POWERPC: Allow 32-bit hashed pgtable code to support 36-bit physical

This rearranges a bit of code, and adds support for
36-bit physical addressing for configs that use a
hashed page table.  The 36b physical support is not
enabled by default on any config - it must be
explicitly enabled via the config system.

This patch *only* expands the page table code to accomodate
large physical addresses on 32-bit systems and enables the
PHYS_64BIT config option for 86xx.  It does *not*
allow you to boot a board with more than about 3.5GB of
RAM - for that, SWIOTLB support is also required (and
coming soon).

Signed-off-by: Becky Bruce <becky.bruce@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>

10 files changed, 109 insertions, 33 deletions

diff --git a/arch/powerpc/include/asm/io.h b/arch/powerpc/include/asm/io.h
index 77c7fa025e6..08266d2728b 100644
--- a/arch/powerpc/include/asm/io.h
+++ b/arch/powerpc/include/asm/io.h
@@ -711,7 +711,7 @@ static inline void * phys_to_virt(unsigned long address)
 /*
  * Change "struct page" to physical address.
  */
-#define page_to_phys(page)	(page_to_pfn(page) << PAGE_SHIFT)
+#define page_to_phys(page)	((phys_addr_t)page_to_pfn(page) << PAGE_SHIFT)
 
 /* We do NOT want virtual merging, it would put too much pressure on
  * our iommu allocator. Instead, we want drivers to be smart enough
diff --git a/arch/powerpc/include/asm/page_32.h b/arch/powerpc/include/asm/page_32.h
index ebfae530a37..d77072a32cc 100644
--- a/arch/powerpc/include/asm/page_32.h
+++ b/arch/powerpc/include/asm/page_32.h
@@ -13,10 +13,16 @@
 #define ARCH_KMALLOC_MINALIGN	L1_CACHE_BYTES
 #endif
 
+#ifdef CONFIG_PTE_64BIT
+#define PTE_FLAGS_OFFSET	4	/* offset of PTE flags, in bytes */
+#else
+#define PTE_FLAGS_OFFSET	0
+#endif
+
 #ifndef __ASSEMBLY__
 /*
  * The basic type of a PTE - 64 bits for those CPUs with > 32 bit
- * physical addressing.  For now this just the IBM PPC440.
+ * physical addressing.
  */
 #ifdef CONFIG_PTE_64BIT
 typedef unsigned long long pte_basic_t;
diff --git a/arch/powerpc/include/asm/pgtable-ppc32.h b/arch/powerpc/include/asm/pgtable-ppc32.h
index c2e58b41bc7..29c83d85b04 100644
--- a/arch/powerpc/include/asm/pgtable-ppc32.h
+++ b/arch/powerpc/include/asm/pgtable-ppc32.h
@@ -369,7 +369,12 @@ extern int icache_44x_need_flush;
 #define _PAGE_RW	0x400	/* software: user write access allowed */
 #define _PAGE_SPECIAL	0x800	/* software: Special page */
 
+#ifdef CONFIG_PTE_64BIT
+/* We never clear the high word of the pte */
+#define _PTE_NONE_MASK	(0xffffffff00000000ULL | _PAGE_HASHPTE)
+#else
 #define _PTE_NONE_MASK	_PAGE_HASHPTE
+#endif
 
 #define _PMD_PRESENT	0
 #define _PMD_PRESENT_MASK (PAGE_MASK)
@@ -587,6 +592,10 @@ extern int flush_hash_pages(unsigned context, unsigned long va,
 extern void add_hash_page(unsigned context, unsigned long va,
 			  unsigned long pmdval);
 
