Vivante x11 support (xserver-xorg-video-imx-viv) 3.10.17-1.0.0 GA

[ro...@tbi.univie.ac.at]

Hi all,

I just updated my wandboard quad to kernel 3.10.17 GA release (the branch from github.com/wandboard-org/linux) and installed the appropriate x11 binaries provided by freescale. Since I wanted to test the x11 support, I also compiled the vivante EXA driver for 3.10.17 GA release which is used in Yocto. Using the xorg EXA driver, however, was not possible, because it constantly allocated to few fb memory to hold the shadow buffer which resulted in the following error message:

 (EE) VIVANTE(0): fb memory is not big enough to hold shadow buffer!

Here is my Xorg.log: http://pastebin.com/iSHNxgET

After crawling through the differences between the 3.10.17-1.0.0_beta and 3.10.17-1.0.0 GA release I found that the framebuffers virtual size is set differently in GA (it seems to need twice the y-size to properly allocate double the fb memory of one regular screen). So I just patched the beta code part of this into the GA sources. Here is my patch: http://pastebin.com/AENk7x4T

Now, xorg starts fine without complaining about framebuffer memory size: http://pastebin.com/Je88d16P

So the question now is, is there another way to get the x11 driver working? Do I have to set some weird stuff in my xorg.conf or somewhere else?

Also, I get some rendering artefacts in chromium using vivante x11 driver. The browser window does not always refresh properly upon scrolling so that a single line of text is repeated multiple times filling the whole window. Selecting the window content then issues a refresh and the content is drawn correctly again.

Best

Ronny

[Chiyosuke Fujii]

Hi Ronny,

can you try to add fbmem phrase to u-boot bootargs like this:

 video=mxcfb0:dev=hdmi,1920x1080M@60,if=RGB24,bpp=32 fbmem=28M

I hope it will work...

Chris

2014-05-10 17:36 GMT+09:00 <ro...@tbi.univie.ac.at>:

Hi all,

I just updated my wandboard quad to kernel 3.10.17 GA release (the branch from github.com/wandboard.org/linux) and installed the appropriate x11 binaries provided by freescale. Since I wanted to test the x11 support, I also compiled the vivante EXA driver for 3.10.17 GA release which is used in Yocto. Using the xorg EXA driver, however, was not possible, because it constantly allocated to few fb memory to hold the shadow buffer which resulted in the following error message:

--
You received this message because you are subscribed to the Google Groups "Wandboard" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wandboard+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ro...@tbi.univie.ac.at]

I already tried this without luck.

I tested several combinations of fbmem (28, 48, 96M), vmalloc etc. (see my first Xorg.log)

They all work fine in taking away RAM from the system, but the EXA driver is the actual problem. As soon as the EXA driver sets the virtual size of the framebuffer, its memory gets reallocated.

The problem also does not depend on the actual fb size. I checked 640x480, 1280x720 with 16bpp and 32bpp. All do not work unless I apply the patch mentioned in my posting above.