[PATCH] ppc64: support 64k pages

Adds a new CONFIG_PPC_64K_PAGES which, when enabled, changes the kernel
base page size to 64K.  The resulting kernel still boots on any
hardware.  On current machines with 4K pages support only, the kernel
will maintain 16 "subpages" for each 64K page transparently.

Note that while real 64K capable HW has been tested, the current patch
will not enable it yet as such hardware is not released yet, and I'm
still verifying with the firmware architects the proper to get the
information from the newer hypervisors.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

arch/powerpc/Kconfig

@@ -603,6 +603,15 @@ config NODES_SPAN_OTHER_NODES
 	def_bool y
 	depends on NEED_MULTIPLE_NODES
 
+config PPC_64K_PAGES
+	bool "64k page size"
+	help
+	  This option changes the kernel logical page size to 64k. On machines
+          without processor support for 64k pages, the kernel will simulate
+          them by loading each individual 4k page on demand transparently,
+          while on hardware with such support, it will be used to map
+          normal application pages.
+
 config SCHED_SMT
 	bool "SMT (Hyperthreading) scheduler support"
 	depends on PPC64 && SMP

arch/powerpc/kernel/asm-offsets.c

@@ -125,6 +125,9 @@ int main(void)
 	DEFINE(PACASLBCACHE, offsetof(struct paca_struct, slb_cache));
 	DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr));
 	DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
+#ifdef CONFIG_PPC_64K_PAGES
+	DEFINE(PACAPGDIR, offsetof(struct paca_struct, pgdir));
+#endif
 #ifdef CONFIG_HUGETLB_PAGE
 	DEFINE(PACALOWHTLBAREAS, offsetof(struct paca_struct, context.low_htlb_areas));
 	DEFINE(PACAHIGHHTLBAREAS, offsetof(struct paca_struct, context.high_htlb_areas));

arch/powerpc/kernel/cputable.c

@@ -240,7 +240,7 @@ struct cpu_spec	cpu_specs[] = {
 		.oprofile_model		= &op_model_power4,
 #endif
 	},
-	{	/* Power5 */
+	{	/* Power5 GR */
 		.pvr_mask		= 0xffff0000,
 		.pvr_value		= 0x003a0000,
 		.cpu_name		= "POWER5 (gr)",
@@ -255,7 +255,7 @@ struct cpu_spec	cpu_specs[] = {
 		.oprofile_model		= &op_model_power4,
 #endif
 	},
-	{	/* Power5 */
+	{	/* Power5 GS */
 		.pvr_mask		= 0xffff0000,
 		.pvr_value		= 0x003b0000,
 		.cpu_name		= "POWER5 (gs)",

arch/powerpc/kernel/head_64.S

@@ -195,11 +195,11 @@ exception_marker:
 #define EX_R12		24
 #define EX_R13		32
 #define EX_SRR0		40
-#define EX_R3		40	/* SLB miss saves R3, but not SRR0 */
 #define EX_DAR		48
-#define EX_LR		48	/* SLB miss saves LR, but not DAR */
 #define EX_DSISR	56
 #define EX_CCR		60
+#define EX_R3		64
+#define EX_LR		72
 
 #define EXCEPTION_PROLOG_PSERIES(area, label)				\
 	mfspr	r13,SPRN_SPRG3;		/* get paca address into r13 */	\
@@ -419,17 +419,22 @@ data_access_slb_pSeries:
 	mtspr	SPRN_SPRG1,r13
 	RUNLATCH_ON(r13)
 	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
+	std	r3,PACA_EXSLB+EX_R3(r13)
+	mfspr	r3,SPRN_DAR
 	std	r9,PACA_EXSLB+EX_R9(r13)	/* save r9 - r12 */
+	mfcr	r9
+#ifdef __DISABLED__
+	/* Keep that around for when we re-implement dynamic VSIDs */
+	cmpdi	r3,0
+	bge	slb_miss_user_pseries
+#endif /* __DISABLED__ */
 	std	r10,PACA_EXSLB+EX_R10(r13)
 	std	r11,PACA_EXSLB+EX_R11(r13)
 	std	r12,PACA_EXSLB+EX_R12(r13)
-	std	r3,PACA_EXSLB+EX_R3(r13)
-	mfspr	r9,SPRN_SPRG1
-	std	r9,PACA_EXSLB+EX_R13(r13)
-	mfcr	r9
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
 	mfspr	r12,SPRN_SRR1		/* and SRR1 */
-	mfspr	r3,SPRN_DAR
-	b	.do_slb_miss		/* Rel. branch works in real mode */
+	b	.slb_miss_realmode	/* Rel. branch works in real mode */
 
 	STD_EXCEPTION_PSERIES(0x400, instruction_access)
 
@@ -440,17 +445,22 @@ instruction_access_slb_pSeries:
 	mtspr	SPRN_SPRG1,r13
 	RUNLATCH_ON(r13)
 	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
+	std	r3,PACA_EXSLB+EX_R3(r13)
+	mfspr	r3,SPRN_SRR0		/* SRR0 is faulting address */
 	std	r9,PACA_EXSLB+EX_R9(r13)	/* save r9 - r12 */
+	mfcr	r9
+#ifdef __DISABLED__
+	/* Keep that around for when we re-implement dynamic VSIDs */
+	cmpdi	r3,0
+	bge	slb_miss_user_pseries
+#endif /* __DISABLED__ */
 	std	r10,PACA_EXSLB+EX_R10(r13)
 	std	r11,PACA_EXSLB+EX_R11(r13)
 	std	r12,PACA_EXSLB+EX_R12(r13)
-	std	r3,PACA_EXSLB+EX_R3(r13)
-	mfspr	r9,SPRN_SPRG1
-	std	r9,PACA_EXSLB+EX_R13(r13)
-	mfcr	r9
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
 	mfspr	r12,SPRN_SRR1		/* and SRR1 */
-	mfspr	r3,SPRN_SRR0			/* SRR0 is faulting address */
-	b	.do_slb_miss		/* Rel. branch works in real mode */
+	b	.slb_miss_realmode	/* Rel. branch works in real mode */
 
 	STD_EXCEPTION_PSERIES(0x500, hardware_interrupt)
 	STD_EXCEPTION_PSERIES(0x600, alignment)
@@ -508,6 +518,38 @@ _GLOBAL(do_stab_bolted_pSeries)
 	mfspr	r12,SPRN_SPRG2
 	EXCEPTION_PROLOG_PSERIES(PACA_EXSLB, .do_stab_bolted)
 
+/*
+ * We have some room here  we use that to put
+ * the peries slb miss user trampoline code so it's reasonably
+ * away from slb_miss_user_common to avoid problems with rfid
+ *
+ * This is used for when the SLB miss handler has to go virtual,
+ * which doesn't happen for now anymore but will once we re-implement
+ * dynamic VSIDs for shared page tables
+ */
+#ifdef __DISABLED__
+slb_miss_user_pseries:
+	std	r10,PACA_EXGEN+EX_R10(r13)
+	std	r11,PACA_EXGEN+EX_R11(r13)
+	std	r12,PACA_EXGEN+EX_R12(r13)
+	mfspr	r10,SPRG1
+	ld	r11,PACA_EXSLB+EX_R9(r13)
+	ld	r12,PACA_EXSLB+EX_R3(r13)
+	std	r10,PACA_EXGEN+EX_R13(r13)
+	std	r11,PACA_EXGEN+EX_R9(r13)
+	std	r12,PACA_EXGEN+EX_R3(r13)
+	clrrdi	r12,r13,32
+	mfmsr	r10
+	mfspr	r11,SRR0			/* save SRR0 */
+	ori	r12,r12,slb_miss_user_common@l	/* virt addr of handler */
+	ori	r10,r10,MSR_IR|MSR_DR|MSR_RI
+	mtspr	SRR0,r12
+	mfspr	r12,SRR1			/* and SRR1 */
+	mtspr	SRR1,r10
+	rfid
+	b	.				/* prevent spec. execution */
+#endif /* __DISABLED__ */
+
 /*
  * Vectors for the FWNMI option.  Share common code.
  */
@@ -559,22 +601,59 @@ END_FTR_SECTION_IFCLR(CPU_FTR_SLB)
 	.globl	data_access_slb_iSeries
 data_access_slb_iSeries:
 	mtspr	SPRN_SPRG1,r13		/* save r13 */
-	EXCEPTION_PROLOG_ISERIES_1(PACA_EXSLB)
+	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
 	std	r3,PACA_EXSLB+EX_R3(r13)
-	ld	r12,PACALPPACA+LPPACASRR1(r13)
 	mfspr	r3,SPRN_DAR
-	b	.do_slb_miss
+	std	r9,PACA_EXSLB+EX_R9(r13)
+	mfcr	r9
+#ifdef __DISABLED__
+	cmpdi	r3,0
+	bge	slb_miss_user_iseries
+#endif
+	std	r10,PACA_EXSLB+EX_R10(r13)
+	std	r11,PACA_EXSLB+EX_R11(r13)
+	std	r12,PACA_EXSLB+EX_R12(r13)
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
+	ld	r12,PACALPPACA+LPPACASRR1(r13);
+	b	.slb_miss_realmode
 
 	STD_EXCEPTION_ISERIES(0x400, instruction_access, PACA_EXGEN)
 
 	.globl	instruction_access_slb_iSeries
 instruction_access_slb_iSeries:
 	mtspr	SPRN_SPRG1,r13		/* save r13 */
-	EXCEPTION_PROLOG_ISERIES_1(PACA_EXSLB)
+	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
 	std	r3,PACA_EXSLB+EX_R3(r13)
-	ld	r12,PACALPPACA+LPPACASRR1(r13)
-	ld	r3,PACALPPACA+LPPACASRR0(r13)
-	b	.do_slb_miss
+	ld	r3,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
+	std	r9,PACA_EXSLB+EX_R9(r13)
+	mfcr	r9
+#ifdef __DISABLED__
+	cmpdi	r3,0
+	bge	.slb_miss_user_iseries
+#endif
+	std	r10,PACA_EXSLB+EX_R10(r13)
+	std	r11,PACA_EXSLB+EX_R11(r13)
+	std	r12,PACA_EXSLB+EX_R12(r13)
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
+	ld	r12,PACALPPACA+LPPACASRR1(r13);
+	b	.slb_miss_realmode
+
+#ifdef __DISABLED__
+slb_miss_user_iseries:
+	std	r10,PACA_EXGEN+EX_R10(r13)
+	std	r11,PACA_EXGEN+EX_R11(r13)
+	std	r12,PACA_EXGEN+EX_R12(r13)
+	mfspr	r10,SPRG1
+	ld	r11,PACA_EXSLB+EX_R9(r13)
+	ld	r12,PACA_EXSLB+EX_R3(r13)
+	std	r10,PACA_EXGEN+EX_R13(r13)
+	std	r11,PACA_EXGEN+EX_R9(r13)
+	std	r12,PACA_EXGEN+EX_R3(r13)
+	EXCEPTION_PROLOG_ISERIES_2
+	b	slb_miss_user_common
+#endif
 
 	MASKABLE_EXCEPTION_ISERIES(0x500, hardware_interrupt)
 	STD_EXCEPTION_ISERIES(0x600, alignment, PACA_EXGEN)
@@ -809,6 +888,126 @@ instruction_access_common:
 	li	r5,0x400
 	b	.do_hash_page		/* Try to handle as hpte fault */
 
+/*
+ * Here is the common SLB miss user that is used when going to virtual
+ * mode for SLB misses, that is currently not used
+ */
+#ifdef __DISABLED__
+	.align	7
+	.globl	slb_miss_user_common
+slb_miss_user_common:
+	mflr	r10
+	std	r3,PACA_EXGEN+EX_DAR(r13)
+	stw	r9,PACA_EXGEN+EX_CCR(r13)
+	std	r10,PACA_EXGEN+EX_LR(r13)
+	std	r11,PACA_EXGEN+EX_SRR0(r13)
+	bl	.slb_allocate_user
+
+	ld	r10,PACA_EXGEN+EX_LR(r13)
+	ld	r3,PACA_EXGEN+EX_R3(r13)
+	lwz	r9,PACA_EXGEN+EX_CCR(r13)
+	ld	r11,PACA_EXGEN+EX_SRR0(r13)
+	mtlr	r10
+	beq-	slb_miss_fault
+
+	andi.	r10,r12,MSR_RI		/* check for unrecoverable exception */
+	beq-	unrecov_user_slb
+	mfmsr	r10
+
+.machine push
+.machine "power4"
+	mtcrf	0x80,r9
+.machine pop
+
+	clrrdi	r10,r10,2		/* clear RI before setting SRR0/1 */
+	mtmsrd	r10,1
+
+	mtspr	SRR0,r11
+	mtspr	SRR1,r12
+
+	ld	r9,PACA_EXGEN+EX_R9(r13)
+	ld	r10,PACA_EXGEN+EX_R10(r13)
+	ld	r11,PACA_EXGEN+EX_R11(r13)
+	ld	r12,PACA_EXGEN+EX_R12(r13)
+	ld	r13,PACA_EXGEN+EX_R13(r13)
+	rfid
+	b	.
+
+slb_miss_fault:
+	EXCEPTION_PROLOG_COMMON(0x380, PACA_EXGEN)
+	ld	r4,PACA_EXGEN+EX_DAR(r13)
+	li	r5,0
+	std	r4,_DAR(r1)
+	std	r5,_DSISR(r1)
+	b	.handle_page_fault
+
+unrecov_user_slb:
+	EXCEPTION_PROLOG_COMMON(0x4200, PACA_EXGEN)
+	DISABLE_INTS
+	bl	.save_nvgprs
+1:	addi	r3,r1,STACK_FRAME_OVERHEAD
+	bl	.unrecoverable_exception
+	b	1b
+
+#endif /* __DISABLED__ */
+
+
+/*
+ * r13 points to the PACA, r9 contains the saved CR,
+ * r12 contain the saved SRR1, SRR0 is still ready for return
+ * r3 has the faulting address
+ * r9 - r13 are saved in paca->exslb.
+ * r3 is saved in paca->slb_r3
+ * We assume we aren't going to take any exceptions during this procedure.
+ */
+_GLOBAL(slb_miss_realmode)
+	mflr	r10
+
+	stw	r9,PACA_EXSLB+EX_CCR(r13)	/* save CR in exc. frame */
+	std	r10,PACA_EXSLB+EX_LR(r13)	/* save LR */
+
+	bl	.slb_allocate_realmode
+
+	/* All done -- return from exception. */
+
+	ld	r10,PACA_EXSLB+EX_LR(r13)
+	ld	r3,PACA_EXSLB+EX_R3(r13)
+	lwz	r9,PACA_EXSLB+EX_CCR(r13)	/* get saved CR */
+#ifdef CONFIG_PPC_ISERIES
+	ld	r11,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
+#endif /* CONFIG_PPC_ISERIES */
+
+	mtlr	r10
+
+	andi.	r10,r12,MSR_RI	/* check for unrecoverable exception */
+	beq-	unrecov_slb
+
+.machine	push
+.machine	"power4"
+	mtcrf	0x80,r9
+	mtcrf	0x01,r9		/* slb_allocate uses cr0 and cr7 */
+.machine	pop
+
+#ifdef CONFIG_PPC_ISERIES
+	mtspr	SPRN_SRR0,r11
+	mtspr	SPRN_SRR1,r12
+#endif /* CONFIG_PPC_ISERIES */
+	ld	r9,PACA_EXSLB+EX_R9(r13)
+	ld	r10,PACA_EXSLB+EX_R10(r13)
+	ld	r11,PACA_EXSLB+EX_R11(r13)
+	ld	r12,PACA_EXSLB+EX_R12(r13)
+	ld	r13,PACA_EXSLB+EX_R13(r13)
+	rfid
+	b	.	/* prevent speculative execution */
+
+unrecov_slb:
+	EXCEPTION_PROLOG_COMMON(0x4100, PACA_EXSLB)
+	DISABLE_INTS
+	bl	.save_nvgprs
+1:	addi	r3,r1,STACK_FRAME_OVERHEAD
+	bl	.unrecoverable_exception
+	b	1b
+
 	.align	7
 	.globl hardware_interrupt_common
 	.globl hardware_interrupt_entry
@@ -1138,62 +1337,6 @@ _GLOBAL(do_stab_bolted)
 	rfid
 	b	.	/* prevent speculative execution */
 
-/*
- * r13 points to the PACA, r9 contains the saved CR,
- * r11 and r12 contain the saved SRR0 and SRR1.
- * r3 has the faulting address
- * r9 - r13 are saved in paca->exslb.
- * r3 is saved in paca->slb_r3
- * We assume we aren't going to take any exceptions during this procedure.
- */
-_GLOBAL(do_slb_miss)
-	mflr	r10
-
-	stw	r9,PACA_EXSLB+EX_CCR(r13)	/* save CR in exc. frame */
-	std	r10,PACA_EXSLB+EX_LR(r13)	/* save LR */
-
-	bl	.slb_allocate			/* handle it */
-
-	/* All done -- return from exception. */
-
-	ld	r10,PACA_EXSLB+EX_LR(r13)
-	ld	r3,PACA_EXSLB+EX_R3(r13)
-	lwz	r9,PACA_EXSLB+EX_CCR(r13)	/* get saved CR */
-#ifdef CONFIG_PPC_ISERIES
-	ld	r11,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
-#endif /* CONFIG_PPC_ISERIES */
-
-	mtlr	r10
-
-	andi.	r10,r12,MSR_RI	/* check for unrecoverable exception */
-	beq-	unrecov_slb
-
-.machine	push
-.machine	"power4"
-	mtcrf	0x80,r9
-	mtcrf	0x01,r9		/* slb_allocate uses cr0 and cr7 */
-.machine	pop
-
-#ifdef CONFIG_PPC_ISERIES
-	mtspr	SPRN_SRR0,r11
-	mtspr	SPRN_SRR1,r12
-#endif /* CONFIG_PPC_ISERIES */
-	ld	r9,PACA_EXSLB+EX_R9(r13)
-	ld	r10,PACA_EXSLB+EX_R10(r13)
-	ld	r11,PACA_EXSLB+EX_R11(r13)
-	ld	r12,PACA_EXSLB+EX_R12(r13)
-	ld	r13,PACA_EXSLB+EX_R13(r13)
-	rfid
-	b	.	/* prevent speculative execution */
-
-unrecov_slb:
-	EXCEPTION_PROLOG_COMMON(0x4100, PACA_EXSLB)
-	DISABLE_INTS
-	bl	.save_nvgprs
-1:	addi	r3,r1,STACK_FRAME_OVERHEAD
-	bl	.unrecoverable_exception
-	b	1b
-
 /*
  * Space for CPU0's segment table.
  *
@@ -1569,7 +1712,10 @@ _GLOBAL(__secondary_start)
 #endif
 	/* Initialize the first segment table (or SLB) entry		 */
 	ld	r3,PACASTABVIRT(r13)	/* get addr of segment table	 */
+BEGIN_FTR_SECTION
 	bl	.stab_initialize
+END_FTR_SECTION_IFCLR(CPU_FTR_SLB)
+	bl	.slb_initialize
 
 	/* Initialize the kernel stack.  Just a repeat for iSeries.	 */
 	LOADADDR(r3,current_set)

arch/powerpc/kernel/lparmap.c

@@ -25,7 +25,7 @@ const struct LparMap __attribute__((__section__(".text"))) xLparMap = {
 	.xRanges = {
 		{ .xPages = HvPagesToMap,
 		  .xOffset = 0,
-		  .xVPN = KERNEL_VSID(KERNELBASE) << (SID_SHIFT - PAGE_SHIFT),
+		  .xVPN = KERNEL_VSID(KERNELBASE) << (SID_SHIFT - HW_PAGE_SHIFT),
 		},
 	},
 };

arch/powerpc/kernel/process.c

@@ -554,12 +554,10 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 #ifdef CONFIG_PPC64
 	if (cpu_has_feature(CPU_FTR_SLB)) {
 		unsigned long sp_vsid = get_kernel_vsid(sp);
+		unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
 
 		sp_vsid <<= SLB_VSID_SHIFT;
-		sp_vsid |= SLB_VSID_KERNEL;
-		if (cpu_has_feature(CPU_FTR_16M_PAGE))
-			sp_vsid |= SLB_VSID_L;
-
+		sp_vsid |= SLB_VSID_KERNEL | llp;
 		p->thread.ksp_vsid = sp_vsid;
 	}
 

arch/powerpc/kernel/prom.c

@@ -724,7 +724,7 @@ static inline char *find_flat_dt_string(u32 offset)
  * used to extract the memory informations at boot before we can
  * unflatten the tree
  */
-static int __init scan_flat_dt(int (*it)(unsigned long node,
+int __init of_scan_flat_dt(int (*it)(unsigned long node,
 				     const char *uname, int depth,
 				     void *data),
 			   void *data)
@@ -784,7 +784,7 @@ static int __init scan_flat_dt(int (*it)(unsigned long node,
  * This  function can be used within scan_flattened_dt callback to get
  * access to properties
  */
-static void* __init get_flat_dt_prop(unsigned long node, const char *name,
+void* __init of_get_flat_dt_prop(unsigned long node, const char *name,
 				 unsigned long *size)
 {
 	unsigned long p = node;
@@ -1087,7 +1087,7 @@ void __init unflatten_device_tree(void)
 static int __init early_init_dt_scan_cpus(unsigned long node,
 					  const char *uname, int depth, void *data)
 {
-	char *type = get_flat_dt_prop(node, "device_type", NULL);
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 	u32 *prop;
 	unsigned long size = 0;
 
@@ -1095,19 +1095,6 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 	if (type == NULL || strcmp(type, "cpu") != 0)
 		return 0;
 
-#ifdef CONFIG_PPC_PSERIES
-	/* On LPAR, look for the first ibm,pft-size property for the  hash table size
-	 */
-	if (systemcfg->platform == PLATFORM_PSERIES_LPAR && ppc64_pft_size == 0) {
-		u32 *pft_size;
-		pft_size = get_flat_dt_prop(node, "ibm,pft-size", NULL);
-		if (pft_size != NULL) {
-			/* pft_size[0] is the NUMA CEC cookie */
-			ppc64_pft_size = pft_size[1];
-		}
-	}
-#endif
-
 	boot_cpuid = 0;
 	boot_cpuid_phys = 0;
 	if (initial_boot_params && initial_boot_params->version >= 2) {
@@ -1117,8 +1104,9 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 		boot_cpuid_phys = initial_boot_params->boot_cpuid_phys;
 	} else {
 		/* Check if it's the boot-cpu, set it's hw index now */
-		if (get_flat_dt_prop(node, "linux,boot-cpu", NULL) != NULL) {
-			prop = get_flat_dt_prop(node, "reg", NULL);
+		if (of_get_flat_dt_prop(node,
+					"linux,boot-cpu", NULL) != NULL) {
+			prop = of_get_flat_dt_prop(node, "reg", NULL);
 			if (prop != NULL)
 				boot_cpuid_phys = *prop;
 		}
@@ -1127,14 +1115,14 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 
 #ifdef CONFIG_ALTIVEC
 	/* Check if we have a VMX and eventually update CPU features */
-	prop = (u32 *)get_flat_dt_prop(node, "ibm,vmx", &size);
+	prop = (u32 *)of_get_flat_dt_prop(node, "ibm,vmx", &size);
 	if (prop && (*prop) > 0) {
 		cur_cpu_spec->cpu_features |= CPU_FTR_ALTIVEC;
 		cur_cpu_spec->cpu_user_features |= PPC_FEATURE_HAS_ALTIVEC;
 	}
 
 	/* Same goes for Apple's "altivec" property */
-	prop = (u32 *)get_flat_dt_prop(node, "altivec", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "altivec", NULL);
 	if (prop) {
 		cur_cpu_spec->cpu_features |= CPU_FTR_ALTIVEC;
 		cur_cpu_spec->cpu_user_features |= PPC_FEATURE_HAS_ALTIVEC;
@@ -1147,7 +1135,7 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 	 * this by looking at the size of the ibm,ppc-interrupt-server#s
 	 * property
 	 */
-	prop = (u32 *)get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s",
+	prop = (u32 *)of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s",
 				       &size);
 	cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT;
 	if (prop && ((size / sizeof(u32)) > 1))
@@ -1170,7 +1158,7 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
 		return 0;
 
 	/* get platform type */
-	prop = (u32 *)get_flat_dt_prop(node, "linux,platform", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "linux,platform", NULL);
 	if (prop == NULL)
 		return 0;
 #ifdef CONFIG_PPC64
@@ -1183,21 +1171,21 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
 
 #ifdef CONFIG_PPC64
 	/* check if iommu is forced on or off */
-	if (get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
+	if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
 		iommu_is_off = 1;
-	if (get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
+	if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
 		iommu_force_on = 1;
 #endif
 
- 	lprop = get_flat_dt_prop(node, "linux,memory-limit", NULL);
+ 	lprop = of_get_flat_dt_prop(node, "linux,memory-limit", NULL);
  	if (lprop)
  		memory_limit = *lprop;
 
 #ifdef CONFIG_PPC64
- 	lprop = get_flat_dt_prop(node, "linux,tce-alloc-start", NULL);
+ 	lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-start", NULL);
  	if (lprop)
  		tce_alloc_start = *lprop;
- 	lprop = get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
+ 	lprop = of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
  	if (lprop)
  		tce_alloc_end = *lprop;
 #endif
@@ -1209,9 +1197,9 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
 	{
 		u64 *basep, *entryp;
 
-		basep = get_flat_dt_prop(node, "linux,rtas-base", NULL);
-		entryp = get_flat_dt_prop(node, "linux,rtas-entry", NULL);
-		prop = get_flat_dt_prop(node, "linux,rtas-size", NULL);
+		basep = of_get_flat_dt_prop(node, "linux,rtas-base", NULL);
+		entryp = of_get_flat_dt_prop(node, "linux,rtas-entry", NULL);
+		prop = of_get_flat_dt_prop(node, "linux,rtas-size", NULL);
 		if (basep && entryp && prop) {
 			rtas.base = *basep;
 			rtas.entry = *entryp;
@@ -1232,11 +1220,11 @@ static int __init early_init_dt_scan_root(unsigned long node,
 	if (depth != 0)
 		return 0;
 
-	prop = get_flat_dt_prop(node, "#size-cells", NULL);
+	prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
 	dt_root_size_cells = (prop == NULL) ? 1 : *prop;
 	DBG("dt_root_size_cells = %x\n", dt_root_size_cells);
 
-	prop = get_flat_dt_prop(node, "#address-cells", NULL);
+	prop = of_get_flat_dt_prop(node, "#address-cells", NULL);
 	dt_root_addr_cells = (prop == NULL) ? 2 : *prop;
 	DBG("dt_root_addr_cells = %x\n", dt_root_addr_cells);
 	
@@ -1271,7 +1259,7 @@ static unsigned long __init dt_mem_next_cell(int s, cell_t **cellp)
 static int __init early_init_dt_scan_memory(unsigned long node,
 					    const char *uname, int depth, void *data)
 {
-	char *type = get_flat_dt_prop(node, "device_type", NULL);
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 	cell_t *reg, *endp;
 	unsigned long l;
 
@@ -1279,7 +1267,7 @@ static int __init early_init_dt_scan_memory(unsigned long node,
 	if (type == NULL || strcmp(type, "memory") != 0)
 		return 0;
 
-	reg = (cell_t *)get_flat_dt_prop(node, "reg", &l);
+	reg = (cell_t *)of_get_flat_dt_prop(node, "reg", &l);
 	if (reg == NULL)
 		return 0;
 
@@ -1343,12 +1331,12 @@ void __init early_init_devtree(void *params)
 	 * device-tree, including the platform type, initrd location and
 	 * size, TCE reserve, and more ...
 	 */
-	scan_flat_dt(early_init_dt_scan_chosen, NULL);
+	of_scan_flat_dt(early_init_dt_scan_chosen, NULL);
 
 	/* Scan memory nodes and rebuild LMBs */
 	lmb_init();
-	scan_flat_dt(early_init_dt_scan_root, NULL);
-	scan_flat_dt(early_init_dt_scan_memory, NULL);
+	of_scan_flat_dt(early_init_dt_scan_root, NULL);
+	of_scan_flat_dt(early_init_dt_scan_memory, NULL);
 	lmb_enforce_memory_limit(memory_limit);
 	lmb_analyze();
 #ifdef CONFIG_PPC64
@@ -1363,10 +1351,10 @@ void __init early_init_devtree(void *params)
 
 	DBG("Scanning CPUs ...\n");
 
-	/* Retreive hash table size from flattened tree plus other
-	 * CPU related informations (altivec support, boot CPU ID, ...)
+	/* Retreive CPU related informations from the flat tree
+	 * (altivec support, boot CPU ID, ...)
 	 */
-	scan_flat_dt(early_init_dt_scan_cpus, NULL);
+	of_scan_flat_dt(early_init_dt_scan_cpus, NULL);
 
 	DBG(" <- early_init_devtree()\n");
 }

arch/powerpc/kernel/setup_64.c

@@ -277,16 +277,21 @@ void __init early_setup(unsigned long dt_ptr)
 	DBG("Found, Initializing memory management...\n");
 
 	/*
-	 * Initialize stab / SLB management
+	 * Initialize the MMU Hash table and create the linear mapping
+	 * of memory. Has to be done before stab/slb initialization as
+	 * this is currently where the page size encoding is obtained
 	 */
-	if (!firmware_has_feature(FW_FEATURE_ISERIES))
-		stab_initialize(lpaca->stab_real);
+	htab_initialize();
 
 	/*
-	 * Initialize the MMU Hash table and create the linear mapping
-	 * of memory
+	 * Initialize stab / SLB management except on iSeries
 	 */
-	htab_initialize();
+	if (!firmware_has_feature(FW_FEATURE_ISERIES)) {
+		if (cpu_has_feature(CPU_FTR_SLB))
+			slb_initialize();
+		else
+			stab_initialize(lpaca->stab_real);
+	}
 
 	DBG(" <- early_setup()\n");
 }
@@ -552,9 +557,11 @@ static void __init irqstack_early_init(void)
 	 * SLB misses on them.
 	 */
 	for_each_cpu(i) {
-		softirq_ctx[i] = (struct thread_info *)__va(lmb_alloc_base(THREAD_SIZE,
+		softirq_ctx[i] = (struct thread_info *)
+			__va(lmb_alloc_base(THREAD_SIZE,
 					    THREAD_SIZE, 0x10000000));
-		hardirq_ctx[i] = (struct thread_info *)__va(lmb_alloc_base(THREAD_SIZE,
+		hardirq_ctx[i] = (struct thread_info *)
+			__va(lmb_alloc_base(THREAD_SIZE,
 					    THREAD_SIZE, 0x10000000));
 	}
 }
@@ -583,8 +590,8 @@ static void __init emergency_stack_init(void)
 	limit = min(0x10000000UL, lmb.rmo_size);
 
 	for_each_cpu(i)
-		paca[i].emergency_sp = __va(lmb_alloc_base(PAGE_SIZE, 128,
-						limit)) + PAGE_SIZE;
+		paca[i].emergency_sp =
+		__va(lmb_alloc_base(HW_PAGE_SIZE, 128, limit)) + HW_PAGE_SIZE;
 }
 
 /*

arch/powerpc/lib/copypage_64.S

@@ -11,7 +11,7 @@
 #include <asm/processor.h>
 #include <asm/ppc_asm.h>
 
-_GLOBAL(copy_page)
+_GLOBAL(copy_4K_page)
 	std	r31,-8(1)
 	std	r30,-16(1)
 	std	r29,-24(1)

arch/powerpc/lib/copyuser_64.S

@@ -24,7 +24,7 @@ _GLOBAL(__copy_tofrom_user)
 	std	r4,-16(r1)
 	std	r5,-8(r1)
 	dcbt	0,r4
-	beq	.Lcopy_page
+	beq	.Lcopy_page_4K
 	andi.	r6,r6,7
 	mtcrf	0x01,r5
 	blt	cr1,.Lshort_copy
@@ -366,7 +366,7 @@ _GLOBAL(__copy_tofrom_user)
  * above (following the .Ldst_aligned label) but it runs slightly
  * slower on POWER3.
  */
-.Lcopy_page:
+.Lcopy_page_4K:
 	std	r31,-32(1)
 	std	r30,-40(1)
 	std	r29,-48(1)

arch/powerpc/mm/hash_low_64.S

 /*
  * ppc64 MMU hashtable management routines
  *
- * (c) Copyright IBM Corp. 2003
+ * (c) Copyright IBM Corp. 2003, 2005
  *
  * Maintained by: Benjamin Herrenschmidt
  *                <benh@kernel.crashing.org>
@@ -10,6 +10,7 @@
  * described in the kernel's COPYING file.
  */
 
