[OS] HIPSTER STRETCHY PANTS (XU4 debian stretch gnome3-wl)

by **memeka** » Sat Mar 04, 2017 3:03 pm

After a lot of teasers, I have decided to share my Gnome3 Wayland XU4 image.

link: http://oph.mdrjr.net/memeka/images/hips ... 032017.img
md5: 041ea0447ed1fd97da125f43960aa8fb
login: odroid (Odroid User)
password: odroid
root password: odroid

There is a lot of work I have put in this image for many years:

* starting with installing gnome3 on the Odroid U3 in 2014: https://www.youtube.com/watch?v=COyZJLHMGDs
* trying gnome3 on the XU4 in 2015: https://www.youtube.com/watch?v=fvhu_OkGKtA
* realising gnome3 can work well only with wayland, and playing with wayland on the original odroid XU in 2014: https://www.youtube.com/watch?v=DRBzOpxEaiU and 2015: https://www.youtube.com/watch?v=5Ha6vFol8ck
* finally getting wayland working on the odroid XU3 at the beginning of 2016 using the TIZEN drivers and the 4.0 kernel: https://www.youtube.com/watch?v=Eza6nIgyZA0
* developing first Mali integration for the 4.x mainline kernel later on in 2016 (kernel 4.7): https://www.youtube.com/watch?v=o0aFehweAyk
* getting gnome3 working after a lot of patching 6 months ago (gnome 3.20 in ubuntu): https://www.youtube.com/watch?v=RbMLkXbB3Gs

This is not the final release I wanted, but more of a preview of Gnome on Wayland. Here is what you're getting:

* kernel 4.9.12 custom - mostly similar to the HK 4.9 kernel, with the main difference that it has my own Mali integration that uses devfreq, with a special governor for mali. Currently it's set at boot to "performance" and uses Mali at 600Mhz, but it can be changed to "simple_ondemand" where it lower the Mali clock to 177Mhz. The HK kernel should work, but I find the GPU performance better with my integration.
* debian 7 stretch rootfs (testing but all packages are frozen)
* Wayland Mali userspace drivers from ARM r12p0
* gnome 3.22 and several apps installed
* BE AWARE THE IMAGE IS SET TO BOOT FROM SD-CARD! if you want emmc, you need to replace /dev/mmcblk1 with /dev/mmcblk0 in /boot/boot.ini and /etc/fstab. Also the monitor is set to DVI mode in boot.init so there's no HDMI audio device -- edit boot.ini to enable it!

* Netflix 1080p streaming with kodi: viewtopic.php?f=96&t=26172&p=185684#p185684

What you should be looking for (good and bad):

* gdm3 working. should start automatically on boot
* the image boots in graphical mode mode, but you can run sudo systemctl set-default multi-user.target and at next boot it goes to text-mode -- where you can start gdm3 with sudo service gdm3 start
* from gdm3 login, you can choose Gnome on Wayland session or Wayland session (from the little gear button)
* most gnome3 apps working, with a few caveats: gnome-maps crashes the session (gnome-shell) on exit; gnome-videos (totem) sometimes crashes gnome-shell on exit
* gnome-videos (totem) has full video acceleration: can play video with no issues (using egl and mfc accelerations). like i said above, due to some gnome-shell issue (it works well in weston) sometimes it crashes the session when you exit it.
* gnome videos can play youtube videos accelerated (egl and mfc): https://www.youtube.com/watch?v=IGk9V5tBOCQ
* mpv can play videos very well using egl without crashing using ffmpeg. However, since I did not include my custom mfc-enabled ffmpeg, mpv uses software decoding. You can try applying the patches to ffmpeg and get mpv fully egl and mfc accelerated: https://www.youtube.com/watch?v=Ie7BR3WgXWQ&t=9s
* mpv can play youtube too: https://www.youtube.com/watch?v=ydWF7ADj6jU
* mfc-accelerated decoding on gstreamer (using the default repository package, not my custom encoding-enabled gst) that can use egl accelerated clutterautovideosink and glimagesink (gst-plugins-bad required a simple patch to enable gles2)
* browser: epiphany -- can do egl & gstreamer decoding !!! -- BUT due to a bug in the mali drivers, tabs that use egl (or gstreamer, since it calls for egl) will not be rendered! I have complained about this to ARM, but of course they have not replied. Apparently, there's an internal patch that fixes this bug, which will allow FULL browser acceleration for webgl and video decoding!!! - but these drivers are not published yet. See https://community.arm.com/graphics/f/di ... -in-webkit and https://community.arm.com/graphics/f/di ... ce-drivers
* There are some GTK3 firefox builds that might work well, but also I don't think they have EGL yet.
* XWayland works well, but without EGL. This means you can run chromium and should work well, but without EGL.
* LibreOffice should work well since it's using GTK3: https://www.youtube.com/watch?v=4WT1i9NU7Tk

Well, this turned out to be a long summary. Please let me know your findings. And please try and debug issues and contribute.
Cheers and thanks to all that helped and made this possible!

