Re: Adding options INSECURE to GENERIC

Hello,

On Mon, 20 Apr 2020 07:05:58 +0200
Martin Husemann <[email protected]> wrote:

> On Sun, Apr 19, 2020 at 08:18:05PM -0400, Michael wrote:
> > That's for accessing IO space. I don't think any of the common PCI
> > graphics cards on sparc64 really need it, but the drivers may request
> > it because x86.
> > So there might be fallout. Trivial to fix fallout but still fallout.  
> 
> Sounds correct, here is an example:
> 
> radeonfb0 at pci4 dev 2 function 0: ATI Technologies Radeon 7000/VE QY (rev. 0x00)
> trying to read disabled BIOS...
> radeonfb0: No video BIOS, using default clocks
> radeonfb0: refclk = 27.000 MHz, refdiv = 12 minpll = 125000, maxpll = 400000
> no data for est. mode 640x480x67
> radeonfb0: 64 MB aperture at 0x08000000, 64 KB registers at 0x00100000
> 
> and IIRC mach64 is similar (with a mem type register window used instead
> of the PCish IO space).
> 
> So we could either #ifndef __sparc__ the IO access in xsrc, or leave it
> there, make it fail in the kernel, and have xsrc deal with failure. The
> latter would probably be easier to upstream.

I know radeon and mach64 do not need IO access, I'm about 99% sure
Permedia 2 doesn't need it either. That covers about 99.9% of all PCI
graphics card you'll find on sparc which we have drivers for ( as in,
not a wildcat )
Actually I'm fairly sure I added it only so we can use some Matrox
cards which the mga driver can cold boot.

have fun
Michael