**** BEGIN LOGGING AT Thu Nov 22 02:59:58 2018 Nov 22 03:05:10 i have a bcm47xx/arm board, with console cable attached and am trying to get an initramfs to boot on it Nov 22 03:08:18 in CFE, i'm first trying to send over an .elf/initramfs image found in build_dir/target-arm*/linux-bcm53xx/ Nov 22 03:08:55 mostly because the *initramfs.trx files arent working as expected Nov 22 03:10:44 i'm trying combinations of: boot -tftp -elf -raw -addr=0x00008000 -max=0x800000 192.168.1.35:vmlinux-initramfs.elf Nov 22 06:31:39 karlp: to achieve readonly access on the file api, you must only allow read, list, stat and md5 Nov 22 06:32:44 karlp: you do need acls to use uhttpd-mod-ubus even if you're logged in via luci Nov 22 06:33:04 karlp: luci itself needs no acls since it talks to ubus locally from Lua code, bypassing uhttpd-mod-ubus Nov 22 08:53:57 ldir: stintel: mind if I add your tested-by tags to the bumps? Nov 22 08:58:52 xback: sure Nov 22 09:01:01 xback: yes, tested onx86/64 too now Nov 22 09:26:27 wow, my ERL is still on 4.14.59 Nov 22 09:29:46 stintel, 0/ Nov 22 09:30:51 nitroshift: yo Nov 22 09:31:16 stintel, do you have a minute? Nov 22 09:31:30 sure Nov 22 09:36:20 igel: that used to work years agou on selected devices Nov 22 09:36:31 igel: not all CFE include support for ELF afair Nov 22 09:37:13 igel: i was using command boot -elf -tftp 192.168.1.2:/openwrt-bcm53xx-vmlinux-initramfs.elf Nov 22 09:37:18 which is pretty much the same as yours Nov 22 09:40:27 blogic: do you have some ath79 device on your desk or just around? could you let me know what model is that and possibly test a patch for me? Nov 22 09:44:13 actually... Nov 22 09:44:30 anyone here with ath79 supported hardware? looking for someone to test a patch for me Nov 22 09:44:38 no real need to bother blogic in particular :) Nov 22 09:46:36 rmilecki: I have one Nov 22 09:46:42 stintel: what device is that? Nov 22 09:47:01 Unifi AP AC Pro Nov 22 09:48:39 stintel: is there something like: Nov 22 09:48:40 2 uimage-fw partitions found on MTD device firmware Nov 22 09:48:44 in your boot log? Nov 22 09:49:22 [ 0.425128] 2 uimage-fw partitions found on MTD device firmware Nov 22 09:49:46 jow: re acls, ok, but how do I tie a user then? did the one I paste allow any user, as long as they can login properly via the session ubus calls? the name in the acl defines a role, but if /etc/config/rpcd is "plain" (just the single root login section) do I need to add roles there too? or will the acl by itself allow any user to work? Nov 22 09:50:06 stintel: thx Nov 22 09:52:25 the "list read *" in the root login means, they can read in any acl? trying to follow: https://openwrt.org/docs/techref/ubus?s[]=acl#full_example Nov 22 10:06:48 karlp: yes, /etc/config/rpcd ties user credentials to acl groups Nov 22 10:07:03 karlp: "list read *" will mean the "read" credential of any group in any file is applied Nov 22 10:07:19 karlp: "list write *" will mean the "write" credentials of any group in any are applied Nov 22 10:07:25 karlp: write implies read Nov 22 10:17:42 ok, so adding an acl with read * write * will basically give the root user the same access via uhttpd-mod-ubus as they have onthe console then. Nov 22 10:20:21 rmilecki: i looked at my ath79 devices, and some do and some don't, e.g. Buffalo WZR-HP-AG300H/WZR-600DHP and Ubiquiti AirRouter do; TP-Link WDR3600 and Netgear WNDR3800 don't have a uimage-fw in their dmesg. Nov 22 10:20:55 russell--: do you have a boot log of 3600 or 3800? Nov 22 10:21:13 rmilecki: hang on ... Nov 22 10:23:47 rmilecki: wdr3600's dmesg -> http://sprunge.us/Sgpay9 Nov 22 10:23:49 russell--: i'm quite curious about it, as in DTS there is partition@20000 { label = "firmware"; reg = <0x020000 0x7d0000>; }; Nov 22 10:23:52 oh, thanks, checking Nov 22 10:24:26 russell--: ah, there