arc: cache - accommodate different L1 cache line lengths

ARC core could be configured with different L1 and L2 (AKA SLC) cache
line lengths. At least these values are possible and were really used:
32, 64 or 128 bytes.

Current implementation requires cache line to be selected upon U-Boot
configuration and then it will only work on matching hardware. Indeed
this is quite efficient because cache line length gets hardcoded during
code compilation. But OTOH it makes binary less portable.

With this commit we allow U-Boot to determine real L1 cache line length
early in runtime and use this value later on. This extends portability
of U-Boot binary a lot.

Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>

3 files changed, 89 insertions, 54 deletions

diff --git a/arch/arc/Kconfig b/arch/arc/Kconfig
index 264c83d..f1dc6c8 100644
--- a/arch/arc/Kconfig
+++ b/arch/arc/Kconfig
@@ -116,17 +116,6 @@ config SYS_DCACHE_OFF
 	bool "Do not use Data Cache"
 	default n
 
-config ARC_CACHE_LINE_SHIFT
-	int "Cache Line Length (as power of 2)"
-	range 5 7
-	default "6"
-	depends on !SYS_DCACHE_OFF || !SYS_ICACHE_OFF
-	help
-	  Starting with ARC700 4.9, Cache line length is configurable,
-	  This option specifies "N", with Line-len = 2 power N
-	  So line lengths of 32, 64, 128 are specified by 5,6,7, respectively
-	  Linux only supports same line lengths for I and D caches.
-
 choice
 	prompt "Target select"
 	default TARGET_AXS101
diff --git a/arch/arc/include/asm/cache.h b/arch/arc/include/asm/cache.h
index 432606a..d26d9fb 100644
--- a/arch/arc/include/asm/cache.h
+++ b/arch/arc/include/asm/cache.h
@@ -9,13 +9,13 @@
 
 #include <config.h>
 
-#ifdef CONFIG_ARC_CACHE_LINE_SHIFT
-#define CONFIG_SYS_CACHELINE_SIZE	(1 << CONFIG_ARC_CACHE_LINE_SHIFT)
-#define ARCH_DMA_MINALIGN		CONFIG_SYS_CACHELINE_SIZE
-#else
-/* Satisfy users of ARCH_DMA_MINALIGN */
-#define ARCH_DMA_MINALIGN		128
-#endif
+/*
+ * As of today we may handle any L1 cache line length right in software.
+ * For that essentially cache line length is a variable not constant.
+ * And to satisfy users of ARCH_DMA_MINALIGN we just use largest line length
+ * that may exist in either L1 or L2 (AKA SLC) caches on ARC.
+ */
+#define ARCH_DMA_MINALIGN	128
 
 #if defined(ARC_MMU_ABSENT)
 #define CONFIG_ARC_MMU_VER 0
diff --git a/arch/arc/lib/cache.c b/arch/arc/lib/cache.c
index ed8e8e7..8089f9d 100644
--- a/arch/arc/lib/cache.c
+++ b/arch/arc/lib/cache.c
@@ -5,13 +5,12 @@
  */
 
 #include <config.h>
+#include <common.h>
 #include <linux/compiler.h>
 #include <linux/kernel.h>
 #include <asm/arcregs.h>
 #include <asm/cache.h>
 
-#define CACHE_LINE_MASK		(~(CONFIG_SYS_CACHELINE_SIZE - 1))
-
 /* Bit values in IC_CTRL */
 #define IC_CTRL_CACHE_DISABLE	(1 << 0)
 
@@ -26,12 +25,18 @@
 #define OP_FLUSH	0x2
 #define OP_INV_IC	0x3
 
-#ifdef CONFIG_ISA_ARCV2
 /*
  * By default that variable will fall into .bss section.
  * But .bss section is not relocated and so it will be initilized before
  * relocation but will be used after being zeroed.
  */
+int l1_line_sz __section(".data");
+int dcache_exists __section(".data");
+int icache_exists __section(".data");
+
+#define CACHE_LINE_MASK		(~(l1_line_sz - 1))
+
+#ifdef CONFIG_ISA_ARCV2
 int slc_line_sz __section(".data");
 int slc_exists __section(".data");
 
@@ -111,46 +116,87 @@ static inline void __slc_line_op(unsigned long paddr, unsigned long sz,
 #define __slc_line_op(paddr, sz, cacheop)
 #endif
 
-static inline int icache_exists(void)
+#ifdef CONFIG_ISA_ARCV2
+static void read_decode_cache_bcr_arcv2(void)
 {
-	/* Check if Instruction Cache is available */
-	if (read_aux_reg(ARC_BCR_IC_BUILD) & CACHE_VER_NUM_MASK)
-		return 1;
-	else
-		return 0;
+	union {
+		struct {
+#ifdef CONFIG_CPU_BIG_ENDIAN
+			unsigned int pad:24, way:2, lsz:2, sz:4;
+#else
+			unsigned int sz:4, lsz:2, way:2, pad:24;
+#endif
+		} fields;
+		unsigned int word;
+	} slc_cfg;
+
+	union {
+		struct {
+#ifdef CONFIG_CPU_BIG_ENDIAN
+			unsigned int pad:24, ver:8;
+#else
+			unsigned int ver:8, pad:24;
+#endif
+		} fields;
+		unsigned int word;
+	} sbcr;
+
+	sbcr.word = read_aux_reg(ARC_BCR_SLC);
+	if (sbcr.fields.ver) {
+		slc_cfg.word = read_aux_reg(ARC_AUX_SLC_CONFIG);
+		slc_exists = 1;
+		slc_line_sz = (slc_cfg.fields.lsz == 0) ? 128 : 64;
+	}
 }
+#endif
 
