**** BEGIN LOGGING AT Wed Mar 27 02:59:57 2019 Mar 27 04:22:15 one hundo packages woo Mar 27 06:41:18 rmilecki: please comapre "fw3 print" on both 4.4 and 4.4.1 builds Mar 27 06:41:38 to me it looks as if there's simply no netfilter kmods loaded on 4.4 Mar 27 06:42:49 jow: trying it now Mar 27 06:42:54 ynezz: that PR #1944 escalated fast ;) Mar 27 06:44:21 jow: ok, interesting.... "fw3 print" is empty with 4.4 (fast routing kernel) Mar 27 06:44:58 jow: does it mean I was seeing fast routing because kernel wasn't doing any network traffic filtering? Mar 27 06:48:01 at least I'd expect a somewhat lower overhead due to no conntrcking Mar 27 06:48:12 how mzch in practise I don't know Mar 27 06:48:28 but should be easiy to verify by completely disabling gw3 on boot on your 4.4.1 kernel Mar 27 06:48:36 *fw3 Mar 27 06:53:26 /etc/init.d/firewall stop ? Mar 27 06:53:31 disable Mar 27 06:53:52 but during runtime a simple fw3 stop; fw3 flush should suffice Mar 27 06:54:08 it will not unload netfilter kmods though, which might also introduce overhead simply by being loaded Mar 27 06:54:45 jow: i have modules loaded with 4.4 Mar 27 06:54:49 ok Mar 27 06:54:53 in teresting Mar 27 06:55:18 jow: bingo... Mar 27 06:55:37 oh, wait, i'm on wrong kernel Mar 27 06:55:40 testing again Mar 27 06:57:01 jow: no, I was wrong... "lsmod" shows nothing on 4.4 Mar 27 06:57:21 jow: there are modules in the /lib/modules/4.4.0/ just not loaded Mar 27 06:57:47 another quick way to check is cat /proc/net/ip_tables_names, /proc/net/ip_tables_matches, /proc/net/ip_tables_targets Mar 27 06:57:55 root@LEDE:~# insmod /lib/modules/4.4.0/crc-ccitt.ko Mar 27 06:57:57 Failed to find the folder holding the modules Mar 27 06:57:59 fw3 uses these files to probe which netfilter facilities are available Mar 27 06:58:26 rmilecki: please strace it. I could imagine that its an uname() discrepancy Mar 27 06:58:41 either due to CONFIG_LOCALVERSION or 4.4 vs. 4.4.0 Mar 27 06:59:08 insmod derives the name from uname() Mar 27 06:59:47 specifically from the "release" field of "struct utsname" as filled by uname(2) Mar 27 07:00:30 maybe we have a packaging bug and never noticed because there rarely are .0 kernels Mar 27 07:03:24 jow: https://pastebin.com/wqnGDH6s Mar 27 07:04:49 jow: oh, it's about a path Mar 27 07:04:56 "4.4" dir vs. "4.4.0" Mar 27 07:05:16 stat64("/lib/modules/4.4/", 0xbe8699a8) = -1 ENOENT (No such file or directory) Mar 27 07:05:42 it's possibly sometihng that's long fixed in the master Mar 27 07:05:53 please remember I'm testing old OpenWrt commit that is capable of building 4.4 Mar 27 07:06:30 ah ok Mar 27 07:06:45 well in this case ln -s /lib/modules/4.4.0 /lib/modules/4.4 should suffice Mar 27 07:08:38 rmilecki: do you use an external kernel tree? Mar 27 07:08:47 jow: not currently Mar 27 07:09:27 rmilecki: the issue might still be present in master then Mar 27 07:09:29 jow: ok, doing "ln -s" fixed loading modules, fixed fw3 and "broke" my routing/NAT performance Mar 27 07:09:54 is any X.0.0 kernel calling itself X.0 in uname? Mar 27 07:10:10 e.g. 4.4 instead of 4.4.0 Mar 27 07:10:35 rmilecki@elitebook:~> uname -r Mar 27 07:10:37 5.0.0-rc1+ Mar 27 07:10:41 it was probably changed later Mar 27 07:10:45 hm okay Mar 27 07:33:12 anyone with a ubifs using device here? Mar 27 07:34:07 jow: what do you need? Mar 27 07:35:19 rmilecki: wanted to know if /sys/sys/devices/virtual/ubi/ubi*/ubi*_*/name is null terminated Mar 27 07:35:27 or newline terminated Mar 27 07:35:29 se Mar 