intel/skylake: Implement native Cache-as-RAM (CAR)

Now coreboot should do BIOS CAR setup along with NEM
mode setup.

This patch also provides a mechanism to use 16MB code caching
benefit although LLC still limited to 1M/1.5M based
on SOC LLC limit.
Here with unlimited cache line gets replaced. Now we could use
unlimited cache size along with well defined data size

[pg: updated to current upstream #defines]

BUG=chrome-os-partner:48412
BRANCH=glados
TEST=Builds and Boots on FAB4 SKU2/3.

Signed-off-by: Subrata Banik <subrata.banik@intel.com>
Signed-off-by: pchandri <preetham.chandrian@intel.com>
Signed-off-by: Dhaval Sharma <dhaval.v.sharma@intel.com>

Change-Id: I96a9cf3a6e41cae9619c683dca28ad31dcaa2536
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Original-Commit-Id: 2ec51f15c874ad2f1f4fad52fa8deced7b27a24b
Original-Change-Id: Id62c15799d98bc27b5e558adfa7c7b3468aa153a
Original-Reviewed-on: https://chromium-review.googlesource.com/320855
Original-Commit-Ready: Subrata Banik <subrata.banik@intel.com>
Original-Tested-by: Subrata Banik <subrata.banik@intel.com>
Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: https://review.coreboot.org/13138
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martinroth@google.com>

7 files changed, 427 insertions, 17 deletions

diff --git a/src/drivers/intel/fsp1_1/Kconfig b/src/drivers/intel/fsp1_1/Kconfig
index c959e57cd1..9b2c463878 100644
--- a/src/drivers/intel/fsp1_1/Kconfig
+++ b/src/drivers/intel/fsp1_1/Kconfig
@@ -23,14 +23,6 @@ if PLATFORM_USES_FSP1_1
 
 comment "Intel FSP 1.1"
 
-config DCACHE_RAM_BASE
-	hex
-	default 0xfef00000
-
-config DCACHE_RAM_SIZE
-	hex
-	default 0x4000
-
 config HAVE_FSP_BIN
 	bool "Should the Intel FSP binary be added to the flash image"
 	help
diff --git a/src/drivers/intel/fsp1_1/after_raminit.S b/src/drivers/intel/fsp1_1/after_raminit.S
index 85a0a8c200..eb99157959 100644
--- a/src/drivers/intel/fsp1_1/after_raminit.S
+++ b/src/drivers/intel/fsp1_1/after_raminit.S
@@ -20,7 +20,6 @@
 #include <cpu/x86/cache.h>
 #include <cpu/x86/post_code.h>
 
-.extern fih_car
 /*
  * This is the common entry point after DRAM has been initialized.
  */
@@ -31,6 +30,14 @@
 	/* Switch to the stack in RAM */
 	movl	%eax, %esp
 
+#if IS_ENABLED(CONFIG_SKIP_FSP_CAR)
+
+	/* SOC specific NEM */
+	#include <soc/car_teardown.S>
+
+#else
+.extern fih_car
+
 	post_code(POST_FSP_TEMP_RAM_EXIT)
 
 	/* Calculate TempRamExit entry into FSP */
@@ -56,8 +63,8 @@
 	 */
 	movb	$0xBC, %ah
 	jmp	.Lhlt
-
 1:
+#endif
 	/* Display the MTRRs */
 	call	soc_display_mtrrs
 
diff --git a/src/drivers/intel/fsp1_1/cache_as_ram.inc b/src/drivers/intel/fsp1_1/cache_as_ram.inc
index 35abdb48da..6e7e50b992 100644
--- a/src/drivers/intel/fsp1_1/cache_as_ram.inc
+++ b/src/drivers/intel/fsp1_1/cache_as_ram.inc
@@ -33,17 +33,22 @@
 	 * mm0:  low 32-bits of TSC value
 	 * mm1:  high 32-bits of TSC value
 	 */
-
-	mov	%eax, %edi
-
+	movl	%eax, %edi
 cache_as_ram:
 	post_code(0x20)
 
+#if IS_ENABLED(CONFIG_SKIP_FSP_CAR)
+
 	/*
-	 * edi:  BIST value
-	 * mm0:  low 32-bits of TSC value
-	 * mm1:  high 32-bits of TSC value
+	 * SOC specific setup
+	 * NOTE: This has to preserve the registers
+	 * mm0, mm1 and edi.
 	 */
+	#include <soc/car_setup.S>
+
+	post_code(0x28)
+
+#endif
 