+#include <linux/config.h>
 #include <asm/reg.h>
 #include <asm/pgtable.h>
 #include <asm/mmu.h>
@@ -42,14 +43,24 @@
 /* Save non-volatile offsets */
 #define STK_REG(i)	(112 + ((i)-14)*8)
 
+
+#ifndef CONFIG_PPC_64K_PAGES
+
+/*****************************************************************************
+ *                                                                           *
+ *           4K SW & 4K HW pages implementation                              *
+ *                                                                           *
+ *****************************************************************************/
+
+
 /*
- * _hash_page(unsigned long ea, unsigned long access, unsigned long vsid,
+ * _hash_page_4K(unsigned long ea, unsigned long access, unsigned long vsid,
  *		 pte_t *ptep, unsigned long trap, int local)
  *
- * Adds a page to the hash table. This is the non-LPAR version for now
+ * Adds a 4K page to the hash table in a segment of 4K pages only
  */
 
-_GLOBAL(__hash_page)
+_GLOBAL(__hash_page_4K)
 	mflr	r0
 	std	r0,16(r1)
 	stdu	r1,-STACKFRAMESIZE(r1)
@@ -88,7 +99,8 @@ _GLOBAL(__hash_page)
 	/* If so, just bail out and refault if needed. Someone else
 	 * is changing this PTE anyway and might hash it.
 	 */
-	bne-	bail_ok
+	bne-	htab_bail_ok
+
 	/* Prepare new PTE value (turn access RW into DIRTY, then
 	 * add BUSY,HASHPTE and ACCESSED)
 	 */
@@ -118,10 +130,10 @@ _GLOBAL(__hash_page)
 
 	/* Convert linux PTE bits into HW equivalents */
 	andi.	r3,r30,0x1fe		/* Get basic set of flags */
-	xori	r3,r3,HW_NO_EXEC	/* _PAGE_EXEC -> NOEXEC */
+	xori	r3,r3,HPTE_R_N		/* _PAGE_EXEC -> NOEXEC */
 	rlwinm	r0,r30,32-9+1,30,30	/* _PAGE_RW -> _PAGE_USER (r0) */
 	rlwinm	r4,r30,32-7+1,30,30	/* _PAGE_DIRTY -> _PAGE_USER (r4) */
-	and	r0,r0,r4		/* _PAGE_RW & _PAGE_DIRTY -> r0 bit 30 */
+	and	r0,r0,r4		/* _PAGE_RW & _PAGE_DIRTY ->r0 bit 30*/
 	andc	r0,r30,r0		/* r0 = pte & ~r0 */
 	rlwimi	r3,r0,32-1,31,31	/* Insert result into PP lsb */
 
@@ -158,19 +170,21 @@ htab_insert_pte:
 	andc	r30,r30,r0
 	ori	r30,r30,_PAGE_HASHPTE
 
-	/* page number in r5 */
-	rldicl	r5,r31,64-PTE_SHIFT,PTE_SHIFT
+	/* physical address r5 */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT
 
 	/* Calculate primary group hash */
 	and	r0,r28,r27
-	rldicr	r3,r0,3,63-3	/* r0 = (hash & mask) << 3 */
+	rldicr	r3,r0,3,63-3		/* r3 = (hash & mask) << 3 */
 
 	/* Call ppc_md.hpte_insert */
-	ld	r7,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
 	mr	r4,r29			/* Retreive va */
-	li	r6,0			/* no vflags */
+	li	r7,0			/* !bolted, !secondary */
+	li	r8,MMU_PAGE_4K		/* page size */
 _GLOBAL(htab_call_hpte_insert1)
-	bl	.			/* Will be patched by htab_finish_init() */
+	bl	.			/* Patched by htab_finish_init() */
 	cmpdi	0,r3,0
 	bge	htab_pte_insert_ok	/* Insertion successful */
 	cmpdi	0,r3,-2			/* Critical failure */
@@ -178,19 +192,21 @@ _GLOBAL(htab_call_hpte_insert1)
 
 	/* Now try secondary slot */
 	
-	/* page number in r5 */
-	rldicl	r5,r31,64-PTE_SHIFT,PTE_SHIFT
+	/* physical address r5 */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT
 
 	/* Calculate secondary group hash */
 	andc	r0,r27,r28
 	rldicr	r3,r0,3,63-3	/* r0 = (~hash & mask) << 3 */
 	
 	/* Call ppc_md.hpte_insert */
-	ld	r7,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
 	mr	r4,r29			/* Retreive va */
-	li	r6,HPTE_V_SECONDARY@l	/* secondary slot */
+	li	r7,HPTE_V_SECONDARY	/* !bolted, secondary */
+	li	r8,MMU_PAGE_4K		/* page size */
 _GLOBAL(htab_call_hpte_insert2)
-	bl	.			/* Will be patched by htab_finish_init() */
+	bl	.			/* Patched by htab_finish_init() */
 	cmpdi	0,r3,0
 	bge+	htab_pte_insert_ok	/* Insertion successful */
 	cmpdi	0,r3,-2			/* Critical failure */
@@ -207,14 +223,14 @@ _GLOBAL(htab_call_hpte_insert2)
 	rldicr	r3,r0,3,63-3	/* r0 = (hash & mask) << 3 */	
 	/* Call ppc_md.hpte_remove */
 _GLOBAL(htab_call_hpte_remove)
-	bl	.			/* Will be patched by htab_finish_init() */
+	bl	.			/* Patched by htab_finish_init() */
 
 	/* Try all again */
 	b	htab_insert_pte	
 
-bail_ok:
+htab_bail_ok:
 	li	r3,0
-	b	bail
+	b	htab_bail
 
 htab_pte_insert_ok:
 	/* Insert slot number & secondary bit in PTE */
@@ -227,7 +243,7 @@ htab_write_out_pte:
 	ld	r6,STK_PARM(r6)(r1)
 	std	r30,0(r6)
 	li	r3, 0
-bail:
+htab_bail:
 	ld	r27,STK_REG(r27)(r1)
 	ld	r28,STK_REG(r28)(r1)
 	ld	r29,STK_REG(r29)(r1)
@@ -256,10 +272,10 @@ htab_modify_pte:
 
 	/* Call ppc_md.hpte_updatepp */
 	mr	r5,r29			/* va */
-	li	r6,0			/* large is 0 */
+	li	r6,MMU_PAGE_4K		/* page size */
 	ld	r7,STK_PARM(r8)(r1)	/* get "local" param */
 _GLOBAL(htab_call_hpte_updatepp)
-	bl	.			/* Will be patched by htab_finish_init() */
+	bl	.			/* Patched by htab_finish_init() */
 
 	/* if we failed because typically the HPTE wasn't really here
 	 * we try an insertion. 
@@ -276,13 +292,556 @@ htab_wrong_access:
 	/* Bail out clearing reservation */
 	stdcx.	r31,0,r6
 	li	r3,1
-	b	bail
+	b	htab_bail
 
 htab_pte_insert_failure:
 	/* Bail out restoring old PTE */
 	ld	r6,STK_PARM(r6)(r1)
 	std	r31,0(r6)
 	li	r3,-1
-	b	bail
+	b	htab_bail
+
+
+#else /* CONFIG_PPC_64K_PAGES */
+
+
+/*****************************************************************************
+ *                                                                           *
+ *           64K SW & 4K or 64K HW in a 4K segment pages implementation      *
+ *                                                                           *
+ *****************************************************************************/
+
+/* _hash_page_4K(unsigned long ea, unsigned long access, unsigned long vsid,
+ *		 pte_t *ptep, unsigned long trap, int local)
+ */
+
+/*
+ * For now, we do NOT implement Admixed pages
+ */
+_GLOBAL(__hash_page_4K)
+	mflr	r0
+	std	r0,16(r1)
+	stdu	r1,-STACKFRAMESIZE(r1)
+	/* Save all params that we need after a function call */
+	std	r6,STK_PARM(r6)(r1)
+	std	r8,STK_PARM(r8)(r1)
+
+	/* Add _PAGE_PRESENT to access */
+	ori	r4,r4,_PAGE_PRESENT
+
+	/* Save non-volatile registers.
+	 * r31 will hold "old PTE"
+	 * r30 is "new PTE"
+	 * r29 is "va"
+	 * r28 is a hash value
+	 * r27 is hashtab mask (maybe dynamic patched instead ?)
+	 * r26 is the hidx mask
+	 * r25 is the index in combo page
+	 */
+	std	r25,STK_REG(r25)(r1)
+	std	r26,STK_REG(r26)(r1)
+	std	r27,STK_REG(r27)(r1)
+	std	r28,STK_REG(r28)(r1)
+	std	r29,STK_REG(r29)(r1)
+	std	r30,STK_REG(r30)(r1)
+	std	r31,STK_REG(r31)(r1)
+
+	/* Step 1:
+	 *
+	 * Check permissions, atomically mark the linux PTE busy
+	 * and hashed.
+	 */
+1:
+	ldarx	r31,0,r6
+	/* Check access rights (access & ~(pte_val(*ptep))) */
+	andc.	r0,r4,r31
+	bne-	htab_wrong_access
+	/* Check if PTE is busy */
+	andi.	r0,r31,_PAGE_BUSY
+	/* If so, just bail out and refault if needed. Someone else
+	 * is changing this PTE anyway and might hash it.
+	 */
+	bne-	htab_bail_ok
+	/* Prepare new PTE value (turn access RW into DIRTY, then
+	 * add BUSY and ACCESSED)
+	 */
+	rlwinm	r30,r4,32-9+7,31-7,31-7	/* _PAGE_RW -> _PAGE_DIRTY */
+	or	r30,r30,r31
+	ori	r30,r30,_PAGE_BUSY | _PAGE_ACCESSED | _PAGE_HASHPTE
+	/* Write the linux PTE atomically (setting busy) */
+	stdcx.	r30,0,r6
+	bne-	1b
+	isync
+
+	/* Step 2:
+	 *
+	 * Insert/Update the HPTE in the hash table. At this point,
+	 * r4 (access) is re-useable, we use it for the new HPTE flags
+	 */
+
+	/* Load the hidx index */
+	rldicl	r25,r3,64-12,60
+
+	/* Calc va and put it in r29 */
+	rldicr	r29,r5,28,63-28		/* r29 = (vsid << 28) */
+	rldicl	r3,r3,0,36		/* r3 = (ea & 0x0fffffff) */
+	or	r29,r3,r29		/* r29 = va
+
+	/* Calculate hash value for primary slot and store it in r28 */
+	rldicl	r5,r5,0,25		/* vsid & 0x0000007fffffffff */
+	rldicl	r0,r3,64-12,48		/* (ea >> 12) & 0xffff */
+	xor	r28,r5,r0
+
+	/* Convert linux PTE bits into HW equivalents */
+	andi.	r3,r30,0x1fe		/* Get basic set of flags */
+	xori	r3,r3,HPTE_R_N		/* _PAGE_EXEC -> NOEXEC */
+	rlwinm	r0,r30,32-9+1,30,30	/* _PAGE_RW -> _PAGE_USER (r0) */
+	rlwinm	r4,r30,32-7+1,30,30	/* _PAGE_DIRTY -> _PAGE_USER (r4) */
+	and	r0,r0,r4		/* _PAGE_RW & _PAGE_DIRTY ->r0 bit 30*/
+	andc	r0,r30,r0		/* r0 = pte & ~r0 */
+	rlwimi	r3,r0,32-1,31,31	/* Insert result into PP lsb */
+
+	/* We eventually do the icache sync here (maybe inline that
+	 * code rather than call a C function...)
+	 */
+BEGIN_FTR_SECTION
+	mr	r4,r30
+	mr	r5,r7
+	bl	.hash_page_do_lazy_icache
+END_FTR_SECTION(CPU_FTR_NOEXECUTE|CPU_FTR_COHERENT_ICACHE, CPU_FTR_NOEXECUTE)
+
+	/* At this point, r3 contains new PP bits, save them in
+	 * place of "access" in the param area (sic)
+	 */
+	std	r3,STK_PARM(r4)(r1)
+
+	/* Get htab_hash_mask */
+	ld	r4,htab_hash_mask@got(2)
+	ld	r27,0(r4)	/* htab_hash_mask -> r27 */
+
+	/* Check if we may already be in the hashtable, in this case, we
+	 * go to out-of-line code to try to modify the HPTE. We look for
+	 * the bit at (1 >> (index + 32))
+	 */
+	andi.	r0,r31,_PAGE_HASHPTE
+	li	r26,0			/* Default hidx */
+	beq	htab_insert_pte
+	ld	r6,STK_PARM(r6)(r1)
+	ori	r26,r6,0x8000		/* Load the hidx mask */
+	ld	r26,0(r26)
+	addi	r5,r25,36		/* Check actual HPTE_SUB bit, this */
+	rldcr.	r0,r31,r5,0		/* must match pgtable.h definition */
+	bne	htab_modify_pte
+
+htab_insert_pte:
+	/* real page number in r5, PTE RPN value + index */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT-HW_PAGE_SHIFT
+	add	r5,r5,r25
+	sldi	r5,r5,HW_PAGE_SHIFT
+
+	/* Calculate primary group hash */
+	and	r0,r28,r27
+	rldicr	r3,r0,3,63-3		/* r0 = (hash & mask) << 3 */
+
+	/* Call ppc_md.hpte_insert */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	mr	r4,r29			/* Retreive va */
+	li	r7,0			/* !bolted, !secondary */
+	li	r8,MMU_PAGE_4K		/* page size */
+_GLOBAL(htab_call_hpte_insert1)
+	bl	.			/* patched by htab_finish_init() */
+	cmpdi	0,r3,0
+	bge	htab_pte_insert_ok	/* Insertion successful */
+	cmpdi	0,r3,-2			/* Critical failure */
+	beq-	htab_pte_insert_failure
+
+	/* Now try secondary slot */
+
+	/* real page number in r5, PTE RPN value + index */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT-HW_PAGE_SHIFT
+	add	r5,r5,r25
+	sldi	r5,r5,HW_PAGE_SHIFT
+
+	/* Calculate secondary group hash */
+	andc	r0,r27,r28
+	rldicr	r3,r0,3,63-3		/* r0 = (~hash & mask) << 3 */
+
+	/* Call ppc_md.hpte_insert */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	mr	r4,r29			/* Retreive va */
+	li	r7,HPTE_V_SECONDARY	/* !bolted, secondary */
+	li	r8,MMU_PAGE_4K		/* page size */
+_GLOBAL(htab_call_hpte_insert2)
+	bl	.			/* patched by htab_finish_init() */
+	cmpdi	0,r3,0
+	bge+	htab_pte_insert_ok	/* Insertion successful */
+	cmpdi	0,r3,-2			/* Critical failure */
+	beq-	htab_pte_insert_failure
+
+	/* Both are full, we need to evict something */
+	mftb	r0
+	/* Pick a random group based on TB */
+	andi.	r0,r0,1
+	mr	r5,r28
+	bne	2f
+	not	r5,r5
+2:	and	r0,r5,r27
+	rldicr	r3,r0,3,63-3		/* r0 = (hash & mask) << 3 */
+	/* Call ppc_md.hpte_remove */
+_GLOBAL(htab_call_hpte_remove)
+	bl	.			/* patched by htab_finish_init() */
+
+	/* Try all again */
+	b	htab_insert_pte
+
+htab_bail_ok:
+	li	r3,0
+	b	htab_bail
+
+htab_pte_insert_ok:
+	/* Insert slot number & secondary bit in PTE second half,
+	 * clear _PAGE_BUSY and set approriate HPTE slot bit
+	 */
+	ld	r6,STK_PARM(r6)(r1)
+	li	r0,_PAGE_BUSY
+	andc	r30,r30,r0
+	/* HPTE SUB bit */
+	li	r0,1
+	subfic	r5,r25,27		/* Must match bit position in */
+	sld	r0,r0,r5		/* pgtable.h */
+	or	r30,r30,r0
+	/* hindx */
+	sldi	r5,r25,2
+	sld	r3,r3,r5
+	li	r4,0xf
+	sld	r4,r4,r5
+	andc	r26,r26,r4
+	or	r26,r26,r3
+	ori	r5,r6,0x8000
+	std	r26,0(r5)
+	lwsync
+	std	r30,0(r6)
+	li	r3, 0
+htab_bail:
+	ld	r25,STK_REG(r25)(r1)
+	ld	r26,STK_REG(r26)(r1)
+	ld	r27,STK_REG(r27)(r1)
+	ld	r28,STK_REG(r28)(r1)
+	ld	r29,STK_REG(r29)(r1)
+	ld      r30,STK_REG(r30)(r1)
+	ld      r31,STK_REG(r31)(r1)
+	addi    r1,r1,STACKFRAMESIZE
+	ld      r0,16(r1)
+	mtlr    r0
+	blr
+
+htab_modify_pte:
+	/* Keep PP bits in r4 and slot idx from the PTE around in r3 */
+	mr	r4,r3
+	sldi	r5,r25,2
+	srd	r3,r26,r5
+
+	/* Secondary group ? if yes, get a inverted hash value */
+	mr	r5,r28
+	andi.	r0,r3,0x8 /* page secondary ? */
+	beq	1f
+	not	r5,r5
+1:	andi.	r3,r3,0x7 /* extract idx alone */
+
+	/* Calculate proper slot value for ppc_md.hpte_updatepp */
+	and	r0,r5,r27
+	rldicr	r0,r0,3,63-3	/* r0 = (hash & mask) << 3 */
+	add	r3,r0,r3	/* add slot idx */
+
+	/* Call ppc_md.hpte_updatepp */
+	mr	r5,r29			/* va */
+	li	r6,MMU_PAGE_4K		/* page size */
+	ld	r7,STK_PARM(r8)(r1)	/* get "local" param */
+_GLOBAL(htab_call_hpte_updatepp)
+	bl	.			/* patched by htab_finish_init() */
+
+	/* if we failed because typically the HPTE wasn't really here
+	 * we try an insertion.
+	 */
+	cmpdi	0,r3,-1
+	beq-	htab_insert_pte
+
+	/* Clear the BUSY bit and Write out the PTE */
+	li	r0,_PAGE_BUSY
+	andc	r30,r30,r0
+	ld	r6,STK_PARM(r6)(r1)
+	std	r30,0(r6)
+	li	r3,0
+	b	htab_bail
+
+htab_wrong_access:
+	/* Bail out clearing reservation */
+	stdcx.	r31,0,r6
+	li	r3,1
+	b	htab_bail
+
+htab_pte_insert_failure:
+	/* Bail out restoring old PTE */
+	ld	r6,STK_PARM(r6)(r1)
+	std	r31,0(r6)
+	li	r3,-1
+	b	htab_bail
+
+
+/*****************************************************************************
+ *                                                                           *
+ *           64K SW & 64K HW in a 64K segment pages implementation           *
+ *                                                                           *
+ *****************************************************************************/
+
+_GLOBAL(__hash_page_64K)
+	mflr	r0
+	std	r0,16(r1)
+	stdu	r1,-STACKFRAMESIZE(r1)
+	/* Save all params that we need after a function call */
+	std	r6,STK_PARM(r6)(r1)
+	std	r8,STK_PARM(r8)(r1)
+
+	/* Add _PAGE_PRESENT to access */
+	ori	r4,r4,_PAGE_PRESENT
+
+	/* Save non-volatile registers.
+	 * r31 will hold "old PTE"
+	 * r30 is "new PTE"
+	 * r29 is "va"
+	 * r28 is a hash value
+	 * r27 is hashtab mask (maybe dynamic patched instead ?)
+	 */
+	std	r27,STK_REG(r27)(r1)
+	std	r28,STK_REG(r28)(r1)
+	std	r29,STK_REG(r29)(r1)
+	std	r30,STK_REG(r30)(r1)
+	std	r31,STK_REG(r31)(r1)
+
+	/* Step 1:
+	 *
+	 * Check permissions, atomically mark the linux PTE busy
+	 * and hashed.
+	 */
+1:
+	ldarx	r31,0,r6
+	/* Check access rights (access & ~(pte_val(*ptep))) */
+	andc.	r0,r4,r31
+	bne-	ht64_wrong_access
+	/* Check if PTE is busy */
+	andi.	r0,r31,_PAGE_BUSY
+	/* If so, just bail out and refault if needed. Someone else
+	 * is changing this PTE anyway and might hash it.
+	 */
+	bne-	ht64_bail_ok
+	/* Prepare new PTE value (turn access RW into DIRTY, then
+	 * add BUSY,HASHPTE and ACCESSED)
+	 */
+	rlwinm	r30,r4,32-9+7,31-7,31-7	/* _PAGE_RW -> _PAGE_DIRTY */
+	or	r30,r30,r31
+	ori	r30,r30,_PAGE_BUSY | _PAGE_ACCESSED | _PAGE_HASHPTE
+	/* Write the linux PTE atomically (setting busy) */
+	stdcx.	r30,0,r6
+	bne-	1b
+	isync
+
+	/* Step 2:
+	 *
+	 * Insert/Update the HPTE in the hash table. At this point,
+	 * r4 (access) is re-useable, we use it for the new HPTE flags
+	 */
+
+	/* Calc va and put it in r29 */
+	rldicr	r29,r5,28,63-28
+	rldicl	r3,r3,0,36
+	or	r29,r3,r29
+
+	/* Calculate hash value for primary slot and store it in r28 */
+	rldicl	r5,r5,0,25		/* vsid & 0x0000007fffffffff */
+	rldicl	r0,r3,64-16,52		/* (ea >> 16) & 0xfff */
+	xor	r28,r5,r0
+
+	/* Convert linux PTE bits into HW equivalents */
+	andi.	r3,r30,0x1fe		/* Get basic set of flags */
+	xori	r3,r3,HPTE_R_N		/* _PAGE_EXEC -> NOEXEC */
+	rlwinm	r0,r30,32-9+1,30,30	/* _PAGE_RW -> _PAGE_USER (r0) */
+	rlwinm	r4,r30,32-7+1,30,30	/* _PAGE_DIRTY -> _PAGE_USER (r4) */
+	and	r0,r0,r4		/* _PAGE_RW & _PAGE_DIRTY ->r0 bit 30*/
+	andc	r0,r30,r0		/* r0 = pte & ~r0 */
+	rlwimi	r3,r0,32-1,31,31	/* Insert result into PP lsb */
+
+	/* We eventually do the icache sync here (maybe inline that
+	 * code rather than call a C function...)
+	 */
+BEGIN_FTR_SECTION
+	mr	r4,r30
+	mr	r5,r7
+	bl	.hash_page_do_lazy_icache
+END_FTR_SECTION(CPU_FTR_NOEXECUTE|CPU_FTR_COHERENT_ICACHE, CPU_FTR_NOEXECUTE)
+
+	/* At this point, r3 contains new PP bits, save them in
+	 * place of "access" in the param area (sic)
+	 */
+	std	r3,STK_PARM(r4)(r1)
+
+	/* Get htab_hash_mask */
+	ld	r4,htab_hash_mask@got(2)
+	ld	r27,0(r4)	/* htab_hash_mask -> r27 */
+
+	/* Check if we may already be in the hashtable, in this case, we
+	 * go to out-of-line code to try to modify the HPTE
+	 */
+	andi.	r0,r31,_PAGE_HASHPTE
+	bne	ht64_modify_pte
+
+ht64_insert_pte:
+	/* Clear hpte bits in new pte (we also clear BUSY btw) and
+	 * add _PAGE_HASHPTE
+	 */
+	lis	r0,_PAGE_HPTEFLAGS@h
+	ori	r0,r0,_PAGE_HPTEFLAGS@l
+	andc	r30,r30,r0
+	ori	r30,r30,_PAGE_HASHPTE
+
+	/* Phyical address in r5 */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT
+
+	/* Calculate primary group hash */
+	and	r0,r28,r27
+	rldicr	r3,r0,3,63-3	/* r0 = (hash & mask) << 3 */
+
+	/* Call ppc_md.hpte_insert */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	mr	r4,r29			/* Retreive va */
+	li	r7,0			/* !bolted, !secondary */
+	li	r8,MMU_PAGE_64K
+_GLOBAL(ht64_call_hpte_insert1)
+	bl	.			/* patched by htab_finish_init() */
+	cmpdi	0,r3,0
+	bge	ht64_pte_insert_ok	/* Insertion successful */
+	cmpdi	0,r3,-2			/* Critical failure */
+	beq-	ht64_pte_insert_failure
+
+	/* Now try secondary slot */
+
+	/* Phyical address in r5 */
+	rldicl	r5,r31,64-PTE_RPN_SHIFT,PTE_RPN_SHIFT
+	sldi	r5,r5,PAGE_SHIFT
+
+	/* Calculate secondary group hash */
+	andc	r0,r27,r28
+	rldicr	r3,r0,3,63-3	/* r0 = (~hash & mask) << 3 */
+
+	/* Call ppc_md.hpte_insert */
+	ld	r6,STK_PARM(r4)(r1)	/* Retreive new pp bits */
+	mr	r4,r29			/* Retreive va */
+	li	r7,HPTE_V_SECONDARY	/* !bolted, secondary */
+	li	r8,MMU_PAGE_64K
+_GLOBAL(ht64_call_hpte_insert2)
+	bl	.			/* patched by htab_finish_init() */
+	cmpdi	0,r3,0
+	bge+	ht64_pte_insert_ok	/* Insertion successful */
+	cmpdi	0,r3,-2			/* Critical failure */
+	beq-	ht64_pte_insert_failure
+
+	/* Both are full, we need to evict something */
+	mftb	r0
+	/* Pick a random group based on TB */
+	andi.	r0,r0,1
+	mr	r5,r28
+	bne	2f
+	not	r5,r5
+2:	and	r0,r5,r27
+	rldicr	r3,r0,3,63-3	/* r0 = (hash & mask) << 3 */
+	/* Call ppc_md.hpte_remove */
+_GLOBAL(ht64_call_hpte_remove)
+	bl	.			/* patched by htab_finish_init() */
+
+	/* Try all again */
+	b	ht64_insert_pte
+
+ht64_bail_ok:
+	li	r3,0
+	b	ht64_bail
+
+ht64_pte_insert_ok:
+	/* Insert slot number & secondary bit in PTE */
+	rldimi	r30,r3,12,63-15
+
+	/* Write out the PTE with a normal write
+	 * (maybe add eieio may be good still ?)
+	 */
+ht64_write_out_pte:
+	ld	r6,STK_PARM(r6)(r1)
+	std	r30,0(r6)
+	li	r3, 0
+ht64_bail:
+	ld	r27,STK_REG(r27)(r1)
+	ld	r28,STK_REG(r28)(r1)
+	ld	r29,STK_REG(r29)(r1)
+	ld      r30,STK_REG(r30)(r1)
+	ld      r31,STK_REG(r31)(r1)
+	addi    r1,r1,STACKFRAMESIZE
+	ld      r0,16(r1)
+	mtlr    r0
+	blr
+
+ht64_modify_pte:
+	/* Keep PP bits in r4 and slot idx from the PTE around in r3 */
+	mr	r4,r3
+	rlwinm	r3,r31,32-12,29,31
+
+	/* Secondary group ? if yes, get a inverted hash value */
+	mr	r5,r28
+	andi.	r0,r31,_PAGE_F_SECOND
+	beq	1f
+	not	r5,r5
+1:
+	/* Calculate proper slot value for ppc_md.hpte_updatepp */
+	and	r0,r5,r27
+	rldicr	r0,r0,3,63-3	/* r0 = (hash & mask) << 3 */
+	add	r3,r0,r3	/* add slot idx */
+
+	/* Call ppc_md.hpte_updatepp */
+	mr	r5,r29			/* va */
+	li	r6,MMU_PAGE_64K
+	ld	r7,STK_PARM(r8)(r1)	/* get "local" param */
+_GLOBAL(ht64_call_hpte_updatepp)
+	bl	.			/* patched by htab_finish_init() */
+
+	/* if we failed because typically the HPTE wasn't really here
+	 * we try an insertion.
+	 */
+	cmpdi	0,r3,-1
+	beq-	ht64_insert_pte
+
+	/* Clear the BUSY bit and Write out the PTE */
+	li	r0,_PAGE_BUSY
+	andc	r30,r30,r0
+	b	ht64_write_out_pte
+
+ht64_wrong_access:
+	/* Bail out clearing reservation */
+	stdcx.	r31,0,r6
+	li	r3,1
+	b	ht64_bail
+
+ht64_pte_insert_failure:
+	/* Bail out restoring old PTE */
+	ld	r6,STK_PARM(r6)(r1)
+	std	r31,0(r6)
+	li	r3,-1
+	b	ht64_bail
+
+
+#endif /* CONFIG_PPC_64K_PAGES */
 
 
+/*****************************************************************************
+ *                                                                           *
+ *           Huge pages implementation is in hugetlbpage.c                   *
+ *                                                                           *
+ *****************************************************************************/

arch/powerpc/mm/hash_native_64.c

@@ -9,6 +9,9 @@
  * as published by the Free Software Foundation; either version
  * 2 of the License, or (at your option) any later version.
  */
+
+#undef DEBUG_LOW
+
 #include <linux/spinlock.h>
 #include <linux/bitops.h>
 #include <linux/threads.h>
@@ -22,11 +25,84 @@
 #include <asm/tlbflush.h>
 #include <asm/tlb.h>
 #include <asm/cputable.h>
+#include <asm/udbg.h>
+
+#ifdef DEBUG_LOW
+#define DBG_LOW(fmt...) udbg_printf(fmt)
+#else
+#define DBG_LOW(fmt...)
+#endif
 
 #define HPTE_LOCK_BIT 3
 
 static DEFINE_SPINLOCK(native_tlbie_lock);
 
+static inline void __tlbie(unsigned long va, unsigned int psize)
+{
+	unsigned int penc;
+
+	/* clear top 16 bits, non SLS segment */
+	va &= ~(0xffffULL << 48);
+
+	switch (psize) {
+	case MMU_PAGE_4K:
+		va &= ~0xffful;
+		asm volatile("tlbie %0,0" : : "r" (va) : "memory");
+		break;
+	default:
+		penc = mmu_psize_defs[psize].penc;
+		va &= ~((1ul << mmu_psize_defs[psize].shift) - 1);
+		va |= (0x7f >> (8 - penc)) << 12;
+		asm volatile("tlbie %0,1" : : "r" (va) : "memory");
+		break;
+	}
+}
+
+static inline void __tlbiel(unsigned long va, unsigned int psize)
+{
+	unsigned int penc;
+
+	/* clear top 16 bits, non SLS segment */
+	va &= ~(0xffffULL << 48);
+
+	switch (psize) {
+	case MMU_PAGE_4K:
+		va &= ~0xffful;
+		asm volatile(".long 0x7c000224 | (%0 << 11) | (0 << 21)"
+			     : : "r"(va) : "memory");
+		break;
+	default:
+		penc = mmu_psize_defs[psize].penc;
+		va &= ~((1ul << mmu_psize_defs[psize].shift) - 1);
+		va |= (0x7f >> (8 - penc)) << 12;
+		asm volatile(".long 0x7c000224 | (%0 << 11) | (1 << 21)"
+			     : : "r"(va) : "memory");
+		break;
+	}
+
+}
+
+static inline void tlbie(unsigned long va, int psize, int local)
+{
+	unsigned int use_local = local && cpu_has_feature(CPU_FTR_TLBIEL);
+	int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
+
+	if (use_local)
+		use_local = mmu_psize_defs[psize].tlbiel;
+	if (lock_tlbie && !use_local)
+		spin_lock(&native_tlbie_lock);
+	asm volatile("ptesync": : :"memory");
+	if (use_local) {
+		__tlbiel(va, psize);
+		asm volatile("ptesync": : :"memory");
+	} else {
+		__tlbie(va, psize);
+		asm volatile("eieio; tlbsync; ptesync": : :"memory");
+	}
+	if (lock_tlbie && !use_local)
+		spin_unlock(&native_tlbie_lock);
+}
+
 static inline void native_lock_hpte(hpte_t *hptep)
 {
 	unsigned long *word = &hptep->v;
@@ -48,13 +124,19 @@ static inline void native_unlock_hpte(hpte_t *hptep)
 }
 
 long native_hpte_insert(unsigned long hpte_group, unsigned long va,
-			unsigned long prpn, unsigned long vflags,
-			unsigned long rflags)
+			unsigned long pa, unsigned long rflags,
+			unsigned long vflags, int psize)
 {
 	hpte_t *hptep = htab_address + hpte_group;
 	unsigned long hpte_v, hpte_r;
 	int i;
 
+	if (!(vflags & HPTE_V_BOLTED)) {
+		DBG_LOW("    insert(group=%lx, va=%016lx, pa=%016lx,"
+			" rflags=%lx, vflags=%lx, psize=%d)\n",
+			hpte_group, va, pa, rflags, vflags, psize);
+	}
+
 	for (i = 0; i < HPTES_PER_GROUP; i++) {
 		if (! (hptep->v & HPTE_V_VALID)) {
 			/* retry with lock held */
@@ -70,10 +152,13 @@ long native_hpte_insert(unsigned long hpte_group, unsigned long va,
 	if (i == HPTES_PER_GROUP)
 		return -1;
 
