Re: XFree86: ati driver problem

Marc Aurele La France <tsi@xxxxxxxxxxx> · Fri, 1 Feb 2008 13:57:32 -0700 (Mountain Standard Time)

On Fri, 1 Feb 2008, Anton 'FaioN' Mayorov wrote:

Marc Aurele La France wrote:

I'm not sure if this bug is connected with XFree86 at all. There are 

several bits of information that it is a kernel bug. It seems to me that 

affected kernel versions are 2.6.22 and 2.6.23. Check this:

http://lkml.org/lkml/2007/6/21/216 .

atyfb has __ALWAYS__ been broken, especially for laptops or adapters not 

initialised by BIOS/firmware.  And it is unfixable primarily because its 

maintainers over the years don't really know what they're doing.  That 

sounds harsh, but it is true.  On ix86, I recommend you use vesafb instead, 

or (better) no kernel framebuffer at all.

Sorry, I've rather misled you. I don't use atyfb at all.

However, now that you have some kind of display, please send me a 

"-logverbose 5" log, and I'll see what I can do to mitigate atyfb's latest 

brain damage.

"XFree86 -logverbose 5" lock my system entirely. Even SysRq magic doesn't 

work. I've failed to trace the reason out.

BTW, I've managed to synchronize my screen. I've overridden pATI->LCDClock 

and used "Option "LCDSync"" in XF86Config (although I don't know for certain 

what it means). It seems to work, but XFree86 still corrupts console mode 

(i.e. after X server shutdown, all local ttys are in unusable state). And I 

haven't found the reason why clock values are set wrong.

And one more question. This is the part of atimode.c:
-----------------------------------------------------------------
static void
ATISwap
(
   int      iScreen,
   ATIPtr   pATI,
   ATIHWPtr pATIHW,
   Bool     ToFB
)
{
   pointer save, *from, *to;
   unsigned int iPlane = 0, PlaneMask = 1;
   CARD8 seq2, seq4, gra1, gra3, gra4, gra5, gra6, gra8;

   /*
    * This is only done for non-accelerator modes.  If the video state on

    * server entry was an accelerator mode, the application that 

relinquished

    * the console had better do the Right Thing (tm) anyway by saving and
    * restoring its own video memory contents.
    */
   if (pATIHW->crtc != ATI_CRTC_VGA) 	        return;
-----------------------------------------------------------------

I don't use kernel framebuffer, so I think my console mode isn't accelerated. 

But at this point, pATIHW->crtc is ATI_CRTC_MACH64. May it be the reason that 

my console is not restored on exit? Is this the only method to restore video 

state?

That pATIHW->crtc == ATI_CRTC_MACH64 at that point is proof positive that 

you are lying to me WRT to not using atyfb.

Marc.

+----------------------------------+----------------------------------+
|  Marc Aurele La France           |  work:   1-780-492-9310          |
|  Academic Information and        |  fax:    1-780-492-1729          |
|    Communications Technologies   |  email:  tsi@xxxxxxxxxxx         |
|  352 General Services Building   +----------------------------------+
|  University of Alberta           |                                  |
|  Edmonton, Alberta               |    Standard disclaimers apply    |
|  T6G 2H1                         |                                  |
|  CANADA                          |                                  |
+----------------------------------+----------------------------------+
_______________________________________________
XFree86 mailing list
XFree86@xxxxxxxxxxx
http://XFree86.Org/mailman/listinfo/xfree86