2016-07-02 00:01:49	@kido	Lrs121: this USB gadget Kconfig turns me mad ! it works if you provide a full defconfig like this one: http://hastebin.com/xalofijubu and also this patch: http://hastebin.com/hiwohuyofa.vhdl
2016-07-02 00:02:31	@kido	but gosh why does it remove some config options when using a small defconfig, and we meet the same problem with every 3.10 kernel we use
2016-07-02 00:02:48	@kido	that's why I spent two hours on this, I would have liked to find a clean definitive answer
2016-07-02 00:04:23	Lrs121	Thats strange. if it doesnt work with a stripped defconfig, i have to wonder why google ships them like that.
2016-07-02 00:04:42	@kido	no it's not google's fault
2016-07-02 00:05:43	@kido	anyway, you can use that for now until I find a proper solution
2016-07-02 00:05:53	@kido	if I ever find one
2016-07-02 00:06:40	Lrs121	alright i get the patch and new defconfig pushed into git
2016-07-02 00:07:25	@kido	it seems to come from here http://www.crashcourse.ca/wiki/index.php/OE_Kernel_Build_Tasks#do_kernel_configme
2016-07-02 00:10:43	@kido	if you check your log.do_configure you can see that OE blindly accepts a shitload of config values
2016-07-02 00:11:30	@kido	and some of them seem to disable CONFIG_USB_G_ANDROID as a side effect
2016-07-02 00:17:29	@kido	that's why I said to locusf that he should enable this option https://github.com/locusf/meta-athias-hybris/blob/master/recipes-kernel/linux/linux-anthias.bb#L19 but in the end this is of no use
2016-07-02 00:18:08	@kido	we've no choice but using full defconfigs on recalcitrant platforms
2016-07-02 00:21:47	Lrs121	that sounds like a pain to deal with
2016-07-02 00:23:50	@kido	it is but we always have the same problem
2016-07-02 00:24:29	@kido	and it is still stopping locusf from using his watch so it kinda pisses me off
2016-07-02 00:26:30	Lrs121	the kernel build is preventing it or is packing the boot image the problem
2016-07-02 00:28:38	@kido	no this is the android's composite usb gadget driver
2016-07-02 00:32:34	Lrs121	ah ok
2016-07-02 01:56:51	@kido	Lrs121: so how far does the build go ? ;)
2016-07-02 02:08:16	Lrs121	still running, im starting with a clean build
2016-07-02 02:08:43	Lrs121	i think i have yet to reach the kernel
2016-07-02 02:47:23	Lrs121	Its made it much further. made it to compile kernel modules and still going where before itd hang at do compile
2016-07-02 03:54:27	Lrs121	We have a successful build for the zenwatch 2 larger aka sparrow. havent tested the boot yet but its next
2016-07-02 09:41:53	@kido	Lrs121: great ! so we need to test this now :)
2016-07-02 11:43:57	locusf	wohoo
2016-07-02 12:01:10	@kido	:)
2016-07-02 12:01:46	@kido	how are you locusf ? was the sfos event nice ?
2016-07-02 12:10:13	locusf	yeah it was great
2016-07-02 12:10:20	locusf	also my first vacation day
2016-07-02 12:10:24	locusf	15h of sleep
2016-07-02 12:11:31	locusf	I've been pretty spent the last month or so
2016-07-02 12:12:53	@kido	ohoh nice ! I did the same thing but for my first day of work  ahah... I guess that we just take strange sleep cycles in holidays :)
2016-07-02 12:13:50	@kido	are you doing anything special for those holidays ?
2016-07-02 12:44:16	locusf	not really
2016-07-02 12:44:23	locusf	just resting
2016-07-02 12:44:30	locusf	thats all I need so far
2016-07-02 12:45:25	@kido	pretty important :)
2016-07-02 13:10:01	@kido	by the way, I updated Qt to 5.6, if anyone wants to rebuild asteroid you will need to git pull your meta-qt5 and meta-smartphone layers
2016-07-02 13:11:58	locusf	ok
2016-07-02 19:18:22	Lrs121	Morning all. Well its morning for me
2016-07-02 19:18:50	@kido	ah? :)
2016-07-02 19:18:59	@kido	good morning then
2016-07-02 19:31:36	Lrs121	I have tested the build. there was a some success, however it stops at the splash screen
2016-07-02 19:32:45	@kido	good :)
2016-07-02 19:32:49	@kido	do you have adb?
