Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6: (230 commits)
  [SPARC64]: Update defconfig.
  [SPARC64]: Fix 2 bugs in huge page support.
  [SPARC64]: CONFIG_BLK_DEV_RAM fix
  [SPARC64]: Optimized TSB table initialization.
  [SPARC64]: Allow CONFIG_MEMORY_HOTPLUG to build.
  [SPARC64]: Use SLAB caches for TSB tables.
  [SPARC64]: Don't kill the page allocator when growing a TSB.
  [SPARC64]: Randomize mm->mmap_base when PF_RANDOMIZE is set.
  [SPARC64]: Increase top of 32-bit process stack.
  [SPARC64]: Top-down address space allocation for 32-bit tasks.
  [SPARC64] bbc_i2c: Fix cpu check and add missing module license.
  [SPARC64]: Fix and re-enable dynamic TSB sizing.
  [SUNSU]: Fix missing spinlock initialization.
  [TG3]: Do not try to access NIC_SRAM_DATA_SIG on Sun parts.
  [SPARC64]: First cut at VIS simulator for Niagara.
  [SPARC64]: Fix system type in /proc/cpuinfo and remove bogus OBP check.
  [SPARC64]: Add SMT scheduling support for Niagara.
  [SPARC64]: Fix 32-bit truncation which broke sparsemem.
  [SPARC64]: Move over to sparsemem.
  [SPARC64]: Fix new context version SMP handling.
  ...

38 files changed, 3797 insertions, 682 deletions

diff --git a/include/asm-sparc/idprom.h b/include/asm-sparc/idprom.h
index d856e640acd3..59083ed85232 100644
--- a/include/asm-sparc/idprom.h
+++ b/include/asm-sparc/idprom.h
@@ -7,27 +7,19 @@
 #ifndef _SPARC_IDPROM_H
 #define _SPARC_IDPROM_H
 
-/* Offset into the EEPROM where the id PROM is located on the 4c */
-#define IDPROM_OFFSET  0x7d8
+#include <linux/types.h>
 
-/* On sun4m; physical. */
-/* MicroSPARC(-II) does not decode 31rd bit, but it works. */
-#define IDPROM_OFFSET_M  0xfd8
-
-struct idprom
-{
-	unsigned char	id_format;	/* Format identifier (always 0x01) */
-	unsigned char	id_machtype;	/* Machine type */
-	unsigned char	id_ethaddr[6];	/* Hardware ethernet address */
-	long		id_date;	/* Date of manufacture */
-	unsigned int	id_sernum:24;	/* Unique serial number */
-	unsigned char	id_cksum;	/* Checksum - xor of the data bytes */
-	unsigned char	reserved[16];
+struct idprom {
+	u8		id_format;	/* Format identifier (always 0x01) */
+	u8		id_machtype;	/* Machine type */
+	u8		id_ethaddr[6];	/* Hardware ethernet address */
+	s32		id_date;	/* Date of manufacture */
+	u32		id_sernum:24;	/* Unique serial number */
+	u8		id_cksum;	/* Checksum - xor of the data bytes */
+	u8		reserved[16];
 };
 
 extern struct idprom *idprom;
 extern void idprom_init(void);
 
-#define IDPROM_SIZE  (sizeof(struct idprom))
-
 #endif /* !(_SPARC_IDPROM_H) */
diff --git a/include/asm-sparc/oplib.h b/include/asm-sparc/oplib.h
index d0d76b30eb4c..f283f8aaf6a9 100644
--- a/include/asm-sparc/oplib.h
+++ b/include/asm-sparc/oplib.h
@@ -165,6 +165,7 @@ enum prom_input_device {
 	PROMDEV_ITTYA,			/* input from ttya */
 	PROMDEV_ITTYB,			/* input from ttyb */
 	PROMDEV_IRSC,			/* input from rsc */
+	PROMDEV_IVCONS,			/* input from virtual-console */
 	PROMDEV_I_UNK,
 };
 
