libpayload arm64: update mmu translation table granule size, logic and macros

1. change mmu granule size from 64KB to 4KB
2. correct level 1 translation table creation logic
3. automatically calculate granule size related macros

BRANCH=none
BUG=none
TEST=boot to kernel on oak board

Change-Id: Ic62c7863dff53f566b82b68ff1d1ad9ec5d0698d
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Original-Commit-Id: e5de7d942e42a8202fb879ce64b871864b1b9d38
Original-Change-Id: I78d7838921fa82a670e18ddc2de6d766dc7a2146
Original-Signed-off-by: Jimmy Huang <jimmy.huang@mediatek.com>
Original-Reviewed-on: https://chromium-review.googlesource.com/266010
Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Original-Tested-by: Yidi Lin <yidi.lin@mediatek.com>
Original-Commit-Queue: Yidi Lin <yidi.lin@mediatek.com>
Reviewed-on: http://review.coreboot.org/10010
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Tested-by: build bot (Jenkins)

3 files changed, 40 insertions, 49 deletions

diff --git a/payloads/libpayload/arch/arm64/mmu.c b/payloads/libpayload/arch/arm64/mmu.c
index 204412efd5..f0781f5ec1 100644
--- a/payloads/libpayload/arch/arm64/mmu.c
+++ b/payloads/libpayload/arch/arm64/mmu.c
@@ -200,12 +200,24 @@ static uint64_t init_xlat_table(uint64_t base_addr,
 
 	/*
 	 * L1 table lookup
-	 * If VA has bits more than 41, lookup starts at L1
+	 * If VA has bits more than L2 can resolve, lookup starts at L1
+	 * Assumption: we don't need L0 table in coreboot
 	 */
-	if (l1_index) {
-		table = get_next_level_table(&table[l1_index]);
-		if (!table)
-			return 0;
+	if (BITS_PER_VA > L1_ADDR_SHIFT) {
+		if ((size >= L1_XLAT_SIZE) &&
+		    IS_ALIGNED(base_addr, (1UL << L1_ADDR_SHIFT))) {
+			/* If block address is aligned and size is greater than
+			 * or equal to size addressed by each L1 entry, we can
+			 * directly store a block desc */
+			desc = base_addr | BLOCK_DESC | attr;
+			table[l1_index] = desc;
+			/* L2 lookup is not required */
+			return L1_XLAT_SIZE;
+		} else {
+			table = get_next_level_table(&table[l1_index]);
+			if (!table)
+				return 0;
+		}
 	}
 
 	/*
@@ -213,10 +225,11 @@ static uint64_t init_xlat_table(uint64_t base_addr,
 	 * If lookup was performed at L1, L2 table addr is obtained from L1 desc
 	 * else, lookup starts at ttbr address
 	 */
-	if (!l3_index && (size >= L2_XLAT_SIZE)) {
+	if ((size >= L2_XLAT_SIZE) &&
+	    IS_ALIGNED(base_addr, (1UL << L2_ADDR_SHIFT))) {
 		/*
 		 * If block address is aligned and size is greater than or equal
-		 * to 512MiB i.e. size addressed by each L2 entry, we can
+		 * to size addressed by each L2 entry, we can
 		 * directly store a block desc
 		 */
 		desc = base_addr | BLOCK_DESC | attr;
@@ -369,7 +382,7 @@ void mmu_enable(void)
 
 	/* Initialize TCR flags */
 	raw_write_tcr_current(TCR_TOSZ | TCR_IRGN0_NM_WBWAC | TCR_ORGN0_NM_WBWAC |
-			      TCR_SH0_IS | TCR_TG0_64KB | TCR_PS_64GB |
+			      TCR_SH0_IS | TCR_TG0_4KB | TCR_PS_64GB |
 			      TCR_TBI_USED);
 
 	/* Initialize TTBR */
diff --git a/payloads/libpayload/include/arm64/arch/mmu.h b/payloads/libpayload/include/arm64/arch/mmu.h
index e241d7593f..79bc783709 100644
--- a/payloads/libpayload/include/arm64/arch/mmu.h
+++ b/payloads/libpayload/include/arm64/arch/mmu.h
@@ -49,14 +49,6 @@ struct mmu_ranges {
  */
 extern char _start[], _end[];
 
-/* IMPORTANT!!!!!!!
- * Assumptions made:
- * Granule size is 64KiB
- * BITS per Virtual address is 33
- * All the calculations for tables L1,L2 and L3 are based on these assumptions
- * If these values are changed, recalculate the other macros as well
- */
-
 /* Memory attributes for mmap regions
  * These attributes act as tag values for memrange regions
  */
@@ -89,47 +81,32 @@ extern char _start[], _end[];
 /* XLAT Table Init Attributes */
 