PS: kodi and retroarch should work on this image if they are configured for the wayland backend (even if the latest kodi doesn't support officially wayland anymore, there are unofficial repositories that still maintain wayland compatibility). You've already seen kodi on wayland working on my image: https://www.youtube.com/watch?v=o0aFehweAyk

PSS: if you're wondering about the naming:
* hipster - well, according to @mdrjr, wayland is for hipsters

* stretchy - based on debian stretch
* pants - first release, getting you all with your pants down

by **memeka** » Sat Mar 04, 2017 3:06 pm

PATCHES:

Gnome3/Mutter
==============

Gnome3 Wayland support has improved A LOT and continues to improve each version. For example, in Gnome 3.18, I had to patch mutter, clutter, and cogl. Now, in 3.22, there is only mutter to apply a few patches. In 3.24, I've seen the patching is way simpler - just replace one string for another (the string checks an OpenGL-ES extension), and that's it!
Here are the patches for 3.22:
1: use software cursor (else the mouse cursor will not be rendered)

Code: Select all: commit d2a6c6a55ba3d7aaf88a494de3e4e84a08325cac Author: memeka <mihailescu2m@gmail.com> Date: Tue Nov 1 09:39:01 2016 +1030 exynos: use software cursor Signed-off-by: memeka <mihailescu2m@gmail.com> diff --git a/src/backends/native/meta-cursor-renderer-native.c b/src/backends/native/meta-cursor-renderer-native.c index f4ea8d2..aad315c 100644 --- a/src/backends/native/meta-cursor-renderer-native.c +++ b/src/backends/native/meta-cursor-renderer-native.c @@ -329,7 +329,9 @@ should_have_hw_cursor (MetaCursorRenderer *renderer, if (!has_valid_cursor_sprite_gbm_bo (cursor_sprite)) return FALSE; - return TRUE; + // exynos hack: use software cursor + return FALSE; + // return TRUE; } static gboolean

2. enable GLES2 driver as default (or else it will try OpenGL and fail)

Code: Select all: commit 220e562b0f92d457c2b06a5d04ace86219fb7dfa Author: memeka <mihailescu2m@gmail.com> Date: Tue Nov 1 10:03:18 2016 +1030 mali on wayland: enable gles2 driver as default Signed-off-by: memeka <mihailescu2m@gmail.com> diff --git a/debian/rules b/debian/rules index 6aba938..92611a8 100755 --- a/debian/rules +++ b/debian/rules @@ -13,7 +13,14 @@ override_dh_autoreconf: ifeq ($(DEB_HOST_ARCH_OS),linux) CONFFLAGS = \ --enable-wayland \ - --enable-native-backend + --enable-native-backend \ + --enable-kms-egl-platform \ + --enable-wayland-egl-platform \ + --enable-wayland-egl-server \ + --with-default-driver=gles2 \ + --enable-cogl-gles2 \ + --enable-gles2 \ + --with-gles2-libname=libGLESv2.so.2 else CONFFLAGS += \ --disable-wayland-egl-platform \

This patch is for the debian packaging; for other systems you need to adapt it and add the configure flags from above.

3. use eglGetPlatformDisplayEXT instead of eglGetDisplay to initialize EGL (because eglGetDisplay fails)

Code: Select all: commit 250520ace3c953667e8e9d10b2568be851797815 Author: memeka <mihailescu2m@gmail.com> Date: Mon Feb 27 12:45:57 2017 +1030 mali on wayland: try eglGetPlatformDisplayEXT before eglGetDisplay Signed-off-by: memeka <mihailescu2m@gmail.com> diff --git a/src/backends/native/meta-renderer-native.c b/src/backends/native/meta-renderer-native.c index 103e6ad..ae7f247 100644 --- a/src/backends/native/meta-renderer-native.c +++ b/src/backends/native/meta-renderer-native.c @@ -235,31 +235,15 @@ static EGLDisplay meta_egl_get_display (void *native) { EGLDisplay dpy = NULL; - const char *client_exts = eglQueryString (NULL, EGL_EXTENSIONS); - if (g_strstr_len (client_exts, -1, "EGL_KHR_platform_base")) - { - PFNEGLGETPLATFORMDISPLAYEXTPROC get_platform_display = - (void *) eglGetProcAddress ("eglGetPlatformDisplay"); - - if (get_platform_display) - dpy = get_platform_display (EGL_PLATFORM_GBM_MESA, native, NULL); - - if (dpy) - return dpy; - } - - if (g_strstr_len (client_exts, -1, "EGL_EXT_platform_base")) - { - PFNEGLGETPLATFORMDISPLAYEXTPROC get_platform_display = + PFNEGLGETPLATFORMDISPLAYEXTPROC get_platform_display = (void *) eglGetProcAddress ("eglGetPlatformDisplayEXT"); - if (get_platform_display) - dpy = get_platform_display (EGL_PLATFORM_GBM_MESA, native, NULL); + if (get_platform_display) + dpy = get_platform_display (EGL_PLATFORM_GBM_KHR, native, NULL); - if (dpy) + if (dpy) return dpy; - } return eglGetDisplay ((EGLNativeDisplayType) native); }

That's it! Pretty easy

Linux kernel
===========

Like I said, everything should be ok (albeit maybe slower EGL) with the HK kernel. But for those that want to keep track of my own kernel, you can find it at https://github.com/mihailescu2m/linux/t ... dxu4-4.9.y.
Please not I am always rebasing the patches, so you'll need to force updates

My config: https://transfer.sh/t5zqE/memeka.config

by **mad_ady** » Sun Mar 05, 2017 1:04 am

Yes, excellent, but is it gluten-free? [emoji14]
Lots of people were waiting for this

by **memeka** » Sun Mar 05, 2017 4:55 am

All uploaded, enjoy

by **fleixi** » Sun Mar 05, 2017 11:58 pm

Nice. It seems to be that im a hipster :-)

I have tried to get wayland working for years too but without much sucess, nice to see its working.

How about kodi. Do i need to compile for my self and if yes could you post the link to the source with wayland and mfc buildin? I have only found the source without mfc patch

do have kodi 17 have the same problems on wayland than on x11

by **memeka** » Mon Mar 06, 2017 3:14 am

fleixi wrote:Nice. It seems to be that im a hipster

I have tried to get wayland working for years too but without much sucess, nice to see its working.