 	/*
 	 * Find the FSP binary in cbfs.
@@ -143,7 +148,7 @@ CAR_init_done:
 	rep	stosl
 
 before_romstage:
-	post_code(0x23)
+	post_code(0x2A)
 
 	/* Call cache_as_ram_main(struct cache_as_ram_params *) */
 	call	cache_as_ram_main
diff --git a/src/soc/intel/skylake/Kconfig b/src/soc/intel/skylake/Kconfig
index 58b828fd17..35b2a18ee6 100644
--- a/src/soc/intel/skylake/Kconfig
+++ b/src/soc/intel/skylake/Kconfig
@@ -176,4 +176,14 @@ config NHLT_SSM4567
 	help
 	  Include DSP firmware settings for ssm4567 smart amplifier.
 
+config DCACHE_RAM_SIZE_TOTAL
+	hex
+	default 0x40000
+
+config SKIP_FSP_CAR
+        bool "Skip cache as RAM setup in FSP"
+        default y
+        help
+          Skip Cache as RAM setup in FSP.
+
 endif
diff --git a/src/soc/intel/skylake/include/soc/car_setup.S b/src/soc/intel/skylake/include/soc/car_setup.S
new file mode 100644
index 0000000000..4b881a18b3
--- /dev/null
+++ b/src/soc/intel/skylake/include/soc/car_setup.S
@@ -0,0 +1,334 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2013-2014 Sage Electronic Engineering, LLC.
+ * Copyright (C) 2016 Intel Corp.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+#include <cpu/x86/mtrr.h>
+#include <cpu/x86/cache.h>
+#include <cpu/x86/post_code.h>
+
+/*
+ * MTRR definitions
+ */
+.equ		IA32_MTRR_CAP,			0x00fe
+
+.equ		NO_EVICT_MODE,			0x02e0
+
+.equ		IA32_PQR_ASSOC,			0x0c8f
+.equ		IA32_L3_MASK_1,			0x0c91
+.equ		IA32_L3_MASK_2,			0x0c92
+.equ		CACHE_INIT_VALUE,		0
+
+/*
+ * See BWG - chapter "Determining Cacheable Code Region Base Addresses and Ranges".
+ *
+ */
+
+	movl	%edi, %ebp		/* Put BIST value in a safe place */
+	/*
+	 * Ensure that all variable-range MTRR valid flags are clear and
+	 * IA32_MTRR_DEF_TYPE MSR E flag is clear.  Note: This is the default state
+	 * after hardware reset.
+	 *
+	 * Initialize all fixed-range and variable-range MTRR register fields to 0.
+	 */
+	mov	$(MtrrByteCountFixed), %ebx		/* EBX = size of  Fixed MTRRs */
+
+	xorl	%eax, %eax			/* Clear the low dword to write */
+	xorl	%edx, %edx			/* Clear the high dword to write */
+	xorl	%ecx, %ecx
+	/* Clearing Fixed Range MTRRs */
+clear_mtrr_fixed:
+	addl	$(-2), %ebx			/* need to check it */
+	movw	(MtrrInitTable)(%ebx), %cx	/* cx <- address of mtrr to zero */
+	wrmsr
+	jnz	clear_mtrr_fixed			/* loop through the whole table */
+
+	post_code(0x21)
+
+	/* Clearing Variable Range MTRRs */
+	movl	$MTRR_CAP_MSR, %ecx
+	rdmsr
+	movzx	%al, %ebx
+	clr	%eax
+	clr	%edx
+	movl	$MTRR_PHYS_BASE(0), %ecx
+clear_var_mtrr:
+	wrmsr
+	inc	%ecx
+	wrmsr
+	inc	%ecx
+	dec	%ebx
+	jnz	clear_var_mtrr
+
+	post_code(0x22)
+
+	/*
+	 * Configure the default memory type to un-cacheable (UC) in the
+	 * IA32_MTRR_DEF_TYPE MSR.
+	 */
+
+	movl	$MTRR_DEF_TYPE_MSR, %ecx	/* Load the MTRR default type index */
+	rdmsr
