sh: Support for extended ASIDs on PTEAEX-capable SH-X3 cores.

This adds support for extended ASIDs (up to 16-bits) on newer SH-X3 cores
that implement the PTAEX register and respective functionality. Presently
only the 65nm SH7786 (90nm only supports legacy 8-bit ASIDs).

The main change is in how the PTE is written out when loading the entry
in to the TLB, as well as in how the TLB entry is selectively flushed.

While SH-X2 extended mode splits out the memory-mapped U and I-TLB data
arrays for extra bits, extended ASID mode splits out the address arrays.
While we don't use the memory-mapped data array access, the address
array accesses are necessary for selective TLB flushes, so these are
implemented newly and replace the generic SH-4 implementation.

With this, TLB flushes in switch_mm() are almost non-existent on newer
parts.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

10 files changed, 148 insertions, 28 deletions

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index a0c879d17fd..6c56495fd15 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -365,6 +365,7 @@ config CPU_SUBTYPE_SH7786
 	bool "Support SH7786 processor"
 	select CPU_SH4A
 	select CPU_SHX3
+	select CPU_HAS_PTEAEX
 	select ARCH_SPARSEMEM_ENABLE
 	select SYS_SUPPORTS_NUMA
 
diff --git a/arch/sh/Kconfig.cpu b/arch/sh/Kconfig.cpu
index 0e27fe3b182..c7d704381a6 100644
--- a/arch/sh/Kconfig.cpu
+++ b/arch/sh/Kconfig.cpu
@@ -104,6 +104,9 @@ config CPU_HAS_SR_RB
 config CPU_HAS_PTEA
 	bool
 
+config CPU_HAS_PTEAEX
+	bool
+
 config CPU_HAS_DSP
 	bool
 
diff --git a/arch/sh/include/asm/cpu-features.h b/arch/sh/include/asm/cpu-features.h
index 86308aa3973..694abe490ed 100644
--- a/arch/sh/include/asm/cpu-features.h
+++ b/arch/sh/include/asm/cpu-features.h
@@ -21,5 +21,6 @@
 #define CPU_HAS_LLSC		0x0040	/* movli.l/movco.l */
 #define CPU_HAS_L2_CACHE	0x0080	/* Secondary cache / URAM */
 #define CPU_HAS_OP32		0x0100	/* 32-bit instruction support */
+#define CPU_HAS_PTEAEX		0x0200	/* PTE ASID Extension support */
 
 #endif /* __ASM_SH_CPU_FEATURES_H */
diff --git a/arch/sh/include/asm/mmu_context.h b/arch/sh/include/asm/mmu_context.h
index 5d9157bd474..2a9c55f1a83 100644
--- a/arch/sh/include/asm/mmu_context.h
+++ b/arch/sh/include/asm/mmu_context.h
@@ -19,13 +19,18 @@
  *    (a) TLB cache version (or round, cycle whatever expression you like)
  *    (b) ASID (Address Space IDentifier)
  */
+#ifdef CONFIG_CPU_HAS_PTEAEX
+#define MMU_CONTEXT_ASID_MASK		0x0000ffff
+#else
 #define MMU_CONTEXT_ASID_MASK		0x000000ff
-#define MMU_CONTEXT_VERSION_MASK	0xffffff00
-#define MMU_CONTEXT_FIRST_VERSION	0x00000100
-#define NO_CONTEXT			0UL
+#endif
 
-/* ASID is 8-bit value, so it can't be 0x100 */
-#define MMU_NO_ASID			0x100
+#define MMU_CONTEXT_VERSION_MASK	(~0UL & ~MMU_CONTEXT_ASID_MASK)
+#define MMU_CONTEXT_FIRST_VERSION	(MMU_CONTEXT_ASID_MASK + 1)
+
+/* Impossible ASID value, to differentiate from NO_CONTEXT. */
+#define MMU_NO_ASID			MMU_CONTEXT_FIRST_VERSION
+#define NO_CONTEXT			0UL
 
 #define asid_cache(cpu)		(cpu_data[cpu].asid_cache)
 
diff --git a/arch/sh/include/asm/mmu_context_32.h b/arch/sh/include/asm/mmu_context_32.h
index f4f9aebd68b..8ef800c549a 100644
--- a/arch/sh/include/asm/mmu_context_32.h
+++ b/arch/sh/include/asm/mmu_context_32.h
@@ -10,6 +10,17 @@ static inline void destroy_context(struct mm_struct *mm)
 	/* Do nothing */
 }
 
+#ifdef CONFIG_CPU_HAS_PTEAEX
+static inline void set_asid(unsigned long asid)
+{
+	__raw_writel(asid, MMU_PTEAEX);
+}
+
+static inline unsigned long get_asid(void)
+{
+	return __raw_readl(MMU_PTEAEX) & MMU_CONTEXT_ASID_MASK;
+}
+#else
 static inline void set_asid(unsigned long asid)
 {
 	unsigned long __dummy;
@@ -33,6 +44,7 @@ static inline unsigned long get_asid(void)
 	asid &= MMU_CONTEXT_ASID_MASK;
 	return asid;
 }
+#endif /* CONFIG_CPU_HAS_PTEAEX */
 
