Re:

[Dave Airlie]

On Tue, 23 Oct 2018 at 16:13, Michael Tirado <mtirado418@xxxxxxxxx> wrote:
>
> That preprocessor define worked but I'm still confused about this
> DRM_FILE_PAGE_OFFSET thing.  Check out drivers/gpu/drm/drm_gem.c
> right above drm_gem_init.
>
> ---
>
> /*
>  * We make up offsets for buffer objects so we can recognize them at
>  * mmap time.
>  */
>
> /* pgoff in mmap is an unsigned long, so we need to make sure that
>  * the faked up offset will fit
>  */
>
> #if BITS_PER_LONG == 64
> #define DRM_FILE_PAGE_OFFSET_START ((0xFFFFFFFFUL >> PAGE_SHIFT) + 1)
> #define DRM_FILE_PAGE_OFFSET_SIZE ((0xFFFFFFFFUL >> PAGE_SHIFT) * 16)
> #else
> #define DRM_FILE_PAGE_OFFSET_START ((0xFFFFFFFUL >> PAGE_SHIFT) + 1)
> #define DRM_FILE_PAGE_OFFSET_SIZE ((0xFFFFFFFUL >> PAGE_SHIFT) * 16)
> #endif
>
>
> ---
>
> Why is having a 64-bit file offsets critical, causing -EINVAL on mmap?
> What problems might be associated with using (0x10000000UL >>
> PAGE_SHIFT) ?

a) it finds people not using the correct userspace defines. mostly
libdrm should handle this,
and possibly mesa.

b) there used to be legacy maps below that address on older drivers,
so we decided to never put stuff in the first 32-bit range that they
could clash with.

Dave.