Update from upstream with manual merge of Yasunori Goto's

changes to swiotlb.c made in commit 281dd25cdc0d6903929b79183816d151ea626341
since this file has been moved from arch/ia64/lib/swiotlb.c to
lib/swiotlb.c

Signed-off-by: Tony Luck <tony.luck@intel.com>

5 files changed, 404 insertions, 727 deletions

diff --git a/arch/sparc64/mm/Makefile b/arch/sparc64/mm/Makefile
index cda87333a77..9d0960e69f4 100644
--- a/arch/sparc64/mm/Makefile
+++ b/arch/sparc64/mm/Makefile
@@ -5,6 +5,6 @@
 EXTRA_AFLAGS := -ansi
 EXTRA_CFLAGS := -Werror
 
-obj-y    := ultra.o tlb.o fault.o init.o generic.o extable.o
+obj-y    := ultra.o tlb.o fault.o init.o generic.o
 
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
diff --git a/arch/sparc64/mm/extable.c b/arch/sparc64/mm/extable.c
deleted file mode 100644
index ec334297ff4..00000000000
--- a/arch/sparc64/mm/extable.c
+++ /dev/null
@@ -1,80 +0,0 @@
-/*
- * linux/arch/sparc64/mm/extable.c
- */
-
-#include <linux/config.h>
-#include <linux/module.h>
-#include <asm/uaccess.h>
-
-extern const struct exception_table_entry __start___ex_table[];
-extern const struct exception_table_entry __stop___ex_table[];
-
-void sort_extable(struct exception_table_entry *start,
-		  struct exception_table_entry *finish)
-{
-}
-
-/* Caller knows they are in a range if ret->fixup == 0 */
-const struct exception_table_entry *
-search_extable(const struct exception_table_entry *start,
-	       const struct exception_table_entry *last,
-	       unsigned long value)
-{
-	const struct exception_table_entry *walk;
-
-	/* Single insn entries are encoded as:
-	 *	word 1:	insn address
-	 *	word 2:	fixup code address
-	 *
-	 * Range entries are encoded as:
-	 *	word 1: first insn address
-	 *	word 2: 0
-	 *	word 3: last insn address + 4 bytes
-	 *	word 4: fixup code address
-	 *
-	 * See asm/uaccess.h for more details.
-	 */
-
-	/* 1. Try to find an exact match. */
-	for (walk = start; walk <= last; walk++) {
-		if (walk->fixup == 0) {
-			/* A range entry, skip both parts. */
-			walk++;
-			continue;
-		}
-
-		if (walk->insn == value)
-			return walk;
-	}
-
-	/* 2. Try to find a range match. */
-	for (walk = start; walk <= (last - 1); walk++) {
-		if (walk->fixup)
-			continue;
-
-		if (walk[0].insn <= value && walk[1].insn > value)
-			return walk;
-
-		walk++;
-	}
-
-        return NULL;
-}
-
-/* Special extable search, which handles ranges.  Returns fixup */
-unsigned long search_extables_range(unsigned long addr, unsigned long *g2)
-{
-	const struct exception_table_entry *entry;
-
-	entry = search_exception_tables(addr);
-	if (!entry)
-		return 0;
-
-	/* Inside range?  Fix g2 and return correct fixup */
-	if (!entry->fixup) {
-		*g2 = (addr - entry->insn) / 4;
-		return (entry + 1)->fixup;
-	}
-
-	return entry->fixup;
-}
diff --git a/arch/sparc64/mm/fault.c b/arch/sparc64/mm/fault.c
index db1e3310e90..31fbc67719a 100644
--- a/arch/sparc64/mm/fault.c
+++ b/arch/sparc64/mm/fault.c
@@ -32,8 +32,6 @@
 
 #define ELEMENTS(arr) (sizeof (arr)/sizeof (arr[0]))
 
-extern struct sparc_phys_banks sp_banks[SPARC_PHYS_BANKS];
-
 /*
  * To debug kernel to catch accesses to certain virtual/physical addresses.
  * Mode = 0 selects physical watchpoints, mode = 1 selects virtual watchpoints.
@@ -71,53 +69,6 @@ void set_brkpt(unsigned long addr, unsigned char mask, int flags, int mode)
 			     : "memory");
 }
 
-/* Nice, simple, prom library does all the sweating for us. ;) */
-unsigned long __init prom_probe_memory (void)
-{
-	register struct linux_mlist_p1275 *mlist;
-	register unsigned long bytes, base_paddr, tally;
-	register int i;
-
-	i = 0;
-	mlist = *prom_meminfo()->p1275_available;
-	bytes = tally = mlist->num_bytes;
-	base_paddr = mlist->start_adr;
-  
-	sp_banks[0].base_addr = base_paddr;
-	sp_banks[0].num_bytes = bytes;
-
-	while (mlist->theres_more != (void *) 0) {
-		i++;
-		mlist = mlist->theres_more;
-		bytes = mlist->num_bytes;
-		tally += bytes;
-		if (i >= SPARC_PHYS_BANKS-1) {
-			printk ("The machine has more banks than "
-				"this kernel can support\n"
-				"Increase the SPARC_PHYS_BANKS "
-				"setting (currently %d)\n",
-				SPARC_PHYS_BANKS);
-			i = SPARC_PHYS_BANKS-1;
-			break;
-		}
-    
-		sp_banks[i].base_addr = mlist->start_adr;
-		sp_banks[i].num_bytes = mlist->num_bytes;
-	}
-
-	i++;
-	sp_banks[i].base_addr = 0xdeadbeefbeefdeadUL;
-	sp_banks[i].num_bytes = 0;
-
-	/* Now mask all bank sizes on a page boundary, it is all we can
-	 * use anyways.
-	 */
-	for (i = 0; sp_banks[i].num_bytes != 0; i++)
-		sp_banks[i].num_bytes &= PAGE_MASK;
-
-	return tally;
-}
-
 static void __kprobes unhandled_fault(unsigned long address,
 				      struct task_struct *tsk,
 				      struct pt_regs *regs)
@@ -242,7 +193,6 @@ static unsigned int get_fault_insn(struct pt_regs *regs, unsigned int insn)
 static void do_kernel_fault(struct pt_regs *regs, int si_code, int fault_code,
 			    unsigned int insn, unsigned long address)
 {
-	unsigned long g2;
 	unsigned char asi = ASI_P;
  
 	if ((!insn) && (regs->tstate & TSTATE_PRIV))
@@ -273,11 +223,9 @@ static void do_kernel_fault(struct pt_regs *regs, int si_code, int fault_code,
 		}
 	}
 		
-	g2 = regs->u_regs[UREG_G2];
-
 	/* Is this in ex_table? */
 	if (regs->tstate & TSTATE_PRIV) {
-		unsigned long fixup;
+		const struct exception_table_entry *entry;
 
 		if (asi == ASI_P && (insn & 0xc0800000) == 0xc0800000) {
 			if (insn & 0x2000)
@@ -288,10 +236,9 @@ static void do_kernel_fault(struct pt_regs *regs, int si_code, int fault_code,
 	
 		/* Look in asi.h: All _S asis have LS bit set */
 		if ((asi & 0x1) &&
-		    (fixup = search_extables_range(regs->tpc, &g2))) {
-			regs->tpc = fixup;
+		    (entry = search_exception_tables(regs->tpc))) {
+			regs->tpc = entry->fixup;
 			regs->tnpc = regs->tpc + 4;
-			regs->u_regs[UREG_G2] = g2;
 			return;
 		}
 	} else {
@@ -461,7 +408,7 @@ good_area:
 	}
 
 	up_read(&mm->mmap_sem);
-	goto fault_done;
+	return;
 
