Re: [PATCH 06/11] drm/xe/xe2hpg: Determine flat ccs offset for vram

[Matt Roper <matthew.d.roper@intel.com>]

On Tue, Apr 02, 2024 at 06:17:19PM +0530, Balasubramani Vivekanandan wrote:
> From: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
> 
> on Xe2 dgfx platform determine the offset using Flat CCS size
> bitfield of XE2_FLAT_CCS_BASE_RANGE_[UPPER/LOWER] mcr registers.
> 
> Bspec: 68023
> 
> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
> Signed-off-by: Akshata Jahagirdar <akshata.jahagirdar@intel.com>
> Signed-off-by: Matthew Auld <matthew.auld@intel.com>
> Signed-off-by: Balasubramani Vivekanandan <balasubramani.vivekanandan@intel.com>
> ---
>  drivers/gpu/drm/xe/regs/xe_gt_regs.h |  5 ++++
>  drivers/gpu/drm/xe/xe_mmio.c         | 39 ++++++++++++++++++++++++++--
>  2 files changed, 42 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/regs/xe_gt_regs.h b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> index d5b21f03beaa..ea75c1f0ebd0 100644
> --- a/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> +++ b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> @@ -69,6 +69,7 @@
>  
>  #define XEHP_TILE_ADDR_RANGE(_idx)		XE_REG_MCR(0x4900 + (_idx) * 4)
>  #define XEHP_FLAT_CCS_BASE_ADDR			XE_REG_MCR(0x4910)
> +#define XEHP_FLAT_CCS_PTR			REG_GENMASK(31, 8)
>  
>  #define WM_CHICKEN3				XE_REG_MCR(0x5588, XE_REG_OPTION_MASKED)
>  #define   HIZ_PLANE_COMPRESSION_DIS		REG_BIT(10)
> @@ -141,6 +142,10 @@
>  
>  #define XE2_FLAT_CCS_BASE_RANGE_LOWER		XE_REG_MCR(0x8800)
>  #define   XE2_FLAT_CCS_ENABLE			REG_BIT(0)
> +#define XE2_FLAT_CCS_BASE_LOWER_ADDR_MASK	REG_GENMASK(31, 6)
> +
> +#define XE2_FLAT_CCS_BASE_RANGE_UPPER		XE_REG_MCR(0x8804)
> +#define XE2_FLAT_CCS_BASE_UPPER_ADDR_MASK	REG_GENMASK(7, 0)
>  
>  #define GSCPSMI_BASE				XE_REG(0x880c)
>  
> diff --git a/drivers/gpu/drm/xe/xe_mmio.c b/drivers/gpu/drm/xe/xe_mmio.c
> index 5d13fc7cb9d2..49a32b03fcf8 100644
> --- a/drivers/gpu/drm/xe/xe_mmio.c
> +++ b/drivers/gpu/drm/xe/xe_mmio.c
> @@ -163,6 +163,42 @@ static int xe_determine_lmem_bar_size(struct xe_device *xe)
>  	return 0;
>  }
>  
> +static inline u64 get_flat_ccs_offset(struct xe_gt *gt, u64 *tile_size)

It doesn't seem to be necessary to pass by reference for tile_size.

> +{
> +	struct xe_device *xe = gt_to_xe(gt);
> +	u64 offset;
> +	u32 reg;
> +
> +	if (GRAPHICS_VER(xe) >= 20) {
> +		u64 ccs_size = *tile_size / 512;

This value only gets used for the assertion below, but should we be 
taking the value from 0x8804[23:14] rather than deriving it from
TILE_ADDR_RANGE?

> +		u64 offset_hi, offset_lo;
> +		u32 nodes, num_enabled;
> +
> +		reg = xe_mmio_read32(gt, MIRROR_FUSE3);
> +		nodes = REG_FIELD_GET(XE2_NODE_ENABLE_MASK, reg);
> +		num_enabled = hweight32(nodes); /* Number of enabled l3 nodes */
> +
> +		reg = xe_gt_mcr_unicast_read_any(gt, XE2_FLAT_CCS_BASE_RANGE_LOWER);
> +		offset_lo = REG_FIELD_GET(XE2_FLAT_CCS_BASE_LOWER_ADDR_MASK, reg);

This register also appears to have a general "enable" bit for FlatCCS in
bit(0).  Should we be reading that to make sure FlatCCS hasn't been
fused off or disabled by the IFWI?


Matt

> +
> +		reg = xe_gt_mcr_unicast_read_any(gt, XE2_FLAT_CCS_BASE_RANGE_UPPER);
> +		offset_hi = REG_FIELD_GET(XE2_FLAT_CCS_BASE_UPPER_ADDR_MASK, reg);
> +
> +		offset = offset_hi << 32; /* HW view bits 39:32 */
> +		offset |= offset_lo << 6; /* HW view bits 31:6 */
> +		offset *= num_enabled; /* convert to SW view */
> +
> +		/* We don't expect any holes */
> +		xe_assert_msg(xe, offset == (xe_mmio_read64_2x32(gt, GSMBASE) - ccs_size),
> +			      "Hole between CCS and GSM.\n");
> +	} else {
> +		reg = xe_gt_mcr_unicast_read_any(gt, XEHP_FLAT_CCS_BASE_ADDR);
> +		offset = (u64)REG_FIELD_GET(XEHP_FLAT_CCS_PTR, reg) * SZ_64K;
> +	}
> +
> +	return offset;
> +}
> +
>  /**
>   * xe_mmio_tile_vram_size() - Collect vram size and offset information
>   * @tile: tile to get info for
> @@ -207,8 +243,7 @@ static int xe_mmio_tile_vram_size(struct xe_tile *tile, u64 *vram_size,
>  
>  	/* minus device usage */
>  	if (xe->info.has_flat_ccs) {
> -		reg = xe_gt_mcr_unicast_read_any(gt, XEHP_FLAT_CCS_BASE_ADDR);
> -		offset = (u64)REG_FIELD_GET(GENMASK(31, 8), reg) * SZ_64K;
> +		offset = get_flat_ccs_offset(gt, tile_size);
>  	} else {
>  		offset = xe_mmio_read64_2x32(gt, GSMBASE);
>  	}
> -- 
> 2.25.1
> 

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