xf86-video-intel: 2 commits - src/i830_driver.c src/i830_uxa.c

[Michel Dänzer]

On Wed, 2009-07-08 at 09:18 -0700, Eric Anholt wrote:
> 
> New commits:
> commit 40e7c9505265823786cf730214db84812a5e494e
> Author: Eric Anholt <eric at anholt.net>
> Date:   Mon Jul 6 11:54:50 2009 -0700
> 
>     Refuse to allocate giant BOs on 32-bit systems.
>     
>     The overcommit of address space combined with these buffers hitting SW
>     fallbacks all the time means that we're probably better off telling the
>     application "no" instead of likely silently failing later.
>     
>     Bug #22601.
> 
> diff --git a/src/i830_uxa.c b/src/i830_uxa.c
> index eb35014..ef7ac84 100644
> --- a/src/i830_uxa.c
> +++ b/src/i830_uxa.c
> @@ -627,6 +627,22 @@ i830_uxa_create_pixmap (ScreenPtr screen, int w, int h, int depth, unsigned usag
>  	     * with drm_intel_bufmgr_check_aperture().
>  	     */
>  	    size = i830_get_fence_size(i830, stride * h);
> +	    assert(size >= stride * h);
> +	}
> +
> +	/* Fail very large allocations on 32-bit systems.  Large BOs will
> +	 * tend to hit SW fallbacks frequently, and also will tend to fail
> +	 * to successfully map when doing SW fallbacks because we overcommit
> +	 * address space for BO access.
> +	 *
> +	 * Note that size should fit in 32 bits.  We throw out >32767x32767x4,
> +	 * and pitch alignment could get us up to 32768x32767x4.
> +	 */
> +	if (sizeof(unsigned int) == 4 &&
> [...]

FWIW, this test is true on 64-bit systems as well.


-- 
Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer