csky: Add memory layout 2.5G(user):1.5G(kernel)

There are two ways for translating va to pa for csky:
 - Use TLB(Translate Lookup Buffer) and PTW (Page Table Walk)
 - Use SSEG0/1 (Simple Segment Mapping)

We use tlb mapping 0-2G and 3G-4G virtual address area and SSEG0/1
are for 2G-2.5G and 2.5G-3G translation. We could disable SSEG0
to use 2G-2.5G as TLB user mapping.

Signed-off-by: Guo Ren <guoren@linux.alibaba.com>

6 files changed, 40 insertions, 11 deletions

diff --git a/arch/csky/kernel/atomic.S b/arch/csky/kernel/atomic.S
index 3821ef9b7567..f03db1d8d060 100644
--- a/arch/csky/kernel/atomic.S
+++ b/arch/csky/kernel/atomic.S
@@ -14,6 +14,10 @@
  */
 ENTRY(csky_cmpxchg)
 	USPTOKSP
+
+	RD_MEH	a3
+	WR_MEH	a3
+
 	mfcr	a3, epc
 	addi	a3, TRAP0_SIZE
 
diff --git a/arch/csky/kernel/entry.S b/arch/csky/kernel/entry.S
index 5a5cabd076e1..d5f6d04b21a8 100644
--- a/arch/csky/kernel/entry.S
+++ b/arch/csky/kernel/entry.S
@@ -49,6 +49,7 @@ ENTRY(csky_\name)
 
 	RD_PGDR	r6
 	RD_MEH	a3
+	WR_MEH	a3
 #ifdef CONFIG_CPU_HAS_TLBI
 	tlbi.vaas a3
 	sync.is
@@ -64,10 +65,11 @@ ENTRY(csky_\name)
 	WR_MCIR	a2
 #endif
 	bclri   r6, 0
+	lrw	a2, PAGE_OFFSET
+	add	r6, a2
 	lrw	a2, va_pa_offset
 	ld.w	a2, (a2, 0)
 	subu	r6, a2
-	bseti	r6, 31
 
 	mov     a2, a3
 	lsri    a2, _PGDIR_SHIFT
@@ -75,10 +77,11 @@ ENTRY(csky_\name)
 	addu    r6, a2
 	ldw     r6, (r6)
 
+	lrw	a2, PAGE_OFFSET
+	add	r6, a2
 	lrw	a2, va_pa_offset
 	ld.w	a2, (a2, 0)
 	subu	r6, a2
-	bseti	r6, 31
 
 	lsri    a3, PTE_INDX_SHIFT
 	lrw     a2, PTE_INDX_MSK
@@ -314,6 +317,9 @@ ENTRY(csky_trap)
 ENTRY(csky_get_tls)
 	USPTOKSP
 
+	RD_MEH	a0
+	WR_MEH	a0
+
 	/* increase epc for continue */
 	mfcr	a0, epc
 	addi	a0, TRAP0_SIZE
diff --git a/arch/csky/kernel/head.S b/arch/csky/kernel/head.S
index 17ed9d250480..7e3e4f15b052 100644
--- a/arch/csky/kernel/head.S
+++ b/arch/csky/kernel/head.S
@@ -21,10 +21,16 @@ END(_start)
 ENTRY(_start_smp_secondary)
 	SETUP_MMU
 
-	/* copy msa1 from CPU0 */
-	lrw     r6, secondary_msa1
+#ifdef CONFIG_PAGE_OFFSET_80000000
+	lrw	r6, secondary_msa1
 	ld.w	r6, (r6, 0)
 	mtcr	r6, cr<31, 15>
+#endif
+
+	lrw     r6, secondary_pgd
+	ld.w	r6, (r6, 0)
+	mtcr	r6, cr<28, 15>
+	mtcr	r6, cr<29, 15>
 
