drm/amd: Add GFX11 modifiers support to AMDGPU (v3)

GFX11 IP introduces new tiling mode. Various combinations of DCC
settings are possible and the most preferred settings must be exposed
for optimal use of the hardware.

add_gfx11_modifiers() is based on recommendation from Marek for the
preferred tiling modifier that are most efficient for the hardware.

v2: microtiling fix noticed by Marek
v3: keep Z tiling check

Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

2 files changed, 115 insertions, 13 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
index 17c9bbe0cbc5..8ef489fad707 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
@@ -30,6 +30,9 @@
 #include "atom.h"
 #include "amdgpu_connectors.h"
 #include "amdgpu_display.h"
+#include "soc15_common.h"
+#include "gc/gc_11_0_0_offset.h"
+#include "gc/gc_11_0_0_sh_mask.h"
 #include <asm/div64.h>
 
 #include <linux/pci.h>
@@ -663,6 +666,11 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 {
 	struct amdgpu_device *adev = drm_to_adev(afb->base.dev);
 	uint64_t modifier = 0;
+	int num_pipes = 0;
+	int num_pkrs = 0;
+
+	num_pkrs = adev->gfx.config.gb_addr_config_fields.num_pkrs;
+	num_pipes = adev->gfx.config.gb_addr_config_fields.num_pipes;
 
 	if (!afb->tiling_flags || !AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE)) {
 		modifier = DRM_FORMAT_MOD_LINEAR;
@@ -675,7 +683,7 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 		int bank_xor_bits = 0;
 		int packers = 0;
 		int rb = 0;
-		int pipes = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes);
+		int pipes = ilog2(num_pipes);
 		uint32_t dcc_offset = AMDGPU_TILING_GET(afb->tiling_flags, DCC_OFFSET_256B);
 
 		switch (swizzle >> 2) {
@@ -691,12 +699,17 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 		case 6: /* 64 KiB _X */
 			block_size_bits = 16;
 			break;
+		case 7: /* 256 KiB */
+			block_size_bits = 18;
+			break;
 		default:
 			/* RESERVED or VAR */
 			return -EINVAL;
 		}
 
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0))
+			version = AMD_FMT_MOD_TILE_VER_GFX11;
+		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS;
 		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 0, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10;
@@ -707,19 +720,33 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 		case 0: /* Z microtiling */
 			return -EINVAL;
 		case 1: /* S microtiling */
-			if (!has_xor)
-				version = AMD_FMT_MOD_TILE_VER_GFX9;
+			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+				if (!has_xor)
+					version = AMD_FMT_MOD_TILE_VER_GFX9;
+			}
 			break;
 		case 2:
-			if (!has_xor && afb->base.format->cpp[0] != 4)
-				version = AMD_FMT_MOD_TILE_VER_GFX9;
+			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+				if (!has_xor && afb->base.format->cpp[0] != 4)
+					version = AMD_FMT_MOD_TILE_VER_GFX9;
+			}
 			break;
 		case 3:
 			break;
 		}
 
 		if (has_xor) {
+			if (num_pipes == num_pkrs && num_pkrs == 0) {
+				DRM_ERROR("invalid number of pipes and packers\n");
+				return -EINVAL;
+			}
+
 			switch (version) {
+			case AMD_FMT_MOD_TILE_VER_GFX11:
+				pipe_xor_bits = min(block_size_bits - 8, pipes);
+				packers = min(block_size_bits - 8 - pipe_xor_bits,
+						ilog2(adev->gfx.config.gb_addr_config_fields.num_pkrs));
+				break;
 			case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS:
 				pipe_xor_bits = min(block_size_bits - 8, pipes);
 				packers = min(block_size_bits - 8 - pipe_xor_bits,
@@ -753,9 +780,10 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 			u64 render_dcc_offset;
 
 			/* Enable constant encode on RAVEN2 and later. */
-			bool dcc_constant_encode = adev->asic_type > CHIP_RAVEN ||
+			bool dcc_constant_encode = (adev->asic_type > CHIP_RAVEN ||
 						   (adev->asic_type == CHIP_RAVEN &&
-						    adev->external_rev_id >= 0x81);
+						    adev->external_rev_id >= 0x81)) &&
+						    adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0);
 
 			int max_cblock_size = dcc_i64b ? AMD_FMT_MOD_DCC_BLOCK_64B :
 					      dcc_i128b ? AMD_FMT_MOD_DCC_BLOCK_128B :
@@ -870,10 +898,11 @@ static unsigned int get_dcc_block_size(uint64_t modifier, bool rb_aligned,
 		return max(10 + (rb_aligned ? (int)AMD_FMT_MOD_GET(RB, modifier) : 0), 12);
 	}
 	case AMD_FMT_MOD_TILE_VER_GFX10:
-	case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: {
+	case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS:
+	case AMD_FMT_MOD_TILE_VER_GFX11: {
 		int pipes_log2 = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier);
 
-		if (ver == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 &&
+		if (ver >= AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 &&
 		    AMD_FMT_MOD_GET(PACKERS, modifier) == pipes_log2)
 			++pipes_log2;
 