27 07:35:31 sec Mar 27 07:35:48 currently looking into https://bugs.openwrt.org/index.php?do=details&task_id=2196 and reviewing other strncmp() calls in fstools Mar 27 07:36:00 there's some places where we shouldn't use it like that Mar 27 07:36:06 flashing Mar 27 07:36:21 I think ubi_volume_match() is also prone to confusing volumes Mar 27 07:36:24 i didn't have that file on ancient 4.4 ;) Mar 27 07:36:49 if we're looking for "foo" and there is a volume with name "foo2", fstools would wrongly attempt to use it Mar 27 07:37:22 due to strncmp("foo", value_from_sysfs, strlen("foo") + 1) Mar 27 07:37:53 root@OpenWrt:/# ls /sys/sys/devices/virtual/ubi/ubi*/ubi*_*/name Mar 27 07:37:55 ls: /sys/sys/devices/virtual/ubi/ubi*/ubi*_*/name: No such file or directory Mar 27 07:38:08 ah, doble /sys/ Mar 27 07:38:24 root@OpenWrt:/# hexdump -C sys/devices/virtual/ubi/ubi0/ubi0_0/name Mar 27 07:38:26 00000000 72 6f 6f 74 66 73 0a |rootfs.| Mar 27 07:38:27 00000007 Mar 27 07:38:58 there seems to be always 0x0a Mar 27 07:39:00 root@OpenWrt:/# hexdump -C sys/devices/virtual/ubi/ubi0/ubi0_1/name Mar 27 07:39:01 00000000 72 6f 6f 74 66 73 5f 64 61 74 61 0a |rootfs_data.| Mar 27 07:39:03 so newline, thanks! Mar 27 07:39:16 np Mar 27 07:44:11 jow, btw: root@OpenWrt:/# uname -r Mar 27 07:44:12 4.14 Mar 27 07:44:14 root@OpenWrt:/# ls /lib/modules/ Mar 27 07:44:15 4.14 Mar 27 07:44:17 anyway, it works Mar 27 08:18:05 dedeckeh: ping Mar 27 08:31:24 jow:pong Mar 27 08:40:55 dedeckeh: are there some hardcoded limits in odhcpd that prevent handing out prefixes larger than a /61 ? A user in the forum wondered. His client is requesting a /60 but only getting a /61 in reply Mar 27 08:43:44 jow: there're ni hard coded limits in odhcpd; odhcpd just takes the prefix size which is put into place by netifd Mar 27 08:45:35 jow:ip6assign in network interface config allows to tweak the prefix size Mar 27 08:46:42 https://www.youtube.com/watch?v=3dfYcQ_r_x8 Mar 27 09:26:20 Good morning Mar 27 09:26:23 anyone knows where dsl logs are stored? Mar 27 09:27:12 dsl logs? not in logread? Mar 27 09:27:23 cant find any Mar 27 09:27:29 not in logread Mar 27 09:27:54 I have noticed my internet connection goes down often, but not the local one Mar 27 09:28:16 so I would like to keep an eye on the pppoe disconnection/ reconnection frequency Mar 27 09:28:45 I can pass you logread if you want to see Mar 27 09:29:11 well /var/log/ would be next best bet, I'd say without knowing any better that if not even there, then there is none Mar 27 09:29:32 :S Mar 27 09:29:54 Ill give another look at it, but couldnt spot anything useful in there either Mar 27 09:30:22 there is rare occasions an thing in openwrt writes to elsewhere than standard log, which equates to then logread... and even those services tell something in logread Mar 27 09:30:54 easy test is to logread -f and yank the phonecable out and see Mar 27 09:32:57 will do that thanks :) Mar 27 09:38:41 dedeckeh: hm, the user claims that ip6assign is 60 Mar 27 09:38:42 https://forum.openwrt.org/t/odhcpd-gets-wrong-prefix-from-dhcpv6-solicit-messages/34130 Mar 27 10:19:02 are there any sane nl80211 apis to query the restrictions for a particular channel without actually switching to that channel? Mar 27 10:19:39 my main problem is that switching to certain channels will clamp down the configured txpower, which is not so ideal from a ui ov Mar 27 10:19:43 *ui pov Mar 27 10:21:09 jow:for downstream prefix delegation odhcpd picks a subprefix; it won't