**** BEGIN LOGGING AT Thu Nov 15 03:00:00 2018 Nov 15 03:22:49 greearb_: If you just want the LED on hard you disable the trigger and simply set the brightness Nov 15 03:32:02 I have a device with a watchdog, to keep it from rebooting I have this shell script: https://pastebin.com/D8NqZg8i Nov 15 03:32:25 this script is placed in /bin/feeder.sh Nov 15 03:32:43 now if I wish to perform a sysupgrade, it will kill all processes including the feeder Nov 15 03:33:41 whats a clean way to keep this running even during a sysupgrade? From first glance I was thinking of simply respawning the script after processes are killed (in /lib/upgrade/stage2 script). Is this the right way? Nov 15 03:37:51 abenz: The correct solution is to let the kernel handle the watchdog - see gpio-wdt Nov 15 03:44:16 Monkeh: thanks for the tip. seems all thats needed is a small modification to dts. Will try it now Nov 15 03:55:25 Monkeh: are you familiar with the flags? I'm not sure I fully understand always-running flag Nov 15 03:55:27 https://www.kernel.org/doc/Documentation/devicetree/bindings/watchdog/gpio-wdt.txt Nov 15 03:58:48 abenz: Not entirely clear myself, not that familiar Nov 15 06:10:40 stintel: which mail exactly? Nov 15 07:39:07 https://github.com/torvalds/linux/blob/5929a1f0ff30d04ccf4b0f9c648e7aa8bc816bbd/Documentation/devicetree/bindings/watchdog/gpio-wdt.txt#L25 Nov 15 07:39:33 can someone help me with this line, I know in the device I'm working on its on gpio2, but how do I identify the pin? (in example its 9) Nov 15 07:39:58 I tried looking around in /sys/class/gpio/gpio2 but can't find anything related.. Nov 15 07:41:10 I have the device running btw (with a manual script toggling the value so the device wont reboot). So with above I'm trying to do this cleanly from kernel Nov 15 08:24:12 jow: see query Nov 15 08:24:38 stintel, o/ Nov 15 08:25:05 nitroshift: yo Nov 15 08:25:16 bbl Nov 15 08:25:35 * nitroshift throws stintel a pillow Nov 15 08:25:44 Hi people any one know what to put for subnet in the https over dns proxy app for luci? Nov 15 08:27:07 I set listserver to 27.0.0.1#5053 in /etc/dhcp but if I stop the proxy dns still works Nov 15 08:31:19 stintel: done Nov 15 08:32:00 Tapper: I think you must also tell dnsmasq to ignore /tmp/resolv.conf.auto Nov 15 08:32:52 Tapper: uci set dhcp.@dnsmasq[0].noresolv=1 Nov 15 08:32:58 Tapper: uci commit dhcp Nov 15 08:33:05 Tapper: /etc/init.d/dnsmasq restart Nov 15 08:33:31 the it will use only whats specified as server=... Nov 15 08:36:57 ok jow thanks Nov 15 08:43:40 dns is now not working Nov 15 08:43:53 jow Nov 15 08:44:08 which is expected, dnsmasq does not know where to get its dns from Nov 15 08:44:34 try this: Nov 15 08:44:40 The proxy is running and should be handing out dns on 27.0.0.1#5053 Nov 15 08:44:49 uci add_list dhcp.@dnsmasq[0].server=8.8.8.8 Nov 15 08:44:56 uci commit dhcp Nov 15 08:45:03 /etc/init.d/dnsmasq restart Nov 15 08:45:06 is it working then? Nov 15 08:45:36 if yes, then there seems to be an issue with the proxy Nov 15 08:46:03 jow yes working again Nov 15 08:46:44 I don't know what to put for subnet and proxy address in the luci page for the proxy Nov 15 08:46:54 me neither Nov 15 08:46:58 lol Nov 15 08:47:08 I took a look at it today but couldn't make sense of it Nov 15 08:47:08 thanks mate fro helping. Nov 15 08:47:11 lol Nov 15 08:47:13 kk Nov 15 08:47:38 I am going to remove it and ask the orther to rite up a wiki Nov 15 08:48:09 the proxy appears to be optional Nov 15 08:48:35 ah and edns client subnet appears to be a dns extension Nov 15 08:50:15 Tapper: so the purpose of this subnet is this: it tells the dns server you're querying about your real subnet (I suppose your wan ip) Nov 15 08:50:59 Tapper: this is useful for CDNs and geoip based load balancers and ensures that you get dns replies for servers close to your actual location Nov 15 08:51:23 so set my sub net to 192.168.1.1? Nov 15 08:51:49 for example if you're in the UK and proxy your DNS-over-HTTPS request via a proxy server in the US, you can then tell the DNS server that you want to get