**** BEGIN LOGGING AT Thu Oct 25 03:00:00 2018 **** BEGIN LOGGING AT Thu Oct 25 07:27:14 2018 Oct 25 07:53:10 nbd, the mt7688 linkit station mode seems *much* better in my build of today's master branch Oct 25 08:00:37 $ git describe Oct 25 08:00:38 reboot-8378-g9ac7350240 Oct 25 08:02:27 at least, my initial test of connecting to one ap is stable where it was not previously Oct 25 08:04:20 gah, reports were premature. trying another ap, not getting DHCP successfully Oct 25 11:21:21 I'm trying to follow https://wiki.openwrt.org/inbox/mikrotik/mikrotik_rb750gr3 to install openwrt on my RB750GR3. I have soldered on serial console that works. I can boot that an LEDE image someone else compiled, but it's lacking support for my newer rev flash chip. I tried to forward-port the patch to 18.06.1 and use that vmlinux-initramfs.elf but after tftpboot then I just get "[ 0.000000] Oct 25 11:21:27 Kernel panic - not syncing: Failed to find mtk,mt7621-sysc node" Oct 25 11:21:46 I then tried to compile an old LEDE image, but that seems to come with a 4.4 kernel, whereas the patch is against 4.9 Oct 25 11:22:34 and 4.4 lacks one of the properties of the support-for-newer-flash-chip patch I found in https://forum.openwrt.org/t/how-to-install-to-rb750gr3/4654 Oct 25 11:23:00 so basically, what am I doing wrong when I am checking out lede and getting a 4.4 kernel instead of 4.9 kernel? Oct 25 11:24:22 my plan was to update https://openwrt.org/toh/hwdata/mikrotik/mikrotik_rb750gr3 with text on how to do this now Oct 25 11:25:58 SwedeMike: did you also tried the last snapshot via tftpboot? http://downloads.openwrt.org/snapshots/targets/ramips/mt7621/ Oct 25 11:26:24 Rene__: I did not. The instructions said to use that elf binary. Oct 25 11:26:47 current mt7621 platform using 4.14 kernel Oct 25 11:28:25 Rene__: the mikrotik bootloader won't boot a standard rb750gr3-initramfs-kernel.bin Oct 25 11:28:38 ok Oct 25 11:28:48 Rene__: so I need to boot that vmlinux-initramfs.elf so I can re-flash the boot loader Oct 25 11:29:04 (I just tried it to be sure) Oct 25 11:31:03 so just to re-iterate. I have an LEDE kernel that boots, but doesn't have the 4 line patch to enable support for my flash chip so there is no /dev/mtd. I also am able to get that vmlinux-initramfs.elf from 18.06.1 and get the boot loader to boot the kernel, but it then panics because it can't find the correct filesystem. Oct 25 11:32:22 https://pastebin.com/KX3ActgZ is the kernel panic Oct 25 11:33:13 seems more like a DTS file issue Oct 25 11:34:27 Rene__: https://wiki.openwrt.org/inbox/mikrotik/mikrotik_rb750gr3 has a patch one is supposed to put in target/linux/ramips/patches-4.9/ to make it work, but as I said before, when I check out LEDE I get a 4.4 kernel that doesn't support one of the flash properties Oct 25 11:36:24 I feel that if I can just follow that instruction I'd be able to create a LEDE image with support for my flash mentioned in https://forum.openwrt.org/t/how-to-install-to-rb750gr3/4654/34 (4.4 doesn't seem to have "SPI_NOR_HAS_TB") Oct 25 11:37:24 https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ramips/dts/mt7621.dtsi;h=aa85aee6cb38935d9478fc395934b216a0781a5f;hb=HEAD#l63 Oct 25 11:37:41 The dt node Oct 25 11:39:40 Rene__: ok, what to do about it? Oct 25 11:40:33 Good question. Unfortunately I can't help you with that. Oct 25 11:41:26 when it does boot with the LEDE image I downloaded, I get: m25p80 spi32766.0: unrecognized JEDEC id bytes: ef, 70, 18 m25p80: probe of spi32766.0 failed with error -2 Oct 25 11:43:10 which does seem to jive with the ""w25q128j", INFO(0xef7018, 0, 64 * 1024, 256, SECT_4K | SPI_NOR_DUAL_READ | SPI_NOR_QUAD_READ | SPI_NOR_HAS_LOCK | SPI_NOR_HAS_TB) patch I feel I need to apply to LEDE to get this to work Oct 25 11:43:42 I think so too Oct 25 11:43:51 but all of this already exists in 18.06, all the patches seem to have been applied... so I don't understand why it won't boot 18.06.1 that I compiled. Oct 25 11:49:33 ./arch/mips/ralink/mt7621.c: rt_sysc_membase = plat_of_remap_node("mtk,mt7621-sysc"); Oct 25 11:49:53 this is the only match that I get with grepping the kernel source Oct 25 11:51:22 This is the problem I think [ 0.000000] OF: fdt: No valid device tree found, continuing without Oct 25 