just give the /60 to the downstream client as it would not be able to delegate prefixes to other clients Mar 27 10:21:28 if they request for it Mar 27 10:21:36 is there a way to disable ipv6 dns ? Mar 27 10:21:48 one of my voip phones is not registering Mar 27 10:21:59 * karlp grins Mar 27 10:22:00 turns out its getting a ipv6 ip for sipgate.de Mar 27 10:22:20 and the phone is on a ipv4 subnet Mar 27 10:22:36 karlp: at me ? :-) Mar 27 10:22:39 broken software, broken software everywhere :) Mar 27 10:22:51 thats its default state right ? Mar 27 10:23:11 blogic:you can set ra_dns to false Mar 27 10:23:21 it's default state is, "it's easy, just some software, shouldn't be more than a week to finished" Mar 27 10:23:46 blogic:assuming the client is only doing slaac Mar 27 10:23:57 its not doinf any ipv6 Mar 27 10:24:06 sorry let me rephrase Mar 27 10:24:15 (inet) - (openwrt) - voip phone Mar 27 10:24:34 the openwrt router is doing dns resolve via dnsmasq for the phone Mar 27 10:25:08 so i want openwrt to disable v6 in dnsmasq Mar 27 10:25:25 and i have the small version istalled, like without IPV6 enabled Mar 27 10:26:42 blogic:so you received IPv6 dns servers on the wan ? Mar 27 10:27:41 no Mar 27 10:27:54 well, its pointing at the DT server Mar 27 10:27:57 using ipv4 Mar 27 10:28:21 and it uses AAA records via IPv4 ? Mar 27 10:28:25 AAAA Mar 27 10:29:14 looks like it Mar 27 10:30:36 Well the thing is that AAAA record is just a record.. those in itself are ip version agnostic thing Mar 27 10:30:37 AFAIK dnsmasq just forwards it the records; I'm not aware of any setting to drop AAAA records Mar 27 10:32:27 Usually there is something else wrong along the way, receiving a AAAA record is not the rootcause, but client believing an ipv6 address can work if its not available is the thing one needs to think Mar 27 10:33:55 olmari: correct Mar 27 10:34:05 but the image comes from a realtek SDK Mar 27 10:34:23 damnit .. build error .. checking Mar 27 10:36:51 Package kmod-nf-ipvs is missing dependencies for the following libraries: Mar 27 10:36:53 nf_defrag_ipv6.ko Mar 27 10:44:25 jow: any idea what happened with the fsf-dock buildslaves? Mar 27 11:02:49 xback: need to be repurposed, will do so later today Mar 27 11:05:53 jow: can I request a minute of your expertise regarding packages? Mar 27 11:06:07 the build error is due to package kmod-nf-ipvs Mar 27 11:06:32 xback: wneed to run now, will be back in 2hours or so Mar 27 11:06:39 ok, thanks Mar 27 11:08:14 Who is the bcm63xx mantainer? I have a fiew question to ask.... so I can ping him! Mar 27 11:10:15 hitech95: I usually poke KanjiMonster :) Mar 27 11:10:46 ynezz, thanks! Mar 27 11:12:05 Have you ever had to implement VOIP drivers? I would like to know what is the best approach ... Should I use ASOC? I have a PCM controller on the SoC and SPI+PCM codec :D Mar 27 11:22:27 xback: yeah, stable kernels bumps :) de7f08cfd53daeb893b928e9a5f9daf71ff8b95d ipvs: fix dependency on nf_defrag_ipv6 Mar 27 11:26:29 working on it Mar 27 11:26:44 the fun part here is that it doesnt occur on all targets .. Mar 27 11:26:57 doing full-kmod built on ex. cns3xxx doesnt trigger the error .. sigh .. Mar 27 11:28:51 those are fun :) Mar 27 11:38:22 Are those the clock sources for the bcm63xx ? https://elixir.bootlin.com/linux/latest/source/arch/mips/bcm63xx/clk.c If so are always are thegates all enabled? Mar 27 11:52:55 xback: seems to build here with http://paste.debian.net/plain/1074879 but I'm not sure if it's