**** BEGIN LOGGING AT Thu May 17 03:00:02 2018
May 17 06:49:09 <Tofe> Herrie: adding setns helped a bit, but now it seems I'm missing other bits of namespace (pid in this case). I guess more patching would be required.
May 17 07:06:21 <Herrie|Laptop> Morning!
May 17 07:08:41 <Tofe> Morning :)
May 17 07:58:56 <Herrie|Laptop> Tofe: I thought we had the pid namespaces already?
May 17 08:03:25 <Herrie|Laptop> I mean with https://github.com/Halium/android_kernel_lge_hammerhead/commit/bd221854de33b75db7a7fa01cb34274b62a7cbf8 and https://github.com/Halium/android_kernel_lge_hammerhead/commit/5754614eb43dea44a99e54898e3b83d4d96d8b83
May 17 08:03:38 <Herrie|Laptop> Or there are some missing bits?
May 17 08:15:35 <Tofe> Herrie|Laptop: maybe I'm confusing it with user namespace, I don't remember well. But from http://lxc.sourceforge.net/patches/linux/3.0.0/3.0.0-lxc1/patches/, we should try to have 0006 to 0008 at least
May 17 08:16:15 <Tofe> unfortunately these patches don't apply that well on our 3.4 kernel
May 17 08:17:53 <Herrie|Laptop> Tofe: Found this for 3.3 kernel
May 17 08:17:54 <Herrie|Laptop> https://github.com/confine-project/openwrt/blob/master/target/linux/generic/patches-3.3/19914-ns-proc-Add-support-for-the-mount-namespace.patch
May 17 08:18:00 <Herrie|Laptop> Might be closer to 3.4 ;)
May 17 08:18:36 <Herrie|Laptop> https://github.com/confine-project/openwrt/blob/master/target/linux/generic/patches-3.3/19915-ns-proc-add-support-for-the-pid-namespace.patch
May 17 08:21:54 <Tofe> ah, yes, thanks!
May 17 08:31:18 <Herrie|Laptop> Tofe: This refers to snap, but has some useful links in it that might work for LXC as well: https://forum.snapcraft.io/t/snap-confine-issue-when-trying-to-run-hello-world-snap-on-a-3-4-kernel/538/34
May 17 08:32:15 <Tofe> we have the mount namespace now, that's fine
May 17 08:32:33 <Herrie|Laptop> Patches like https://github.com/torvalds/linux/commit/332a2e1244bd08b9e3ecd378028513396a004a24 and https://lists.linuxfoundation.org/pipermail/containers/2012-November/031023.html
May 17 08:34:57 <Tofe> there are some commits also here http://visa.lab.asu.edu/gitlab/fstrace/android-kernel-msm-hammerhead-3.4-marshmallow-mr3/commits/android-msm-hammerhead-3.4-marshmallow-mr3
May 17 08:42:26 <Herrie|Laptop> Tofe: Here's also a Trello with 3.4 attempt to get Anbox working with some useful links (including to Tofee github :P) https://trello.com/c/JvkxnGMr/4-port-sf-droid-anbox
May 17 08:46:41 <Tofe> Yes, I saw that one too :) Ah, fame !
May 17 09:08:49 <Herrie|Laptop> LOL
May 17 09:14:48 <saidinesh5> @Herrie|Laptop tried a bunch of patches like that for backporting userns support to 3.4 kernel.. its a mess i tell you
May 17 09:15:02 <Herrie|Laptop> saidinesh5: OK
May 17 09:15:07 <saidinesh5> the problem is one little patch depends on another patch on some other code..
May 17 09:15:09 <Herrie|Laptop> So no luck with Anbox on 3.4 at your end?
May 17 09:15:52 <saidinesh5> on 3.4 ? nope. i had to disable devpts ns, and other namespaces.. because even their implementation were buggy
May 17 09:16:37 <Herrie|Laptop> I see
May 17 09:16:46 <Herrie|Laptop> My Mido has 3.18 which should behaving better
May 17 09:16:49 <saidinesh5> at last got tired of the thing after trying to fix the namespace patches for overlayfs and decided to look into it once my device has 3.10 kernel
May 17 09:16:51 <Herrie|Laptop> Just UID 100000 seems to be an issue
May 17 09:17:02 <saidinesh5> Ahh that sounds like the user namespace patches
May 17 09:17:14 <saidinesh5> i think for 3.18 you should be able to get  it up and running
May 17 09:17:44 <saidinesh5> krnlying worked on anbox on 3.18 itself iirc
May 17 09:19:19 <saidinesh5> @Herrie|Laptop https://github.com/sfdroid/anbox/projects/1
May 17 09:19:19 <Herrie|Laptop> It seems to work mostly, just not compeltely yet
May 17 09:19:35 <saidinesh5> yeah mal fixed the input bug recently iirc
May 17 09:19:52 <saidinesh5> (note that sfdroid's anbox is different from "standard anbox")
May 17 09:20:19 <saidinesh5> we expose libegl etc.. to the container and tried to get full acceleration
May 17 09:21:12 <saidinesh5> and directly use libwayland_client to draw windows ..