-	hpte_v = (va >> 23) << HPTE_V_AVPN_SHIFT | vflags | HPTE_V_VALID;
-	if (vflags & HPTE_V_LARGE)
-		va &= ~(1UL << HPTE_V_AVPN_SHIFT);
-	hpte_r = (prpn << HPTE_R_RPN_SHIFT) | rflags;
+	hpte_v = hpte_encode_v(va, psize) | vflags | HPTE_V_VALID;
+	hpte_r = hpte_encode_r(pa, psize) | rflags;
+
+	if (!(vflags & HPTE_V_BOLTED)) {
+		DBG_LOW(" i=%x hpte_v=%016lx, hpte_r=%016lx\n",
+			i, hpte_v, hpte_r);
+	}
 
 	hptep->r = hpte_r;
 	/* Guarantee the second dword is visible before the valid bit */
@@ -96,6 +181,8 @@ static long native_hpte_remove(unsigned long hpte_group)
 	int slot_offset;
 	unsigned long hpte_v;
 
+	DBG_LOW("    remove(group=%lx)\n", hpte_group);
+
 	/* pick a random entry to start at */
 	slot_offset = mftb() & 0x7;
 
@@ -126,34 +213,51 @@ static long native_hpte_remove(unsigned long hpte_group)
 	return i;
 }
 
-static inline void set_pp_bit(unsigned long pp, hpte_t *addr)
+static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
+				 unsigned long va, int psize, int local)
 {
-	unsigned long old;
-	unsigned long *p = &addr->r;
-
-	__asm__ __volatile__(
-	"1:	ldarx	%0,0,%3\n\
-		rldimi	%0,%2,0,61\n\
-		stdcx.	%0,0,%3\n\
-		bne	1b"
-	: "=&r" (old), "=m" (*p)
-	: "r" (pp), "r" (p), "m" (*p)
-	: "cc");
+	hpte_t *hptep = htab_address + slot;
+	unsigned long hpte_v, want_v;
+	int ret = 0;
+
+	want_v = hpte_encode_v(va, psize);
+
+	DBG_LOW("    update(va=%016lx, avpnv=%016lx, hash=%016lx, newpp=%x)",
+		va, want_v & HPTE_V_AVPN, slot, newpp);
+
+	native_lock_hpte(hptep);
+
+	hpte_v = hptep->v;
+
+	/* Even if we miss, we need to invalidate the TLB */
+	if (!HPTE_V_COMPARE(hpte_v, want_v) || !(hpte_v & HPTE_V_VALID)) {
+		DBG_LOW(" -> miss\n");
+		native_unlock_hpte(hptep);
+		ret = -1;
+	} else {
+		DBG_LOW(" -> hit\n");
+		/* Update the HPTE */
+		hptep->r = (hptep->r & ~(HPTE_R_PP | HPTE_R_N)) |
+			(newpp & (HPTE_R_PP | HPTE_R_N));
+		native_unlock_hpte(hptep);
+	}
+
+	/* Ensure it is out of the tlb too. */
+	tlbie(va, psize, local);
+
+	return ret;
 }
 
-/*
- * Only works on small pages. Yes its ugly to have to check each slot in
- * the group but we only use this during bootup.
- */
-static long native_hpte_find(unsigned long vpn)
+static long native_hpte_find(unsigned long va, int psize)
 {
 	hpte_t *hptep;
 	unsigned long hash;
 	unsigned long i, j;
 	long slot;
-	unsigned long hpte_v;
+	unsigned long want_v, hpte_v;
 
-	hash = hpt_hash(vpn, 0);
+	hash = hpt_hash(va, mmu_psize_defs[psize].shift);
+	want_v = hpte_encode_v(va, psize);
 
 	for (j = 0; j < 2; j++) {
 		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
@@ -161,7 +265,7 @@ static long native_hpte_find(unsigned long vpn)
 			hptep = htab_address + slot;
 			hpte_v = hptep->v;
 
-			if ((HPTE_V_AVPN_VAL(hpte_v) == (vpn >> 11))
+			if (HPTE_V_COMPARE(hpte_v, want_v)
 			    && (hpte_v & HPTE_V_VALID)
 			    && ( !!(hpte_v & HPTE_V_SECONDARY) == j)) {
 				/* HPTE matches */
@@ -177,120 +281,92 @@ static long native_hpte_find(unsigned long vpn)
 	return -1;
 }
 
-static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
-				 unsigned long va, int large, int local)
-{
-	hpte_t *hptep = htab_address + slot;
-	unsigned long hpte_v;
-	unsigned long avpn = va >> 23;
-	int ret = 0;
-
-	if (large)
-		avpn &= ~1;
-
-	native_lock_hpte(hptep);
-
-	hpte_v = hptep->v;
-
-	/* Even if we miss, we need to invalidate the TLB */
-	if ((HPTE_V_AVPN_VAL(hpte_v) != avpn)
-	    || !(hpte_v & HPTE_V_VALID)) {
-		native_unlock_hpte(hptep);
-		ret = -1;
-	} else {
-		set_pp_bit(newpp, hptep);
-		native_unlock_hpte(hptep);
-	}
-
-	/* Ensure it is out of the tlb too */
-	if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) {
-		tlbiel(va);
-	} else {
-		int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
-
-		if (lock_tlbie)
-			spin_lock(&native_tlbie_lock);
-		tlbie(va, large);
-		if (lock_tlbie)
-			spin_unlock(&native_tlbie_lock);
-	}
-
-	return ret;
-}
-
 /*
  * Update the page protection bits. Intended to be used to create
  * guard pages for kernel data structures on pages which are bolted
  * in the HPT. Assumes pages being operated on will not be stolen.
- * Does not work on large pages.
  *
  * No need to lock here because we should be the only user.
  */
-static void native_hpte_updateboltedpp(unsigned long newpp, unsigned long ea)
+static void native_hpte_updateboltedpp(unsigned long newpp, unsigned long ea,
+				       int psize)
 {
-	unsigned long vsid, va, vpn, flags = 0;
+	unsigned long vsid, va;
 	long slot;
 	hpte_t *hptep;
-	int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
 
 	vsid = get_kernel_vsid(ea);
 	va = (vsid << 28) | (ea & 0x0fffffff);
-	vpn = va >> PAGE_SHIFT;
 
-	slot = native_hpte_find(vpn);
+	slot = native_hpte_find(va, psize);
 	if (slot == -1)
 		panic("could not find page to bolt\n");
 	hptep = htab_address + slot;
 
-	set_pp_bit(newpp, hptep);
+	/* Update the HPTE */
+	hptep->r = (hptep->r & ~(HPTE_R_PP | HPTE_R_N)) |
+		(newpp & (HPTE_R_PP | HPTE_R_N));
 
-	/* Ensure it is out of the tlb too */
-	if (lock_tlbie)
-		spin_lock_irqsave(&native_tlbie_lock, flags);
-	tlbie(va, 0);
-	if (lock_tlbie)
-		spin_unlock_irqrestore(&native_tlbie_lock, flags);
+	/* Ensure it is out of the tlb too. */
+	tlbie(va, psize, 0);
 }
 
 static void native_hpte_invalidate(unsigned long slot, unsigned long va,
-				    int large, int local)
+				   int psize, int local)
 {
 	hpte_t *hptep = htab_address + slot;
 	unsigned long hpte_v;
-	unsigned long avpn = va >> 23;
+	unsigned long want_v;
 	unsigned long flags;
-	int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
-
-	if (large)
-		avpn &= ~1;
 
 	local_irq_save(flags);
-	native_lock_hpte(hptep);
 
+	DBG_LOW("    invalidate(va=%016lx, hash: %x)\n", va, slot);
+
+	want_v = hpte_encode_v(va, psize);
+	native_lock_hpte(hptep);
 	hpte_v = hptep->v;
 
 	/* Even if we miss, we need to invalidate the TLB */
-	if ((HPTE_V_AVPN_VAL(hpte_v) != avpn)
-	    || !(hpte_v & HPTE_V_VALID)) {
+	if (!HPTE_V_COMPARE(hpte_v, want_v) || !(hpte_v & HPTE_V_VALID))
 		native_unlock_hpte(hptep);
-	} else {
+	else
 		/* Invalidate the hpte. NOTE: this also unlocks it */
 		hptep->v = 0;
-	}
 
-	/* Invalidate the tlb */
-	if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) {
-		tlbiel(va);
-	} else {
-		if (lock_tlbie)
-			spin_lock(&native_tlbie_lock);
-		tlbie(va, large);
-		if (lock_tlbie)
-			spin_unlock(&native_tlbie_lock);
-	}
+	/* Invalidate the TLB */
+	tlbie(va, psize, local);
+
 	local_irq_restore(flags);
 }
 
+/*
+ * XXX This need fixing based on page size. It's only used by
+ * native_hpte_clear() for now which needs fixing too so they
+ * make a good pair...
+ */
+static unsigned long slot2va(unsigned long hpte_v, unsigned long slot)
+{
+	unsigned long avpn = HPTE_V_AVPN_VAL(hpte_v);
+	unsigned long va;
+
+	va = avpn << 23;
+
+	if (! (hpte_v & HPTE_V_LARGE)) {
+		unsigned long vpi, pteg;
+
+		pteg = slot / HPTES_PER_GROUP;
+		if (hpte_v & HPTE_V_SECONDARY)
+			pteg = ~pteg;
+
+		vpi = ((va >> 28) ^ pteg) & htab_hash_mask;
+
+		va |= vpi << PAGE_SHIFT;
+	}
+
+	return va;
+}
+
 /*
  * clear all mappings on kexec.  All cpus are in real mode (or they will
  * be when they isi), and we are the only one left.  We rely on our kernel
@@ -298,6 +374,8 @@ static void native_hpte_invalidate(unsigned long slot, unsigned long va,
  *
  * TODO: add batching support when enabled.  remember, no dynamic memory here,
  * athough there is the control page available...
+ *
+ * XXX FIXME: 4k only for now !
  */
 static void native_hpte_clear(void)
 {
@@ -327,7 +405,7 @@ static void native_hpte_clear(void)
 
 		if (hpte_v & HPTE_V_VALID) {
 			hptep->v = 0;
-			tlbie(slot2va(hpte_v, slot), hpte_v & HPTE_V_LARGE);
+			tlbie(slot2va(hpte_v, slot), MMU_PAGE_4K, 0);
 		}
 	}
 
@@ -335,59 +413,59 @@ static void native_hpte_clear(void)
 	local_irq_restore(flags);
 }
 
+/*
+ * Batched hash table flush, we batch the tlbie's to avoid taking/releasing
+ * the lock all the time
+ */
 static void native_flush_hash_range(unsigned long number, int local)
 {
-	unsigned long va, vpn, hash, secondary, slot, flags, avpn;
-	int i, j;
+	unsigned long va, hash, index, hidx, shift, slot;
 	hpte_t *hptep;
 	unsigned long hpte_v;
+	unsigned long want_v;
+	unsigned long flags;
+	real_pte_t pte;
 	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
-	unsigned long large = batch->large;
+	unsigned long psize = batch->psize;
+	int i;
 
 	local_irq_save(flags);
 
-	j = 0;
 	for (i = 0; i < number; i++) {
-		va = batch->vaddr[j];
-		if (large)
-			vpn = va >> HPAGE_SHIFT;
-		else
-			vpn = va >> PAGE_SHIFT;
-		hash = hpt_hash(vpn, large);
-		secondary = (pte_val(batch->pte[i]) & _PAGE_SECONDARY) >> 15;
-		if (secondary)
+		va = batch->vaddr[i];
+		pte = batch->pte[i];
+
+		pte_iterate_hashed_subpages(pte, psize, va, index, shift) {
+			hash = hpt_hash(va, shift);
+			hidx = __rpte_to_hidx(pte, index);
+			if (hidx & _PTEIDX_SECONDARY)
 				hash = ~hash;
 			slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
-		slot += (pte_val(batch->pte[i]) & _PAGE_GROUP_IX) >> 12;
-
+			slot += hidx & _PTEIDX_GROUP_IX;
 			hptep = htab_address + slot;
-
-		avpn = va >> 23;
-		if (large)
-			avpn &= ~0x1UL;
-
+			want_v = hpte_encode_v(va, psize);
 			native_lock_hpte(hptep);
-
 			hpte_v = hptep->v;
-
-		/* Even if we miss, we need to invalidate the TLB */
-		if ((HPTE_V_AVPN_VAL(hpte_v) != avpn)
-		    || !(hpte_v & HPTE_V_VALID)) {
+			if (!HPTE_V_COMPARE(hpte_v, want_v) ||
+			    !(hpte_v & HPTE_V_VALID))
 				native_unlock_hpte(hptep);
-		} else {
-			/* Invalidate the hpte. NOTE: this also unlocks it */
+			else
 				hptep->v = 0;
+		} pte_iterate_hashed_end();
 	}
 
-		j++;
-	}
-
-	if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) {
+	if (cpu_has_feature(CPU_FTR_TLBIEL) &&
+	    mmu_psize_defs[psize].tlbiel && local) {
 		asm volatile("ptesync":::"memory");
+		for (i = 0; i < number; i++) {
+			va = batch->vaddr[i];
+			pte = batch->pte[i];
 
-		for (i = 0; i < j; i++)
-			__tlbiel(batch->vaddr[i]);
-
+			pte_iterate_hashed_subpages(pte, psize, va, index,
+						    shift) {
+				__tlbiel(va, psize);
+			} pte_iterate_hashed_end();
+		}
 		asm volatile("ptesync":::"memory");
 	} else {
 		int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE);
@@ -396,10 +474,15 @@ static void native_flush_hash_range(unsigned long number, int local)
 			spin_lock(&native_tlbie_lock);
 
 		asm volatile("ptesync":::"memory");
+		for (i = 0; i < number; i++) {
+			va = batch->vaddr[i];
+			pte = batch->pte[i];
 
-		for (i = 0; i < j; i++)
-			__tlbie(batch->vaddr[i], large);
-
+			pte_iterate_hashed_subpages(pte, psize, va, index,
+						    shift) {
+				__tlbie(va, psize);
+			} pte_iterate_hashed_end();
+		}
 		asm volatile("eieio; tlbsync; ptesync":::"memory");
 
 		if (lock_tlbie)

arch/powerpc/mm/hash_utils_64.c

@@ -19,6 +19,7 @@
  */
 
 #undef DEBUG
+#undef DEBUG_LOW
 
 #include <linux/config.h>
 #include <linux/spinlock.h>
@@ -59,6 +60,15 @@
 #define DBG(fmt...)
 #endif
 
+#ifdef DEBUG_LOW
+#define DBG_LOW(fmt...) udbg_printf(fmt)
+#else
+#define DBG_LOW(fmt...)
+#endif
+
+#define KB (1024)
+#define MB (1024*KB)
+
 /*
  * Note:  pte   --> Linux PTE
  *        HPTE  --> PowerPC Hashed Page Table Entry
@@ -77,91 +87,290 @@ extern unsigned long dart_tablebase;
 
 hpte_t *htab_address;
 unsigned long htab_hash_mask;
-
 unsigned long _SDR1;
+struct mmu_psize_def mmu_psize_defs[MMU_PAGE_COUNT];
+int mmu_linear_psize = MMU_PAGE_4K;
+int mmu_virtual_psize = MMU_PAGE_4K;
+#ifdef CONFIG_HUGETLB_PAGE
+int mmu_huge_psize = MMU_PAGE_16M;
+unsigned int HPAGE_SHIFT;
+#endif
 
-#define KB (1024)
-#define MB (1024*KB)
-
-static inline void loop_forever(void)
-{
-	volatile unsigned long x = 1;
-	for(;x;x|=1)
-		;
-}
+/* There are definitions of page sizes arrays to be used when none
+ * is provided by the firmware.
+ */
 
-static inline void create_pte_mapping(unsigned long start, unsigned long end,
-				      unsigned long mode, int large)
+/* Pre-POWER4 CPUs (4k pages only)
+ */
+struct mmu_psize_def mmu_psize_defaults_old[] = {
+	[MMU_PAGE_4K] = {
+		.shift	= 12,
+		.sllp	= 0,
+		.penc	= 0,
+		.avpnm	= 0,
+		.tlbiel = 0,
+	},
+};
+
+/* POWER4, GPUL, POWER5
+ *
+ * Support for 16Mb large pages
+ */
+struct mmu_psize_def mmu_psize_defaults_gp[] = {
+	[MMU_PAGE_4K] = {
+		.shift	= 12,
+		.sllp	= 0,
+		.penc	= 0,
+		.avpnm	= 0,
+		.tlbiel = 1,
+	},
+	[MMU_PAGE_16M] = {
+		.shift	= 24,
+		.sllp	= SLB_VSID_L,
+		.penc	= 0,
+		.avpnm	= 0x1UL,
+		.tlbiel = 0,
+	},
+};
+
+
+int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
+		      unsigned long pstart, unsigned long mode, int psize)
 {
-	unsigned long addr;
-	unsigned int step;
+	unsigned long vaddr, paddr;
+	unsigned int step, shift;
 	unsigned long tmp_mode;
-	unsigned long vflags;
+	int ret = 0;
 
-	if (large) {
-		step = 16*MB;
-		vflags = HPTE_V_BOLTED | HPTE_V_LARGE;
-	} else {
-		step = 4*KB;
-		vflags = HPTE_V_BOLTED;
-	}
+	shift = mmu_psize_defs[psize].shift;
+	step = 1 << shift;
 
-	for (addr = start; addr < end; addr += step) {
+	for (vaddr = vstart, paddr = pstart; vaddr < vend;
+	     vaddr += step, paddr += step) {
 		unsigned long vpn, hash, hpteg;
-		unsigned long vsid = get_kernel_vsid(addr);
-		unsigned long va = (vsid << 28) | (addr & 0xfffffff);
-		int ret = -1;
-
-		if (large)
-			vpn = va >> HPAGE_SHIFT;
-		else
-			vpn = va >> PAGE_SHIFT;
-
+		unsigned long vsid = get_kernel_vsid(vaddr);
+		unsigned long va = (vsid << 28) | (vaddr & 0x0fffffff);
 
+		vpn = va >> shift;
 		tmp_mode = mode;
 		
 		/* Make non-kernel text non-executable */
-		if (!in_kernel_text(addr))
-			tmp_mode = mode | HW_NO_EXEC;
-
-		hash = hpt_hash(vpn, large);
+		if (!in_kernel_text(vaddr))
+			tmp_mode = mode | HPTE_R_N;
 
+		hash = hpt_hash(va, shift);
 		hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP);
 
+		/* The crap below can be cleaned once ppd_md.probe() can
+		 * set up the hash callbacks, thus we can just used the
+		 * normal insert callback here.
+		 */
 #ifdef CONFIG_PPC_ISERIES
-		if (systemcfg->platform & PLATFORM_ISERIES_LPAR)
-			ret = iSeries_hpte_bolt_or_insert(hpteg, va,
-				virt_to_abs(addr) >> PAGE_SHIFT,
-				vflags, tmp_mode);
+		if (systemcfg->platform == PLATFORM_ISERIES_LPAR)
+			ret = iSeries_hpte_insert(hpteg, va,
+						  virt_to_abs(paddr),
+						  tmp_mode,
+						  HPTE_V_BOLTED,
+						  psize);
 		else
 #endif
 #ifdef CONFIG_PPC_PSERIES
 		if (systemcfg->platform & PLATFORM_LPAR)
 			ret = pSeries_lpar_hpte_insert(hpteg, va,
-				virt_to_abs(addr) >> PAGE_SHIFT,
-				vflags, tmp_mode);
+						       virt_to_abs(paddr),
+						       tmp_mode,
+						       HPTE_V_BOLTED,
+						       psize);
 		else
 #endif
 #ifdef CONFIG_PPC_MULTIPLATFORM
 			ret = native_hpte_insert(hpteg, va,
-				virt_to_abs(addr) >> PAGE_SHIFT,
-				vflags, tmp_mode);
+						 virt_to_abs(paddr),
+						 tmp_mode, HPTE_V_BOLTED,
+						 psize);
 #endif
+		if (ret < 0)
+			break;
+	}
+	return ret < 0 ? ret : 0;
+}
+
+static int __init htab_dt_scan_page_sizes(unsigned long node,
+					  const char *uname, int depth,
+					  void *data)
+{
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
+	u32 *prop;
+	unsigned long size = 0;
+
+	/* We are scanning "cpu" nodes only */
+	if (type == NULL || strcmp(type, "cpu") != 0)
+		return 0;
+
+	prop = (u32 *)of_get_flat_dt_prop(node,
+					  "ibm,segment-page-sizes", &size);
+	if (prop != NULL) {
+		DBG("Page sizes from device-tree:\n");
+		size /= 4;
+		cur_cpu_spec->cpu_features &= ~(CPU_FTR_16M_PAGE);
+		while(size > 0) {
+			unsigned int shift = prop[0];
+			unsigned int slbenc = prop[1];
+			unsigned int lpnum = prop[2];
+			unsigned int lpenc = 0;
+			struct mmu_psize_def *def;
+			int idx = -1;
+
+			size -= 3; prop += 3;
+			while(size > 0 && lpnum) {
+				if (prop[0] == shift)
+					lpenc = prop[1];
+				prop += 2; size -= 2;
+				lpnum--;
+			}
+			switch(shift) {
+			case 0xc:
+				idx = MMU_PAGE_4K;
+				break;
+			case 0x10:
+				idx = MMU_PAGE_64K;
+				break;
+			case 0x14:
+				idx = MMU_PAGE_1M;
+				break;
+			case 0x18:
+				idx = MMU_PAGE_16M;
+				cur_cpu_spec->cpu_features |= CPU_FTR_16M_PAGE;
+				break;
+			case 0x22:
+				idx = MMU_PAGE_16G;
+				break;
+			}
+			if (idx < 0)
+				continue;
+			def = &mmu_psize_defs[idx];
+			def->shift = shift;
+			if (shift <= 23)
+				def->avpnm = 0;
+			else
+				def->avpnm = (1 << (shift - 23)) - 1;
+			def->sllp = slbenc;
+			def->penc = lpenc;
+			/* We don't know for sure what's up with tlbiel, so
+			 * for now we only set it for 4K and 64K pages
+			 */
+			if (idx == MMU_PAGE_4K || idx == MMU_PAGE_64K)
+				def->tlbiel = 1;
+			else
+				def->tlbiel = 0;
 
-		if (ret == -1) {
-			ppc64_terminate_msg(0x20, "create_pte_mapping");
-			loop_forever();
+			DBG(" %d: shift=%02x, sllp=%04x, avpnm=%08x, "
+			    "tlbiel=%d, penc=%d\n",
+			    idx, shift, def->sllp, def->avpnm, def->tlbiel,
+			    def->penc);
 		}
+		return 1;
 	}
+	return 0;
+}
+
+
+static void __init htab_init_page_sizes(void)
+{
+	int rc;
+
+	/* Default to 4K pages only */
+	memcpy(mmu_psize_defs, mmu_psize_defaults_old,
+	       sizeof(mmu_psize_defaults_old));
+
+	/*
+	 * Try to find the available page sizes in the device-tree
+	 */
+	rc = of_scan_flat_dt(htab_dt_scan_page_sizes, NULL);
+	if (rc != 0)  /* Found */
+		goto found;
+
+	/*
+	 * Not in the device-tree, let's fallback on known size
+	 * list for 16M capable GP & GR
+	 */
+	if ((systemcfg->platform != PLATFORM_ISERIES_LPAR) &&
+	    cpu_has_feature(CPU_FTR_16M_PAGE))
+		memcpy(mmu_psize_defs, mmu_psize_defaults_gp,
+		       sizeof(mmu_psize_defaults_gp));
+ found:
+	/*
+	 * Pick a size for the linear mapping. Currently, we only support
+	 * 16M, 1M and 4K which is the default
+	 */
+	if (mmu_psize_defs[MMU_PAGE_16M].shift)
+		mmu_linear_psize = MMU_PAGE_16M;
+	else if (mmu_psize_defs[MMU_PAGE_1M].shift)
+		mmu_linear_psize = MMU_PAGE_1M;
+
+	/*
+	 * Pick a size for the ordinary pages. Default is 4K, we support
+	 * 64K if cache inhibited large pages are supported by the
+	 * processor
+	 */
+#ifdef CONFIG_PPC_64K_PAGES
+	if (mmu_psize_defs[MMU_PAGE_64K].shift &&
+	    cpu_has_feature(CPU_FTR_CI_LARGE_PAGE))
+		mmu_virtual_psize = MMU_PAGE_64K;
+#endif
+
+	printk(KERN_INFO "Page orders: linear mapping = %d, others = %d\n",
+	       mmu_psize_defs[mmu_linear_psize].shift,
+	       mmu_psize_defs[mmu_virtual_psize].shift);
+
+#ifdef CONFIG_HUGETLB_PAGE
+	/* Init large page size. Currently, we pick 16M or 1M depending
+	 * on what is available
+	 */
+	if (mmu_psize_defs[MMU_PAGE_16M].shift)
+		mmu_huge_psize = MMU_PAGE_16M;
+	else if (mmu_psize_defs[MMU_PAGE_1M].shift)
+		mmu_huge_psize = MMU_PAGE_1M;
+
+	/* Calculate HPAGE_SHIFT and sanity check it */
+	if (mmu_psize_defs[mmu_huge_psize].shift > 16 &&
+	    mmu_psize_defs[mmu_huge_psize].shift < 28)
+		HPAGE_SHIFT = mmu_psize_defs[mmu_huge_psize].shift;
+	else
+		HPAGE_SHIFT = 0; /* No huge pages dude ! */
+#endif /* CONFIG_HUGETLB_PAGE */
 }
 
-static unsigned long get_hashtable_size(void)
+static int __init htab_dt_scan_pftsize(unsigned long node,
+				       const char *uname, int depth,
+				       void *data)
+{
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
+	u32 *prop;
+
+	/* We are scanning "cpu" nodes only */
+	if (type == NULL || strcmp(type, "cpu") != 0)
+		return 0;
+
+	prop = (u32 *)of_get_flat_dt_prop(node, "ibm,pft-size", NULL);
+	if (prop != NULL) {
+		/* pft_size[0] is the NUMA CEC cookie */
+		ppc64_pft_size = prop[1];
+		return 1;
+	}
+	return 0;
+}
+
+static unsigned long __init htab_get_table_size(void)
 {
 	unsigned long rnd_mem_size, pteg_count;
 
-	/* If hash size wasn't obtained in prom.c, we calculate it now based on
-	 * the total RAM size
+	/* If hash size isn't already provided by the platform, we try to
+	 * retreive it from the device-tree. If it's not there neither, we
+	 * calculate it now based on the total RAM size
 	 */
+	if (ppc64_pft_size == 0)
+		of_scan_flat_dt(htab_dt_scan_pftsize, NULL);
 	if (ppc64_pft_size)
 		return 1UL << ppc64_pft_size;
 
@@ -181,17 +390,21 @@ void __init htab_initialize(void)
 	unsigned long table, htab_size_bytes;
 	unsigned long pteg_count;
 	unsigned long mode_rw;
-	int i, use_largepages = 0;
 	unsigned long base = 0, size = 0;
+	int i;
+
 	extern unsigned long tce_alloc_start, tce_alloc_end;
 
 	DBG(" -> htab_initialize()\n");
 
+	/* Initialize page sizes */
+	htab_init_page_sizes();
+
 	/*
 	 * Calculate the required size of the htab.  We want the number of
 	 * PTEGs to equal one half the number of real pages.
 	 */ 
-	htab_size_bytes = get_hashtable_size();
+	htab_size_bytes = htab_get_table_size();
 	pteg_count = htab_size_bytes >> 7;
 
 	/* For debug, make the HTAB 1/8 as big as it normally would be. */
@@ -211,14 +424,11 @@ void __init htab_initialize(void)
 		 * the absolute address space.
 		 */
 		table = lmb_alloc(htab_size_bytes, htab_size_bytes);
+		BUG_ON(table == 0);
 
 		DBG("Hash table allocated at %lx, size: %lx\n", table,
 		    htab_size_bytes);
 
-		if ( !table ) {
-			ppc64_terminate_msg(0x20, "hpt space");
-			loop_forever();
-		}
 		htab_address = abs_to_virt(table);
 
 		/* htab absolute addr + encoded htabsize */
@@ -234,8 +444,6 @@ void __init htab_initialize(void)
 	 * _NOT_ map it to avoid cache paradoxes as it's remapped non
 	 * cacheable later on
 	 */
-	if (cpu_has_feature(CPU_FTR_16M_PAGE))
-		use_largepages = 1;
 
 	/* create bolted the linear mapping in the hash table */
 	for (i=0; i < lmb.memory.cnt; i++) {
@@ -246,26 +454,31 @@ void __init htab_initialize(void)
 
 #ifdef CONFIG_U3_DART
 		/* Do not map the DART space. Fortunately, it will be aligned
-		 * in such a way that it will not cross two lmb regions and will
-		 * fit within a single 16Mb page.
-		 * The DART space is assumed to be a full 16Mb region even if we
-		 * only use 2Mb of that space. We will use more of it later for
-		 * AGP GART. We have to use a full 16Mb large page.
+		 * in such a way that it will not cross two lmb regions and
+		 * will fit within a single 16Mb page.
+		 * The DART space is assumed to be a full 16Mb region even if
+		 * we only use 2Mb of that space. We will use more of it later
+		 * for AGP GART. We have to use a full 16Mb large page.
 		 */
 		DBG("DART base: %lx\n", dart_tablebase);
 
 		if (dart_tablebase != 0 && dart_tablebase >= base
 		    && dart_tablebase < (base + size)) {
 			if (base != dart_tablebase)
-				create_pte_mapping(base, dart_tablebase, mode_rw,
-						   use_largepages);
+				BUG_ON(htab_bolt_mapping(base, dart_tablebase,
+							 base, mode_rw,
+							 mmu_linear_psize));
 			if ((base + size) > (dart_tablebase + 16*MB))
-				create_pte_mapping(dart_tablebase + 16*MB, base + size,
-						   mode_rw, use_largepages);
+				BUG_ON(htab_bolt_mapping(dart_tablebase+16*MB,
+							 base + size,
+							 dart_tablebase+16*MB,
+							 mode_rw,
+							 mmu_linear_psize));
 			continue;
 		}
 #endif /* CONFIG_U3_DART */
-		create_pte_mapping(base, base + size, mode_rw, use_largepages);
+		BUG_ON(htab_bolt_mapping(base, base + size, base,
+					 mode_rw, mmu_linear_psize));
        }
 
 	/*
@@ -282,8 +495,9 @@ void __init htab_initialize(void)
 		if (base + size >= tce_alloc_start)
 			tce_alloc_start = base + size + 1;
 
-		create_pte_mapping(tce_alloc_start, tce_alloc_end,
-			mode_rw, use_largepages);
+ 		BUG_ON(htab_bolt_mapping(tce_alloc_start, tce_alloc_end,
+ 					 tce_alloc_start, mode_rw,
+					 mmu_linear_psize));
 	}
 
 	DBG(" <- htab_initialize()\n");
@@ -298,9 +512,6 @@ unsigned int hash_page_do_lazy_icache(unsigned int pp, pte_t pte, int trap)
 {
 	struct page *page;
 
-	if (!pfn_valid(pte_pfn(pte)))
-		return pp;
-
 	page = pte_page(pte);
 
 	/* page is dirty */
@@ -309,7 +520,7 @@ unsigned int hash_page_do_lazy_icache(unsigned int pp, pte_t pte, int trap)
 			__flush_dcache_icache(page_address(page));
 			set_bit(PG_arch_1, &page->flags);
 		} else
-			pp |= HW_NO_EXEC;
+			pp |= HPTE_R_N;
 	}
 	return pp;
 }
@@ -325,94 +536,169 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 	unsigned long vsid;
 	struct mm_struct *mm;
 	pte_t *ptep;
-	int ret;
-	int user_region = 0;
-	int local = 0;
 	cpumask_t tmp;
+	int rc, user_region = 0, local = 0;
 
-	if ((ea & ~REGION_MASK) >= PGTABLE_RANGE)
+	DBG_LOW("hash_page(ea=%016lx, access=%lx, trap=%lx\n",
+		ea, access, trap);
+
+	if ((ea & ~REGION_MASK) >= PGTABLE_RANGE) {
+		DBG_LOW(" out of pgtable range !\n");
  		return 1;
+	}
 