 	/*
 	 * Something tried to access memory that isn't in our memory map..
@@ -473,8 +420,7 @@ bad_area:
 
 handle_kernel_fault:
 	do_kernel_fault(regs, si_code, fault_code, insn, address);
-
-	goto fault_done;
+	return;
 
 /*
  * We ran out of memory, or some other thing happened to us that made
@@ -505,9 +451,4 @@ do_sigbus:
 	/* Kernel mode? Handle exceptions or die */
 	if (regs->tstate & TSTATE_PRIV)
 		goto handle_kernel_fault;
-
-fault_done:
-	/* These values are no longer needed, clear them. */
-	set_thread_fault_code(0);
-	current_thread_info()->fault_address = 0;
 }
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index fdb1ebb308c..1e44ee26cee 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -20,6 +20,8 @@
 #include <linux/fs.h>
 #include <linux/seq_file.h>
 #include <linux/kprobes.h>
+#include <linux/cache.h>
+#include <linux/sort.h>
 
 #include <asm/head.h>
 #include <asm/system.h>
@@ -40,24 +42,80 @@
 
 extern void device_scan(void);
 
-struct sparc_phys_banks sp_banks[SPARC_PHYS_BANKS];
+#define MAX_BANKS	32
 
-unsigned long *sparc64_valid_addr_bitmap;
+static struct linux_prom64_registers pavail[MAX_BANKS] __initdata;
+static struct linux_prom64_registers pavail_rescan[MAX_BANKS] __initdata;
+static int pavail_ents __initdata;
+static int pavail_rescan_ents __initdata;
+
+static int cmp_p64(const void *a, const void *b)
+{
+	const struct linux_prom64_registers *x = a, *y = b;
+
+	if (x->phys_addr > y->phys_addr)
+		return 1;
+	if (x->phys_addr < y->phys_addr)
+		return -1;
+	return 0;
+}
+
+static void __init read_obp_memory(const char *property,
+				   struct linux_prom64_registers *regs,
+				   int *num_ents)
+{
+	int node = prom_finddevice("/memory");
+	int prop_size = prom_getproplen(node, property);
+	int ents, ret, i;
+
+	ents = prop_size / sizeof(struct linux_prom64_registers);
+	if (ents > MAX_BANKS) {
+		prom_printf("The machine has more %s property entries than "
+			    "this kernel can support (%d).\n",
+			    property, MAX_BANKS);
+		prom_halt();
+	}
+
+	ret = prom_getproperty(node, property, (char *) regs, prop_size);
+	if (ret == -1) {
+		prom_printf("Couldn't get %s property from /memory.\n");
+		prom_halt();
+	}
+
+	*num_ents = ents;
+
+	/* Sanitize what we got from the firmware, by page aligning
+	 * everything.
+	 */
+	for (i = 0; i < ents; i++) {
+		unsigned long base, size;
+
+		base = regs[i].phys_addr;
+		size = regs[i].reg_size;
+
+		size &= PAGE_MASK;
+		if (base & ~PAGE_MASK) {
+			unsigned long new_base = PAGE_ALIGN(base);
+
+			size -= new_base - base;
+			if ((long) size < 0L)
+				size = 0UL;
+			base = new_base;
+		}
+		regs[i].phys_addr = base;
+		regs[i].reg_size = size;
+	}
+	sort(regs, ents, sizeof(struct linux_prom64_registers),
+	     cmp_p64, NULL);
+}
+
+unsigned long *sparc64_valid_addr_bitmap __read_mostly;
 
 /* Ugly, but necessary... -DaveM */
-unsigned long phys_base;
-unsigned long kern_base;
-unsigned long kern_size;
-unsigned long pfn_base;
-
-/* This is even uglier. We have a problem where the kernel may not be
- * located at phys_base. However, initial __alloc_bootmem() calls need to
- * be adjusted to be within the 4-8Megs that the kernel is mapped to, else
- * those page mappings wont work. Things are ok after inherit_prom_mappings
- * is called though. Dave says he'll clean this up some other time.
- * -- BenC
- */
-static unsigned long bootmap_base;
+unsigned long phys_base __read_mostly;
+unsigned long kern_base __read_mostly;
+unsigned long kern_size __read_mostly;
+unsigned long pfn_base __read_mostly;
 
 /* get_new_mmu_context() uses "cache + 1".  */
 DEFINE_SPINLOCK(ctx_alloc_lock);
@@ -73,7 +131,13 @@ extern unsigned long sparc_ramdisk_image64;
 extern unsigned int sparc_ramdisk_image;
 extern unsigned int sparc_ramdisk_size;
 
-struct page *mem_map_zero;
+struct page *mem_map_zero __read_mostly;
+
+unsigned int sparc64_highest_unlocked_tlb_ent __read_mostly;
+
+unsigned long sparc64_kern_pri_context __read_mostly;
+unsigned long sparc64_kern_pri_nuc_bits __read_mostly;
+unsigned long sparc64_kern_sec_context __read_mostly;
 
 int bigkernel = 0;
 
@@ -179,8 +243,6 @@ static __inline__ void clear_dcache_dirty_cpu(struct page *page, unsigned long c
 			     : "g1", "g7");
 }
 
-extern void __update_mmu_cache(unsigned long mmu_context_hw, unsigned long address, pte_t pte, int code);
-
 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t pte)
 {
 	struct page *page;
@@ -207,10 +269,6 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t p
 
 		put_cpu();
 	}
-
-	if (get_thread_fault_code())
-		__update_mmu_cache(CTX_NRBITS(vma->vm_mm->context),
-				   address, pte, get_thread_fault_code());
 }
 
 void flush_dcache_page(struct page *page)
@@ -310,6 +368,11 @@ struct linux_prom_translation {
 	unsigned long data;
 };
 
+/* Exported for kernel TLB miss handling in ktlb.S */
+struct linux_prom_translation prom_trans[512] __read_mostly;
+unsigned int prom_trans_ents __read_mostly;
+unsigned int swapper_pgd_zero __read_mostly;
+
 extern unsigned long prom_boot_page;
 extern void prom_remap(unsigned long physpage, unsigned long virtpage, int mmu_ihandle);
 extern int prom_get_mmu_ihandle(void);
@@ -318,297 +381,162 @@ extern void register_prom_callbacks(void);
 /* Exported for SMP bootup purposes. */
 unsigned long kern_locked_tte_data;
 
-void __init early_pgtable_allocfail(char *type)
-{
-	prom_printf("inherit_prom_mappings: Cannot alloc kernel %s.\n", type);
-	prom_halt();
-}
-
-#define BASE_PAGE_SIZE 8192
-static pmd_t *prompmd;
-
 /*
  * Translate PROM's mapping we capture at boot time into physical address.
  * The second parameter is only set from prom_callback() invocations.
  */
 unsigned long prom_virt_to_phys(unsigned long promva, int *error)
 {
-	pmd_t *pmdp = prompmd + ((promva >> 23) & 0x7ff);
-	pte_t *ptep;
-	unsigned long base;
-
-	if (pmd_none(*pmdp)) {
-		if (error)
-			*error = 1;
-		return(0);
-	}
-	ptep = (pte_t *)__pmd_page(*pmdp) + ((promva >> 13) & 0x3ff);
-	if (!pte_present(*ptep)) {
-		if (error)
-			*error = 1;
-		return(0);
-	}
-	if (error) {
-		*error = 0;
-		return(pte_val(*ptep));
+	int i;
+
+	for (i = 0; i < prom_trans_ents; i++) {
+		struct linux_prom_translation *p = &prom_trans[i];
+
+		if (promva >= p->virt &&
+		    promva < (p->virt + p->size)) {
+			unsigned long base = p->data & _PAGE_PADDR;
+
+			if (error)
+				*error = 0;
+			return base + (promva & (8192 - 1));
+		}
 	}
-	base = pte_val(*ptep) & _PAGE_PADDR;
-	return(base + (promva & (BASE_PAGE_SIZE - 1)));
+	if (error)
+		*error = 1;
+	return 0UL;
 }
 