+	andl	$0xFFFFF300, %eax		/* Clear the enable bits and def type UC. */
+	wrmsr
+
+	/* Configure MTRR_PHYS_MASK_HIGH for proper addressing above 4GB
+	 * based on the physical address size supported for this processor
+	 * This is based on read from CPUID EAX = 080000008h, EAX bits [7:0]
+	 *
+	 * Examples:
+	 *  MTRR_PHYS_MASK_HIGH = 00000000Fh  For 36 bit addressing
+	 *  MTRR_PHYS_MASK_HIGH = 0000000FFh  For 40 bit addressing
+	 */
+
+	movl	$0x80000008, %eax 	/* Address sizes leaf */
+	cpuid
+	sub	$32, %al
+	movzx	%al, %eax
+	xorl	%esi, %esi
+	bts	%eax, %esi
+	dec	%esi			/* esi <- MTRR_PHYS_MASK_HIGH */
+
+	/*
+	 *   Configure the DataStack region as write-back (WB) cacheable memory type
+	 *   using the variable range MTRRs.
+	 *
+	 *
+	 * Set the base address of the DataStack cache range
+	 */
+
+	movl	$CONFIG_DCACHE_RAM_BASE, %eax
+	orl	$MTRR_TYPE_WRBACK, %eax	/* Load the write-back cache value */
+	xorl	%edx, %edx			/* clear upper dword */
+	movl	$MTRR_PHYS_BASE(0), %ecx	/* Load the MTRR index */
+	wrmsr					/* the value in MTRR_PHYS_BASE_0 */
+
+	/*
+	 * Set the mask for the DataStack cache range
+	 * Compute MTRR mask value:  Mask = NOT (Size - 1)
+	 */
+	movl	$CONFIG_DCACHE_RAM_SIZE_TOTAL, %eax
+	dec	%eax
+	not	%eax
+	orl	$MTRR_PHYS_MASK_VALID, %eax	/* turn on the Valid flag */
+	movl	%esi, %edx			/* edx <- MTRR_PHYS_MASK_HIGH */
+	inc	%ecx
+	wrmsr					/* the value in MTRR_PHYS_MASK_0 */
+
+	post_code(0x23)
+
+	/*
+	 * Enable the MTRRs by setting the IA32_MTRR_DEF_TYPE MSR E flag.
+	 */
+	movl	$MTRR_DEF_TYPE_MSR, %ecx               /* Load the MTRR default type index */
+	rdmsr
+	orl	$MTRR_DEF_TYPE_EN, %eax             /* Enable variable range MTRRs */
+	wrmsr
+
+	post_code(0x24)
+
+	/*
+	 *   Enable the logical processor's (BSP) cache: execute INVD and set
+	 *   CR0.CD = 0, CR0.NW = 0.
+	 */
+	movl	%cr0, %eax
+	and	$(~(CR0_CD + CR0_NW)), %eax
+	invd
+	movl	%eax, %cr0
+
+	/*
+	 *   Enable No-Eviction Mode Setup State by setting
+	 *   NO_EVICT_MODE  MSR 2E0h bit [0] = '1'.
+	 */
+	movl	$NO_EVICT_MODE, %ecx
+	rdmsr
+	orl	$0x01, %eax
+	wrmsr
+
+	/* Create n-way set associativity of cache */
+	xorl	%edi, %edi
+Find_LLC_subleaf:
+	movl	%edi, %ecx
+	movl	$0x04, %eax
+	cpuid
+	inc	%edi
+	and	$0xe0, %al	/* EAX[7:5] = Cache Level */
+	cmp	$0x60, %al	/* Check to see if it is LLC */
+	jnz	Find_LLC_subleaf
+
+	/*
+	 * Set MSR 0xC91 IA32_L3_MASK_! = 0xE/0xFE/0xFFE/0xFFFE
+	 * for 4/8/16 way of LLC
+	*/
+	shr	$22, %ebx
+	inc	%ebx
+	/* Calculate n-way associativity of LLC */
+	mov	%bl, %cl
+
+	/*
+	 * Maximizing RO cacheability while locking in the CAR to a
+	 * single way since that particular way won't be victim candidate
+	 * for evictions.
+	 * This has been done after programing LLC_WAY_MASK_1 MSR
+	 * with desired LLC way as mentioned below.
+	 *
+	 * Hence create Code and Data Size as per request