proper fix Mar 27 11:54:28 ynezz: i'm currently looking at include/netfilter.mk Mar 27 11:54:39 trying to see how other modules do it Mar 27 11:56:44 xback: yeah, that looks like a proper place Mar 27 11:57:14 xback: there was something changed regarding nf_defrag_ipv6 with kernel 4.19 Mar 27 11:58:41 Looks like that on brcm63xx is still using custom apis to manage stuff! No reset controller... Mar 27 11:59:04 hauke: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v4.14.109&id=de7f08cfd53daeb893b928e9a5f9daf71ff8b95d Mar 27 12:00:29 basically, I'm trying to find out how to properly fix this. Either by adding a dependency to the module, or if the symbol NF_DEFRAG_IPV6 should be added .. Mar 27 12:02:38 :q Mar 27 12:04:48 I think, that you can't fix it in include/netfilter.mk as ipvs is not there Mar 27 12:06:44 and adding DEPENDS += +!(LINUX_3_18||LINUX_4_9):kmod-nf-defrag-ipv6 into kmod-nf-ipvs didn't helped either Mar 27 12:12:00 brb, quick lunch Mar 27 12:20:26 devm_request_irq vs request_irq? Mar 27 12:42:48 KanjiMonster, I'm working on a SPI/PCM driver to add voip functionality on my router. I have a few question about the bcm63xx hardware and kernel. the first one is: How is the CS line choosed? How can I map the GPIOs used for IRQ without destroying my hardware? (each codec have a INT and RST line) Mar 27 12:44:19 hitech95: the gpio lines usable as irqs are fixed; there are 4 to 6 per chip Mar 27 12:45:48 KanjiMonster, I have 2 slave SPI codecs, and 1 PCI device on a 6358. Lots of LEDs but I have no idea how many GPIOs are free and not yet mapped. https://github.com/openwrt/openwrt/blob/12310f05b7b080b016ec515796be437f4cd30b62/target/linux/brcm63xx/dts/agpf-s0.dts Mar 27 12:46:27 what a hack :) Mar 27 12:46:55 ynezz, ? Mar 27 12:48:33 hitech95: are you trying to design/layout a board, or are you trying to make an existing device work? Mar 27 12:49:00 hitech95: well, I've thought, that it's some postmarket modification/hack of the device, but it seems like you're just probably trying to make existing device working Mar 27 12:49:15 KanjiMonster, Making a existing device work. Openwrt run on it fine I'm adding support for SLIC Mar 27 12:49:27 Or at least tring Mar 27 12:49:30 :D Mar 27 12:50:08 hitech95: then you need to look at the device's vendor GPL tarball; you can find the GPIO lines used there Mar 27 12:50:46 KanjiMonster, Ha ha ha ... no gpl for that! Old hardware OpenRG based! Mar 27 12:51:37 ynezz: would you have a minute for me? :) Mar 27 12:52:45 Pepe, how is going with the libre router project? Mar 27 12:52:59 Pepe: yeah, whats up? Mar 27 12:53:24 hitech95: I don't think that I'm the Pepe what you think. :( O'm not related to Libre router anyhow. Mar 27 12:53:37 ynezz: I wrote you PM, may you check that? Mar 27 12:53:50 Pepe, Oh ok :D Mar 27 12:54:58 pepe the frog? Mar 27 12:57:33 KanjiMonster, If I use the clk_get(), clk_prepare_enable(), clk_disable_unprepare() the gates/sources are set correctly or I have to make my own implementation? Mar 27 12:58:57 hitech95: they should be set correctly. the difference between the devm_* and * versions is that the former do the _put/free/etc automatically on driver removal/deattach, with the latter you need to do it manually on error Mar 27 12:59:23 so the devm_ versions are strongly preferred Mar 27 12:59:37 KanjiMonster, perfect! Working on the PCM driver right now! Mar 27 13:00:51 hitech95: please also work on upstreaming the driver; there are already too many local patches for