-static void inherit_prom_mappings(void)
+/* The obp translations are saved based on 8k pagesize, since obp can
+ * use a mixture of pagesizes. Misses to the LOW_OBP_ADDRESS ->
+ * HI_OBP_ADDRESS range are handled in ktlb.S and do not use the vpte
+ * scheme (also, see rant in inherit_locked_prom_mappings()).
+ */
+static inline int in_obp_range(unsigned long vaddr)
 {
-	struct linux_prom_translation *trans;
-	unsigned long phys_page, tte_vaddr, tte_data;
-	void (*remap_func)(unsigned long, unsigned long, int);
-	pmd_t *pmdp;
-	pte_t *ptep;
-	int node, n, i, tsz;
-	extern unsigned int obp_iaddr_patch[2], obp_daddr_patch[2];
+	return (vaddr >= LOW_OBP_ADDRESS &&
+		vaddr < HI_OBP_ADDRESS);
+}
+
+static int cmp_ptrans(const void *a, const void *b)
+{
+	const struct linux_prom_translation *x = a, *y = b;
+
+	if (x->virt > y->virt)
+		return 1;
+	if (x->virt < y->virt)
+		return -1;
+	return 0;
+}
+
+/* Read OBP translations property into 'prom_trans[]'.  */
+static void __init read_obp_translations(void)
+{
+	int n, node, ents, first, last, i;
 
 	node = prom_finddevice("/virtual-memory");
 	n = prom_getproplen(node, "translations");
-	if (n == 0 || n == -1) {
-		prom_printf("Couldn't get translation property\n");
+	if (unlikely(n == 0 || n == -1)) {
+		prom_printf("prom_mappings: Couldn't get size.\n");
 		prom_halt();
 	}
-	n += 5 * sizeof(struct linux_prom_translation);
-	for (tsz = 1; tsz < n; tsz <<= 1)
-		/* empty */;
-	trans = __alloc_bootmem(tsz, SMP_CACHE_BYTES, bootmap_base);
-	if (trans == NULL) {
-		prom_printf("inherit_prom_mappings: Cannot alloc translations.\n");
+	if (unlikely(n > sizeof(prom_trans))) {
+		prom_printf("prom_mappings: Size %Zd is too big.\n", n);
 		prom_halt();
 	}
-	memset(trans, 0, tsz);
 
-	if ((n = prom_getproperty(node, "translations", (char *)trans, tsz)) == -1) {
-		prom_printf("Couldn't get translation property\n");
+	if ((n = prom_getproperty(node, "translations",
+				  (char *)&prom_trans[0],
+				  sizeof(prom_trans))) == -1) {
+		prom_printf("prom_mappings: Couldn't get property.\n");
 		prom_halt();
 	}
-	n = n / sizeof(*trans);
 
-	/*
-	 * The obp translations are saved based on 8k pagesize, since obp can
-	 * use a mixture of pagesizes. Misses to the 0xf0000000 - 0x100000000,
-	 * ie obp range, are handled in entry.S and do not use the vpte scheme
-	 * (see rant in inherit_locked_prom_mappings()).
-	 */
-#define OBP_PMD_SIZE 2048
-	prompmd = __alloc_bootmem(OBP_PMD_SIZE, OBP_PMD_SIZE, bootmap_base);
-	if (prompmd == NULL)
-		early_pgtable_allocfail("pmd");
-	memset(prompmd, 0, OBP_PMD_SIZE);
-	for (i = 0; i < n; i++) {
-		unsigned long vaddr;
-
-		if (trans[i].virt >= LOW_OBP_ADDRESS && trans[i].virt < HI_OBP_ADDRESS) {
-			for (vaddr = trans[i].virt;
-			     ((vaddr < trans[i].virt + trans[i].size) && 
-			     (vaddr < HI_OBP_ADDRESS));
-			     vaddr += BASE_PAGE_SIZE) {
-				unsigned long val;
-
-				pmdp = prompmd + ((vaddr >> 23) & 0x7ff);
-				if (pmd_none(*pmdp)) {
-					ptep = __alloc_bootmem(BASE_PAGE_SIZE,
-							       BASE_PAGE_SIZE,
-							       bootmap_base);
-					if (ptep == NULL)
-						early_pgtable_allocfail("pte");
-					memset(ptep, 0, BASE_PAGE_SIZE);
-					pmd_set(pmdp, ptep);
-				}
-				ptep = (pte_t *)__pmd_page(*pmdp) +
-						((vaddr >> 13) & 0x3ff);
+	n = n / sizeof(struct linux_prom_translation);
 
-				val = trans[i].data;
+	ents = n;
 
-				/* Clear diag TTE bits. */
-				if (tlb_type == spitfire)
-					val &= ~0x0003fe0000000000UL;
+	sort(prom_trans, ents, sizeof(struct linux_prom_translation),
+	     cmp_ptrans, NULL);
 
-				set_pte_at(&init_mm, vaddr,
-					   ptep, __pte(val | _PAGE_MODIFIED));
-				trans[i].data += BASE_PAGE_SIZE;
-			}
-		}
+	/* Now kick out all the non-OBP entries.  */
+	for (i = 0; i < ents; i++) {
+		if (in_obp_range(prom_trans[i].virt))
+			break;
+	}
+	first = i;
+	for (; i < ents; i++) {
+		if (!in_obp_range(prom_trans[i].virt))
+			break;
 	}
-	phys_page = __pa(prompmd);
-	obp_iaddr_patch[0] |= (phys_page >> 10);
-	obp_iaddr_patch[1] |= (phys_page & 0x3ff);
-	flushi((long)&obp_iaddr_patch[0]);
-	obp_daddr_patch[0] |= (phys_page >> 10);
-	obp_daddr_patch[1] |= (phys_page & 0x3ff);
-	flushi((long)&obp_daddr_patch[0]);
+	last = i;
 
-	/* Now fixup OBP's idea about where we really are mapped. */
-	prom_printf("Remapping the kernel... ");
+	for (i = 0; i < (last - first); i++) {
+		struct linux_prom_translation *src = &prom_trans[i + first];
+		struct linux_prom_translation *dest = &prom_trans[i];
 