+	/* Get region & vsid */
  	switch (REGION_ID(ea)) {
 	case USER_REGION_ID:
 		user_region = 1;
 		mm = current->mm;
-		if (! mm)
+		if (! mm) {
+			DBG_LOW(" user region with no mm !\n");
 			return 1;
-
+		}
 		vsid = get_vsid(mm->context.id, ea);
 		break;
 	case VMALLOC_REGION_ID:
 		mm = &init_mm;
 		vsid = get_kernel_vsid(ea);
 		break;
-#if 0
-	case KERNEL_REGION_ID:
-		/*
-		 * Should never get here - entire 0xC0... region is bolted.
-		 * Send the problem up to do_page_fault 
-		 */
-#endif
 	default:
 		/* Not a valid range
 		 * Send the problem up to do_page_fault 
 		 */
 		return 1;
-		break;
 	}
+	DBG_LOW(" mm=%p, mm->pgdir=%p, vsid=%016lx\n", mm, mm->pgd, vsid);
 
+	/* Get pgdir */
 	pgdir = mm->pgd;
-
 	if (pgdir == NULL)
 		return 1;
 
+	/* Check CPU locality */
 	tmp = cpumask_of_cpu(smp_processor_id());
 	if (user_region && cpus_equal(mm->cpu_vm_mask, tmp))
 		local = 1;
 
-	/* Is this a huge page ? */
-	if (unlikely(in_hugepage_area(mm->context, ea)))
-		ret = hash_huge_page(mm, access, ea, vsid, local);
-	else {
+	/* Handle hugepage regions */
+	if (unlikely(in_hugepage_area(mm->context, ea))) {
+		DBG_LOW(" -> huge page !\n");
+		return hash_huge_page(mm, access, ea, vsid, local);
+	}
+
+	/* Get PTE and page size from page tables */
 	ptep = find_linux_pte(pgdir, ea);
-		if (ptep == NULL)
+	if (ptep == NULL || !pte_present(*ptep)) {
+		DBG_LOW(" no PTE !\n");
+		return 1;
+	}
+
+#ifndef CONFIG_PPC_64K_PAGES
+	DBG_LOW(" i-pte: %016lx\n", pte_val(*ptep));
+#else
+	DBG_LOW(" i-pte: %016lx %016lx\n", pte_val(*ptep),
+		pte_val(*(ptep + PTRS_PER_PTE)));
+#endif
+	/* Pre-check access permissions (will be re-checked atomically
+	 * in __hash_page_XX but this pre-check is a fast path
+	 */
+	if (access & ~pte_val(*ptep)) {
+		DBG_LOW(" no access !\n");
 		return 1;
-		ret = __hash_page(ea, access, vsid, ptep, trap, local);
 	}
 
-	return ret;
+	/* Do actual hashing */
+#ifndef CONFIG_PPC_64K_PAGES
+	rc = __hash_page_4K(ea, access, vsid, ptep, trap, local);
+#else
+	if (mmu_virtual_psize == MMU_PAGE_64K)
+		rc = __hash_page_64K(ea, access, vsid, ptep, trap, local);
+	else
+		rc = __hash_page_4K(ea, access, vsid, ptep, trap, local);
+#endif /* CONFIG_PPC_64K_PAGES */
+
+#ifndef CONFIG_PPC_64K_PAGES
+	DBG_LOW(" o-pte: %016lx\n", pte_val(*ptep));
+#else
+	DBG_LOW(" o-pte: %016lx %016lx\n", pte_val(*ptep),
+		pte_val(*(ptep + PTRS_PER_PTE)));
+#endif
+	DBG_LOW(" -> rc=%d\n", rc);
+	return rc;
 }
 
-void flush_hash_page(unsigned long va, pte_t pte, int local)
+void hash_preload(struct mm_struct *mm, unsigned long ea,
+		  unsigned long access, unsigned long trap)
 {
-	unsigned long vpn, hash, secondary, slot;
-	unsigned long huge = pte_huge(pte);
+	unsigned long vsid;
+	void *pgdir;
+	pte_t *ptep;
+	cpumask_t mask;
+	unsigned long flags;
+	int local = 0;
+
+	/* We don't want huge pages prefaulted for now
+	 */
+	if (unlikely(in_hugepage_area(mm->context, ea)))
+		return;
+
+	DBG_LOW("hash_preload(mm=%p, mm->pgdir=%p, ea=%016lx, access=%lx,"
+		" trap=%lx\n", mm, mm->pgd, ea, access, trap);
 
-	if (huge)
-		vpn = va >> HPAGE_SHIFT;
+	/* Get PTE, VSID, access mask */
+	pgdir = mm->pgd;
+	if (pgdir == NULL)
+		return;
+	ptep = find_linux_pte(pgdir, ea);
+	if (!ptep)
+		return;
+	vsid = get_vsid(mm->context.id, ea);
+
+	/* Hash it in */
+	local_irq_save(flags);
+	mask = cpumask_of_cpu(smp_processor_id());
+	if (cpus_equal(mm->cpu_vm_mask, mask))
+		local = 1;
+#ifndef CONFIG_PPC_64K_PAGES
+	__hash_page_4K(ea, access, vsid, ptep, trap, local);
+#else
+	if (mmu_virtual_psize == MMU_PAGE_64K)
+		__hash_page_64K(ea, access, vsid, ptep, trap, local);
 	else
-		vpn = va >> PAGE_SHIFT;
-	hash = hpt_hash(vpn, huge);
-	secondary = (pte_val(pte) & _PAGE_SECONDARY) >> 15;
-	if (secondary)
+		__hash_page_4K(ea, access, vsid, ptep, trap, local);
+#endif /* CONFIG_PPC_64K_PAGES */
+	local_irq_restore(flags);
+}
+
+void flush_hash_page(unsigned long va, real_pte_t pte, int psize, int local)
+{
+	unsigned long hash, index, shift, hidx, slot;
+
+	DBG_LOW("flush_hash_page(va=%016x)\n", va);
+	pte_iterate_hashed_subpages(pte, psize, va, index, shift) {
+		hash = hpt_hash(va, shift);
+		hidx = __rpte_to_hidx(pte, index);
+		if (hidx & _PTEIDX_SECONDARY)
 			hash = ~hash;
 		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
-	slot += (pte_val(pte) & _PAGE_GROUP_IX) >> 12;
-
-	ppc_md.hpte_invalidate(slot, va, huge, local);
+		slot += hidx & _PTEIDX_GROUP_IX;
+		DBG_LOW(" sub %d: hash=%x, hidx=%x\n", index, slot, hidx);
+		ppc_md.hpte_invalidate(slot, va, psize, local);
+	} pte_iterate_hashed_end();
 }
 
 void flush_hash_range(unsigned long number, int local)
 {
-	if (ppc_md.flush_hash_range) {
+	if (ppc_md.flush_hash_range)
 		ppc_md.flush_hash_range(number, local);
-	} else {
+	else {
 		int i;
 		struct ppc64_tlb_batch *batch =
 			&__get_cpu_var(ppc64_tlb_batch);
 
 		for (i = 0; i < number; i++)
-			flush_hash_page(batch->vaddr[i], batch->pte[i], local);
+			flush_hash_page(batch->vaddr[i], batch->pte[i],
+					batch->psize, local);
 	}
 }
 
@@ -452,6 +738,18 @@ void __init htab_finish_init(void)
 	extern unsigned int *htab_call_hpte_remove;
 	extern unsigned int *htab_call_hpte_updatepp;
 
+#ifdef CONFIG_PPC_64K_PAGES
+	extern unsigned int *ht64_call_hpte_insert1;
+	extern unsigned int *ht64_call_hpte_insert2;
+	extern unsigned int *ht64_call_hpte_remove;
+	extern unsigned int *ht64_call_hpte_updatepp;
+
+	make_bl(ht64_call_hpte_insert1, ppc_md.hpte_insert);
+	make_bl(ht64_call_hpte_insert2, ppc_md.hpte_insert);
+	make_bl(ht64_call_hpte_remove, ppc_md.hpte_remove);
+	make_bl(ht64_call_hpte_updatepp, ppc_md.hpte_updatepp);
+#endif /* CONFIG_PPC_64K_PAGES */
+
 	make_bl(htab_call_hpte_insert1, ppc_md.hpte_insert);
 	make_bl(htab_call_hpte_insert2, ppc_md.hpte_insert);
 	make_bl(htab_call_hpte_remove, ppc_md.hpte_remove);

arch/powerpc/mm/hugetlbpage.c

@@ -47,10 +47,25 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
 		pu = pud_offset(pg, addr);
 		if (!pud_none(*pu)) {
 			pm = pmd_offset(pu, addr);
+#ifdef CONFIG_PPC_64K_PAGES
+			/* Currently, we use the normal PTE offset within full
+			 * size PTE pages, thus our huge PTEs are scattered in
+			 * the PTE page and we do waste some. We may change
+			 * that in the future, but the current mecanism keeps
+			 * things much simpler
+			 */
+			if (!pmd_none(*pm)) {
+				/* Note: pte_offset_* are all equivalent on
+				 * ppc64 as we don't have HIGHMEM
+				 */
+				pt = pte_offset_kernel(pm, addr);
+				return pt;
+			}
+#else /* CONFIG_PPC_64K_PAGES */
+			/* On 4k pages, we put huge PTEs in the PMD page */
 			pt = (pte_t *)pm;
-			BUG_ON(!pmd_none(*pm)
-			       && !(pte_present(*pt) && pte_huge(*pt)));
 			return pt;
+#endif /* CONFIG_PPC_64K_PAGES */
 		}
 	}
 
@@ -74,9 +89,16 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr)
 	if (pu) {
 		pm = pmd_alloc(mm, pu, addr);
 		if (pm) {
+#ifdef CONFIG_PPC_64K_PAGES
+			/* See comment in huge_pte_offset. Note that if we ever
+			 * want to put the page size in the PMD, we would have
+			 * to open code our own pte_alloc* function in order
+			 * to populate and set the size atomically
+			 */
+			pt = pte_alloc_map(mm, pm, addr);
+#else /* CONFIG_PPC_64K_PAGES */
 			pt = (pte_t *)pm;
-			BUG_ON(!pmd_none(*pm)
-			       && !(pte_present(*pt) && pte_huge(*pt)));
+#endif /* CONFIG_PPC_64K_PAGES */
 			return pt;
 		}
 	}
@@ -84,35 +106,29 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr)
 	return NULL;
 }
 
-#define HUGEPTE_BATCH_SIZE	(HPAGE_SIZE / PMD_SIZE)
-
 void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
 		     pte_t *ptep, pte_t pte)
 {
-	int i;
-
 	if (pte_present(*ptep)) {
-		pte_clear(mm, addr, ptep);
+		/* We open-code pte_clear because we need to pass the right
+		 * argument to hpte_update (huge / !huge)
+		 */
+		unsigned long old = pte_update(ptep, ~0UL);
+		if (old & _PAGE_HASHPTE)
+			hpte_update(mm, addr & HPAGE_MASK, ptep, old, 1);
 		flush_tlb_pending();
 	}
-
-	for (i = 0; i < HUGEPTE_BATCH_SIZE; i++) {
 	*ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
-		ptep++;
-	}
 }
 
 pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 			      pte_t *ptep)
 {
 	unsigned long old = pte_update(ptep, ~0UL);
-	int i;
 
 	if (old & _PAGE_HASHPTE)
-		hpte_update(mm, addr, old, 0);
-
-	for (i = 1; i < HUGEPTE_BATCH_SIZE; i++)
-		ptep[i] = __pte(0);
+		hpte_update(mm, addr & HPAGE_MASK, ptep, old, 1);
+	*ptep = __pte(0);
 
 	return __pte(old);
 }
@@ -563,6 +579,8 @@ unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
 	int lastshift;
 	u16 areamask, curareas;
 
+	if (HPAGE_SHIFT == 0)
+		return -EINVAL;
 	if (len & ~HPAGE_MASK)
 		return -EINVAL;
 
@@ -619,19 +637,15 @@ int hash_huge_page(struct mm_struct *mm, unsigned long access,
 		   unsigned long ea, unsigned long vsid, int local)
 {
 	pte_t *ptep;
-	unsigned long va, vpn;
-	pte_t old_pte, new_pte;
-	unsigned long rflags, prpn;
+	unsigned long old_pte, new_pte;
+	unsigned long va, rflags, pa;
 	long slot;
 	int err = 1;
 
-	spin_lock(&mm->page_table_lock);
-
 	ptep = huge_pte_offset(mm, ea);
 
 	/* Search the Linux page table for a match with va */
 	va = (vsid << 28) | (ea & 0x0fffffff);
-	vpn = va >> HPAGE_SHIFT;
 
 	/*
 	 * If no pte found or not present, send the problem up to
@@ -640,8 +654,6 @@ int hash_huge_page(struct mm_struct *mm, unsigned long access,
 	if (unlikely(!ptep || pte_none(*ptep)))
 		goto out;
 
-/* 	BUG_ON(pte_bad(*ptep)); */
-
 	/* 
 	 * Check the user's access rights to the page.  If access should be
 	 * prevented then send the problem up to do_page_fault.
@@ -661,58 +673,64 @@ int hash_huge_page(struct mm_struct *mm, unsigned long access,
 	 */
 
 
-	old_pte = *ptep;
-	new_pte = old_pte;
+	do {
+		old_pte = pte_val(*ptep);
+		if (old_pte & _PAGE_BUSY)
+			goto out;
+		new_pte = old_pte | _PAGE_BUSY |
+			_PAGE_ACCESSED | _PAGE_HASHPTE;
+	} while(old_pte != __cmpxchg_u64((unsigned long *)ptep,
+					 old_pte, new_pte));
 
-	rflags = 0x2 | (! (pte_val(new_pte) & _PAGE_RW));
+	rflags = 0x2 | (!(new_pte & _PAGE_RW));
  	/* _PAGE_EXEC -> HW_NO_EXEC since it's inverted */
-	rflags |= ((pte_val(new_pte) & _PAGE_EXEC) ? 0 : HW_NO_EXEC);
+	rflags |= ((new_pte & _PAGE_EXEC) ? 0 : HPTE_R_N);
 
 	/* Check if pte already has an hpte (case 2) */
-	if (unlikely(pte_val(old_pte) & _PAGE_HASHPTE)) {
+	if (unlikely(old_pte & _PAGE_HASHPTE)) {
 		/* There MIGHT be an HPTE for this pte */
 		unsigned long hash, slot;
 
-		hash = hpt_hash(vpn, 1);
-		if (pte_val(old_pte) & _PAGE_SECONDARY)
+		hash = hpt_hash(va, HPAGE_SHIFT);
+		if (old_pte & _PAGE_F_SECOND)
 			hash = ~hash;
 		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
-		slot += (pte_val(old_pte) & _PAGE_GROUP_IX) >> 12;
+		slot += (old_pte & _PAGE_F_GIX) >> 12;
 
 		if (ppc_md.hpte_updatepp(slot, rflags, va, 1, local) == -1)
-			pte_val(old_pte) &= ~_PAGE_HPTEFLAGS;
+			old_pte &= ~_PAGE_HPTEFLAGS;
 	}
 
-	if (likely(!(pte_val(old_pte) & _PAGE_HASHPTE))) {
-		unsigned long hash = hpt_hash(vpn, 1);
+	if (likely(!(old_pte & _PAGE_HASHPTE))) {
+		unsigned long hash = hpt_hash(va, HPAGE_SHIFT);
 		unsigned long hpte_group;
 
-		prpn = pte_pfn(old_pte);
+		pa = pte_pfn(__pte(old_pte)) << PAGE_SHIFT;
 
 repeat:
 		hpte_group = ((hash & htab_hash_mask) *
 			      HPTES_PER_GROUP) & ~0x7UL;
 
-		/* Update the linux pte with the HPTE slot */
-		pte_val(new_pte) &= ~_PAGE_HPTEFLAGS;
-		pte_val(new_pte) |= _PAGE_HASHPTE;
+		/* clear HPTE slot informations in new PTE */
+		new_pte = (new_pte & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE;
 
 		/* Add in WIMG bits */
 		/* XXX We should store these in the pte */
+		/* --BenH: I think they are ... */
 		rflags |= _PAGE_COHERENT;
 
-		slot = ppc_md.hpte_insert(hpte_group, va, prpn,
-					  HPTE_V_LARGE, rflags);
+		/* Insert into the hash table, primary slot */
+		slot = ppc_md.hpte_insert(hpte_group, va, pa, rflags, 0,
+					  mmu_huge_psize);
 
 		/* Primary is full, try the secondary */
 		if (unlikely(slot == -1)) {
-			pte_val(new_pte) |= _PAGE_SECONDARY;
+			new_pte |= _PAGE_F_SECOND;
 			hpte_group = ((~hash & htab_hash_mask) *
 				      HPTES_PER_GROUP) & ~0x7UL; 
-			slot = ppc_md.hpte_insert(hpte_group, va, prpn,
-						  HPTE_V_LARGE |
+			slot = ppc_md.hpte_insert(hpte_group, va, pa, rflags,
 						  HPTE_V_SECONDARY,
-						  rflags);
+						  mmu_huge_psize);
 			if (slot == -1) {
 				if (mftb() & 0x1)
 					hpte_group = ((hash & htab_hash_mask) *
@@ -726,20 +744,18 @@ int hash_huge_page(struct mm_struct *mm, unsigned long access,
 		if (unlikely(slot == -2))
 			panic("hash_huge_page: pte_insert failed\n");
 
-		pte_val(new_pte) |= (slot<<12) & _PAGE_GROUP_IX;
+		new_pte |= (slot << 12) & _PAGE_F_GIX;
+	}
 
 	/*
 	 * No need to use ldarx/stdcx here because all who
 	 * might be updating the pte will hold the
 	 * page_table_lock
 	 */
-		*ptep = new_pte;
-	}
+	*ptep = __pte(new_pte & ~_PAGE_BUSY);
 
 	err = 0;
 
  out:
-	spin_unlock(&mm->page_table_lock);
-
 	return err;
 }

arch/powerpc/mm/init_64.c

@@ -188,12 +188,21 @@ static void zero_ctor(void *addr, kmem_cache_t *cache, unsigned long flags)
 	memset(addr, 0, kmem_cache_size(cache));
 }
 
+#ifdef CONFIG_PPC_64K_PAGES
+static const int pgtable_cache_size[2] = {
+	PTE_TABLE_SIZE, PGD_TABLE_SIZE
+};
+static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = {
+	"pte_pmd_cache", "pgd_cache",
+};
+#else
 static const int pgtable_cache_size[2] = {
 	PTE_TABLE_SIZE, PMD_TABLE_SIZE
 };
 static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = {
 	"pgd_pte_cache", "pud_pmd_cache",
 };
+#endif /* CONFIG_PPC_64K_PAGES */
 
 kmem_cache_t *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)];
 
@@ -201,19 +210,14 @@ void pgtable_cache_init(void)
 {
 	int i;
 
-	BUILD_BUG_ON(PTE_TABLE_SIZE != pgtable_cache_size[PTE_CACHE_NUM]);
-	BUILD_BUG_ON(PMD_TABLE_SIZE != pgtable_cache_size[PMD_CACHE_NUM]);
-	BUILD_BUG_ON(PUD_TABLE_SIZE != pgtable_cache_size[PUD_CACHE_NUM]);
-	BUILD_BUG_ON(PGD_TABLE_SIZE != pgtable_cache_size[PGD_CACHE_NUM]);
-
 	for (i = 0; i < ARRAY_SIZE(pgtable_cache_size); i++) {
 		int size = pgtable_cache_size[i];
 		const char *name = pgtable_cache_name[i];
 
 		pgtable_cache[i] = kmem_cache_create(name,
 						     size, size,
-						     SLAB_HWCACHE_ALIGN
-						     | SLAB_MUST_HWCACHE_ALIGN,
+						     SLAB_HWCACHE_ALIGN |
+						     SLAB_MUST_HWCACHE_ALIGN,
 						     zero_ctor,
 						     NULL);
 		if (! pgtable_cache[i])

arch/powerpc/mm/mem.c

@@ -61,6 +61,9 @@ int init_bootmem_done;
 int mem_init_done;
 unsigned long memory_limit;
 
+extern void hash_preload(struct mm_struct *mm, unsigned long ea,
+			 unsigned long access, unsigned long trap);
+
 /*
  * This is called by /dev/mem to know if a given address has to
  * be mapped non-cacheable or not
@@ -493,18 +496,10 @@ EXPORT_SYMBOL(flush_icache_user_range);
 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
 		      pte_t pte)
 {
-	/* handle i-cache coherency */
-	unsigned long pfn = pte_pfn(pte);
-#ifdef CONFIG_PPC32
-	pmd_t *pmd;
-#else
-	unsigned long vsid;
-	void *pgdir;
-	pte_t *ptep;
-	int local = 0;
-	cpumask_t tmp;
-	unsigned long flags;
+#ifdef CONFIG_PPC_STD_MMU
+	unsigned long access = 0, trap;
 #endif
+	unsigned long pfn = pte_pfn(pte);
 
 	/* handle i-cache coherency */
 	if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE) &&
@@ -535,30 +530,21 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address,
 	/* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */
 	if (!pte_young(pte) || address >= TASK_SIZE)
 		return;
-#ifdef CONFIG_PPC32
-	if (Hash == 0)
-		return;
-	pmd = pmd_offset(pgd_offset(vma->vm_mm, address), address);
-	if (!pmd_none(*pmd))
-		add_hash_page(vma->vm_mm->context, address, pmd_val(*pmd));
-#else
-	pgdir = vma->vm_mm->pgd;
-	if (pgdir == NULL)
-		return;
 
-	ptep = find_linux_pte(pgdir, address);
-	if (!ptep)
+	/* We try to figure out if we are coming from an instruction
+	 * access fault and pass that down to __hash_page so we avoid
+	 * double-faulting on execution of fresh text. We have to test
+	 * for regs NULL since init will get here first thing at boot
+	 *
+	 * We also avoid filling the hash if not coming from a fault
+	 */
+	if (current->thread.regs == NULL)
 		return;
-
-	vsid = get_vsid(vma->vm_mm->context.id, address);
-
-	local_irq_save(flags);
-	tmp = cpumask_of_cpu(smp_processor_id());
-	if (cpus_equal(vma->vm_mm->cpu_vm_mask, tmp))
-		local = 1;
-
-	__hash_page(address, 0, vsid, ptep, 0x300, local);
-	local_irq_restore(flags);
-#endif
-#endif
+	trap = TRAP(current->thread.regs);
+	if (trap == 0x400)
+		access |= _PAGE_EXEC;
+	else if (trap != 0x300)
+		return;
+	hash_preload(vma->vm_mm, address, access, trap);
+#endif /* CONFIG_PPC_STD_MMU */
 }

arch/powerpc/mm/pgtable_64.c

@@ -101,7 +101,6 @@ static int map_io_page(unsigned long ea, unsigned long pa, int flags)
 	pud_t *pudp;
 	pmd_t *pmdp;
 	pte_t *ptep;
-	unsigned long vsid;
 
 	if (mem_init_done) {
 		pgdp = pgd_offset_k(ea);
@@ -117,28 +116,15 @@ static int map_io_page(unsigned long ea, unsigned long pa, int flags)
 		set_pte_at(&init_mm, ea, ptep, pfn_pte(pa >> PAGE_SHIFT,
 							  __pgprot(flags)));
 	} else {
-		unsigned long va, vpn, hash, hpteg;
-
 		/*
 		 * If the mm subsystem is not fully up, we cannot create a
 		 * linux page table entry for this mapping.  Simply bolt an
 		 * entry in the hardware page table.
+		 *
 		 */
-		vsid = get_kernel_vsid(ea);
-		va = (vsid << 28) | (ea & 0xFFFFFFF);
-		vpn = va >> PAGE_SHIFT;
-
-		hash = hpt_hash(vpn, 0);
-
-		hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP);
-
-		/* Panic if a pte grpup is full */
-		if (ppc_md.hpte_insert(hpteg, va, pa >> PAGE_SHIFT,
-				       HPTE_V_BOLTED,
-				       _PAGE_NO_CACHE|_PAGE_GUARDED|PP_RWXX)
-		    == -1) {
-			panic("map_io_page: could not insert mapping");
-		}
+		if (htab_bolt_mapping(ea, ea + PAGE_SIZE, pa, flags,
+				      mmu_virtual_psize))
+			panic("Can't map bolted IO mapping");
 	}
 	return 0;
 }

arch/powerpc/mm/ppc_mmu_32.c

@@ -178,6 +178,21 @@ void __init setbat(int index, unsigned long virt, unsigned long phys,
 	bat_addrs[index].phys = phys;
 }
 
+/*
+ * Preload a translation in the hash table
+ */
+void hash_preload(struct mm_struct *mm, unsigned long ea,
+		  unsigned long access, unsigned long trap)
+{
+	pmd_t *pmd;
+
+	if (Hash == 0)
+		return;
+	pmd = pmd_offset(pgd_offset(vma->vm_mm, address), address);
+	if (!pmd_none(*pmd))
+		add_hash_page(vma->vm_mm->context, address, pmd_val(*pmd));
+}
+
 /*
  * Initialize the hash table and patch the instructions in hashtable.S.
  */

arch/powerpc/mm/slb.c

@@ -14,14 +14,32 @@
  *      2 of the License, or (at your option) any later version.
  */
 
+#undef DEBUG
+
 #include <linux/config.h>
 #include <asm/pgtable.h>
 #include <asm/mmu.h>
 #include <asm/mmu_context.h>
 #include <asm/paca.h>
 #include <asm/cputable.h>
+#include <asm/cacheflush.h>
+
+#ifdef DEBUG
+#define DBG(fmt...) udbg_printf(fmt)
+#else
+#define DBG(fmt...)
+#endif
 
-extern void slb_allocate(unsigned long ea);
+extern void slb_allocate_realmode(unsigned long ea);
+extern void slb_allocate_user(unsigned long ea);
+
+static void slb_allocate(unsigned long ea)
+{
+	/* Currently, we do real mode for all SLBs including user, but
+	 * that will change if we bring back dynamic VSIDs
+	 */
+	slb_allocate_realmode(ea);
+}
 
 static inline unsigned long mk_esid_data(unsigned long ea, unsigned long slot)
 {
@@ -46,13 +64,15 @@ static void slb_flush_and_rebolt(void)
 {
 	/* If you change this make sure you change SLB_NUM_BOLTED
 	 * appropriately too. */
-	unsigned long ksp_flags = SLB_VSID_KERNEL;
+	unsigned long linear_llp, virtual_llp, lflags, vflags;
 	unsigned long ksp_esid_data;
 
 	WARN_ON(!irqs_disabled());
 
-	if (cpu_has_feature(CPU_FTR_16M_PAGE))
-		ksp_flags |= SLB_VSID_L;
+	linear_llp = mmu_psize_defs[mmu_linear_psize].sllp;
+	virtual_llp = mmu_psize_defs[mmu_virtual_psize].sllp;
+	lflags = SLB_VSID_KERNEL | linear_llp;
+	vflags = SLB_VSID_KERNEL | virtual_llp;
 
 	ksp_esid_data = mk_esid_data(get_paca()->kstack, 2);
 	if ((ksp_esid_data & ESID_MASK) == KERNELBASE)
@@ -67,9 +87,9 @@ static void slb_flush_and_rebolt(void)
 		     /* Slot 2 - kernel stack */
 		     "slbmte	%2,%3\n"
 		     "isync"
-		     :: "r"(mk_vsid_data(VMALLOCBASE, SLB_VSID_KERNEL)),
+		     :: "r"(mk_vsid_data(VMALLOCBASE, vflags)),
 		        "r"(mk_esid_data(VMALLOCBASE, 1)),
-		        "r"(mk_vsid_data(ksp_esid_data, ksp_flags)),
+		        "r"(mk_vsid_data(ksp_esid_data, lflags)),
 		        "r"(ksp_esid_data)
 		     : "memory");
 }
@@ -102,6 +122,9 @@ void switch_slb(struct task_struct *tsk, struct mm_struct *mm)
 
 	get_paca()->slb_cache_ptr = 0;
 	get_paca()->context = mm->context;
+#ifdef CONFIG_PPC_64K_PAGES
+	get_paca()->pgdir = mm->pgd;
+#endif /* CONFIG_PPC_64K_PAGES */
 
 	/*
 	 * preload some userspace segments into the SLB.
@@ -131,28 +154,77 @@ void switch_slb(struct task_struct *tsk, struct mm_struct *mm)
 	slb_allocate(unmapped_base);
 }
 
+static inline void patch_slb_encoding(unsigned int *insn_addr,
+				      unsigned int immed)
+{
+	/* Assume the instruction had a "0" immediate value, just
+	 * "or" in the new value
+	 */
+	*insn_addr |= immed;
+	flush_icache_range((unsigned long)insn_addr, 4+
+			   (unsigned long)insn_addr);
+}
+
 void slb_initialize(void)
 {
+	unsigned long linear_llp, virtual_llp;
+	static int slb_encoding_inited;
+	extern unsigned int *slb_miss_kernel_load_linear;
+	extern unsigned int *slb_miss_kernel_load_virtual;
+	extern unsigned int *slb_miss_user_load_normal;
+#ifdef CONFIG_HUGETLB_PAGE
+	extern unsigned int *slb_miss_user_load_huge;
+	unsigned long huge_llp;
+
+	huge_llp = mmu_psize_defs[mmu_huge_psize].sllp;
+#endif
+
+	/* Prepare our SLB miss handler based on our page size */
+	linear_llp = mmu_psize_defs[mmu_linear_psize].sllp;
+	virtual_llp = mmu_psize_defs[mmu_virtual_psize].sllp;
+	if (!slb_encoding_inited) {
+		slb_encoding_inited = 1;
+		patch_slb_encoding(slb_miss_kernel_load_linear,
+				   SLB_VSID_KERNEL | linear_llp);
+		patch_slb_encoding(slb_miss_kernel_load_virtual,
+				   SLB_VSID_KERNEL | virtual_llp);
+		patch_slb_encoding(slb_miss_user_load_normal,
+				   SLB_VSID_USER | virtual_llp);
+
+		DBG("SLB: linear  LLP = %04x\n", linear_llp);
+		DBG("SLB: virtual LLP = %04x\n", virtual_llp);
+#ifdef CONFIG_HUGETLB_PAGE
+		patch_slb_encoding(slb_miss_user_load_huge,
+				   SLB_VSID_USER | huge_llp);
+		DBG("SLB: huge    LLP = %04x\n", huge_llp);
+#endif
+	}
+
 	/* On iSeries the bolted entries have already been set up by
 	 * the hypervisor from the lparMap data in head.S */
 #ifndef CONFIG_PPC_ISERIES
-	unsigned long flags = SLB_VSID_KERNEL;
+ {
+	unsigned long lflags, vflags;
 
- 	/* Invalidate the entire SLB (even slot 0) & all the ERATS */
- 	if (cpu_has_feature(CPU_FTR_16M_PAGE))
- 		flags |= SLB_VSID_L;
+	lflags = SLB_VSID_KERNEL | linear_llp;
+	vflags = SLB_VSID_KERNEL | virtual_llp;
 
+	/* Invalidate the entire SLB (even slot 0) & all the ERATS */
 	asm volatile("isync":::"memory");
 	asm volatile("slbmte  %0,%0"::"r" (0) : "memory");
 	asm volatile("isync; slbia; isync":::"memory");
-	create_slbe(KERNELBASE, flags, 0);
-	create_slbe(VMALLOCBASE, SLB_VSID_KERNEL, 1);
+	create_slbe(KERNELBASE, lflags, 0);
+
+	/* VMALLOC space has 4K pages always for now */
+	create_slbe(VMALLOCBASE, vflags, 1);
+
 	/* We don't bolt the stack for the time being - we're in boot,
 	 * so the stack is in the bolted segment.  By the time it goes
 	 * elsewhere, we'll call _switch() which will bolt in the new
 	 * one. */
 	asm volatile("isync":::"memory");
-#endif
+ }
+#endif /* CONFIG_PPC_ISERIES */
 
 	get_paca()->stab_rr = SLB_NUM_BOLTED;
 }

arch/powerpc/mm/slb_low.S

@@ -18,61 +18,28 @@
 
 #include <linux/config.h>
 #include <asm/processor.h>
-#include <asm/page.h>
-#include <asm/mmu.h>
 #include <asm/ppc_asm.h>
 #include <asm/asm-offsets.h>
 #include <asm/cputable.h>
+#include <asm/page.h>
+#include <asm/mmu.h>
+#include <asm/pgtable.h>
 
