diff mbox series

[v5,1/2] drm/xe: Align all VRAM scanout buffers to 64k physical pages when needed.

Message ID 20240826095041.75416-2-maarten.lankhorst@linux.intel.com (mailing list archive)
State New, archived
Headers show
Series drm/xe: Align all VRAM scanout buffers to 64k physical pages when needed. | expand

Commit Message

Maarten Lankhorst Aug. 26, 2024, 9:50 a.m. UTC
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: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Co-developed-by: Zbigniew Kempczyński <zbigniew.kempczynski@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Juha-Pekka Heikkilä <juha-pekka.heikkila@intel.com>
---
 drivers/gpu/drm/xe/display/intel_fb_bo.c |  5 +++++
 drivers/gpu/drm/xe/xe_bo.c               |  7 +++++++
 drivers/gpu/drm/xe/xe_vm.c               | 11 ++++++++++-
 3 files changed, 22 insertions(+), 1 deletion(-)

Comments

Zbigniew Kempczyński Aug. 26, 2024, 1:44 p.m. UTC | #1
On Mon, Aug 26, 2024 at 11:50:40AM +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.

For me patch and explanation looks correct. According to submitting
patches rules please add my s-o-b before the merge.

--
Zbigniew


> 
> Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Co-developed-by: Zbigniew Kempczyński <zbigniew.kempczynski@intel.com>
> Cc: Matthew Auld <matthew.auld@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Cc: Juha-Pekka Heikkilä <juha-pekka.heikkila@intel.com>
> ---
>  drivers/gpu/drm/xe/display/intel_fb_bo.c |  5 +++++
>  drivers/gpu/drm/xe/xe_bo.c               |  7 +++++++
>  drivers/gpu/drm/xe/xe_vm.c               | 11 ++++++++++-
>  3 files changed, 22 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..6775c2557b9df 100644
> --- a/drivers/gpu/drm/xe/display/intel_fb_bo.c
> +++ b/drivers/gpu/drm/xe/display/intel_fb_bo.c
> @@ -28,6 +28,11 @@ int intel_fb_bo_framebuffer_init(struct intel_framebuffer *intel_fb,
>  	struct xe_device *xe = to_xe_device(bo->ttm.base.dev);
>  	int ret;
>  
> +	/* Only this specific format is affected, and it's only available on VRAM */
> +	if (XE_IOCTL_DBG(xe, mode_cmd->modifier[0] == I915_FORMAT_MOD_4_TILED_BMG_CCS &&
> +			     !(bo->flags & XE_BO_FLAG_NEEDS_64K)))
> +		return -EINVAL;
> +
>  	xe_bo_get(bo);
>  
>  	ret = ttm_bo_reserve(&bo->ttm, true, false, NULL);
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index cbe7bf098970f..41297b5797173 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -2019,6 +2019,13 @@ int xe_gem_create_ioctl(struct drm_device *dev, void *data,
>  
>  	bo_flags |= args->placement << (ffs(XE_BO_FLAG_SYSTEM) - 1);
>  
> +	/* CCS formats need physical placement at a 64K alignment in VRAM. */
> +	if ((bo_flags & XE_BO_FLAG_VRAM_MASK) &&
> +	    (bo_flags & XE_BO_FLAG_SCANOUT) &&
> +	    !(xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K) &&
> +	    !(args->size % SZ_64K))
> +		bo_flags |= XE_BO_FLAG_NEEDS_64K;
> +
>  	if (args->flags & DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM) {
>  		if (XE_IOCTL_DBG(xe, !(bo_flags & XE_BO_FLAG_VRAM_MASK)))
>  			return -EINVAL;
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 4cc13eddb6b32..3eb76d874eb28 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -2878,7 +2878,16 @@ static int xe_vm_bind_ioctl_validate_bo(struct xe_device *xe, struct xe_bo *bo,
>  		return -EINVAL;
>  	}
>  
> -	if (bo->flags & XE_BO_FLAG_INTERNAL_64K) {
> +	/*
> +	 * Some platforms require 64k VM_BIND alignment,
> +	 * specifically those with XE_VRAM_FLAGS_NEED64K.
> +	 *
> +	 * Other platforms may have BO's set to 64k physical placement,
> +	 * but can be mapped at 4k offsets anyway. This check is only
> +	 * there for the former case.
> +	 */
> +	if ((bo->flags & XE_BO_FLAG_INTERNAL_64K) &&
> +	    (xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)) {
>  		if (XE_IOCTL_DBG(xe, obj_offset &
>  				 XE_64K_PAGE_MASK) ||
>  		    XE_IOCTL_DBG(xe, addr & XE_64K_PAGE_MASK) ||
> -- 
> 2.45.2
>
Thomas Hellstrom Aug. 26, 2024, 2:01 p.m. UTC | #2
On Mon, 2024-08-26 at 11:50 +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: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Co-developed-by: Zbigniew Kempczyński
> <zbigniew.kempczynski@intel.com>
> Cc: Matthew Auld <matthew.auld@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Cc: Juha-Pekka Heikkilä <juha-pekka.heikkila@intel.com>
> ---
>  drivers/gpu/drm/xe/display/intel_fb_bo.c |  5 +++++
>  drivers/gpu/drm/xe/xe_bo.c               |  7 +++++++
>  drivers/gpu/drm/xe/xe_vm.c               | 11 ++++++++++-
>  3 files changed, 22 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..6775c2557b9df 100644
> --- a/drivers/gpu/drm/xe/display/intel_fb_bo.c
> +++ b/drivers/gpu/drm/xe/display/intel_fb_bo.c
> @@ -28,6 +28,11 @@ int intel_fb_bo_framebuffer_init(struct
> intel_framebuffer *intel_fb,
>  	struct xe_device *xe = to_xe_device(bo->ttm.base.dev);
>  	int ret;
>  
> +	/* Only this specific format is affected, and it's only
> available on VRAM */