+/* Flush an entry from the TLB/hash table */
+extern void flush_hash_entry(struct mm_struct *mm, pte_t *ptep,
+			     unsigned long address);
+
 /*
  * Atomic PTE updates.
  *
@@ -665,9 +674,13 @@ static inline unsigned long long pte_update(pte_t *p,
 static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
 			      pte_t *ptep, pte_t pte)
 {
-#if _PAGE_HASHPTE != 0
+#if (_PAGE_HASHPTE != 0) && defined(CONFIG_SMP) && !defined(CONFIG_PTE_64BIT)
 	pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte) & ~_PAGE_HASHPTE);
 #elif defined(CONFIG_PTE_64BIT) && defined(CONFIG_SMP)
+#if _PAGE_HASHPTE != 0
+	if (pte_val(*ptep) & _PAGE_HASHPTE)
+		flush_hash_entry(mm, ptep, addr);
+#endif
 	__asm__ __volatile__("\
 		stw%U0%X0 %2,%0\n\
 		eieio\n\
@@ -675,7 +688,7 @@ static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
 	: "=m" (*ptep), "=m" (*((unsigned char *)ptep+4))
 	: "r" (pte) : "memory");
 #else
-	*ptep = pte;
+	*ptep = (*ptep & _PAGE_HASHPTE) | (pte & ~_PAGE_HASHPTE);
 #endif
 }
 
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index e9c4044012b..09febc58258 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -352,6 +352,7 @@ int main(void)
 #endif
 
 	DEFINE(PGD_TABLE_SIZE, PGD_TABLE_SIZE);
+	DEFINE(PTE_SIZE, sizeof(pte_t));
 
 #ifdef CONFIG_KVM
 	DEFINE(TLBE_BYTES, sizeof(struct tlbe));
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 8bb65751929..a6de6dbc5ed 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -369,13 +369,13 @@ i##n:								\
 DataAccess:
 	EXCEPTION_PROLOG
 	mfspr	r10,SPRN_DSISR
+	stw	r10,_DSISR(r11)
 	andis.	r0,r10,0xa470		/* weird error? */
 	bne	1f			/* if not, try to put a PTE */
 	mfspr	r4,SPRN_DAR		/* into the hash table */
 	rlwinm	r3,r10,32-15,21,21	/* DSISR_STORE -> _PAGE_RW */
 	bl	hash_page
-1:	stw	r10,_DSISR(r11)
-	mr	r5,r10
+1:	lwz	r5,_DSISR(r11)		/* get DSISR value */
 	mfspr	r4,SPRN_DAR
 	EXC_XFER_EE_LITE(0x300, handle_page_fault)
 
diff --git a/arch/powerpc/kernel/head_fsl_booke.S b/arch/powerpc/kernel/head_fsl_booke.S
index 377e0c155c9..18c0093f932 100644
--- a/arch/powerpc/kernel/head_fsl_booke.S
+++ b/arch/powerpc/kernel/head_fsl_booke.S
@@ -422,7 +422,6 @@ skpinv:	addi	r6,r6,1				/* Increment */
  *   r12 is pointer to the pte
  */
 #ifdef CONFIG_PTE_64BIT
-#define PTE_FLAGS_OFFSET	4
 #define FIND_PTE	\
 	rlwinm	r12, r10, 13, 19, 29;	/* Compute pgdir/pmd offset */	\
 	lwzx	r11, r12, r11;		/* Get pgd/pmd entry */		\
@@ -431,7 +430,6 @@ skpinv:	addi	r6,r6,1				/* Increment */
 	rlwimi	r12, r10, 23, 20, 28;	/* Compute pte address */	\
 	lwz	r11, 4(r12);		/* Get pte entry */
 #else
-#define PTE_FLAGS_OFFSET	0
 #define FIND_PTE	\
 	rlwimi	r11, r10, 12, 20, 29;	/* Create L1 (pgdir/pmd) address */	\
 	lwz	r11, 0(r11);		/* Get L1 entry */			\
diff --git a/arch/powerpc/mm/hash_low_32.S b/arch/powerpc/mm/hash_low_32.S
index c41d658176a..7bffb70b9fe 100644
--- a/arch/powerpc/mm/hash_low_32.S
+++ b/arch/powerpc/mm/hash_low_32.S
@@ -75,7 +75,7 @@ _GLOBAL(hash_page_sync)
  * Returns to the caller if the access is illegal or there is no
  * mapping for the address.  Otherwise it places an appropriate PTE
  * in the hash table and returns from the exception.
- * Uses r0, r3 - r8, ctr, lr.
+ * Uses r0, r3 - r8, r10, ctr, lr.
  */
 	.text
 _GLOBAL(hash_page)
@@ -106,9 +106,15 @@ _GLOBAL(hash_page)
 	addi	r5,r5,swapper_pg_dir@l	/* kernel page table */
 	rlwimi	r3,r9,32-12,29,29	/* MSR_PR -> _PAGE_USER */
 112:	add	r5,r5,r7		/* convert to phys addr */
+#ifndef CONFIG_PTE_64BIT
 	rlwimi	r5,r4,12,20,29		/* insert top 10 bits of address */
 	lwz	r8,0(r5)		/* get pmd entry */
 	rlwinm.	r8,r8,0,0,19		/* extract address of pte page */
+#else
+	rlwinm	r8,r4,13,19,29		/* Compute pgdir/pmd offset */
+	lwzx	r8,r8,r5		/* Get L1 entry */
+	rlwinm.	r8,r8,0,0,20		/* extract pt base address */
+#endif
 #ifdef CONFIG_SMP
 	beq-	hash_page_out		/* return if no mapping */
 #else
@@ -118,7 +124,11 @@ _GLOBAL(hash_page)
 	   to the address following the rfi. */
 	beqlr-
 #endif
+#ifndef CONFIG_PTE_64BIT
 	rlwimi	r8,r4,22,20,29		/* insert next 10 bits of address */
+#else
+	rlwimi	r8,r4,23,20,28		/* compute pte address */
+#endif
 	rlwinm	r0,r3,32-3,24,24	/* _PAGE_RW access -> _PAGE_DIRTY */
 	ori	r0,r0,_PAGE_ACCESSED|_PAGE_HASHPTE
 