 #define VA_START                   0x0
-/* If BITS_PER_VA or GRANULE_SIZE are changed, recalculate and change the
-   macros following them */
 #define BITS_PER_VA                33
-/* Granule size of 64KB is being used */
 #define MIN_64_BIT_ADDR            (1UL << 32)
-#define XLAT_TABLE_MASK            ~(0xffffUL)
-#define GRANULE_SIZE_SHIFT         16
+/* Granule size of 4KB is being used */
+#define GRANULE_SIZE_SHIFT         12
 #define GRANULE_SIZE               (1 << GRANULE_SIZE_SHIFT)
-#define GRANULE_SIZE_MASK          ((1 << 16) - 1)
-
-#define L1_ADDR_SHIFT              42
-#define L2_ADDR_SHIFT              29
-#define L3_ADDR_SHIFT              16
-
-#define L1_ADDR_MASK               (0UL << L1_ADDR_SHIFT)
-#define L2_ADDR_MASK               (0xfUL << L2_ADDR_SHIFT)
-#define L3_ADDR_MASK               (0x1fffUL << L3_ADDR_SHIFT)
-
-/* Dependent on BITS_PER_VA and GRANULE_SIZE */
-#define INIT_LEVEL                 2
-#define XLAT_MAX_LEVEL             3
-
-/* Each entry in XLAT table is 8 bytes */
-#define XLAT_ENTRY_SHIFT           3
-#define XLAT_ENTRY_SIZE            (1 << XLAT_ENTRY_SHIFT)
+#define XLAT_TABLE_MASK            (~(0UL) << GRANULE_SIZE_SHIFT)
+#define GRANULE_SIZE_MASK          ((1 << GRANULE_SIZE_SHIFT) - 1)
 
-#define XLAT_TABLE_SHIFT           GRANULE_SIZE_SHIFT
-#define XLAT_TABLE_SIZE            (1 << XLAT_TABLE_SHIFT)
+#define BITS_RESOLVED_PER_LVL   (GRANULE_SIZE_SHIFT - 3)
+#define L1_ADDR_SHIFT           (GRANULE_SIZE_SHIFT + BITS_RESOLVED_PER_LVL * 2)
+#define L2_ADDR_SHIFT           (GRANULE_SIZE_SHIFT + BITS_RESOLVED_PER_LVL * 1)
+#define L3_ADDR_SHIFT           (GRANULE_SIZE_SHIFT + BITS_RESOLVED_PER_LVL * 0)
 
-#define XLAT_NUM_ENTRIES_SHIFT     (XLAT_TABLE_SHIFT - XLAT_ENTRY_SHIFT)
-#define XLAT_NUM_ENTRIES           (1 << XLAT_NUM_ENTRIES_SHIFT)
+#if BITS_PER_VA > L1_ADDR_SHIFT + BITS_RESOLVED_PER_LVL
+  #error "BITS_PER_VA too large (we don't have L0 table support)"
+#endif
 
-#define L3_XLAT_SIZE_SHIFT         (GRANULE_SIZE_SHIFT)
-#define L2_XLAT_SIZE_SHIFT         (GRANULE_SIZE_SHIFT + XLAT_NUM_ENTRIES_SHIFT)
-#define L1_XLAT_SIZE_SHIFT         (GRANULE_SIZE_SHIFT + XLAT_NUM_ENTRIES_SHIFT)
+#define L1_ADDR_MASK     (((1UL << BITS_RESOLVED_PER_LVL) - 1) << L1_ADDR_SHIFT)
+#define L2_ADDR_MASK     (((1UL << BITS_RESOLVED_PER_LVL) - 1) << L2_ADDR_SHIFT)
+#define L3_ADDR_MASK     (((1UL << BITS_RESOLVED_PER_LVL) - 1) << L3_ADDR_SHIFT)
 
 /* These macros give the size of the region addressed by each entry of a xlat
    table at any given level */
-#define L3_XLAT_SIZE               (1 << L3_XLAT_SIZE_SHIFT)
-#define L2_XLAT_SIZE               (1 << L2_XLAT_SIZE_SHIFT)
-#define L1_XLAT_SIZE               (1 << L1_XLAT_SIZE_SHIFT)
+#define L3_XLAT_SIZE               (1UL << L3_ADDR_SHIFT)
+#define L2_XLAT_SIZE               (1UL << L2_ADDR_SHIFT)
+#define L1_XLAT_SIZE               (1UL << L1_ADDR_SHIFT)
 
 /* Block indices required for MAIR */
 #define BLOCK_INDEX_MEM_DEV_NGNRNE 0
@@ -184,7 +161,7 @@ extern char _start[], _end[];
 #define TCR_TBI_USED               (0x0 << TCR_TBI_SHIFT)
 #define TCR_TBI_IGNORED            (0x1 << TCR_TBI_SHIFT)
 
-#define DMA_DEFAULT_SIZE           (0x200 * GRANULE_SIZE)
+#define DMA_DEFAULT_SIZE           (32 * MiB)
 #define TTB_DEFAULT_SIZE           0x100000
 
 #define MB_SIZE			   (1UL << 20)
diff --git a/payloads/libpayload/include/stdlib.h b/payloads/libpayload/include/stdlib.h
index deb5e3dac4..08c7d06c3e 100644
--- a/payloads/libpayload/include/stdlib.h
+++ b/payloads/libpayload/include/stdlib.h
@@ -39,6 +39,7 @@
 #define __ALIGN_MASK(x,mask)    (((x)+(mask))&~(mask))
 #define ALIGN_UP(x,a)           ALIGN((x),(a))
 #define ALIGN_DOWN(x,a)         ((x) & ~((typeof(x))(a)-1UL))
+#define IS_ALIGNED(x,a)         (((x) & ((typeof(x))(a)-1UL)) == 0)
 
 /**
  * @defgroup malloc Memory allocation functions