**** BEGIN LOGGING AT Mon Jul 09 03:00:02 2018 Jul 09 04:54:22 morning Jul 09 08:50:58 Ola guys, how is with kmod-bridge, kmod-8021q, etc in 18.06? (https://github.com/openwrt/packages/issues/6409 I did not find email of package maintainer: https://openwrt.org/packages/pkgdata/lxc-checkconfig Jul 09 09:56:14 Ola guys, how is with kmod-bridge, kmod-8021q, etc in 18.06? -> https://github.com/openwrt/packages/issues/6409 I did not find email of package maintainer: https://openwrt.org/packages/pkgdata/lxc-checkconfig Jul 09 11:31:02 Hi. I have a quick question about OPKG: does install overwrite symbolic links, or does it overwrite the target file? Jul 09 11:31:55 good question :) Jul 09 11:32:31 so, why would disabling openvpn lead to a memory leak (afaict a kernel memory leak, since processes don't look responsible)? Jul 09 11:34:10 ryanmeulenkamp: opkg performs lstat(2) on the target path, if it exists and is not a directory, it attempts to unlink Jul 09 11:34:30 ryanmeulenkamp: if the target path exists and is a directory, it will error out Jul 09 11:34:47 disabling on ubnt-erx doesn't seem to do it, but seems pretty reliable on dir860l-b1 Jul 09 11:35:01 ryanmeulenkamp: so short answer, symlinks are "overwritten" as in unlinked and replaced with whatever happens to be in the archive (another symlink, plain file, ...) Jul 09 11:35:19 russell--: maybe it is related to flow offloading? Jul 09 11:35:32 did you check your conntrack table yet? Jul 09 11:35:40 no Jul 09 11:38:01 looks pretty empty Jul 09 11:38:05 ok Jul 09 11:40:53 jow: Thanks. This is the behaviour I was hoping for. Jul 09 12:05:19 if i "wifi down", with openvpn stopped, the leak seems to go away. Jul 09 12:05:52 bizarre Jul 09 12:08:42 clues welcome ;-) Jul 09 12:19:02 ldir, mkresin: I can look at wpad-basic. Jul 09 12:19:19 what about wpa_supplicant? Should wpa-supplicant-mini exist? Or should that be wpa-supplicant-basic? Jul 09 12:23:19 dwmw2_gone: personally I think that we do not need mini / micro / tiny variants of wpa_supplicant and hostapd as long as there are mini / micro / tiny variants of wpad Jul 09 12:23:50 if someone, for whatever reason, replaces wpad with the standalone utils then likely not to save space Jul 09 12:24:12 we already have wpa-supplicant-mini Jul 09 12:24:19 I know Jul 09 12:24:32 if we're introducing a new "basic" variant of wpad, I was just wondering if wpa-supplicant-mini would be better off renamed, for consistency Jul 09 12:24:37 ah Jul 09 12:24:41 other than that I don't really want to touch it Jul 09 12:25:03 I'd say don't rename it and eventually just trop it Jul 09 12:25:09 ok Jul 09 12:25:12 I don't think there's need for an additional variant of wpad, either - it's just confusing and stripping small but useful features out seems counterproductive Jul 09 12:25:12 that works for me Jul 09 12:25:44 the objection is that even the small addition of 11r (mostly compensated for by adding -flto) breaks the bank on some platforms Jul 09 12:25:45 s/trop/drop/ Jul 09 12:25:59 and I haven't actually added 11w either, and probably should Jul 09 12:26:11 11w was mgmt frame protection? Jul 09 12:26:14 11w's tiny and more useful than 11r, imo Jul 09 12:26:14 yeah Jul 09 12:26:16 jow: Yes Jul 09 12:26:17 my personal view...if the mini(malist) package provides the functionality then stick with that. Only if finesse is required is a basic variant required. Jul 09 12:26:21 is that even supported by anything? Jul 09 12:26:21 It's about 9K uncompressed. Jul 09 12:26:38 jow: I have.. two devices left which don't use it? Jul 09 12:26:42 Monkeh: sure, its always just a few K Jul 09 12:27:01 blogic: ping? Jul 09 12:27:04 I recently had to work about three weeks to strip "a few K" from luci Jul 09 12:27:14 