8:07 AM <jstultz> pundir: so yea.. how'd the last week go for you?
8:08 AM <pundir> it was good.. boot tested preview-4.1 on ifc6410
8:09 AM <jstultz> pundir: do you have a basic test process for that?
8:09 AM <pundir> run into few new bugs in android-3.18, switch class driver and extcon conflicts to speak of
8:10 AM ⇐ jamzhou quit (~jamzhou@104.237.91.35) Ping timeout: 250 seconds
8:10 AM <pundir> so plan was to run core cts tests but usb gadget is broken on qcomlt-4.1 for ifc6410 and ethernet port is broken on my machine
8:10 AM <pundir> so couldn't get adb working
8:10 AM <jstultz> pundir: heh.. ok.. so boot and i'm guessing twidddle a bit around in the ui?
8:10 AM <pundir> so it was more of a smoke test
8:10 AM <pundir> yes ran monkey for a bit
8:10 AM <jstultz> ok..
8:10 AM <jstultz> just curious.. sorry for the distraction
8:10 AM <pundir> but with s/w graphics it had its own issues
8:11 AM <pundir> moved llct to 4.2-rc8
8:12 AM <pundir> and looked gbsim and ehci debug port for configfs session planning
8:12 AM <jstultz> pundir: yea, its probably an interesting example
8:12 AM <jstultz> pundir: that it?
8:13 AM <pundir> gbsim yes.. i did not try it but it seem to be a very good use case of configfs gadgets
8:13 AM <pundir> yes thats all i had
8:14 AM <jstultz> pundir: ok.. so the configfs gadget session and lpc wasn't quite what i was hoping for initially..  andrejz and badhri got into a bunch of details on configfs implementation questions
8:14 AM <jstultz> basically when in the process of configuring the various configfs filesystem state should allocations and binds occur.
8:14 AM <jstultz> and could it be made more atomic, and what tradeoffs that would have
8:15 AM <jstultz> it was also only a 15 minute session, so it wasn't until minute 20 or so when badhri got started asking questions
8:15 AM <jstultz> i wasn't following much of the details.. i figure its really good the two were talking, but i wasn't able to get to any of our agenda items
8:16 AM <jstultz> :(
8:16 AM <jstultz> pundir: that said.. i did chat with badhri before the session, and he seemed ok with all the work you've done..
8:16 AM <jstultz> they are a little hesitant to the idea of moving mtp to userspace
8:16 AM <jstultz> (mostly because the churn of moving to ConfigFS gadget already has them anxious)
8:17 AM <pundir>  i looked at etherpad and though it was informative but hard to understand the context
8:17 AM <pundir> yes i'm doubt they would move mtp to userspace
8:17 AM <jstultz> pundir: yea.. honestly, i was lost for most of the talk.. i interrupted (probably a bit more rudely then i'd like) and asked how it all related to android..
8:17 AM <pundir> :)
8:18 AM <jstultz> and i'm not sure i understood the answer, but badhri had some feedback then so it felt like that was good.
8:18 AM <pundir> ok.. also anything on android accessory and other android gadgets?
8:18 AM <jstultz> badhri is also really focused on usb-c issues right now, so that's something we'll have to think about
8:19 AM <jstultz> pundir: nothing really.. he commented that there aren't many devices that uses them, but they're still important to the platform folks, so they have to be maintained..
8:19 AM <jstultz> pundir: he seemed to have a hard time being able to find devices to test with, so i'm not sure how we'll do..
8:20 AM <pundir> ok.. so his usb-c work, is it related to the usb dual role patch of his on aosp gerrit?
8:20 AM <jstultz> pundir: so i think the main item we should focus on is the notification issue
8:20 AM <jstultz> pundir: dunno.. maybe? haven't seen that one..
8:20 AM <pundir> yes my next question was on notification :)
8:20 AM <pundir> sorry joeskb7 for keep you waiting
8:20 AM <jstultz> pundir: if we can get the android userspace working w/ the notification side that would be helpful
8:20 AM <pundir> :)
8:20 AM <jstultz> joeskb7: :) yes sorry!