-	/* Spitfire Errata #32 workaround */
-	/* NOTE: Using plain zero for the context value is
-	 *       correct here, we are not using the Linux trap
-	 *       tables yet so we should not use the special
-	 *       UltraSPARC-III+ page size encodings yet.
-	 */
-	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
-			     "flush	%%g6"
-			     : /* No outputs */
-			     : "r" (0), "r" (PRIMARY_CONTEXT), "i" (ASI_DMMU));
-
-	switch (tlb_type) {
-	default:
-	case spitfire:
-		phys_page = spitfire_get_dtlb_data(sparc64_highest_locked_tlbent());
-		break;
-
-	case cheetah:
-	case cheetah_plus:
-		phys_page = cheetah_get_litlb_data(sparc64_highest_locked_tlbent());
-		break;
-	};
-
-	phys_page &= _PAGE_PADDR;
-	phys_page += ((unsigned long)&prom_boot_page -
-		      (unsigned long)KERNBASE);
+		*dest = *src;
+	}
+	for (; i < ents; i++) {
+		struct linux_prom_translation *dest = &prom_trans[i];
+		dest->virt = dest->size = dest->data = 0x0UL;
+	}
+
+	prom_trans_ents = last - first;
 
 	if (tlb_type == spitfire) {
-		/* Lock this into i/d tlb entry 59 */
-		__asm__ __volatile__(
-			"stxa	%%g0, [%2] %3\n\t"
-			"stxa	%0, [%1] %4\n\t"
-			"membar	#Sync\n\t"
-			"flush	%%g6\n\t"
-			"stxa	%%g0, [%2] %5\n\t"
-			"stxa	%0, [%1] %6\n\t"
-			"membar	#Sync\n\t"
-			"flush	%%g6"
-			: : "r" (phys_page | _PAGE_VALID | _PAGE_SZ8K | _PAGE_CP |
-				 _PAGE_CV | _PAGE_P | _PAGE_L | _PAGE_W),
-			"r" (59 << 3), "r" (TLB_TAG_ACCESS),
-			"i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS),
-			"i" (ASI_IMMU), "i" (ASI_ITLB_DATA_ACCESS)
-			: "memory");
-	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
-		/* Lock this into i/d tlb-0 entry 11 */
-		__asm__ __volatile__(
-			"stxa	%%g0, [%2] %3\n\t"
-			"stxa	%0, [%1] %4\n\t"
-			"membar	#Sync\n\t"
-			"flush	%%g6\n\t"
-			"stxa	%%g0, [%2] %5\n\t"
-			"stxa	%0, [%1] %6\n\t"
-			"membar	#Sync\n\t"
-			"flush	%%g6"
-			: : "r" (phys_page | _PAGE_VALID | _PAGE_SZ8K | _PAGE_CP |
-				 _PAGE_CV | _PAGE_P | _PAGE_L | _PAGE_W),
-			"r" ((0 << 16) | (11 << 3)), "r" (TLB_TAG_ACCESS),
-			"i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS),
-			"i" (ASI_IMMU), "i" (ASI_ITLB_DATA_ACCESS)
-			: "memory");
-	} else {
-		/* Implement me :-) */
-		BUG();
+		/* Clear diag TTE bits. */
+		for (i = 0; i < prom_trans_ents; i++)
+			prom_trans[i].data &= ~0x0003fe0000000000UL;
 	}
+}
 
-	tte_vaddr = (unsigned long) KERNBASE;
+static void __init remap_kernel(void)
+{
+	unsigned long phys_page, tte_vaddr, tte_data;
+	int tlb_ent = sparc64_highest_locked_tlbent();
 
-	/* Spitfire Errata #32 workaround */
-	/* NOTE: Using plain zero for the context value is
-	 *       correct here, we are not using the Linux trap
-	 *       tables yet so we should not use the special
-	 *       UltraSPARC-III+ page size encodings yet.
-	 */
-	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
-			     "flush	%%g6"
-			     : /* No outputs */
-			     : "r" (0),
-			     "r" (PRIMARY_CONTEXT), "i" (ASI_DMMU));
-
-	if (tlb_type == spitfire)
-		tte_data = spitfire_get_dtlb_data(sparc64_highest_locked_tlbent());
-	else
-		tte_data = cheetah_get_ldtlb_data(sparc64_highest_locked_tlbent());
+	tte_vaddr = (unsigned long) KERNBASE;
+	phys_page = (prom_boot_mapping_phys_low >> 22UL) << 22UL;
+	tte_data = (phys_page | (_PAGE_VALID | _PAGE_SZ4MB |
+				 _PAGE_CP | _PAGE_CV | _PAGE_P |
+				 _PAGE_L | _PAGE_W));
 
 	kern_locked_tte_data = tte_data;
 
-	remap_func = (void *)  ((unsigned long) &prom_remap -
-				(unsigned long) &prom_boot_page);
-
-
-	/* Spitfire Errata #32 workaround */
-	/* NOTE: Using plain zero for the context value is
-	 *       correct here, we are not using the Linux trap
-	 *       tables yet so we should not use the special
-	 *       UltraSPARC-III+ page size encodings yet.
-	 */
-	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
-			     "flush	%%g6"
-			     : /* No outputs */
-			     : "r" (0),
-			     "r" (PRIMARY_CONTEXT), "i" (ASI_DMMU));
-
-	remap_func((tlb_type == spitfire ?
-		    (spitfire_get_dtlb_data(sparc64_highest_locked_tlbent()) & _PAGE_PADDR) :
-		    (cheetah_get_litlb_data(sparc64_highest_locked_tlbent()) & _PAGE_PADDR)),
-		   (unsigned long) KERNBASE,
-		   prom_get_mmu_ihandle());
-
-	if (bigkernel)
-		remap_func(((tte_data + 0x400000) & _PAGE_PADDR),
-			(unsigned long) KERNBASE + 0x400000, prom_get_mmu_ihandle());
-
-	/* Flush out that temporary mapping. */
-	spitfire_flush_dtlb_nucleus_page(0x0);
-	spitfire_flush_itlb_nucleus_page(0x0);
-
-	/* Now lock us back into the TLBs via OBP. */
-	prom_dtlb_load(sparc64_highest_locked_tlbent(), tte_data, tte_vaddr);
-	prom_itlb_load(sparc64_highest_locked_tlbent(), tte_data, tte_vaddr);
+	/* Now lock us into the TLBs via OBP. */
+	prom_dtlb_load(tlb_ent, tte_data, tte_vaddr);
+	prom_itlb_load(tlb_ent, tte_data, tte_vaddr);
 	if (bigkernel) {
-		prom_dtlb_load(sparc64_highest_locked_tlbent()-1, tte_data + 0x400000, 
-								tte_vaddr + 0x400000);
-		prom_itlb_load(sparc64_highest_locked_tlbent()-1, tte_data + 0x400000, 
-								tte_vaddr + 0x400000);
+		tlb_ent -= 1;
+		prom_dtlb_load(tlb_ent,
+			       tte_data + 0x400000, 
+			       tte_vaddr + 0x400000);
+		prom_itlb_load(tlb_ent,
+			       tte_data + 0x400000, 
+			       tte_vaddr + 0x400000);
 	}
-
-	/* Re-read translations property. */
-	if ((n = prom_getproperty(node, "translations", (char *)trans, tsz)) == -1) {
-		prom_printf("Couldn't get translation property\n");
-		prom_halt();
+	sparc64_highest_unlocked_tlb_ent = tlb_ent - 1;
+	if (tlb_type == cheetah_plus) {
+		sparc64_kern_pri_context = (CTX_CHEETAH_PLUS_CTX0 |
+					    CTX_CHEETAH_PLUS_NUC);
+		sparc64_kern_pri_nuc_bits = CTX_CHEETAH_PLUS_NUC;
+		sparc64_kern_sec_context = CTX_CHEETAH_PLUS_CTX0;
 	}
-	n = n / sizeof(*trans);
-
-	for (i = 0; i < n; i++) {
-		unsigned long vaddr = trans[i].virt;
-		unsigned long size = trans[i].size;
-
-		if (vaddr < 0xf0000000UL) {
-			unsigned long avoid_start = (unsigned long) KERNBASE;
-			unsigned long avoid_end = avoid_start + (4 * 1024 * 1024);
-
-			if (bigkernel)
-				avoid_end += (4 * 1024 * 1024);
-			if (vaddr < avoid_start) {
-				unsigned long top = vaddr + size;
+}
 