Fair point Jul 09 12:27:16 these savings were immediately aten away within a matter of days Jul 09 12:27:29 Reminds me, I must get v2 of this patch out Jul 09 12:27:35 I can save more than 9K with it. ;) Jul 09 12:27:35 what happens if we set the default cflags to -flto? :) Jul 09 12:27:41 dwmw2_gone: Things break Jul 09 12:28:02 ...and we file bugs and ... Jul 09 12:28:10 code bugs, or tools bugs? Jul 09 12:28:12 no filing bugs is boring Jul 09 12:28:22 cursing upstream and complaining in irc os more fun Jul 09 12:28:27 :) Jul 09 12:28:27 *is Jul 09 12:28:46 Upstream often deserve it.. :P Jul 09 12:28:59 * ldir slaps jow - you've gone delirious with heat :-) Jul 09 12:29:03 jokes aside, if we tie 11r and 11w to CONFIG_SMALL_FLASH I'm fine with either some wpad-basic or simply adding functionality to wpad-mini Jul 09 12:29:39 as a general rule I don't like packages varying based on platform specifics Jul 09 12:29:52 then wpad-basic it is Jul 09 12:30:05 yeah Jul 09 12:30:07 I'm a 'fan' of wpad-basic simply because we can put in some 'this supports 11r/w blah blah' help text Jul 09 12:30:22 and make it the default for most platforms Jul 09 12:30:51 'cos I found the choice and 'how to enable 11r' mysterious Jul 09 12:31:35 So long as we can get 11r and 11w without installing the full wpad I'll be happy :) Jul 09 12:31:55 jow: my idea expressed in the PR is to create wpad-basic with 11r + 11w and set it as default for all (sub)targets !CONFIG_SMALL_FLASH Jul 09 12:32:01 anything that makes it more idiot proof (and I'm the idiot!) will be helpful Jul 09 12:32:24 mkresin: yeah, I just rgrep'd wpad in target/linux/ and its going to be ... interesting Jul 09 12:32:37 * ldir acks mkresin's idea. And also agrees with Monkeh . Jul 09 12:32:37 I somehow remembered it to be part of DEFAULT_PACKAGES Jul 09 12:32:53 but seems that every target matintains its own lists nowadays Jul 09 12:33:54 PKG_CONFIG_DEPENDS:= \ Jul 09 12:33:58 CONFIG_PACKAGE_hostapd \ Jul 09 12:33:58 CONFIG_PACKAGE_hostapd-mini \ Jul 09 12:34:04 what is that doing? Jul 09 12:34:27 dwmw2_gone: it means invalidate the make goals (.configured, .built etc.) if any of the mentioned symbols changed its choice state Jul 09 12:34:29 isn't it the DEFAULT_PACKAGES of the target? which might be extended by the subtarget? Jul 09 12:34:40 so basically trigger a rebuild of the package if any of the variants is toggled Jul 09 12:35:02 ok, thanks. In this case we build all binary variants in parallel and make different packages Jul 09 12:35:08 I know that there are some exception to it for targets where we have a SoC instead of a WiSoC Jul 09 12:35:12 so no rebuild if the variants are toggled. But yep, gotcha Jul 09 12:36:39 while we are talking about CONFIG_SMALL_FLASH, any opinions on https://git.openwrt.org/daf00d707e3f697c6c236517ff3ab949b463b7e9 ? Jul 09 12:36:55 mkresin: $ grep -rE 'PACKAGES.*wpad' target/linux/ | wc -l Jul 09 12:36:55 103 Jul 09 12:36:57 :( Jul 09 12:37:50 mkresin: ack Jul 09 12:38:18 should probably be done for ath25 too Jul 09 12:39:16 mkresin: also shouldn't it be "config *KERNEL_*CC_OPTIMIZE_FOR_SIZE" in config/Config-kernel.in ? Jul 09 12:40:14 jow: let me check. have to admit, I havn't tested it properly yet. Jul 09 12:40:43 jow: I want to collect some feedback about the idea before investing more time Jul 09 12:41:07 It should be KERNEL_, yes Jul 09 12:41:40 well, it works this way Jul 09 12:42:18 probably because its exposed as make variable and picked up by the inner kbuild system Jul 09 12:42:37 but usually you should prefix symbols with KERNEL_ if you want to auto-inject them into the Linux .config Jul 09 12:43:05 yeah, it has to be read as: wow, for some reason it works anyway Jul 09 12:43:57 So while