Re: [Dri-devel] mgaSwapBuffers() xserver lock 2

[Gareth H. <ga...@va...>]

Gareth Hughes wrote:
> 
> Mike Lincoln wrote:
> >
> > OK I've changed :-
> >
> > mmesa->status_offset = mmesa->mgaScreen->primary.handle;
> >
> > to
> >
> > mmesa->status_offset = mmesa->mgaScreen->status.handle;
> >
> > in mga_xmesa.c, XMesaCreateContext(...)
> >
> > And  now my Xserver runs fine.  I need to remove all the debuging code
> > and test some more because it could just be a timing thing, but it
> > *seems* to have made things better for me. Either way its time to go to
> > bed.

Actually, this is wrong.  I'm undoing the change in CVS now.

The card writes out the current primary DMA pointer to the status page. 
This is a full 32-bit address, and so we need to subtract the primary
buffer address from the status value to get the offset from the start of
primary DMA space.  Changing the offset to a regular PCI address instead
of an AGP address will basically make the test always pass.  If this is
working for you (ie. not locking up), then we need to work out why the
status page value isn't being updated correctly.

-- Gareth