[PATCH v3 1/6] drm/xe: Add functions to convert regular address to canonical address and back

[Matt Roper]

On Mon, Jan 29, 2024 at 10:17:37AM -0800, José Roberto de Souza wrote:
> Some instructions requires canonical address like
> MI_BATCH_BUFFER_START(UMDs must call xe_exec with a canonical address
> for Xe2+).
> 
> So here adding functions to convert regular address to canonical
> address and back, the first user of this functions will be added
> in the next patches.
> 
> v3:
> - inline removed
> - rename highest_address_bit_get() to ppgtt_msb_get()
> 
> Cc: Rodrigo Vivi <rodrigo.vivi at intel.com>
> Cc: Maarten Lankhorst <dev at lankhorst.se>
> Cc: Stuart Summers <stuart.summers at intel.com>
> Cc: Jani Nikula <jani.nikula at intel.com>
> Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> ---
>  drivers/gpu/drm/xe/xe_device.c | 15 +++++++++++++++
>  drivers/gpu/drm/xe/xe_device.h |  3 +++
>  2 files changed, 18 insertions(+)
> 
> diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> index 6faa7865b1aab..6172d4f987937 100644
> --- a/drivers/gpu/drm/xe/xe_device.c
> +++ b/drivers/gpu/drm/xe/xe_device.c
> @@ -747,3 +747,18 @@ void xe_device_snapshot_print(struct xe_device *xe, struct drm_printer *p)
>  		drm_printf(p, "\tCS reference clock: %u\n", gt->info.reference_clock);
>  	}
>  }
> +
> +static int ppgtt_msb_get(struct xe_device *xe)
> +{
> +	return xe->info.dma_mask_size > 48 ? 57 : 47;

Can we just use (xe->info.va_bits - 1)?  As far as I can tell, what we
really care about for canonical form is the size of the PPGTT virtual
address space, not the DMA mask, right?  That also matches platforms
like Xe_LP better where:

        .dma_mask_size = 39,                                                                                   
        .va_bits = 48,                                                                                         


Matt

> +}
> +
> +u64 xe_device_canonicalize_addr(struct xe_device *xe, u64 address)
> +{
> +	return sign_extend64(address, ppgtt_msb_get(xe));
> +}
> +
> +u64 xe_device_uncanonicalize_addr(struct xe_device *xe, u64 address)
> +{
> +	return address & GENMASK_ULL(ppgtt_msb_get(xe), 0);
> +}
> diff --git a/drivers/gpu/drm/xe/xe_device.h b/drivers/gpu/drm/xe/xe_device.h
> index 270124da1e00e..462f59e902b12 100644
> --- a/drivers/gpu/drm/xe/xe_device.h
> +++ b/drivers/gpu/drm/xe/xe_device.h
> @@ -177,4 +177,7 @@ u32 xe_device_ccs_bytes(struct xe_device *xe, u64 size);
>  
>  void xe_device_snapshot_print(struct xe_device *xe, struct drm_printer *p);
>  
> +u64 xe_device_canonicalize_addr(struct xe_device *xe, u64 address);
> +u64 xe_device_uncanonicalize_addr(struct xe_device *xe, u64 address);
> +
>  #endif
> -- 
> 2.43.0
> 

-- 
Matt Roper
Graphics Software Engineer
Linux GPU Platform Enablement
Intel Corporation