-/* void slb_allocate(unsigned long ea);
+/* void slb_allocate_realmode(unsigned long ea);
  *
  * Create an SLB entry for the given EA (user or kernel).
  * 	r3 = faulting address, r13 = PACA
  *	r9, r10, r11 are clobbered by this function
  * No other registers are examined or changed.
  */
-_GLOBAL(slb_allocate)
-	/*
-	 * First find a slot, round robin. Previously we tried to find
-	 * a free slot first but that took too long. Unfortunately we
-	 * dont have any LRU information to help us choose a slot.
-	 */
-#ifdef CONFIG_PPC_ISERIES
-	/*
-	 * On iSeries, the "bolted" stack segment can be cast out on
-	 * shared processor switch so we need to check for a miss on
-	 * it and restore it to the right slot.
-	 */
-	ld	r9,PACAKSAVE(r13)
-	clrrdi	r9,r9,28
-	clrrdi	r11,r3,28
-	li	r10,SLB_NUM_BOLTED-1	/* Stack goes in last bolted slot */
-	cmpld	r9,r11
-	beq	3f
-#endif /* CONFIG_PPC_ISERIES */
-
-	ld	r10,PACASTABRR(r13)
-	addi	r10,r10,1
-	/* use a cpu feature mask if we ever change our slb size */
-	cmpldi	r10,SLB_NUM_ENTRIES
-
-	blt+	4f
-	li	r10,SLB_NUM_BOLTED
-
-4:
-	std	r10,PACASTABRR(r13)
-3:
-	/* r3 = faulting address, r10 = entry */
+_GLOBAL(slb_allocate_realmode)
+	/* r3 = faulting address */
 
 	srdi	r9,r3,60		/* get region */
-	srdi	r3,r3,28		/* get esid */
+	srdi	r10,r3,28		/* get esid */
 	cmpldi	cr7,r9,0xc		/* cmp KERNELBASE for later use */
 
-	rldimi	r10,r3,28,0		/* r10= ESID<<28 | entry */
-	oris	r10,r10,SLB_ESID_V@h	/* r10 |= SLB_ESID_V */
-
-	/* r3 = esid, r10 = esid_data, cr7 = <>KERNELBASE */
-
+	/* r3 = address, r10 = esid, cr7 = <>KERNELBASE */
 	blt	cr7,0f			/* user or kernel? */
 
 	/* kernel address: proto-VSID = ESID */
@@ -81,43 +48,161 @@ _GLOBAL(slb_allocate)
 	 * top segment.  That's ok, the scramble below will translate
 	 * it to VSID 0, which is reserved as a bad VSID - one which
 	 * will never have any pages in it.  */
-	li	r11,SLB_VSID_KERNEL
-BEGIN_FTR_SECTION
-	bne	cr7,9f
-	li	r11,(SLB_VSID_KERNEL|SLB_VSID_L)
-END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE)
-	b	9f
 
-0:	/* user address: proto-VSID = context<<15 | ESID */
-	srdi.	r9,r3,USER_ESID_BITS
+	/* Check if hitting the linear mapping of the vmalloc/ioremap
+	 * kernel space
+	*/
+	bne	cr7,1f
+
+	/* Linear mapping encoding bits, the "li" instruction below will
+	 * be patched by the kernel at boot
+	 */
+_GLOBAL(slb_miss_kernel_load_linear)
+	li	r11,0
+	b	slb_finish_load
+
+1:	/* vmalloc/ioremap mapping encoding bits, the "li" instruction below
+	 * will be patched by the kernel at boot
+	 */
+_GLOBAL(slb_miss_kernel_load_virtual)
+	li	r11,0
+	b	slb_finish_load
+
+
+0:	/* user address: proto-VSID = context << 15 | ESID. First check
+	 * if the address is within the boundaries of the user region
+	 */
+	srdi.	r9,r10,USER_ESID_BITS
 	bne-	8f			/* invalid ea bits set */
 
+	/* Figure out if the segment contains huge pages */
 #ifdef CONFIG_HUGETLB_PAGE
 BEGIN_FTR_SECTION
+	b	1f
+END_FTR_SECTION_IFCLR(CPU_FTR_16M_PAGE)
 	lhz	r9,PACAHIGHHTLBAREAS(r13)
-	srdi	r11,r3,(HTLB_AREA_SHIFT-SID_SHIFT)
+	srdi	r11,r10,(HTLB_AREA_SHIFT-SID_SHIFT)
 	srd	r9,r9,r11
 	lhz	r11,PACALOWHTLBAREAS(r13)
-	srd	r11,r11,r3
-	or	r9,r9,r11
-END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE)
+	srd	r11,r11,r10
+	or.	r9,r9,r11
+	beq	1f
+_GLOBAL(slb_miss_user_load_huge)
+	li	r11,0
+	b	2f
+1:
 #endif /* CONFIG_HUGETLB_PAGE */
 
-	li	r11,SLB_VSID_USER
+_GLOBAL(slb_miss_user_load_normal)
+	li	r11,0
 
-#ifdef CONFIG_HUGETLB_PAGE
-BEGIN_FTR_SECTION
-	rldimi	r11,r9,8,55		/* shift masked bit into SLB_VSID_L */
-END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE)
-#endif /* CONFIG_HUGETLB_PAGE */
+2:
+	ld	r9,PACACONTEXTID(r13)
+	rldimi	r10,r9,USER_ESID_BITS,0
+	b	slb_finish_load
+
+8:	/* invalid EA */
+	li	r10,0			/* BAD_VSID */
+	li	r11,SLB_VSID_USER	/* flags don't much matter */
+	b	slb_finish_load
+
+#ifdef __DISABLED__
 
+/* void slb_allocate_user(unsigned long ea);
+ *
+ * Create an SLB entry for the given EA (user or kernel).
+ * 	r3 = faulting address, r13 = PACA
+ *	r9, r10, r11 are clobbered by this function
+ * No other registers are examined or changed.
+ *
+ * It is called with translation enabled in order to be able to walk the
+ * page tables. This is not currently used.
+ */
+_GLOBAL(slb_allocate_user)
+	/* r3 = faulting address */
+	srdi	r10,r3,28		/* get esid */
+
+	crset	4*cr7+lt		/* set "user" flag for later */
+
+	/* check if we fit in the range covered by the pagetables*/
+	srdi.	r9,r3,PGTABLE_EADDR_SIZE
+	crnot	4*cr0+eq,4*cr0+eq
+	beqlr
+
+	/* now we need to get to the page tables in order to get the page
+	 * size encoding from the PMD. In the future, we'll be able to deal
+	 * with 1T segments too by getting the encoding from the PGD instead
+	 */
+	ld	r9,PACAPGDIR(r13)
+	cmpldi	cr0,r9,0
+	beqlr
+	rlwinm	r11,r10,8,25,28
+	ldx	r9,r9,r11		/* get pgd_t */
+	cmpldi	cr0,r9,0
+	beqlr
+	rlwinm	r11,r10,3,17,28
+	ldx	r9,r9,r11		/* get pmd_t */
+	cmpldi	cr0,r9,0
+	beqlr
+
+	/* build vsid flags */
+	andi.	r11,r9,SLB_VSID_LLP
+	ori	r11,r11,SLB_VSID_USER
+
+	/* get context to calculate proto-VSID */
 	ld	r9,PACACONTEXTID(r13)
-	rldimi	r3,r9,USER_ESID_BITS,0
+	rldimi	r10,r9,USER_ESID_BITS,0
 
-9:	/* r3 = protovsid, r11 = flags, r10 = esid_data, cr7 = <>KERNELBASE */
-	ASM_VSID_SCRAMBLE(r3,r9)
+	/* fall through slb_finish_load */
 
-	rldimi	r11,r3,SLB_VSID_SHIFT,16	/* combine VSID and flags */
+#endif /* __DISABLED__ */
+
+
+/*
+ * Finish loading of an SLB entry and return
+ *
+ * r3 = EA, r10 = proto-VSID, r11 = flags, clobbers r9, cr7 = <>KERNELBASE
+ */
+slb_finish_load:
+	ASM_VSID_SCRAMBLE(r10,r9)
+	rldimi	r11,r10,SLB_VSID_SHIFT,16	/* combine VSID and flags */
+
+	/* r3 = EA, r11 = VSID data */
+	/*
+	 * Find a slot, round robin. Previously we tried to find a
+	 * free slot first but that took too long. Unfortunately we
+ 	 * dont have any LRU information to help us choose a slot.
+ 	 */
+#ifdef CONFIG_PPC_ISERIES
+	/*
+	 * On iSeries, the "bolted" stack segment can be cast out on
+	 * shared processor switch so we need to check for a miss on
+	 * it and restore it to the right slot.
+	 */
+	ld	r9,PACAKSAVE(r13)
+	clrrdi	r9,r9,28
+	clrrdi	r3,r3,28
+	li	r10,SLB_NUM_BOLTED-1	/* Stack goes in last bolted slot */
+	cmpld	r9,r3
+	beq	3f
+#endif /* CONFIG_PPC_ISERIES */
+
+	ld	r10,PACASTABRR(r13)
+	addi	r10,r10,1
+	/* use a cpu feature mask if we ever change our slb size */
+	cmpldi	r10,SLB_NUM_ENTRIES
+
+	blt+	4f
+	li	r10,SLB_NUM_BOLTED
+
+4:
+	std	r10,PACASTABRR(r13)
+
+3:
+	rldimi	r3,r10,0,36		/* r3= EA[0:35] | entry */
+	oris	r10,r3,SLB_ESID_V@h	/* r3 |= SLB_ESID_V */
+
+	/* r3 = ESID data, r11 = VSID data */
 
 	/*
 	 * No need for an isync before or after this slbmte. The exception
@@ -125,7 +210,9 @@ END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE)
 	 */
 	slbmte	r11,r10
 
-	bgelr	cr7			/* we're done for kernel addresses */
+	/* we're done for kernel addresses */
+	crclr	4*cr0+eq		/* set result to "success" */
+	bgelr	cr7
 
 	/* Update the slb cache */
 	lhz	r3,PACASLBCACHEPTR(r13)	/* offset = paca->slb_cache_ptr */
@@ -143,9 +230,6 @@ END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE)
 	li	r3,SLB_CACHE_ENTRIES+1
 2:
 	sth	r3,PACASLBCACHEPTR(r13)	/* paca->slb_cache_ptr = offset */
+	crclr	4*cr0+eq		/* set result to "success" */
 	blr
 
-8:	/* invalid EA */
-	li	r3,0			/* BAD_VSID */
-	li	r11,SLB_VSID_USER	/* flags don't much matter */
-	b	9b

arch/powerpc/mm/stab.c

@@ -26,7 +26,6 @@ struct stab_entry {
 	unsigned long vsid_data;
 };
 
-/* Both the segment table and SLB code uses the following cache */
 #define NR_STAB_CACHE_ENTRIES 8
 DEFINE_PER_CPU(long, stab_cache_ptr);
 DEFINE_PER_CPU(long, stab_cache[NR_STAB_CACHE_ENTRIES]);
@@ -186,7 +185,7 @@ void switch_stab(struct task_struct *tsk, struct mm_struct *mm)
 		/* Never flush the first entry. */
 		ste += 1;
 		for (entry = 1;
-		     entry < (PAGE_SIZE / sizeof(struct stab_entry));
+		     entry < (HW_PAGE_SIZE / sizeof(struct stab_entry));
 		     entry++, ste++) {
 			unsigned long ea;
 			ea = ste->esid_data & ESID_MASK;
@@ -200,6 +199,10 @@ void switch_stab(struct task_struct *tsk, struct mm_struct *mm)
 
 	__get_cpu_var(stab_cache_ptr) = 0;
 
+#ifdef CONFIG_PPC_64K_PAGES
+	get_paca()->pgdir = mm->pgd;
+#endif /* CONFIG_PPC_64K_PAGES */
+
 	/* Now preload some entries for the new task */
 	if (test_tsk_thread_flag(tsk, TIF_32BIT))
 		unmapped_base = TASK_UNMAPPED_BASE_USER32;
@@ -223,8 +226,6 @@ void switch_stab(struct task_struct *tsk, struct mm_struct *mm)
 	asm volatile("sync" : : : "memory");
 }
 
-extern void slb_initialize(void);
-
 /*
  * Allocate segment tables for secondary CPUs.  These must all go in
  * the first (bolted) segment, so that do_stab_bolted won't get a
@@ -243,18 +244,21 @@ void stabs_alloc(void)
 		if (cpu == 0)
 			continue; /* stab for CPU 0 is statically allocated */
 
-		newstab = lmb_alloc_base(PAGE_SIZE, PAGE_SIZE, 1<<SID_SHIFT);
+		newstab = lmb_alloc_base(HW_PAGE_SIZE, HW_PAGE_SIZE,
+					 1<<SID_SHIFT);
 		if (! newstab)
 			panic("Unable to allocate segment table for CPU %d.\n",
 			      cpu);
 
 		newstab += KERNELBASE;
 
-		memset((void *)newstab, 0, PAGE_SIZE);
+		memset((void *)newstab, 0, HW_PAGE_SIZE);
 
 		paca[cpu].stab_addr = newstab;
 		paca[cpu].stab_real = virt_to_abs(newstab);
-		printk(KERN_DEBUG "Segment table for CPU %d at 0x%lx virtual, 0x%lx absolute\n", cpu, paca[cpu].stab_addr, paca[cpu].stab_real);
+		printk(KERN_DEBUG "Segment table for CPU %d at 0x%lx "
+		       "virtual, 0x%lx absolute\n",
+		       cpu, paca[cpu].stab_addr, paca[cpu].stab_real);
 	}
 }
 
@@ -267,13 +271,9 @@ void stab_initialize(unsigned long stab)
 {
 	unsigned long vsid = get_kernel_vsid(KERNELBASE);
 
-	if (cpu_has_feature(CPU_FTR_SLB)) {
-		slb_initialize();
-	} else {
 	asm volatile("isync; slbia; isync":::"memory");
 	make_ste(stab, GET_ESID(KERNELBASE), vsid);
 
 	/* Order update */
 	asm volatile("sync":::"memory");
-	}
 }

arch/powerpc/mm/tlb_64.c

@@ -21,6 +21,7 @@
  *  as published by the Free Software Foundation; either version
  *  2 of the License, or (at your option) any later version.
  */
+
 #include <linux/config.h>
 #include <linux/kernel.h>
 #include <linux/mm.h>
@@ -30,7 +31,7 @@
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 #include <asm/tlb.h>
-#include <linux/highmem.h>
+#include <asm/bug.h>
 
 DEFINE_PER_CPU(struct ppc64_tlb_batch, ppc64_tlb_batch);
 
@@ -126,28 +127,46 @@ void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf)
  * (if we remove it we should clear the _PTE_HPTEFLAGS bits).
  */
 void hpte_update(struct mm_struct *mm, unsigned long addr,
-		 unsigned long pte, int wrprot)
+		 pte_t *ptep, unsigned long pte, int huge)
 {
 	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
 	unsigned long vsid;
+	unsigned int psize = mmu_virtual_psize;
 	int i;
 
 	i = batch->index;
 
+	/* We mask the address for the base page size. Huge pages will
+	 * have applied their own masking already
+	 */
+	addr &= PAGE_MASK;
+
+	/* Get page size (maybe move back to caller) */
+	if (huge) {
+#ifdef CONFIG_HUGETLB_PAGE
+		psize = mmu_huge_psize;
+#else
+		BUG();
+#endif
+	}
+
 	/*
 	 * This can happen when we are in the middle of a TLB batch and
 	 * we encounter memory pressure (eg copy_page_range when it tries
 	 * to allocate a new pte). If we have to reclaim memory and end
 	 * up scanning and resetting referenced bits then our batch context
 	 * will change mid stream.
+	 *
+	 * We also need to ensure only one page size is present in a given
+	 * batch
 	 */
-	if (i != 0 && (mm != batch->mm || batch->large != pte_huge(pte))) {
+	if (i != 0 && (mm != batch->mm || batch->psize != psize)) {
 		flush_tlb_pending();
 		i = 0;
 	}
 	if (i == 0) {
 		batch->mm = mm;
-		batch->large = pte_huge(pte);
+		batch->psize = psize;
 	}
 	if (addr < KERNELBASE) {
 		vsid = get_vsid(mm->context.id, addr);
@@ -155,7 +174,7 @@ void hpte_update(struct mm_struct *mm, unsigned long addr,
 	} else
 		vsid = get_kernel_vsid(addr);
 	batch->vaddr[i] = (vsid << 28 ) | (addr & 0x0fffffff);
-	batch->pte[i] = __pte(pte);
+	batch->pte[i] = __real_pte(__pte(pte), ptep);
 	batch->index = ++i;
 	if (i >= PPC64_TLB_BATCH_NR)
 		flush_tlb_pending();
@@ -177,7 +196,8 @@ void __flush_tlb_pending(struct ppc64_tlb_batch *batch)
 		local = 1;
 
 	if (i == 1)
-		flush_hash_page(batch->vaddr[0], batch->pte[0], local);
+		flush_hash_page(batch->vaddr[0], batch->pte[0],
+				batch->psize, local);
 	else
 		flush_hash_range(i, local);
 	batch->index = 0;

arch/powerpc/platforms/iseries/htab.c

@@ -39,15 +39,16 @@ static inline void iSeries_hunlock(unsigned long slot)
 	spin_unlock(&iSeries_hlocks[(slot >> 4) & 0x3f]);
 }
 
-static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
-				unsigned long prpn, unsigned long vflags,
-				unsigned long rflags)
+long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
+			 unsigned long pa, unsigned long rflags,
+			 unsigned long vflags, int psize)
 {
-	unsigned long arpn;
 	long slot;
 	hpte_t lhpte;
 	int secondary = 0;
 
+	BUG_ON(psize != MMU_PAGE_4K);
+
 	/*
 	 * The hypervisor tries both primary and secondary.
 	 * If we are being called to insert in the secondary,
@@ -59,8 +60,19 @@ static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
 
 	iSeries_hlock(hpte_group);
 
-	slot = HvCallHpt_findValid(&lhpte, va >> PAGE_SHIFT);
-	BUG_ON(lhpte.v & HPTE_V_VALID);
+	slot = HvCallHpt_findValid(&lhpte, va >> HW_PAGE_SHIFT);
+	if (unlikely(lhpte.v & HPTE_V_VALID)) {
+		if (vflags & HPTE_V_BOLTED) {
+			HvCallHpt_setSwBits(slot, 0x10, 0);
+			HvCallHpt_setPp(slot, PP_RWXX);
+			iSeries_hunlock(hpte_group);
+			if (slot < 0)
+				return 0x8 | (slot & 7);
+			else
+				return slot & 7;
+		}
+		BUG();
+	}
 
 	if (slot == -1)	{ /* No available entry found in either group */
 		iSeries_hunlock(hpte_group);
@@ -73,10 +85,9 @@ static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
 		slot &= 0x7fffffffffffffff;
 	}
 
-	arpn = phys_to_abs(prpn << PAGE_SHIFT) >> PAGE_SHIFT;
 
-	lhpte.v = (va >> 23) << HPTE_V_AVPN_SHIFT | vflags | HPTE_V_VALID;
-	lhpte.r = (arpn << HPTE_R_RPN_SHIFT) | rflags;
+ 	lhpte.v = hpte_encode_v(va, MMU_PAGE_4K) | vflags | HPTE_V_VALID;
+	lhpte.r = hpte_encode_r(phys_to_abs(pa), MMU_PAGE_4K) | rflags;
 
 	/* Now fill in the actual HPTE */
 	HvCallHpt_addValidate(slot, secondary, &lhpte);
@@ -86,25 +97,6 @@ static long iSeries_hpte_insert(unsigned long hpte_group, unsigned long va,
 	return (secondary << 3) | (slot & 7);
 }
 
-long iSeries_hpte_bolt_or_insert(unsigned long hpte_group,
-		unsigned long va, unsigned long prpn, unsigned long vflags,
-		unsigned long rflags)
-{
-	long slot;
-	hpte_t lhpte;
-
-	slot = HvCallHpt_findValid(&lhpte, va >> PAGE_SHIFT);
-
-	if (lhpte.v & HPTE_V_VALID) {
-		/* Bolt the existing HPTE */
-		HvCallHpt_setSwBits(slot, 0x10, 0);
-		HvCallHpt_setPp(slot, PP_RWXX);
-		return 0;
-	}
-
-	return iSeries_hpte_insert(hpte_group, va, prpn, vflags, rflags);
-}
-
 static unsigned long iSeries_hpte_getword0(unsigned long slot)
 {
 	hpte_t hpte;
@@ -150,15 +142,17 @@ static long iSeries_hpte_remove(unsigned long hpte_group)
  *	bits 61..63 : PP2,PP1,PP0
  */
 static long iSeries_hpte_updatepp(unsigned long slot, unsigned long newpp,
-				  unsigned long va, int large, int local)
+				  unsigned long va, int psize, int local)
 {
 	hpte_t hpte;
-	unsigned long avpn = va >> 23;
+	unsigned long want_v;
 
 	iSeries_hlock(slot);
 
 	HvCallHpt_get(&hpte, slot);
-	if ((HPTE_V_AVPN_VAL(hpte.v) == avpn) && (hpte.v & HPTE_V_VALID)) {
+	want_v = hpte_encode_v(va, MMU_PAGE_4K);
+
+	if (HPTE_V_COMPARE(hpte.v, want_v) && (hpte.v & HPTE_V_VALID)) {
 		/*
 		 * Hypervisor expects bits as NPPP, which is
 		 * different from how they are mapped in our PP.
@@ -210,14 +204,17 @@ static long iSeries_hpte_find(unsigned long vpn)
  *
  * No need to lock here because we should be the only user.
  */
-static void iSeries_hpte_updateboltedpp(unsigned long newpp, unsigned long ea)
+static void iSeries_hpte_updateboltedpp(unsigned long newpp, unsigned long ea,
+					int psize)
 {
 	unsigned long vsid,va,vpn;
 	long slot;
 
+	BUG_ON(psize != MMU_PAGE_4K);
+
 	vsid = get_kernel_vsid(ea);
 	va = (vsid << 28) | (ea & 0x0fffffff);
-	vpn = va >> PAGE_SHIFT;
+	vpn = va >> HW_PAGE_SHIFT;
 	slot = iSeries_hpte_find(vpn);
 	if (slot == -1)
 		panic("updateboltedpp: Could not find page to bolt\n");
@@ -225,7 +222,7 @@ static void iSeries_hpte_updateboltedpp(unsigned long newpp, unsigned long ea)
 }
 
 static void iSeries_hpte_invalidate(unsigned long slot, unsigned long va,
-				    int large, int local)
+				    int psize, int local)
 {
 	unsigned long hpte_v;
 	unsigned long avpn = va >> 23;

arch/powerpc/platforms/iseries/hvlog.c

@@ -22,7 +22,7 @@ void HvCall_writeLogBuffer(const void *buffer, u64 len)
 
 	while (len) {
 		hv_buf.addr = cur;
-		left_this_page = ((cur & PAGE_MASK) + PAGE_SIZE) - cur;
+		left_this_page = ((cur & HW_PAGE_MASK) + HW_PAGE_SIZE) - cur;
 		if (left_this_page > len)
 			left_this_page = len;
 		hv_buf.len = left_this_page;
@@ -30,6 +30,6 @@ void HvCall_writeLogBuffer(const void *buffer, u64 len)
 		HvCall2(HvCallBaseWriteLogBuffer,
 				virt_to_abs(&hv_buf),
 				left_this_page);
-		cur = (cur & PAGE_MASK) + PAGE_SIZE;
+		cur = (cur & HW_PAGE_MASK) + HW_PAGE_SIZE;
 	}
 }

arch/powerpc/platforms/iseries/iommu.c

@@ -43,9 +43,12 @@ static void tce_build_iSeries(struct iommu_table *tbl, long index, long npages,
 	u64 rc;
 	union tce_entry tce;
 
+	index <<= TCE_PAGE_FACTOR;
+	npages <<= TCE_PAGE_FACTOR;
+
 	while (npages--) {
 		tce.te_word = 0;
-		tce.te_bits.tb_rpn = virt_to_abs(uaddr) >> PAGE_SHIFT;
+		tce.te_bits.tb_rpn = virt_to_abs(uaddr) >> TCE_SHIFT;
 
 		if (tbl->it_type == TCE_VB) {
 			/* Virtual Bus */
@@ -66,7 +69,7 @@ static void tce_build_iSeries(struct iommu_table *tbl, long index, long npages,
 			panic("PCI_DMA: HvCallXm_setTce failed, Rc: 0x%lx\n",
 					rc);
 		index++;
-		uaddr += PAGE_SIZE;
+		uaddr += TCE_PAGE_SIZE;
 	}
 }
 
@@ -74,6 +77,9 @@ static void tce_free_iSeries(struct iommu_table *tbl, long index, long npages)
 {
 	u64 rc;
 
+	npages <<= TCE_PAGE_FACTOR;
+	index <<= TCE_PAGE_FACTOR;
+
 	while (npages--) {
 		rc = HvCallXm_setTce((u64)tbl->it_index, (u64)index, 0);
 		if (rc)
@@ -83,27 +89,6 @@ static void tce_free_iSeries(struct iommu_table *tbl, long index, long npages)
 	}
 }
 
-#ifdef CONFIG_PCI
-/*
- * This function compares the known tables to find an iommu_table
- * that has already been built for hardware TCEs.
- */
-static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
-{
-	struct pci_dn *pdn;
-
-	list_for_each_entry(pdn, &iSeries_Global_Device_List, Device_List) {
-		struct iommu_table *it = pdn->iommu_table;
-		if ((it != NULL) &&
-		    (it->it_type == TCE_PCI) &&
-		    (it->it_offset == tbl->it_offset) &&
-		    (it->it_index == tbl->it_index) &&
-		    (it->it_size == tbl->it_size))
-			return it;
-	}
-	return NULL;
-}
-
 /*
  * Call Hv with the architected data structure to get TCE table info.
  * info. Put the returned data into the Linux representation of the
@@ -113,7 +98,9 @@ static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
  * 2. TCE table per Bus.
  * 3. TCE Table per IOA.
  */
-static void iommu_table_getparms(struct pci_dn *pdn,
+void iommu_table_getparms_iSeries(unsigned long busno,
+				  unsigned char slotno,
+				  unsigned char virtbus,
 				  struct iommu_table* tbl)
 {
 	struct iommu_table_cb *parms;
@@ -124,9 +111,9 @@ static void iommu_table_getparms(struct pci_dn *pdn,
 
 	memset(parms, 0, sizeof(*parms));
 
-	parms->itc_busno = pdn->busno;
-	parms->itc_slotno = pdn->LogicalSlot;
-	parms->itc_virtbus = 0;
+	parms->itc_busno = busno;
+	parms->itc_slotno = slotno;
+	parms->itc_virtbus = virtbus;
 
 	HvCallXm_getTceTableParms(iseries_hv_addr(parms));
 
@@ -134,17 +121,40 @@ static void iommu_table_getparms(struct pci_dn *pdn,
 		panic("PCI_DMA: parms->size is zero, parms is 0x%p", parms);
 
 	/* itc_size is in pages worth of table, it_size is in # of entries */
-	tbl->it_size = (parms->itc_size * PAGE_SIZE) / sizeof(union tce_entry);
+	tbl->it_size = ((parms->itc_size * TCE_PAGE_SIZE) /
+			sizeof(union tce_entry)) >> TCE_PAGE_FACTOR;
 	tbl->it_busno = parms->itc_busno;
-	tbl->it_offset = parms->itc_offset;
+	tbl->it_offset = parms->itc_offset >> TCE_PAGE_FACTOR;
 	tbl->it_index = parms->itc_index;
 	tbl->it_blocksize = 1;
-	tbl->it_type = TCE_PCI;
+	tbl->it_type = virtbus ? TCE_VB : TCE_PCI;
 
 	kfree(parms);
 }
 
 
+#ifdef CONFIG_PCI
+/*
+ * This function compares the known tables to find an iommu_table
+ * that has already been built for hardware TCEs.
+ */
+static struct iommu_table *iommu_table_find(struct iommu_table * tbl)
+{
+	struct pci_dn *pdn;
+
+	list_for_each_entry(pdn, &iSeries_Global_Device_List, Device_List) {
+		struct iommu_table *it = pdn->iommu_table;
+		if ((it != NULL) &&
+		    (it->it_type == TCE_PCI) &&
+		    (it->it_offset == tbl->it_offset) &&
+		    (it->it_index == tbl->it_index) &&
+		    (it->it_size == tbl->it_size))
+			return it;
+	}
+	return NULL;
+}
+
+
 void iommu_devnode_init_iSeries(struct device_node *dn)
 {
 	struct iommu_table *tbl;
@@ -152,7 +162,7 @@ void iommu_devnode_init_iSeries(struct device_node *dn)
 
 	tbl = kmalloc(sizeof(struct iommu_table), GFP_KERNEL);
 
-	iommu_table_getparms(pdn, tbl);
+	iommu_table_getparms_iSeries(pdn->busno, pdn->LogicalSlot, 0, tbl);
 
 	/* Look for existing tce table */
 	pdn->iommu_table = iommu_table_find(tbl);

arch/powerpc/platforms/iseries/setup.c

@@ -320,11 +320,11 @@ static void __init iSeries_init_early(void)
 	 */
 	if (naca.xRamDisk) {
 		initrd_start = (unsigned long)__va(naca.xRamDisk);
-		initrd_end = initrd_start + naca.xRamDiskSize * PAGE_SIZE;
+		initrd_end = initrd_start + naca.xRamDiskSize * HW_PAGE_SIZE;
 		initrd_below_start_ok = 1;	// ramdisk in kernel space
 		ROOT_DEV = Root_RAM0;
-		if (((rd_size * 1024) / PAGE_SIZE) < naca.xRamDiskSize)
-			rd_size = (naca.xRamDiskSize * PAGE_SIZE) / 1024;
+		if (((rd_size * 1024) / HW_PAGE_SIZE) < naca.xRamDiskSize)
+			rd_size = (naca.xRamDiskSize * HW_PAGE_SIZE) / 1024;
 	} else
 #endif /* CONFIG_BLK_DEV_INITRD */
 	{
@@ -470,13 +470,14 @@ static void __init build_iSeries_Memory_Map(void)
 	 */
 	hptFirstChunk = (u32)addr_to_chunk(HvCallHpt_getHptAddress());
 	hptSizePages = (u32)HvCallHpt_getHptPages();
-	hptSizeChunks = hptSizePages >> (MSCHUNKS_CHUNK_SHIFT - PAGE_SHIFT);
+	hptSizeChunks = hptSizePages >>
+		(MSCHUNKS_CHUNK_SHIFT - HW_PAGE_SHIFT);
 	hptLastChunk = hptFirstChunk + hptSizeChunks - 1;
 
 	printk("HPT absolute addr = %016lx, size = %dK\n",
 			chunk_to_addr(hptFirstChunk), hptSizeChunks * 256);
 
-	ppc64_pft_size = __ilog2(hptSizePages * PAGE_SIZE);
+	ppc64_pft_size = __ilog2(hptSizePages * HW_PAGE_SIZE);
 
 	/*
 	 * The actual hashed page table is in the hypervisor,
@@ -629,7 +630,7 @@ static void __init iSeries_fixup_klimit(void)
 	 */
 	if (naca.xRamDisk)
 		klimit = KERNELBASE + (u64)naca.xRamDisk +
-			(naca.xRamDiskSize * PAGE_SIZE);
+			(naca.xRamDiskSize * HW_PAGE_SIZE);
 	else {
 		/*
 		 * No ram disk was included - check and see if there

arch/powerpc/platforms/iseries/vio.c

@@ -30,41 +30,14 @@ static struct iommu_table vio_iommu_table;
 
 static void __init iommu_vio_init(void)
 {
-	struct iommu_table *t;
-	struct iommu_table_cb cb;
-	unsigned long cbp;
-	unsigned long itc_entries;
+	iommu_table_getparms_iSeries(255, 0, 0xff, &veth_iommu_table);
+	veth_iommu_table.it_size /= 2;
+	vio_iommu_table = veth_iommu_table;
+	vio_iommu_table.it_offset += veth_iommu_table.it_size;
 