8:21 AM <joeskb7> jstultz: ok, so here is my status for the last week
8:21 AM <jstultz> pundir: then the in-kernel gadgets would likely have to stay in the android tree for now..
8:21 AM <pundir> notification side as in moving state notifications to UDC-core?
8:21 AM <jstultz> joeskb7: ok :)
8:21 AM <pundir> sorry again
8:21 AM <joeskb7> ok, here I go :)
8:22 AM <joeskb7> i made some progress with PPP task. figured that "can't find tunnel id" issue on server side was caused by wrong code in mtpd
8:23 AM <jstultz> joeskb7: nice!
8:23 AM <joeskb7> after I added code for using L2TP implementation from mainline kernel to mtpd, server failed on some point due to pppd was starting (by mtpd) with incorrect params
8:24 AM <joeskb7> particularly, pppd starting code needs to be reworked to use pppol2tp plugin
8:24 AM <joeskb7> this plugin should be ported from mainline ppp project to Android first
8:25 AM <joeskb7> btw, I added all my status to wiki, here: https://wiki.linaro.org/LMG/Kernel/PPP#Current_development_status
8:25 AM <jstultz> joeskb7: ok... careful there.. looks like the pppol2tp plugin is gpl?
8:25 AM <joeskb7> because it's some sort of milestone, I guess, so I decided it's important enough to be added
8:26 AM <joeskb7> need to check
8:27 AM <jstultz> joeskb7: yea.. i suspect android's ppp userspace is apache2.. but you might confirm
8:27 AM <joeskb7> what I figured is the whole ppp implementation (in Android) is nothing but original ppp project (old version of it, though). plugin is a part of ppp, if I understand correctly. so plugin should be shipped under the same license as ppp
8:28 AM <jstultz> joeskb7: yea. the android guys really don't want any gpl code in userspace.. there are a few exceptions, but they have been rewriting some of that..
8:28 AM <jstultz> if the pppd in android is gpl that makes it easier..
8:28 AM <jstultz> but do check carefully there
8:28 AM <joeskb7> one sec, i can check it really fast
8:30 AM <joeskb7> yeah, at the first glance looks like ppp itself goes under BSD-like license, but plugins are GPL
8:30 AM <jstultz> huh..
8:30 AM <joeskb7> but all plugins except pppol2tp are in Android anyway
8:31 AM <jstultz> joeskb7: ok.. so they do seem to have gpl plugins in there.. so that's at least a good precedence.
8:31 AM <joeskb7> hmm, tell you what. let me do this work first as it is, with pppol2tp plugin, and then we will figure it out
8:32 AM <joeskb7> because it is really tricky as it already is :)
8:32 AM <jstultz> joeskb7: sure.. integrating the exsting code is likely to be much easier then rewriting it.. so go ahead and get it working and we'll talk w/ the android devs about it
8:32 AM <joeskb7> jstultz: ok. in fact, this is what i'm trying to do right now :)
8:32 AM <jstultz> joeskb7: who were you chatting w/ earlier on the android team about this stuff?
8:33 AM <joeskb7> jstultz: JP Abgrall <jpa@google.com>
8:33 AM <joeskb7> jstultz: but he's not working on Android anymore
8:33 AM <jstultz> joeskb7: ok. so from what i heard.. yes.. he's not on android anymore..
8:34 AM <joeskb7> jstultz: yeah, just happened to know the answer to my question
8:34 AM <jstultz> joeskb7: there's a new guy Muhammad, who is covering some of JPA's work, so i'll try to get a contact there...
8:35 AM <jstultz> joeskb7: ok.. so sounds like good progress!
8:35 AM <jstultz> joeskb7: anything else you wanted to cover?
8:35 AM <joeskb7> jstultz: ok. so, right now i'm porting pppol2tp plugin to android, trying to get it build as shared library. get some collisions between ppp headers and bionic libc headers, but hopefully i will manage to fix it
8:36 AM <joeskb7> jstultz: once i have plugin built as library, i'm gonna rework mtpd so it runs pppd with pppol2tp plugin, like it's done in multihop.c
8:36 AM <jstultz> joeskb7: and what platform are you using for testing?