How about kodi. Do i need to compile for my self and if yes could you post the link to the source with wayland and mfc buildin? I have only found the source without mfc patch

do have kodi 17 have the same problems on wayland than on x11

i haven't tried kodi yet (well, i tried kodi 16 6 months ago, but don't have that source).
i would make a diff between the owersun source and mainline kodi (to get the mfc patch), apply that patch to the source you found, and compile

by **memeka** » Mon Mar 06, 2017 12:46 pm

totem and gnome-maps crashing are because of issues in the mali driver - probably, the driver frees an object too early:

Code: Select all: Core was generated by `totem bbb_720p.mov'. Program terminated with signal SIGSEGV, Segmentation fault. #0 get_next_argument (signature=0x2c <error: Cannot access memory at address 0x2c>, details=details@entry=0xbee39a9c) at ../src/connection.c:430 430 for(; *signature; ++signature) { [Current thread is 1 (Thread 0xb213cd70 (LWP 12827))] (gdb) bt #0 get_next_argument (signature=0x2c <error: Cannot access memory at address 0x2c>, details=details@entry=0xbee39a9c) at ../src/connection.c:430 #1 0xb4ce69ba in wl_argument_from_va_list (signature=<optimized out>, args=args@entry=0xbee39acc, count=count@entry=20, ap=..., ap@entry=...) at ../src/connection.c:493 #2 0xb4ce5598 in wl_proxy_marshal (proxy=0x7f6bedb0, opcode=1) at ../src/wayland-client.c:692 #3 0xb4f8685e in window_surface_delete () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #4 0xb4f7e1e4 in eglp_window_surface_specific_deinitialization () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #5 0xb4f7cd14 in eglp_delete_surface () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #6 0xb4f7ce74 in eglp_destroy_all_non_current_surfaces () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #7 0xb4f7a71a in eglp_try_display_finish_terminating () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #8 0xb4f7b1e2 in eglTerminate () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #9 0xb4f7b22c in eglp_unload_callback () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #10 0xb4decc24 in osup_term_unload_hooks () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #11 0xb4dde4ca in osup_c_unload_hook () from /usr/lib/arm-linux-gnueabihf/egl-current/libwayland-egl.so.1 #12 0xb6fd3f42 in ?? () from /lib/ld-linux-armhf.so.3 Backtrace stopped: previous frame identical to this frame (corrupt stack?)

The window_surface_delete method at #3 sends an invalid proxy object (proxy=0x7f6bedb0) to the wayland library:

Code: Select all: (gdb) print (struct wl_proxy) *0x7f6bedb0 $3 = {object = {interface = 0x7fe1bfc8, implementation = 0x7fb51c30, id = 44}, display = 0x7f660ec0, queue = 0x7f660f2c, flags = 2, refcount = 1, user_data = 0x0, dispatcher = 0x0, version = 3} (gdb) print (struct wl_interface) *0x7fe1bfc8 # => this is proxy->interface - you can see the name is garbage already $4 = {name = 0xa93e931d "iXh\377\367Һ\022KP!0\265{D\021L\205\260\025F\034Y#h\003\223\377\367\f\354\016IjF", version = 49, method_count = -2147421248, methods = 0x7f6beda8, event_count = 0, events = 0x0} (gdb) print (struct wl_message) *0x7f6beda8 # => this is proxy->interface->methods => you can see the signature field cannot be accessed (0x31 is invalid) leading to the segmentation fault $5 = {name = 0x0, signature = 0x31 <error: Cannot access memory at address 0x31>, types = 0x7fe1bfc8}

I've reported this to ARM and let's hope it will be fixed in the next driver release. https://community.arm.com/graphics/f/di ... segm-fault

by **rooted** » Mon Mar 06, 2017 1:52 pm

Subscribed

by **memeka** » Wed Mar 08, 2017 8:35 pm

after installing the image, you should resize your partition to maximum space available:

Code: Select all: sudo fdisk /dev/mmcblk1 p (see starting block of p2, e.g. 266240) d (delete) 2 (partition 2) n (new) p (primary) 2 (partition 2) 266240 (starting block) <enter> (it will take ending block as end partition, so you have entire space) w (write partitions)

exit fdisk, reboot, then run:

Code: Select all: sudo resize2fs /dev/mmcblk1p2

that's it, check out space with:

Code: Select all: df -h

by **AreaScout** » Thu Mar 09, 2017 2:00 am

So for XU3 i only have to replace the kernel or is there some other special thing ? Maybe you can provide your custom kernal sources or binary for XU3 ?

RG

by **memeka** » Thu Mar 09, 2017 2:26 am

Although I did not test, the image should work on XU3 too.

The kernel is at https://github.com/mihailescu2m/linux and the one used in the image is at commit https://github.com/mihailescu2m/linux/t ... aee8bb8e7c

I will publish the config i used, since it's not in the repository.

If you have a XU3 to test, let me know if the kernel works and if drivers are missing; I can try and make some config changes in the future to include all XU3 features.

by **blu** » Thu Mar 09, 2017 9:01 am

Hi memeka,

I'm the guy who was nagging you to share your work @ cnx ; )

First of all, big thanks for actually sharing it - you've made a difference! Just to give you an idea:

On r14 @ x11 (stock XU4 xenial) (vsync on):
Dancing pirate @ 640x640 8x msaa: 51 fps
Material sphere @ 640x640 8x msaa: 55 fps

..where one big CPU core constantly stays at 85-100% load.