2016-07-02 19:35:13	Lrs121	unfortunately no, i believe im in a similar situation as locusf with adb
2016-07-02 19:35:23	@kido	:(
2016-07-02 19:35:44	@kido	so yes, you are at the same point now
2016-07-02 19:38:13	@kido	we have no choice but debugging what's going on between adbd and the kernel's gadget features
2016-07-02 19:38:30	@kido	but debugging without shell is painful
2016-07-02 19:40:18	Lrs121	well im open to learning new things
2016-07-02 19:44:20	@kido	sure but I don't know where things can go wrong
2016-07-02 19:44:51	Lrs121	would getting non volatile logs be a good start?
2016-07-02 19:46:05	@kido	that can help yes, but we will need to output more things to those logs by adding commands in the ramdisk's init script
2016-07-02 19:46:47	Lrs121	ah fun
2016-07-02 19:48:54	@kido	mhhh I just noticed that USB_G_ANDROID is also disabled on the g watch's .config ...
2016-07-02 19:51:39	Lrs121	and adb is functional with the g watch?
2016-07-02 19:54:39	@kido	yes
2016-07-02 19:59:44	@kido	mhh no, I'm wrong OE just wiped dory's source by sparrow and I was looking at the sparrow's defconfig
2016-07-02 20:00:12	Lrs121	ah ok
2016-07-02 20:27:31	@kido	http://hastebin.com/remiqeluzo <- diff config-dory config-sparrow
2016-07-02 20:33:13	@kido	the defconfig seems ok, something must be wrong elsewhere
2016-07-02 20:33:33	@kido	Lrs121: do you have the android's boot.img somewhere ?
2016-07-02 20:36:41	Lrs121	http://lrsservers.ddns.net/download/devices/asus/sparrow/boot.emmc.win
2016-07-02 20:36:59	Lrs121	its a twrp backup but it can be unpacked normally
2016-07-02 20:37:35	@kido	alright, thank you
2016-07-02 20:38:08	Lrs121	youre welcome
2016-07-02 20:44:55	@kido	init.usb.configfs.rc this is interesting
2016-07-02 20:45:03	@kido	locusf didn't have that
2016-07-02 20:46:39	@kido	but configfs seems to be disabled at line 13 of init.usb.rc
2016-07-02 20:46:51	@kido	so functionfs should be used instead
2016-07-02 20:48:50	@kido	selinux seems to be disabled so it shouldn't be a problem neither
2016-07-02 21:07:24	Lrs121	the fact that selinux is disabled is the only thing i understand of that
2016-07-02 21:08:11	@kido	well, when you boot up your device the kernel is loaded first and it setup all the devices according to its devicetree and available modules
2016-07-02 21:08:36	@kido	it *seems* that your kernel config is correct so the kernel should setup the USB's OTG mode correctly
2016-07-02 21:09:05	Lrs121	oh ok
2016-07-02 21:09:06	@kido	as soon as the kernel is up and running it loads the ramdisk which contains an init script setting up different kind of things from user space
2016-07-02 21:09:52	@kido	so in your android boot.img you have a ramdisk with a /init binary which does all kind of actions according to the /init*.rc files
2016-07-02 21:10:28	@kido	and for example this init loads configure usb and starts adbd, the adb daemon
2016-07-02 21:11:38	@kido	so if you extract your boot.emmc.win and then if you extract the ramdisk.cpio.gz, there are a couple of interesting files in there: init.rc, init.usb.rc, init.asus.usb.rc and init.usb.configs.rc
2016-07-02 21:12:11	@kido	the main parts being the lines 16-32 from init.usb.rc
2016-07-02 21:12:58	@kido	in asteroid we don't use android's init system in favor of a simple bash script
2016-07-02 21:13:34	@kido	you can notice that this script tries to mimic android's usb setting up procedure here https://github.com/Lrs121/meta-sparrow-hybris/blob/master/recipes-core/initrdscripts/initramfs-boot-android/init.sh#L28
2016-07-02 21:14:18	@kido	and if we have enable_adb in the kernel's command line we start adbd here https://github.com/Lrs121/meta-sparrow-hybris/blob/master/recipes-core/initrdscripts/initramfs-boot-android/init.sh#L50
2016-07-02 21:15:04	@kido	if we don't have it, it is loaded by usb_moded which is loaded by systemd when mounting the rootfs here https://github.com/AsteroidOS/meta-dory-hybris/blob/master/recipes-core/initrdscripts/initramfs-boot-android/init.sh#L114
2016-07-02 21:15:36	@kido	so there are two adbd daemons in asteroid, one in the ramdisk to debug early boot stages and one in the rootfs to develop comfortably
2016-07-02 21:16:39	Lrs121	so if i drop enable_adb=1 into the cmd_line of the img_info it should attempt to bring up adb during ramdisk?
