On Mon, Apr 09, 2012 at 11:30:34AM -0700, ron minnich wrote:
I'm trying an experiment and I welcome anyone who wants to help.
It's a simple idea motivated by a very hard problem. The problem is that we have unending problems with binary video bioses. We'd like to get to a source based video bios on, e.g., chromebook.
[...]
Now why do it this way? Why not just do a one time hack of the driver code to a coreboot driver?
One possibility would be to create your own VGA BIOS with the SeaVGABIOS code. This has been done for Geode vga. I'm not familiar with the i915 chipset, but there's a chance that you just need to copy the C initialization code over.
The vga bios isn't inherently slow - it's just poorly coded implementations that make them slow.
The reason is that the driver keeps changing and getting bug fixes. It most recently got some very useful fixes to the GMBUS driver. We want those fixes but the structure of the code changed a lot.
Not sure the above would help here. If the vga support is needed just for early debug printing, what's the gain in closely tracking upstream?
-Kevin
On Mon, Apr 16, 2012 at 5:06 PM, Kevin O'Connor kevin@koconnor.net wrote: This has been done for Geode vga. I'm not familiar
with the i915 chipset, but there's a chance that you just need to copy the C initialization code over.
no, it's quite complex, especially when you take the aux/ddc channel into account and all the other fun bits. It's not a simple framebuffer.
Not sure the above would help here. If the vga support is needed just for early debug printing, what's the gain in closely tracking upstream?
Because the upstream comes with bugs too, and we've seen them a lot in our systems.
ron
-
Kevin O'Connor -
ron minnich