On your r12 @ gnome3-wayland (vsync on):
Dancing pirate @ 640x640 8x msaa : 60fps
Material sphere @ 640x640 8x msaa : 60fps

..where the above are running in parallel! With no CPU cores stuck at 100%!
https://vimeo.com/207552599

It's good to be a hipster! Seriously, though, if this is not a material proof why Wayland is much needed for this platform, let me add a comparison to another similar platform running unity8-mir (MT8163A):

Dancing pirate @ 1920x1200 : 38 fps
Material sphere @ 1920x1200 : 27 fps

That at 1/3 the ALU and 1/2 the RAM BW of the XU4. All thanks to Mir.

by **rooted** » Thu Mar 09, 2017 9:06 am

Your video link is bad.

by **blu** » Thu Mar 09, 2017 9:53 am

To put things into perspective (and show that Midgard drivers r14 are not the bottleneck), here are the same tests on x11 again, but this time only every 4th frame is shown on screen -- the rest are drawn to completion and thrown away.

r14 @ x11, displayed 1 frame out of 4 (vsync on):
Dancing pirate @ 640x640 8x msaa : 57 fps
Material sphere @ 640x640 8x msaa : 230 fps

Performance of the second test jumps several-fold, as it's really not CPU-dependent in any form, whereas the first increases some, but still the blitting of every 4th frame proves heavier on the fps than wayland's CPU-free surface presentation pipeline.

In conclusion, T628 mp6 on wayland behaves like an entirely different class of GPU compared to the x11 presentation stack.

rooted wrote:Your video link is bad.

Sorry about that. Fixed.

by **memeka** » Thu Mar 09, 2017 1:14 pm

graphics performance is amazing in wayland indeed.
with a few issues to solve, gnome can become extremely usable.
there is elementary os that might be ported, since it's a fork of gnome: https://elementary.io
and also there is enlightenment, which i tried 6 months ago and is faster than gnome - https://www.youtube.com/watch?v=hct0BieTNrM

by **odroid** » Fri Mar 10, 2017 5:59 pm

Yes. I've played with this nice image for several hours.
It is quite impressive and snappy.
I didn't know I can run normal X11 application on gnome3-wayland.
I installed LibreOffice and Chromium 56 from the stock Stretchy ppa and they worked well.
It gave me a very nice user experience.

But sometimes it crashed the system and jumped to the login screen.
In worst case(when I kept playing youtube on Chromium) it rebooted due to the wrong DVFS/TMU in Kernel 4.9.
We will fix it soon.

by **memeka** » Fri Mar 10, 2017 7:12 pm

The crashes are issues with the Wayland driver - I've already reported to arm and they are looking into it.
The kernel in the image is different from the HK one, I did not have all thermal patches applied from hk repository - you might see different behavior with your kernel.

by **fleixi** » Sun Mar 12, 2017 6:29 am

Im trying to mount a sdcard with your image on my odroid-xu with an debian jessie running and get this errors:

mount /dev/sdc2 /mnt
mount: wrong fs type, bad option, bad superblock on /dev/sdc2,
missing codepage or helper program, or other error

In some cases useful info is found in syslog - try
dmesg | tail or so.

and dmesg:

[ 2276.489332] [c7] JBD2: Unrecognised features on journal
[ 2276.501409] [c7] EXT4-fs (sdc2): error loading journal

if i trying mount it on an other device with a ubuntu 16.10 i have no problems. Any idea?

by **memeka** » Sun Mar 12, 2017 6:58 am

fleixi wrote:Im trying to mount a sdcard with your image on my odroid-xu with an debian jessie running and get this errors:

mount /dev/sdc2 /mnt
mount: wrong fs type, bad option, bad superblock on /dev/sdc2,
missing codepage or helper program, or other error

In some cases useful info is found in syslog - try
dmesg | tail or so.

and dmesg:

[ 2276.489332] [c7] JBD2: Unrecognised features on journal
[ 2276.501409] [c7] EXT4-fs (sdc2): error loading journal

if i trying mount it on an other device with a ubuntu 16.10 i have no problems. Any idea?

probably the older kernel in the XU, the ext4 module is too old and does not recognize the ext4 from the sd. Even the XU4 Ubuntu 16.04 doesn't recognize all features (it complains about metadata_csum), although it can mount it.

by **blu** » Mon Mar 13, 2017 6:06 am

memeka wrote:probably the older kernel in the XU, the ext4 module is too old and does not recognize the ext4 from the sd. Even the XU4 Ubuntu 16.04 doesn't recognize all features (it complains about metadata_csum), although it can mount it.

IIRC, it takes e2fsprogs version 1.43.x or newer to either e2fsck or resize2fs hipster stretchy pants, for the metadata_csum alone.

by **fleixi** » Thu Mar 16, 2017 7:57 am

I didnt know ext4 isnt compatible to older versions. Thanks to clearyfy this.

I really like this image. No enduser cleanups but for developers its really fun. Thanks alot

I only still stuck getting mfc kodi running on wayland

by **memeka** » Thu Mar 16, 2017 8:10 am

fleixi wrote:I didnt know ext4 isnt compatible to older versions. Thanks to clearyfy this.

I really like this image. No enduser cleanups but for developers its really fun. Thanks alot

I only still stuck getting mfc kodi running on wayland

Yes, I released it for devs atm

pretty "raw"
Thank you so much for merging the patches for wayland kodi - i was just messaging @oversun yesterday to ask him his compile options since i wanted to start on that. so you took some work off my hands

Instead of getting kodi to work, I can look more closely now why encoding with FFmpeg 3.2.4 is not working as good as 3.0.2

-- i've added the patches to https://github.com/mihailescu2m/FFmpeg/ ... _m2m-3.2.4

Like @oversun said, he's gonna push some patches for kodi, seems that it didn't work on x11 either. So I think that will fix wayland too.