@@ -177,6 +178,7 @@ enum prom_output_device {
 	PROMDEV_OTTYA,			/* to ttya */
 	PROMDEV_OTTYB,			/* to ttyb */
 	PROMDEV_ORSC,			/* to rsc */
+	PROMDEV_OVCONS,			/* to virtual-console */
 	PROMDEV_O_UNK,
 };
 
diff --git a/include/asm-sparc/uaccess.h b/include/asm-sparc/uaccess.h
index f8f1ec1f06e6..3cf132e1aa25 100644
--- a/include/asm-sparc/uaccess.h
+++ b/include/asm-sparc/uaccess.h
@@ -120,17 +120,6 @@ case 8: __put_user_asm(x,d,addr,__pu_ret); break; \
 default: __pu_ret = __put_user_bad(); break; \
 } } else { __pu_ret = -EFAULT; } __pu_ret; })
 
-#define __put_user_check_ret(x,addr,size,retval) ({ \
-register int __foo __asm__ ("l1"); \
-if (__access_ok(addr,size)) { \
-switch (size) { \
-case 1: __put_user_asm_ret(x,b,addr,retval,__foo); break; \
-case 2: __put_user_asm_ret(x,h,addr,retval,__foo); break; \
-case 4: __put_user_asm_ret(x,,addr,retval,__foo); break; \
-case 8: __put_user_asm_ret(x,d,addr,retval,__foo); break; \
-default: if (__put_user_bad()) return retval; break; \
-} } else return retval; })
-
 #define __put_user_nocheck(x,addr,size) ({ \
 register int __pu_ret; \
 switch (size) { \
@@ -141,16 +130,6 @@ case 8: __put_user_asm(x,d,addr,__pu_ret); break; \
 default: __pu_ret = __put_user_bad(); break; \
 } __pu_ret; })
 
-#define __put_user_nocheck_ret(x,addr,size,retval) ({ \
-register int __foo __asm__ ("l1"); \
-switch (size) { \
-case 1: __put_user_asm_ret(x,b,addr,retval,__foo); break; \
-case 2: __put_user_asm_ret(x,h,addr,retval,__foo); break; \
-case 4: __put_user_asm_ret(x,,addr,retval,__foo); break; \
-case 8: __put_user_asm_ret(x,d,addr,retval,__foo); break; \
-default: if (__put_user_bad()) return retval; break; \
-} })
-
 #define __put_user_asm(x,size,addr,ret)					\
 __asm__ __volatile__(							\
 	"/* Put user asm, inline. */\n"					\
@@ -170,32 +149,6 @@ __asm__ __volatile__(							\
        : "=&r" (ret) : "r" (x), "m" (*__m(addr)),			\
 	 "i" (-EFAULT))
 
-#define __put_user_asm_ret(x,size,addr,ret,foo)				\
-if (__builtin_constant_p(ret) && ret == -EFAULT)			\
-__asm__ __volatile__(							\
-	"/* Put user asm ret, inline. */\n"				\
-"1:\t"	"st"#size " %1, %2\n\n\t"					\
-	".section __ex_table,#alloc\n\t"				\
-	".align	4\n\t"							\
-	".word	1b, __ret_efault\n\n\t"					\
-	".previous\n\n\t"						\
-       : "=r" (foo) : "r" (x), "m" (*__m(addr)));			\
-else									\
-__asm__ __volatile(							\
-	"/* Put user asm ret, inline. */\n"				\
-"1:\t"	"st"#size " %1, %2\n\n\t"					\
-	".section .fixup,#alloc,#execinstr\n\t"				\
-	".align	4\n"							\
-"3:\n\t"								\
-	"ret\n\t"							\
-	" restore %%g0, %3, %%o0\n\t"					\
-	".previous\n\n\t"						\
-	".section __ex_table,#alloc\n\t"				\
-	".align	4\n\t"							\
-	".word	1b, 3b\n\n\t"						\
-	".previous\n\n\t"						\
-       : "=r" (foo) : "r" (x), "m" (*__m(addr)), "i" (ret))
-
 extern int __put_user_bad(void);
 
 #define __get_user_check(x,addr,size,type) ({ \
diff --git a/include/asm-sparc64/a.out.h b/include/asm-sparc64/a.out.h
index 02af289e3f46..35cb5c9e0c92 100644
--- a/include/asm-sparc64/a.out.h
+++ b/include/asm-sparc64/a.out.h
@@ -95,7 +95,11 @@ struct relocation_info /* used when header.a_machtype == M_SPARC */
 