8:36 AM <joeskb7> jstultz: it's BeagleBoard X15
8:36 AM <jstultz> ok..
8:37 AM <joeskb7> jstultz: which is described pretty well in wiki page i'm maintaining for this task
8:37 AM <jstultz> joeskb7: sorry.. i just did a quick scan..
8:37 AM <jstultz> joeskb7: that wiki does look really nice, btw..
8:37 AM <joeskb7> jstultz: yeah, it doesn't really matters, you just need any Android board that has Ethernet on it, to repeat what I've done
8:38 AM <joeskb7> jstultz: thanks :)
8:38 AM <jstultz> joeskb7: yea.. it was just for my curiosity..
8:38 AM <joeskb7> jstultz: sure
8:38 AM <joeskb7> jstultz: sad thing is that i still need to split time between Linaro and TI tasks
8:38 AM ⇐ davi quit (~davi@gnu/davi) Ping timeout: 246 seconds
8:39 AM <joeskb7> jstultz: so i'm gonna spend some time on SELinux task starting from tomorrow. will get back to PPP from next week or so, I guess
8:39 AM <jstultz> joeskb7: ok.. is that something you want pushback from our side to try to reudce? or just a reality?
8:40 AM <joeskb7> jstultz: i guess we can't do much about it, since I'm an asignee from TI :)
8:40 AM <jstultz> joeskb7: ok.. though there have been cases in the past where assignees were really only doing internal non-linaro work, and that's not really how being an asignee works..
8:41 AM <jstultz> joeskb7: just wanted to check :)
8:41 AM <joeskb7> jstultz: i guess. but it's not up to me, you know :)
8:41 AM <joeskb7> jstultz: ok. so i think this is about it for me
8:41 AM <jstultz> joeskb7: ok..
8:42 AM <jstultz> pundir:  so i wanted to get back to next actions..
8:42 AM <pundir> jstultz, sure
8:43 AM <jstultz> so i'm thinking the notification bits, if we can convert that to the UDC method, but keep the rest of the android in-kernel mtp gadget for now, that's probably going to be a good step
8:44 AM <jstultz> pundir: then we can look at trying to see if there's a BSD/ASL2.0 licenced mtp library to try to move to using functionfs
8:44 AM <jstultz> but that's likely to be a longer play
8:44 AM <jstultz> pundir: i think getting and android guys to just move off of android gadget is going to be a big enough transition
8:45 AM <jstultz> pundir: i'm thinking we may even consider pushing their mtp gadget to staging... in the meantime..
8:45 AM <jstultz> pundir: does that all make sense?
8:45 AM <pundir> yeah we can look at that mtp-responder from jolla/mer which the reviewer was talking about
8:46 AM <pundir> how does pushing to staging will take place? is there a separate process for that? or we just push it to staging directory and mark it as in process of userspace migration?
8:46 AM <jstultz> pundir: yea.. the buteo-mtp does seem to be BSD
8:47 AM <jstultz> pundir: so i was thinking the current mtp gadget function.. that might be something we can do via staging..
8:47 AM <jstultz> pundir: but i'm not sure yet..
8:47 AM <jstultz> pundir: really i suspect accessory and audio can be ignored for now.
8:48 AM <jstultz> if we can't test them..
8:48 AM <pundir> also i looked at gbsim gadget and it does use functionfs and does all usb endpoint magic in userspace.. so that was quiet encouraging from mtp point of view
8:48 AM <jstultz> though they may be interesting test cases for learning functionfs :)
8:48 AM <jstultz> yea
8:48 AM <pundir> if we can replicate the mtp processing from mtp-gadget driver to that userspace daemon
8:48 AM <jstultz> the big thing with mtp in userspace that the android guys said, was that they don't want any regressions at all.