+	 * Code Size (RO) : Up to 16M
+	 * Data Size (RW) : Up to 256K
+	 */
+	movl	$0x01, %eax
+	/*
+	 * LLC Ways -> LLC_WAY_MASK_1:
+	 *  4: 0x000E
+	 *  8: 0x00FE
+	 * 12: 0x0FFE
+	 * 16: 0xFFFE
+	 *
+	 * These MSRs contain one bit per each way of LLC
+	 * - If this bit is '0' - the way is protected from eviction
+	 * - If this bit is '1' - the way is not protected from eviction
+	 */
+	shl	%cl, %eax
+	subl	$0x02, %eax
+	movl	$IA32_L3_MASK_1, %ecx
+	xorl	%edx, %edx
+	wrmsr
+	/*
+	 * Set MSR 0xC92 IA32_L3_MASK_2 = 0x1
+	 *
+	 * For SKL SOC, data size remains 256K consistently.
+	 * Hence, creating 1-way associative cache for Data
+	*/
+	mov	$IA32_L3_MASK_2, %ecx
+	mov	$0x01, %eax
+	xorl	%edx, %edx
+	wrmsr
+	/*
+	 * Set IA32_PQR_ASSOC = 0x02
+	 *
+	 * Possible values:
+	 * 0: Default value, no way mask should be applied
+	 * 1: Apply way mask 1 to LLC
+	 * 2: Apply way mask 2 to LLC
+	 * 3: Shouldn't be use in NEM Mode
+	 */
+	movl	$IA32_PQR_ASSOC, %ecx
+	movl	$0x02, %eax
+	xorl	%edx, %edx
+	wrmsr
+
+	movl	$CONFIG_DCACHE_RAM_BASE, %edi
+	movl	$CONFIG_DCACHE_RAM_SIZE_TOTAL, %ecx
+	shr	$0x02, %ecx
+	movl	$CACHE_INIT_VALUE, %eax
+	cld
+	rep	stosl
+	/*
+	 * Set IA32_PQR_ASSOC = 0x01
+	 * At this stage we apply LLC_WAY_MASK_1 to the cache.
+	 * i.e. way 0 is protected from eviction.
+	*/
+	movl	$IA32_PQR_ASSOC, %ecx
+	movl	$0x01, %eax
+	xorl	%edx, %edx
+	wrmsr
+
+	/*
+	 * Enable No-Eviction Mode Run State by setting
+	 * NO_EVICT_MODE MSR 2E0h bit [1] = '1'.
+	 */
+
+	movl	$NO_EVICT_MODE, %ecx
+	rdmsr
+	orl	$0x02, %eax
+	wrmsr
+
+	post_code(0x25)
+	/*
+	 *   Configure the BIOS code region as write-protected (WP) cacheable
+	 *   memory type using a single variable range MTRR.
+	 *
+	 *   Ensure region to cache meets MTRR requirements for
+	 *   size and alignment.
+	 */
+	movl	$(0xFFFFFFFF - CONFIG_ROM_SIZE + 1), %edi	/* Code region base */
+	movl	$CONFIG_ROM_SIZE, %eax				/* Code region size */
+	cmpl	$0, %edi
+	jz	InvalidParameter
+	cmpl	$0, %eax
+	jz	InvalidParameter
+	jmp	CheckPass
+
+InvalidParameter:
+	movl	$0x80000002, %eax	/* RETURN_INVALID_PARAMETER */
+	jmp	.Lhlt
+
+CheckPass:
+
+	post_code(0x26)
+
+	/*
+	 * Program base register
+	 */
+	xorl	%edx, %edx			/* clear upper dword */
+	movl	$MTRR_PHYS_BASE(1), %ecx	/* setup variable mtrr */
+	movl	%edi, %eax
+	orl	$MTRR_TYPE_WRPROT, %eax	/* set type to write protect */
+	wrmsr
+
+	movl	$CONFIG_ROM_SIZE, %eax
+
+	/*
+	 * Compute MTRR mask value:  Mask = NOT (Size - 1)
+	 */
+	dec	%eax	/* eax - size to cache less one byte */
+	not	%eax	/* eax contains low 32 bits of mask */
+	or	$MTRR_PHYS_MASK_VALID, %eax
+	/*
+	 * Program mask register
+	 */
+	movl	$MTRR_PHYS_MASK(1) , %ecx	/* setup variable mtrr */
+	movl	%esi, %edx	/* edx <- MTRR_PHYS_MASK_HIGH */
+	wrmsr
+
+	post_code(0x27)
+
+	/*
+	 * edi:  BIST value
+	 * mm0:  low 32-bits of TSC value