 #ifdef __KERNEL__
 
-#define STACK_TOP (test_thread_flag(TIF_32BIT) ? 0xf0000000 : 0x80000000000L)
+#define STACK_TOP32	((1UL << 32UL) - PAGE_SIZE)
+#define STACK_TOP64	(0x0000080000000000UL - (1UL << 32UL))
+
+#define STACK_TOP (test_thread_flag(TIF_32BIT) ? \
+		   STACK_TOP32 : STACK_TOP64)
 
 #endif
 
diff --git a/include/asm-sparc64/asi.h b/include/asm-sparc64/asi.h
index 534855660f2a..662a21107ae6 100644
--- a/include/asm-sparc64/asi.h
+++ b/include/asm-sparc64/asi.h
@@ -25,14 +25,27 @@
 
 /* SpitFire and later extended ASIs.  The "(III)" marker designates
  * UltraSparc-III and later specific ASIs.  The "(CMT)" marker designates
- * Chip Multi Threading specific ASIs.
+ * Chip Multi Threading specific ASIs.  "(NG)" designates Niagara specific
+ * ASIs, "(4V)" designates SUN4V specific ASIs.
  */
 #define ASI_PHYS_USE_EC		0x14 /* PADDR, E-cachable		*/
 #define ASI_PHYS_BYPASS_EC_E	0x15 /* PADDR, E-bit			*/
+#define ASI_BLK_AIUP_4V		0x16 /* (4V) Prim, user, block ld/st	*/
+#define ASI_BLK_AIUS_4V		0x17 /* (4V) Sec, user, block ld/st	*/
 #define ASI_PHYS_USE_EC_L	0x1c /* PADDR, E-cachable, little endian*/
 #define ASI_PHYS_BYPASS_EC_E_L	0x1d /* PADDR, E-bit, little endian	*/
+#define ASI_BLK_AIUP_L_4V	0x1e /* (4V) Prim, user, block, l-endian*/
+#define ASI_BLK_AIUS_L_4V	0x1f /* (4V) Sec, user, block, l-endian	*/
+#define ASI_SCRATCHPAD		0x20 /* (4V) Scratch Pad Registers	*/
+#define ASI_MMU			0x21 /* (4V) MMU Context Registers	*/
+#define ASI_BLK_INIT_QUAD_LDD_AIUS 0x23 /* (NG) init-store, twin load,
+					 * secondary, user
+					 */
 #define ASI_NUCLEUS_QUAD_LDD	0x24 /* Cachable, qword load		*/
+#define ASI_QUEUE		0x25 /* (4V) Interrupt Queue Registers	*/
+#define ASI_QUAD_LDD_PHYS_4V	0x26 /* (4V) Physical, qword load	*/
 #define ASI_NUCLEUS_QUAD_LDD_L	0x2c /* Cachable, qword load, l-endian 	*/
+#define ASI_QUAD_LDD_PHYS_L_4V	0x2e /* (4V) Phys, qword load, l-endian	*/
 #define ASI_PCACHE_DATA_STATUS	0x30 /* (III) PCache data stat RAM diag	*/
 #define ASI_PCACHE_DATA		0x31 /* (III) PCache data RAM diag	*/
 #define ASI_PCACHE_TAG		0x32 /* (III) PCache tag RAM diag	*/
@@ -137,6 +150,9 @@
 #define ASI_FL16_SL		0xdb /* Secondary, 1 16-bit, fpu ld/st,L*/
 #define ASI_BLK_COMMIT_P	0xe0 /* Primary, blk store commit	*/
 #define ASI_BLK_COMMIT_S	0xe1 /* Secondary, blk store commit	*/
+#define ASI_BLK_INIT_QUAD_LDD_P	0xe2 /* (NG) init-store, twin load,
+				      * primary, implicit
+				      */
 #define ASI_BLK_P		0xf0 /* Primary, blk ld/st		*/
 #define ASI_BLK_S		0xf1 /* Secondary, blk ld/st		*/
 #define ASI_BLK_PL		0xf8 /* Primary, blk ld/st, little	*/
diff --git a/include/asm-sparc64/cpudata.h b/include/asm-sparc64/cpudata.h
index 74de79dca915..c66a81bbc84d 100644
--- a/include/asm-sparc64/cpudata.h
+++ b/include/asm-sparc64/cpudata.h
@@ -1,41 +1,224 @@
 /* cpudata.h: Per-cpu parameters.
  *
- * Copyright (C) 2003, 2005 David S. Miller (davem@redhat.com)
+ * Copyright (C) 2003, 2005, 2006 David S. Miller (davem@davemloft.net)
  */
 