@@ -966,6 +995,9 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 			case DC_SW_64KB_S_X:
 				block_size_log2 = 16;
 				break;
+			case DC_SW_VAR_S_X:
+				block_size_log2 = 18;
+				break;
 			default:
 				drm_dbg_kms(rfb->base.dev,
 					    "Swizzle mode with unknown block size: %d\n", swizzle);
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index 7280db2821ef..872b7899b645 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -89,10 +89,14 @@
 #include "dcn/dcn_1_0_offset.h"
 #include "dcn/dcn_1_0_sh_mask.h"
 #include "soc15_hw_ip.h"
+#include "soc15_common.h"
 #include "vega10_ip_offset.h"
 
 #include "soc15_common.h"
 
+#include "gc/gc_11_0_0_offset.h"
+#include "gc/gc_11_0_0_sh_mask.h"
+
 #include "modules/inc/mod_freesync.h"
 #include "modules/power/power_helpers.h"
 #include "modules/inc/mod_info_packet.h"
@@ -4888,7 +4892,9 @@ fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev,
 	unsigned int mod_bank_xor_bits = AMD_FMT_MOD_GET(BANK_XOR_BITS, modifier);
 	unsigned int mod_pipe_xor_bits = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier);
 	unsigned int pkrs_log2 = AMD_FMT_MOD_GET(PACKERS, modifier);
-	unsigned int pipes_log2 = min(4u, mod_pipe_xor_bits);
+	unsigned int pipes_log2;
+
+	pipes_log2 = min(5u, mod_pipe_xor_bits);
 
 	fill_gfx9_tiling_info_from_device(adev, tiling_info);
 
@@ -5224,8 +5230,69 @@ add_gfx10_3_modifiers(const struct amdgpu_device *adev,
 		    AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9));
 }
 
+static void
+add_gfx11_modifiers(struct amdgpu_device *adev,
+		      uint64_t **mods, uint64_t *size, uint64_t *capacity)
+{
+	int num_pipes = 0;
+	int pipe_xor_bits = 0;
+	int num_pkrs = 0;
+	int pkrs = 0;
+	u32 gb_addr_config;
+	unsigned swizzle_r_x;
+	uint64_t modifier_r_x;
+	uint64_t modifier_dcc_best;
+	uint64_t modifier_dcc_4k;
+
+	/* TODO: GFX11 IP HW init hasnt finish and we get zero if we read from
+	 * adev->gfx.config.gb_addr_config_fields.num_{pkrs,pipes} */
+	gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG);
+	ASSERT(gb_addr_config != 0);
+
+	num_pkrs = 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS);
+	pkrs = ilog2(num_pkrs);
+	num_pipes = 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PIPES);
+	pipe_xor_bits = ilog2(num_pipes);
+
+	/* R_X swizzle modes are the best for rendering and DCC requires them. */
+	swizzle_r_x = num_pipes > 16 ? AMD_FMT_MOD_TILE_GFX11_256K_R_X :
+                                              AMD_FMT_MOD_TILE_GFX9_64K_R_X;
+
+	modifier_r_x = AMD_FMT_MOD |
+		AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) |
+		AMD_FMT_MOD_SET(TILE, swizzle_r_x) |
+		AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) |
+		AMD_FMT_MOD_SET(PACKERS, pkrs);
+
+	/* DCC_CONSTANT_ENCODE is not set because it can't vary with gfx11 (it's implied to be 1). */
+	modifier_dcc_best = modifier_r_x |
+		AMD_FMT_MOD_SET(DCC, 1) |
+		AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 0) |
+		AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+		AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B);
+
+	/* DCC settings for 4K and greater resolutions. (required by display hw) */
+	modifier_dcc_4k = modifier_r_x |
+			AMD_FMT_MOD_SET(DCC, 1) |
+			AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) |
+			AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) |
+			AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B);
+
+	add_modifier(mods, size, capacity, modifier_dcc_best);
+	add_modifier(mods, size, capacity, modifier_dcc_4k);
+
+	add_modifier(mods, size, capacity, modifier_dcc_best | AMD_FMT_MOD_SET(DCC_RETILE, 1));
+	add_modifier(mods, size, capacity, modifier_dcc_4k | AMD_FMT_MOD_SET(DCC_RETILE, 1));
+
+	add_modifier(mods, size, capacity, modifier_r_x);
+
+	add_modifier(mods, size, capacity, AMD_FMT_MOD |
+             AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) |
+			 AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D));
+}
+
 static int
-get_plane_modifiers(const struct amdgpu_device *adev, unsigned int plane_type, uint64_t **mods)
+get_plane_modifiers(struct amdgpu_device *adev, unsigned int plane_type, uint64_t **mods)
 {
 	uint64_t size = 0, capacity = 128;
 	*mods = NULL;
@@ -5257,6 +5324,9 @@ get_plane_modifiers(const struct amdgpu_device *adev, unsigned int plane_type, u
 		else
 			add_gfx10_1_modifiers(adev, mods, &size, &capacity);
 		break;
+	case AMDGPU_FAMILY_GC_11_0_0:
+		add_gfx11_modifiers(adev, mods, &size, &capacity);
+		break;
 	}
 
 	add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR);
@@ -5295,7 +5365,7 @@ fill_gfx9_plane_attributes_from_modifiers(struct amdgpu_device *adev,
 		dcc->enable = 1;
 		dcc->meta_pitch = afb->base.pitches[1];
 		dcc->independent_64b_blks = independent_64b_blks;
-		if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) {
+		if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) {
 			if (independent_64b_blks && independent_128b_blks)
 				dcc->dcc_ind_blk = hubp_ind_block_64b_no_128bcl;
 			else if (independent_128b_blks)