is "2 tplink-fw partitions found on MTD device firmware" Nov 22 10:24:31 ah Nov 22 10:24:41 here's the wndr3800: http://sprunge.us/FCGJL2 Nov 22 10:24:55 right, tp-link has its own format instead of using uimage Nov 22 10:25:12 russell--: thanks Nov 22 10:25:59 [ 0.695777] 2 netgear-fw partitions found on MTD device firmware Nov 22 10:26:10 rmilecki: and if it helps - tplink archer c7 v2 "2 tplink-fw partitions found on MTD device firmware" Nov 22 10:26:31 thanks, i'm working on some patches now Nov 22 10:57:53 does anyone know what the phy0radio trigger do? Nov 22 10:58:20 cat /sys/class/leds Nov 22 10:59:35 feckert: I think it blinks on activity for the entire radio (phy) and not just a specific netdev (wlan0) Nov 22 12:27:24 karlp: correct Nov 22 12:27:42 karlp: btw, didn't you report unkillable processes during sysupgrade? Nov 22 12:30:13 yeah, netdata at least seems to be problematic Nov 22 12:30:24 it will always stop if I just do /etc/init.d/netdata stop Nov 22 12:30:31 karlp: https://github.com/openwrt/openwrt/pull/1557 Nov 22 12:30:39 this might be the culprit Nov 22 12:30:39 but if I do sysupgrade it will ~50% of the time fail to stop and abort Nov 22 12:31:18 in any case a nice finding Nov 22 12:31:29 definitely. Nov 22 12:31:50 on rpcd again, is there anyway of passing the timeout from ubus into rpcd? Nov 22 12:32:20 the exec timeout of 30s is at least workable, but "file" is fixed at 3sec, and that's too short for some things I want to use file-exec on, and it's hardcoded. Nov 22 12:32:46 https://git.openwrt.org/?p=project/rpcd.git;a=blob;f=file.c;h=a09e9de7a44e846c0e3d0bcabf89ba56ce7a1f92;hb=HEAD#l42 Nov 22 12:37:17 karlp: I have no issue raising that to 30s Nov 22 12:37:45 karlp: alternatively you can write an rpcd plugin script and simply expose your own procedure for exec Nov 22 12:39:58 yeah, or just the luci controller I already had, Nov 22 12:40:05 was tryign to do _less_ things, not more :) Nov 22 12:40:37 my own rpcd plugin would still be limited to 30sec right? Nov 22 12:40:46 there's simply no way to get the timeotu from ubus across to the rpcd side? Nov 22 12:41:38 the ubus timeout is a client setting Nov 22 12:41:44 while rpcd is a server Nov 22 12:47:47 karlp: actually it seems there will be that 30s limit, yes. But not entirely sure where it comes from Nov 22 12:48:14 nbd: is there a way for an ubus server process to request a larger timeout? Is the 30s limit hardcoded in ubusd or libubus somehwere? Nov 22 12:48:31 it' Nov 22 12:48:44 s from RPC_EXEC_MAX_RUNTIME in rpcd/exec.h I think Nov 22 12:49:06 https://git.openwrt.org/?p=project/rpcd.git;a=blob;f=exec.c;h=f7bfcb2f5e16f550b03e2521c95c3931277e6b91;hb=HEAD#l353 Nov 22 12:49:51 ah right Nov 22 12:50:32 karlp: I'll make it configurable Nov 22 12:50:43 stintel: could you test https://git.openwrt.org/?p=openwrt/staging/rmilecki.git;a=summary on Unifi AP AC Pro? Nov 22 12:50:45 russell--: could you test https://git.openwrt.org/?p=openwrt/staging/rmilecki.git;a=summary on Netgear WNDR3800? Nov 22 12:50:57 there's no particular reason for the 30s limit, it just seemed a good guess when I implemented it Nov 22 12:52:44 karlp: what kind of runtime duration do you need? Nov 22 12:52:54 what do you exec that requires longer than 30s? Nov 22 12:53:21 it /might/ make sense to restructure that logic so that the ubus call only initiates the oepration and the return value is not checked, but depends on your use case obviously Nov 22 12:55:14 I've not got anything longer than 30s at the moment, Nov 22 12:56:21 3s was far too short, ntpdate -q localhost takes ~6 seconds, but just didn't want to run into another arbitrary limit later. Nov 22 12:56:39 luci controller running os.exec() can run as long as the http timeout if it wants Nov 22 12:57:20 I'll bump the max runtime to 120s Nov 22 12:57:29 and add a cli flag to set it Nov 22 12:59:10 yeah, even leaving it at 30 if you can change it via /etc/init.d/rpcd command line options is plenty Nov 22 13:43:59 karlp: pushed Nov 22 13:44:25 uci add rpcd rpcd; uci set rpcd.