On Sunday, September 21, 2003, at 11:49 AM, Steve Gehlbach wrote:
SONE Takeshi wrote:
On Sat, Sep 20, 2003 at 11:25:58PM -0700, Steve Gehlbach wrote:
Nathanael Noblet wrote:
I don't need a framebuffer on boot, but once loaded would I be able to insmod the epiafb module and get that working?
More than likely, since a framebuffer driver made for epia vga must have intimate knowledge of the chipset, so it should completely initialize it.
Unfortunately it does not. Looks like we have to program some more VGA registers for framebuffer to function properly.
Hmmm... this is strange. Has anyone used HAVE_FRAMEBUFFER option without VIDEO_CONSOLE and gotten that to work? The only difference is the 0xFB chipset register, however, the setting I use matches the Award setting when it boots. The HAVE_FRAMEBUFER sets the framebuffer size to something different, but I don't think this is releveant, my setting would be 8M but the frambuffer access bit (bit 3) is off, and the address of the framebuffer is not set (bits 2-0 and reg 0xFA and others).
I suspect it has something to do with the graphics aperture settings (GA). The GA relevant registers are 0x13 and 0x80-0x8B. You can try putting in the settings at the bottom of afteram.inc (this file is not used in my config), the CS_WRITE commands to 0x13, 0x84, 0x80, 0x88. These match the Award settings, except for 0x84, which is 0 in Award (256M aperture) rather than 0xc0 (64M aperture). I know that these registers can cause hangs, though. I would try putting these in the vgainit.inc file replacing the 0x88 setting where it says "turn off GA".
Which epiafb are you using? Maybe I can try it later this week, but I am not a big user of framebuffers.
I've used the epiafb (epiafb.sf.net) and I'm currently working on getting the tridentfb to work. I am noticing some kernel oops? Though the system is still functional, but the vga console is garbled.
I have yet to get the epiafb working or even try the tridentfb with a linuxbios boot. The epia complains of unresolved dependancies which look like missing symbols for the generic framebuffer, even though that is loaded... so I'm not sure what is going on there.
I have been taking lspci -xxx output of each boot, normal boot, filo boot, filo with vgabios extracted boot. I have noticed differences in the register settings between the two. Where do you get the CS_WRITE and register information about this board? And more importantly, can I have a copy?
I have yet to get your changes to work though, mainly because of the difference in the payloads, though that is on my list of getting working. I'd like to see from there how it goes. I haven't looked much at the new code you supplied a diff for, but will be poking around there, though a spec sheet as always would be nice.
Thanks so far for the work and information...