A first time reader would wonder "Affected by what". Could we rephrase
like "This specific format, only available with DGFX needs .." or
something similar self-contained?

> +	if (XE_IOCTL_DBG(xe, mode_cmd->modifier[0] ==
> I915_FORMAT_MOD_4_TILED_BMG_CCS &&
> +			     !(bo->flags & XE_BO_FLAG_NEEDS_64K)))
> +		return -EINVAL;
> +
>  	xe_bo_get(bo);
>  
>  	ret = ttm_bo_reserve(&bo->ttm, true, false, NULL);
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index cbe7bf098970f..41297b5797173 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -2019,6 +2019,13 @@ int xe_gem_create_ioctl(struct drm_device
> *dev, void *data,
>  
>  	bo_flags |= args->placement << (ffs(XE_BO_FLAG_SYSTEM) - 1);
>  
> +	/* CCS formats need physical placement at a 64K alignment in
> VRAM. */
> +	if ((bo_flags & XE_BO_FLAG_VRAM_MASK) &&
> +	    (bo_flags & XE_BO_FLAG_SCANOUT) &&
> +	    !(xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K) &&
> +	    !(args->size % SZ_64K))

This might probably fail on 32-bit compiles? Can we use the IS_ALIGNED
macro?


> +		bo_flags |= XE_BO_FLAG_NEEDS_64K;
> +
>  	if (args->flags & DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM)
> {
>  		if (XE_IOCTL_DBG(xe, !(bo_flags &
> XE_BO_FLAG_VRAM_MASK)))
>  			return -EINVAL;
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 4cc13eddb6b32..3eb76d874eb28 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -2878,7 +2878,16 @@ static int xe_vm_bind_ioctl_validate_bo(struct
> xe_device *xe, struct xe_bo *bo,
>  		return -EINVAL;
>  	}
>  
> -	if (bo->flags & XE_BO_FLAG_INTERNAL_64K) {
> +	/*
> +	 * Some platforms require 64k VM_BIND alignment,
> +	 * specifically those with XE_VRAM_FLAGS_NEED64K.
> +	 *
> +	 * Other platforms may have BO's set to 64k physical
> placement,
> +	 * but can be mapped at 4k offsets anyway. This check is
> only
> +	 * there for the former case.
> +	 */
> +	if ((bo->flags & XE_BO_FLAG_INTERNAL_64K) &&
> +	    (xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)) {
>  		if (XE_IOCTL_DBG(xe, obj_offset &
>  				 XE_64K_PAGE_MASK) ||
>  		    XE_IOCTL_DBG(xe, addr & XE_64K_PAGE_MASK) ||