responses with servers in the UK and not in the US Nov 15 08:52:05 no it should actually be the subnet of your wan interface (which is dynamic) Nov 15 08:52:20 but the way I understand it, it is optional and likely not useful if you do not also use a proxy Nov 15 08:52:52 so you can simply leave proxy and subnet empty Nov 15 08:54:05 O Nov 15 08:54:09 Thanks jow Nov 15 08:55:01 you can also test your proxy without reconfiguring dnsmasq Nov 15 08:55:18 by using nslookup openwrt.org 127.0.0.1#5053 Nov 15 10:31:32 jow I fixt it mate. I had the tick box for dns ecc ticked! Nov 15 10:31:44 silly me! Nov 15 10:32:00 thanks for your help! Nov 15 10:35:25 btw jow thanks for your luci fixes. The onley thing left is the tables. Nov 15 10:36:31 That if they are supposed to look like a table to a sited person. Nov 15 10:36:50 that's if* Nov 15 10:40:38 still investigating solutions for that Nov 15 11:15:24 pfft, 2 colleagues in the same amount of weeks have VPN issues due to being behind bloody CGN Nov 15 11:15:40 * karlp has been having fun with a vpn all morning. Nov 15 11:16:01 I swear some of these "solutions" only seem to serv to keep IT people employed Nov 15 11:16:07 the first one called his ISP, they immediately knew what he was talking about Nov 15 11:16:33 and flagged him with "don't CGN" Nov 15 11:16:54 but the 2nd guy is working from a co-working space where nobody has a clue about the connection Nov 15 11:17:25 luckily both ends have IPv6, so I'm now doing a 4-in-6 IPsec tunnel :P Nov 15 11:17:53 what's a CGN? Nov 15 11:18:06 Carrier Grade NAT Nov 15 11:18:18 anti-carrier, maybe? Nov 15 11:18:27 aka your ISP doing NAT for you, so you share an IP with your neighbors Nov 15 11:18:35 oh, lovely Nov 15 11:19:18 and that really doesn't play well with IPsec VPNs Nov 15 11:19:29 for some reason I don't really understand Nov 15 11:20:32 that's ok, I was running a windows vm so I could rrun the obsolete version of keytalk (obsolete vpn software? sounds great!) so that I could run an activex control that provided a remote terminal services connection to another windows machine so that I could finally ssh/web to my linux machines. Nov 15 11:20:51 this is what I get for asking nicely instead of just installing my own vpn on those machines. Nov 15 11:28:43 ugh Nov 15 11:29:04 yeah, that kind of VPN software they invented to "solve" IPsec problems probably Nov 15 11:32:48 russell--: hey, please test master Nov 15 11:33:06 c6a1bcac16 ("kernel: fix ubifs loosing O_TMPFILE data after power cut") Nov 15 11:33:18 it's a pick of "official" fix Nov 15 11:44:57 rmilecki: ACK Nov 15 11:48:47 rmilecki: there was a V2 of the patch on the linux-mtd mailing list, i haven't looked to see how it differed Nov 15 11:49:04 i missed V2 Nov 15 11:49:13 I asked Richard about it today and he pointed me to the V2 Nov 15 11:49:19 I've tested it and it works ok Nov 15 11:49:50 okay Nov 15 12:05:22 rmilecki: confirmed it fixed my bug Nov 15 12:05:35 hooray :) Nov 15 12:05:46 +1 Nov 15 13:56:14 So I removed "ath10k-ct" and choose "ath10k" instead, but for some reason after flashing the router Im still using ath10k-ct Nov 15 13:58:52 "CONFIG_DEFAULT_kmod-ath10k-ct=y" stays in the .config even though Ive unselected it. Normal? Nov 15 14:12:53 barhom: CONFIG_DEFAULT_* is just the default, not the package selection itself Nov 15 14:22:49 good afternooon Nov 15 14:33:07 nbd: how is it possible that even though ath10k-ct is not installed (I verified with opkg) ethtool -i wlan0 actually shows ath10k-ct being in use Nov 15 14:33:20 some lingering files while building? Nov 15 14:33:36 trying a make clean build now to see if it goes away Nov 15 15:01:24 can someone please explain how does making a module makefile in package/kernel/linux/modules work? Nov 15 15:01:45 I needed a dvb driver, noticed the turris guys made this and it would work for me https://github.com/CZ-NIC/turris-os/blob/test/package/kernel/linux/modules/dvb.mk Nov 15 15:02:29 but when i tried transplanting this into 18.06 it seems that at some point make modules regenerates the kconfig and trips over since