@@ -127,9 +137,15 @@ _GLOBAL(hash_page)
 	 * because almost always, there won't be a permission violation
 	 * and there won't already be an HPTE, and thus we will have
 	 * to update the PTE to set _PAGE_HASHPTE.  -- paulus.
+	 *
+	 * If PTE_64BIT is set, the low word is the flags word; use that
+	 * word for locking since it contains all the interesting bits.
 	 */
+#if (PTE_FLAGS_OFFSET != 0)
+	addi	r8,r8,PTE_FLAGS_OFFSET
+#endif
 retry:
-	lwarx	r6,0,r8			/* get linux-style pte */
+	lwarx	r6,0,r8			/* get linux-style pte, flag word */
 	andc.	r5,r3,r6		/* check access & ~permission */
 #ifdef CONFIG_SMP
 	bne-	hash_page_out		/* return if access not permitted */
@@ -137,6 +153,15 @@ retry:
 	bnelr-
 #endif
 	or	r5,r0,r6		/* set accessed/dirty bits */
+#ifdef CONFIG_PTE_64BIT
+#ifdef CONFIG_SMP
+	subf	r10,r6,r8		/* create false data dependency */
+	subi	r10,r10,PTE_FLAGS_OFFSET
+	lwzx	r10,r6,r10		/* Get upper PTE word */
+#else
+	lwz	r10,-PTE_FLAGS_OFFSET(r8)
+#endif /* CONFIG_SMP */
+#endif /* CONFIG_PTE_64BIT */
 	stwcx.	r5,0,r8			/* attempt to update PTE */
 	bne-	retry			/* retry if someone got there first */
 
@@ -203,9 +228,9 @@ _GLOBAL(add_hash_page)
 	 * we can't take a hash table miss (assuming the code is
 	 * covered by a BAT).  -- paulus
 	 */
-	mfmsr	r10
+	mfmsr	r9
 	SYNC
-	rlwinm	r0,r10,0,17,15		/* clear bit 16 (MSR_EE) */
+	rlwinm	r0,r9,0,17,15		/* clear bit 16 (MSR_EE) */
 	rlwinm	r0,r0,0,28,26		/* clear MSR_DR */
 	mtmsr	r0
 	SYNC_601
@@ -214,14 +239,14 @@ _GLOBAL(add_hash_page)
 	tophys(r7,0)
 