-	cb.itc_busno = 255;    /* Bus 255 is the virtual bus */
-	cb.itc_virtbus = 0xff; /* Ask for virtual bus */
-
-	cbp = virt_to_abs(&cb);
-	HvCallXm_getTceTableParms(cbp);
-
-	itc_entries = cb.itc_size * PAGE_SIZE / sizeof(union tce_entry);
-	veth_iommu_table.it_size        = itc_entries / 2;
-	veth_iommu_table.it_busno       = cb.itc_busno;
-	veth_iommu_table.it_offset      = cb.itc_offset;
-	veth_iommu_table.it_index       = cb.itc_index;
-	veth_iommu_table.it_type        = TCE_VB;
-	veth_iommu_table.it_blocksize	= 1;
-
-	t = iommu_init_table(&veth_iommu_table);
-
-	if (!t)
+	if (!iommu_init_table(&veth_iommu_table))
 		printk("Virtual Bus VETH TCE table failed.\n");
-
-	vio_iommu_table.it_size         = itc_entries - veth_iommu_table.it_size;
-	vio_iommu_table.it_busno        = cb.itc_busno;
-	vio_iommu_table.it_offset       = cb.itc_offset +
-					  veth_iommu_table.it_size;
-	vio_iommu_table.it_index        = cb.itc_index;
-	vio_iommu_table.it_type         = TCE_VB;
-	vio_iommu_table.it_blocksize	= 1;
-
-	t = iommu_init_table(&vio_iommu_table);
-
-	if (!t)
+	if (!iommu_init_table(&vio_iommu_table))
 		printk("Virtual Bus VIO TCE table failed.\n");
 }
 

arch/powerpc/platforms/iseries/viopath.c

@@ -68,7 +68,8 @@ static DEFINE_SPINLOCK(statuslock);
  * For each kind of event we allocate a buffer that is
  * guaranteed not to cross a page boundary
  */
-static unsigned char event_buffer[VIO_MAX_SUBTYPES * 256] __page_aligned;
+static unsigned char event_buffer[VIO_MAX_SUBTYPES * 256]
+	__attribute__((__aligned__(4096)));
 static atomic_t event_buffer_available[VIO_MAX_SUBTYPES];
 static int event_buffer_initialised;
 
@@ -116,12 +117,12 @@ static int proc_viopath_show(struct seq_file *m, void *v)
 	HvLpEvent_Rc hvrc;
 	DECLARE_MUTEX_LOCKED(Semaphore);
 
-	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	buf = kmalloc(HW_PAGE_SIZE, GFP_KERNEL);
 	if (!buf)
 		return 0;
-	memset(buf, 0, PAGE_SIZE);
+	memset(buf, 0, HW_PAGE_SIZE);
 
-	handle = dma_map_single(iSeries_vio_dev, buf, PAGE_SIZE,
+	handle = dma_map_single(iSeries_vio_dev, buf, HW_PAGE_SIZE,
 				DMA_FROM_DEVICE);
 
 	hvrc = HvCallEvent_signalLpEventFast(viopath_hostLp,
@@ -131,7 +132,7 @@ static int proc_viopath_show(struct seq_file *m, void *v)
 			viopath_sourceinst(viopath_hostLp),
 			viopath_targetinst(viopath_hostLp),
 			(u64)(unsigned long)&Semaphore, VIOVERSION << 16,
-			((u64)handle) << 32, PAGE_SIZE, 0, 0);
+			((u64)handle) << 32, HW_PAGE_SIZE, 0, 0);
 
 	if (hvrc != HvLpEvent_Rc_Good)
 		printk(VIOPATH_KERN_WARN "hv error on op %d\n", (int)hvrc);
@@ -140,7 +141,7 @@ static int proc_viopath_show(struct seq_file *m, void *v)
 
 	vlanMap = HvLpConfig_getVirtualLanIndexMap();
 
-	buf[PAGE_SIZE-1] = '\0';
+	buf[HW_PAGE_SIZE-1] = '\0';
 	seq_printf(m, "%s", buf);
 	seq_printf(m, "AVAILABLE_VETH=%x\n", vlanMap);
 	seq_printf(m, "SRLNBR=%c%c%c%c%c%c%c\n",
@@ -152,7 +153,8 @@ static int proc_viopath_show(struct seq_file *m, void *v)
 		   e2a(xItExtVpdPanel.systemSerial[4]),
 		   e2a(xItExtVpdPanel.systemSerial[5]));
 
-	dma_unmap_single(iSeries_vio_dev, handle, PAGE_SIZE, DMA_FROM_DEVICE);
+	dma_unmap_single(iSeries_vio_dev, handle, HW_PAGE_SIZE,
+			 DMA_FROM_DEVICE);
 	kfree(buf);
 
 	return 0;

arch/powerpc/platforms/pseries/lpar.c

@@ -19,7 +19,7 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  */
 
-#define DEBUG
+#undef DEBUG_LOW
 
 #include <linux/config.h>
 #include <linux/kernel.h>
@@ -41,10 +41,10 @@
 
 #include "plpar_wrappers.h"
 
-#ifdef DEBUG
-#define DBG(fmt...) udbg_printf(fmt)
+#ifdef DEBUG_LOW
+#define DBG_LOW(fmt...) do { udbg_printf(fmt); } while(0)
 #else
-#define DBG(fmt...)
+#define DBG_LOW(fmt...) do { } while(0)
 #endif
 
 /* in pSeries_hvCall.S */
@@ -276,8 +276,9 @@ void vpa_init(int cpu)
 }
 
 long pSeries_lpar_hpte_insert(unsigned long hpte_group,
-			      unsigned long va, unsigned long prpn,
-			      unsigned long vflags, unsigned long rflags)
+ 			      unsigned long va, unsigned long pa,
+ 			      unsigned long rflags, unsigned long vflags,
+ 			      int psize)
 {
 	unsigned long lpar_rc;
 	unsigned long flags;
@@ -285,11 +286,28 @@ long pSeries_lpar_hpte_insert(unsigned long hpte_group,
 	unsigned long hpte_v, hpte_r;
 	unsigned long dummy0, dummy1;
 
-	hpte_v = ((va >> 23) << HPTE_V_AVPN_SHIFT) | vflags | HPTE_V_VALID;
-	if (vflags & HPTE_V_LARGE)
-		hpte_v &= ~(1UL << HPTE_V_AVPN_SHIFT);
+	if (!(vflags & HPTE_V_BOLTED))
+		DBG_LOW("hpte_insert(group=%lx, va=%016lx, pa=%016lx, "
+			"rflags=%lx, vflags=%lx, psize=%d)\n",
+		hpte_group, va, pa, rflags, vflags, psize);
 
-	hpte_r = (prpn << HPTE_R_RPN_SHIFT) | rflags;
+ 	hpte_v = hpte_encode_v(va, psize) | vflags | HPTE_V_VALID;
+	hpte_r = hpte_encode_r(pa, psize) | rflags;
+
+	if (!(vflags & HPTE_V_BOLTED))
+		DBG_LOW(" hpte_v=%016lx, hpte_r=%016lx\n", hpte_v, hpte_r);
+
+#if 1
+	{
+		int i;
+		for (i=0;i<8;i++) {
+			unsigned long w0, w1;
+			plpar_pte_read(0, hpte_group, &w0, &w1);
+			BUG_ON (HPTE_V_COMPARE(hpte_v, w0)
+				&& (w0 & HPTE_V_VALID));
+		}
+	}
+#endif
 
 	/* Now fill in the actual HPTE */
 	/* Set CEC cookie to 0         */
@@ -299,23 +317,30 @@ long pSeries_lpar_hpte_insert(unsigned long hpte_group,
 	/* Exact = 0                   */
 	flags = 0;
 
-	/* XXX why is this here? - Anton */
+	/* Make pHyp happy */
 	if (rflags & (_PAGE_GUARDED|_PAGE_NO_CACHE))
 		hpte_r &= ~_PAGE_COHERENT;
 
 	lpar_rc = plpar_hcall(H_ENTER, flags, hpte_group, hpte_v,
 			      hpte_r, &slot, &dummy0, &dummy1);
-
-	if (unlikely(lpar_rc == H_PTEG_Full))
+	if (unlikely(lpar_rc == H_PTEG_Full)) {
+		if (!(vflags & HPTE_V_BOLTED))
+			DBG_LOW(" full\n");
 		return -1;
+	}
 
 	/*
 	 * Since we try and ioremap PHBs we don't own, the pte insert
 	 * will fail. However we must catch the failure in hash_page
 	 * or we will loop forever, so return -2 in this case.
 	 */
-	if (unlikely(lpar_rc != H_Success))
+	if (unlikely(lpar_rc != H_Success)) {
+		if (!(vflags & HPTE_V_BOLTED))
+			DBG_LOW(" lpar err %d\n", lpar_rc);
 		return -2;
+	}
+	if (!(vflags & HPTE_V_BOLTED))
+		DBG_LOW(" -> slot: %d\n", slot & 7);
 
 	/* Because of iSeries, we have to pass down the secondary
 	 * bucket bit here as well
@@ -340,10 +365,8 @@ static long pSeries_lpar_hpte_remove(unsigned long hpte_group)
 		/* don't remove a bolted entry */
 		lpar_rc = plpar_pte_remove(H_ANDCOND, hpte_group + slot_offset,
 					   (0x1UL << 4), &dummy1, &dummy2);
-
 		if (lpar_rc == H_Success)
 			return i;
-
 		BUG_ON(lpar_rc != H_Not_Found);
 
 		slot_offset++;
@@ -371,20 +394,28 @@ static void pSeries_lpar_hptab_clear(void)
  * We can probably optimize here and assume the high bits of newpp are
  * already zero.  For now I am paranoid.
  */
-static long pSeries_lpar_hpte_updatepp(unsigned long slot, unsigned long newpp,
-				       unsigned long va, int large, int local)
+static long pSeries_lpar_hpte_updatepp(unsigned long slot,
+				       unsigned long newpp,
+				       unsigned long va,
+				       int psize, int local)
 {
 	unsigned long lpar_rc;
 	unsigned long flags = (newpp & 7) | H_AVPN;
-	unsigned long avpn = va >> 23;
+	unsigned long want_v;
 
-	if (large)
-		avpn &= ~0x1UL;
+	want_v = hpte_encode_v(va, psize);
 
-	lpar_rc = plpar_pte_protect(flags, slot, (avpn << 7));
+	DBG_LOW("    update: avpnv=%016lx, hash=%016lx, f=%x, psize: %d ... ",
+		want_v & HPTE_V_AVPN, slot, flags, psize);
 
-	if (lpar_rc == H_Not_Found)
+	lpar_rc = plpar_pte_protect(flags, slot, want_v & HPTE_V_AVPN);
+
+	if (lpar_rc == H_Not_Found) {
+		DBG_LOW("not found !\n");
 		return -1;
+	}
+
+	DBG_LOW("ok\n");
 
 	BUG_ON(lpar_rc != H_Success);
 
@@ -410,21 +441,22 @@ static unsigned long pSeries_lpar_hpte_getword0(unsigned long slot)
 	return dword0;
 }
 
-static long pSeries_lpar_hpte_find(unsigned long vpn)
+static long pSeries_lpar_hpte_find(unsigned long va, int psize)
 {
 	unsigned long hash;
 	unsigned long i, j;
 	long slot;
-	unsigned long hpte_v;
+	unsigned long want_v, hpte_v;
 
-	hash = hpt_hash(vpn, 0);
+	hash = hpt_hash(va, mmu_psize_defs[psize].shift);
+	want_v = hpte_encode_v(va, psize);
 
 	for (j = 0; j < 2; j++) {
 		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
 		for (i = 0; i < HPTES_PER_GROUP; i++) {
 			hpte_v = pSeries_lpar_hpte_getword0(slot);
 
-			if ((HPTE_V_AVPN_VAL(hpte_v) == (vpn >> 11))
+			if (HPTE_V_COMPARE(hpte_v, want_v)
 			    && (hpte_v & HPTE_V_VALID)
 			    && (!!(hpte_v & HPTE_V_SECONDARY) == j)) {
 				/* HPTE matches */
@@ -441,17 +473,15 @@ static long pSeries_lpar_hpte_find(unsigned long vpn)
 } 
 
 static void pSeries_lpar_hpte_updateboltedpp(unsigned long newpp,
-					     unsigned long ea)
+					     unsigned long ea,
+					     int psize)
 {
-	unsigned long lpar_rc;
-	unsigned long vsid, va, vpn, flags;
-	long slot;
+	unsigned long lpar_rc, slot, vsid, va, flags;
 
 	vsid = get_kernel_vsid(ea);
 	va = (vsid << 28) | (ea & 0x0fffffff);
-	vpn = va >> PAGE_SHIFT;
 
-	slot = pSeries_lpar_hpte_find(vpn);
+	slot = pSeries_lpar_hpte_find(va, psize);
 	BUG_ON(slot == -1);
 
 	flags = newpp & 7;
@@ -461,18 +491,18 @@ static void pSeries_lpar_hpte_updateboltedpp(unsigned long newpp,
 }
 
 static void pSeries_lpar_hpte_invalidate(unsigned long slot, unsigned long va,
-					 int large, int local)
+					 int psize, int local)
 {
-	unsigned long avpn = va >> 23;
+	unsigned long want_v;
 	unsigned long lpar_rc;
 	unsigned long dummy1, dummy2;
 
-	if (large)
-		avpn &= ~0x1UL;
-
-	lpar_rc = plpar_pte_remove(H_AVPN, slot, (avpn << 7), &dummy1,
-				   &dummy2);
+	DBG_LOW("    inval : slot=%lx, va=%016lx, psize: %d, local: %d",
+		slot, va, psize, local);
 
+	want_v = hpte_encode_v(va, psize);
+	lpar_rc = plpar_pte_remove(H_AVPN, slot, want_v & HPTE_V_AVPN,
+				   &dummy1, &dummy2);
 	if (lpar_rc == H_Not_Found)
 		return;
 
@@ -494,7 +524,8 @@ void pSeries_lpar_flush_hash_range(unsigned long number, int local)
 		spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags);
 
 	for (i = 0; i < number; i++)
-		flush_hash_page(batch->vaddr[i], batch->pte[i], local);
+		flush_hash_page(batch->vaddr[i], batch->pte[i],
+				batch->psize, local);
 
 	if (lock_tlbie)
 		spin_unlock_irqrestore(&pSeries_lpar_tlbie_lock, flags);

arch/ppc64/Kconfig

@@ -47,6 +47,10 @@ config ARCH_MAY_HAVE_PC_FDC
 	bool
 	default y
 
+config PPC_STD_MMU
+	bool
+	default y
+
 # We optimistically allocate largepages from the VM, so make the limit
 # large enough (16MB). This badly named config option is actually
 # max order + 1
@@ -294,6 +298,15 @@ config NODES_SPAN_OTHER_NODES
 	def_bool y
 	depends on NEED_MULTIPLE_NODES
 
+config PPC_64K_PAGES
+	bool "64k page size"
+	help
+	  This option changes the kernel logical page size to 64k. On machines
+          without processor support for 64k pages, the kernel will simulate
+          them by loading each individual 4k page on demand transparently,
+          while on hardware with such support, it will be used to map
+          normal application pages.
+
 config SCHED_SMT
 	bool "SMT (Hyperthreading) scheduler support"
 	depends on SMP

arch/ppc64/kernel/asm-offsets.c

@@ -93,6 +93,9 @@ int main(void)
 	DEFINE(PACASLBCACHE, offsetof(struct paca_struct, slb_cache));
 	DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr));
 	DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
+#ifdef CONFIG_PPC_64K_PAGES
+	DEFINE(PACAPGDIR, offsetof(struct paca_struct, pgdir));
+#endif
 #ifdef CONFIG_HUGETLB_PAGE
 	DEFINE(PACALOWHTLBAREAS, offsetof(struct paca_struct, context.low_htlb_areas));
 	DEFINE(PACAHIGHHTLBAREAS, offsetof(struct paca_struct, context.high_htlb_areas));

arch/ppc64/kernel/head.S

@@ -195,11 +195,11 @@ exception_marker:
 #define EX_R12		24
 #define EX_R13		32
 #define EX_SRR0		40
-#define EX_R3		40	/* SLB miss saves R3, but not SRR0 */
 #define EX_DAR		48
-#define EX_LR		48	/* SLB miss saves LR, but not DAR */
 #define EX_DSISR	56
 #define EX_CCR		60
+#define EX_R3		64
+#define EX_LR		72
 
 #define EXCEPTION_PROLOG_PSERIES(area, label)				\
 	mfspr	r13,SPRN_SPRG3;		/* get paca address into r13 */	\
@@ -419,17 +419,22 @@ data_access_slb_pSeries:
 	mtspr	SPRN_SPRG1,r13
 	RUNLATCH_ON(r13)
 	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
+	std	r3,PACA_EXSLB+EX_R3(r13)
+	mfspr	r3,SPRN_DAR
 	std	r9,PACA_EXSLB+EX_R9(r13)	/* save r9 - r12 */
+	mfcr	r9
+#ifdef __DISABLED__
+	/* Keep that around for when we re-implement dynamic VSIDs */
+	cmpdi	r3,0
+	bge	slb_miss_user_pseries
+#endif /* __DISABLED__ */
 	std	r10,PACA_EXSLB+EX_R10(r13)
 	std	r11,PACA_EXSLB+EX_R11(r13)
 	std	r12,PACA_EXSLB+EX_R12(r13)
-	std	r3,PACA_EXSLB+EX_R3(r13)
-	mfspr	r9,SPRN_SPRG1
-	std	r9,PACA_EXSLB+EX_R13(r13)
-	mfcr	r9
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
 	mfspr	r12,SPRN_SRR1		/* and SRR1 */
-	mfspr	r3,SPRN_DAR
-	b	.do_slb_miss		/* Rel. branch works in real mode */
+	b	.slb_miss_realmode	/* Rel. branch works in real mode */
 
 	STD_EXCEPTION_PSERIES(0x400, instruction_access)
 
@@ -440,17 +445,22 @@ instruction_access_slb_pSeries:
 	mtspr	SPRN_SPRG1,r13
 	RUNLATCH_ON(r13)
 	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
+	std	r3,PACA_EXSLB+EX_R3(r13)
+	mfspr	r3,SPRN_SRR0		/* SRR0 is faulting address */
 	std	r9,PACA_EXSLB+EX_R9(r13)	/* save r9 - r12 */
+	mfcr	r9
+#ifdef __DISABLED__
+	/* Keep that around for when we re-implement dynamic VSIDs */
+	cmpdi	r3,0
+	bge	slb_miss_user_pseries
+#endif /* __DISABLED__ */
 	std	r10,PACA_EXSLB+EX_R10(r13)
 	std	r11,PACA_EXSLB+EX_R11(r13)
 	std	r12,PACA_EXSLB+EX_R12(r13)
-	std	r3,PACA_EXSLB+EX_R3(r13)
-	mfspr	r9,SPRN_SPRG1
-	std	r9,PACA_EXSLB+EX_R13(r13)
-	mfcr	r9
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
 	mfspr	r12,SPRN_SRR1		/* and SRR1 */
-	mfspr	r3,SPRN_SRR0			/* SRR0 is faulting address */
-	b	.do_slb_miss		/* Rel. branch works in real mode */
+	b	.slb_miss_realmode	/* Rel. branch works in real mode */
 
 	STD_EXCEPTION_PSERIES(0x500, hardware_interrupt)
 	STD_EXCEPTION_PSERIES(0x600, alignment)
@@ -508,6 +518,38 @@ _GLOBAL(do_stab_bolted_pSeries)
 	mfspr	r12,SPRN_SPRG2
 	EXCEPTION_PROLOG_PSERIES(PACA_EXSLB, .do_stab_bolted)
 
+/*
+ * We have some room here  we use that to put
+ * the peries slb miss user trampoline code so it's reasonably
+ * away from slb_miss_user_common to avoid problems with rfid
+ *
+ * This is used for when the SLB miss handler has to go virtual,
+ * which doesn't happen for now anymore but will once we re-implement
+ * dynamic VSIDs for shared page tables
+ */
+#ifdef __DISABLED__
+slb_miss_user_pseries:
+	std	r10,PACA_EXGEN+EX_R10(r13)
+	std	r11,PACA_EXGEN+EX_R11(r13)
+	std	r12,PACA_EXGEN+EX_R12(r13)
+	mfspr	r10,SPRG1
+	ld	r11,PACA_EXSLB+EX_R9(r13)
+	ld	r12,PACA_EXSLB+EX_R3(r13)
+	std	r10,PACA_EXGEN+EX_R13(r13)
+	std	r11,PACA_EXGEN+EX_R9(r13)
+	std	r12,PACA_EXGEN+EX_R3(r13)
+	clrrdi	r12,r13,32
+	mfmsr	r10
+	mfspr	r11,SRR0			/* save SRR0 */
+	ori	r12,r12,slb_miss_user_common@l	/* virt addr of handler */
+	ori	r10,r10,MSR_IR|MSR_DR|MSR_RI
+	mtspr	SRR0,r12
+	mfspr	r12,SRR1			/* and SRR1 */
+	mtspr	SRR1,r10
+	rfid
+	b	.				/* prevent spec. execution */
+#endif /* __DISABLED__ */
+
 /*
  * Vectors for the FWNMI option.  Share common code.
  */
@@ -559,22 +601,59 @@ END_FTR_SECTION_IFCLR(CPU_FTR_SLB)
 	.globl	data_access_slb_iSeries
 data_access_slb_iSeries:
 	mtspr	SPRN_SPRG1,r13		/* save r13 */
-	EXCEPTION_PROLOG_ISERIES_1(PACA_EXSLB)
+	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
 	std	r3,PACA_EXSLB+EX_R3(r13)
-	ld	r12,PACALPPACA+LPPACASRR1(r13)
 	mfspr	r3,SPRN_DAR
-	b	.do_slb_miss
+	std	r9,PACA_EXSLB+EX_R9(r13)
+	mfcr	r9
+#ifdef __DISABLED__
+	cmpdi	r3,0
+	bge	slb_miss_user_iseries
+#endif
+	std	r10,PACA_EXSLB+EX_R10(r13)
+	std	r11,PACA_EXSLB+EX_R11(r13)
+	std	r12,PACA_EXSLB+EX_R12(r13)
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
+	ld	r12,PACALPPACA+LPPACASRR1(r13);
+	b	.slb_miss_realmode
 
 	STD_EXCEPTION_ISERIES(0x400, instruction_access, PACA_EXGEN)
 
 	.globl	instruction_access_slb_iSeries
 instruction_access_slb_iSeries:
 	mtspr	SPRN_SPRG1,r13		/* save r13 */
-	EXCEPTION_PROLOG_ISERIES_1(PACA_EXSLB)
+	mfspr	r13,SPRN_SPRG3		/* get paca address into r13 */
 	std	r3,PACA_EXSLB+EX_R3(r13)
-	ld	r12,PACALPPACA+LPPACASRR1(r13)
-	ld	r3,PACALPPACA+LPPACASRR0(r13)
-	b	.do_slb_miss
+	ld	r3,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
+	std	r9,PACA_EXSLB+EX_R9(r13)
+	mfcr	r9
+#ifdef __DISABLED__
+	cmpdi	r3,0
+	bge	.slb_miss_user_iseries
+#endif
+	std	r10,PACA_EXSLB+EX_R10(r13)
+	std	r11,PACA_EXSLB+EX_R11(r13)
+	std	r12,PACA_EXSLB+EX_R12(r13)
+	mfspr	r10,SPRN_SPRG1
+	std	r10,PACA_EXSLB+EX_R13(r13)
+	ld	r12,PACALPPACA+LPPACASRR1(r13);
+	b	.slb_miss_realmode
+
+#ifdef __DISABLED__
+slb_miss_user_iseries:
+	std	r10,PACA_EXGEN+EX_R10(r13)
+	std	r11,PACA_EXGEN+EX_R11(r13)
+	std	r12,PACA_EXGEN+EX_R12(r13)
+	mfspr	r10,SPRG1
+	ld	r11,PACA_EXSLB+EX_R9(r13)
+	ld	r12,PACA_EXSLB+EX_R3(r13)
+	std	r10,PACA_EXGEN+EX_R13(r13)
+	std	r11,PACA_EXGEN+EX_R9(r13)
+	std	r12,PACA_EXGEN+EX_R3(r13)
+	EXCEPTION_PROLOG_ISERIES_2
+	b	slb_miss_user_common
+#endif
 
 	MASKABLE_EXCEPTION_ISERIES(0x500, hardware_interrupt)
 	STD_EXCEPTION_ISERIES(0x600, alignment, PACA_EXGEN)
@@ -809,6 +888,126 @@ instruction_access_common:
 	li	r5,0x400
 	b	.do_hash_page		/* Try to handle as hpte fault */
 
+/*
+ * Here is the common SLB miss user that is used when going to virtual
+ * mode for SLB misses, that is currently not used
+ */
+#ifdef __DISABLED__
+	.align	7
+	.globl	slb_miss_user_common
+slb_miss_user_common:
+	mflr	r10
+	std	r3,PACA_EXGEN+EX_DAR(r13)
+	stw	r9,PACA_EXGEN+EX_CCR(r13)
+	std	r10,PACA_EXGEN+EX_LR(r13)
+	std	r11,PACA_EXGEN+EX_SRR0(r13)
+	bl	.slb_allocate_user
+
+	ld	r10,PACA_EXGEN+EX_LR(r13)
+	ld	r3,PACA_EXGEN+EX_R3(r13)
+	lwz	r9,PACA_EXGEN+EX_CCR(r13)
+	ld	r11,PACA_EXGEN+EX_SRR0(r13)
+	mtlr	r10
+	beq-	slb_miss_fault
+
+	andi.	r10,r12,MSR_RI		/* check for unrecoverable exception */
+	beq-	unrecov_user_slb
+	mfmsr	r10
+
+.machine push
+.machine "power4"
+	mtcrf	0x80,r9
+.machine pop
+
+	clrrdi	r10,r10,2		/* clear RI before setting SRR0/1 */
+	mtmsrd	r10,1
+
+	mtspr	SRR0,r11
+	mtspr	SRR1,r12
+
+	ld	r9,PACA_EXGEN+EX_R9(r13)
+	ld	r10,PACA_EXGEN+EX_R10(r13)
+	ld	r11,PACA_EXGEN+EX_R11(r13)
+	ld	r12,PACA_EXGEN+EX_R12(r13)
+	ld	r13,PACA_EXGEN+EX_R13(r13)
+	rfid
+	b	.
+
+slb_miss_fault:
+	EXCEPTION_PROLOG_COMMON(0x380, PACA_EXGEN)
+	ld	r4,PACA_EXGEN+EX_DAR(r13)
+	li	r5,0
+	std	r4,_DAR(r1)
+	std	r5,_DSISR(r1)
+	b	.handle_page_fault
+
+unrecov_user_slb:
+	EXCEPTION_PROLOG_COMMON(0x4200, PACA_EXGEN)
+	DISABLE_INTS
+	bl	.save_nvgprs
+1:	addi	r3,r1,STACK_FRAME_OVERHEAD
+	bl	.unrecoverable_exception
+	b	1b
+
+#endif /* __DISABLED__ */
+
+
+/*
+ * r13 points to the PACA, r9 contains the saved CR,
+ * r12 contain the saved SRR1, SRR0 is still ready for return
+ * r3 has the faulting address
+ * r9 - r13 are saved in paca->exslb.
+ * r3 is saved in paca->slb_r3
+ * We assume we aren't going to take any exceptions during this procedure.
+ */
+_GLOBAL(slb_miss_realmode)
+	mflr	r10
+
+	stw	r9,PACA_EXSLB+EX_CCR(r13)	/* save CR in exc. frame */
+	std	r10,PACA_EXSLB+EX_LR(r13)	/* save LR */
+
+	bl	.slb_allocate_realmode
+
+	/* All done -- return from exception. */
+
+	ld	r10,PACA_EXSLB+EX_LR(r13)
+	ld	r3,PACA_EXSLB+EX_R3(r13)
+	lwz	r9,PACA_EXSLB+EX_CCR(r13)	/* get saved CR */
+#ifdef CONFIG_PPC_ISERIES
+	ld	r11,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
+#endif /* CONFIG_PPC_ISERIES */
+
+	mtlr	r10
+
+	andi.	r10,r12,MSR_RI	/* check for unrecoverable exception */
+	beq-	unrecov_slb
+
+.machine	push
+.machine	"power4"
+	mtcrf	0x80,r9
+	mtcrf	0x01,r9		/* slb_allocate uses cr0 and cr7 */
+.machine	pop
+
+#ifdef CONFIG_PPC_ISERIES
+	mtspr	SPRN_SRR0,r11
+	mtspr	SPRN_SRR1,r12
+#endif /* CONFIG_PPC_ISERIES */
+	ld	r9,PACA_EXSLB+EX_R9(r13)
+	ld	r10,PACA_EXSLB+EX_R10(r13)
+	ld	r11,PACA_EXSLB+EX_R11(r13)
+	ld	r12,PACA_EXSLB+EX_R12(r13)
+	ld	r13,PACA_EXSLB+EX_R13(r13)
+	rfid
+	b	.	/* prevent speculative execution */
+
+unrecov_slb:
+	EXCEPTION_PROLOG_COMMON(0x4100, PACA_EXSLB)
+	DISABLE_INTS
+	bl	.save_nvgprs
+1:	addi	r3,r1,STACK_FRAME_OVERHEAD
+	bl	.unrecoverable_exception
+	b	1b
+
 	.align	7
 	.globl hardware_interrupt_common
 	.globl hardware_interrupt_entry
@@ -1138,62 +1337,6 @@ _GLOBAL(do_stab_bolted)
 	rfid
 	b	.	/* prevent speculative execution */
 
-/*
- * r13 points to the PACA, r9 contains the saved CR,
- * r11 and r12 contain the saved SRR0 and SRR1.
- * r3 has the faulting address
- * r9 - r13 are saved in paca->exslb.
- * r3 is saved in paca->slb_r3
- * We assume we aren't going to take any exceptions during this procedure.
- */
-_GLOBAL(do_slb_miss)
-	mflr	r10
-
-	stw	r9,PACA_EXSLB+EX_CCR(r13)	/* save CR in exc. frame */
-	std	r10,PACA_EXSLB+EX_LR(r13)	/* save LR */
-
-	bl	.slb_allocate			/* handle it */
-
-	/* All done -- return from exception. */
-
-	ld	r10,PACA_EXSLB+EX_LR(r13)
-	ld	r3,PACA_EXSLB+EX_R3(r13)
-	lwz	r9,PACA_EXSLB+EX_CCR(r13)	/* get saved CR */
-#ifdef CONFIG_PPC_ISERIES
-	ld	r11,PACALPPACA+LPPACASRR0(r13)	/* get SRR0 value */
-#endif /* CONFIG_PPC_ISERIES */
-
-	mtlr	r10
-
-	andi.	r10,r12,MSR_RI	/* check for unrecoverable exception */
-	beq-	unrecov_slb
-
-.machine	push
-.machine	"power4"
-	mtcrf	0x80,r9
-	mtcrf	0x01,r9		/* slb_allocate uses cr0 and cr7 */
-.machine	pop
-
-#ifdef CONFIG_PPC_ISERIES
-	mtspr	SPRN_SRR0,r11
-	mtspr	SPRN_SRR1,r12
-#endif /* CONFIG_PPC_ISERIES */
-	ld	r9,PACA_EXSLB+EX_R9(r13)
-	ld	r10,PACA_EXSLB+EX_R10(r13)
-	ld	r11,PACA_EXSLB+EX_R11(r13)
-	ld	r12,PACA_EXSLB+EX_R12(r13)
-	ld	r13,PACA_EXSLB+EX_R13(r13)
-	rfid
-	b	.	/* prevent speculative execution */
-
-unrecov_slb:
-	EXCEPTION_PROLOG_COMMON(0x4100, PACA_EXSLB)
-	DISABLE_INTS
-	bl	.save_nvgprs
-1:	addi	r3,r1,STACK_FRAME_OVERHEAD
-	bl	.unrecoverable_exception
-	b	1b
-
 /*
  * Space for CPU0's segment table.
  *
@@ -1569,7 +1712,10 @@ _GLOBAL(__secondary_start)
 #endif
 	/* Initialize the first segment table (or SLB) entry		 */
 	ld	r3,PACASTABVIRT(r13)	/* get addr of segment table	 */
+BEGIN_FTR_SECTION
 	bl	.stab_initialize
+END_FTR_SECTION_IFCLR(CPU_FTR_SLB)
+	bl	.slb_initialize
 
 	/* Initialize the kernel stack.  Just a repeat for iSeries.	 */
 	LOADADDR(r3,current_set)

arch/ppc64/kernel/pacaData.c

@@ -23,7 +23,7 @@
 static union {
 	struct systemcfg	data;
 	u8			page[PAGE_SIZE];
-} systemcfg_store __page_aligned;
+} systemcfg_store __attribute__((__section__(".data.page.aligned")));
 struct systemcfg *systemcfg = &systemcfg_store.data;
 EXPORT_SYMBOL(systemcfg);
 

arch/ppc64/kernel/prom.c

@@ -635,7 +635,7 @@ static inline char *find_flat_dt_string(u32 offset)
  * used to extract the memory informations at boot before we can
  * unflatten the tree
  */
-static int __init scan_flat_dt(int (*it)(unsigned long node,
+int __init of_scan_flat_dt(int (*it)(unsigned long node,
 				     const char *uname, int depth,
 				     void *data),
 			   void *data)
@@ -695,7 +695,7 @@ static int __init scan_flat_dt(int (*it)(unsigned long node,
  * This  function can be used within scan_flattened_dt callback to get
  * access to properties
  */
-static void* __init get_flat_dt_prop(unsigned long node, const char *name,
+void* __init of_get_flat_dt_prop(unsigned long node, const char *name,
 				 unsigned long *size)
 {
 	unsigned long p = node;
@@ -996,7 +996,7 @@ void __init unflatten_device_tree(void)
 static int __init early_init_dt_scan_cpus(unsigned long node,
 					  const char *uname, int depth, void *data)
 {
-	char *type = get_flat_dt_prop(node, "device_type", NULL);
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 	u32 *prop;
 	unsigned long size;
 