 /* MMU_TTB is used for optimizing the fault handling. */
 static inline void set_TTB(pgd_t *pgd)
diff --git a/arch/sh/include/cpu-sh4/cpu/mmu_context.h b/arch/sh/include/cpu-sh4/cpu/mmu_context.h
index 9ea8eb27b18..3ce7ef6c297 100644
--- a/arch/sh/include/cpu-sh4/cpu/mmu_context.h
+++ b/arch/sh/include/cpu-sh4/cpu/mmu_context.h
@@ -14,28 +14,35 @@
 #define MMU_PTEL	0xFF000004	/* Page table entry register LOW */
 #define MMU_TTB		0xFF000008	/* Translation table base register */
 #define MMU_TEA		0xFF00000C	/* TLB Exception Address */
-#define MMU_PTEA	0xFF000034	/* Page table entry assistance register */
+#define MMU_PTEA	0xFF000034	/* PTE assistance register */
+#define MMU_PTEAEX	0xFF00007C	/* PTE ASID extension register */
 
 #define MMUCR		0xFF000010	/* MMU Control Register */
 
-#define MMU_ITLB_ADDRESS_ARRAY	0xF2000000
 #define MMU_UTLB_ADDRESS_ARRAY	0xF6000000
+#define MMU_UTLB_ADDRESS_ARRAY2	0xF6800000
 #define MMU_PAGE_ASSOC_BIT	0x80
 
 #define MMUCR_TI		(1<<2)
 
-#ifdef CONFIG_X2TLB
-#define MMUCR_ME		(1 << 7)
-#else
-#define MMUCR_ME		(0)
-#endif
-
 #if defined(CONFIG_32BIT) && defined(CONFIG_CPU_SUBTYPE_ST40)
 #define MMUCR_SE		(1 << 4)
 #else
 #define MMUCR_SE		(0)
 #endif
 
+#ifdef CONFIG_CPU_HAS_PTEAEX
+#define MMUCR_AEX		(1 << 6)
+#else
+#define MMUCR_AEX		(0)
+#endif
+
+#ifdef CONFIG_X2TLB
+#define MMUCR_ME		(1 << 7)
+#else
+#define MMUCR_ME		(0)
+#endif
+
 #ifdef CONFIG_SH_STORE_QUEUES
 #define MMUCR_SQMD		(1 << 9)
 #else
@@ -43,17 +50,7 @@
 #endif
 
 #define MMU_NTLB_ENTRIES	64
-#define MMU_CONTROL_INIT	(0x05|MMUCR_SQMD|MMUCR_ME|MMUCR_SE)
-
-#define MMU_ITLB_DATA_ARRAY	0xF3000000
-#define MMU_UTLB_DATA_ARRAY	0xF7000000
-
-#define MMU_UTLB_ENTRIES	   64
-#define MMU_U_ENTRY_SHIFT	    8
-#define MMU_UTLB_VALID		0x100
-#define MMU_ITLB_ENTRIES	    4
-#define MMU_I_ENTRY_SHIFT	    8
-#define MMU_ITLB_VALID		0x100
+#define MMU_CONTROL_INIT	(0x05|MMUCR_SQMD|MMUCR_ME|MMUCR_SE|MMUCR_AEX)
 
 #define TRA	0xff000020
 #define EXPEVT	0xff000024
diff --git a/arch/sh/kernel/cpu/sh4/probe.c b/arch/sh/kernel/cpu/sh4/probe.c
index 2bd0ec96263..3d3a3c4425a 100644
--- a/arch/sh/kernel/cpu/sh4/probe.c
+++ b/arch/sh/kernel/cpu/sh4/probe.c
@@ -134,7 +134,7 @@ int __init detect_cpu_and_cache_system(void)
 		boot_cpu_data.icache.ways = 4;
 		boot_cpu_data.dcache.ways = 4;
 		boot_cpu_data.flags |= CPU_HAS_FPU | CPU_HAS_PERF_COUNTER |
-			CPU_HAS_LLSC;
+			CPU_HAS_LLSC | CPU_HAS_PTEAEX;
 		break;
 	case 0x3008:
 		boot_cpu_data.icache.ways = 4;
diff --git a/arch/sh/kernel/setup.c b/arch/sh/kernel/setup.c
index 61ab2a7f864..24c60251f68 100644
--- a/arch/sh/kernel/setup.c
+++ b/arch/sh/kernel/setup.c
@@ -449,7 +449,7 @@ EXPORT_SYMBOL(get_cpu_subtype);
 /* Symbolic CPU flags, keep in sync with asm/cpu-features.h */
 static const char *cpu_flags[] = {
 	"none", "fpu", "p2flush", "mmuassoc", "dsp", "perfctr",
-	"ptea", "llsc", "l2", "op32", NULL
+	"ptea", "llsc", "l2", "op32", "pteaex", NULL
 };
 