PS: interface-only, would you say are there any glitches? how's the CPU usage in GUI (comparatively to X11)?

by **fleixi** » Fri Mar 17, 2017 3:09 am

There a only glitches on really fast switching between submenus. it is running really smooth.
I have problems compare x11 kodi 16.2 with wayland kodi 18 Alpha since my skin (cirrus extend) isnt ported to Kodi 18.

But the hardest part in GUI was on x11 the coverflow in the Movie Section. On my old kodi i have had glitches but on wayland kodi 18 there a nearly no glitches.
The only way to get more framedrops is in pvr if i running a HD(720p) programm in background and i move to menus.

Subjective i would say it is better than on x11 Kernel 3.10

The CPU on htop is between 3% - 15% on Kodi idle and thats just small CPUs

by **fleixi** » Mon Mar 20, 2017 1:37 am

After a few days of testing and configuration i could use the system as my daily system (still some errors but nothing i couldnt deal with).

Now i like to know if you could upload your gnome-patches and kernel-source somewhere so i could keep working with your stuff?

by **memeka** » Mon Mar 20, 2017 10:33 am

@fleixi -- check second post.

For those that are unhappy with the fact that epiphany browser is slow/not working on some pages, let me remind you this is because the Wayland driver from ARM is missing the nesting compositing extension. I hope in a future release, this feature will be available, and hw-accelerated browsing will work correctly.

In the meantime, you can disable the non-working HW-accelerated rendering, so that pages will render correctly, by setting the environment variable WEBKIT_DISABLE_COMPOSITING_MODE. For example, starting epiphany with:

Code: Select all: WEBKIT_DISABLE_COMPOSITING_MODE=1 epiphany-browser

will allow you to visit pages like youtube.com and render them in software.

by **fleixi** » Sun Mar 26, 2017 6:24 pm

I have problems with the autologin function of gdm3. If i enable it the system isnt booting into the grafical enviroment. Saddly i dont find any log about it. Any idea

by **memeka** » Sun Mar 26, 2017 9:10 pm

fleixi wrote:I have problems with the autologin function of gdm3. If i enable it the system isnt booting into the grafical enviroment. Saddly i dont find any log about it. Any idea

did you update your system?
if you disable autologin, does gdm3 start?

by **fleixi** » Sun Mar 26, 2017 11:40 pm

I have upgrade my system 5 days ago and gdm is starting without error if autologin is disabled. if autologin is enabled there are a process of gdm in htop too but no WM or console is comming up on my tv. I could only connect via ssh.

by **memeka** » Mon Mar 27, 2017 10:15 am

fleixi wrote:I have upgrade my system 5 days ago and gdm is starting without error if autologin is disabled. if autologin is enabled there are a process of gdm in htop too but no WM or console is comming up on my tv. I could only connect via ssh.

please don't upgrade libmozjs. more recent versions of this package are compiled with more recent versions of gcc, which lead to the package crashing on armhf (probably missing some arm gcc flags... which would enable/disable newer gcc "optimizations").
gnome-shell (and gdm3) uses libmozjs, so this may lead to the GUI not working at all. if this happens, install libmozjs-24-0v5_24.2.0-3.1ubuntu1_armhf.deb (from ubuntu repo, it's the most recent version compiled with older gcc).

but since it works w/o autologin, this means you did not upgrade this package yet. i had a look, and indeed, autologin is failing. there are 2 causes:

1) you need to install libpam-gnome-keyring, such that gdm has access to the user password
2) gdm3 defaults to x session. when you login manually, you can select "gnome-wayland" session from the gear button, and this setting is saved in /var/lib/AccountsService/users/odroid (XSession=gnome-wayland). But for some reason, this setting is ignored when doing autologin, and gdm fails with: /usr/lib/gdm3/gdm-x-session[6053]: Unable to run X server. I've tried symlinking /usr/lib/gdm-x-session to gdm-wayland-session, and while it gets further along in the process, it still fails. I didn't have time enough to investigate, but this is obviously a hack, and there must be a way to set the default session to wayland....
You're not the only one with this issue, e.g. see http://unix.stackexchange.com/questions ... -autologin
There's a debian bug opened for this: https://bugs.debian.org/cgi-bin/bugrepo ... bug=851753

EDIT: found a way to make it work: remove /usr/share/gdm/BuiltInSessions/default.desktop and /usr/share/xsessions/gnome-classic.desktop -- this will force gdm to load gnome-wayland.desktop

by **memeka** » Thu Mar 30, 2017 9:22 am

Updated kernel to 4.9.18 can be found here: https://github.com/mihailescu2m/linux/c ... dxu3-4.9.y
Kernel config: https://transfer.sh/OWjmd/memeka.config

by **Cartridge** » Thu Mar 30, 2017 10:31 am

Is the image updated to all of these changes recently made? If not, are they available through apt-get upgrade? I really like wayland and I'd love to be able to install this and to easily update to all your changes you've made and will make in the future. There's a number of questions I've been wanting to ask since I've seen this while being at work today but I forgot them

by **memeka** » Thu Mar 30, 2017 11:27 am

Cartridge wrote:Is the image updated to all of these changes recently made? If not, are they available through apt-get upgrade? I really like wayland and I'd love to be able to install this and to easily update to all your changes you've made and will make in the future. There's a number of questions I've been wanting to ask since I've seen this while being at work today but I forgot them

The only updates are for the kernel, which is not available via apt, you need to build it.
But you can always use the HK kernel in this image, which you can update via apt.