+	 * mm1:  high 32-bits of TSC value
+	 */
+	movl	%ebp, %edi		/* Restore BIST value */
+
+	.section	.rodata
+
+MtrrInitTable:
+	.word      MTRR_DEF_TYPE_MSR
+	.word      MTRR_FIX_64K_00000
+	.word      MTRR_FIX_16K_80000
+	.word      MTRR_FIX_16K_A0000
+	.word      MTRR_FIX_4K_C0000
+	.word      MTRR_FIX_4K_C8000
+	.word      MTRR_FIX_4K_D0000
+	.word      MTRR_FIX_4K_D8000
+	.word      MTRR_FIX_4K_E0000
+	.word      MTRR_FIX_4K_E8000
+	.word      MTRR_FIX_4K_F0000
+	.word      MTRR_FIX_4K_F8000
+
+.equ MtrrByteCountFixed,  (.-MtrrInitTable)
+
+	.previous
diff --git a/src/soc/intel/skylake/include/soc/car_teardown.S b/src/soc/intel/skylake/include/soc/car_teardown.S
new file mode 100644
index 0000000000..315b3c1deb
--- /dev/null
+++ b/src/soc/intel/skylake/include/soc/car_teardown.S
@@ -0,0 +1,54 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2013-2014 Sage Electronic Engineering, LLC.
+ * Copyright (C) 2016 Intel Corp.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+.equ		IA32_PQR_ASSOC,			0x0c8f
+
+	/* Disable MTRR by clearing the IA32_MTRR_DEF_TYPE MSR E flag. */
+	movl	$MTRR_DEF_TYPE_MSR, %ecx
+	rdmsr
+	andl	$(~MTRR_DEF_TYPE_EN), %eax
+	wrmsr
+
+	/* Invalidate Cache */
+	invd
+
+	/*
+	 * Disable No-Eviction Mode Run State by clearing
+	 * NO_EVICT_MODE MSR 2E0h bit [1] = 0
+	 */
+	movl	$0x000002E0, %ecx
+	rdmsr
+	andl	$~(0x2), %eax
+	wrmsr
+
+	/*
+	 * Disable No-Eviction Mode Setup State by clearing
+	 * NO_EVICT_MODE MSR 2E0h bit [0] = 0
+	 */
+	rdmsr
+	andl	$~(0x1), %eax
+	wrmsr
+
+	/*
+	 * Set IA32_PQR_ASSOC = 0x00
+	 * This step guarantees that no protected way remain in LLC cache,
+	 * all the ways are open for the evictions.
+	 */
+	movl	$IA32_PQR_ASSOC, %ecx
+	movl	$0x00, %eax
+	xorl	%edx, %edx
+	wrmsr
diff --git a/src/soc/intel/skylake/romstage/romstage.c b/src/soc/intel/skylake/romstage/romstage.c
index b872d39fc2..9b95f721ca 100644
--- a/src/soc/intel/skylake/romstage/romstage.c
+++ b/src/soc/intel/skylake/romstage/romstage.c
@@ -96,6 +96,10 @@ void soc_memory_init_params(struct romstage_params *params,
 		upd->SaGv = config->SaGv;
 	upd->RMT = config->Rmt;
 	upd->DdrFreqLimit = config->DdrFreqLimit;
+	if (IS_ENABLED(CONFIG_SKIP_FSP_CAR)) {
+		upd->FspCarBase = CONFIG_DCACHE_RAM_BASE;
+		upd->FspCarSize = CONFIG_DCACHE_RAM_SIZE_TOTAL;
+	}
 }
 
 void soc_display_memory_init_params(const MEMORY_INIT_UPD *old,
@@ -232,6 +236,10 @@ void soc_display_memory_init_params(const MEMORY_INIT_UPD *old,
 		new->ApertureSize);
 	fsp_display_upd_value("SaGv", 1, old->SaGv, new->SaGv);
 	fsp_display_upd_value("RMT", 1, old->RMT, new->RMT);
+	fsp_display_upd_value("FspCarBase", 1, old->FspCarBase,
+		new->FspCarBase);
+	fsp_display_upd_value("FspCarSize", 1, old->FspCarSize,
+		new->FspCarSize);
 }
 
 /* SOC initialization after RAM is enabled. */