-				if (top > avoid_start)
-					top = avoid_start;
-				prom_unmap(top - vaddr, vaddr);
-			}
-			if ((vaddr + size) > avoid_end) {
-				unsigned long bottom = vaddr;
 
-				if (bottom < avoid_end)
-					bottom = avoid_end;
-				prom_unmap((vaddr + size) - bottom, bottom);
-			}
-		}
-	}
+static void __init inherit_prom_mappings(void)
+{
+	read_obp_translations();
 
+	/* Now fixup OBP's idea about where we really are mapped. */
+	prom_printf("Remapping the kernel... ");
+	remap_kernel();
 	prom_printf("done.\n");
 
+	prom_printf("Registering callbacks... ");
 	register_prom_callbacks();
+	prom_printf("done.\n");
 }
 
 /* The OBP specifications for sun4u mark 0xfffffffc00000000 and
@@ -792,8 +720,8 @@ void inherit_locked_prom_mappings(int save_p)
 		}
 	}
 	if (tlb_type == spitfire) {
-		int high = SPITFIRE_HIGHEST_LOCKED_TLBENT - bigkernel;
-		for (i = 0; i < high; i++) {
+		int high = sparc64_highest_unlocked_tlb_ent;
+		for (i = 0; i <= high; i++) {
 			unsigned long data;
 
 			/* Spitfire Errata #32 workaround */
@@ -881,9 +809,9 @@ void inherit_locked_prom_mappings(int save_p)
 			}
 		}
 	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
-		int high = CHEETAH_HIGHEST_LOCKED_TLBENT - bigkernel;
+		int high = sparc64_highest_unlocked_tlb_ent;
 
-		for (i = 0; i < high; i++) {
+		for (i = 0; i <= high; i++) {
 			unsigned long data;
 
 			data = cheetah_get_ldtlb_data(i);
@@ -1276,14 +1204,14 @@ unsigned long __init bootmem_init(unsigned long *pages_avail)
 	int i;
 
 #ifdef CONFIG_DEBUG_BOOTMEM
-	prom_printf("bootmem_init: Scan sp_banks, ");
+	prom_printf("bootmem_init: Scan pavail, ");
 #endif
 
 	bytes_avail = 0UL;
-	for (i = 0; sp_banks[i].num_bytes != 0; i++) {
-		end_of_phys_memory = sp_banks[i].base_addr +
-			sp_banks[i].num_bytes;
-		bytes_avail += sp_banks[i].num_bytes;
+	for (i = 0; i < pavail_ents; i++) {
+		end_of_phys_memory = pavail[i].phys_addr +
+			pavail[i].reg_size;
+		bytes_avail += pavail[i].reg_size;
 		if (cmdline_memory_size) {
 			if (bytes_avail > cmdline_memory_size) {
 				unsigned long slack = bytes_avail - cmdline_memory_size;
@@ -1291,12 +1219,15 @@ unsigned long __init bootmem_init(unsigned long *pages_avail)
 				bytes_avail -= slack;
 				end_of_phys_memory -= slack;
 
-				sp_banks[i].num_bytes -= slack;
-				if (sp_banks[i].num_bytes == 0) {
-					sp_banks[i].base_addr = 0xdeadbeef;
+				pavail[i].reg_size -= slack;
+				if ((long)pavail[i].reg_size <= 0L) {
+					pavail[i].phys_addr = 0xdeadbeefUL;
+					pavail[i].reg_size = 0UL;
+					pavail_ents = i;
 				} else {
-					sp_banks[i+1].num_bytes = 0;
-					sp_banks[i+1].base_addr = 0xdeadbeef;
+					pavail[i+1].reg_size = 0Ul;
+					pavail[i+1].phys_addr = 0xdeadbeefUL;
+					pavail_ents = i + 1;
 				}
 				break;
 			}
@@ -1347,17 +1278,15 @@ unsigned long __init bootmem_init(unsigned long *pages_avail)
 #endif
 	bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap_pfn, pfn_base, end_pfn);
 
-	bootmap_base = bootmap_pfn << PAGE_SHIFT;
-
 	/* Now register the available physical memory with the
 	 * allocator.
 	 */
-	for (i = 0; sp_banks[i].num_bytes != 0; i++) {
+	for (i = 0; i < pavail_ents; i++) {
 #ifdef CONFIG_DEBUG_BOOTMEM
-		prom_printf("free_bootmem(sp_banks:%d): base[%lx] size[%lx]\n",
-			    i, sp_banks[i].base_addr, sp_banks[i].num_bytes);
+		prom_printf("free_bootmem(pavail:%d): base[%lx] size[%lx]\n",
+			    i, pavail[i].phys_addr, pavail[i].reg_size);
 #endif
-		free_bootmem(sp_banks[i].base_addr, sp_banks[i].num_bytes);
+		free_bootmem(pavail[i].phys_addr, pavail[i].reg_size);
 	}
 
 #ifdef CONFIG_BLK_DEV_INITRD
@@ -1398,121 +1327,167 @@ unsigned long __init bootmem_init(unsigned long *pages_avail)
 	return end_pfn;
 }
 
+#ifdef CONFIG_DEBUG_PAGEALLOC
+static unsigned long kernel_map_range(unsigned long pstart, unsigned long pend, pgprot_t prot)
+{
+	unsigned long vstart = PAGE_OFFSET + pstart;
+	unsigned long vend = PAGE_OFFSET + pend;
+	unsigned long alloc_bytes = 0UL;
+
+	if ((vstart & ~PAGE_MASK) || (vend & ~PAGE_MASK)) {
+		prom_printf("kernel_map: Unaligned physmem[%lx:%lx]\n",
+			    vstart, vend);
+		prom_halt();
+	}
+
+	while (vstart < vend) {
+		unsigned long this_end, paddr = __pa(vstart);
+		pgd_t *pgd = pgd_offset_k(vstart);
+		pud_t *pud;
+		pmd_t *pmd;
+		pte_t *pte;
+
+		pud = pud_offset(pgd, vstart);
+		if (pud_none(*pud)) {
+			pmd_t *new;
+
+			new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
+			alloc_bytes += PAGE_SIZE;
+			pud_populate(&init_mm, pud, new);
+		}
+
+		pmd = pmd_offset(pud, vstart);
+		if (!pmd_present(*pmd)) {
+			pte_t *new;
+
+			new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
+			alloc_bytes += PAGE_SIZE;
+			pmd_populate_kernel(&init_mm, pmd, new);
+		}
+
+		pte = pte_offset_kernel(pmd, vstart);
+		this_end = (vstart + PMD_SIZE) & PMD_MASK;
+		if (this_end > vend)
+			this_end = vend;
+
+		while (vstart < this_end) {
+			pte_val(*pte) = (paddr | pgprot_val(prot));
+
+			vstart += PAGE_SIZE;
+			paddr += PAGE_SIZE;
+			pte++;
+		}
+	}
+
+	return alloc_bytes;
+}
+
+static struct linux_prom64_registers pall[MAX_BANKS] __initdata;
+static int pall_ents __initdata;
+
+extern unsigned int kvmap_linear_patch[1];
+
+static void __init kernel_physical_mapping_init(void)
+{
+	unsigned long i, mem_alloced = 0UL;
+
+	read_obp_memory("reg", &pall[0], &pall_ents);
+
+	for (i = 0; i < pall_ents; i++) {
+		unsigned long phys_start, phys_end;
+
+		phys_start = pall[i].phys_addr;
+		phys_end = phys_start + pall[i].reg_size;
+		mem_alloced += kernel_map_range(phys_start, phys_end,
+						PAGE_KERNEL);
+	}
+
+	printk("Allocated %ld bytes for kernel page tables.\n",
+	       mem_alloced);
+
+	kvmap_linear_patch[0] = 0x01000000; /* nop */
+	flushi(&kvmap_linear_patch[0]);
+
+	__flush_tlb_all();
+}
+
+void kernel_map_pages(struct page *page, int numpages, int enable)
+{
+	unsigned long phys_start = page_to_pfn(page) << PAGE_SHIFT;
+	unsigned long phys_end = phys_start + (numpages * PAGE_SIZE);
+
+	kernel_map_range(phys_start, phys_end,
+			 (enable ? PAGE_KERNEL : __pgprot(0)));
+
+	/* we should perform an IPI and flush all tlbs,
+	 * but that can deadlock->flush only current cpu.
+	 */
+	__flush_tlb_kernel_range(PAGE_OFFSET + phys_start,
+				 PAGE_OFFSET + phys_end);
+}
+#endif
+
+unsigned long __init find_ecache_flush_span(unsigned long size)
+{
+	int i;
+
+	for (i = 0; i < pavail_ents; i++) {
+		if (pavail[i].reg_size >= size)
+			return pavail[i].phys_addr;
+	}
+
+	return ~0UL;
+}
+
 /* paging_init() sets up the page tables */
 
