Hi,
Am 14.10.2016 um 12:26 schrieb aTc atc@k-n-p.org:
On 10/14/2016 07:38 AM, H. Nikolaus Schaller wrote:
Hi,
Am 14.10.2016 um 02:45 schrieb aTc atc@k-n-p.org:
Here's a test image with a tiler enabled 4.8.1 kernel, and an u-boot that show the extlinux menu( on the serial output) .
Oh nice! Will give it a try asap.
Would it be difficult to share the source code of the u-boot patches?
You can find them, and the tiler stuff here: http://packages.pyra-handheld.com/images/extra/tiler/
Thanks! I will try to integrate it into the Letux kernel tree. I have started to rebase to v4.9-rc1.
And about the patched u-boot, on my pyra the previous u-boot seemed to have glitched to do a saveenv for some reason ( it probably went into the fastboot stuff), which causes it not to pick up the new config. In that case you need to reset the env to its default (env default -a) and save it.
Yes, I remember that I had to do that once a while.
For your patch I am not sure if it is wise to hard encode the dtb file name. It should be set by uEnv.txt.
But before applying any changes I still have to debug the u-boot 2016.09 source tree I have created last week. It does not even run MLO. So something went wrong or missing during upgrade. Maybe a simple CONFIG.
The default kernel will load after a 3 second timeout
http://packages.pyra-handheld.com/images/extra/tiler/pyra-debian-jessie-mate...
either extract the .img, and dd to an sd card
well, that can fail if the target SD is a little smaller (1-2% variation of "official" 8GB are possible) than baked into the .img.
Yes, that's why I made the image 7GB , i've had similar problems with the rpi image that was just a few bytes larger than my sd card.
Ok, that is of course a workaround. If a resize is required/recommended anyways, the image could also be made as small as possible. I.e. just 20% bigger than the sum of all files.
I've got scripts to output the rootfs as either a .tar.xz/gz , and use a makesd type script to install it to sd, or as a full.img , which is much easier to install for an end user,
Yes, in the long run there should be some image created by applying makesd to a loopback device. On the other hand it is probably only ED and some developers who want to make fresh SD cards without a Pyra.
makesd runs fine on a Pyra and can pull together things on a SD card in the second slot.
I always dream to write a small GUI wrapper for makesd so that people can simply open some "SD card generator" app. This could even either download a full image for simple unpack or run makesd (or a successor).
But I have to many other things to do.
since it requires far less support tools, and doesn't have as many problems running on win/mac.
Well, makesd has indeed a big problem on Mac... It can't write ext2/3/4 without problems. There is MacFUSE/OSXFuse but writing ext file systems is discouraged because it is slow and buggy.
I don't really like the "write a full image to an sd card" method either though :)
Then, we at least agree in that :) My answer was to develop makesd...
I would like to include both, the Tiler kernel patches and this script in the Letux kernel tree so that we can build/modify that independently of user-space code and have a well known location where tiler kernel code and helper scripts are version-managed.
the tiler patches came from Matthijs's kernel at https://github.com/mvduin/linux/tree/work/merged
Thanks! I will check if and how I can update the code in http://git.goldelico.com/?p=gta04-kernel.git;a=shortlog;h=refs/heads/work/te... for 4.9-rc1.
so that the result should be the same as merging.
I just merged them into the newer 4.8.1. There are still a lot of problems and evil code in there. (biggest problem is currently that the screen turns off after 10 seconds if /dev/dri/card0 isn't kept open, which is what tiler.sh tries to do)
Ok. Something that can be fixed as soon as more hardware is available.
BR and thanks, Nikolaus