 #ifdef CONFIG_SMP
-	addis	r9,r7,mmu_hash_lock@ha
-	addi	r9,r9,mmu_hash_lock@l
-10:	lwarx	r0,0,r9			/* take the mmu_hash_lock */
+	addis	r6,r7,mmu_hash_lock@ha
+	addi	r6,r6,mmu_hash_lock@l
+10:	lwarx	r0,0,r6			/* take the mmu_hash_lock */
 	cmpi	0,r0,0
 	bne-	11f
-	stwcx.	r8,0,r9
+	stwcx.	r8,0,r6
 	beq+	12f
-11:	lwz	r0,0(r9)
+11:	lwz	r0,0(r6)
 	cmpi	0,r0,0
 	beq	10b
 	b	11b
@@ -234,10 +259,24 @@ _GLOBAL(add_hash_page)
 	 * HPTE, so we just unlock and return.
 	 */
 	mr	r8,r5
+#ifndef CONFIG_PTE_64BIT
 	rlwimi	r8,r4,22,20,29
+#else
+	rlwimi	r8,r4,23,20,28
+	addi	r8,r8,PTE_FLAGS_OFFSET
+#endif
 1:	lwarx	r6,0,r8
 	andi.	r0,r6,_PAGE_HASHPTE
 	bne	9f			/* if HASHPTE already set, done */
+#ifdef CONFIG_PTE_64BIT
+#ifdef CONFIG_SMP
+	subf	r10,r6,r8		/* create false data dependency */
+	subi	r10,r10,PTE_FLAGS_OFFSET
+	lwzx	r10,r6,r10		/* Get upper PTE word */
+#else
+	lwz	r10,-PTE_FLAGS_OFFSET(r8)
+#endif /* CONFIG_SMP */
+#endif /* CONFIG_PTE_64BIT */
 	ori	r5,r6,_PAGE_HASHPTE
 	stwcx.	r5,0,r8
 	bne-	1b
@@ -246,13 +285,15 @@ _GLOBAL(add_hash_page)
 
 9:
 #ifdef CONFIG_SMP
+	addis	r6,r7,mmu_hash_lock@ha
+	addi	r6,r6,mmu_hash_lock@l
 	eieio
 	li	r0,0
-	stw	r0,0(r9)		/* clear mmu_hash_lock */
+	stw	r0,0(r6)		/* clear mmu_hash_lock */
 #endif
 
 	/* reenable interrupts and DR */
-	mtmsr	r10
+	mtmsr	r9
 	SYNC_601
 	isync
 
@@ -267,7 +308,8 @@ _GLOBAL(add_hash_page)
  * r5 contains the linux PTE, r6 contains the old value of the
  * linux PTE (before setting _PAGE_HASHPTE) and r7 contains the
  * offset to be added to addresses (0 if the MMU is on,
- * -KERNELBASE if it is off).
+ * -KERNELBASE if it is off).  r10 contains the upper half of
+ * the PTE if CONFIG_PTE_64BIT.
  * On SMP, the caller should have the mmu_hash_lock held.
  * We assume that the caller has (or will) set the _PAGE_HASHPTE
  * bit in the linux PTE in memory.  The value passed in r6 should
@@ -313,6 +355,11 @@ _GLOBAL(create_hpte)
 BEGIN_FTR_SECTION
 	ori	r8,r8,_PAGE_COHERENT	/* set M (coherence required) */
 END_FTR_SECTION_IFSET(CPU_FTR_NEED_COHERENT)
+#ifdef CONFIG_PTE_64BIT
+	/* Put the XPN bits into the PTE */
+	rlwimi	r8,r10,8,20,22
+	rlwimi	r8,r10,2,29,29
+#endif
 
 	/* Construct the high word of the PPC-style PTE (r5) */
 	rlwinm	r5,r3,7,1,24		/* put VSID in 0x7fffff80 bits */
@@ -499,14 +546,18 @@ _GLOBAL(flush_hash_pages)
 	isync
 
 	/* First find a PTE in the range that has _PAGE_HASHPTE set */
+#ifndef CONFIG_PTE_64BIT
 	rlwimi	r5,r4,22,20,29
-1:	lwz	r0,0(r5)
+#else
+	rlwimi	r5,r4,23,20,28
+#endif
+1:	lwz	r0,PTE_FLAGS_OFFSET(r5)
 	cmpwi	cr1,r6,1
 	andi.	r0,r0,_PAGE_HASHPTE
 	bne	2f
 	ble	cr1,19f
 	addi	r4,r4,0x1000
-	addi	r5,r5,4
+	addi	r5,r5,PTE_SIZE
 	addi	r6,r6,-1
 	b	1b
 