Otherwise LGTM.
/Thomas
Maarten Lankhorst Aug. 26, 2024, 2:03 p.m. UTC | #3
Hey,

Den 2024-08-26 kl. 16:01, skrev Thomas Hellström:
> On Mon, 2024-08-26 at 11:50 +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: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
>> Co-developed-by: Zbigniew Kempczyński
>> <zbigniew.kempczynski@intel.com>
>> Cc: Matthew Auld <matthew.auld@intel.com>
>> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
>> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
>> Cc: Juha-Pekka Heikkilä <juha-pekka.heikkila@intel.com>
>> ---
>>  drivers/gpu/drm/xe/display/intel_fb_bo.c |  5 +++++
>>  drivers/gpu/drm/xe/xe_bo.c               |  7 +++++++
>>  drivers/gpu/drm/xe/xe_vm.c               | 11 ++++++++++-
>>  3 files changed, 22 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..6775c2557b9df 100644
>> --- a/drivers/gpu/drm/xe/display/intel_fb_bo.c
>> +++ b/drivers/gpu/drm/xe/display/intel_fb_bo.c
>> @@ -28,6 +28,11 @@ int intel_fb_bo_framebuffer_init(struct
>> intel_framebuffer *intel_fb,
>>  	struct xe_device *xe = to_xe_device(bo->ttm.base.dev);
>>  	int ret;
>>  
>> +	/* Only this specific format is affected, and it's only
>> available on VRAM */
> 
> A first time reader would wonder "Affected by what". Could we rephrase
> like "This specific format, only available with DGFX needs .." or
> something similar self-contained?
Yeah, I will reorder with next patch, then this comment goes away.

>> +	if (XE_IOCTL_DBG(xe, mode_cmd->modifier[0] ==
>> I915_FORMAT_MOD_4_TILED_BMG_CCS &&
>> +			     !(bo->flags & XE_BO_FLAG_NEEDS_64K)))
>> +		return -EINVAL;
>> +
>>  	xe_bo_get(bo);
>>  
>>  	ret = ttm_bo_reserve(&bo->ttm, true, false, NULL);
>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>> index cbe7bf098970f..41297b5797173 100644
>> --- a/drivers/gpu/drm/xe/xe_bo.c
>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>> @@ -2019,6 +2019,13 @@ int xe_gem_create_ioctl(struct drm_device
>> *dev, void *data,
>>  
>>  	bo_flags |= args->placement << (ffs(XE_BO_FLAG_SYSTEM) - 1);
>>  
>> +	/* CCS formats need physical placement at a 64K alignment in
>> VRAM. */
>> +	if ((bo_flags & XE_BO_FLAG_VRAM_MASK) &&
>> +	    (bo_flags & XE_BO_FLAG_SCANOUT) &&
>> +	    !(xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K) &&
>> +	    !(args->size % SZ_64K))
> 
> This might probably fail on 32-bit compiles? Can we use the IS_ALIGNED
> macro?
Good point!