I will soon make another release, based on HK kernel, with kodi, hw-encoder gstreamer/ffmpeg + emby server.

by **Cartridge** » Thu Mar 30, 2017 1:13 pm

Do you think Wayland is better solution to faster frames? I personally think it has great potential to surpass what we can already do. I tried the r10p0 drivers with the weston launch desktop and it did give good results somewhat close to original blobs. I can't wait to try this on, maybe try some flash aswell see how it would do in an intensive game.

https://www.phoronix.com/scan.php?page= ... -Chrome-50

by **memeka** » Thu Mar 30, 2017 2:37 pm

Cartridge wrote:Do you think Wayland is better solution to faster frames? I personally think it has great potential to surpass what we can already do. I tried the r10p0 drivers with the weston launch desktop and it did give good results somewhat close to original blobs. I can't wait to try this on, maybe try some flash aswell see how it would do in an intensive game.

https://www.phoronix.com/scan.php?page= ... -Chrome-50

There's no wayland support in chrome (only in ChromeOS, not regular linux).
You can try the image as it is now and see how fast wayland is. What do you mean "original blobs" ?

by **fleixi** » Fri Mar 31, 2017 1:25 am

Im a little bit confused. In the first post you linking to odroidxu4-4.9.y and now to odroidxu3-4.9.y and both are different commit stage. With one are the right (xu3 and xu4 using the same kernel right!?)

by **memeka** » Fri Mar 31, 2017 2:34 am

fleixi wrote:Im a little bit confused. In the first post you linking to odroidxu4-4.9.y and now to odroidxu3-4.9.y and both are different commit stage. With one are the right (xu3 and xu4 using the same kernel right!?)

Xu3 and xu4 are different branches, currently xu3 is more recent. For both branches I am using rebase, so git cannot merge after update, you need force

by **Cartridge** » Fri Mar 31, 2017 3:31 am

how come I can't mount the ext4 part to fix the fstab to be emmc? Ive rewritten the image 3 times and nothing. I can only access the fat partition... is there a way to change fstab at boot?

by **memeka** » Fri Mar 31, 2017 7:42 am

check the discussion few posts above: viewtopic.php?f=96&t=26172&p=184906#p182675
you need very new e2fsprogs packages installed to be able to read the ext4 parition.
i will take this into consideration for the next image and create a more compatible ext4 for older systems. also i will use UUID. since this is development version, i usually work with the same image on both emmc and sd-card, this is why i prefer to hardcode rootfs and not use uuid, since having the same uuid on sd-card and emmc is confusing the system...

to fix your issue:
i think you can boot with wrong fstab, it will dump you to root prompt, manually mount the ext4 partition, change fstab, and reboot. just make sure you modify boot.ini first.

by **Cartridge** » Fri Mar 31, 2017 8:02 am

Gotcha, im gettin the newer version from backports. Thanks for pointing me in the right direction! I've been looking left and right all day! XD

by **Cartridge** » Tue Apr 04, 2017 2:24 pm

Do you think i could get ahold of the linux headers? I need to recompile my wifi dongle source. Archer T2UH wifi. I'm lovin it so far.

by **memeka** » Tue Apr 04, 2017 3:49 pm

Cartridge wrote:Do you think i could get ahold of the linux headers? I need to recompile my wifi dongle source. Archer T2UH wifi. I'm lovin it so far.

see the post above: viewtopic.php?f=96&t=26172&p=185353#p184825 to compile your own kernel

by **Cartridge** » Tue Apr 04, 2017 9:20 pm

Well I just tried the most intensive flash game I could find and....? This is where I find it weird.
Without Wayland, and HW Acceleration enabled I got a laggier experience which came to a stop every now and then (a 6-7 second stop and maybe more at times)
With Wayland and Software Mode, Its faster! Reacts faster and the stopping is gone!
Not to say that before without wayland i did not have a good experience, au contraire, It was so good i kept it for years. But Wayland just makes it better.
So to say, software or not, It's good as it is right now quite surprisingly. I can't wait for more updates

by **fleixi** » Tue Apr 04, 2017 9:29 pm

I try to enable CEC but i dont find a /dev/CEC Devices.

I have allready tried adding setenv cecenable "true" to the boot.ini

What do in need to enable it

PS. I mean the oldfashend way with libcec not the new way you posted here: viewtopic.php?f=146&t=26490

by **memeka** » Tue Apr 04, 2017 10:09 pm

fleixi wrote:I try to enable CEC but i dont find a /dev/CEC Devices.

I have allready tried adding setenv cecenable "true" to the boot.ini

What do in need to enable it

PS. I mean the oldfashend way with libcec not the new way you posted here: viewtopic.php?f=146&t=26490

the original kernel in the image had no cec support because @mdrjr did not port it at the time
in the post above viewtopic.php?f=96&t=26172&p=184906#p184825 you can find the latest kernel and config, which include the legacy s5p kernel (this is one of the reasons it's in the xu3 branch, is using the legacy cec and not the official kernel one).
you can also try the HK kernel by adding the deb source and apt-get install the kernel (I think it will work but not 100% sure).

by **fleixi** » Wed Apr 05, 2017 1:23 am

I am already on the xu3 branch with kernel 4.9.18+ from your git with your config
At the moment i like to stay at your kernel source for wayland.

by **Cartridge** » Wed Apr 05, 2017 9:59 am

No sound? Tried some tutorials online since I installed every alsa file there is, pulseaudio aswell, tried adding stuff to /boot/config file with no success.
What am I doing wrong? LOL XD

by **memeka** » Wed Apr 05, 2017 10:11 am

fleixi wrote:I am already on the xu3 branch with kernel 4.9.18+ from your git with your config
At the moment i like to stay at your kernel source for wayland.