8:49 AM <pundir> but this ^^ was just based  a quick look at gbsim
8:49 AM <jstultz> and that sort of testing might take awhile to get the confidence up
8:49 AM <pundir> yeah thats regression part is true
8:49 AM <pundir> it might take a while to get it in aosp even if we able to pull something like that
8:49 AM <jstultz> pundir: the downside with moving the mtp-gadget to userspace, is that the kerenl code is gpl2 :P
8:49 AM <jstultz> pundir: so they might not take it..
8:50 AM <pundir> :)
8:50 AM <jstultz> pundir: more likely we'd have to see if they would relicense it and let us move it
8:50 AM <pundir> yeah its totally upto them right..
8:50 AM <jstultz> pundir: but we could use it for testing..
8:50 AM <pundir> yes
8:51 AM <pundir> so i'll take a look at it but my priority is that configfs talk which i'm not really confident about as of today..
8:51 AM <jstultz> but anyway.. all stuff for later.. i think fixing the notification thing first is a good plan, because they only did their own thing because 3.10 didnt' support anything, so they wrote
their own solution.
8:51 AM <pundir> may be while looking at different configfs examples I'll learn a trick or two :)
8:51 AM <jstultz> pundir: if you want me to review slides or listen to early dry-runs of the talk, let me kow..
8:52 AM <jstultz> practice is really the key for these talks.. :)
8:52 AM <pundir> about notifications.. the state sysfs attribute is already there in /sys/claa/udc/.. so we are good there
8:52 AM <pundir> i would definitely do that :)
8:52 AM <jstultz> (we so rarely get to do our talks more then once)
8:52 AM <pundir> back to uevent notifications, will upstream be accepting a solution like that
8:53 AM <jstultz> pundir: right.. i'd look to see if userspace can be reworked to support either interface depending..
8:53 AM <pundir> no one commented on uevent part on my patch
8:53 AM <jstultz> so the uevent side is still different then what is upstream?
8:53 AM <jstultz> (sorry, i'm only following this in rough brush-strokes)
8:54 AM <pundir> yes, there is no uevent notification available in upstream
8:54 AM <pundir> only sysfs state reporting is available and that too reported by UDC device class
8:55 AM <jstultz> pundir: ok.. so does it make sense to see about adding uevents  to the upstream udc sysfs hooks?
8:55 AM <pundir> android need both of them ie. uevent notification and sysfs usb state reporting
8:55 AM <pundir> jstultz, yes i tried that "hookin" uevents to drivers/usb/gadget/udc-core and it works
8:56 AM <jstultz> pundir: ok.. great.. i'd rework that patch and send it by itself for comments..
8:56 AM <pundir> remember I said only USB state mapping has to be worked on
8:56 AM <pundir> jstultz, ok
8:56 AM <jstultz> (sorry.. last few weeks have been a blur)
8:56 AM <pundir> so i'll do that
8:56 AM <pundir> no problem :)
8:56 AM <pundir> i'll draft a patch for internal review
8:57 AM <pundir> and if it make sense to you, then we'll send it to upstream for review
8:57 AM <jstultz> pundir: ok.. feel free to send me the patches when you get the first draft and i'll use my "forgot everything" goldfish mind to review it with fresh eyes :)
8:57 AM <jstultz> yep
8:57 AM <pundir> :)
8:58 AM <jstultz> ok.. i think that's it.. i'm working on a summary of what all went on at the plumbers android sessions... hopefully will have a draft in the next few days that you guys can review
8:58 AM <pundir> great!! thanks
8:59 AM <jstultz> i'm hoping it can go to lwn, but it just depends on how long it takes me to write up
8:59 AM <pundir> :)
8:59 AM <jstultz> pundir: so far i'm only ~30 minutes into 6.5 hours and its 2.5 pages
9:00 AM <jstultz> pundir: so editing will likely be required
9:00 AM <pundir> I'd really love to read that unedited version :)
9:00 AM <jstultz> pundir: joeskb7: ok.. thanks so much for the chat.. i'll send the logs to sumit so he's aware
9:01 AM <pundir> thanks jstultz, joeskb7

LMG/Notes/Kernel-2015-08-25-logs (last modified 2015-09-02 14:45:13)