11:51:43 Device tree file is not added / found in your bootfile Oct 25 11:54:38 Did you select "Target Profile (MikroTik RouterBOARD M33G)" ? Oct 25 11:54:59 in "make menuconfig"? Oct 25 11:55:27 Rene__: no, I selected rb750g3 Oct 25 11:56:03 Ok, sorry I picked the wrong one. Oct 25 11:57:47 Now I am out of idea's. Oct 25 11:59:20 any idea how I can override search domain sent by dnsmasq? Oct 25 11:59:29 list dhcp_option '119,domain.tld' doesn't seem to work Oct 25 12:02:45 Rene__: so these .dts files are device trees description files? Hm, wonder if that's what's missing then to make it boot. Because that's one part of the patch that I linked to, it inserts a device tree into the kernel itself. Oct 25 12:03:20 Rene__: this is good, I'll fiddle around a bit more with that and see if I can make it work in kernel_menuconfig the same way the patch does it Oct 25 12:03:38 even under openwrt 18 Oct 25 12:03:51 thanks Oct 25 12:06:21 With U-boot I know how you can add a dts (dtb compiled version) a kernel. But I think it is different with routerboard bootloader? Oct 25 12:07:40 Rene__: right, that's the main problem, the routerboard bootloader doesn't have serial and requires this elf-version to boot. If I can just get this booted with flash support, then I can put regular uboot on it Oct 25 12:09:24 uboot you simply "cat zImage my.dtb > tmp " and using mkimage to adding the right headers to the tmp file. Oct 25 12:10:29 BTW only needed for older uboot version. Oct 25 12:53:32 Rene__: aha! now I got it booting. it actually was the last part of the patch (that I thought was already included), that was needed. the one patching /arch/mips/ralink/of.c Oct 25 12:55:23 https://pastebin.com/t4uz7XcT what does that do? Oct 25 13:00:15 SwedeMike: without being an expert there and without knowing the context I would assume that it parses the dtb appended to the kernel image (&__image_dtb) if no dtb pointers got passed by the bootloader Oct 25 13:01:08 jow_: right, this is to get it to boot on the mikrotik routerboard stock bootloader which doesn't pass this information. Oct 25 13:01:35 SwedeMike: openwrt patches the kernel to include a custom "__image_dtb" symbol Oct 25 13:02:08 which is populated by the "patch-dtb" util during build to append an external dtb file to a bianry kernel image Oct 25 13:02:55 so yes, this patch hooks the openwrt appended-dtb machinery into the platform setup procedure as last resort Oct 25 13:03:36 ldir: ping Oct 25 13:04:17 jow_: right, so https://drive.google.com/file/d/1enQvVKUrtMEYUUYX2x42PgK7pqn5s3-d/view is what I now have "forward-patched" so it now works on 18.06 head (I think it was). What should I do with these patches to help the next guy that comes along? Oct 25 13:05:03 SwedeMike: ideally submit it as two patches Oct 25 13:05:15 this is all about https://wiki.openwrt.org/inbox/mikrotik/mikrotik_rb750gr3 and https://forum.openwrt.org/t/how-to-install-to-rb750gr3/4654/34 Oct 25 13:05:41 jow_: well, I don't know if this actually belongs in "normal openwrt". It's basically to create an image that is bootable by regular Mikrotik bootloader Oct 25 13:05:55 it requires you to go into make kernel_menuconfig Oct 25 13:06:22 jow_: pong Oct 25 13:07:14 ldir: are you still closely following dnsmasq upstream and if yes, do you remember any recent DHCP specific changes? I've got a few reports about static leases getting ignored and stuff and I wonder if it is just coincidential or due to a recent change Oct 25 13:09:36 SwedeMike: first of all I'd break out the arch/mips/ralink/of.c change, this should be a standalone patch file going into either target/linux/generic/hack-4.X/ or target/linux/ath79/patches-4.X/ Oct 25 13:11:10 sorry, target/linux/ramips/patches-4.X/ not ath79 Oct 25 13:11:48 jow_: I keep an eye on it still yes. I don't remember anything specific about static leases and nothing leaps out at me from the commit log. Oct 25 13:12:00 SwedeMike: the dts files should be placed in target/linux/ramips/files-4.14/arch/mips/boot/dts/ralink/ Oct 25 13:12:23 Is this recent...ie. since I bumped 1806 to v2.80 a day/two ago? Oct 25 13:13:23 SwedeMike: and the Makefile / Kconfig changes should be added as target/linux/ramips/patches-4.14/XXXX-add-rb750gr3-support.patch Oct 25 13:13:48 jow_: the obvious change is related to not handing out certain hostnames to leases. Oct 25 13:14:00 ldir: no, I don't think it is related to that particular change/update Oct 25 13:14:51 I have a number of devices here tied down with static mac/name & addresses and they're all fine. Oct 25 13:18:28 jow: time for a copy of /var/etc/dnsmasq.conf.cfg. Oct 25 13:18:58 jow: no other dhcp server on the network? Oct 25 13:27:12 SwedeMike: Great! Oct 25 15:23:56 I added a script in /etc/init.d/, and I soft-linked it from /etc/rc.d/S99foo Oct 25 15:24:27 and a manual run of /etc/rc.d/S99foo start works Oct 25 15:24:59 but, it does not start on boot. Is there a place I can look for errors? Oct 25 15:39:59 ok, got it working. Oct 25 17:20:16 is there a way to have some built-in tool in openwrt change config for a single interface through the command line? Oct 25 17:40:09 greearb_: i think you need to be more specific Oct 25 17:40:24 greearb_: from what I understood, all you need is "uci set network.lan.foo=bar" Oct 25 17:40:39 it's a command line call that changes interface config Oct 25 17:44:10 rmilecki, yes, I think so Oct 25 17:46:05 so if I have an entry: config interface lan2 then, I could do uci set network.lan2.proto=dhcp? Oct 25 17:46:24 will poke at it Oct 25 17:46:35 yes Oct 25 19:16:29 (then uci commit; reload_config).... Oct 25 19:17:56 is it noticeably more efficient to pack files into an openwrt image than to install them later on the overlayfs? Oct 25 19:18:26 for binaries, yes, for text and configs...~mixed in my experience Oct 25 19:18:34 depends more on what's easier for you to manage build system wise. Oct 25 19:19:35 I've personalyl foudn it easier to add files in packages, and edit configs in uci-defaults to avoid having to have a forked openwrt tree as much as possible Oct 25 19:20:02 yeah, I will plan to edit config files, but I have a big set of binaries too...I'll pack them into the image Oct 25 19:20:11 once I get them debugged a bit more Oct 25 19:21:22 just put them in a package in an extra feed. Oct 25 19:23:52 seems using the 'files' overlay is easier? Oct 25 19:24:32 files you have to either add to you fork, or copy in after forking. Oct 25 19:24:47 packages give you nice version numbers for the files too. Oct 25 19:25:02 and lets you group things in chunks if that helps Oct 25 19:25:19 I will very likely have to fork anyway, will want some packets that upstream (kernel, at least) will not accept Oct 25 19:25:44 anyone her going to the summit ? pm me ... Oct 25 19:25:56 trying to organize a dinne on sun Oct 25 20:29:17 I notice the mac-addr of wlanX (at least) keeps changing on reboot. Is that on purpose? Oct 25 20:32:34 greearb_: that shouldn't happen, see https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ipq806x/base-files/etc/hotplug.d/ieee80211/10_fix_wifi_mac for details Oct 25 20:33:17 phy address changed on one of the radios as well Oct 25 20:33:54 that would explain it Oct 25 20:34:39 it only changed by one octed (the last b was an a last time): dc:ef:09:e5:6a:b Oct 25 20:34:50 er, dc:ef:09:e5:6a:bb Oct 25 20:39:50 hmmm, maybe smething else, I have phy0 and phy2 on this boot, instead of phy0, phy1 Oct 25 20:39:56 that would throw off that script Oct 25 20:40:20 maybe something I self inflicted as I am trying to rename phy0 to wiphy0, etc. Oct 25 20:41:11 pkgadd, is there a good place in that hotplug logic to do the rename? Oct 25 20:42:38 greearb_: sorry, not really my area of expertise - I only touched that file once, to fix the wlan mac assignment for my nbg6817 Oct 25 20:43:33 ok Oct 25 20:47:09 that script definitely uses wrong MAC if you reload the driver causing phy3, phy4 to show up :( Oct 25 21:08:46 fixed that and posted the patch, but wlanX still has wrong mac... Oct 25 21:09:12 greearb_: yeah, it's jsut how many patches you want to rebase each time. I've personalyl chosen to have only patches for kernel stuff in the fork, and have all "files" stuff be provided in package feeds, Oct 25 21:09:27 nod, will keep that in mind Oct 25 21:09:59 there's no wrong way, just lots of ways that all hae downsides :) Oct 25 21:10:43 having uci-defaults files in packages lets me detect upgrades and try and migrate things too. **** ENDING LOGGING AT Fri Oct 26 02:59:59 2018