there have been new drivers added which this makefile doesn't have Nov 15 15:03:42 why isn't it selecting the default in such a case? Nov 15 16:54:10 barhom, there is the ath10k-ct driver, and then there is the ath10k-ct firmware. The QCA driver can load my (non-htt) firmware, and my driver can load stock firmware. Nov 15 16:55:02 My driver prints out some CT specific info on load, and my firmware has a unique name that is also printed out in dmesg on boot. If you want to put that somewhere I can find it (pastebin, etc), then I can take a look. Nov 15 17:03:32 movi: it works by defining a new KernelModule section and mentioning the required symbols to enable in KCONFIG:=... Nov 15 17:04:29 movi: in some cases, corresponding symbols are forcibly marked as disabled in target/linux/generic/config-* or target/linux/$target/config-*, it should be removed there, then Nov 15 17:09:41 when installing to micro-sd related systems, is there an easy way to tell it to resize partitions to use all available storage? Nov 15 17:22:36 greearb_: it's probably better to make a 3rd partition and use that Nov 15 17:23:53 actually I still haven't figured out what I would like the most :) Nov 15 17:29:22 we have a resize script we use on SSD on x86-64 platforms, I'll talk to my engineer who wrote that and see if it works on the OpenWRT image. I'd prefer one larger partition than more small ones Nov 15 17:31:14 trying to get python and nanohat OLED working on the neo2 now.... Nov 15 17:31:49 greearb_: heh i was just thinking about that Nov 15 17:31:58 about which? Nov 15 17:32:26 when i clone the image to a hdd or whatever and resize the part the new available space does not get reflected in the packages area Nov 15 17:32:40 or do you have to set that in the build menu Nov 15 17:33:02 if/when I get it working, I'll post a patch... Nov 15 18:02:41 jow: i'm looking at other module makefiles and I can't see a KernelModule section anywhere. And it's not that some of the new ones were disabled, it's that their state was undefined and that tripped up make modules Nov 15 18:02:53 undefined - not even mentioned in .config Nov 15 18:11:37 movi: I mean define KernelPackage/xxx Nov 15 18:11:51 movi: pastebin your current diff Nov 15 18:13:02 jow: openwrt diff or the generated kconfig? Nov 15 18:13:50 git diff Nov 15 18:14:15 what my question amounts to is thiss - if i want to declare an additional kernel module config makefile, do i have to handle each possible module in that area of the kernel? (in this case dvb) Nov 15 18:19:38 jow: https://pastebin.com/ArZG8Zrf Nov 15 18:27:21 movi: no, You don't have to do handle every module of media subsystem, You only need to packagethe ones required by the device and their dependencies. Nov 15 18:31:47 tmn505: that's what i thought, yet this happens https://paste.ubuntu.com/p/gtptCdw8cn/ Nov 15 18:32:43 tmn505: i don't handle CONFIG_DVB_DEMUX_SECTION_LOSS_LOG in my makefile, and it's missing completely in the generated kernel .config, which is why i think it trips up Nov 15 18:35:22 pfff, sysupgrade code is giving me a headache Nov 15 18:38:03 movi: Some kernel symbols were introduced in later kernels and they are dependencies from already defined symbols. You need to specify them, either disable or enable them. Nov 15 18:38:51 tmn505: even when i do (by manually altering the kernel .config) it would keep asking me for new usb devices or tuners that are not dependencies Nov 15 18:40:20 movi: DVB_DEMUX_SECTION_LOSS_LOG is unmasked by DVB_CORE Nov 15 18:40:52 movi: KernelPackage/dvb-core enables DVB_CORE, so the responsibility lies there Nov 15 18:41:00 define unmasked Nov 15 18:41:03 movi: add DVB_DEMUX_SECTION_LOSS_LOG=n to the KCONFIG list of KernelPackage/dvb-core Nov 15 18:41:16 unmasked as in; https://elixir.bootlin.com/linux/latest/source/drivers/media/dvb-core/Kconfig#L33 Nov 15 18:41:31 config DVB_DEMUX_SECTION_LOSS_LOG depends on DVB_CORE != n Nov 15 18:41:48 means whoever enables DVB_CORE also has to take care of disabling DVB_DEMUX_SECTION_LOSS_LOG Nov 15 18:42:19 (or setting it to something) Nov 15 18:44:30 