@@ -545,7 +596,10 @@ _GLOBAL(flush_hash_pages)
 	 * already clear, we're done (for this pte).  If not,
 	 * clear it (atomically) and proceed.  -- paulus.
 	 */
-33:	lwarx	r8,0,r5			/* fetch the pte */
+#if (PTE_FLAGS_OFFSET != 0)
+	addi	r5,r5,PTE_FLAGS_OFFSET
+#endif
+33:	lwarx	r8,0,r5			/* fetch the pte flags word */
 	andi.	r0,r8,_PAGE_HASHPTE
 	beq	8f			/* done if HASHPTE is already clear */
 	rlwinm	r8,r8,0,31,29		/* clear HASHPTE bit */
@@ -590,7 +644,7 @@ _GLOBAL(flush_hash_patch_B)
 
 8:	ble	cr1,9f			/* if all ptes checked */
 81:	addi	r6,r6,-1
-	addi	r5,r5,4			/* advance to next pte */
+	addi	r5,r5,PTE_SIZE
 	addi	r4,r4,0x1000
 	lwz	r0,0(r5)		/* check next pte */
 	cmpwi	cr1,r6,1
diff --git a/arch/powerpc/mm/pgtable_32.c b/arch/powerpc/mm/pgtable_32.c
index 2001abdb191..c31d6d26f0b 100644
--- a/arch/powerpc/mm/pgtable_32.c
+++ b/arch/powerpc/mm/pgtable_32.c
@@ -73,7 +73,7 @@ extern unsigned long p_mapped_by_tlbcam(unsigned long pa);
 #endif /* HAVE_TLBCAM */
 
 #ifdef CONFIG_PTE_64BIT
-/* 44x uses an 8kB pgdir because it has 8-byte Linux PTEs. */
+/* Some processors use an 8kB pgdir because they have 8-byte Linux PTEs. */
 #define PGDIR_ORDER	1
 #else
 #define PGDIR_ORDER	0
@@ -288,7 +288,7 @@ int map_page(unsigned long va, phys_addr_t pa, int flags)
 }
 
 /*
- * Map in all of physical memory starting at KERNELBASE.
+ * Map in a big chunk of physical memory starting at KERNELBASE.
  */
 void __init mapin_ram(void)
 {
diff --git a/arch/powerpc/mm/tlb_32.c b/arch/powerpc/mm/tlb_32.c
index eb4b512d65f..f9a47fee392 100644
--- a/arch/powerpc/mm/tlb_32.c
+++ b/arch/powerpc/mm/tlb_32.c
@@ -45,6 +45,7 @@ void flush_hash_entry(struct mm_struct *mm, pte_t *ptep, unsigned long addr)
 		flush_hash_pages(mm->context.id, addr, ptephys, 1);
 	}
 }
+EXPORT_SYMBOL(flush_hash_entry);
 
 /*
  * Called by ptep_set_access_flags, must flush on CPUs for which the
diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
index 7f651273386..439c5ba34ec 100644
--- a/arch/powerpc/platforms/Kconfig.cputype
+++ b/arch/powerpc/platforms/Kconfig.cputype
@@ -50,6 +50,7 @@ config 44x
 	select PPC_UDBG_16550
 	select 4xx_SOC
 	select PPC_PCI_CHOICE
+	select PHYS_64BIT
 
 config E200
 	bool "Freescale e200"
@@ -128,18 +129,20 @@ config FSL_EMB_PERFMON
 
 config PTE_64BIT
 	bool
-	depends on 44x || E500
-	default y if 44x
-	default y if E500 && PHYS_64BIT
+	depends on 44x || E500 || PPC_86xx
+	default y if PHYS_64BIT
 
 config PHYS_64BIT
-	bool 'Large physical address support' if E500
-	depends on 44x || E500
+	bool 'Large physical address support' if E500 || PPC_86xx
+	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
 	select RESOURCES_64BIT
-	default y if 44x
 	---help---
 	  This option enables kernel support for larger than 32-bit physical
-	  addresses.  This features is not be available on all e500 cores.
+	  addresses.  This feature may not be available on all cores.
+
+	  If you have more than 3.5GB of RAM or so, you also need to enable
+	  SWIOTLB under Kernel Options for this to work.  The actual number
+	  is platform-dependent.
 
 	  If in doubt, say N here.