-static inline int dcache_exists(void)
+void read_decode_cache_bcr(void)
 {
-	/* Check if Data Cache is available */
-	if (read_aux_reg(ARC_BCR_DC_BUILD) & CACHE_VER_NUM_MASK)
-		return 1;
-	else
-		return 0;
+	int dc_line_sz = 0, ic_line_sz = 0;
+
+	union {
+		struct {
+#ifdef CONFIG_CPU_BIG_ENDIAN
+			unsigned int pad:12, line_len:4, sz:4, config:4, ver:8;
+#else
+			unsigned int ver:8, config:4, sz:4, line_len:4, pad:12;
+#endif
+		} fields;
+		unsigned int word;
+	} ibcr, dbcr;
+
+	ibcr.word = read_aux_reg(ARC_BCR_IC_BUILD);
+	if (ibcr.fields.ver) {
+		icache_exists = 1;
+		l1_line_sz = ic_line_sz = 8 << ibcr.fields.line_len;
+		if (!ic_line_sz)
+			panic("Instruction exists but line length is 0\n");
+	}
+
+	dbcr.word = read_aux_reg(ARC_BCR_DC_BUILD);
+	if (dbcr.fields.ver){
+		dcache_exists = 1;
+		l1_line_sz = dc_line_sz = 16 << dbcr.fields.line_len;
+		if (!dc_line_sz)
+			panic("Data cache exists but line length is 0\n");
+	}
+
+	if (ic_line_sz && dc_line_sz && (ic_line_sz != dc_line_sz))
+		panic("Instruction and data cache line lengths differ\n");
 }
 
 void cache_init(void)
 {
+	read_decode_cache_bcr();
+
 #ifdef CONFIG_ISA_ARCV2
-	/* Check if System-Level Cache (SLC) is available */
-	if (read_aux_reg(ARC_BCR_SLC) & CACHE_VER_NUM_MASK) {
-#define LSIZE_OFFSET	4
-#define LSIZE_MASK	3
-		if (read_aux_reg(ARC_AUX_SLC_CONFIG) &
-		    (LSIZE_MASK << LSIZE_OFFSET))
-			slc_line_sz = 64;
-		else
-			slc_line_sz = 128;
-		slc_exists = 1;
-	} else {
-		slc_exists = 0;
-	}
+	read_decode_cache_bcr_arcv2();
 #endif
 }
 
 int icache_status(void)
 {
-	if (!icache_exists())
+	if (!icache_exists)
 		return 0;
 
 	if (read_aux_reg(ARC_AUX_IC_CTRL) & IC_CTRL_CACHE_DISABLE)
@@ -161,14 +207,14 @@ int icache_status(void)
 
 void icache_enable(void)
 {
-	if (icache_exists())
+	if (icache_exists)
 		write_aux_reg(ARC_AUX_IC_CTRL, read_aux_reg(ARC_AUX_IC_CTRL) &
 			      ~IC_CTRL_CACHE_DISABLE);
 }
 
 void icache_disable(void)
 {
-	if (icache_exists())
+	if (icache_exists)
 		write_aux_reg(ARC_AUX_IC_CTRL, read_aux_reg(ARC_AUX_IC_CTRL) |
 			      IC_CTRL_CACHE_DISABLE);
 }
@@ -190,7 +236,7 @@ void invalidate_icache_all(void)
 
 int dcache_status(void)
 {
-	if (!dcache_exists())
+	if (!dcache_exists)
 		return 0;
 
 	if (read_aux_reg(ARC_AUX_DC_CTRL) & DC_CTRL_CACHE_DISABLE)
@@ -201,7 +247,7 @@ int dcache_status(void)
 
 void dcache_enable(void)
 {
-	if (!dcache_exists())
+	if (!dcache_exists)
 		return;
 
 	write_aux_reg(ARC_AUX_DC_CTRL, read_aux_reg(ARC_AUX_DC_CTRL) &
@@ -210,7 +256,7 @@ void dcache_enable(void)
 
 void dcache_disable(void)
 {
-	if (!dcache_exists())
+	if (!dcache_exists)
 		return;
 
 	write_aux_reg(ARC_AUX_DC_CTRL, read_aux_reg(ARC_AUX_DC_CTRL) |
@@ -246,14 +292,14 @@ static inline void __cache_line_loop(unsigned long paddr, unsigned long sz,
 	sz += paddr & ~CACHE_LINE_MASK;
 	paddr &= CACHE_LINE_MASK;
 
-	num_lines = DIV_ROUND_UP(sz, CONFIG_SYS_CACHELINE_SIZE);
+	num_lines = DIV_ROUND_UP(sz, l1_line_sz);
 
 	while (num_lines-- > 0) {
 #if (CONFIG_ARC_MMU_VER == 3)
 		write_aux_reg(aux_tag, paddr);
 #endif
 		write_aux_reg(aux_cmd, paddr);
-		paddr += CONFIG_SYS_CACHELINE_SIZE;
+		paddr += l1_line_sz;
 	}
 }