Then it should work, if you're using the config from that post, since the xu3 branch has this commit:

Code: Select all: s5p-cec: Add legacy s5p-cec driver … mdrjr committed with mihailescu2m on 24 Feb

and the config has
CONFIG_EXYNOS_HDMI_CEC=y

make sure your user has access to /dev/CEC...

Cartridge wrote:No sound? Tried some tutorials online since I installed every alsa file there is, pulseaudio aswell, tried adding stuff to /boot/config file with no success.
What am I doing wrong? LOL XD

the original kernel had no sound because it was before @mdrjr added the sound drivers

so you need the newer kernel from the source above

Also boot.ini has by default "dvi" mode for the monitor, which i think disables audio. Try commenting that out.

by **fleixi** » Wed Apr 05, 2017 7:51 pm

Sorry but there is no device /dev/CEC

Code: Select all: fleixi@odroid:~$ ls /dev/ autofs gpiochip31 network_throughput ptyc6 ptyp2 ptyre ptyua ptyx6 ram10 tty19 tty59 ttyc3 ttyef ttyrb ttytf ttywa ttyz6 block gpiochip32 null ptyc7 ptyp3 ptyrf ptyub ptyx7 ram11 tty2 tty6 ttyc4 ttyp0 ttyrc ttyu0 ttywb ttyz7 bsg gpiochip33 ppp ptyc8 ptyp4 ptys0 ptyuc ptyx8 ram12 tty20 tty60 ttyc5 ttyp1 ttyrd ttyu1 ttywc ttyz8 btrfs-control gpiochip34 psaux ptyc9 ptyp5 ptys1 ptyud ptyx9 ram13 tty21 tty61 ttyc6 ttyp2 ttyre ttyu2 ttywd ttyz9 bus gpiochip35 ptmx ptyca ptyp6 ptys2 ptyue ptyxa ram14 tty22 tty62 ttyc7 ttyp3 ttyrf ttyu3 ttywe ttyza cdrom gpiochip4 pts ptycb ptyp7 ptys3 ptyuf ptyxb ram15 tty23 tty63 ttyc8 ttyp4 ttys0 ttyu4 ttywf ttyzb cdrw gpiochip5 ptya0 ptycc ptyp8 ptys4 ptyv0 ptyxc ram2 tty24 tty7 ttyc9 ttyp5 ttyS0 ttyu5 ttyx0 ttyzc char gpiochip6 ptya1 ptycd ptyp9 ptys5 ptyv1 ptyxd ram3 tty25 tty8 ttyca ttyp6 ttys1 ttyu6 ttyx1 ttyzd console gpiochip7 ptya2 ptyce ptypa ptys6 ptyv2 ptyxe ram4 tty26 tty9 ttycb ttyp7 ttyS1 ttyu7 ttyx2 ttyze cpu_dma_latency gpiochip8 ptya3 ptycf ptypb ptys7 ptyv3 ptyxf ram5 tty27 ttya0 ttycc ttyp8 ttys2 ttyu8 ttyx3 ttyzf disk gpiochip9 ptya4 ptyd0 ptypc ptys8 ptyv4 ptyy0 ram6 tty28 ttya1 ttycd ttyp9 ttyS2 ttyu9 ttyx4 uinput dri gpiomem ptya5 ptyd1 ptypd ptys9 ptyv5 ptyy1 ram7 tty29 ttya2 ttyce ttypa ttys3 ttyua ttyx5 urandom dvb hwrng ptya6 ptyd2 ptype ptysa ptyv6 ptyy2 ram8 tty3 ttya3 ttycf ttypb ttyS3 ttyub ttyx6 v4l dvd i2c-2 ptya7 ptyd3 ptypf ptysb ptyv7 ptyy3 ram9 tty30 ttya4 ttyd0 ttypc ttys4 ttyuc ttyx7 vcs dvdrw i2c-4 ptya8 ptyd4 ptyq0 ptysc ptyv8 ptyy4 random tty31 ttya5 ttyd1 ttypd ttys5 ttyud ttyx8 vcs1 fb0 iio:device0 ptya9 ptyd5 ptyq1 ptysd ptyv9 ptyy5 rfkill tty32 ttya6 ttyd2 ttype ttys6 ttyue ttyx9 vcs2 fd initctl ptyaa ptyd6 ptyq2 ptyse ptyva ptyy6 rtc tty33 ttya7 ttyd3 ttypf ttys7 ttyuf ttyxa vcs3 full input ptyab ptyd7 ptyq3 ptysf ptyvb ptyy7 rtc0 tty34 ttya8 ttyd4 ttyq0 ttys8 ttyUSB0 ttyxb vcs4 fuse kmem ptyac ptyd8 ptyq4 ptyt0 ptyvc ptyy8 rtc1 tty35 ttya9 ttyd5 ttyq1 ttys9 ttyv0 ttyxc vcs5 gpiochip0 kmsg ptyad ptyd9 ptyq5 ptyt1 ptyvd ptyy9 sda tty36 ttyaa ttyd6 ttyq2 ttysa ttyv1 ttyxd vcs6 gpiochip1 log ptyae ptyda ptyq6 ptyt2 ptyve ptyya sda1 tty37 ttyab ttyd7 ttyq3 ttySAC0 ttyv2 ttyxe vcsa gpiochip10 loop0 ptyaf ptydb ptyq7 ptyt3 ptyvf ptyyb serial tty38 ttyac ttyd8 ttyq4 ttySAC1 ttyv3 ttyxf vcsa1 gpiochip11 loop1 ptyb0 ptydc ptyq8 ptyt4 ptyw0 ptyyc sg0 tty39 ttyad ttyd9 ttyq5 ttySAC2 ttyv4 ttyy0 vcsa2 gpiochip12 loop2 ptyb1 ptydd ptyq9 ptyt5 ptyw1 ptyyd sg1 tty4 ttyae ttyda ttyq6 ttySAC3 ttyv5 ttyy1 vcsa3 gpiochip13 loop3 ptyb2 ptyde ptyqa ptyt6 ptyw2 ptyye shm tty40 ttyaf ttydb ttyq7 ttysb ttyv6 ttyy2 vcsa4 gpiochip14 loop4 ptyb3 ptydf ptyqb ptyt7 ptyw3 ptyyf snd tty41 ttyb0 ttydc ttyq8 ttysc ttyv7 ttyy3 vcsa5 gpiochip15 loop5 ptyb4 ptye0 ptyqc ptyt8 ptyw4 ptyz0 spidev1.0 tty42 ttyb1 ttydd ttyq9 ttysd ttyv8 ttyy4 vcsa6 gpiochip16 loop6 ptyb5 ptye1 ptyqd ptyt9 ptyw5 ptyz1 sr0 tty43 ttyb2 ttyde ttyqa ttyse ttyv9 ttyy5 video0 gpiochip17 loop7 ptyb6 ptye2 ptyqe ptyta ptyw6 ptyz2 stderr tty44 ttyb3 ttydf ttyqb ttysf ttyva ttyy6 video1 gpiochip18 loop-control ptyb7 ptye3 ptyqf ptytb ptyw7 ptyz3 stdin tty45 ttyb4 ttye0 ttyqc ttyt0 ttyvb ttyy7 video10 gpiochip19 mali0 ptyb8 ptye4 ptyr0 ptytc ptyw8 ptyz4 stdout tty46 ttyb5 ttye1 ttyqd ttyt1 ttyvc ttyy8 video11 gpiochip2 mapper ptyb9 ptye5 ptyr1 ptytd ptyw9 ptyz5 sundtek tty47 ttyb6 ttye2 ttyqe ttyt2 ttyvd ttyy9 watchdog gpiochip20 mediainput0 ptyba ptye6 ptyr2 ptyte ptywa ptyz6 tty tty48 ttyb7 ttye3 ttyqf ttyt3 ttyve ttyya watchdog0 gpiochip21 mem ptybb ptye7 ptyr3 ptytf ptywb ptyz7 tty0 tty49 ttyb8 ttye4 ttyr0 ttyt4 ttyvf ttyyb zero gpiochip22 memory_bandwidth ptybc ptye8 ptyr4 ptyu0 ptywc ptyz8 tty1 tty5 ttyb9 ttye5 ttyr1 ttyt5 ttyw0 ttyyc gpiochip23 mmcblk0 ptybd ptye9 ptyr5 ptyu1 ptywd ptyz9 tty10 tty50 ttyba ttye6 ttyr2 ttyt6 ttyw1 ttyyd gpiochip24 mmcblk0boot0 ptybe ptyea ptyr6 ptyu2 ptywe ptyza tty11 tty51 ttybb ttye7 ttyr3 ttyt7 ttyw2 ttyye gpiochip25 mmcblk0boot1 ptybf ptyeb ptyr7 ptyu3 ptywf ptyzb tty12 tty52 ttybc ttye8 ttyr4 ttyt8 ttyw3 ttyyf gpiochip26 mmcblk0p1 ptyc0 ptyec ptyr8 ptyu4 ptyx0 ptyzc tty13 tty53 ttybd ttye9 ttyr5 ttyt9 ttyw4 ttyz0 gpiochip27 mmcblk0p2 ptyc1 ptyed ptyr9 ptyu5 ptyx1 ptyzd tty14 tty54 ttybe ttyea ttyr6 ttyta ttyw5 ttyz1 gpiochip28 mmcblk0rpmb ptyc2 ptyee ptyra ptyu6 ptyx2 ptyze tty15 tty55 ttybf ttyeb ttyr7 ttytb ttyw6 ttyz2 gpiochip29 mqueue ptyc3 ptyef ptyrb ptyu7 ptyx3 ptyzf tty16 tty56 ttyc0 ttyec ttyr8 ttytc ttyw7 ttyz3 gpiochip3 net ptyc4 ptyp0 ptyrc ptyu8 ptyx4 ram0 tty17 tty57 ttyc1 ttyed ttyr9 ttytd ttyw8 ttyz4 gpiochip30 network_latency ptyc5 ptyp1 ptyrd ptyu9 ptyx5 ram1 tty18 tty58 ttyc2 ttyee ttyra ttyte ttyw9 ttyz5

I have checked the /proc/config.gz and the module for CEC is enabled:

CONFIG_EXYNOS_HDMI_CEC=y

Do you have the device in /dev/?

by **memeka** » Wed Apr 05, 2017 8:39 pm

yes

Code: Select all: odroid@odroid:~/src/linux$ ls -lah /dev/CEC crwxrwxrwx 1 root root 10, 243 Nov 4 03:46 /dev/CEC odroid@odroid:~/src/linux$ uname -a Linux odroid 4.9.18 #1 SMP PREEMPT Thu Mar 30 11:29:33 ACDT 2017 armv7l GNU/Linux odroid@odroid:~/src/linux$ dmesg | grep -i cec [ 2.352162] of_get_named_gpiod_flags: parsed 'gpios' property of node '/cec@101B0000[0]' - status (0)

by **fleixi** » Wed Apr 05, 2017 8:51 pm

thats strange. Have you set something in the boot.ini or somewere else?

fleixi@odroid:~$ dmesg | grep -i cec
fleixi@odroid:~$

i dont know what i could have missing?