 extern void cheetah_ecache_flush_init(void);
 
 static unsigned long last_valid_pfn;
+pgd_t swapper_pg_dir[2048];
 
 void __init paging_init(void)
 {
-	extern pmd_t swapper_pmd_dir[1024];
-	extern unsigned int sparc64_vpte_patchme1[1];
-	extern unsigned int sparc64_vpte_patchme2[1];
-	unsigned long alias_base = kern_base + PAGE_OFFSET;
-	unsigned long second_alias_page = 0;
-	unsigned long pt, flags, end_pfn, pages_avail;
-	unsigned long shift = alias_base - ((unsigned long)KERNBASE);
-	unsigned long real_end;
+	unsigned long end_pfn, pages_avail, shift;
+	unsigned long real_end, i;
+
+	/* Find available physical memory... */
+	read_obp_memory("available", &pavail[0], &pavail_ents);
+
+	phys_base = 0xffffffffffffffffUL;
+	for (i = 0; i < pavail_ents; i++)
+		phys_base = min(phys_base, pavail[i].phys_addr);
+
+	pfn_base = phys_base >> PAGE_SHIFT;
+
+	kern_base = (prom_boot_mapping_phys_low >> 22UL) << 22UL;
+	kern_size = (unsigned long)&_end - (unsigned long)KERNBASE;
 
 	set_bit(0, mmu_context_bmap);
 
+	shift = kern_base + PAGE_OFFSET - ((unsigned long)KERNBASE);
+
 	real_end = (unsigned long)_end;
 	if ((real_end > ((unsigned long)KERNBASE + 0x400000)))
 		bigkernel = 1;
-#ifdef CONFIG_BLK_DEV_INITRD
-	if (sparc_ramdisk_image || sparc_ramdisk_image64)
-		real_end = (PAGE_ALIGN(real_end) + PAGE_ALIGN(sparc_ramdisk_size));
-#endif
-
-	/* We assume physical memory starts at some 4mb multiple,
-	 * if this were not true we wouldn't boot up to this point
-	 * anyways.
-	 */
-	pt  = kern_base | _PAGE_VALID | _PAGE_SZ4MB;
-	pt |= _PAGE_CP | _PAGE_CV | _PAGE_P | _PAGE_L | _PAGE_W;
-	local_irq_save(flags);
-	if (tlb_type == spitfire) {
-		__asm__ __volatile__(
-	"	stxa	%1, [%0] %3\n"
-	"	stxa	%2, [%5] %4\n"
-	"	membar	#Sync\n"
-	"	flush	%%g6\n"
-	"	nop\n"
-	"	nop\n"
-	"	nop\n"
-		: /* No outputs */
-		: "r" (TLB_TAG_ACCESS), "r" (alias_base), "r" (pt),
-		  "i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS), "r" (61 << 3)
-		: "memory");
-		if (real_end >= KERNBASE + 0x340000) {
-			second_alias_page = alias_base + 0x400000;
-			__asm__ __volatile__(
-		"	stxa	%1, [%0] %3\n"
-		"	stxa	%2, [%5] %4\n"
-		"	membar	#Sync\n"
-		"	flush	%%g6\n"
-		"	nop\n"
-		"	nop\n"
-		"	nop\n"
-			: /* No outputs */
-			: "r" (TLB_TAG_ACCESS), "r" (second_alias_page), "r" (pt + 0x400000),
-			  "i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS), "r" (60 << 3)
-			: "memory");
-		}
-	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
-		__asm__ __volatile__(
-	"	stxa	%1, [%0] %3\n"
-	"	stxa	%2, [%5] %4\n"
-	"	membar	#Sync\n"
-	"	flush	%%g6\n"
-	"	nop\n"
-	"	nop\n"
-	"	nop\n"
-		: /* No outputs */
-		: "r" (TLB_TAG_ACCESS), "r" (alias_base), "r" (pt),
-		  "i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS), "r" ((0<<16) | (13<<3))
-		: "memory");
-		if (real_end >= KERNBASE + 0x340000) {
-			second_alias_page = alias_base + 0x400000;
-			__asm__ __volatile__(
-		"	stxa	%1, [%0] %3\n"
-		"	stxa	%2, [%5] %4\n"
-		"	membar	#Sync\n"
-		"	flush	%%g6\n"
-		"	nop\n"
-		"	nop\n"
-		"	nop\n"
-			: /* No outputs */
-			: "r" (TLB_TAG_ACCESS), "r" (second_alias_page), "r" (pt + 0x400000),
-			  "i" (ASI_DMMU), "i" (ASI_DTLB_DATA_ACCESS), "r" ((0<<16) | (12<<3))
-			: "memory");
-		}
+	if ((real_end > ((unsigned long)KERNBASE + 0x800000))) {
+		prom_printf("paging_init: Kernel > 8MB, too large.\n");
+		prom_halt();
 	}
-	local_irq_restore(flags);
-	
-	/* Now set kernel pgd to upper alias so physical page computations
+
+	/* Set kernel pgd to upper alias so physical page computations
 	 * work.
 	 */
 	init_mm.pgd += ((shift) / (sizeof(pgd_t)));
 	
-	memset(swapper_pmd_dir, 0, sizeof(swapper_pmd_dir));
+	memset(swapper_low_pmd_dir, 0, sizeof(swapper_low_pmd_dir));
 
 	/* Now can init the kernel/bad page tables. */
 	pud_set(pud_offset(&swapper_pg_dir[0], 0),
-		swapper_pmd_dir + (shift / sizeof(pgd_t)));
+		swapper_low_pmd_dir + (shift / sizeof(pgd_t)));
 	
-	sparc64_vpte_patchme1[0] |=
-		(((unsigned long)pgd_val(init_mm.pgd[0])) >> 10);
-	sparc64_vpte_patchme2[0] |=
-		(((unsigned long)pgd_val(init_mm.pgd[0])) & 0x3ff);
-	flushi((long)&sparc64_vpte_patchme1[0]);
+	swapper_pgd_zero = pgd_val(swapper_pg_dir[0]);
 	