@rpcd[0].timeout=30 Nov 22 13:44:28 default is 120 Nov 22 13:53:32 niceee Nov 22 13:53:34 thanks. Nov 22 13:53:47 I think the sysupgrade failure I was having was related to this aspect of netdata: https://zerobin.net/?5501959303540825#M60rYmrC/4PzlfKm0tCYx40Z4v+ha4JCOyPUW7z/ykw= Nov 22 13:53:59 maybe, will try that other TERM/KILL patch too. Nov 22 13:54:58 at least, I can run /etc/init.d/netdata restart, and it still receives TERM signals. Nov 22 14:27:35 I am still working on Ubiquiti ER-X-SFP to get SFP port to work. I can get it to work via switch P5 but not 2nd GMAC. But to make more use of SFP kernel features, it is best to support phylink. Nov 22 14:27:59 I wonder if you can mix phylib and phylink devices? Nov 22 15:11:33 I believe https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=41055551151e12abf8efe710efa4dc025a7b7b6a broke rpcd, LuCI doesn't work anymore ;), cya Nov 22 18:57:43 rmilecki: right? thats exactly where i'm coming from. idea is to first boot initramfs, see what else is needed, then flash a known-working kernel/rootfs to nvram Nov 22 19:01:18 dedeckeh: nice 20 ping Nov 22 19:57:12 what should we look for if an usb lte modem doesnt show up in lsusb? Nov 22 19:58:55 usb driver support for *hci Nov 22 20:00:10 first are there usb drivers, any indication a usb device was plugged in? Nov 22 20:01:00 some lte modems show up as cdrom Nov 22 20:02:23 so, re: usb modeswitch Nov 22 20:03:16 usb modeswitch is installed Nov 22 20:03:26 under lsusb we only have the rot hub Nov 22 20:03:28 *root Nov 22 20:03:35 few months ago we had the modem: Nov 22 20:03:43 Bus 001 Device 002: ID 12d1:1573 Huawei Technologies Co., Ltd. Nov 22 20:03:48 but today it wont show up Nov 22 20:04:48 list of installed kmods/pkgs related to usb: https://paste.debian.net/plain/1052856 Nov 22 20:04:54 do we miss anything maybe? Nov 22 20:10:26 ldir:pong Nov 22 20:19:23 Hi all Nov 22 20:25:29 what changed? Nov 22 20:26:17 ive seen devices go in to a wierd debug mode, different vendor/device id's too Nov 22 20:29:58 keep troubleshooting it, make sure it works in a laptop too, etc.. Nov 22 20:38:09 I have a bit of a problem :) Nov 22 20:38:43 I'm working on a move from ar71xx to ath79 on a router Nov 22 20:38:55 It boots, I can connect to it Nov 22 20:39:20 But for some reason the scripts that make up your initial network config don't seem to run Nov 22 20:39:31 Like /etc/board.d/02_network Nov 22 20:40:13 I mean the file exists, and I can see that it was actually used and it had the right board (I put an echo to a file in there) Nov 22 20:40:27 But for some reason I always end up with a default config Nov 22 20:41:01 The one from files/etc/config/network in git Nov 22 21:07:57 wat? it shows up in "lsusb" with 17.01.6, only installing kmod-usb-dwc2 Nov 22 21:08:13 but not with 18.06 Nov 22 21:16:27 oh, thats new information Nov 22 21:17:09 its probably usb/lte drivers? Nov 22 21:20:56 #openwrt prob more help avail Nov 22 21:52:34 huh? thought this is the main channel? Nov 22 21:53:34 we got it completely running with 17.01.6, but it doesnt even show up in "lsusb" using 18.06 or newer Nov 22 21:55:10 the vendor firmware is very old though, openwrt 12 or something Nov 22 21:56:50 maybe some change in kmod-dwc2 from 17.01 to 18.06 or kernel 4.4 to 4.14 Nov 22 21:57:03 *kmod-usb-dwc2 Nov 22 22:18:05 For two days I'm wondering why owrt doesn't create a network config after flashing. Nov 22 22:18:38 And now I find that I had files/etc/config/network in my buildroot. Nov 22 22:18:42 Argh Nov 23 00:16:42 https://www.hardkernel.com/shop/odroid-h2/ Nov 23 00:16:44 it's here! Nov 23 00:17:42 Realtek RTL8111G with 2x nics Nov 23 00:18:15 yes.... Nov 23 00:27:32 jwh: and it's sold out already Nov 23 00:27:37 afaik Nov 23 00:27:53 yeah Nov 23 00:27:59 not unexpected Nov 23 00:28:13 kinda sucks though, I missed the announcement Nov 23 00:28:28 but maybe it's a good thing, first run as the potential to be a bit shady Nov 23 00:28:31 has* Nov 23 00:28:52 not ubiquiti shady, but you never know Nov 23 00:29:19 will take at least 2 months for the 2nd batch, perhaps longer, as intel cant supply CPUs earlier from what i read Nov 23 00:29:27 other thing... Nov 23 00:29:37 it's a pretty sucky time for electronics Nov 23 00:30:23 how can i create a build as close to the official builds as possible? have a build here, where some usb lte modem doesnt work when i build an image myself but it works with official image… Nov 23 00:30:36 i compared "opkg list-installed" already Nov 23 00:30:47 kernel version is also the same Nov 23 00:45:04 got it. always the things you dont think matter. Nov 23 00:45:28 oh? Nov 23 00:46:04 only issues remaining: doesnt work with newer than 17.01 (18.06 or master) and normal reboot doesnt sufficiently initialize usb Nov 23 00:46:45 but 17.01 with kernel 4.4 is waaaay better already than vendors owrt 12 with kernel 3.6.6 Nov 23 01:52:33 what's exciting about the h2? Nov 23 01:53:08 cheap, x86 Nov 23 01:53:15 also sata and m.2 Nov 23 01:53:46 what's the use case? htpc? Nov 23 01:54:03 I guess you could, or full pc Nov 23 01:54:07 but it has dual gig nic too Nov 23 01:54:34 but its a tiny x86 board, and it has sodimm slots so at least its not something dumb like 1G ram Nov 23 01:54:35 * mangix has never understood dual nics on computers Nov 23 01:54:38 will take up to 2g Nov 23 01:54:41 32g* Nov 23 01:54:56 yeah with ram prices as they are Nov 23 01:55:00 heh quite Nov 23 01:55:03 the ram costs more than the board Nov 23 01:55:17 hardkernel are selling 8G for 64 bucks Nov 23 01:55:20 not sooooo bad Nov 23 01:55:31 lol i was thinking the full 32gb Nov 23 01:55:35 i mean not like it used to be, but not crazy Nov 23 01:56:06 what were you planning on using the 32g for? Nov 23 01:56:08 meh i guess it's nice. I have no use for it though Nov 23 01:56:15 for most applications you'd probably run out of cpu time first Nov 23 01:56:17 or i/o Nov 23 01:56:44 true Nov 23 01:57:02 I've been waiting for half decent x86 embedded boards though Nov 23 01:57:03 it's funny how people do that on laptops and desktops though Nov 23 01:57:16 ehhhh actually nvm Nov 23 01:57:23 some games take a massive amount of RAM Nov 23 01:57:40 arm is all well and good but rather limited by software and OS availability Nov 23 01:58:00 is it? Nov 23 01:58:14 for hipster stuff yeah Nov 23 01:58:19 unless you build it yourself Nov 23 01:58:39 won't see debian near any of my stuff :D Nov 23 01:58:47 or ubuntu, which is even worse Nov 23 01:59:16 speaking of which, dealing with a hipster project that won't build working binaries without docker Nov 23 01:59:20 wha is wrong with the world Nov 23 01:59:39 * mangix has no idea Nov 23 02:00:06 autotools-in-cmake-in-custom hand rolled make files Nov 23 02:00:08 its madness Nov 23 02:00:11 oh, and go Nov 23 02:00:40 don't worry, it will rust down ruby red as well Nov 23 02:00:45 lol Nov 23 02:01:01 the only method that produced working musl binaries has been removed Nov 23 02:01:01 :D Nov 23 02:01:19 but they still build binaries for download unofficially that work, but don't reproduce the steps required Nov 23 02:02:48 but then it enforces such flags as -funroll-loops Nov 23 02:02:51 maybe they're gentoo ricers Nov 23 02:03:31 lol Nov 23 02:03:51 and -O3 for good measure Nov 23 02:27:48 anyone good with CFE bootloaders **** ENDING LOGGING AT Fri Nov 23 03:00:01 2018