow: thanks for the feedback, what to you mean with "make targetinfo"? Like targetinfo instead of jsondevicemerge? Jul 15 17:32:12 also lynxis suggested to use per target files, as it would likely fill up devices with small ram Jul 15 17:32:21 we have ~1400 devices right Jul 15 17:32:46 I also think it's easier to work with files for every subtarget as we build every subtarget on it's own Jul 15 17:42:43 lynxis: I'll update you once the script is updated Jul 15 19:46:07 cotequeiroz: https://forum.openwrt.org/t/dev-crypto-no-ciphers-or-hashes/40736 Jul 15 20:16:59 cotequeiroz - thank's, I'll have a look at it later. Jul 15 20:17:15 ... just didn't expect to compile a kernel module for it. Jul 15 20:17:40 Did you benchmark cha2020 on some devices (w.r.t. aes) Jul 15 20:17:42 = Jul 15 20:17:44 ? Jul 15 20:17:58 chacha20 ... args Jul 15 20:31:44 Anyone on have experience with ath10k devices not supporting MTU over 1500, even though set with netifd / ip link? Jul 15 20:31:56 (or at least net seeming to support it) Jul 15 20:32:42 . /sys/devices/platform/soc/40000000.pci/pci0000:00/0000:00:00.0/0000:01:00.0/net/mesh0$ cat mtu Jul 15 20:33:19 shows the "right number", but transport over the link fails as soon as the packet size gets too large Jul 15 20:34:40 802.11 should support 2304 MTU, "2048 payload plus a possible header of up to 256 bytes" as the legend has it Jul 15 20:35:17 mangix: It looks like he does not have any hw-crypto, but he was assuming that he did because of the /dev/crypto presence. I'll reply there. Jul 15 20:35:47 No "mtu" in hostapd.conf Jul 15 20:36:56 (IPQ4019) Jul 15 20:36:57 jeffsf: not saying they don't exist, but rarely wifi devices really support mich larger than 1500, is my experience Jul 15 20:38:08 The open-mesh people (batman-adv) specifically call out setting `option mtu '1536'` Jul 15 20:38:39 and the compiled-in error message when the MTU is 1500 is "Setting the MTU to 1532 would solve the problem." Jul 15 20:38:59 Might have to poke BG on it Jul 15 20:40:06 http://lists.infradead.org/pipermail/ath10k/2016-October/008621.html Jul 15 20:40:15 "It seems it accepts up to a value of 2300 by configuration, Jul 15 20:40:15 but practically the limit is 1595 as the thread commented." Jul 15 20:40:32 ..tough like you said, 802.11 should support said values indeed Jul 15 20:40:40 yanosz: there are two parts of strongswan that use crypto. The IPsec itself is handled by the kernel (the data travels there), while strongswan itself will handle IKE (basically, IPsec negotiation). See https://wiki.strongswan.org/projects/strongswan/wiki/IntroductionTostrongSwan Jul 15 20:42:30 So, IKE can be encrypted with chacha20-poly1305 with the userspace strongswan module, but to be able to use it for your actual traffic, you need the kernel module. Jul 15 20:44:33 The kernel module was crashing my server (x86_64), so I haven't used it for a while now, and won't have a useful benchmark for you, sorry. Jul 15 21:26:53 cotequeiroz: don Jul 15 21:28:12 't remember if i asked you, but did I show you this? https://gitlab.labs.nic.cz/turris/turris-os-packages/issues/398 It seems OPENSSL_PKCS1_OAEP_PADDING doesn't work right with cryptodev Jul 15 21:29:25 I'm leaving right now, but will take a look at it tomorrow. Jul 15 23:30:40 Is anyone else seeing `NOHZ: local_softirq_pending 08` in their logs? Jul 15 23:49:59 hmm, I have faint memories of that, but not on OpenWrt and pretty much a decade ago... Jul 15 23:50:20 Yea, found some stuff from Linux 3.x on it, but not much since Jul 15 23:50:42 Might be batman-adv Jul 15 23:51:20 mac802154 patch back in 3.6 Jul 15 23:51:44 my memories might have still been on kernel 2.6, really a long time ago Jul 15 23:52:04 Yeah, I've got a few ugly memories of Samsung kernels from back then Jul 15 23:52:33 They make Broadcom's open-source support seem fantastic Jul 15 23:54:02 don't even dare to think that (in particularly considering that Samsung has often used broadcom wlan/ bt/ nfc for their phones) Jul 15 23:54:35 I had nightmares around trying to fix backlight under their touch-sensitive Android keypad Jul 15 23:55:15 Finally just shut the thing off, as I recall Jul 15 23:55:58 yeah, adaptive backlight is ... Jul 16 00:53:48 Broadcom does open source? Jul 16 00:54:08 since when?! Jul 16 00:55:33 SOC support and wired ethernet tends to be pretty good, fullmac wlan support (brcmfmac) is also reasonably active. anything beyond that... Jul 16 01:02:59 * hellsenberg points at some cursed broadcom docsis router SoC, then at some other cursed broadcom SoC inside some paid tv player Jul 16 01:03:18 don't think there's much stuff for these... **** ENDING LOGGING AT Tue Jul 16 03:02:17 2019