-	/* Setup bootmem... */
-	pages_avail = 0;
-	last_valid_pfn = end_pfn = bootmem_init(&pages_avail);
-
-	/* Inherit non-locked OBP mappings. */
 	inherit_prom_mappings();
 	
 	/* Ok, we can use our TLB miss and window trap handlers safely.
@@ -1527,13 +1502,16 @@ void __init paging_init(void)
 
 	inherit_locked_prom_mappings(1);
 
-	/* We only created DTLB mapping of this stuff. */
-	spitfire_flush_dtlb_nucleus_page(alias_base);
-	if (second_alias_page)
-		spitfire_flush_dtlb_nucleus_page(second_alias_page);
-
 	__flush_tlb_all();
 
+	/* Setup bootmem... */
+	pages_avail = 0;
+	last_valid_pfn = end_pfn = bootmem_init(&pages_avail);
+
+#ifdef CONFIG_DEBUG_PAGEALLOC
+	kernel_physical_mapping_init();
+#endif
+
 	{
 		unsigned long zones_size[MAX_NR_ZONES];
 		unsigned long zholes_size[MAX_NR_ZONES];
@@ -1554,128 +1532,35 @@ void __init paging_init(void)
 	device_scan();
 }
 
-/* Ok, it seems that the prom can allocate some more memory chunks
- * as a side effect of some prom calls we perform during the
- * boot sequence.  My most likely theory is that it is from the
- * prom_set_traptable() call, and OBP is allocating a scratchpad
- * for saving client program register state etc.
- */
-static void __init sort_memlist(struct linux_mlist_p1275 *thislist)
-{
-	int swapi = 0;
-	int i, mitr;
-	unsigned long tmpaddr, tmpsize;
-	unsigned long lowest;
-
-	for (i = 0; thislist[i].theres_more != 0; i++) {
-		lowest = thislist[i].start_adr;
-		for (mitr = i+1; thislist[mitr-1].theres_more != 0; mitr++)
-			if (thislist[mitr].start_adr < lowest) {
-				lowest = thislist[mitr].start_adr;
-				swapi = mitr;
-			}
-		if (lowest == thislist[i].start_adr)
-			continue;
-		tmpaddr = thislist[swapi].start_adr;
-		tmpsize = thislist[swapi].num_bytes;
-		for (mitr = swapi; mitr > i; mitr--) {
-			thislist[mitr].start_adr = thislist[mitr-1].start_adr;
-			thislist[mitr].num_bytes = thislist[mitr-1].num_bytes;
-		}
-		thislist[i].start_adr = tmpaddr;
-		thislist[i].num_bytes = tmpsize;
-	}
-}
-
-void __init rescan_sp_banks(void)
-{
-	struct linux_prom64_registers memlist[64];
-	struct linux_mlist_p1275 avail[64], *mlist;
-	unsigned long bytes, base_paddr;
-	int num_regs, node = prom_finddevice("/memory");
-	int i;
-
-	num_regs = prom_getproperty(node, "available",
-				    (char *) memlist, sizeof(memlist));
-	num_regs = (num_regs / sizeof(struct linux_prom64_registers));
-	for (i = 0; i < num_regs; i++) {
-		avail[i].start_adr = memlist[i].phys_addr;
-		avail[i].num_bytes = memlist[i].reg_size;
-		avail[i].theres_more = &avail[i + 1];
-	}
-	avail[i - 1].theres_more = NULL;
-	sort_memlist(avail);
-
-	mlist = &avail[0];
-	i = 0;
-	bytes = mlist->num_bytes;
-	base_paddr = mlist->start_adr;
-  
-	sp_banks[0].base_addr = base_paddr;
-	sp_banks[0].num_bytes = bytes;
-
-	while (mlist->theres_more != NULL){
-		i++;
-		mlist = mlist->theres_more;
-		bytes = mlist->num_bytes;
-		if (i >= SPARC_PHYS_BANKS-1) {
-			printk ("The machine has more banks than "
-				"this kernel can support\n"
-				"Increase the SPARC_PHYS_BANKS "
-				"setting (currently %d)\n",
-				SPARC_PHYS_BANKS);
-			i = SPARC_PHYS_BANKS-1;
-			break;
-		}
-    
-		sp_banks[i].base_addr = mlist->start_adr;
-		sp_banks[i].num_bytes = mlist->num_bytes;
-	}
-
-	i++;
-	sp_banks[i].base_addr = 0xdeadbeefbeefdeadUL;
-	sp_banks[i].num_bytes = 0;
-
-	for (i = 0; sp_banks[i].num_bytes != 0; i++)
-		sp_banks[i].num_bytes &= PAGE_MASK;
-}
-
 static void __init taint_real_pages(void)
 {
-	struct sparc_phys_banks saved_sp_banks[SPARC_PHYS_BANKS];
 	int i;
 
-	for (i = 0; i < SPARC_PHYS_BANKS; i++) {
-		saved_sp_banks[i].base_addr =
-			sp_banks[i].base_addr;
-		saved_sp_banks[i].num_bytes =
-			sp_banks[i].num_bytes;
-	}
-
-	rescan_sp_banks();
+	read_obp_memory("available", &pavail_rescan[0], &pavail_rescan_ents);
 
-	/* Find changes discovered in the sp_bank rescan and
+	/* Find changes discovered in the physmem available rescan and
 	 * reserve the lost portions in the bootmem maps.
 	 */
-	for (i = 0; saved_sp_banks[i].num_bytes; i++) {
+	for (i = 0; i < pavail_ents; i++) {
 		unsigned long old_start, old_end;
 
-		old_start = saved_sp_banks[i].base_addr;
+		old_start = pavail[i].phys_addr;
 		old_end = old_start +
-			saved_sp_banks[i].num_bytes;
+			pavail[i].reg_size;
 		while (old_start < old_end) {
 			int n;
 
-			for (n = 0; sp_banks[n].num_bytes; n++) {
+			for (n = 0; pavail_rescan_ents; n++) {
 				unsigned long new_start, new_end;
 
-				new_start = sp_banks[n].base_addr;
-				new_end = new_start + sp_banks[n].num_bytes;
+				new_start = pavail_rescan[n].phys_addr;
+				new_end = new_start +
+					pavail_rescan[n].reg_size;
 
 				if (new_start <= old_start &&
 				    new_end >= (old_start + PAGE_SIZE)) {
-					set_bit (old_start >> 22,
-						 sparc64_valid_addr_bitmap);
+					set_bit(old_start >> 22,
+						sparc64_valid_addr_bitmap);
 					goto do_next_page;
 				}
 			}
@@ -1695,8 +1580,7 @@ void __init mem_init(void)
 
 	i = last_valid_pfn >> ((22 - PAGE_SHIFT) + 6);
 	i += 1;
-	sparc64_valid_addr_bitmap = (unsigned long *)
-		__alloc_bootmem(i << 3, SMP_CACHE_BYTES, bootmap_base);
+	sparc64_valid_addr_bitmap = (unsigned long *) alloc_bootmem(i << 3);
 	if (sparc64_valid_addr_bitmap == NULL) {
 		prom_printf("mem_init: Cannot alloc valid_addr_bitmap.\n");
 		prom_halt();
@@ -1749,7 +1633,7 @@ void __init mem_init(void)
 		cheetah_ecache_flush_init();
 }
 
