Fatdog64 700 is released
Linux 3.18.7 and many enhancements and bug fixes over the 700rc release.Release notes
Forum announcement
Get it as usual from ibiblio or one of its mirrors: aarnet, uoc.gr, and nluug.nl.
No comments - Edit - Delete
Fatdog64 700rc Release
Linux 3.18.6, Seamonkey 2.32, flash player 11.2.202.442 and other bug fixes.Release notes
Forum announcement
Get it as usual from ibiblio or one of its mirrors: aarnet, uoc.gr, and nluug.nl.
No comments - Edit - Delete
Maximum size of initramfs
Well, I'm still around .Just want to document this often asked problem: what is the maximum size allowed, for an initramfs? Obviously this will be related to how much RAM you have, but what is exactly the relationship?
I have answered the question here, but I didn't explain why the numbers are what they are. This post attempts to correct that, if you're interested.
When considering this question, there are two things to remember:
a) initramfs will be uncompressed to memory during boot-up, so the amount of RAM needed would be the size of the initial initramfs + size of uncompressed initramfs
b) how much RAM the kernel will want to allocate for holding initramfs.
For the system to successfully boot, (a) must be the same or less than (b).
(a) is easy - by rule of thumb, it is twice the uncompressed size of initramfs. It is exact for uncompressed initramfs, and approximate for compressed initramfs; in either way it will give you the upper bound.
(b) is a bit more complicated. Traditionally the kernel will use "ramfs" for initramfs (that's why it is called init-ramfs ... ), and it will happily allocate all of your RAM for it, which is unwise, so you should at least subtract about 0.5GB from your RAM size if you want a usable system (unless your idea of fun is dealing with kernel OOM-killer that kills your important system services right after boot - not me).
But somewhere around kernel 3.x (circa 2012), the kernel merged "inittmpfs" patch from Rob Landley (of busybox and toybox and aboriginal Linux fame). It means that from that day onwards, initramfs will use "tmpfs" as opposed to "ramfs" by default (you can still switch back to using ramfs by specifying "rootfstype=ramfs" on your kernel command line). In case you're not familiar with the difference, "tmpfs" is "ramfs" with a size-limit (among others); and by default the limit is 50% of your RAM.
But wait, you say, doesn't "tmpfs" come with knobs to specify the maximum size? Yes, it does, but not at boot-time, no. Sure, you can re-mount it and changes it size, but we're talking here the situation during initramfs loading, *before* any code in initramfs itself is started ...
Rob actually has a patch to do exactly this, using "rootfsflags", but don't count on it going to mainline kernel soon.
But I digress. So, to come back to the subject, with "ramfs" you get all of your RAM (minus your 0.5GB), and with "inittmpfs" you get 50% of your RAM. This is (b). And since (a) is twice the size of your initramfs, it means that the maximum allowable size of your initramfs is 1/2 of (b).
Which means, if you use "inittmpfs" (which is the default in modern kernels), your initramfs size is limited to 1/2 of 50% of your RAM, or 25% of your RAM. If you use "ramfs" by specifying "rootfstype=ramfs" in your kernel boot command line, when this limit can increase to 1/2 of your RAM (minus 0.5GB), or, about 40%-45% of your RAM (depending on how big it is). Q.E.D.
1 Comment - Edit - Delete
jamesbond3142.no-ip.org no longer used.
I have relinquished control over the old domain of this blog and wiki, which was "jamesbond3142.no-ip.org". I have put a notification here but apparently I still have a few links pointing to the old domains.I am informed today that the old domain has been taken over and is being used to distribute malware. Please avoid that domain from now on.
Because of that, I have urgently changed all links that I know to point to the new domains. If you still know some old links to the no-longer used domain, please let me know and I will update them if it is under my control.
No comments - Edit - Delete
Fatdog64 700 beta2 is released
Linux 3.17.1 and other updates, fixes problem with missing fonts on opera/chrome and others.Release notes
Forum announcement
Get it as usual from ibiblio or one of its mirrors: aarnet, uoc.gr, and nluug.nl.
No comments - Edit - Delete
FatdogArm Beta2 sfs re-uploaded
The main change is that xorg-server is now built with these two patches: http://forum.odroid.com/viewtopic.php?p=37553&f=7#p37553 and http://forum.odroid.com/viewtopic.php?p=35833&f=7#p35833 for better performance and stability. Thanks Daniel (aka dsd). This change only improves Odroid, and has no impact in other platforms as only Odroid uses DRI2 interface at the moment (Allwinner: fbdev, OLPC/Marvell: DRI1, Cubox-i/Freescale: DRI1).Coupled with a few updates in the repo and youtube-dl, it is now possible to write a simple script that will stream video and play it fullscreen on Odroid/Cubox-i directly from youtube. Perhaps I'll share that script later.
3 Comments - Edit - Delete
3D hardware acceleration for cubox-i
I have finally managed to get 3D acceleration working on cubox-i. This has been long-time coming.Initial testing of the 3D acceleration always caused a hard lock-up on Xorg. System was still alive, and was still responsive (if you use SSH), but Xorg completey died, you can't even switch VT anymore.
Today I have finally found the solution - Freescale provided a patch for libdrm, *and* Xorg must be re-compiled against this patched libdrm. This was found as early as March by forum member "notwa", but I didn't take this seriously because:
a) it was intended for kernel 3.0.35 while I used 3.10.30 kernel
b) I actually patched libdrm but didn't rebuild xorg-server as I thought it was unnecessary (I reviewed xorg-server source to find where exactly the patched change could have impact, and couldn't see it).