we're on small_flash again, anyone have comments on the memory impact of changing the squashfs block size and cache? Jul 09 12:44:16 Monkeh: there is a small ram feature flag Jul 09 12:44:42 Monkeh: maybe use it to limit the change to small flash && !small ram targets Jul 09 12:45:20 Is anyone even using LOW_MEMORY_FOOTPRINT currently? Jul 09 12:45:41 Monkeh: yes, saw it somewhere Jul 09 12:46:34 jow: do you know by any chance if the CONFIG_CC_OPTIMIZE_FOR_SIZE is used for packages as well? Jul 09 12:47:07 jow: if I limit to the kernel the config symbol will be CONFIG_CC_OPTIMIZE_FOR_SIZE vs. CONFIG_KERNEL_CC_OPTIMIZE_FOR_SIZE Jul 09 12:48:12 Is there anybody there :-)) Jul 09 12:49:33 mkresin: Okay, so select 1M block size and say, 2 fragments if SMALL_FLASH and !LOW_MEMORY_FOOTPRINT? Jul 09 12:49:59 I don't think the extra RAM usage should be fatal for most 32M users, but I only use mine as APs and bridges.. Jul 09 12:50:34 Monkeh: hehe. I only pointed to the low mem feature flag. no idea how to do it the right way(tm) Jul 09 12:51:57 I have a tp-link 703N router with Openwrt Jul 09 12:52:10 using as a print server through the USB port Jul 09 12:52:29 anyone know if I need to open port 9100 to get it to print? Jul 09 12:53:28 This is the site with info http://blog.nguyenvq.com/blog/2015/03/15/make-a-printer-wireless-using-a-router-with-usb-running-openwrt/ Jul 09 12:56:51 jow: Monkeh: nope, I can't make KERNEL_CC_OPTIMIZE_FOR_SIZE selecting CC_OPTIMIZE_FOR_SIZE in the kernel config. fails with missing symbol Jul 09 12:57:42 * Monkeh blinks Jul 09 13:22:24 Hi guys! How can I set second DNS that will be used only if first one is offline? Jul 09 13:23:11 I'm using OpenWrt Chaos Calmer 15.05.1 Jul 09 13:39:22 hehe, choice option to the rescue to unset one and set the other optimization Jul 09 13:50:52 f00b4r0: yes ? Jul 09 14:11:58 blogic: i'm considering buying one of those and porting owrt to it: https://mikrotik.com/product/hap_ac2 - it's ipq4018 and I know nothing of that platform (dealt with ar71xx and mtk before). Anything I should be aware of? Jul 09 14:12:09 any obvious (to you) showstopper? Jul 09 14:17:06 * ldir can already hear the board screaming 'I submit!' :-) Jul 09 14:17:38 i wish ;) Jul 09 14:18:09 I'm also curious how ipq4018 compares to qca9533 in terms of wifi performance/range, if anyone has input on that topic Jul 09 14:18:49 I can't speak as to portability, but as usual from Mikrotik that's rather an attractive price. Jul 09 14:18:59 that it is Jul 09 14:19:54 too bad it has only 2x2 antennae Jul 09 14:20:22 Eh. Jul 09 14:21:24 i wonder what it would take to support the "usa" and "international" versions equally Jul 09 14:21:45 forgetting about the usa ? ;-) Jul 09 14:22:24 tax deals! Jul 09 14:22:26 building a wall ? Jul 09 14:24:30 https://www.streakwave.com/Itemdesc.asp?ic=hAP_ac2-US&o1=0 Jul 09 15:24:59 f00b4r0: none Jul 09 15:25:15 f00b4r0: hang on i recently perf tested an ipq4019 Jul 09 15:25:55 f00b4r0: so on ipq4019 i am getting .. using vht80 and 512byte frames .... Jul 09 15:26:02 TCP - 267 MBit Jul 09 15:26:02 TCP - offload - 318 MBit Jul 09 15:26:02 UDP - 60 MBit Jul 09 15:26:02 UDP - offload - 65 MBit Jul 09 15:26:22 and on wire i am getting wirespeed but not for a single flow Jul 09 15:26:24 how many streams? Jul 09 15:26:35 if you do a single flow its pegged on 1 CPU core and caps at ~580mbit Jul 09 15:26:49 DonkeyHotei: this was with 100 Jul 09 15:27:05 i mean tx/rx streams Jul 09 15:27:23 3T3R? Jul 09 15:27:27 yep Jul 09 15:27:40 and 100 parralell traffic flows Jul 09 15:27:57 this was using a tuned RPS/XPS/affinity setup Jul 09 15:28:06 i should really make that setting default Jul 09 15:28:09 I thought those chips were all 2T2R Jul 09 15:28:19 Monkeh: let me check Jul 09 15:28:34 the above device certainly looks to be 2x2, but it doesn't say Jul 09 15:28:36 indeed, the board only has 2 antennas Jul 09 15:28:43 so 2x2 Jul 09 15:29:01 however this is all done with 512 byte frame size Jul 09 16:28:36 blogic: sounds good to me :) I'll order one and see how it goes. Will keep you posted. Jul 09 16:46:40 * f00b4r0 adds 1 hAP mini, expects support to be a matter of adding a couple lines Jul 09 16:51:15 f00b4r0: Now that's just silly cheap Jul 09 16:53:45 because single-band Jul 09 17:12:36 Monkeh: ack :) Jul 09 17:22:26 Monkeh: i also expect to not need serial for either device work, thanks to routerboot's convenient embedded netboot Jul 09 17:44:25 I did some SW forwarding on lantiq xrx200 and was able to do about 300MBit/s, but SW NAT only worked with up to 60MBit/s Jul 09 17:44:41 why is this so different Jul 09 17:44:59 thi was with kernel 4.18 + DSA driver Jul 09 17:46:24 nat is more work, makes sense no? Jul 09 17:47:55 yes I know that it has to change the packet, but I would not expect that to make a difference of 300 MBit/s vs. 60 MBit/s Jul 09 17:48:44 true Jul 09 17:49:52 hi! Jul 09 17:53:24 I have a tplink router with a file to which I connected an internet hilink key, and an ATA for the voip connected via LAN. I can not connect with the voip if I use the hilink key, but if I use a key not hilink the voip works. what can it be? Jul 09 18:19:00 morning Jul 09 18:57:42 is there lxc-checkconfig sourcode https://git.openwrt.org/?p=feed/packages.git ? Jul 09 19:05:39 ping nbd lynxis others in berlin Jul 09 19:06:01 anyone with a spare usb-to-3.3v-serial adapter i could quickly borrow tonight? Jul 09 19:12:35 aw, nobody invited me :D Jul 09 19:24:30 dangole: yes => query Jul 09 19:33:25 Hi jow: When nls.mk is included, neither libiconv-stub nor libiconv-full are meant to override the libc iconv, correct? Jul 09 19:34:41 I mean the includes and libs are added after the system includes and libs, so I guess the system includes and libs will be used first. Jul 09 20:14:11 Hey! I was looking what to do with htmode and this article helped me https://wiki.openwrt.org/doc/uci/wireless but the "new" article is not that concrete with that option https://openwrt.org/docs/guide-user/network/wifi/basic should I update it with the information? Jul 09 20:29:18 Any plan to bump LXC to LXC 3.0 LTS? Jul 09 20:36:13 the update of mt76 is in the clear, my revert didn't change the leak behavior Jul 09 20:48:34 so whats going on in berlin Jul 09 20:49:19 ldir: ping Jul 09 20:49:30 is there a way for a wifi AP to announce to STAs that it has a metered uplink? Jul 09 20:49:42 also: are there standards for clients to respect this? Jul 09 20:51:50 mmm, none that I know of Jul 09 20:52:27 maybe something based on upnp? Jul 09 20:53:39 sounds like someone needs to draft an rfc or something :D Jul 09 20:55:07 yeah, five idevices connected to their chargers at night can be a PITA Jul 09 20:55:15 oh, look, wifi Jul 09 20:55:15 what the client actually ends up doing or not doing is always up to its own interpretation Jul 09 20:55:20 let's download 1.5gb each Jul 09 20:55:27 a generic way of signalling a metered connection would be kinda useful I guess Jul 09 20:55:44 wifi isn't really an appropriate place to do it since its just a generic access medium Jul 09 20:55:45 not to mention wintendo boxes Jul 09 20:55:54 you can set windows connections to metered though Jul 09 20:56:01 and it'll reduce its usage Jul 09 20:56:07 yeah, but users will have to do it per box Jul 09 20:56:07 (and apps can, if they bother checking) Jul 09 20:56:13 yeah Jul 09 20:56:23 I'm surprised microsoft haven't come up with a way to signal it either tbh Jul 09 20:56:32 win10 is a prime example that even if you do manually configure a metered link, windows update will mostly ignore it Jul 09 20:56:33 seems the obvious thing to do once you add support in the OS Jul 09 20:56:48 heh, its actually quite annoying Jul 09 20:56:55 mine won't let me set cellular to unmetered Jul 09 20:57:24 the router has a cellular uplink, pulling 100mbps easy Jul 09 20:57:32 yeah Jul 09 20:57:42 got caps? Jul 09 20:57:47 meaning clients will start doing chores on a metered uplink Jul 09 20:58:02 perhaps time based rate limits? Jul 09 20:58:18 something like that Jul 09 20:58:24 yeah, right now i just have cron kill the public ap at night Jul 09 20:59:09 sounds like a task for iptables quota or somesuch. Jul 09 20:59:10 I don't think I could survive with usage limits heh Jul 09 20:59:27 there really is no point in expecting clients to voluntarily do what you want ;P Jul 09 20:59:29 i was looking at wshaper Jul 09 20:59:37 sigh Jul 09 20:59:37 have some uci and cron-fu Jul 09 20:59:40 wshaper is wrong Jul 09 20:59:57 is no so many ways it's sad. Jul 09 21:00:11 i was using it five years ago Jul 09 21:00:14 guess you could just figure out the hosts they connect to and shape/filter them Jul 09 21:00:19 yeah. Even back then it was wrong Jul 09 21:00:26 punishing a tenant who left his shit in the hallway Jul 09 21:00:33 avoid the update sort of things Jul 09 21:00:39 trash in hallway? two days of 128kbps Jul 09 21:00:41 lol Jul 09 21:01:11 speaking of which, how broken is the internets in rural prague Jul 09 21:01:57 jwh, he got with the program rather fast once he figured out the correlation Jul 09 21:02:01 ha Jul 09 21:03:02 months of letters and notes, not a single frak was given. mess with his tube and fakebook? insta attitude adjustment Jul 09 21:03:11 was doing it on AA to, if mem serves Jul 09 21:03:15 needs the megabits for pornhub Jul 09 21:03:16 :D Jul 09 21:03:19 no, backfire or kamikaze Jul 09 21:04:08 If I've got a question regarding setting a different interface (not eth0) as WAN, best to ask here or on LEDE forums? Jul 09 21:04:33 you define the ifname in /etc/config/network Jul 09 21:41:16 anybody used the IPQ4019 before? Jul 09 21:41:22 how does it compare to the mtk7621 Jul 09 21:55:33 biangbiangmian: from all I've read so far, it should be quite decent (only ipq8065 here though) Jul 09 21:56:17 thanks Jul 09 22:16:52 biangbiangmian: quad-core vs. dual-core, and arm vs. mips Jul 09 22:20:21 i don't see anywhere that it supports hardware nat Jul 09 22:20:28 not too big of a deal though Jul 09 22:27:20 hardware nat is not implemented on ipq4* atm afaik Jul 09 23:26:07 Slimey: so, did anything ever come of your gpl source code request to adtran? Jul 10 02:13:55 wtf spacerat is STILL at it lol Jul 10 02:14:32 obviously a bot, so should be removed anyway Jul 10 02:14:56 yeah i think it might be a cracked out normal irc client with some kind of script too, idk Jul 10 02:15:07 if there was a human they'd notice Jul 10 02:15:16 CTCP reply from SpaceRat: VERSION xchat 2.8.9 Windows NT 6.2 [i386/3.49GHz] :GemX/MSWin32 2.7.5 + EDK/X 0.01 + PornKing Jul 10 02:15:22 some people are perpetually afk Jul 10 02:15:29 yeah thats fake Jul 10 02:15:29 :D Jul 10 02:15:32 lol Jul 10 02:15:40 nobody uses xchat anymore Jul 10 02:47:26 netflix refuses to work with my router. believes there to be a proxy. pile of merde Jul 10 02:48:34 triggered by my IPv6 tunnel. Gotta love garbage engineering Jul 10 02:48:54 he.net is considered a vpn provider Jul 10 02:49:12 yeah it is Jul 10 02:49:17 but they can work around it Jul 10 02:49:25 and fall back to ipv4 Jul 10 02:51:12 use /etc/hosts Jul 10 02:51:42 make the netflix hosts resolve only to ipv4 **** ENDING LOGGING AT Tue Jul 10 03:00:01 2018