@@ -1004,17 +1004,6 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 	if (type == NULL || strcmp(type, "cpu") != 0)
 		return 0;
 
-	/* On LPAR, look for the first ibm,pft-size property for the  hash table size
-	 */
-	if (systemcfg->platform == PLATFORM_PSERIES_LPAR && ppc64_pft_size == 0) {
-		u32 *pft_size;
-		pft_size = (u32 *)get_flat_dt_prop(node, "ibm,pft-size", NULL);
-		if (pft_size != NULL) {
-			/* pft_size[0] is the NUMA CEC cookie */
-			ppc64_pft_size = pft_size[1];
-		}
-	}
-
 	if (initial_boot_params && initial_boot_params->version >= 2) {
 		/* version 2 of the kexec param format adds the phys cpuid
 		 * of booted proc.
@@ -1023,8 +1012,9 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 		boot_cpuid = 0;
 	} else {
 		/* Check if it's the boot-cpu, set it's hw index in paca now */
-		if (get_flat_dt_prop(node, "linux,boot-cpu", NULL) != NULL) {
-			u32 *prop = get_flat_dt_prop(node, "reg", NULL);
+		if (of_get_flat_dt_prop(node, "linux,boot-cpu", NULL)
+		    != NULL) {
+			u32 *prop = of_get_flat_dt_prop(node, "reg", NULL);
 			set_hard_smp_processor_id(0, prop == NULL ? 0 : *prop);
 			boot_cpuid_phys = get_hard_smp_processor_id(0);
 		}
@@ -1032,14 +1022,14 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 
 #ifdef CONFIG_ALTIVEC
 	/* Check if we have a VMX and eventually update CPU features */
-	prop = (u32 *)get_flat_dt_prop(node, "ibm,vmx", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "ibm,vmx", NULL);
 	if (prop && (*prop) > 0) {
 		cur_cpu_spec->cpu_features |= CPU_FTR_ALTIVEC;
 		cur_cpu_spec->cpu_user_features |= PPC_FEATURE_HAS_ALTIVEC;
 	}
 
 	/* Same goes for Apple's "altivec" property */
-	prop = (u32 *)get_flat_dt_prop(node, "altivec", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "altivec", NULL);
 	if (prop) {
 		cur_cpu_spec->cpu_features |= CPU_FTR_ALTIVEC;
 		cur_cpu_spec->cpu_user_features |= PPC_FEATURE_HAS_ALTIVEC;
@@ -1051,7 +1041,7 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
 	 * this by looking at the size of the ibm,ppc-interrupt-server#s
 	 * property
 	 */
-	prop = (u32 *)get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s",
+	prop = (u32 *)of_get_flat_dt_prop(node, "ibm,ppc-interrupt-server#s",
 				       &size);
 	cur_cpu_spec->cpu_features &= ~CPU_FTR_SMT;
 	if (prop && ((size / sizeof(u32)) > 1))
@@ -1072,26 +1062,26 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
 		return 0;
 
 	/* get platform type */
-	prop = (u32 *)get_flat_dt_prop(node, "linux,platform", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "linux,platform", NULL);
 	if (prop == NULL)
 		return 0;
 	systemcfg->platform = *prop;
 
 	/* check if iommu is forced on or off */
-	if (get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
+	if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
 		iommu_is_off = 1;
-	if (get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
+	if (of_get_flat_dt_prop(node, "linux,iommu-force-on", NULL) != NULL)
 		iommu_force_on = 1;
 
- 	prop64 = (u64*)get_flat_dt_prop(node, "linux,memory-limit", NULL);
+ 	prop64 = (u64*)of_get_flat_dt_prop(node, "linux,memory-limit", NULL);
  	if (prop64)
  		memory_limit = *prop64;
 
- 	prop64 = (u64*)get_flat_dt_prop(node, "linux,tce-alloc-start", NULL);
+ 	prop64 = (u64*)of_get_flat_dt_prop(node, "linux,tce-alloc-start",NULL);
  	if (prop64)
  		tce_alloc_start = *prop64;
 
- 	prop64 = (u64*)get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
+ 	prop64 = (u64*)of_get_flat_dt_prop(node, "linux,tce-alloc-end", NULL);
  	if (prop64)
  		tce_alloc_end = *prop64;
 
@@ -1102,9 +1092,12 @@ static int __init early_init_dt_scan_chosen(unsigned long node,
 	{
 		u64 *basep, *entryp;
 
-		basep = (u64*)get_flat_dt_prop(node, "linux,rtas-base", NULL);
-		entryp = (u64*)get_flat_dt_prop(node, "linux,rtas-entry", NULL);
-		prop = (u32*)get_flat_dt_prop(node, "linux,rtas-size", NULL);
+		basep = (u64*)of_get_flat_dt_prop(node,
+						  "linux,rtas-base", NULL);
+		entryp = (u64*)of_get_flat_dt_prop(node,
+						   "linux,rtas-entry", NULL);
+		prop = (u32*)of_get_flat_dt_prop(node,
+						 "linux,rtas-size", NULL);
 		if (basep && entryp && prop) {
 			rtas.base = *basep;
 			rtas.entry = *entryp;
@@ -1125,11 +1118,11 @@ static int __init early_init_dt_scan_root(unsigned long node,
 	if (depth != 0)
 		return 0;
 
-	prop = (u32 *)get_flat_dt_prop(node, "#size-cells", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "#size-cells", NULL);
 	dt_root_size_cells = (prop == NULL) ? 1 : *prop;
 	DBG("dt_root_size_cells = %x\n", dt_root_size_cells);
 
-	prop = (u32 *)get_flat_dt_prop(node, "#address-cells", NULL);
+	prop = (u32 *)of_get_flat_dt_prop(node, "#address-cells", NULL);
 	dt_root_addr_cells = (prop == NULL) ? 2 : *prop;
 	DBG("dt_root_addr_cells = %x\n", dt_root_addr_cells);
 	
@@ -1161,7 +1154,7 @@ static unsigned long __init dt_mem_next_cell(int s, cell_t **cellp)
 static int __init early_init_dt_scan_memory(unsigned long node,
 					    const char *uname, int depth, void *data)
 {
-	char *type = get_flat_dt_prop(node, "device_type", NULL);
+	char *type = of_get_flat_dt_prop(node, "device_type", NULL);
 	cell_t *reg, *endp;
 	unsigned long l;
 
@@ -1169,7 +1162,7 @@ static int __init early_init_dt_scan_memory(unsigned long node,
 	if (type == NULL || strcmp(type, "memory") != 0)
 		return 0;
 
-	reg = (cell_t *)get_flat_dt_prop(node, "reg", &l);
+	reg = (cell_t *)of_get_flat_dt_prop(node, "reg", &l);
 	if (reg == NULL)
 		return 0;
 
@@ -1225,19 +1218,16 @@ void __init early_init_devtree(void *params)
 	/* Setup flat device-tree pointer */
 	initial_boot_params = params;
 
-	/* By default, hash size is not set */
-	ppc64_pft_size = 0;
-
 	/* Retreive various informations from the /chosen node of the
 	 * device-tree, including the platform type, initrd location and
 	 * size, TCE reserve, and more ...
 	 */
-	scan_flat_dt(early_init_dt_scan_chosen, NULL);
+	of_scan_flat_dt(early_init_dt_scan_chosen, NULL);
 
 	/* Scan memory nodes and rebuild LMBs */
 	lmb_init();
-	scan_flat_dt(early_init_dt_scan_root, NULL);
-	scan_flat_dt(early_init_dt_scan_memory, NULL);
+	of_scan_flat_dt(early_init_dt_scan_root, NULL);
+	of_scan_flat_dt(early_init_dt_scan_memory, NULL);
 	lmb_enforce_memory_limit(memory_limit);
 	lmb_analyze();
 	systemcfg->physicalMemorySize = lmb_phys_mem_size();
@@ -1253,26 +1243,8 @@ void __init early_init_devtree(void *params)
 	/* Retreive hash table size from flattened tree plus other
 	 * CPU related informations (altivec support, boot CPU ID, ...)
 	 */
-	scan_flat_dt(early_init_dt_scan_cpus, NULL);
-
-	/* If hash size wasn't obtained above, we calculate it now based on
-	 * the total RAM size
-	 */
-	if (ppc64_pft_size == 0) {
-		unsigned long rnd_mem_size, pteg_count;
-
-		/* round mem_size up to next power of 2 */
-		rnd_mem_size = 1UL << __ilog2(systemcfg->physicalMemorySize);
-		if (rnd_mem_size < systemcfg->physicalMemorySize)
-			rnd_mem_size <<= 1;
-
-		/* # pages / 2 */
-		pteg_count = max(rnd_mem_size >> (12 + 1), 1UL << 11);
-
-		ppc64_pft_size = __ilog2(pteg_count << 7);
-	}
+	of_scan_flat_dt(early_init_dt_scan_cpus, NULL);
 
-	DBG("Hash pftSize: %x\n", (int)ppc64_pft_size);
 	DBG(" <- early_init_devtree()\n");
 }
 

include/asm-powerpc/cputable.h

@@ -101,6 +101,7 @@ extern void do_cpu_ftr_fixups(unsigned long offset);
 #define CPU_FTR_COHERENT_ICACHE  	ASM_CONST(0x0000020000000000)
 #define CPU_FTR_LOCKLESS_TLBIE		ASM_CONST(0x0000040000000000)
 #define CPU_FTR_MMCRA_SIHV		ASM_CONST(0x0000080000000000)
+#define CPU_FTR_CI_LARGE_PAGE		ASM_CONST(0x0000100000000000)
 #else
 /* ensure on 32b processors the flags are available for compiling but
  * don't do anything */
@@ -116,6 +117,7 @@ extern void do_cpu_ftr_fixups(unsigned long offset);
 #define CPU_FTR_COHERENT_ICACHE  	ASM_CONST(0x0)
 #define CPU_FTR_LOCKLESS_TLBIE		ASM_CONST(0x0)
 #define CPU_FTR_MMCRA_SIHV		ASM_CONST(0x0)
+#define CPU_FTR_CI_LARGE_PAGE		ASM_CONST(0x0)
 #endif
 
 #ifndef __ASSEMBLY__
@@ -339,6 +341,7 @@ enum {
 #ifdef __powerpc64__
 	    CPU_FTRS_POWER3 | CPU_FTRS_RS64 | CPU_FTRS_POWER4 |
 	    CPU_FTRS_PPC970 | CPU_FTRS_POWER5 | CPU_FTRS_CELL |
+            CPU_FTR_CI_LARGE_PAGE |
 #endif
 	    0,
 

include/asm-powerpc/iommu.h

@@ -74,6 +74,11 @@ extern void iommu_devnode_init_pSeries(struct device_node *dn);
 
 /* Creates table for an individual device node */
 extern void iommu_devnode_init_iSeries(struct device_node *dn);
+/* Get table parameters from HV */
+extern void iommu_table_getparms_iSeries(unsigned long busno,
+					 unsigned char slotno,
+					 unsigned char virtbus,
+					 struct iommu_table* tbl);
 
 #endif /* CONFIG_PPC_ISERIES */
 

include/asm-powerpc/machdep.h

@@ -47,20 +47,22 @@ struct machdep_calls {
 #ifdef CONFIG_PPC64
 	void            (*hpte_invalidate)(unsigned long slot,
 					   unsigned long va,
-					   int large,
+					   int psize,
 					   int local);
 	long		(*hpte_updatepp)(unsigned long slot, 
 					 unsigned long newpp, 
 					 unsigned long va,
-					 int large,
+					 int pize,
 					 int local);
 	void            (*hpte_updateboltedpp)(unsigned long newpp, 
-					       unsigned long ea);
+					       unsigned long ea,
+					       int psize);
 	long		(*hpte_insert)(unsigned long hpte_group,
 				       unsigned long va,
 				       unsigned long prpn,
+				       unsigned long rflags,
 				       unsigned long vflags,
-				       unsigned long rflags);
+				       int psize);
 	long		(*hpte_remove)(unsigned long hpte_group);
 	void		(*flush_hash_range)(unsigned long number, int local);
 

include/asm-powerpc/prom.h

@@ -178,6 +178,14 @@ extern struct device_node *of_get_next_child(const struct device_node *node,
 extern struct device_node *of_node_get(struct device_node *node);
 extern void of_node_put(struct device_node *node);
 
+/* For scanning the flat device-tree at boot time */
+int __init of_scan_flat_dt(int (*it)(unsigned long node,
+				     const char *uname, int depth,
+				     void *data),
+			   void *data);
+void* __init of_get_flat_dt_prop(unsigned long node, const char *name,
+				 unsigned long *size);
+
 /* For updating the device tree at runtime */
 extern void of_attach_node(struct device_node *);
 extern void of_detach_node(const struct device_node *);

include/asm-powerpc/system.h

@@ -289,7 +289,7 @@ __cmpxchg_u32(volatile unsigned int *p, unsigned long old, unsigned long new)
 
 #ifdef CONFIG_PPC64
 static __inline__ unsigned long
-__cmpxchg_u64(volatile long *p, unsigned long old, unsigned long new)
+__cmpxchg_u64(volatile unsigned long *p, unsigned long old, unsigned long new)
 {
 	unsigned long prev;
 

include/asm-powerpc/thread_info.h

@@ -65,23 +65,27 @@ struct thread_info {
 
 /* thread information allocation */
 
-#ifdef CONFIG_DEBUG_STACK_USAGE
-#define THREAD_INFO_GFP		GFP_KERNEL | __GFP_ZERO
-#else
-#define THREAD_INFO_GFP		GFP_KERNEL
-#endif
-
 #if THREAD_SHIFT >= PAGE_SHIFT
 
 #define THREAD_ORDER	(THREAD_SHIFT - PAGE_SHIFT)
 
+#ifdef CONFIG_DEBUG_STACK_USAGE
 #define alloc_thread_info(tsk)	\
-	((struct thread_info *)__get_free_pages(THREAD_INFO_GFP, THREAD_ORDER))
+	((struct thread_info *)__get_free_pages(GFP_KERNEL | \
+		__GFP_ZERO, THREAD_ORDER))
+#else
+#define alloc_thread_info(tsk)	\
+	((struct thread_info *)__get_free_pages(GFP_KERNEL, THREAD_ORDER))
+#endif
 #define free_thread_info(ti)	free_pages((unsigned long)ti, THREAD_ORDER)
 
 #else /* THREAD_SHIFT < PAGE_SHIFT */
 
-#define alloc_thread_info(tsk)	kmalloc(THREAD_SIZE, THREAD_INFO_GFP)
+#ifdef CONFIG_DEBUG_STACK_USAGE
+#define alloc_thread_info(tsk)	kzalloc(THREAD_SIZE, GFP_KERNEL)
+#else
+#define alloc_thread_info(tsk)	kmalloc(THREAD_SIZE, GFP_KERNEL)
+#endif
 #define free_thread_info(ti)	kfree(ti)
 
 #endif /* THREAD_SHIFT < PAGE_SHIFT */

include/asm-powerpc/tlbflush.h

@@ -31,9 +31,9 @@ struct mm_struct;
 struct ppc64_tlb_batch {
 	unsigned long index;
 	struct mm_struct *mm;
-	pte_t pte[PPC64_TLB_BATCH_NR];
+	real_pte_t pte[PPC64_TLB_BATCH_NR];
 	unsigned long vaddr[PPC64_TLB_BATCH_NR];
-	unsigned int large;
+	unsigned int psize;
 };
 DECLARE_PER_CPU(struct ppc64_tlb_batch, ppc64_tlb_batch);
 
@@ -48,8 +48,9 @@ static inline void flush_tlb_pending(void)
 	put_cpu_var(ppc64_tlb_batch);
 }
 
-extern void flush_hash_page(unsigned long va, pte_t pte, int local);
-void flush_hash_range(unsigned long number, int local);
+extern void flush_hash_page(unsigned long va, real_pte_t pte, int psize,
+			    int local);
+extern void flush_hash_range(unsigned long number, int local);
 
 #else /* CONFIG_PPC64 */
 

include/asm-ppc64/mmu.h

@@ -48,12 +48,20 @@ extern char initial_stab[];
 
 /* Bits in the SLB VSID word */
 #define SLB_VSID_SHIFT		12
+#define SLB_VSID_B		ASM_CONST(0xc000000000000000)
+#define SLB_VSID_B_256M		ASM_CONST(0x0000000000000000)
+#define SLB_VSID_B_1T		ASM_CONST(0x4000000000000000)
 #define SLB_VSID_KS		ASM_CONST(0x0000000000000800)
 #define SLB_VSID_KP		ASM_CONST(0x0000000000000400)
 #define SLB_VSID_N		ASM_CONST(0x0000000000000200) /* no-execute */
-#define SLB_VSID_L		ASM_CONST(0x0000000000000100) /* largepage */
+#define SLB_VSID_L		ASM_CONST(0x0000000000000100)
 #define SLB_VSID_C		ASM_CONST(0x0000000000000080) /* class */
-#define SLB_VSID_LS		ASM_CONST(0x0000000000000070) /* size of largepage */
+#define SLB_VSID_LP		ASM_CONST(0x0000000000000030)
+#define SLB_VSID_LP_00		ASM_CONST(0x0000000000000000)
+#define SLB_VSID_LP_01		ASM_CONST(0x0000000000000010)
+#define SLB_VSID_LP_10		ASM_CONST(0x0000000000000020)
+#define SLB_VSID_LP_11		ASM_CONST(0x0000000000000030)
+#define SLB_VSID_LLP		(SLB_VSID_L|SLB_VSID_LP)
 
 #define SLB_VSID_KERNEL		(SLB_VSID_KP)
 #define SLB_VSID_USER		(SLB_VSID_KP|SLB_VSID_KS|SLB_VSID_C)
@@ -69,6 +77,7 @@ extern char initial_stab[];
 #define HPTE_V_AVPN_SHIFT	7
 #define HPTE_V_AVPN		ASM_CONST(0xffffffffffffff80)
 #define HPTE_V_AVPN_VAL(x)	(((x) & HPTE_V_AVPN) >> HPTE_V_AVPN_SHIFT)
+#define HPTE_V_COMPARE(x,y)	(!(((x) ^ (y)) & HPTE_V_AVPN))
 #define HPTE_V_BOLTED		ASM_CONST(0x0000000000000010)
 #define HPTE_V_LOCK		ASM_CONST(0x0000000000000008)
 #define HPTE_V_LARGE		ASM_CONST(0x0000000000000004)
@@ -81,6 +90,7 @@ extern char initial_stab[];
 #define HPTE_R_RPN		ASM_CONST(0x3ffffffffffff000)
 #define HPTE_R_FLAGS		ASM_CONST(0x00000000000003ff)
 #define HPTE_R_PP		ASM_CONST(0x0000000000000003)
+#define HPTE_R_N		ASM_CONST(0x0000000000000004)
 
 /* Values for PP (assumes Ks=0, Kp=1) */
 /* pp0 will always be 0 for linux     */
@@ -99,100 +109,120 @@ typedef struct {
 extern hpte_t *htab_address;
 extern unsigned long htab_hash_mask;
 
-static inline unsigned long hpt_hash(unsigned long vpn, int large)
+/*
+ * Page size definition
+ *
+ *    shift : is the "PAGE_SHIFT" value for that page size
+ *    sllp  : is a bit mask with the value of SLB L || LP to be or'ed
+ *            directly to a slbmte "vsid" value
+ *    penc  : is the HPTE encoding mask for the "LP" field:
+ *
+ */
+struct mmu_psize_def
 {
-	unsigned long vsid;
-	unsigned long page;
-
-	if (large) {
-		vsid = vpn >> 4;
-		page = vpn & 0xf;
-	} else {
-		vsid = vpn >> 16;
-		page = vpn & 0xffff;
-	}
+	unsigned int	shift;	/* number of bits */
+	unsigned int	penc;	/* HPTE encoding */
+	unsigned int	tlbiel;	/* tlbiel supported for that page size */
+	unsigned long	avpnm;	/* bits to mask out in AVPN in the HPTE */
+	unsigned long	sllp;	/* SLB L||LP (exact mask to use in slbmte) */
+};
 
-	return (vsid & 0x7fffffffffUL) ^ page;
-}
+#endif /* __ASSEMBLY__ */
 
-static inline void __tlbie(unsigned long va, int large)
-{
-	/* clear top 16 bits, non SLS segment */
-	va &= ~(0xffffULL << 48);
-
-	if (large) {
-		va &= HPAGE_MASK;
-		asm volatile("tlbie %0,1" : : "r"(va) : "memory");
-	} else {
-		va &= PAGE_MASK;
-		asm volatile("tlbie %0,0" : : "r"(va) : "memory");
-	}
-}
+/*
+ * The kernel use the constants below to index in the page sizes array.
+ * The use of fixed constants for this purpose is better for performances
+ * of the low level hash refill handlers.
+ *
+ * A non supported page size has a "shift" field set to 0
+ *
+ * Any new page size being implemented can get a new entry in here. Whether
+ * the kernel will use it or not is a different matter though. The actual page
+ * size used by hugetlbfs is not defined here and may be made variable
+ */
 
-static inline void tlbie(unsigned long va, int large)
-{
-	asm volatile("ptesync": : :"memory");
-	__tlbie(va, large);
-	asm volatile("eieio; tlbsync; ptesync": : :"memory");
-}
+#define MMU_PAGE_4K		0	/* 4K */
+#define MMU_PAGE_64K		1	/* 64K */
+#define MMU_PAGE_64K_AP		2	/* 64K Admixed (in a 4K segment) */
+#define MMU_PAGE_1M		3	/* 1M */
+#define MMU_PAGE_16M		4	/* 16M */
+#define MMU_PAGE_16G		5	/* 16G */
+#define MMU_PAGE_COUNT		6
 
-static inline void __tlbiel(unsigned long va)
-{
-	/* clear top 16 bits, non SLS segment */
-	va &= ~(0xffffULL << 48);
-	va &= PAGE_MASK;
-
-	/* 
-	 * Thanks to Alan Modra we are now able to use machine specific 
-	 * assembly instructions (like tlbiel) by using the gas -many flag.
-	 * However we have to support older toolchains so for the moment 
-	 * we hardwire it.
+#ifndef __ASSEMBLY__
+
+/*
+ * The current system page sizes
  */
-#if 0
-	asm volatile("tlbiel %0" : : "r"(va) : "memory");
-#else
-	asm volatile(".long 0x7c000224 | (%0 << 11)" : : "r"(va) : "memory");
-#endif
-}
+extern struct mmu_psize_def mmu_psize_defs[MMU_PAGE_COUNT];
+extern int mmu_linear_psize;
+extern int mmu_virtual_psize;
 
-static inline void tlbiel(unsigned long va)
+#ifdef CONFIG_HUGETLB_PAGE
+/*
+ * The page size index of the huge pages for use by hugetlbfs
+ */
+extern int mmu_huge_psize;
+
+#endif /* CONFIG_HUGETLB_PAGE */
+
+/*
+ * This function sets the AVPN and L fields of the HPTE  appropriately
+ * for the page size
+ */
+static inline unsigned long hpte_encode_v(unsigned long va, int psize)
 {
-	asm volatile("ptesync": : :"memory");
-	__tlbiel(va);
-	asm volatile("ptesync": : :"memory");
+	unsigned long v =
+	v = (va >> 23) & ~(mmu_psize_defs[psize].avpnm);
+	v <<= HPTE_V_AVPN_SHIFT;
+	if (psize != MMU_PAGE_4K)
+		v |= HPTE_V_LARGE;
+	return v;
 }
 
-static inline unsigned long slot2va(unsigned long hpte_v, unsigned long slot)
+/*
+ * This function sets the ARPN, and LP fields of the HPTE appropriately
+ * for the page size. We assume the pa is already "clean" that is properly
+ * aligned for the requested page size
+ */
+static inline unsigned long hpte_encode_r(unsigned long pa, int psize)
 {
-	unsigned long avpn = HPTE_V_AVPN_VAL(hpte_v);
-	unsigned long va;
-
-	va = avpn << 23;
-
-	if (! (hpte_v & HPTE_V_LARGE)) {
-		unsigned long vpi, pteg;
-
-		pteg = slot / HPTES_PER_GROUP;
-		if (hpte_v & HPTE_V_SECONDARY)
-			pteg = ~pteg;
-
-		vpi = ((va >> 28) ^ pteg) & htab_hash_mask;
+	unsigned long r;
 
-		va |= vpi << PAGE_SHIFT;
+	/* A 4K page needs no special encoding */
+	if (psize == MMU_PAGE_4K)
+		return pa & HPTE_R_RPN;
+	else {
+		unsigned int penc = mmu_psize_defs[psize].penc;
+		unsigned int shift = mmu_psize_defs[psize].shift;
+		return (pa & ~((1ul << shift) - 1)) | (penc << 12);
 	}
-
-	return va;
+	return r;
 }
 
 /*
- * Handle a fault by adding an HPTE. If the address can't be determined
- * to be valid via Linux page tables, return 1. If handled return 0
+ * This hashes a virtual address for a 256Mb segment only for now
  */
-extern int __hash_page(unsigned long ea, unsigned long access,
+
+static inline unsigned long hpt_hash(unsigned long va, unsigned int shift)
+{
+	return ((va >> 28) & 0x7fffffffffUL) ^ ((va & 0x0fffffffUL) >> shift);
+}
+
+extern int __hash_page_4K(unsigned long ea, unsigned long access,
+			  unsigned long vsid, pte_t *ptep, unsigned long trap,
+			  unsigned int local);
+extern int __hash_page_64K(unsigned long ea, unsigned long access,
 			   unsigned long vsid, pte_t *ptep, unsigned long trap,
-		       int local);
+			   unsigned int local);
+struct mm_struct;
+extern int hash_huge_page(struct mm_struct *mm, unsigned long access,
+			  unsigned long ea, unsigned long vsid, int local);
 
 extern void htab_finish_init(void);
+extern int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
+			     unsigned long pstart, unsigned long mode,
+			     int psize);
 
 extern void hpte_init_native(void);
 extern void hpte_init_lpar(void);
@@ -200,17 +230,21 @@ extern void hpte_init_iSeries(void);
 
 extern long pSeries_lpar_hpte_insert(unsigned long hpte_group,
 				     unsigned long va, unsigned long prpn,
-				     unsigned long vflags,
-				     unsigned long rflags);
-extern long native_hpte_insert(unsigned long hpte_group, unsigned long va,
-			       unsigned long prpn,
-			       unsigned long vflags, unsigned long rflags);
+				     unsigned long rflags,
+				     unsigned long vflags, int psize);
+
+extern long native_hpte_insert(unsigned long hpte_group,
+			       unsigned long va, unsigned long prpn,
+			       unsigned long rflags,
+			       unsigned long vflags, int psize);
 
-extern long iSeries_hpte_bolt_or_insert(unsigned long hpte_group,
+extern long iSeries_hpte_insert(unsigned long hpte_group,
 				unsigned long va, unsigned long prpn,
-		unsigned long vflags, unsigned long rflags);
+				unsigned long rflags,
+				unsigned long vflags, int psize);
 
 extern void stabs_alloc(void);
+extern void slb_initialize(void);
 
 #endif /* __ASSEMBLY__ */
 

include/asm-ppc64/mmu_context.h

@@ -16,8 +16,16 @@
  * 2 of the License, or (at your option) any later version.
  */
 
-static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk)
+/*
+ * Getting into a kernel thread, there is no valid user segment, mark
+ * paca->pgdir NULL so that SLB miss on user addresses will fault
+ */
+static inline void enter_lazy_tlb(struct mm_struct *mm,
+				  struct task_struct *tsk)
 {
+#ifdef CONFIG_PPC_64K_PAGES
+	get_paca()->pgdir = NULL;
+#endif /* CONFIG_PPC_64K_PAGES */
 }
 
 #define NO_CONTEXT	0
@@ -40,8 +48,13 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		cpu_set(smp_processor_id(), next->cpu_vm_mask);
 
 	/* No need to flush userspace segments if the mm doesnt change */
+#ifdef CONFIG_PPC_64K_PAGES
+	if (prev == next && get_paca()->pgdir == next->pgd)
+		return;
+#else
 	if (prev == next)
 		return;
+#endif /* CONFIG_PPC_64K_PAGES */
 
 #ifdef CONFIG_ALTIVEC
 	if (cpu_has_feature(CPU_FTR_ALTIVEC))

include/asm-ppc64/paca.h

@@ -72,10 +72,15 @@ struct paca_struct {
 	/*
 	 * Now, starting in cacheline 2, the exception save areas
 	 */
-	u64 exgen[8] __attribute__((aligned(0x80))); /* used for most interrupts/exceptions */
-	u64 exmc[8];		/* used for machine checks */
-	u64 exslb[8];		/* used for SLB/segment table misses
+	/* used for most interrupts/exceptions */
+	u64 exgen[10] __attribute__((aligned(0x80)));
+	u64 exmc[10];		/* used for machine checks */
+	u64 exslb[10];		/* used for SLB/segment table misses
  				 * on the linear mapping */
+#ifdef CONFIG_PPC_64K_PAGES
+	pgd_t *pgdir;
+#endif /* CONFIG_PPC_64K_PAGES */
+
 	mm_context_t context;
 	u16 slb_cache[SLB_CACHE_ENTRIES];
 	u16 slb_cache_ptr;

include/asm-ppc64/page.h

@@ -13,23 +13,50 @@
 #include <linux/config.h>
 #include <asm/ppc_asm.h> /* for ASM_CONST */
 
-/* PAGE_SHIFT determines the page size */
+/*
+ * We support either 4k or 64k software page size. When using 64k pages
+ * however, wether we are really supporting 64k pages in HW or not is
+ * irrelevant to those definitions. We always define HW_PAGE_SHIFT to 12
+ * as use of 64k pages remains a linux kernel specific, every notion of
+ * page number shared with the firmware, TCEs, iommu, etc... still assumes
+ * a page size of 4096.
+ */
+#ifdef CONFIG_PPC_64K_PAGES
+#define PAGE_SHIFT		16
+#else
 #define PAGE_SHIFT		12
+#endif
+
 #define PAGE_SIZE		(ASM_CONST(1) << PAGE_SHIFT)
 #define PAGE_MASK		(~(PAGE_SIZE-1))
 
+/* HW_PAGE_SHIFT is always 4k pages */
+#define HW_PAGE_SHIFT		12
+#define HW_PAGE_SIZE		(ASM_CONST(1) << HW_PAGE_SHIFT)
+#define HW_PAGE_MASK		(~(HW_PAGE_SIZE-1))
+
+/* PAGE_FACTOR is the number of bits factor between PAGE_SHIFT and
+ * HW_PAGE_SHIFT, that is 4k pages
+ */
+#define PAGE_FACTOR		(PAGE_SHIFT - HW_PAGE_SHIFT)
+
+/* Segment size */
 #define SID_SHIFT       	28
 #define SID_MASK        	0xfffffffffUL
 #define ESID_MASK		0xfffffffff0000000UL
 #define GET_ESID(x)     	(((x) >> SID_SHIFT) & SID_MASK)
 
-#define HPAGE_SHIFT	24
+/* Large pages size */
+
+#ifndef __ASSEMBLY__
+extern unsigned int HPAGE_SHIFT;
 #define HPAGE_SIZE		((1UL) << HPAGE_SHIFT)
 #define HPAGE_MASK		(~(HPAGE_SIZE - 1))
+#define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
+#endif /* __ASSEMBLY__ */
 
 #ifdef CONFIG_HUGETLB_PAGE
 
-#define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
 
 #define HTLB_AREA_SHIFT		40
 #define HTLB_AREA_SIZE		(1UL << HTLB_AREA_SHIFT)
@@ -114,7 +141,25 @@ static __inline__ void clear_page(void *addr)
 	: "ctr", "memory");
 }
 