2016-07-02 21:16:55	@kido	if the usb device is setup correctly (both by the kernel and the init script), adbd should be able to do its job correctly and your host computer should show you a new device in "dmesg -w"
2016-07-02 21:17:14	@kido	and then be usable by the adb command from your computer
2016-07-02 21:17:18	@kido	but it doesn't seem to be the case
2016-07-02 21:17:43	@kido	yep if you add this line it will start adbd in the ramdisk *and stop there* so you won't get splashscreen anymore since psplash is ran in the rootfs
2016-07-02 21:18:01	@kido	but this is actually quite handy, it is helpful for debugging what's going on in the early stages of your boot process
2016-07-02 21:19:44	@kido	it won't try to mount any partition or whatever, only /proc /sys/ /dev/ /dev/pts and /dev/usb-ffs should have been setup and you should get a system "fresh out of the kernel"
2016-07-02 21:21:21	@kido	if you want to add any other setting up command, just add them in your init.sh, re-run bitbake linux-sparrow and it will pack you a new boot.img containing your ramdisk
2016-07-02 21:25:02	@kido	there are a couple of handy things to know here if you want to know what's actually going on in your ramdisk. if you enable CONFIG_TTY_PRINTK in your kernel (not done yet) you will get a nice /dev/ttyprintk device. you can then redirect the output of commands to this file and they will appear in the kernel's dmesg. if you manage to get persistent logs you will get the output of those commands
2016-07-02 21:26:04	@kido	if you use CONFIG_ANDROID_RAM_CONSOLE in both android and asteroid you can also access previous boot's logs from android
2016-07-02 21:27:20	@kido	this is how I worked in the early days of asteroid but that would be cool if porters don't have to mess with this method anymore
2016-07-02 21:33:00	@kido	so basically the procedure to debug usb now is as follow: 1) make sure the kernel has everything that is needed to setup the usb device (it *should* be the case but we might be missing something...) <- once again you might be able to get some info about this if you get persistent logs 2) make sure that our init.sh does the same initialisation procedure as android's init*.rc config files <- for example I
2016-07-02 21:33:00	@kido	noticed that asus plays with the permissions of /sys/devices/platform/msm_hsusb/gadget/wakeup ... the permissions by themselves shouldn't be important but /sys/devices/platform/msm_hsusb/gadget/ might be an interesting place to look at
2016-07-02 21:34:06	@kido	3) get info on what's going on with commands output redirections to /dev/ttyprintk
2016-07-02 21:34:58	@kido	if you have any question, as usual, you know you can always ask it... ;)
2016-07-02 21:35:52	@kido	and if this problem gets solved, two devices should get unlocked in a row so this is super exciting! :) good luck and have fun Lrs121!
2016-07-02 21:54:30	@kido	for example you should check /sys/class/dual_role_usb/otg_default/power_role (can be sink or source) and /sys/class/dual_role_usb/otg_default/data_role (can be host or device) and /sys/class/dual_role_usb/otg_default/mode (can be dfp or ufp) the values that are used in android are probably correct so you can make sure the same are used in asteroid
2016-07-02 22:11:51	Lrs121	is that within android
2016-07-02 22:11:55	@kido	Lrs121: you could also try to boot init.sh with the android's kernel (I'm not sure if selinux will complain but well... you can try) and enable_adb. it might work and lead us into a direction
2016-07-02 22:12:13	@kido	you should have those files both in android and asteroid as they are virtual files exposed by the kernel
2016-07-02 22:14:14	@kido	actually, trying the android's kernel with the asteroid's ramdisk can be a really good way to know if the problem is in kernel space or user space
2016-07-02 22:15:00	@kido	and if you use enable_adb=1 you also also use selinux=0 so it shouldn't be a problem, it should work
2016-07-02 22:15:16	@kido	s/also also/can also/
2016-07-02 22:21:20	Lrs121	i dont see a dual_role_usb folder in android under /sys/class
2016-07-02 23:07:28	@kido	alright then it might not be setup, but as long as it is not used by android it shouldn't have to be used by asteroid ;)
2016-07-02 23:20:21	Lrs121	ok