> 
>> +		bo_flags |= XE_BO_FLAG_NEEDS_64K;
>> +
>>  	if (args->flags & DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM)
>> {
>>  		if (XE_IOCTL_DBG(xe, !(bo_flags &
>> XE_BO_FLAG_VRAM_MASK)))
>>  			return -EINVAL;
>> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
>> index 4cc13eddb6b32..3eb76d874eb28 100644
>> --- a/drivers/gpu/drm/xe/xe_vm.c
>> +++ b/drivers/gpu/drm/xe/xe_vm.c
>> @@ -2878,7 +2878,16 @@ static int xe_vm_bind_ioctl_validate_bo(struct
>> xe_device *xe, struct xe_bo *bo,
>>  		return -EINVAL;
>>  	}
>>  
>> -	if (bo->flags & XE_BO_FLAG_INTERNAL_64K) {
>> +	/*
>> +	 * Some platforms require 64k VM_BIND alignment,
>> +	 * specifically those with XE_VRAM_FLAGS_NEED64K.
>> +	 *
>> +	 * Other platforms may have BO's set to 64k physical
>> placement,
>> +	 * but can be mapped at 4k offsets anyway. This check is
>> only
>> +	 * there for the former case.
>> +	 */
>> +	if ((bo->flags & XE_BO_FLAG_INTERNAL_64K) &&
>> +	    (xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)) {
>>  		if (XE_IOCTL_DBG(xe, obj_offset &
>>  				 XE_64K_PAGE_MASK) ||
>>  		    XE_IOCTL_DBG(xe, addr & XE_64K_PAGE_MASK) ||
> 
> Otherwise LGTM.
> /Thomas
> 
Thanks,
~Maarten
diff mbox series

Patch

diff --git a/drivers/gpu/drm/xe/display/intel_fb_bo.c b/drivers/gpu/drm/xe/display/intel_fb_bo.c
index f835492f73fb4..6775c2557b9df 100644
--- a/drivers/gpu/drm/xe/display/intel_fb_bo.c
+++ b/drivers/gpu/drm/xe/display/intel_fb_bo.c
@@ -28,6 +28,11 @@  int intel_fb_bo_framebuffer_init(struct intel_framebuffer *intel_fb,
 	struct xe_device *xe = to_xe_device(bo->ttm.base.dev);
 	int ret;
 
+	/* Only this specific format is affected, and it's only available on VRAM */
+	if (XE_IOCTL_DBG(xe, mode_cmd->modifier[0] == I915_FORMAT_MOD_4_TILED_BMG_CCS &&
+			     !(bo->flags & XE_BO_FLAG_NEEDS_64K)))
+		return -EINVAL;
+
 	xe_bo_get(bo);
 
 	ret = ttm_bo_reserve(&bo->ttm, true, false, NULL);
diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index cbe7bf098970f..41297b5797173 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -2019,6 +2019,13 @@  int xe_gem_create_ioctl(struct drm_device *dev, void *data,
 
 	bo_flags |= args->placement << (ffs(XE_BO_FLAG_SYSTEM) - 1);
 
+	/* CCS formats need physical placement at a 64K alignment in VRAM. */
+	if ((bo_flags & XE_BO_FLAG_VRAM_MASK) &&
+	    (bo_flags & XE_BO_FLAG_SCANOUT) &&
+	    !(xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K) &&
+	    !(args->size % SZ_64K))
+		bo_flags |= XE_BO_FLAG_NEEDS_64K;
+
 	if (args->flags & DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM) {
 		if (XE_IOCTL_DBG(xe, !(bo_flags & XE_BO_FLAG_VRAM_MASK)))
 			return -EINVAL;
diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index 4cc13eddb6b32..3eb76d874eb28 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -2878,7 +2878,16 @@  static int xe_vm_bind_ioctl_validate_bo(struct xe_device *xe, struct xe_bo *bo,
 		return -EINVAL;
 	}
 
-	if (bo->flags & XE_BO_FLAG_INTERNAL_64K) {
+	/*
+	 * Some platforms require 64k VM_BIND alignment,
+	 * specifically those with XE_VRAM_FLAGS_NEED64K.
+	 *
+	 * Other platforms may have BO's set to 64k physical placement,
+	 * but can be mapped at 4k offsets anyway. This check is only
+	 * there for the former case.
+	 */
+	if ((bo->flags & XE_BO_FLAG_INTERNAL_64K) &&
+	    (xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K)) {
 		if (XE_IOCTL_DBG(xe, obj_offset &
 				 XE_64K_PAGE_MASK) ||
 		    XE_IOCTL_DBG(xe, addr & XE_64K_PAGE_MASK) ||