movi: the media subsystem guys heavily use "select" in their Kconfig files, so that all known devices have already built all modules (example: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/media/usb/dvb-usb/Kconfig#n80). You'll have to define all symbols which are selected. Nov 15 18:44:30 now i fail on CONFIG_DVB_USB_AU6610 https://paste.ubuntu.com/p/tvrrWMjDjx/ Nov 15 18:45:56 movi: basically add each symbol it quieries as XXX=n to KCONFIG:= until there's nothing left Nov 15 18:46:33 tmn505: and these are evaluated and need to match regardless if they're =m =n or =y ? Nov 15 18:46:46 movi: try with this PR: https://github.com/openwrt/openwrt/pull/1451. It should be updated, i didn't test this. Nov 15 18:46:59 yes Nov 15 18:47:30 tmn505: right, now i understand everything :) Nov 15 18:47:37 thank you Nov 15 18:47:54 and thanks for that PR, will check it out Nov 15 19:02:30 jow: in sysupgrade, there is export_bootdevice and export_partdevice Nov 15 19:02:48 jow: the name partdevice is unclear to me where it comes from Nov 15 19:03:26 stintel: iirc export_partdevice is supposed to figure out the /dev/* blockdev node refering to the rootfs partition Nov 15 19:04:10 stintel: since we specify the rootpart using a partuuid in grub, we cannot directly scrape it from /proc/cmdline anymore but have to probe the partition magic of each potential /dev/sd* candidate Nov 15 19:04:40 iirc that shell procedure is supposed to do exactly that Nov 15 19:04:54 jow: I'll stare at the code some more. I would like to completely ditch the offset in that export_partdevice function (if possible), and further consolidate duplicate upgrade code in targets Nov 15 19:05:09 I'm thinking about adding /lib/upgrade/blkdev.sh Nov 15 19:06:03 I am messing around with UDOO Quad (imx6) and OpenWrt, that target has no sysupgrade support for devices with mmc storage Nov 15 19:06:46 stintel: since the partition probe / uuid stuff is really low level, the natural place to put it would be either block-mount of fstools (imho) Nov 15 19:07:18 something like lookup_blkdev {uuid|partuuid} Nov 15 19:07:25 and I would like to avoid duplicating code that is already used in x86/sunxi/brcm2708/omap/apm821xx Nov 15 19:08:27 does what I'm saying make sense to you ? Nov 15 19:09:53 partly :D Nov 15 19:10:24 I do understand that you want to reuse sysupgrade code used on other platforms with ext4 fs Nov 15 19:10:40 and apparently there's something unclear about export_partdevice Nov 15 19:11:00 jow: might be a language thing ;) Nov 15 19:11:05 but I am not entirely sure what you mean exactly with "the name partdevice is unclear to me where it comes from" Nov 15 19:11:23 do you wonder about the name of the function or how it produces its result? Nov 15 19:11:29 the name Nov 15 19:11:51 okay, well sometimes I'm really bad at naming things Nov 15 19:11:52 ok, so going on your explanation its long name could be root_partition_device_node Nov 15 19:12:04 so I take two somewhat related things in the vincinity and join them Nov 15 19:12:26 there was an export_bootdevice and I needed a partition device Nov 15 19:12:31 so born was export_partdev Nov 15 19:12:34 ok Nov 15 19:12:40 thanks Nov 15 19:12:57 I think a good name would be lookup_blkdev_by_partuuid Nov 15 19:12:57 let's see if I can come up with something Nov 15 19:13:30 but for me personally, blkdev is something like /dev/sda Nov 15 19:13:39 that's pretty explanatory Nov 15 19:13:47 while I wanted to emphatize that the function looks up /dev/sda2 Nov 15 19:13:54 yep, I get it now Nov 15 19:14:01 *empathize Nov 15 19:14:14 man I could have just copied some platform.sh and be done with it :D Nov 15 19:33:44 jow: basename $(readlink -f /sys/class/block/mmcblk0p2/..) Nov 15 19:34:39 stintel: hm? Nov 15 19:34:58 ah, it's the other way around Nov 15 19:35:05 lol, maybe I should go do something else Nov 15 19:47:02 * Slimey sprays everyone in the mouth with coffee. Nov 15 19:50:57 * mamarley loses 50 HP from burn damage. Nov 15 19:52:20 oh no it was for that afternoon boost **** ENDING LOGGING AT Fri Nov 16 02:59:58 2018