EXA & vtSema

[Thomas Winischhofer]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Jon Smirl wrote:
> The fun rules of VT swap are that the swap can do anything it wants to
> the hardware. That includes resetting it, turning off the GPU, wiping
> VRAM, disabling it's PCI IO/MEM access, etc. When you get swapped back
> you have to recover from all of these things.
> 
> For example a framebuffer driver can not leave interrupts running
> across a VT swap. I found out the hard way (my machine locked on VT
> swap) that X is going to turn off the PCI IO/MEM access and my ISR
> then isn't able to ack the interrupt because it can't write to the
> registers.
> 
> You can get away with not restoring some things on a swap. That's
> because there are some unwritten agreements between various pairs of X
> and fbdev drivers to not stomp on certain registers or parts of
> memory.
> 

Erm.. yes :) But all that doesn't change the mere fact that X with EXA
running is accessing video ram and eventually calling acceleration
routines while "swapped away" as you put it....

EXA needs a complete overhaul with regard to NOT access video memory and
NOT access driver-provided acceleration functions while vtSema is FALSE.
Just checking vtSema in the acceleration routines doesn't help, because
the software fallbacks will access the video RAM in any case then.

- --
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net	       *** http://www.winischhofer.net
twini AT xfree86 DOT org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFCy+FpzydIRAktyUcRAg/WAJ43NIeWYnR8uHJdGlOS5pIdPnpFWgCghQFX
I0PYL6rIaSoxP3jkNY7XqYE=
=zAs0
-----END PGP SIGNATURE-----