**** BEGIN LOGGING AT Mon Aug 03 02:59:58 2015 Aug 03 07:46:50 anybody around to help me figure out how to overlay the overlay? that is, i want to put a tmpfs overlay on top of the jffs2 overlay Aug 03 07:50:02 mkdir /tmp/root ; mount -t tmpfs -o root /tmp/root ; mkdir /tmp/root/upper ; mkdir /tmp/root/work ; mount -o noatime,lowerdir=/,upperdir=/tmp/root/upper,workdir=/tmp/root/work -t overlayfs "overlayfs:/tmp/root" /mnt/test /overlay Aug 03 07:53:16 or something like that, mount complains, gives me mount command help Aug 03 07:59:44 maybe i can't stack Aug 03 08:01:42 mount -t tmpfs -o root /tmp/root Aug 03 08:01:49 something is missing Aug 03 08:02:59 and what is -o root? Aug 03 08:08:56 no idea, i found that in /lib/functions/preinit.sh Aug 03 08:24:44 sure that "-o root" is right ? its not in the documentation of tmpfs Aug 03 08:24:54 http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/filesystems/tmpfs.txt Aug 03 08:59:47 plntyk: i left part out, see: http://git.openwrt.org/?p=openwrt.git;a=blob;f=package/base-files/files/lib/functions/preinit.sh;h=db1f298c8bd25c9cb21ac66df1cddea79118ad55;hb=a51528c01015fc3f13ab4d7d5474db0c0f26ce3d#l85 Aug 03 09:00:05 line 85 Aug 03 09:00:27 /bin/mount -t tmpfs -o noatime,mode=0755 root /tmp/root Aug 03 09:01:06 but i typed that part in correctly, it is the final mount that fails Aug 03 09:04:04 this is part of a production process, i want to make ephemeral filesystem changes for the purposes of testing, then have them go poof and disappear on a power cycle. Aug 03 09:04:08 rusell - ah prbly it works because tmpfs may be a "special filesystem" like proc (device indication section in man mount) Aug 03 09:04:51 that sounds plausible Aug 03 09:05:20 * russell-- doesn't quite fully understand all the mechanics of pivots and such Aug 03 09:07:17 worst case, i include a factory reset in the production process after scribbling on the jffs2 Aug 03 09:07:33 but a tmpfs overlay sounded so sweet Aug 03 09:08:55 hm... the overlayfs doc explicitly states "The lower filesystem can even be another Aug 03 09:08:55 overlayfs" Aug 03 09:12:10 https://lkml.org/lkml/2015/6/25/105 did it by accident Aug 03 09:16:13 the patch series that is removing overlayfs on overlayfs is not accepted atm so it should still work Aug 03 10:23:13 blogic r46548 branches/ (10 files in 10 dirs) * ar71xx: Backport support for Onion Omega Aug 03 10:23:15 blogic r46549 branches/chaos_calmer/target/linux/ar71xx/image/Makefile * ar71xx: use automatic firmware partition splitting for devolo dLAN devices Aug 03 17:19:52 trine: waht device are you talking about? Aug 03 18:00:19 Hauke, this one http://www.aliexpress.com/item/300Mbps-high-power-long-range-wireless-n-cpe-repeater-ATHEROS-AR9341-Comfast-CF-E316NV2-0-with/32259599919.html Aug 03 18:40:36 build #68 of ar71xx is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/ar71xx/builds/68 Aug 03 19:10:18 Hauke: the device in question is a chinese clone of ubnt nano, by "COMFAST". Aug 03 19:11:10 Hauke: the symptoms is that it reboot shortly after a self-compiled image is booted, and it's not too consistent when it comes to timing, different images produce different results, but always at exactly the same spot. Aug 03 19:11:57 Hauke: e.g. an initramfs image reboots after "NET: Registered protocol family 1" (here it didn't have a chance to write 7) Aug 03 19:12:51 Hauke: same squashfs build when flashed reboots just after the userspace is passed control to, so if init=/bin/sh, one can see the shell prompt produced (then the board reboots). Aug 03 19:14:12 watchdog issue? Aug 03 19:14:13 Hauke: when I added some delays and prints to board init code, it rebooted too, inbetween that additional (print+delay)*N