-void free_initmem (void)
+void free_initmem(void)
 {
 	unsigned long addr, initend;
 
diff --git a/arch/sparc64/mm/ultra.S b/arch/sparc64/mm/ultra.S
index b2ee9b53227..e4c9151fa11 100644
--- a/arch/sparc64/mm/ultra.S
+++ b/arch/sparc64/mm/ultra.S
@@ -144,42 +144,29 @@ __flush_icache_page:	/* %o0 = phys_page */
 
 #define DTAG_MASK 0x3
 
+	/* This routine is Spitfire specific so the hardcoded
+	 * D-cache size and line-size are OK.
+	 */
 	.align		64
 	.globl		__flush_dcache_page
 __flush_dcache_page:	/* %o0=kaddr, %o1=flush_icache */
 	sethi		%uhi(PAGE_OFFSET), %g1
 	sllx		%g1, 32, %g1
-	sub		%o0, %g1, %o0
-	clr		%o4
-	srlx		%o0, 11, %o0
-	sethi		%hi(1 << 14), %o2
-1:	ldxa		[%o4] ASI_DCACHE_TAG, %o3	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	ldxa		[%o4] ASI_DCACHE_TAG, %g1	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	ldxa		[%o4] ASI_DCACHE_TAG, %g2	! LSU	Group	o3 available
-	add		%o4, (1 << 5), %o4		! IEU0
-	andn		%o3, DTAG_MASK, %o3		! IEU1
-	ldxa		[%o4] ASI_DCACHE_TAG, %g3	! LSU	Group
-	add		%o4, (1 << 5), %o4		! IEU0
-	andn		%g1, DTAG_MASK, %g1		! IEU1
-	cmp		%o0, %o3			! IEU1	Group
-	be,a,pn		%xcc, dflush1			! CTI
-	 sub		%o4, (4 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g1			! IEU1	Group
-	andn		%g2, DTAG_MASK, %g2		! IEU0
-	be,a,pn		%xcc, dflush2			! CTI
-	 sub		%o4, (3 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g2			! IEU1	Group
-	andn		%g3, DTAG_MASK, %g3		! IEU0
-	be,a,pn		%xcc, dflush3			! CTI
-	 sub		%o4, (2 << 5), %o4		! IEU0	(Group)
-	cmp		%o0, %g3			! IEU1	Group
-	be,a,pn		%xcc, dflush4			! CTI
-	 sub		%o4, (1 << 5), %o4		! IEU0
-2:	cmp		%o4, %o2			! IEU1	Group
-	bne,pt		%xcc, 1b			! CTI
-	 nop						! IEU0
+	sub		%o0, %g1, %o0			! physical address
+	srlx		%o0, 11, %o0			! make D-cache TAG
+	sethi		%hi(1 << 14), %o2		! D-cache size
+	sub		%o2, (1 << 5), %o2		! D-cache line size
+1:	ldxa		[%o2] ASI_DCACHE_TAG, %o3	! load D-cache TAG
+	andcc		%o3, DTAG_MASK, %g0		! Valid?
+	be,pn		%xcc, 2f			! Nope, branch
+	 andn		%o3, DTAG_MASK, %o3		! Clear valid bits
+	cmp		%o3, %o0			! TAG match?
+	bne,pt		%xcc, 2f			! Nope, branch
+	 nop
+	stxa		%g0, [%o2] ASI_DCACHE_TAG	! Invalidate TAG
+	membar		#Sync
+2:	brnz,pt		%o2, 1b
+	 sub		%o2, (1 << 5), %o2		! D-cache line size
 
 	/* The I-cache does not snoop local stores so we
 	 * better flush that too when necessary.
@@ -189,48 +176,9 @@ __flush_dcache_page:	/* %o0=kaddr, %o1=flush_icache */
 	retl
 	 nop
 
-dflush1:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush2:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush3:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-dflush4:stxa		%g0, [%o4] ASI_DCACHE_TAG
-	add		%o4, (1 << 5), %o4
-	membar		#Sync
-	ba,pt		%xcc, 2b
-	 nop
 #endif /* DCACHE_ALIASING_POSSIBLE */
 
-	.previous .text
-	.align		32
-__prefill_dtlb:
-	rdpr		%pstate, %g7
-	wrpr		%g7, PSTATE_IE, %pstate
-	mov		TLB_TAG_ACCESS, %g1
-	stxa		%o5, [%g1] ASI_DMMU
-	stxa		%o2, [%g0] ASI_DTLB_DATA_IN
-	flush		%g6
-	retl
-	 wrpr		%g7, %pstate
-__prefill_itlb:
-	rdpr		%pstate, %g7
-	wrpr		%g7, PSTATE_IE, %pstate
-	mov		TLB_TAG_ACCESS, %g1
-	stxa		%o5, [%g1] ASI_IMMU
-	stxa		%o2, [%g0] ASI_ITLB_DATA_IN
-	flush		%g6
-	retl
-	 wrpr		%g7, %pstate
-
-	.globl		__update_mmu_cache
-__update_mmu_cache:	/* %o0=hw_context, %o1=address, %o2=pte, %o3=fault_code */
-	srlx		%o1, PAGE_SHIFT, %o1
-	andcc		%o3, FAULT_CODE_DTLB, %g0
-	sllx		%o1, PAGE_SHIFT, %o5
-	bne,pt		%xcc, __prefill_dtlb
-	 or		%o5, %o0, %o5
-	ba,a,pt		%xcc, __prefill_itlb
+	.previous
 
 	/* Cheetah specific versions, patched at boot time. */
 __cheetah_flush_tlb_mm: /* 18 insns */
@@ -283,7 +231,7 @@ __cheetah_flush_tlb_pending:	/* 26 insns */
 	 wrpr		%g7, 0x0, %pstate
 
 #ifdef DCACHE_ALIASING_POSSIBLE
-flush_dcpage_cheetah: /* 11 insns */
+__cheetah_flush_dcache_page: /* 11 insns */
 	sethi		%uhi(PAGE_OFFSET), %g1
 	sllx		%g1, 32, %g1
 	sub		%o0, %g1, %o0
@@ -329,8 +277,8 @@ cheetah_patch_cachetlbops:
 #ifdef DCACHE_ALIASING_POSSIBLE
 	sethi		%hi(__flush_dcache_page), %o0
 	or		%o0, %lo(__flush_dcache_page), %o0
-	sethi		%hi(flush_dcpage_cheetah), %o1
-	or		%o1, %lo(flush_dcpage_cheetah), %o1
+	sethi		%hi(__cheetah_flush_dcache_page), %o1
+	or		%o1, %lo(__cheetah_flush_dcache_page), %o1
 	call		cheetah_patch_one
 	 mov		11, %o2
 #endif /* DCACHE_ALIASING_POSSIBLE */
@@ -505,22 +453,6 @@ xcall_flush_dcache_page_spitfire: /* %g1 == physical page address
 	nop
 	nop
 
-	.globl		xcall_promstop
-xcall_promstop:
-	rdpr		%pstate, %g2
-	wrpr		%g2, PSTATE_IG | PSTATE_AG, %pstate
-	rdpr		%pil, %g2
-	wrpr		%g0, 15, %pil
-	sethi		%hi(109f), %g7
-	b,pt		%xcc, etrap_irq
-109:	 or		%g7, %lo(109b), %g7
-	flushw
-	call		prom_stopself
-	 nop
-	/* We should not return, just spin if we do... */
-1:	b,a,pt		%xcc, 1b
-	nop
-
 	.data
 
 errata32_hwbug: