**** BEGIN LOGGING AT Mon Jul 16 03:00:01 2018 Jul 16 03:34:20 pair of questions wrt. WNDR4300. (a) how do I use the full 128 MB flash instead of just 32MB (b) how to i build a sysupgrade.bin to flash with mtd instead of via sysupgrade (I don't want any old cruft to copy over) - I'm using the imagebuilder - any ideas? Jul 16 03:39:03 you can't Jul 16 03:39:42 a) with imagebuilder not at all, because you'd have to modify and recompile the kernel - and with that the whole firmware Jul 16 03:40:28 b) the OEM firmware is structured in a way that doesn't allow 'easy' changes, so you'd have to go the long way round Jul 16 03:59:33 so (a) isn't just a matter of changing the partition listing on the kernel commandline? Jul 16 04:00:17 could you clarify wrt. (b) what do you mean by long way round? older versions of openwrt appeared to build sysupgrade.bin files by default, now I only seem to get a sysupgrade.tar Jul 16 04:00:17 not on NAND Jul 16 04:00:54 ah. I have no idea what's special about NAND... is there something/site to read about the diffs? I'm a curious sort Jul 16 04:01:30 NAND needs error correction and wear levelling, e.g. using ubi Jul 16 04:01:50 and other flashes don't because they do it for you? Jul 16 04:03:33 does it actually use the full 128 mb to provide the wear levelling/error correction? or is there actually space left over? Jul 16 04:05:25 netgear in all their wisdom left ~96 MB unused, but with important data between rootfs and the reserved space - so you can't expand the rootfs/ overlay without also moving caldata_backup, aka the (multi-step) long way round Jul 16 04:06:07 possible, yes - advisable, no Jul 16 04:06:15 ah, yes I did see the caldata_backup partition, I assumed it didn't matter for openwrt... Jul 16 04:08:03 in general: "board=WNDR4300 console=ttyS0,115200 mtdparts=ar934x-nfc:256k(u-boot)ro,256k(u-boot-env)ro,256k(caldata)ro,512k(pot),2048k(language),512k(config),3072k(traffic_meter),2048k(kernel),23552k(ubi),25600k@0x6c0000(firmware),256k(caldata_backup),-(reserved) rootfstype=squashfs noinitrd" - there seems to be a lot of cruft. Jul 16 04:20:39 is ./bin/targets/ar71xx/nand/openwrt-18.06.0-rc1-ar71xx-nand-vmlinux.bin the sysupgrade.bin file I should mtd write to firmware partition? Jul 16 04:21:05 at 4542980 bytes it looks to be roughly the right size Jul 16 04:24:26 no Jul 16 04:27:01 hmm, the only relevant build outputs appear to be: -rw-r--r--. 1 root root 4730880 Jul 15 14:44 bin/targets/ar71xx/nand/openwrt-18.06.0-rc1-ar71xx-nand-wndr4300-squashfs-sysupgrade.tar -rw-r--r--. 1 root root 5505153 Jul 15 14:44 bin/targets/ar71xx/nand/openwrt-18.06.0-rc1-ar71xx-nand-wndr4300-ubi-factory.img Jul 16 04:33:55 I guess that just means there is no sysupgrade.bin build output. https://downloads.openwrt.org/releases/18.06.0-rc1/targets/ar71xx/nand/ also doesn't have a sysupgrade.bin for wndr4300 Jul 16 07:00:01 ldir: https://patchwork.ozlabs.org/patch/916121/ is assigned to you Jul 16 07:00:17 had a chat with nbd about it, we think adding a wpad-tiny might be a good choice Jul 16 07:00:44 then strip mini down further and set all the small targets to use tiny variant Jul 16 07:32:44 mkresin: ping Jul 16 07:37:23 jow: pong Jul 16 07:37:52 blogic: it will be the other way round. we add a wpad-basic and use it for !small_flash by default Jul 16 07:40:03 nbd: pong (from Friday or sth) Jul 16 07:47:09 mkresin: while debugging a luci timeout issue I noticed that the dsl_control lucistat command takes about 5-6 seconds to complete Jul 16 07:47:22 mkresin: is this expected? Jul 16 07:47:53 jow: not sure, but yes it is slow Jul 16 07:47:58 okay Jul 16 07:48:12 then I'll probably lower the status poll interval for it Jul 16 07:48:16 its currently queried every 5s Jul 16 07:51:57 jow: we send a query for each value we are interested via userspace. it wasn't much an issue at the time we were interested in only a handful of values Jul 16 07:58:43 mkresin: ok Jul 16 07:59:17 jow: time /etc/init.d/dsl_control lucistat: real 0m 2.03s; user 0m 0.81s; sys 0m 1.25s Jul 16 07:59:52 jow: but there is some locking involved since the interface can have only one consumer Jul 16 08:00:07 mkresin: yeah, that is significant, and I guess on some really slow lantiq hw it can easily exceed the 5s grace period Jul 16 08:00:43 jow: if the locking happens at a bad moment, we are easily up to 4s Jul 16 08:12:46 hi all, quick question, where in the openwrt source does it specify if 4.14 or 4.9 kernel will be built for a particular board in 18.06? Jul 16 08:13:23 leperet: $ grep -r KERNEL_PATCHVER target/linux/ Jul 16 08:13:42 dedeckeh: ping Jul 16 08:14:00 dedeckeh: ping from me too Jul 16 08:14:33 dedeckeh: https://patchwork.ozlabs.org/patch/877640/ Jul 16 08:14:45 jow: any comments on that ^ Jul 16 08:16:06 imho was should just check strlen(dev->ifname) >= IFNAMSIZ - 6 Jul 16 08:16:35 at the start of get_vlan_device() Jul 16 08:16:47 and drop the over engineering Jul 16 08:17:21 his commits shortens the interface base name to make the vlan tag fit Jul 16 08:17:45 I actually like the idea but I think we need to come up with some kind predictable name hashing if the target name gets too long Jul 16 08:18:13 he does not shorten the name Jul 16 08:18:53 ah, ok Jul 16 08:19:47 and there is an off-by one aswell Jul 16 08:19:56 ah no there is not Jul 16 08:20:09 strlen wont include the \0 Jul 16 08:23:25 blogic: Can you please look also on the other patch in http://lists.infradead.org/pipermail/openwrt-devel/2018-July/013226.html ? I noticed only one commit/merge for gpio, not for leds. Thanks Jul 16 08:24:00 because you sent it as an attchement Jul 16 08:24:13 why dont you use git ? Jul 16 08:25:29 muhaha: merged into my local tree Jul 16 08:25:51 I have blocked traffic to smtp ... Its hard to setup proxy or vpn just for sending it trough git command Jul 16 08:26:03 ? Jul 16 08:26:24 i wont ask why .... Jul 16 08:26:28 https://git-scm.com/docs/git-send-email Jul 16 08:26:37 I can not use it :d Jul 16 08:28:07 blogic:pong Jul 16 08:28:12 jow:pong Jul 16 08:30:50 blogic:I'm not in favor of the patch; as you said it's over engineered Jul 16 08:31:57 vlan_dev_set_name should check the return code of device_set_ifname iso doing the strnlen check Jul 16 08:34:04 dedeckeh: sent an alternate patch Jul 16 08:34:26 blogic:will do Jul 16 08:34:30 no Jul 16 08:34:35 i've just mailed one :-D Jul 16 08:34:41 ah ok Jul 16 08:34:45 just noticed it Jul 16 08:35:03 ah wait Jul 16 08:35:06 should return NULL Jul 16 08:35:12 should compile test stuff :-D lol Jul 16 08:35:20 :) Jul 16 08:35:38 but yeah, that patch but return NULL should be fine Jul 16 08:36:09 blogic: thinking out loud but instead of having multiple flavors of wpad, wouldn't it be better to have a modular wpad instead? One core daemon, extra plugins available? Jul 16 08:36:23 f00b4r0: yes Jul 16 08:36:28 would give a lot more flexibility for small devices in particular Jul 16 08:36:28 but the codebase does not support that Jul 16 08:36:31 ah ok Jul 16 08:36:44 time to rewrite codebase? :D Jul 16 08:36:48 so we are stuck with feature creap or missing features Jul 16 08:37:07 makes sense Jul 16 08:38:08 blogic:the check can be moved to line 122; check needs only to be done when the interface needs to be created Jul 16 08:38:20 muhaha: you don't have to use git to send inline attachments. You just need to configure your mailer correctly Jul 16 08:39:06 blogic:will rework your patch Jul 16 08:39:42 f00b4r0: I know, but I would have to setup tunnel over 443 or 80 to outside smtp to send an email Jul 16 08:40:05 jow:does rework patch http://patchwork.ozlabs.org/patch/942852/ look sane to you ? Jul 16 08:40:13 *reworked Jul 16 08:40:16 muhaha: oh, you're using webmail. Jul 16 08:42:48 blogic:could you give feedback on patch http://patchwork.ozlabs.org/patch/924714/ ? Jul 16 08:44:47 Does it makes sense ? echo "CONFIG_SYMBOL_THAT_WILL_EXPAND_ANOTHER_NON_CONFIGURED_SYMBOLS=y" >> ./target/linux/x86/64/config-default && yes n | make kernel_oldconfig CONFIG_TARGET=subtarget && make ... Jul 16 08:46:21 alternative to global config: echo "CONFIG_SYMBOL_THAT_WILL_EXPAND_ANOTHER_NON_CONFIGURED_SYMBOLS=y" >> .config && make defconfig; make Jul 16 08:51:33 dedeckeh: yes, exactly what I had in mind - Acked-by Jul 16 08:51:55 dedeckeh: hopefully Alin is not too pissed off by my change requests ;) Jul 16 09:01:29 dedeckeh: something else; https://github.com/openwrt/luci/issues/1947 (tldr: luci shows local interface ip6addr as reported by "local-address" but netifd configures a /60 instead of a /64 on the iface) Jul 16 09:02:03 dedeckeh: apart from showing the delegated prefix portion for downstream ifaces, how to solve the /64 vs. /60 thing? Shall I display the local-address in conjunction with the prefix size of the delegated prefix portion? Jul 16 09:02:26 and if so, what significance has the mask option of local-address? Jul 16 09:08:12 blogic: thx for merging. Don't merge my other PR just yet please :) Jul 16 09:10:24 Hi! I'm trying to figure out the UART 4 pin headers on my routers PCB Jul 16 09:10:46 I know the Vcc and GND but can't figure out the TX and RX Jul 16 09:11:18 the remaining two headers do not have any voltage with ground Jul 16 09:11:42 and tried both one by one as TX but no output Jul 16 09:12:03 Is it posssible TP-link has the UART disabled in the stock firmware? Jul 16 09:16:06 https://github.com/openwrt/openwrt/pull/1182 Jul 16 09:16:07 boom Jul 16 09:16:26 erm, shall we rik it ? i checked 10 random ones and they were all correct Jul 16 09:16:47 f00b4r0: this one https://github.com/openwrt/openwrt/pull/1181 Jul 16 09:17:05 if you dont want it merged, then dont send it or put a note under it saying it should not be merged Jul 16 09:17:32 blogic: risk it, that cleanup looks very worthwhile Jul 16 09:17:38 indeed Jul 16 09:17:44 and will break with the next ramips commit Jul 16 09:22:02 blogic: jow: can we get such DTS changes upsteam? Jul 16 09:22:18 rmilecki: we are upstream Jul 16 09:23:13 heh Jul 16 09:23:39 i guess it means "no" Jul 16 09:23:56 rmilecki: :-) Jul 16 09:47:30 * f00b4r0 notices 17.01.5 has been 'released', might be worth updating the topic ;) Jul 16 09:48:45 now I have to update my personal 17.01 tree Jul 16 09:49:00 so my 'unsupported' devices can enjoy mainstream release ;) Jul 16 09:56:26 jow:netifd needs be fixed for the reported issue Jul 16 09:58:09 jow:the reported mask in ubus has now a check => assign->length < 64 ? 64 : assign->length Jul 16 09:58:18 I will look into this Jul 16 09:59:45 f00b4r0: its not released yet Jul 16 09:59:55 ah Jul 16 10:00:09 well it's tagged in the git tree so the code base is frozen now? Jul 16 10:00:18 yes, but there'll likely be a .6 Jul 16 10:00:28 ok Jul 16 10:00:30 at least I plan to backport firewall updates, luci updates and some other Jul 16 10:00:39 cake would be nice Jul 16 10:00:40 rmilecki has pending bcm fixes Jul 16 10:00:43 * f00b4r0 winks at ldir Jul 16 10:01:36 right, I just want to wait a day before I cherry pick them to the 18.06 and 17.01 Jul 16 10:03:15 f00b4r0: sod off! ;-D Jul 16 10:03:20 jow: fwiw I've opened #1181 with 3 patches for 18.06 (just so it's on the radar). I'll also probably have a minor fix for a ramips routerboard dts when I hear back from original submitter. Jul 16 10:03:50 ldir: ;D Jul 16 10:54:04 Does it makes sense ? echo "KERNEL_CONFIG_SYMBOL_THAT_WILL_EXPAND_ANOTHER_NON_CONFIGURED_SYMBOLS=y" >> ./target/linux/x86/64/config-default && yes n | make kernel_oldconfig CONFIG_TARGET=subtarget && make ... Jul 16 10:54:04 as alternative to global config: echo "GLOBAL_CONFIG_SYMBOL_THAT_WILL_EXPAND_ANOTHER_NON_CONFIGURED_SYMBOLS=y" >> .config && make defconfig && make ? Jul 16 10:58:54 no Jul 16 10:59:24 :] Jul 16 11:02:12 f00b4r0: is that all? :D Jul 16 11:02:26 yes Jul 16 11:03:01 Of course I need to edit it noninteractively, so kernel_menuconfig is not an option :] Jul 16 11:50:48 mkresin: there's a report on the forum stating that on lantiq/xway, the wifi only starts up a fter a reboot, any idea what might cause it? Jul 16 11:50:58 https://forum.lede-project.org/t/openwrt-18-06-rc1-lantiq-xway-no-wifi-after-sysupgrade/16078 Jul 16 11:51:01 rmounce: ping? Jul 16 12:11:40 on openwrt /dev/shm is mounted under /dev which is a 512K tmpfs, this is too small and where is the best place I can increase it? could not find corresponding setting under base-files or preinit yet Jul 16 12:12:34 tmpfs 512.0K 0 512.0K 0% /dev Jul 16 12:17:09 interesting new error from make package/foo/download PKG_HASH=skip. https://pastebin.com/DumV1NiL Jul 16 12:23:58 anyone around who has worked on/seen work on the Edimax BR6478AC v2? Jul 16 12:38:54 there was some work done on it prior to the merge of LEDE and openwrt, but it was never fully integrated. patches were made for kernel 4.4, but the patches don't port fully over to 4.14 due to some changes in mtdsplit_uimage.c Jul 16 12:39:49 the need for the changes comes from the header requirements (something about csys) Jul 16 12:47:57 warrick_: from what i saw, the kernel patches are most likely not needed at all, just need to change the image generation code instead Jul 16 12:51:32 DonkeyHotei: thanks for continuing to look through that stuff. I think the part I'm stalled on has to do with the image generation code, though, in terms of correctly calculating the header size of the final image Jul 16 12:52:09 or am I not understanding how this is structured (very possible) Jul 16 12:53:37 near as i can figure, the code should look more like the code for the br-6475nd Jul 16 12:57:19 just to make sure, are we talking about the code in the mtdsplit_uimage.c file? Jul 16 12:57:25 or just in general Jul 16 12:58:13 most likely, *none* of the kernel changes are needed, only dts and image generation Jul 16 12:58:49 ok Jul 16 12:58:57 do you have serial console access to the device? Jul 16 12:59:42 not yet, but will set it up when I'm ready to flash Jul 16 13:04:02 i could try whipping up an experimental image for you to test, but for all i know, it could brick your device Jul 16 13:05:22 so the changed files will amount to: base-files/ramips.sh; base-files/lib/upgrade/platform.sh; dts/BR-6478AC-V2.dts; image/mt7620.mk Jul 16 13:05:53 is the image/mt7620.mk where the 'image generation code' resides? Jul 16 13:06:16 several more in the patch file prior to the one you linked Jul 16 13:06:28 appreciate the offer, but I'll take the risks myself. Jul 16 13:06:32 right, that was just from jmm.patch Jul 16 13:06:41 the Ulli Kroll file, forgot about that one Jul 16 13:08:30 etc/board.d/[02_network,01_leds]; etc/diag.sh Jul 16 13:08:57 so those all look like sane files to be touching if adding support for a board? Jul 16 13:09:43 I'm also heading out of town for a couple of weeks starting today, so no point in trying to troubleshoot ATM Jul 16 13:11:19 should i just roll it up into a git commit for you to try before heading out? Jul 16 13:12:39 that would be more efficient than me trying to do it. please do Jul 16 13:20:46 hmmm, creating a new package, have luajit in its DEPENDS, yet getting this: Package suricata is missing dependencies for the following libraries: Jul 16 13:20:47 libluajit-5.1.so.2 Jul 16 13:20:54 that will also get it one step closer to merging Jul 16 13:22:01 any idea why that's happening? Jul 16 13:27:35 stintel: one of the deps is pulling that ? Jul 16 13:27:44 or you are linking a host lib ? Jul 16 13:28:03 Hi! I have a custom auto channel scanning script which only use non-DFS channels for faster boot time. How do I implement it so its selectable in the Wifi Channels dropdown menu in Luci, such as "auto (non-DFS)"? Jul 16 13:37:04 blogic: I guess I should us ldd on the binary Jul 16 13:39:35 readelf Jul 16 13:43:02 Hi! I have a custom auto channel scanning script which only use non-DFS channels for faster boot time. How do I implement it so its selectable in the Wifi Channels dropdown menu in Luci, such as "auto (non-DFS)"? Jul 16 13:43:10 readelf Jul 16 13:43:32 alx_: you have to edit the luci model (wireless_modefreq.htm iirc) to offer "auto (non-DFS)" as choice, then you need to edit /lib/netifd/hostapd.sh to handle it Jul 16 13:44:38 you probably want the hsotapd.sh script to invoke your scanning scriptm then write the resulting fixed freq into hostapd Jul 16 13:45:06 however, iirc there is some hostapd / uci option to restrict the channels to scan for acs, might be simpler to just set that Jul 16 13:46:27 ah right, `chanlist=100 104 108 112 116` Jul 16 13:48:23 alx_: https://git.openwrt.org/d293f33558e3d0a08135c48c8d8bcd079e4fd69b Jul 16 13:49:53 thanks jow, I´ll have a look at it :) Jul 16 14:06:33 WARNING: Makefile 'package/feeds/luci/luci-nginx/Makefile' has a dependency on 'nginx-mod-luci', which does not exist Jul 16 14:06:44 you can ignore that for now Jul 16 14:06:48 ok Jul 16 14:28:52 warrick_: are you still there? Jul 16 14:30:22 blogic: readelf isn't making me any wiser :( Jul 16 14:30:39 0x0000000000000001 (NEEDED) Shared library: [libluajit-5.1.so.2] Jul 16 14:31:04 stintel: what package are you building Jul 16 14:31:33 suricata Jul 16 14:32:23 and I have staging_dir/target-x86_64_musl/usr/lib/libluajit-5.1.so.2 build_dir/target-x86_64_musl/luajit-2017-01-17-71ff7ef/ipkg-install/usr/lib/libluajit-5.1.so.2 Jul 16 14:32:31 but also /usr/lib64/libluajit-5.1.so.2 Jul 16 14:32:36 so it might be picking up the host Jul 16 14:32:55 there's also option acs_chan_bias '1:0.7 5:0.7 9:0.7 13:0.7' Jul 16 14:59:02 DonkeyHotei: pong Jul 16 14:59:32 warrick_: https://github.com/danielg4/openwrt Jul 16 15:01:31 wicked. will try shortly Jul 16 15:19:29 SO this is my first time with using git like this. do I clone your repository? Jul 16 15:20:41 f00b4r0: Just read you mail, hope my reply is helpful :) Jul 16 15:21:13 you could "git add remote danielg4 https://github.com/danielg4/openwrt" then "git fetch danielg4" and "git checkout danielg4/BRANCH_YOU_WANT_TO_CHECKOUT" Jul 16 15:21:16 tobleminer-tSYS: it is very much, thanks. I see why you came up with this "incorrect" partition map: the 750Gr3 is unfortunately a bastardized device Jul 16 15:21:26 the current support for it in tree is for a device with a reflashed bootloader Jul 16 15:21:30 i.e. not the original bootloader Jul 16 15:21:38 Yes, indeed Jul 16 15:21:45 I really wish this dts could be wiped but I'm afraid I'll get no support on that ;P Jul 16 15:22:25 So there is nothing like a seperate "factory" partition then? Jul 16 15:22:40 tobleminer-tSYS: i'm going to take a look at your flash dump and will prepare ASAP a revised DTS for you to test. If it works I'll open a PR to get the fix before 18.06 is released. Jul 16 15:22:53 Sure, that would be great Jul 16 15:23:06 tobleminer-tSYS: did you confirm that the third regulator does anything? How did you find its gpio? I see no reference in the mikrotik source Jul 16 15:23:45 warrick_: yes Jul 16 15:23:58 Well, I looked up the datasheet for the step down regulator used, hooked myself up to the enable pin and started switching GPIOs Jul 16 15:24:21 so it does work ;) Jul 16 15:24:40 interesting. I suppose mikrotik has no use for it and that's probably why it's not used in their source. Jul 16 15:24:51 Do you have access to RouterBOOT source, too? Jul 16 15:24:57 Maybe it's handled in there Jul 16 15:25:00 i wish ;) Jul 16 15:25:25 (or maybe I really don't, given how ugly their linux patch is) Jul 16 15:25:52 All the regulators are turned on by RouterBOOT so it's possible they just don't bother handling some of them in linux, i guess Jul 16 15:26:00 tobleminer-tSYS: there was someone here earlier reporting partial success with a ubnt nanobeam ac gen2 and wanting to talk to you about it Jul 16 15:26:13 Oh, ok Jul 16 15:26:21 DonkyHotel: I totally missed that Jul 16 15:26:59 I just slept through most of the day because I spent all night hacking on a Mikrotik wAP60 G wireless wire Jul 16 15:27:45 whoever pinged me, it scrolled out of my buffer Jul 16 15:28:24 f00bar0: Do you have any special sources for Mikrotik stuff or just the normal "give me your GPL sources" stuff? Jul 16 15:28:39 greearb_: it was Hauke who ponged you Jul 16 15:28:52 nano Jul 16 15:29:19 Hauke, you around? Jul 16 15:29:22 stintel, thanks Jul 16 15:29:33 tobleminer-tSYS: I ask politely. And when they send me outdated garbage (which they usually do), I insist ;) Jul 16 15:30:04 I had a backports project question....after debugging, I noticed that coccinelle and/or some dependency is broken on F27, took a while to figure that out though. Jul 16 15:30:36 anyone know an OS where backports works out of the box? Jul 16 15:30:52 (I failed to compile the latest coccinnelle due to some other bug) Jul 16 15:30:58 tobleminer-tSYS: tail end of the page, http://logs.nslu2-linux.org/livelogs/openwrt-devel/openwrt-devel.20180714.txt Jul 16 15:31:15 f00b4r0: Ah, ok. I should probably try that, too. I'm currently working on the wireless wire (wAP60 G) and got a major headache form a lot of unknowns like PCIe and RGMII GPIOs Jul 16 15:31:38 I'm waiting for their 'updated' patch. I can link you up when I get it Jul 16 15:31:50 they sent me one from Feb which doesn't have some of the newer devices Jul 16 15:32:19 f00bar0: That would be absoluteley awesome Jul 16 15:32:33 yep, suricata is using host libs :( Jul 16 15:32:35 damnit Jul 16 15:32:40 DonkeyHotel: Thanks! Jul 16 15:33:33 or not Jul 16 15:34:02 ugh, I am just going to disable the luajit support Jul 16 15:34:07 whitewolf: Does ethernet work on 10 Mbit/s, 100 Mbit/s and 1000 Mbit/s ? I had some PLL issues on some devices. Jul 16 15:37:02 whitewolf: I guess the NanoBeam is probably exactly the same as the NanoStation AC as far as the 'non-HF' part goes Jul 16 15:38:12 whitewolf: Do all the LEDs work, too? Do the signal leds light in the correct order? Jul 16 15:41:42 DonkeyHotei, many thanks for putting those changes into git. I'm out of time now, but will try it when I'm back and let you know how it goes Jul 16 15:43:49 tobleminer-tSYS: you dumps reveal what I expected. The ramips flash is sequenced in a similar way as the ar71xx ones, with different offsets Jul 16 15:44:26 in an ideal world we would be able to detect these offsets the way it's done on ar71xx but in a DTS context I'm going to hardcode them for now, so that at least we have a 'sane' basis for future work Jul 16 15:46:40 f00b4r0: Ok, I'll be happy to test any changes Jul 16 15:46:53 thanks. I'll cook something up right away Jul 16 16:04:10 ok i think i've identified everything Jul 16 16:04:12 * f00b4r0 double checks Jul 16 16:21:28 tobleminer-tSYS: if you could test the patch on this tree: https://github.com/f00b4r0/openwrt/commits/rbramipsfixes Jul 16 16:22:34 it appears rbcfg can be made to work on these devices, however I think the source code will not work as is due to endianness. Jul 16 16:24:25 i don't thikn I'll have time to fix that before 18.06 Jul 16 16:24:34 blogic: it appears as if mt7622 fails to build on openwrt-18.06, will investigate later Jul 16 16:24:39 seems to be due to missing dts or something Jul 16 16:25:45 build #1 of mediatek/mt7622 is complete: Success [build successful] Build details are at http://release-builds.openwrt.org/18.06/images/builders/mediatek%2Fmt7622/builds/1 Jul 16 16:31:46 jow: out of curiosity, a PR that removes a DTS for a 'supported' device that requires a bootloader reflash (when the original bootloader _can_ be supported) has no chance of being accepted? Jul 16 16:33:59 f00b4r0: I don't see why not Jul 16 16:34:10 well it was part of 17.01 Jul 16 16:34:51 are we ok to antagonize whatever users there are out there? :P Jul 16 16:35:41 ideally I'd like it to go extinct before 18.06, so we can work on proper support going forward Jul 16 16:35:55 f00b4r0: Looks good except for duplicate nodenanems Jul 16 16:35:58 *names Jul 16 16:36:25 tobleminer-tSYS: by that you mean I can't have overlapping partitions in DTS? Jul 16 16:36:27 18.06 will drop a bunch of devices that previously worked, don't think one more would make a huge difference Jul 16 16:36:37 jow: great. Let me open a PR right away then Jul 16 16:36:49 also 18.06.0 isn't the end of the world, its quite likely that we'll follow up with a .1 and of july Jul 16 16:36:56 *end of Jul 16 16:36:58 f00b4r0: At least the can not both have the node name partition@0 Jul 16 16:36:59 bbl Jul 16 16:37:01 *they Jul 16 16:37:06 ha Jul 16 16:37:35 i understand but I'm not sure what the correct naming convention would be there. Care to suggest one? Jul 16 16:39:06 * f00b4r0 goes with 'partition@1' for the second one Jul 16 16:39:43 f00b4r0: I do only have limited device tree experince myself. I guess partition@1 will be ok Jul 16 16:39:54 tobleminer-tSYS: i'd like you to confirm that you still get the correct MAC, and that the mtd partition scheme does look like what the DTS suggests Jul 16 16:40:08 Yep, I'm on it Jul 16 16:40:10 if both are good then I'll open a PR, in particular for the 33 which is already in 18.06 Jul 16 16:44:50 * tobleminer-tSYS boots rbm33g Jul 16 16:50:07 tobleminer-tSYS: the suspense is killing me ;) Jul 16 16:52:26 f00b4r0: Looks pretty goot so far, nice work Jul 16 16:52:34 excellent. I'll open a PR Jul 16 16:52:50 then when I have a bit more time I'll work on making rbcfg endianness-agnostic Jul 16 16:53:01 That would be nice Jul 16 16:53:16 tobleminer-tSYS: ok to add "Tested-by"? Jul 16 16:53:47 f00b4r0: Sure Jul 16 16:58:30 blogic: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=d8d2133c35c9c9b410e16cdebe878acd0da6382f looks weird, shouldn't this be an InstallDev instead of a host build? Jul 16 17:03:51 tobleminer-tSYS: can you post the mtd partitioning somewhere? I don't fully trust myself ;) Jul 16 17:04:01 (output from dmesg) Jul 16 17:06:49 f00b4r0: Here you go https://gist.github.com/TobleMiner/da3a2a098ac983c20ba191c68991e70b Jul 16 17:07:56 awsome. Looks like what I expected :) Jul 16 17:07:59 thanks Jul 16 17:10:34 np Jul 16 17:11:30 tobleminer-tSYS: #1185 Jul 16 17:11:34 thanks for your help! Jul 16 17:13:09 f00b4r0: No problem, thank you for fixing my butchered partition layouts :) Jul 16 17:13:19 you're welcome :) Jul 16 17:15:06 bbiab Jul 16 17:18:30 jow: no idea. look like the firmware userspace helper fail at first boot Jul 16 17:22:49 mmm Jul 16 17:22:55 did gcc 8 get merged yet? Jul 16 17:31:10 jow: now that I'm thinking about it, I rather suspect the owl-loader kernel module isn't loaded at firstboot Jul 16 17:32:23 jow: I've seen something similar on my 1st gen RPi. A kernel module backed into the image doesn't load on first boot but on all subsequently boots Jul 16 17:33:01 smells like some kind of timeout is hit Jul 16 17:35:01 boo. github died Jul 16 17:36:52 all those unboxing videos and none of the idiots^Whosts care to plug the damn device... *sigh* Jul 16 17:37:44 mangix: ? Jul 16 17:37:58 getting the unicorn Jul 16 17:38:37 hmm, yes. worked a minute ago Jul 16 17:38:49 mangix: Yeah, me too Jul 16 17:45:48 it's back. trés bien. Jul 16 17:46:29 aaaand...it's gone Jul 16 17:46:55 and back again Jul 16 17:47:39 Why is PACKAGE_lxc-unprivileged deselecting packages shadow-newuidmap and shadow-newgidmap which are needed by lxc-checkconfig ? Jul 16 17:48:10 crypto devices aren't supported on ramips, correct? Jul 16 17:49:58 f00b4r0: hardware? Jul 16 17:50:11 mangix: aye Jul 16 17:50:18 i'm assuming it's a SoC feature Jul 16 17:50:23 but I don't really know Jul 16 17:50:26 not supported. a user on the forums is working on it Jul 16 17:50:31 ok Jul 16 17:50:48 * f00b4r0 will leave a comment in DTS for future reference Jul 16 17:50:56 sweet Jul 16 17:52:43 tobleminer-tSYS: well I think something else's missing: USB. Jul 16 17:52:51 needed for LTE modems I suppose Jul 16 17:53:29 or does it work oob Jul 16 17:53:53 ah it's in the dtsi Jul 16 17:53:57 * f00b4r0 is overheating again Jul 16 17:54:37 * mamarley hands f00b4r0 an air conditioner. Jul 16 17:54:48 i'd kill for one ;P Jul 16 18:12:09 Seems that CONFIG_LXC_BUSYBOX_OPTIONS=y deselect PACKAGE_shadow-new{uid,gid}map which is needed for unprivileged. Busybox alternative is BUSYBOX_CONFIG_FEATURE_SHADOWPASSWDS and BUSYBOX_CONFIG_USE_BB_SHADOW ?? Jul 16 18:12:21 unprivileged lxc Jul 16 18:12:50 DonkeyHotei: ping Jul 16 18:13:05 hi Jul 16 18:14:07 DonkeyHotei: hi. Do you have enabled unprivileged support for lxc ? Jul 16 18:14:42 yes Jul 16 18:15:49 Are you using CONFIG_LXC_BUSYBOX_OPTIONS=y ? Jul 16 18:16:11 CONFIG_LXC_BUSYBOX_OPTIONS=y Jul 16 18:17:00 ok I'm confused Jul 16 18:17:07 I defined KernelPackage/accel-ppp-ipoe Jul 16 18:17:14 but its bailing with.. Jul 16 18:17:20 Makefile:91: *** Package/accel-ppp-ipoe is missing the TITLE field. Stop. Jul 16 18:17:25 (it has TITLE) Jul 16 18:17:40 or should I be putting kernel bits in seperate kmod packages? Jul 16 18:17:59 DonkeyHotei: and what about newuidmap, newgidmap ... ? Are you using any symbol to extend this utilities to busybox? Jul 16 18:18:06 lmfao openwrt's build system tried to run rm -rf / Jul 16 18:18:28 mangix: uhhhhhhhhh Jul 16 18:21:13 CONFIG_PACKAGE_shadow-newgidmap=y Jul 16 18:21:13 CONFIG_PACKAGE_shadow-newuidmap=y Jul 16 18:21:18 muhaha: ^ Jul 16 18:22:02 greearb_: hi Jul 16 18:22:51 greearb_: last time I also had some problems with coccinelle Jul 16 18:23:09 DonkeyHotei: weird, these are deselected by some other symbol for me... But seems that CONFIG_LXC_BUSYBOX_OPTION is not source of the problem. Jul 16 18:23:12 let me check if I fixed it Jul 16 18:23:12 Hauke, good to hear from you. I reverted the patch that required the latest version, and that seems to be working mostly-ish Jul 16 18:24:43 f00b4r0: https://forum.lede-project.org/t/hardware-crypto-for-mediatek-missing/3314/56 Jul 16 18:27:17 mangix: i see, thanks Jul 16 18:27:22 DonkeyHotei: Can you send me your .config, please? Jul 16 18:28:06 greearb_: yes this change caused some problems for me: https://git.kernel.org/pub/scm/linux/kernel/git/backports/backports.git/commit/?id=98272f479c2126a135dfcb12484e93d5888164ab Jul 16 18:28:35 that is the one I reverted Jul 16 18:28:49 greearb_: Johannes asked Julia for help but he hasn't received an answer Jul 16 18:29:16 I wanted to bring backports in sync with the recent kernel so we can use a more recent version in openwrt Jul 16 18:29:22 no way to conditionally make the code work with either version? Jul 16 18:29:39 I am not very good with coccinelle Jul 16 18:29:51 yeah, not very promising when it won't compile Jul 16 18:35:13 mangix: does 'ralink_ts_device' mean anything to you, mt7621 context? Jul 16 18:35:43 greearb_: yes that is still on my todo list Jul 16 18:35:53 f00b4r0: absolutely nothing Jul 16 18:35:58 I think currently backports is only used by some companies internally Jul 16 18:35:59 heh Jul 16 18:37:20 Hauke, it would be great if there were some way to make latest backport able to handle older upstream kernels. Like with special 'aux' patches for older kernels. Even if it didn't handle all of them, users could contribute their own aux patches to enable kernels they care about. Jul 16 18:37:34 I had to hack a few patches to make my 4.16 kernel work as upstream, for instance Jul 16 18:39:07 greearb_: currently we do not even have enought people to keep it up with the latest development in time Jul 16 18:39:18 muhaha: https://pastebin.com/sMGvVByV Jul 16 18:39:24 originally the plan was to have a branch for each kernel version Jul 16 18:40:24 Most users are going to want something other than just the latest, and if latest is broken for their kernel, they are never going to have useful patches to push upstream. Jul 16 18:41:07 If someone that understood the code a bit made a way to contribute to other 'upstream' kernels, then you might get a lot more patches/contributors Jul 16 18:41:43 Hauke, this mean anything to you? backports//compat/Kconfig:57: symbol 'BACKPORT_BP_MODULES' redefines option 'modules' already defined by symbol 'MODULES' Jul 16 18:41:56 I guess my attempt had some issue Jul 16 18:45:07 DonkeyHotei: Thanks, I will try it with CONFIG_PACKAGE_shadow=y Jul 16 18:48:32 greearb_: when I was still maintaining backports, there was also a branch for older kernels and I backported important patches to it Jul 16 18:50:06 greearb_: I do not know why you get this error Jul 16 18:52:29 commenting out "option modules" allows it to go forward Jul 16 18:53:42 tobleminer-tSYS: why did you perform a 'magic check' in platform.sh for the devices? Jul 16 18:53:46 they use metadata images Jul 16 18:57:30 * f00b4r0 notices metadata images don't appear to be exempted from extra checks on ramips as they are on ar71xx Jul 16 19:03:27 correct Jul 16 19:23:35 f00b4r0: keep in mind that metadata are not yet enforced in ramips Jul 16 19:24:11 f00b4r0: there are still three legacy images which prevent doing so Jul 16 19:24:36 I'm thinking about dropping the image build code for these three boards Jul 16 19:30:56 mkresin: ok, I'll tag along then Jul 16 19:33:18 mkresin: I've added a little change to mt7621.mk in #1185. Not sure all the variables defined there are strictly necessary but I feel much less confident about the image build system for this platform than I am for ar71xx so I won't touch it Jul 16 19:34:16 f00b4r0: PR#1185? Jul 16 19:34:20 yes Jul 16 19:34:29 in fact I think I should probably define SUPPORTED_DEVICES Jul 16 19:34:29 will check in a few minutes Jul 16 19:34:55 f00b4r0: something else I already spotted are two partitions at offset 0x0 Jul 16 19:35:11 yes. There's a big 'top level' one and sub-ones Jul 16 19:35:48 there are holes between the sub-ones and I want to make it plain that the whole area covered by the top level one is never to be touched Jul 16 19:35:56 f00b4r0: can't we get rid of the toplevel one? everytime I look at it, it looks like a mistake Jul 16 19:36:33 i don't see how it looks anymore like a mistake than having a top-level 'firmware' that's split into sublevels? Jul 16 19:36:38 f00b4r0: and you should see dtc warning because of partition@1 vs. reg = <0x0> Jul 16 19:37:03 I didn't test that code I don't have the hardware... Jul 16 19:37:08 * f00b4r0 throws tobleminer-tSYS under the bus Jul 16 19:37:27 f00b4r0: compile an image and check the device tree compilers output Jul 16 19:37:51 f00b4r0: where do we have a firmware + child partitions? Jul 16 19:38:03 everywhere there's a firmware partition? Jul 16 19:38:12 firmware is split into kernel + rootfs Jul 16 19:38:18 rootfs is split into rootfs_data and user data Jul 16 19:38:25 f00b4r0: yes, but it isn't defined in the dts files Jul 16 19:38:52 i'm not writing a routerboot splitter. Jul 16 19:39:40 f00b4r0: I'm not sure if we're talking about the same Jul 16 19:40:27 f00b4r0: this partition is IMHO fine to remove: https://github.com/f00b4r0/openwrt/blob/7b927f75de23ea90c411ede3e13a801dba1442f1/target/linux/ramips/dts/RBM11G.dts#L96 Jul 16 19:40:40 it's not imo Jul 16 19:40:54 the oem code defines that whole segment as reserved Jul 16 19:41:03 what we extract from that is pure reverse engineering Jul 16 19:41:09 nothing says we're correct. Jul 16 19:41:19 nothing says the current holes won't be filled later Jul 16 19:41:38 if the top level isn't defined, no userspace tool will ever be able to look at the holes. Jul 16 19:42:00 f00b4r0: if the holes contain data, they should be added as partitions Jul 16 19:42:10 read again. Jul 16 19:42:34 i see a lot of ar71xx hardcoding overlapping partitions in the image makefiles. How's that not a problem there and one here? Jul 16 19:42:50 Hauke, after hacking some bcom Kcofig to not create loops, I now get this in a loop: Jul 16 19:42:56 scripts/kconfig/conf --silentoldconfig Kconfig Jul 16 19:42:56 GEN ./Makefile Jul 16 19:43:04 f00b4r0: long story short, ar71xx is a mess Jul 16 19:43:09 I'm about to give up no backports for now unless perhaps someone knows how to fix this? Jul 16 19:43:26 more to the point: what is the correct way to define top-level partitions in DTS? Jul 16 19:44:10 f00b4r0: there is no way, as it is usually not required/useful Jul 16 19:45:48 I think it can be useful in the present case Jul 16 19:46:36 alternatively I can define the holes as separate partitions. That will be absolutely ugly, of course. Jul 16 19:47:29 f00b4r0: or you leave to space for now and add the partitions at the time the space contains values Jul 16 19:47:50 and how would i know if the space contains values if I can't ask a user to dump the segment? Jul 16 19:49:25 make it easy and set the partition sizes to val which aligns with the start of the partition Jul 16 19:50:35 no. That's not going to make it easy (still several partitions to dump one at a time) and it doesn't match the current partition scheme Jul 16 19:51:04 ... Jul 16 19:51:50 are you objecting to top level mtd partitions or are you objecting to the fact that DTS can't handle them properly? Jul 16 19:53:30 f00b4r0: both. neither is something like that intended for fixed-partitions, nor do I see the need. Jul 16 19:54:08 f00b4r0: in such cases, I've added a reserved or empty partition Jul 16 19:54:37 Hauke, this fixes the loop, but then compile blows up all over the place once it tries to compile backports stuff. So, I'm off to manual hacking. Just possibly the patch in this email is worth applying though: Jul 16 19:54:40 https://www.spinics.net/lists/backports/msg04115.html Jul 16 19:55:27 f00b4r0: your image build code looks fine Jul 16 19:55:41 mkresin: I'm getting out of my way to fix broken commits. I'd like a modicum of convenience and safety (as in making sure it's plain and obvious to anyone that the whole segment should be considered "OEM reserved") Jul 16 19:55:56 if you disagree with that, feel free to ignore or edit my PR. But don't expect me to provide further support if you do. Jul 16 19:58:29 in the longer term I expect it will be possible to write a proper runtime splitter for that partition Jul 16 19:58:38 I'm not signing up for that task now. Jul 16 20:01:39 f00b4r0: I still don't get it. I pointed out a formal issue, provided alternative ways to get want you want to do with the result that you are upset Jul 16 20:02:03 i want to be able to request a single, coherent dump of the whole OEM segment Jul 16 20:02:14 I can eventually build a database of such segments, Jul 16 20:02:22 against which to test reverse engineered tools Jul 16 20:02:24 f00b4r0: which isn't what the device tree is for. Jul 16 20:02:35 you offer no alternative to that Jul 16 20:03:35 the alternative to define partitions with an empty label, dump /dev/mtd* and concat the result Jul 16 20:03:35 is save & apply supposed to look different in rc2: https://pasteboard.co/HuLHIni.png ? Jul 16 20:03:41 I mean luci Jul 16 20:03:48 mkresin: that's more work and error prone. Jul 16 20:04:00 hence me saying I won't deal with that. Jul 16 20:04:03 everything else has "transparent" background Jul 16 20:10:16 mkresin: if top level partitions are so unsightly and unnecessary, I'll point out that the splitter code can be adjusted to make them go away. Jul 16 20:10:47 f00b4r0: you miss the point that the dts should describe the hardware and not the usecase Jul 16 20:12:29 partitions can hardly be considered a hardware description. Jul 16 20:13:12 and should I mention the recent patch that make use of DTS to handle led _use cases_ ? Jul 16 20:17:12 greearb_: what are you doing? Jul 16 20:17:18 breaking shit Jul 16 20:17:38 I would like to reproduce your problem Jul 16 20:17:40 real goal is to backport my ath10k-ct kernel to a customer's 3.19 kernel Jul 16 20:18:01 ok Jul 16 20:18:11 your ath10-ct is based on kernel 4.16? Jul 16 20:18:16 yes Jul 16 20:18:20 f00b4r0: https://elinux.org/Device_Tree_Usage#aliases_Node Jul 16 20:18:48 f00b4r0: the use case is still in userspace Jul 16 20:19:18 Hauke, I used this as starting point: ./gentree.py --integrate --clean --gitdebug --git-revision 08044ae21d2df7385fbb2363d855d3244e060994 ~/git/linux-ct-4.16 ~/git/linux.3.19-mg90/ --verbose Jul 16 20:19:50 I would guess any 3.19-ish kernel would reproduce the pain Jul 16 20:19:54 f00b4r0: anyway, I left a comment on your PR Jul 16 20:20:02 greearb_: I used this in my last experiment: ./gentree.py --clean /home/hauke/linux/linux-next-clean/ ../backports-4.18-rc2-2 --verbose Jul 16 20:20:34 is --integrate not a good idea maybe? Jul 16 20:20:47 greearb_: --integarte is not so good Jul 16 20:20:52 mkresin: the use-case is in userspace and the DTS conveniently provides a way to make that use case easier to implement. 'led-status' alias isn't hardware description. Jul 16 20:20:53 sounds familiar? Jul 16 20:21:16 this will help you to integarte it into a kernel tree when you want to compile this stuff into the kernel Jul 16 20:21:22 regardless, i've had it. Do what you want, I'll stay away from platforms you're involved with. It's obvious you have no problem accepting some half-broken commits and then bikeshed me when I try to fix them. Jul 16 20:21:38 greearb_: when you want to have it as an external module, please use it without --integarte Jul 16 20:21:52 f00b4r0: whatever you mean Jul 16 20:22:05 I have never used --integrate Luis added it some years ago I think Jul 16 20:22:31 it will be easiest for my user if it is part of the kernel I think, but maybe that option is just broken Jul 16 20:22:58 yes Jul 16 20:23:13 you can also remove stuff you do not need from the copy-list file Jul 16 20:23:28 but this could also trigger other problems ;-) Jul 16 20:24:11 it can't be partition@0 since there would be two nodes with the same name Jul 16 21:08:59 Hauke, once I run that gentree cmd, then what? The target kernel was blown away :P Jul 16 21:09:21 I guess I re-clone the tree and somehow build backports against it? Jul 16 21:10:10 jow: i've problem backporting my mtd fixes due to 6734f32d91cd31ec5c2cc44a021a9839232fe587 ("mtd: add build hack to reintroduce shared mtd for older releases") Jul 16 21:10:20 jow: cherry-picking should normaly bump relase 21 to 22 Jul 16 21:10:33 but now there is that magic: Jul 16 21:10:35 PKG_RELEASE:=21$(if $(SDK),,.1) Jul 16 21:10:52 when cherry-picking commit bumping that to 22, should i use: Jul 16 21:11:00 1) PKG_RELEASE:=22 Jul 16 21:11:03 2) PKG_RELEASE:=22$(if $(SDK),,.1) Jul 16 21:23:37 wow, i had no idea it can be done: kernel: insmod phy-broadcom module during preinit Jul 16 21:39:35 Probably really off topic, but I was looking to purchase a new router that can actually handle my 1gbps download connection. My current TP-LINK AC1200 is only really giving me 250-300 mbps with a wired connection. Any suggestions/thoughts for good routers that can run LEDE/openWRT? Jul 16 21:40:52 greearb_: ~/git/linux.3.19-mg90/ contains a backports tree you can build against a kernel like 3.19 Jul 16 21:42:01 Astray: NETGEAR R7800 or Linksys WRT3200ACM Jul 16 21:42:10 Astray: As long as you /just need routing/ at 1 GBit/s (and not additional services like VPN on top), mt7621 and mvebu should provide that performance (mvebu has a faster CPU, so probably providing more margin for additional services than mt7621); but with a 1 GBit/s WAN connection I'd seriously consider x86 as routing platform. Jul 16 21:42:32 pkgadd: +1 Jul 16 21:42:56 mt7261 is getting/has gotten hw accel too yeah? Jul 16 21:43:03 I would still recommend that or x86 (apu2) Jul 16 21:43:22 apu2 has 2 mpcie slots for WNICs Jul 16 21:43:23 koops: no, ipq806x can't do WAN-to-LAN routing 1 GBit/s (yet, the hardware may offer that in the future if the two nss cores could be used under OpenWrt, but right now, it's limited to roughly 350-400 MBit/s) Jul 16 21:43:46 I stand corrected, Linksys only. Jul 16 21:44:11 m4t: yes Jul 16 21:44:19 Linksys was practically designed with purpose of being compatible with LEDE Jul 16 21:44:37 if you just want wired and separate wireless, this guy is on sale. mt7261 powered. https://slickdeals.net/f/11774203-ubiquiti-5-port-edgerouter-x-advanced-gigabit-router-w-poe-sfp-er-x-sfp-us-55ac-newegg Jul 16 21:45:16 If you want VPN, you would need at least desktop i3 CPU though. Jul 16 21:45:34 that for wireguard. Jul 16 21:45:47 For OpenVPN, forget it, nothing will run gigabit on it. Jul 16 21:46:22 I'm looking to future proof myself pretty well if possible Jul 16 21:46:23 pkgadd: can it do gigabit lan to lan? Jul 16 21:46:29 Then go x86 Jul 16 21:46:31 threadripper then Jul 16 21:46:37 ;) Jul 16 21:46:49 no. get i3 Jul 16 21:46:49 kmod-ipt-ipsec doesnt compile xt_policy.ko module Jul 16 21:47:32 koops: LAN to LAN happens exclusively on the hardware switch, without the CPU/ SOC being involved - that means (almost) line speed, so 1 GBit/s (at least more than 950 MBit/s) Jul 16 21:48:05 Pentium should be fine even Jul 16 21:48:27 Can I somehow make even LAN to LAN go through the SoC for firewalling? Jul 16 21:48:31 I'm a little confused by what you mean go x86 or i3. Is that hardware to be looking for in the routers? Jul 16 21:49:01 No. Get a desktop mITX mobo and Intel core i3 cpu. Jul 16 21:49:13 Oh build a PC to be the router? Jul 16 21:49:25 Yeah, for power you can use picoPSU. Jul 16 21:49:52 Installing OpenWRT is possible on any PC with at least Ethernet ports. Jul 16 21:50:08 For wireless, get Ubiquity AP Jul 16 21:50:21 Okay I'm not looking for that level of future proofing. Just a router I can purchase and place on my desk would be plenty fine. Jul 16 21:50:31 Hauke, yeah, I realized that...but I have never build backports against a tree before. And, in the end, it needs to go into a yocto build system, so I've pinged my customer to see if they know how to integrate it. Jul 16 21:50:43 Then WRT3200ACM is the only option really. Jul 16 21:51:15 "only option"? :( Jul 16 21:51:35 For future-proof open-source router? Jul 16 21:51:57 nope Jul 16 21:52:14 If by future-proof you mean "can run VPN with at least a sizable fraction of gigabit" Jul 16 21:52:25 There's nothing to future-proof otherwise. Jul 16 21:52:39 There's no routers for speeds higher than 1Gbps. Jul 16 21:52:59 gotcha Jul 16 21:53:00 koops, there are, btw...some can bond two 1G interfaces Jul 16 21:53:07 at least the ones available for consumers lol. Jul 16 21:53:10 This is on sale right: https://www.amazon.com/Linksys-Max-Stream-MU-MIMO-Tri-Band-EA9300/dp/B071DKY952/ref=gbps_tit_m-8_60d4_deb468f9?smid=ATVPDKIKX0DER&pf_rd_p=d4d850bb-9bd6-4e1a-8756-8ad85fd760d4&pf_rd_s=merchandised-search-8&pf_rd_t=101&pf_rd_i=13887280011&pf_rd_m=ATVPDKIKX0DER&pf_rd_r=BRGZ8KJKEDV9BBR3GXMC Jul 16 21:53:28 but I don't think it's supported yet Jul 16 21:53:45 some are starting to ship w/10gbe e.g. https://wikidevi.com/wiki/Netgear_R9000 Jul 16 21:54:12 i highly doubt it could handle all of that, maybe > 1GB/s tho Jul 16 21:54:13 Doesn't look like the one Astray sent supports OpenWRT Jul 16 21:56:21 Future proof = 10Gbps ports and 802.11ax IMO, and open-source supported. None of these are currently available. Jul 16 21:56:41 I may be wrong but there's no ax routers on the market at all currently. Jul 16 21:56:44 only ad Jul 16 21:57:05 depends what you mean by future-proof heh Jul 16 21:57:24 Astray: what do you mean by future proof? Jul 16 21:57:27 R9000 has 10Gbps ports but really, do you think it's going to do 10Gbps LAN to WAN? Jul 16 21:57:49 And what's the point of these ports if it doesn't lol. Jul 16 21:58:04 Consumer level future proof, 1 gig down handled either through LAN or WLAN Jul 16 21:58:25 there are, not many - not yet supported, but there are Jul 16 21:58:38 https://wikidevi.com/wiki/Special:Ask?title=Special%3AAsk&q=%3Cq%3E[[CPU1+model::~IPQ807*]] Jul 16 21:58:54 10 gbps is not something I'll likely need for next decade Jul 16 21:59:05 Astray: QoS/traffic shaping or just nat/firewall? Jul 16 21:59:06 for me it's a router I can run for 10 years. Jul 16 21:59:20 let's say 5 years then Jul 16 21:59:58 IMO router should be able to run until it stops being supported by openwrt. Jul 16 22:00:02 like 4mb ones. Jul 16 22:00:04 I have an arris SB8200 modem currently and don't expect to need to replace it within the next five years since I've only just got access to gigabit Jul 16 22:00:22 Astray: in practice, the linksys wrt32acm/ wrt32x, cz.nic turris omnia or a nice x86 system, e.g. PC-Engines APU2/ APU3/ APU4 or some cheap 4+ port x86 systems from aliexpress/ alibaba etc. are the best option for your use case Jul 16 22:00:23 I have a 2MB router currently. Jul 16 22:00:38 It runs proprietary firmware lol. Jul 16 22:01:05 i looked into the omnia, it seems very difficult to obtain in the US Jul 16 22:01:43 or something mvebu based from clearfog or espressobin (x86), or ... Jul 16 22:01:54 It's unlikely that in event BIOS vulnerability will be found in a box by QOTOM or Partaker it will be patched. Jul 16 22:01:59 s/x86/mvebu/ Jul 16 22:01:59 their bios is one-off. Jul 16 22:02:11 what are the chances of the one I posted eventually getting WRT support? Does Linksys lock their non open source stuff down hard? Jul 16 22:02:20 I don't think so. Jul 16 22:02:56 These Chinese manufactured firewall appliances are unlikely to be supported by manufacturer for long too. Jul 16 22:02:57 Astray: very, very low - probably zero Jul 16 22:03:07 Better go with APU2. It uses coreboot. Jul 16 22:03:25 If it's broadcom, then 0. Jul 16 22:04:15 The ones by protectli are rebranded QOTOM. Jul 16 22:04:34 * m4t likes the espressobin. lack of (non 3d printed) case though :| Jul 16 22:05:00 I wish Novena was supported. That's fully open source hardware. Jul 16 22:05:12 apu2? Jul 16 22:05:16 huh, that netgear R9000 thing Jul 16 22:05:25 https://www.pcengines.ch/apu2.htm Jul 16 22:05:31 Yeah, APU2 will work. Jul 16 22:07:34 I think you can order 3d printing of the case. Jul 16 22:07:41 (m4t) Jul 16 22:07:59 yeah there are services for that. i saw trhis https://upon2020.com/blog/2017/12/3d-printable-box-for-the-espressobin-plus-a-hard-disk/ Jul 16 22:08:09 the one I posted uses broadcom cpus Jul 16 22:08:18 Then forget about it. Jul 16 22:08:27 Broadcom is no. Jul 16 22:08:28 https://wikidevi.com/wiki/Linksys_EA9300 Jul 16 22:08:36 gotcha Jul 16 22:08:40 Astray: +1 for the wrt3200acm Jul 16 22:08:44 Get something with Marvell or MediaTek. Jul 16 22:08:57 Alrighty, I think I'll look into the WRT3200acm Jul 16 22:08:58 anyone else feel like we dodged a bullet with the broadcom qualcomm thing? Jul 16 22:09:03 Or Intel (x86) Jul 16 22:09:05 seems like the easiest and most recommended Jul 16 22:09:31 Don't really feel like going down a whole rabbit whole of learning to build and set up a PC specifically to be a router Jul 16 22:09:35 It won't run gigabit VPN though. Good luck with 200Mbps Jul 16 22:09:49 that's fine, I don't have a VPN yet Jul 16 22:10:06 Get one if you torrent (in any country, not only US) Jul 16 22:10:28 I will if I start running into issues while torrenting the things that I do Jul 16 22:10:30 In Russia we already learned that the hard way (lawsuits) Jul 16 22:11:00 You will eventually get DMCA notice or maybe even get sued outright. Jul 16 22:11:22 APU2 is easier to use than building a PC. Jul 16 22:11:29 Go with it if you can. Jul 16 22:11:31 I've only been served DMCA notices a few times on things that things I should've been protecting myself on Jul 16 22:11:52 Even better get a seedbox on OVH or Online.net Jul 16 22:13:43 These APU things aren't that expensive Jul 16 22:13:47 is there a guide somewhere? Jul 16 22:13:49 m4t> anyone else feel like we dodged a bullet with the broadcom qualcomm thing? --- what do you mean? Jul 16 22:14:04 broadcom is terrible for open source Jul 16 22:14:15 if broadcom acquired qualcomm, what would happen? Jul 16 22:14:19 That if qualcomm did the same, we would have been screwed. Jul 16 22:14:29 doom Jul 16 22:14:41 qualcomm is surprisingly friendly to open-source Jul 16 22:14:51 really? Jul 16 22:14:52 you think? Jul 16 22:15:11 kind of Jul 16 22:15:11 well they don't try to block open source from their devices Jul 16 22:15:23 most of the firmware are closed source Jul 16 22:15:26 they also don't support anything post atheros Jul 16 22:15:35 and they QCA6xxx are pieces of junk Jul 16 22:15:41 their Jul 16 22:15:42 basically broadcom at this point Jul 16 22:16:12 so, it's not too bad, but we're doomed if broadcom buys qualcomm Jul 16 22:16:13 So Marvell and MediaTek are the only ones that are friendly? Jul 16 22:16:25 marvell aren't either Jul 16 22:16:47 ask them for the code to drive their proper chips Jul 16 22:17:00 Astray: the APU2 is basically a normal x86 system, without graphics card (serial console) and a slightly weird 'BIOS' (coreboot), you basically just write the OpenWrt to the mass storage - insert it and off you go Jul 16 22:18:02 running linux (OpenWrt) it is supposed to do 1 GBit/s (not quite using FreeBSD/ pfsense/ opnsense though) Jul 16 22:18:23 MediaTek isn't very friendly too actually Jul 16 22:18:36 mediatek are only "friendly" for their iot crap Jul 16 22:18:43 Avoid pfsense, use OPN Jul 16 22:18:54 Company that owns pf is bonkers Jul 16 22:19:03 they resell APU2 for $499 Jul 16 22:19:08 for one lot Jul 16 22:19:12 why? coz they sell fast equipment? :P Jul 16 22:19:22 simply for preinstalling pf on it Jul 16 22:19:30 R7800, wrt3200acm, or APU thingy? Jul 16 22:19:39 No R7800 Jul 16 22:19:43 koops: not quite Jul 16 22:19:43 it won't handle the load Jul 16 22:19:51 okay Jul 16 22:20:02 Better get APU then. Jul 16 22:20:04 how well does the APU thing work with VPN? Jul 16 22:20:23 At least you can plug new wireless card with ax support into it. Jul 16 22:20:30 when it will be available Jul 16 22:20:52 with wrt3200acm you will be stuck with ac until you get a new router (probably very expensive) Jul 16 22:21:03 Maybe it will run 500Mbps Jul 16 22:21:10 Maybe (wireguard) Jul 16 22:21:29 But for OpenVPN good luck with 20Mbps Jul 16 22:21:48 Wireguard is supported by Mullvad for example Jul 16 22:21:52 Seems reputable Jul 16 22:21:59 or set up on you own server. Jul 16 22:22:01 wireguard solves all the worlds problems :DDDD Jul 16 22:22:21 the rest are solved by shadowsocks xD Jul 16 22:22:33 nah you're alright :D Jul 16 22:24:54 At least you can plug new wireless card with ax support into it. --- is there even a driver for that chipset yet? Jul 16 22:25:42 wireguard still needs an audit Jul 16 22:25:47 it's not out yet. Jul 16 22:27:03 koops: you will be never satisfied, an audit won't change anything Jul 16 22:27:04 :) Jul 16 22:27:31 Of course, there needs to be many audits. Jul 16 22:29:26 Truecrypt had an audit, and it missed a serious vulnerability Jul 16 22:29:44 And it was trivial. Jul 16 22:30:24 I also want a router that can be upgraded to ax like you upgrade a PC. Jul 16 22:30:47 is this about development? :D Jul 16 22:30:49 Or maybe a wired-only router with wireless functionality being a