May 17 09:22:02 <saidinesh5> https://github.com/mlehtima/anbox/commits/rebase (up to date one)
May 17 09:25:09 <Herrie|Laptop> saidinesh5: Well ours will be different again since we use the bitbake etc bits
May 17 09:25:27 <Herrie|Laptop> But seems morphis has some more time on his hands now so he can more actively merge Anbox bits again too ;)
May 17 09:25:30 <saidinesh5> bitbake is just packaging no?
May 17 09:27:07 <Tofe> saidinesh5: yes
May 17 09:27:29 <Tofe> saidinesh5: for now we're simply on vanilla-ish anbox, through SDL (wayland variant)
May 17 09:27:49 <saidinesh5> Tofe: Ahh so you have anbox running?
May 17 09:28:17 <Herrie|Laptop> saidinesh5: Yeah just like you have OBS ;)
May 17 09:28:19 <Tofe> nope, I have only a hammerhead on my side to test with, so... wel...
May 17 09:28:28 <saidinesh5> hahha i see
May 17 09:28:37 <Tofe> but the container is starting
May 17 09:28:47 <Herrie|Laptop> I have a Mido with 3.18 but not much luck for the rest
May 17 09:28:57 <Tofe> just surfaceflinger crashes all the time, so it nevers reaches the "booted" state
May 17 09:28:58 <Herrie|Laptop> And our VirtualBox has too old Mesa to be useful
May 17 09:29:04 <saidinesh5> Tofe: container starts but the problems occur inside that.. yeah
May 17 09:29:44 <saidinesh5> Tofe: btw. your hammerhead now has upstream kernel no?
May 17 09:29:50 <Tofe> I'm not even sure why it crashes, so it's a bit frustrating - might be a minor thing
May 17 09:29:55 <saidinesh5> upstreamish*
May 17 09:30:03 <Tofe> saidinesh5: I use halium's
May 17 09:30:27 <saidinesh5> Ahh iirc there is even an upstreamish kernel for you .. been wanting to look at it myself: https://github.com/flto/linux/wiki/hammerhead-upstream
May 17 09:30:44 <bshah> That kernel is not really in useful state tbh
May 17 09:30:51 <saidinesh5> Ahh
May 17 09:30:58 <bshah> I mean well..
May 17 09:31:02 <saidinesh5> new phone it is for me too then
May 17 09:31:05 <bshah> it is still under heavy work in progress
May 17 09:31:17 <saidinesh5> i wanted to try this before getting another toy phone to play with
May 17 09:31:26 <saidinesh5> (for my cancro..)
May 17 09:32:15 <saidinesh5> Tofe: so just surfaceflinger crashes? or arbitrary commands crash?
May 17 09:32:20 <Tofe> I'll probably get a mido at some point
May 17 09:32:34 <Tofe> saidinesh5: just SF it seems
May 17 09:32:47 * saidinesh5 has to look up his logs
May 17 09:32:55 <Herrie|Laptop> Tofe: Well there are other Xiaomi devices that should be fine too ;)
May 17 09:33:12 <Tofe> Herrie|Laptop: true, I'll wander a bit :)
May 17 09:33:13 <saidinesh5> i was looking forward for that Mi A2 .. but no headphone jack
May 17 09:33:39 <saidinesh5> maybe Redmi Note 5 pro or a Oneplus 5 if it's price falls
May 17 09:34:17 <Herrie|Laptop> The Redmi devices are good value for money
May 17 09:34:24 <Tofe> yeah a redmi note 5 pro might be my next personal device very soon - but I won't hack my everyday device :)
May 17 09:34:36 <Herrie|Laptop> And seem pretty OK in terms of support, community etc
May 17 09:34:36 <saidinesh5> lol
May 17 09:35:41 <saidinesh5> yeah honestly they seem nice. plus for me only xiaomi/oneplus devices because they wont void the warranty on rooting
May 17 09:36:01 <Tofe> saidinesh5: would you have an idea how to debug a surfaceflinger crash ? Are there symbols for the images provided by Anbox's website ?
May 17 09:36:03 <Tofe> morphis: ^
May 17 09:36:21 <saidinesh5> Tofe: how to debug - yes.. i was looking at my chat logs
May 17 09:36:21 <Herrie|Laptop> Well rooting on Xiaomi is a pain
May 17 09:36:33 <saidinesh5> rooting lenovo is a bigger pain..
May 17 09:36:38 <Herrie|Laptop> With their central authorization server etc
May 17 09:36:39 <Herrie|Laptop> But well
May 17 09:36:41 <Herrie|Laptop> Doable
May 17 09:36:42 <Tofe> s/is a pain/takes a couple of days/
May 17 09:36:53 <Herrie|Laptop> I managed to hard brick mine by flashing LuneOS LOL
May 17 09:37:47 <saidinesh5> Tofe: https://trello.com/c/JvkxnGMr/4-port-sf-droid-anbox ^ i used sshfs to remote mount my large symbols folder from my desktop and used the libhybris gdb scripts to debug
May 17 09:37:54 <Tofe> yep, flashing the wrong boot partition usually ends with a scary status
May 17 09:38:09 <Herrie|Laptop> Luckily fixing it was pretty straight forward too :P
May 17 09:38:12 <saidinesh5> our surfaceflinger crashes happened because the vendor blobs provided some older library versions
May 17 09:39:06 <Tofe> could be my case too, I don't even have surfaceless support
May 17 09:39:27 <Tofe> saidinesh5: but I'll need the debug symbol from anbox's build, right?
May 17 09:39:36 <saidinesh5> but you guys aren't using vendor blobs right? i mean your rendering happens inside SDL -
May 17 09:39:57 <saidinesh5> Tofe: yep. you have to build the anbox-image first
May 17 09:40:08 <saidinesh5> took like 5-6 hours on my laptop
May 17 09:40:24 <Tofe> saidinesh5: well... anbox->SDL->wayland->hybris->vendor blobs
May 17 09:41:11 <Tofe> maybe the devs can upload a symbols archive
May 17 09:41:15 <saidinesh5> in our case, it was anbox->vendor blobs->wayland_client -> wayland
May 17 09:41:44 <saidinesh5> wayland/lipstick
May 17 09:43:00 <Tofe> ok
May 17 09:43:42 <Tofe> I could try inserting traces in the host gles driver, but I'm not even sure that'll help
May 17 09:44:26 <saidinesh5> oh chances are your crashes are happening because surfaceflinger was expecting some /dev/ something - and nothing to do with vendor blobs
May 17 09:44:49 <saidinesh5> look at strace logs - they tell you a lot about the crashes
May 17 09:46:05 <Tofe> saidinesh5: how do you insert strace between the container init and surfaceflinger ?
May 17 09:46:16 <Tofe> (and get the output !)
May 17 09:46:41 <saidinesh5> well you can simply chroot into container and strace init.. but android docuemntation has info on this
May 17 09:46:57 <Tofe> oh, ok, good
May 17 09:47:10 <saidinesh5> https://source.android.com/devices/tech/debug/strace
May 17 09:47:38 <morphis> Tofe: there should be a tombstone Android generates in /data/tombstones
May 17 09:47:48 <morphis> that should give details
May 17 09:47:57 <Tofe> morphis: ah yes, good idea, thanks
May 17 09:48:39 <morphis> most likely its crashing because it can't talk to the anbox session-manager
May 17 09:48:50 <morphis> you have a dump of the logcat output?
May 17 09:49:35 <Tofe> morphis: not here; but I didn't remark anything weird last time I checked
May 17 09:49:48 <Tofe> morphis: logcats are merged for all the containers, right?
May 17 09:50:15 <Tofe> (i.e. there's only one logcat space in the kernel)
May 17 09:50:51 <Tofe> because I'm starting logcat from luneos directly
May 17 10:00:09 <morphis> no, the Anbox container one is on its own
May 17 10:00:19 <morphis> it should
May 17 10:58:28 <Tofe> ah, ok, that's why I didn't see anything... :)
May 17 16:57:35 <Tofe> no tombstone, all file are size zero...
May 17 17:05:14 <Tofe> saidinesh5: ping
May 17 17:14:13 <saidinesh5> Tofe: pong
May 17 17:14:17 <saidinesh5> nice timing!
May 17 17:14:32 <saidinesh5> just came back home
May 17 17:15:43 <Tofe> :)
May 17 17:15:55 <Tofe> saidinesh5: I wondered how you chrooted into the container
May 17 17:43:07 <saidinesh5> Tofe: ahh.. sorry afk .. lxe mounts it's rootfs to /var/something.. hold
May 17 17:43:53 <Tofe> to /var/lib/anbox/root yes, but it's not the real lxc rootfs, that one is in /proc/<init pid>/root
May 17 17:45:20 <saidinesh5> oh i chrooted into that path, after manually mounting anbox image and devpts etc.. there
May 17 17:45:25 <saidinesh5> and then straced into there
May 17 17:45:47 <Tofe> ah ok
May 17 17:45:53 <saidinesh5> (without starting the container that is)
May 17 17:46:07 <saidinesh5> also we used overlayfs to save a lot of time - not having to rebuild anbox-image
May 17 17:46:40 <saidinesh5> any changes , extra config files , vendor blobs are all exposed to anbox root via anbox-image + overlayfs
**** ENDING LOGGING AT Fri May 18 03:00:03 2018