loop. Aug 03 19:14:38 nbd: tried making internal watchdog driver issue keepalive early in the board code, no change. Aug 03 19:14:50 maybe an external watchdog? Aug 03 19:15:04 When stock firmware is booted, there's no additional gpio watchdog driver loaded, just the regular integrated ath79 one. Aug 03 19:15:54 maybe it's built into the kernel Aug 03 19:16:10 or hacked into the regular watchdog driver Aug 03 19:16:28 Yes, probably. Aug 03 19:16:43 Unfortunately, I do not have direct access to the board to inspect it. Aug 03 19:20:45 I even can't unlzma the vendor kernel, the lzma tool says data is corrupt after processing about 5% of it. Aug 03 19:23:01 Heh, probably worth trying gpio watchdog triggering all the available gpios just in case :) Aug 03 19:24:27 if it works, you can start bisecting ;) Aug 03 19:24:34 Sure Aug 03 19:24:44 Thanks for the idea, I was about to give up ;) Aug 03 19:24:54 The timing is not too consistent though. Aug 03 19:25:29 maybe the boot loader has a timer based refresh and doesn't do one last refresh before booting the kernel Aug 03 19:25:41 that would make the timing inconsistent Aug 03 19:25:41 With initramfs image the last message is "[ 0.570000] NET: Registered protocol family 1", but when I added the print+msleep loop it worked till about 1.1. Aug 03 19:25:54 or does it depend only on the image? Aug 03 19:26:03 Always the same with same image. Aug 03 19:26:12 hm, that doesn't sound like a watchdog issue then Aug 03 19:26:44 maybe it registers some driver that crashes and it depends on the initialization order Aug 03 19:26:45 My next thought was that some GPIO controls power supply. Aug 03 19:27:01 what mach file are you using as reference? Aug 03 19:27:34 carambola2 but I had everything disabled there. Then tried tplink tl-wr841n-v8. Aug 03 19:28:01 (or whatever wr841 version it was that's ar9341 based) Aug 03 19:29:23 same result? Aug 03 19:29:59 afaict, yes. Aug 03 19:30:12 did you try to remove all gpio stuff from there? Aug 03 19:30:39 I even tried to remove ath79 gpio registration from setup.c so it won't touch it at all. Aug 03 19:30:45 And that made no difference. Aug 03 19:33:30 (registers some driver that crashes) with carambola2 squashfs image and init=/bin/sh it surely went as far as starting the userspace. Aug 03 19:33:53 build #68 of mpc83xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/mpc83xx/builds/68 Aug 03 19:58:29 build #68 of ep93xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/ep93xx/builds/68 Aug 03 20:18:54 fwiw, i have a recent image on an ubnt airrouter that is rebooting frequently, unfortunately i don't have a console on it at the moment to see what the panics are about Aug 03 20:34:36 hauke r46550 trunk/package/network/services/dnsmasq/Makefile * dnsmasq: Bump to dnsmasq2.75 Aug 03 20:35:05 hauke r46551 trunk/package/libs/cyassl/Makefile * cyassl: the upstream package in version 4.6.0 changed Aug 03 20:50:15 hauke r46552 trunk/package/kernel/acx-mac80211/Makefile * acx-mac80211: update git url Aug 03 21:16:11 PaulFertser: it has intelligent restart Aug 03 21:19:15 PaulFertser: they are marketing their watchdog with the problems you described I assume something unusable ;-) Aug 03 21:19:23 PaulFertser: they are marketing their watchdog with the problems you described I assume something unusable ;-) Aug 03 21:20:03 so one image always reboots at the same position, but different images are rebooting at different positions? Aug 03 21:20:22 have you tried an image which just has serial? Aug 03 21:20:26 so no ethernet Aug 03 21:29:42 build #69 of au1000 is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/au1000/builds/69 **** ENDING LOGGING AT Tue Aug 04 02:59:58 2015