[PATCH v6 2/2] drm/xe: Align all VRAM scanout buffers to 64k physical pages when needed.

[Matthew Brost]

On Tue, Aug 27, 2024 at 08:43:36AM +0200, Maarten Lankhorst wrote:
> Hey,
> 
> Den 2024-08-27 kl. 05:11, skrev Matthew Brost:
> > On Mon, Aug 26, 2024 at 09:42:54PM +0200, Maarten Lankhorst wrote:
> >> Hey,
> >>
> >> Den 2024-08-26 kl. 21:30, skrev Matthew Brost:
> >>> On Mon, Aug 26, 2024 at 07:01:16PM +0200, Maarten Lankhorst wrote:
> >>>> For CCS formats on affected platforms, CCS can be used freely, but
> >>>> display engine requires a multiple of 64k physical pages. No other
> >>>> changes are needed.
> >>>>
> >>>> At the BO creation time we don't know if the BO will be used for CCS
> >>>> or not. If the scanout flag is set, and the BO is a multiple of 64k,
> >>>> we take the safe route and force the physical alignment of 64k pages.
> >>>>
> >>>> If the BO is not a multiple of 64k, or the scanout flag was not set
> >>>> at BO creation, we reject it for usage as CCS in display. The physical
> >>>> pages are likely not aligned correctly, and this will cause corruption
> >>>> when used as FB.
> >>>>
> >>>> The scanout flag and size being a multiple of 64k are used together
> >>>> to enforce 64k physical placement.
> >>>>
> >>>> VM_BIND is completely unaffected, mappings to a VM can still be aligned
> >>>> to 4k, just like for normal buffers.
> >>>>
> >>>> Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> >>>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> >>>> Cc: Matthew Auld <matthew.auld at intel.com>
> >>>> Cc: Rodrigo Vivi <rodrigo.vivi at intel.com>
> >>>> Cc: Thomas Hellström <thomas.hellstrom at linux.intel.com>
> >>>> Cc: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> >>>> Cc: Juha-Pekka Heikkilä <juha-pekka.heikkila at intel.com>
> >>>> ---
> >>>>  drivers/gpu/drm/xe/display/intel_fb_bo.c |  9 +++++++++
> >>>>  drivers/gpu/drm/xe/xe_bo.c               |  7 +++++++
> >>>>  drivers/gpu/drm/xe/xe_vm.c               | 11 ++++++++++-
> >>>>  3 files changed, 26 insertions(+), 1 deletion(-)
> >>>>
> >>>> diff --git a/drivers/gpu/drm/xe/display/intel_fb_bo.c b/drivers/gpu/drm/xe/display/intel_fb_bo.c
> >>>> index f835492f73fb4..63ce97cc4cfef 100644
> >>>> --- a/drivers/gpu/drm/xe/display/intel_fb_bo.c
> >>>> +++ b/drivers/gpu/drm/xe/display/intel_fb_bo.c
> >>>> @@ -7,6 +7,7 @@
> >>>>  #include <drm/ttm/ttm_bo.h>
> >>>>  
> >>>>  #include "intel_display_types.h"
> >>>> +#include "intel_fb.h"
> >>>>  #include "intel_fb_bo.h"
> >>>>  #include "xe_bo.h"
> >>>>  
> >>>> @@ -28,6 +29,14 @@ int intel_fb_bo_framebuffer_init(struct intel_framebuffer *intel_fb,
> >>>>  	struct xe_device *xe = to_xe_device(bo->ttm.base.dev);
> >>>>  	int ret;
> >>>>  
> >>>> +	/*
> >>>> +	 * Some modifiers require physical alignment of 64KiB VRAM pages;
> >>>> +	 * require that the BO in those cases is created correctly.
> >>>> +	 */
> >>>> +	if (XE_IOCTL_DBG(xe, intel_fb_needs_64k_phys(mode_cmd->modifier[0]) &&
> >>>> +			     !(bo->flags & XE_BO_FLAG_NEEDS_64K)))
> >>>> +		return -EINVAL;
> >>>
> >>> I don't think this is correct use of this macro as XE_BO_FLAG_NEEDS_64K
> >>> is an internal flag we set and typically this macro is used to santize
> >>> user input. An assert here or WARN would make more sense.
> >> Ideally we'd use 'is bo created as scanout', but that flag can be set by fb_init too, so if the BO was used for normal 4-tiled before, then as CCS it would pass when it wouldn't be valid.
> >>
> >> I could change it to bo_created_with_scanout_flag_on_64k_platform inline, but I doubt that's more readable. :)
> >>
> > 
> > Not trying to block the patch and really don't know anything about
> > display but still think XE_IOCTL_DBG should replaced by either an
> > assert or WARN (or Xe flavor of warn). Kinda pedantic but we really are
> > trying hard to uniformly use these types of macros and this just doesn't
> > look correct.
> 
> mode_cmd->modifier[0] is passed from userspace without validation, and this function is called very early on in fb creation. Anything more than XE_IOCTL_DBG would be invalid here.

Ok, that makes this usage more clear then. Fine with it as is then. Sorry for the noise.

Matt