Today I found that another forum member "catwich" had the same problem I had, and confirmed that "notwa" solution worked - if followed in full. I decided to have a go myself on FatdogArm, and look and behold - it worked! es2gears, glmark2 and glmark2-es2 all worked. What a joy!
The updated drivers will be in the repository as usual. As a closing note, I also built gst-plugins-gl (containing "glimagesink" - see my earlier post about hardware accelerated video playback pipeline), and it works well as an alternative to mfw_v4lsink.
With this, support for cubox-i platform is now considered complete (apart from bluetooth support, but that's minor stuff ... may be later). All these work with FatdogArm beta2. I updated the release note to reflect this fact but if you have already downloaded the SFS and the kernel package, there is no need to download again as there is no change there: all the drivers are in the repository and those are the ones that got updated.
No comments - Edit - Delete
Hardware-accelerated video playback on Odroid U2/U3 - part 2
Near the end of my previous post, I said that while the video playback was decent, it was not as smooth as expected, and that "memeka" from the odroid forum suggested to use cluttersink instead of glimagesink.I have now built and tested cluttersink, and indeed cluttersink shows smoother playback than glimagesink. There is one small difference: cluttersink doesn't respect aspect ratio, it will just scale up everything to its window size; while glimagesink does honour aspect ratio (but you can disable it if you don't need it).
Here is an example pipeline that plays an mp4 file, both audio and video, with full video acceleration using cluttersink:
gst-launch-1.0 filesrc location=/path/to/media.mp4 ! qtdemux name=m m. ! queue ! h264parse ! video/x-h264, stream-format=byte-stream, alignment=au ! v4l2video8dec ! v4l2video10convert ! cluttersink m. ! queue ! faad ! audioconvert ! alsasink
The same pipeline, simplified using decodebin:
gst-launch-1.0 filesrc location=/path/to/media.mp4 ! decodebin name=m m. ! queue ! v4l2video10convert ! cluttersink m. ! queue ! audioconvert ! autoaudiosink
cluttersink is available in "gst2-clutter" package in the repository.
I also re-uploaded Odroid U2/U3 kernel package for FatdogArm Beta2, now with MFC firmware (which is needed for hardware acceleration). Thanks to forum member "mories" who informed me that I forgot to do that. Beta2 sfs is also re-uploaded, now with updated glib (2.38.2 now, it was 2.36.3 previously) which is needed for cluttersink and its libraries, among other small fixes.
No comments - Edit - Delete
Hardware-accelerated video playback on Odroid U2/U3
Thanks to the effort of many people, I have uploaded gstreamer 1.x libraries (1.4.3) (gstreamer2 and gst2-plugins-* in the repository) which provide support for video playback acceleration on Odroid U2/U3 on FatdogArm.The accelerated pipeline in Odroid U2/U3 goes like this:
- hardware decoding (Samsung S5P MFC codec) - use v4l2video8dec
- hardware colourspace conversion (Samsung FIMC) - use v4l2video10convert
- GPU-assisted rendering - use glimagesink
There were a few hassles to finally get to this.
The hardware decoding was quite a breeze. The kernel has supported the MFC codec for a while. The userspace support was done by a gstreamer plugin called mfcdec, but this was finally deprecated and replaced with v4l2-based decoder in the most recent gstreamer version.
Hardware colourspace conversion didn't work due to a bug in libv4l2, as found out by forum member "memeka" in this post. The library basically assumes that all v4l2 devices can use DMABUF interface to v4l2 devices, but Samsung FIMC kernel drivers don't support that, only normal mmap-ed access. Disabling libv4l2 forces gstreamer to use its own internal code which works with either. So, following his advice, FatdogArm gstreamer libraries are built without libv4l2 dependency. However, the default colourspace converter for gst-play and playbin is unchanged and is still the software decoder (videoconvert) - because forcing the default to Samsung converter results in colour artifacts. If you need to use hardware colourspace converter, you will need to build the pipeline yourself (and if you do this yourself, there is no artifact - the artifact is only seen when using gst-playback/playbin with hardware converter, which is odd).
GPU-assisted rendering is done by glimagesink. This component replaces eglglessink in earlier gstreamer version. Due to a long-standing bug in Mali driver (documented here, found in r3p2 and still exist in r4p0), it doesn't always work (well - it *mostly* doesn't work on my tests). To make it work, I patched the binary driver (it's also in the repo as libmail-odroid).
The result of the acceleration is decent, but it is not as smooth as I hope. "memeka" in another forum post said that the best sink is cluttersink (better than glimagesink). We'll see about that later.
To test that it is all working, try this pipeline (it is one long line):
gst-launch-1.0 filesrc location=/path/to/media.mp4 typefind=true ! qtdemux name=m m.video_0 ! h264parse ! video/x-h264, stream-format=byte-stream, alignment=au ! v4l2video8dec ! v4l2video10convert ! glimagesink
This pipeline will only play the video and there is no buffering, but you get the idea.
No comments - Edit - Delete
FatdogArm Beta2 release
FatdogArm Beta2 is a minor release, mainly:- bugfixes and upstream update from Fatdog64 700
- updated kernel package for Odroid U2/U3 with r4p0 Mali driver
- introducing support for cubox-i (Freescale i.MX6DL/i.MX6Q SoC)
Release note here.
Get it from here or any of ibiblio mirrors.
No comments - Edit - Delete