 	/* set stack point */
 	lrw     r6, secondary_stack
diff --git a/arch/csky/kernel/setup.c b/arch/csky/kernel/setup.c
index e4cab16056d6..e93bc6f74432 100644
--- a/arch/csky/kernel/setup.c
+++ b/arch/csky/kernel/setup.c
@@ -45,13 +45,17 @@ static void __init csky_memblock_init(void)
 
 	if (size >= lowmem_size) {
 		max_low_pfn = min_low_pfn + lowmem_size;
+#ifdef CONFIG_PAGE_OFFSET_80000000
 		write_mmu_msa1(read_mmu_msa0() + SSEG_SIZE);
+#endif
 	} else if (size > sseg_size) {
 		max_low_pfn = min_low_pfn + sseg_size;
 	}
 
 	max_zone_pfn[ZONE_NORMAL] = max_low_pfn;
 
+	mmu_init(min_low_pfn, max_low_pfn);
+
 #ifdef CONFIG_HIGHMEM
 	max_zone_pfn[ZONE_HIGHMEM] = max_pfn;
 
@@ -101,16 +105,26 @@ void __init setup_arch(char **cmdline_p)
 unsigned long va_pa_offset;
 EXPORT_SYMBOL(va_pa_offset);
 
+static inline unsigned long read_mmu_msa(void)
+{
+#ifdef CONFIG_PAGE_OFFSET_80000000
+	return read_mmu_msa0();
+#endif
+
+#ifdef CONFIG_PAGE_OFFSET_A0000000
+	return read_mmu_msa1();
+#endif
+}
+
 asmlinkage __visible void __init csky_start(unsigned int unused,
 					    void *dtb_start)
 {
 	/* Clean up bss section */
 	memset(__bss_start, 0, __bss_stop - __bss_start);
 
-	va_pa_offset = read_mmu_msa0() & ~(SSEG_SIZE - 1);
+	va_pa_offset = read_mmu_msa() & ~(SSEG_SIZE - 1);
 
 	pre_trap_init();
-	pre_mmu_init();
 
 	if (dtb_start == NULL)
 		early_init_dt_scan(__dtb_start);
diff --git a/arch/csky/kernel/smp.c b/arch/csky/kernel/smp.c
index 041d0de6a1b6..0f9f5eef9338 100644
--- a/arch/csky/kernel/smp.c
+++ b/arch/csky/kernel/smp.c
@@ -203,8 +203,8 @@ volatile unsigned int secondary_hint;
 volatile unsigned int secondary_hint2;
 volatile unsigned int secondary_ccr;
 volatile unsigned int secondary_stack;
-
-unsigned long secondary_msa1;
+volatile unsigned int secondary_msa1;
+volatile unsigned int secondary_pgd;
 
 int __cpu_up(unsigned int cpu, struct task_struct *tidle)
 {
@@ -216,6 +216,7 @@ int __cpu_up(unsigned int cpu, struct task_struct *tidle)
 	secondary_hint2 = mfcr("cr<21, 1>");
 	secondary_ccr  = mfcr("cr18");
 	secondary_msa1 = read_mmu_msa1();
+	secondary_pgd = mfcr("cr<29, 15>");
 
 	/*
 	 * Because other CPUs are in reset status, we must flush data
@@ -262,8 +263,6 @@ void csky_start_secondary(void)
 
 	flush_tlb_all();
 	write_mmu_pagemask(0);
-	TLBMISS_HANDLER_SETUP_PGD(swapper_pg_dir);
-	TLBMISS_HANDLER_SETUP_PGD_KERNEL(swapper_pg_dir);
 
 #ifdef CONFIG_CPU_HAS_FPU
 	init_fpu();
diff --git a/arch/csky/kernel/vmlinux.lds.S b/arch/csky/kernel/vmlinux.lds.S
index f03033e17c29..e8b1a4a49798 100644
--- a/arch/csky/kernel/vmlinux.lds.S
+++ b/arch/csky/kernel/vmlinux.lds.S
@@ -33,6 +33,7 @@ SECTIONS
 
 	.text : AT(ADDR(.text) - LOAD_OFFSET) {
 		_text = .;
+		VBR_BASE
 		IRQENTRY_TEXT
 		SOFTIRQENTRY_TEXT
 		TEXT_TEXT
@@ -104,7 +105,6 @@ SECTIONS
 
 	EXCEPTION_TABLE(L1_CACHE_BYTES)
 	BSS_SECTION(L1_CACHE_BYTES, PAGE_SIZE, L1_CACHE_BYTES)
-	VBR_BASE
 	_end = . ;
 
 	STABS_DEBUG