 static void show_cpuflags(struct seq_file *m, struct sh_cpuinfo *c)
diff --git a/arch/sh/mm/Makefile_32 b/arch/sh/mm/Makefile_32
index 469ff167245..986a1e05583 100644
--- a/arch/sh/mm/Makefile_32
+++ b/arch/sh/mm/Makefile_32
@@ -25,8 +25,10 @@ obj-$(CONFIG_CPU_SH4)	+= cache-debugfs.o
 endif
 
 ifdef CONFIG_MMU
-obj-$(CONFIG_CPU_SH3)	+= tlb-sh3.o
-obj-$(CONFIG_CPU_SH4)	+= tlb-sh4.o
+tlb-$(CONFIG_CPU_SH3)		:= tlb-sh3.o
+tlb-$(CONFIG_CPU_SH4)		:= tlb-sh4.o
+tlb-$(CONFIG_CPU_HAS_PTEAEX)	:= tlb-pteaex.o
+obj-y				+= $(tlb-y)
 ifndef CONFIG_CACHE_OFF
 obj-$(CONFIG_CPU_SH4)		+= pg-sh4.o
 obj-$(CONFIG_SH7705_CACHE_32KB)	+= pg-sh7705.o
diff --git a/arch/sh/mm/tlb-pteaex.c b/arch/sh/mm/tlb-pteaex.c
new file mode 100644
index 00000000000..5c9b2d781e0
--- /dev/null
+++ b/arch/sh/mm/tlb-pteaex.c
@@ -0,0 +1,99 @@
+/*
+ * arch/sh/mm/tlb-pteaex.c
+ *
+ * TLB operations for SH-X3 CPUs featuring PTE ASID Extensions.
+ *
+ * Copyright (C) 2009 Paul Mundt
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <asm/system.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+void update_mmu_cache(struct vm_area_struct * vma,
+		      unsigned long address, pte_t pte)
+{
+	unsigned long flags;
+	unsigned long pteval;
+	unsigned long vpn;
+
+	/* Ptrace may call this routine. */
+	if (vma && current->active_mm != vma->vm_mm)
+		return;
+
+#ifndef CONFIG_CACHE_OFF
+	{
+		unsigned long pfn = pte_pfn(pte);
+
+		if (pfn_valid(pfn)) {
+			struct page *page = pfn_to_page(pfn);
+
+			if (!test_bit(PG_mapped, &page->flags)) {
+				unsigned long phys = pte_val(pte) & PTE_PHYS_MASK;
+				__flush_wback_region((void *)P1SEGADDR(phys),
+						     PAGE_SIZE);
+				__set_bit(PG_mapped, &page->flags);
+			}
+		}
+	}
+#endif
+
+	local_irq_save(flags);
+
+	/* Set PTEH register */
+	vpn = address & MMU_VPN_MASK;
+	__raw_writel(vpn, MMU_PTEH);
+
+	/* Set PTEAEX */
+	__raw_writel(get_asid(), MMU_PTEAEX);
+
+	pteval = pte.pte_low;
+
+	/* Set PTEA register */
+#ifdef CONFIG_X2TLB
+	/*
+	 * For the extended mode TLB this is trivial, only the ESZ and
+	 * EPR bits need to be written out to PTEA, with the remainder of
+	 * the protection bits (with the exception of the compat-mode SZ
+	 * and PR bits, which are cleared) being written out in PTEL.
+	 */
+	__raw_writel(pte.pte_high, MMU_PTEA);
+#else
+	/* TODO: make this look less hacky */
+	__raw_writel(((pteval >> 28) & 0xe) | (pteval & 0x1), MMU_PTEA);
+#endif
+
+	/* Set PTEL register */
+	pteval &= _PAGE_FLAGS_HARDWARE_MASK; /* drop software flags */
+#ifdef CONFIG_CACHE_WRITETHROUGH
+	pteval |= _PAGE_WT;
+#endif
+	/* conveniently, we want all the software flags to be 0 anyway */
+	__raw_writel(pteval, MMU_PTEL);
+
+	/* Load the TLB */
+	asm volatile("ldtlb": /* no output */ : /* no input */ : "memory");
+	local_irq_restore(flags);
+}
+
+/*
+ * While SH-X2 extended TLB mode splits out the memory-mapped I/UTLB
+ * data arrays, SH-X3 cores with PTEAEX split out the memory-mapped
+ * address arrays. In compat mode the second array is inaccessible, while
+ * in extended mode, the legacy 8-bit ASID field in address array 1 has
+ * undefined behaviour.
+ */
+void __uses_jump_to_uncached local_flush_tlb_one(unsigned long asid,
+						 unsigned long page)
+{
+	jump_to_uncached();
+	__raw_writel(page, MMU_UTLB_ADDRESS_ARRAY | MMU_PAGE_ASSOC_BIT);
+	__raw_writel(asid, MMU_UTLB_ADDRESS_ARRAY2 | MMU_PAGE_ASSOC_BIT);
+	back_to_cached();
+}