-extern void copy_page(void *to, void *from);
+extern void copy_4K_page(void *to, void *from);
+
+#ifdef CONFIG_PPC_64K_PAGES
+static inline void copy_page(void *to, void *from)
+{
+	unsigned int i;
+	for (i=0; i < (1 << (PAGE_SHIFT - 12)); i++) {
+		copy_4K_page(to, from);
+		to += 4096;
+		from += 4096;
+	}
+}
+#else /* CONFIG_PPC_64K_PAGES */
+static inline void copy_page(void *to, void *from)
+{
+	copy_4K_page(to, from);
+}
+#endif /* CONFIG_PPC_64K_PAGES */
+
 struct page;
 extern void clear_user_page(void *page, unsigned long vaddr, struct page *pg);
 extern void copy_user_page(void *to, void *from, unsigned long vaddr, struct page *p);
@@ -124,43 +169,75 @@ extern void copy_user_page(void *to, void *from, unsigned long vaddr, struct pag
  * These are used to make use of C type-checking.  
  * Entries in the pte table are 64b, while entries in the pgd & pmd are 32b.
  */
-typedef struct { unsigned long pte; } pte_t;
-typedef struct { unsigned long pmd; } pmd_t;
-typedef struct { unsigned long pud; } pud_t;
-typedef struct { unsigned long pgd; } pgd_t;
-typedef struct { unsigned long pgprot; } pgprot_t;
 
+/* PTE level */
+typedef struct { unsigned long pte; } pte_t;
 #define pte_val(x)	((x).pte)
-#define pmd_val(x)	((x).pmd)
-#define pud_val(x)	((x).pud)
-#define pgd_val(x)	((x).pgd)
-#define pgprot_val(x)	((x).pgprot)
-
 #define __pte(x)	((pte_t) { (x) })
+
+/* 64k pages additionally define a bigger "real PTE" type that gathers
+ * the "second half" part of the PTE for pseudo 64k pages
+ */
+#ifdef CONFIG_PPC_64K_PAGES
+typedef struct { pte_t pte; unsigned long hidx; } real_pte_t;
+#else
+typedef struct { pte_t pte; } real_pte_t;
+#endif
+
+/* PMD level */
+typedef struct { unsigned long pmd; } pmd_t;
+#define pmd_val(x)	((x).pmd)
 #define __pmd(x)	((pmd_t) { (x) })
+
+/* PUD level exusts only on 4k pages */
+#ifndef CONFIG_PPC_64K_PAGES
+typedef struct { unsigned long pud; } pud_t;
+#define pud_val(x)	((x).pud)
 #define __pud(x)	((pud_t) { (x) })
+#endif
+
+/* PGD level */
+typedef struct { unsigned long pgd; } pgd_t;
+#define pgd_val(x)	((x).pgd)
 #define __pgd(x)	((pgd_t) { (x) })
+
+/* Page protection bits */
+typedef struct { unsigned long pgprot; } pgprot_t;
+#define pgprot_val(x)	((x).pgprot)
 #define __pgprot(x)	((pgprot_t) { (x) })
 
 #else
+
 /*
  * .. while these make it easier on the compiler
  */
-typedef unsigned long pte_t;
-typedef unsigned long pmd_t;
-typedef unsigned long pud_t;
-typedef unsigned long pgd_t;
-typedef unsigned long pgprot_t;
 
+typedef unsigned long pte_t;
 #define pte_val(x)	(x)
+#define __pte(x)	(x)
+
+#ifdef CONFIG_PPC_64K_PAGES
+typedef struct { pte_t pte; unsigned long hidx; } real_pte_t;
+#else
+typedef unsigned long real_pte_t;
+#endif
+
+
+typedef unsigned long pmd_t;
 #define pmd_val(x)	(x)
+#define __pmd(x)	(x)
+
+#ifndef CONFIG_PPC_64K_PAGES
+typedef unsigned long pud_t;
 #define pud_val(x)	(x)
+#define __pud(x)	(x)
+#endif
+
+typedef unsigned long pgd_t;
 #define pgd_val(x)	(x)
 #define pgprot_val(x)	(x)
 
-#define __pte(x)	(x)
-#define __pmd(x)	(x)
-#define __pud(x)	(x)
+typedef unsigned long pgprot_t;
 #define __pgd(x)	(x)
 #define __pgprot(x)	(x)
 

include/asm-ppc64/pgalloc.h

@@ -8,10 +8,16 @@
 
 extern kmem_cache_t *pgtable_cache[];
 
+#ifdef CONFIG_PPC_64K_PAGES
+#define PTE_CACHE_NUM	0
+#define PMD_CACHE_NUM	0
+#define PGD_CACHE_NUM	1
+#else
 #define PTE_CACHE_NUM	0
 #define PMD_CACHE_NUM	1
 #define PUD_CACHE_NUM	1
 #define PGD_CACHE_NUM	0
+#endif
 
 /*
  * This program is free software; you can redistribute it and/or
@@ -30,6 +36,8 @@ static inline void pgd_free(pgd_t *pgd)
 	kmem_cache_free(pgtable_cache[PGD_CACHE_NUM], pgd);
 }
 
+#ifndef CONFIG_PPC_64K_PAGES
+
 #define pgd_populate(MM, PGD, PUD)	pgd_set(PGD, PUD)
 
 static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
@@ -43,7 +51,30 @@ static inline void pud_free(pud_t *pud)
 	kmem_cache_free(pgtable_cache[PUD_CACHE_NUM], pud);
 }
 
-#define pud_populate(MM, PUD, PMD)	pud_set(PUD, PMD)
+static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
+{
+	pud_set(pud, (unsigned long)pmd);
+}
+
+#define pmd_populate(mm, pmd, pte_page) \
+	pmd_populate_kernel(mm, pmd, page_address(pte_page))
+#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, (unsigned long)(pte))
+
+
+#else /* CONFIG_PPC_64K_PAGES */
+
+#define pud_populate(mm, pud, pmd)	pud_set(pud, (unsigned long)pmd)
+
+static inline void pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd,
+				       pte_t *pte)
+{
+	pmd_set(pmd, (unsigned long)pte);
+}
+
+#define pmd_populate(mm, pmd, pte_page) \
+	pmd_populate_kernel(mm, pmd, page_address(pte_page))
+
+#endif /* CONFIG_PPC_64K_PAGES */
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
 {
@@ -56,17 +87,15 @@ static inline void pmd_free(pmd_t *pmd)
 	kmem_cache_free(pgtable_cache[PMD_CACHE_NUM], pmd);
 }
 
-#define pmd_populate_kernel(mm, pmd, pte) pmd_set(pmd, pte)
-#define pmd_populate(mm, pmd, pte_page) \
-	pmd_populate_kernel(mm, pmd, page_address(pte_page))
-
-static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
+static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
+					  unsigned long address)
 {
 	return kmem_cache_alloc(pgtable_cache[PTE_CACHE_NUM],
 				GFP_KERNEL|__GFP_REPEAT);
 }
 
-static inline struct page *pte_alloc_one(struct mm_struct *mm, unsigned long address)
+static inline struct page *pte_alloc_one(struct mm_struct *mm,
+					 unsigned long address)
 {
 	return virt_to_page(pte_alloc_one_kernel(mm, address));
 }
@@ -103,7 +132,7 @@ static inline void pgtable_free(pgtable_free_t pgf)
 	kmem_cache_free(pgtable_cache[cachenum], p);
 }
 
-void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
+extern void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
 
 #define __pte_free_tlb(tlb, ptepage)	\
 	pgtable_free_tlb(tlb, pgtable_free_cache(page_address(ptepage), \
@@ -111,9 +140,11 @@ void pgtable_free_tlb(struct mmu_gather *tlb, pgtable_free_t pgf);
 #define __pmd_free_tlb(tlb, pmd) 	\
 	pgtable_free_tlb(tlb, pgtable_free_cache(pmd, \
 		PMD_CACHE_NUM, PMD_TABLE_SIZE-1))
+#ifndef CONFIG_PPC_64K_PAGES
 #define __pud_free_tlb(tlb, pmd)	\
 	pgtable_free_tlb(tlb, pgtable_free_cache(pud, \
 		PUD_CACHE_NUM, PUD_TABLE_SIZE-1))
+#endif /* CONFIG_PPC_64K_PAGES */
 
 #define check_pgt_cache()	do { } while (0)
 

include/asm-ppc64/pgtable-4k.h

+/*
+ * Entries per page directory level.  The PTE level must use a 64b record
+ * for each page table entry.  The PMD and PGD level use a 32b record for
+ * each entry by assuming that each entry is page aligned.
+ */
+#define PTE_INDEX_SIZE  9
+#define PMD_INDEX_SIZE  7
+#define PUD_INDEX_SIZE  7
+#define PGD_INDEX_SIZE  9
+
+#define PTE_TABLE_SIZE	(sizeof(pte_t) << PTE_INDEX_SIZE)
+#define PMD_TABLE_SIZE	(sizeof(pmd_t) << PMD_INDEX_SIZE)
+#define PUD_TABLE_SIZE	(sizeof(pud_t) << PUD_INDEX_SIZE)
+#define PGD_TABLE_SIZE	(sizeof(pgd_t) << PGD_INDEX_SIZE)
+
+#define PTRS_PER_PTE	(1 << PTE_INDEX_SIZE)
+#define PTRS_PER_PMD	(1 << PMD_INDEX_SIZE)
+#define PTRS_PER_PUD	(1 << PMD_INDEX_SIZE)
+#define PTRS_PER_PGD	(1 << PGD_INDEX_SIZE)
+
+/* PMD_SHIFT determines what a second-level page table entry can map */
+#define PMD_SHIFT	(PAGE_SHIFT + PTE_INDEX_SIZE)
+#define PMD_SIZE	(1UL << PMD_SHIFT)
+#define PMD_MASK	(~(PMD_SIZE-1))
+
+/* PUD_SHIFT determines what a third-level page table entry can map */
+#define PUD_SHIFT	(PMD_SHIFT + PMD_INDEX_SIZE)
+#define PUD_SIZE	(1UL << PUD_SHIFT)
+#define PUD_MASK	(~(PUD_SIZE-1))
+
+/* PGDIR_SHIFT determines what a fourth-level page table entry can map */
+#define PGDIR_SHIFT	(PUD_SHIFT + PUD_INDEX_SIZE)
+#define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
+#define PGDIR_MASK	(~(PGDIR_SIZE-1))
+
+/* PTE bits */
+#define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */
+#define _PAGE_GROUP_IX  0x7000 /* software: HPTE index within group */
+#define _PAGE_F_SECOND  _PAGE_SECONDARY
+#define _PAGE_F_GIX     _PAGE_GROUP_IX
+
+/* PTE flags to conserve for HPTE identification */
+#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | \
+			 _PAGE_SECONDARY | _PAGE_GROUP_IX)
+
+/* PAGE_MASK gives the right answer below, but only by accident */
+/* It should be preserving the high 48 bits and then specifically */
+/* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */
+#define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | \
+                         _PAGE_HPTEFLAGS)
+
+/* Bits to mask out from a PMD to get to the PTE page */
+#define PMD_MASKED_BITS		0
+/* Bits to mask out from a PUD to get to the PMD page */
+#define PUD_MASKED_BITS		0
+/* Bits to mask out from a PGD to get to the PUD page */
+#define PGD_MASKED_BITS		0
+
+/* shift to put page number into pte */
+#define PTE_RPN_SHIFT	(17)
+
+#define __real_pte(e,p)		((real_pte_t)(e))
+#define __rpte_to_pte(r)	(r)
+#define __rpte_to_hidx(r,index)	(pte_val((r)) >> 12)
+
+#define pte_iterate_hashed_subpages(rpte, psize, va, index, shift)       \
+	do {							         \
+		index = 0;					         \
+		shift = mmu_psize_defs[psize].shift;		         \
+
+#define pte_iterate_hashed_end() } while(0)
+
+/*
+ * 4-level page tables related bits
+ */
+
+#define pgd_none(pgd)		(!pgd_val(pgd))
+#define pgd_bad(pgd)		(pgd_val(pgd) == 0)
+#define pgd_present(pgd)	(pgd_val(pgd) != 0)
+#define pgd_clear(pgdp)		(pgd_val(*(pgdp)) = 0)
+#define pgd_page(pgd)		(pgd_val(pgd) & ~PGD_MASKED_BITS)
+
+#define pud_offset(pgdp, addr)	\
+  (((pud_t *) pgd_page(*(pgdp))) + \
+    (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
+
+#define pud_ERROR(e) \
+	printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pud_val(e))

include/asm-ppc64/pgtable-64k.h

+#include <asm-generic/pgtable-nopud.h>
+
+
+#define PTE_INDEX_SIZE  12
+#define PMD_INDEX_SIZE  12
+#define PUD_INDEX_SIZE	0
+#define PGD_INDEX_SIZE  4
+
+#define PTE_TABLE_SIZE	(sizeof(real_pte_t) << PTE_INDEX_SIZE)
+#define PMD_TABLE_SIZE	(sizeof(pmd_t) << PMD_INDEX_SIZE)
+#define PGD_TABLE_SIZE	(sizeof(pgd_t) << PGD_INDEX_SIZE)
+
+#define PTRS_PER_PTE	(1 << PTE_INDEX_SIZE)
+#define PTRS_PER_PMD	(1 << PMD_INDEX_SIZE)
+#define PTRS_PER_PGD	(1 << PGD_INDEX_SIZE)
+
+/* PMD_SHIFT determines what a second-level page table entry can map */
+#define PMD_SHIFT	(PAGE_SHIFT + PTE_INDEX_SIZE)
+#define PMD_SIZE	(1UL << PMD_SHIFT)
+#define PMD_MASK	(~(PMD_SIZE-1))
+
+/* PGDIR_SHIFT determines what a third-level page table entry can map */
+#define PGDIR_SHIFT	(PMD_SHIFT + PMD_INDEX_SIZE)
+#define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
+#define PGDIR_MASK	(~(PGDIR_SIZE-1))
+
+/* Additional PTE bits (don't change without checking asm in hash_low.S) */
+#define _PAGE_HPTE_SUB	0x0ffff000 /* combo only: sub pages HPTE bits */
+#define _PAGE_HPTE_SUB0	0x08000000 /* combo only: first sub page */
+#define _PAGE_COMBO	0x10000000 /* this is a combo 4k page */
+#define _PAGE_F_SECOND  0x00008000 /* full page: hidx bits */
+#define _PAGE_F_GIX     0x00007000 /* full page: hidx bits */
+
+/* PTE flags to conserve for HPTE identification */
+#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_HPTE_SUB |\
+                         _PAGE_COMBO)
+
+/* Shift to put page number into pte.
+ *
+ * That gives us a max RPN of 32 bits, which means a max of 48 bits
+ * of addressable physical space.
+ * We could get 3 more bits here by setting PTE_RPN_SHIFT to 29 but
+ * 32 makes PTEs more readable for debugging for now :)
+ */
+#define PTE_RPN_SHIFT	(32)
+#define PTE_RPN_MAX	(1UL << (64 - PTE_RPN_SHIFT))
+#define PTE_RPN_MASK	(~((1UL<<PTE_RPN_SHIFT)-1))
+
+/* _PAGE_CHG_MASK masks of bits that are to be preserved accross
+ * pgprot changes
+ */
+#define _PAGE_CHG_MASK	(PTE_RPN_MASK | _PAGE_HPTEFLAGS | _PAGE_DIRTY | \
+                         _PAGE_ACCESSED)
+
+/* Bits to mask out from a PMD to get to the PTE page */
+#define PMD_MASKED_BITS		0x1ff
+/* Bits to mask out from a PGD/PUD to get to the PMD page */
+#define PUD_MASKED_BITS		0x1ff
+
+#ifndef __ASSEMBLY__
+
+/* Manipulate "rpte" values */
+#define __real_pte(e,p) 	((real_pte_t) { \
+	(e), pte_val(*((p) + PTRS_PER_PTE)) })
+#define __rpte_to_hidx(r,index)	((pte_val((r).pte) & _PAGE_COMBO) ? \
+        (((r).hidx >> ((index)<<2)) & 0xf) : ((pte_val((r).pte) >> 12) & 0xf))
+#define __rpte_to_pte(r)	((r).pte)
+#define __rpte_sub_valid(rpte, index) \
+	(pte_val(rpte.pte) & (_PAGE_HPTE_SUB0 >> (index)))
+
+
+/* Trick: we set __end to va + 64k, which happens works for
+ * a 16M page as well as we want only one iteration
+ */
+#define pte_iterate_hashed_subpages(rpte, psize, va, index, shift)	    \
+        do {                                                                \
+                unsigned long __end = va + PAGE_SIZE;                       \
+                unsigned __split = (psize == MMU_PAGE_4K ||                 \
+				    psize == MMU_PAGE_64K_AP);              \
+                shift = mmu_psize_defs[psize].shift;                        \
+	        for (index = 0; va < __end; index++, va += (1 << shift)) {  \
+		        if (!__split || __rpte_sub_valid(rpte, index)) do { \
+
+#define pte_iterate_hashed_end() } while(0); } } while(0)
+
+
+#endif /*  __ASSEMBLY__ */

include/asm-ppc64/pgtable.h

@@ -15,40 +15,11 @@
 #include <asm/tlbflush.h>
 #endif /* __ASSEMBLY__ */
 
-/*
- * Entries per page directory level.  The PTE level must use a 64b record
- * for each page table entry.  The PMD and PGD level use a 32b record for 
- * each entry by assuming that each entry is page aligned.
- */
-#define PTE_INDEX_SIZE  9
-#define PMD_INDEX_SIZE  7
-#define PUD_INDEX_SIZE  7
-#define PGD_INDEX_SIZE  9
-
-#define PTE_TABLE_SIZE	(sizeof(pte_t) << PTE_INDEX_SIZE)
-#define PMD_TABLE_SIZE	(sizeof(pmd_t) << PMD_INDEX_SIZE)
-#define PUD_TABLE_SIZE	(sizeof(pud_t) << PUD_INDEX_SIZE)
-#define PGD_TABLE_SIZE	(sizeof(pgd_t) << PGD_INDEX_SIZE)
-
-#define PTRS_PER_PTE	(1 << PTE_INDEX_SIZE)
-#define PTRS_PER_PMD	(1 << PMD_INDEX_SIZE)
-#define PTRS_PER_PUD	(1 << PMD_INDEX_SIZE)
-#define PTRS_PER_PGD	(1 << PGD_INDEX_SIZE)
-
-/* PMD_SHIFT determines what a second-level page table entry can map */
-#define PMD_SHIFT	(PAGE_SHIFT + PTE_INDEX_SIZE)
-#define PMD_SIZE	(1UL << PMD_SHIFT)
-#define PMD_MASK	(~(PMD_SIZE-1))
-
-/* PUD_SHIFT determines what a third-level page table entry can map */
-#define PUD_SHIFT	(PMD_SHIFT + PMD_INDEX_SIZE)
-#define PUD_SIZE	(1UL << PUD_SHIFT)
-#define PUD_MASK	(~(PUD_SIZE-1))
-
-/* PGDIR_SHIFT determines what a fourth-level page table entry can map */
-#define PGDIR_SHIFT	(PUD_SHIFT + PUD_INDEX_SIZE)
-#define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
-#define PGDIR_MASK	(~(PGDIR_SIZE-1))
+#ifdef CONFIG_PPC_64K_PAGES
+#include <asm/pgtable-64k.h>
+#else
+#include <asm/pgtable-4k.h>
+#endif
 
 #define FIRST_USER_ADDRESS	0
 
@@ -75,8 +46,9 @@
 #define VMALLOC_END   (VMALLOC_START + VMALLOC_SIZE)
 
 /*
- * Bits in a linux-style PTE.  These match the bits in the
- * (hardware-defined) PowerPC PTE as closely as possible.
+ * Common bits in a linux-style PTE.  These match the bits in the
+ * (hardware-defined) PowerPC PTE as closely as possible. Additional
+ * bits may be defined in pgtable-*.h
  */
 #define _PAGE_PRESENT	0x0001 /* software: pte contains a translation */
 #define _PAGE_USER	0x0002 /* matches one of the PP bits */
@@ -91,15 +63,6 @@
 #define _PAGE_RW	0x0200 /* software: user write access allowed */
 #define _PAGE_HASHPTE	0x0400 /* software: pte has an associated HPTE */
 #define _PAGE_BUSY	0x0800 /* software: PTE & hash are busy */ 
-#define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */
-#define _PAGE_GROUP_IX  0x7000 /* software: HPTE index within group */
-#define _PAGE_HUGE	0x10000 /* 16MB page */
-/* Bits 0x7000 identify the index within an HPT Group */
-#define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_SECONDARY | _PAGE_GROUP_IX)
-/* PAGE_MASK gives the right answer below, but only by accident */
-/* It should be preserving the high 48 bits and then specifically */
-/* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */
-#define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HPTEFLAGS)
 
 #define _PAGE_BASE	(_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
 
@@ -122,10 +85,10 @@
 #define PAGE_AGP	__pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
 #define HAVE_PAGE_AGP
 
-/*
- * This bit in a hardware PTE indicates that the page is *not* executable.
- */
-#define HW_NO_EXEC	_PAGE_EXEC
+/* PTEIDX nibble */
+#define _PTEIDX_SECONDARY	0x8
+#define _PTEIDX_GROUP_IX	0x7
+
 
 /*
  * POWER4 and newer have per page execute protection, older chips can only
@@ -164,21 +127,10 @@ extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
 #endif /* __ASSEMBLY__ */
 
-/* shift to put page number into pte */
-#define PTE_SHIFT (17)
-
 #ifdef CONFIG_HUGETLB_PAGE
 
-#ifndef __ASSEMBLY__
-int hash_huge_page(struct mm_struct *mm, unsigned long access,
-		   unsigned long ea, unsigned long vsid, int local);
-#endif /* __ASSEMBLY__ */
-
 #define HAVE_ARCH_UNMAPPED_AREA
 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
-#else
-
-#define hash_huge_page(mm,a,ea,vsid,local)	-1
 
 #endif
 
@@ -197,7 +149,7 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 	pte_t pte;
 
 
-	pte_val(pte) = (pfn << PTE_SHIFT) | pgprot_val(pgprot);
+	pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
 	return pte;
 }
 
@@ -209,30 +161,25 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 
 /* pte_clear moved to later in this file */
 
-#define pte_pfn(x)		((unsigned long)((pte_val(x) >> PTE_SHIFT)))
+#define pte_pfn(x)		((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
 #define pte_page(x)		pfn_to_page(pte_pfn(x))
 
-#define pmd_set(pmdp, ptep) 	({BUG_ON((u64)ptep < KERNELBASE); pmd_val(*(pmdp)) = (unsigned long)(ptep);})
+#define pmd_set(pmdp, pmdval) 	(pmd_val(*(pmdp)) = (pmdval))
 #define pmd_none(pmd)		(!pmd_val(pmd))
 #define	pmd_bad(pmd)		(pmd_val(pmd) == 0)
 #define	pmd_present(pmd)	(pmd_val(pmd) != 0)
 #define	pmd_clear(pmdp)		(pmd_val(*(pmdp)) = 0)
-#define pmd_page_kernel(pmd)	(pmd_val(pmd))
+#define pmd_page_kernel(pmd)	(pmd_val(pmd) & ~PMD_MASKED_BITS)
 #define pmd_page(pmd)		virt_to_page(pmd_page_kernel(pmd))
 
-#define pud_set(pudp, pmdp)	(pud_val(*(pudp)) = (unsigned long)(pmdp))
+#define pud_set(pudp, pudval)	(pud_val(*(pudp)) = (pudval))
 #define pud_none(pud)		(!pud_val(pud))
 #define pud_bad(pud)		((pud_val(pud)) == 0)
 #define pud_present(pud)	(pud_val(pud) != 0)
 #define pud_clear(pudp)		(pud_val(*(pudp)) = 0)
-#define pud_page(pud)		(pud_val(pud))
+#define pud_page(pud)		(pud_val(pud) & ~PUD_MASKED_BITS)
 
 #define pgd_set(pgdp, pudp)	({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
-#define pgd_none(pgd)		(!pgd_val(pgd))
-#define pgd_bad(pgd)		(pgd_val(pgd) == 0)
-#define pgd_present(pgd)	(pgd_val(pgd) != 0)
-#define pgd_clear(pgdp)		(pgd_val(*(pgdp)) = 0)
-#define pgd_page(pgd)		(pgd_val(pgd))
 
 /* 
  * Find an entry in a page-table-directory.  We combine the address region 
@@ -243,9 +190,6 @@ static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
 
 #define pgd_offset(mm, address)	 ((mm)->pgd + pgd_index(address))
 
-#define pud_offset(pgdp, addr)	\
-  (((pud_t *) pgd_page(*(pgdp))) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
-
 #define pmd_offset(pudp,addr) \
   (((pmd_t *) pud_page(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
 
@@ -271,7 +215,6 @@ static inline int pte_exec(pte_t pte)  { return pte_val(pte) & _PAGE_EXEC;}
 static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;}
 static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;}
 static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;}
-static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_HUGE;}
 
 static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
 static inline void pte_cache(pte_t pte)   { pte_val(pte) &= ~_PAGE_NO_CACHE; }
@@ -286,7 +229,6 @@ static inline pte_t pte_mkclean(pte_t pte) {
 	pte_val(pte) &= ~(_PAGE_DIRTY); return pte; }
 static inline pte_t pte_mkold(pte_t pte) {
 	pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
-
 static inline pte_t pte_mkread(pte_t pte) {
 	pte_val(pte) |= _PAGE_USER; return pte; }
 static inline pte_t pte_mkexec(pte_t pte) {
@@ -298,7 +240,7 @@ static inline pte_t pte_mkdirty(pte_t pte) {
 static inline pte_t pte_mkyoung(pte_t pte) {
 	pte_val(pte) |= _PAGE_ACCESSED; return pte; }
 static inline pte_t pte_mkhuge(pte_t pte) {
-	pte_val(pte) |= _PAGE_HUGE; return pte; }
+	return pte; }
 
 /* Atomic PTE updates */
 static inline unsigned long pte_update(pte_t *p, unsigned long clr)
@@ -321,11 +263,13 @@ static inline unsigned long pte_update(pte_t *p, unsigned long clr)
 /* PTE updating functions, this function puts the PTE in the
  * batch, doesn't actually triggers the hash flush immediately,
  * you need to call flush_tlb_pending() to do that.
+ * Pass -1 for "normal" size (4K or 64K)
  */
-extern void hpte_update(struct mm_struct *mm, unsigned long addr, unsigned long pte,
-			int wrprot);
+extern void hpte_update(struct mm_struct *mm, unsigned long addr,
+			pte_t *ptep, unsigned long pte, int huge);
 
-static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
+					      unsigned long addr, pte_t *ptep)
 {
 	unsigned long old;
 
@@ -333,7 +277,7 @@ static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned lon
 		return 0;
 	old = pte_update(ptep, _PAGE_ACCESSED);
 	if (old & _PAGE_HASHPTE) {
-		hpte_update(mm, addr, old, 0);
+		hpte_update(mm, addr, ptep, old, 0);
 		flush_tlb_pending();
 	}
 	return (old & _PAGE_ACCESSED) != 0;
@@ -351,7 +295,8 @@ static inline int __ptep_test_and_clear_young(struct mm_struct *mm, unsigned lon
  * moment we always flush but we need to fix hpte_update and test if the
  * optimisation is worth it.
  */
-static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm,
+					      unsigned long addr, pte_t *ptep)
 {
 	unsigned long old;
 
@@ -359,7 +304,7 @@ static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned lon
 		return 0;
 	old = pte_update(ptep, _PAGE_DIRTY);
 	if (old & _PAGE_HASHPTE)
-		hpte_update(mm, addr, old, 0);
+		hpte_update(mm, addr, ptep, old, 0);
 	return (old & _PAGE_DIRTY) != 0;
 }
 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
@@ -371,7 +316,8 @@ static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm, unsigned lon
 })
 
 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
-static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
+				      pte_t *ptep)
 {
 	unsigned long old;
 
@@ -379,7 +325,7 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
        		return;
 	old = pte_update(ptep, _PAGE_RW);
 	if (old & _PAGE_HASHPTE)
-		hpte_update(mm, addr, old, 0);
+		hpte_update(mm, addr, ptep, old, 0);
 }
 
 /*
@@ -408,21 +354,23 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
 })
 
 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
-static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
+				       unsigned long addr, pte_t *ptep)
 {
 	unsigned long old = pte_update(ptep, ~0UL);
 
 	if (old & _PAGE_HASHPTE)
-		hpte_update(mm, addr, old, 0);
+		hpte_update(mm, addr, ptep, old, 0);
 	return __pte(old);
 }
 
-static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t * ptep)
+static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
+			     pte_t * ptep)
 {
 	unsigned long old = pte_update(ptep, ~0UL);
 
 	if (old & _PAGE_HASHPTE)
-		hpte_update(mm, addr, old, 0);
+		hpte_update(mm, addr, ptep, old, 0);
 }
 
 /*
@@ -435,7 +383,14 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
 		pte_clear(mm, addr, ptep);
 		flush_tlb_pending();
 	}
-	*ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
+	pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
+
+#ifdef CONFIG_PPC_64K_PAGES
+	if (mmu_virtual_psize != MMU_PAGE_64K)
+		pte = __pte(pte_val(pte) | _PAGE_COMBO);
+#endif /* CONFIG_PPC_64K_PAGES */
+
+	*ptep = pte;
 }
 
 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
@@ -482,8 +437,6 @@ extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
 	printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
 #define pmd_ERROR(e) \
 	printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
-#define pud_ERROR(e) \
-	printk("%s:%d: bad pud %08lx.\n", __FILE__, __LINE__, pud_val(e))
 #define pgd_ERROR(e) \
 	printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
 
@@ -509,12 +462,12 @@ extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t);
 /* Encode and de-code a swap entry */
 #define __swp_type(entry)	(((entry).val >> 1) & 0x3f)
 #define __swp_offset(entry)	((entry).val >> 8)
-#define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
-#define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) >> PTE_SHIFT })
-#define __swp_entry_to_pte(x)	((pte_t) { (x).val << PTE_SHIFT })
-#define pte_to_pgoff(pte)	(pte_val(pte) >> PTE_SHIFT)
-#define pgoff_to_pte(off)	((pte_t) {((off) << PTE_SHIFT)|_PAGE_FILE})
-#define PTE_FILE_MAX_BITS	(BITS_PER_LONG - PTE_SHIFT)
+#define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
+#define __pte_to_swp_entry(pte)	((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
+#define __swp_entry_to_pte(x)	((pte_t) { (x).val << PTE_RPN_SHIFT })
+#define pte_to_pgoff(pte)	(pte_val(pte) >> PTE_RPN_SHIFT)
+#define pgoff_to_pte(off)	((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
+#define PTE_FILE_MAX_BITS	(BITS_PER_LONG - PTE_RPN_SHIFT)
 
 /*
  * kern_addr_valid is intended to indicate whether an address is a valid
@@ -532,29 +485,22 @@ void pgtable_cache_init(void);
 /*
  * find_linux_pte returns the address of a linux pte for a given 
  * effective address and directory.  If not found, it returns zero.
- */
-static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
+ */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
 {
 	pgd_t *pg;
 	pud_t *pu;
 	pmd_t *pm;
 	pte_t *pt = NULL;
-	pte_t pte;
 
 	pg = pgdir + pgd_index(ea);
 	if (!pgd_none(*pg)) {
 		pu = pud_offset(pg, ea);
 		if (!pud_none(*pu)) {
 			pm = pmd_offset(pu, ea);
-			if (pmd_present(*pm)) {
+			if (pmd_present(*pm))
 				pt = pte_offset_kernel(pm, ea);
-				pte = *pt;
-				if (!pte_present(pte))
-					pt = NULL;
 		}
 	}
-	}
-
 	return pt;
 }
 

include/asm-ppc64/prom.h

@@ -188,6 +188,14 @@ extern struct device_node *of_get_next_child(const struct device_node *node,
 extern struct device_node *of_node_get(struct device_node *node);
 extern void of_node_put(struct device_node *node);
 
+/* For scanning the flat device-tree at boot time */
+int __init of_scan_flat_dt(int (*it)(unsigned long node,
+				     const char *uname, int depth,
+				     void *data),
+			   void *data);
+void* __init of_get_flat_dt_prop(unsigned long node, const char *name,
+				 unsigned long *size);
+
 /* For updating the device tree at runtime */
 extern void of_attach_node(struct device_node *);
 extern void of_detach_node(const struct device_node *);

include/asm-ppc64/system.h

@@ -248,7 +248,7 @@ __cmpxchg_u32(volatile unsigned int *p, unsigned long old, unsigned long new)
 }
 
 static __inline__ unsigned long
-__cmpxchg_u64(volatile long *p, unsigned long old, unsigned long new)
+__cmpxchg_u64(volatile unsigned long *p, unsigned long old, unsigned long new)
 {
 	unsigned long prev;
 

mm/hugetlb.c

@@ -103,6 +103,9 @@ static int __init hugetlb_init(void)
 	unsigned long i;
 	struct page *page;
 
+	if (HPAGE_SHIFT == 0)
+		return 0;
+
 	for (i = 0; i < MAX_NUMNODES; ++i)
 		INIT_LIST_HEAD(&hugepage_freelists[i]);