bcm63xx Mar 27 13:01:26 KanjiMonster, AFIK there is no way to create a telephone driver... unless dandi can hook into asoc! Mar 27 13:04:28 xback: back, in case your package question is still relevant Mar 27 13:06:40 jow: yes Mar 27 14:11:42 does anyone know why there's this line https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/x86/image/Makefile;h=84a3d88a7f25c2219c5b30460ce9c06827288fd9;hb=HEAD#l97 ? Mar 27 14:11:46 dd if=/dev/zero bs=128k count=1 >> $(KDIR)/root.squashfs Mar 27 14:12:40 it looks like some additional padding and I'm wondering why the mksquashfs cant produce proper and padded images Mar 27 14:36:48 q Mar 27 15:13:24 KanjiMonster, It's going to be quite hard without the Datasheet. I'm tring to understand DMA right now! (No idea of how it works... doh) Mar 27 15:18:26 lynxis: do you know why there's this line `dd if=/dev/zero bs=128k count=1 >> $(KDIR)/root.squashfs` in https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/x86/image/Makefile;h=84a3d88a7f25c2219c5b30460ce9c06827288fd9;hb=HEAD#l97 ? Why is not mksquashfs producing already usable images? Mar 27 15:49:17 ynezz: it's fixed, just pushed Mar 27 15:49:37 special thanks to jow :) Mar 27 17:24:43 Pepe: why does home assistant need a perl port of unicode handling? what's not in python stuff natively? Mar 27 17:25:05 I was under the impression taht home assistant was a "new" application, not an old perl thing that had been partly ported to python? Mar 27 17:26:34 Yes, Home Assistant is ported to Python3, but in setup.py, there is dependency for python-slugify ( https://github.com/home-assistant/home-assistant/blob/dev/setup.py#L46) , which in it setup.py, there's dependency for text-unidecode: https://github.com/un33k/python-slugify/blob/master/setup.py#L17 Mar 27 17:30:10 no, I ddidn't mean python3 porting, I meant, why the dep on a perl api unicode helper, instead of python unicode stuff. Mar 27 17:30:13 but forget I asked :) Mar 27 17:30:41 No, no, It's fine, that you asked. I'm glad. :) Mar 27 17:31:29 However to properly answer your question, I dont know yet. I want to have all dependencies set and then I can start to look at Home Assistant. The next PR will be for python-slugify and then finally HA. Mar 27 17:34:13 HA probably doesn't care what slugify uses internally, and never looked. Mar 27 17:34:22 it's really, "hwy did slufigy need a perl api" Mar 27 17:35:09 But good finding. I will ask them. I'm curious as it as well. Mar 27 17:36:34 https://github.com/kmike/text-unidecode lists the other one as a better alternative anyway, but gpl? Mar 27 17:37:25 I have somewhere prepared Makefile for unidecode, but there's some thing, which I need to sort first before I will send PR. Mar 27 17:37:45 heh, seems it's _another_ port of the the perl library anyway Mar 27 17:37:52 should never have looked at this :) Mar 27 17:44:54 mwarning: that strongswan change is neat! Mar 27 17:46:14 stintel: indeed, but didn't know how to collapse the menu initially Mar 27 17:47:13 me neither. had been looking into it again some time ago, but got very lost in the menu code Mar 27 19:52:53 is there anyone who knows the PCM/TDM bus? Mar 27 20:12:30 jow: example of the build system not generating proper tarballs: https://github.com/openwrt/packages/pull/8513 Mar 27 21:24:39 I hate this mirror_hash junk. it's been error prone for everyone Mar 27 21:27:57 and codeload was thhe "solution" which is also not doing thigns stably? **** ENDING LOGGING AT Thu Mar 28 02:59:57 2019