cheap router I can replace. Jul 16 22:31:35 Yeah, Astray: your question probably should also be asked in #openwrt Jul 16 22:35:37 Astray: You may duplicate DonkeyHotei's setup: x86 main router connected to one or more cheap routers that function as APs. Jul 16 22:36:02 replace as needed. Jul 16 22:47:13 jwh: Oh, why do you think I won't be satisfied with an audit? Jul 16 22:47:24 "there still may be backdoors"? lol Jul 16 22:48:42 The TrueCrypt vulnerability is so severe it may as well can be a backdoor. Same effect. Jul 16 22:48:59 It allows detecting the presence of a hidden volume. Jul 16 22:49:03 koops: please tone it down a notch, no one is saying that what you're saying would be wrong, but it's getting a bit excessive Jul 16 22:49:18 Thanks. Jul 16 22:49:31 Will take a note. Jul 16 22:49:53 ah someone more diplomatic heh Jul 16 22:54:31 is there a 17.01.5 changelog out there? Jul 16 22:54:52 alright, well I'll look into these other options, but wrt3200acm is definitely the easiest. If I find a decent price on a used one I'll probably just do that Jul 16 22:55:35 Avoid used routers or alternatively reflash them directly somehow. Jul 16 22:55:47 Ah, you can't reflash bootloader. Jul 16 22:55:50 So avoid. Jul 16 22:56:06 nobody is as paranoid as ou Jul 16 22:56:07 you Jul 16 22:56:25 I would avoid personally. Jul 16 22:56:51 yes, keep the paranoia to yourself, thats all Jul 16 22:56:53 There's also ART table which has settings specific to a router, which also shouldn't be reflashed. Jul 16 22:57:09 ok. get used but don't blame me when your pc gets filled up with gay porn. Jul 16 22:57:17 or something worse happens. Jul 16 22:57:22 that might be the intention anyway xd Jul 16 22:58:38 I would never be satisfied with a used device if I couldn't reflash it completely (directly, without using the device's possibly compromised reflash function). Jul 16 22:58:46 * mangix grabs popcorn Jul 16 22:58:49 like attaching spi programmer to it. Jul 16 22:59:02 https://openwrt.org/releases/17.01/notes-17.01.5 Jul 16 22:59:10 ntd: nice Jul 16 22:59:42 Oh, how can Meltdown affect routers? Jul 16 22:59:49 They don't run untrusted code. Jul 16 23:00:11 * mangix smiles while chewing popcorn Jul 16 23:00:24 just stop Jul 16 23:00:32 They don't have any type of security features against untrusted code either. Jul 16 23:00:38 So why patch Meltdown lol Jul 16 23:00:47 No one browses directly from a router. Jul 16 23:01:29 Also shouldn't that be Spectre, isn't Meltdown Intel-specific? Jul 16 23:01:35 meltdown affects intel and arm64. nobody else Jul 16 23:01:49 Never seen a router with arm64. Jul 16 23:01:58 it's mainly an issue with VMs Jul 16 23:02:09 Can a router even run VM? Jul 16 23:02:19 it would melt down lol. Jul 16 23:02:21 of course Jul 16 23:02:26 LXC Jul 16 23:03:12 I'm moving to that sort of setup soon for my big x86 boxes prob heh Jul 16 23:06:56 What about Mikrotik routerboards and Ubiquiti's similar devices? Do they have fast CPU? Jul 16 23:08:40 no Jul 16 23:08:44 they're routers Jul 16 23:09:20 day must end in a y, cable is out Jul 16 23:09:23 * jwh rollseyes Jul 16 23:09:57 Broadband No known issues Jul 16 23:09:59 of course Jul 16 23:11:30 jwh: Virgin? Jul 16 23:12:34 MediaTek MT7621AT. That's MIPS, so not fast. Jul 16 23:12:56 Yeah, only APU2 or mini-itx mobo can work for me. Jul 16 23:13:09 I don't think anythign can work for you at this point Jul 16 23:13:38 Don't know if anyone makes an affordable roll of ASA yet.. Jul 16 23:13:45 Why? Jul 16 23:13:52 nothing will work? Jul 16 23:14:00 ASA? Jul 16 23:14:03 oops, wrong channel Jul 16 23:15:50 If I set 2 VLANs in OpenWRT, will traffic go between them by default? Jul 16 23:16:08 Do I have to explicitly block or explicitly allow it? Jul 16 23:17:29 Monkeh: ye Jul 16 23:17:37 jwh: Clue's in the name Jul 16 23:17:40 heh Jul 16 23:17:42 Networking virgins, what do you expect? Jul 16 23:17:58 only got one downstream channel locked, but not getting anything from the network Jul 16 23:18:05 its pretty typical Jul 16 23:18:19 one channel is fine, of coure Jul 16 23:18:21 course Jul 16 23:18:24 its still 50 meg Jul 16 23:18:55 except I don't think its setup to degrade gracefully Jul 16 23:18:57 gg, and such Jul 16 23:21:39 Any guides on making a combined router+NAS? Jul 16 23:21:48 x86 Jul 16 23:25:22 install samba, configure it as normal, /done Jul 16 23:25:39 no black magic involved Jul 16 23:26:37 I want a proper one, with RAID etc. Jul 16 23:26:45 any x86 box Jul 16 23:26:52 also, #openwrt Jul 16 23:26:53 stop it Jul 16 23:27:03 well, mdadm and/ or lvm2 don't really make it any more complex Jul 16 23:28:09 sine we're off topic anyway, real estate agents suck Jul 16 23:28:12 since* Jul 17 00:45:33 and I'd like to add that vacuum cleaner suck too :) Jul 17 00:59:10 Can someone give me a hand with the Piezoelectric buzzer method Jul 17 00:59:26 can you find the Vcc pin with that? Jul 17 01:03:57 Vcc pin should cause a click when you connect it to ground through the speaker Jul 17 01:04:41 Tx pin should make a dial-up sound when you connect it to the ground through the speaker. Jul 17 01:04:47 Rx pin should be quiet. Jul 17 01:05:20 You can't really find a GND pin with it. Jul 17 01:05:45 thank you very much Jul 17 01:06:09 it only makes the sound while the router is booting. Jul 17 01:06:13 tx Jul 17 01:07:13 allright, if I find the Vcc and Tx I know what the others 2 are Jul 17 01:07:38 Ah, Vcc can click when you disconnect it too. Jul 17 01:09:38 awesome Jul 17 01:10:56 Do you remember what sound dial-up makes? Jul 17 01:10:57 lol Jul 17 01:11:17 Some people have no idea internet used to make sound. Jul 17 01:11:31 I'm old enough to know Jul 17 01:13:01 Do you play Skyrim? Jul 17 01:13:07 I really like it. Jul 17 01:13:35 I play the original 2011 version, 2016 one is extremely slow and has bugs 2011 doesn't. Jul 17 01:13:50 NO Jul 17 01:14:59 I thought you were sleeping. Jul 17 01:15:08 ok he went here first. Jul 17 01:15:22 ok answer in pm if you want to. Jul 17 01:16:29 :0 Jul 17 01:16:30 :) Jul 17 01:21:52 I've played the elder scrolls oblivion back in the day, I actually own Skyrim but never play it due to the lack of time Jul 17 02:44:01 latest mt76, clients cant connect to wifi Jul 17 02:44:12 Tue Jul 17 02:35:40 2018 daemon.info hostapd: wlan0: STA 38:a4:ed:d5:c0:ba IEEE 802.11: disassociated due to inactivity Jul 17 02:44:12 Tue Jul 17 02:35:41 2018 daemon.info hostapd: wlan0: STA 38:a4:ed:d5:c0:ba IEEE 802.11: deauthenticated due to inactivity (timer DEAUTH/REMOVE) Jul 17 02:47:34 tried to off wifi Jul 17 02:47:42 now everything hangs Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 daemon.notice hostapd: wlan0: interface state ENABLED->DISABLED Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 daemon.notice hostapd: wlan0: AP-DISABLED Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 daemon.notice hostapd: wlan0: CTRL-EVENT-TERMINATING Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 daemon.notice hostapd: nl80211: deinit ifname=wlan0 disabled_11b_rates=0 Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 kern.info kernel: [70416.666327] device wlan0 left promiscuous mode Jul 17 02:47:57 Tue Jul 17 02:38:54 2018 kern.info kernel: [70416.670981] br-lan: port 2(wlan0) entered disabled state Jul 17 02:48:00 Tue Jul 17 02:38:55 2018 daemon.notice netifd: Network device 'wlan0' link is down Jul 17 02:48:04 Tue Jul 17 02:38:55 2018 kern.warn kernel: [70417.742822] MCU message 8 (seq 9) timed out Jul 17 02:48:06 Tue Jul 17 02:39:24 2018 daemon.notice netifd: radio0 (3905): Command failed: Request timed out Jul 17 02:54:15 is that on 5ghz, or 2.4? Jul 17 02:57:28 root@OpenWrt:~# reboot Jul 17 02:57:28 -ash: can't fork: Out of memory Jul 17 02:57:32 2.4 Jul 17 02:57:46 :/ **** ENDING LOGGING AT Tue Jul 17 03:00:03 2018