 #ifndef _SPARC64_CPUDATA_H
 #define _SPARC64_CPUDATA_H
 
+#include <asm/hypervisor.h>
+#include <asm/asi.h>
+
+#ifndef __ASSEMBLY__
+
 #include <linux/percpu.h>
+#include <linux/threads.h>
 
 typedef struct {
 	/* Dcache line 1 */
 	unsigned int	__softirq_pending; /* must be 1st, see rtrap.S */
 	unsigned int	multiplier;
 	unsigned int	counter;
-	unsigned int	idle_volume;
+	unsigned int	__pad1;
 	unsigned long	clock_tick;	/* %tick's per second */
 	unsigned long	udelay_val;
 
-	/* Dcache line 2 */
-	unsigned int	pgcache_size;
-	unsigned int	__pad1;
-	unsigned long	*pte_cache[2];
-	unsigned long	*pgd_cache;
-
-	/* Dcache line 3, rarely used */
+	/* Dcache line 2, rarely used */
 	unsigned int	dcache_size;
 	unsigned int	dcache_line_size;
 	unsigned int	icache_size;
 	unsigned int	icache_line_size;
 	unsigned int	ecache_size;
 	unsigned int	ecache_line_size;
-	unsigned int	__pad2;
 	unsigned int	__pad3;
+	unsigned int	__pad4;
 } cpuinfo_sparc;
 
 DECLARE_PER_CPU(cpuinfo_sparc, __cpu_data);
 #define cpu_data(__cpu)		per_cpu(__cpu_data, (__cpu))
 #define local_cpu_data()	__get_cpu_var(__cpu_data)
 
+/* Trap handling code needs to get at a few critical values upon
+ * trap entry and to process TSB misses.  These cannot be in the
+ * per_cpu() area as we really need to lock them into the TLB and
+ * thus make them part of the main kernel image.  As a result we
+ * try to make this as small as possible.
+ *
+ * This is padded out and aligned to 64-bytes to avoid false sharing
+ * on SMP.
+ */
+
+/* If you modify the size of this structure, please update
+ * TRAP_BLOCK_SZ_SHIFT below.
+ */
+struct thread_info;
+struct trap_per_cpu {
+/* D-cache line 1: Basic thread information, cpu and device mondo queues */
+	struct thread_info	*thread;
+	unsigned long		pgd_paddr;
+	unsigned long		cpu_mondo_pa;
+	unsigned long		dev_mondo_pa;
+
+/* D-cache line 2: Error Mondo Queue and kernel buffer pointers */
+	unsigned long		resum_mondo_pa;
+	unsigned long		resum_kernel_buf_pa;
+	unsigned long		nonresum_mondo_pa;
+	unsigned long		nonresum_kernel_buf_pa;
+
+/* Dcache lines 3, 4, 5, and 6: Hypervisor Fault Status */
+	struct hv_fault_status	fault_info;
+
+/* Dcache line 7: Physical addresses of CPU send mondo block and CPU list.  */
+	unsigned long		cpu_mondo_block_pa;
+	unsigned long		cpu_list_pa;
+	unsigned long		__pad1[2];
+
+/* Dcache line 8: Unused, needed to keep trap_block a power-of-2 in size.  */
+	unsigned long		__pad2[4];
+} __attribute__((aligned(64)));
+extern struct trap_per_cpu trap_block[NR_CPUS];
+extern void init_cur_cpu_trap(struct thread_info *);
+extern void setup_tba(void);
+
+struct cpuid_patch_entry {
+	unsigned int	addr;
+	unsigned int	cheetah_safari[4];
+	unsigned int	cheetah_jbus[4];
+	unsigned int	starfire[4];
+	unsigned int	sun4v[4];
+};
+extern struct cpuid_patch_entry __cpuid_patch, __cpuid_patch_end;
+
+struct sun4v_1insn_patch_entry {
+	unsigned int	addr;
+	unsigned int	insn;
+};
+extern struct sun4v_1insn_patch_entry __sun4v_1insn_patch,
+	__sun4v_1insn_patch_end;
+
+struct sun4v_2insn_patch_entry {
+	unsigned int	addr;
+	unsigned int	insns[2];
+};
+extern struct sun4v_2insn_patch_entry __sun4v_2insn_patch,
+	__sun4v_2insn_patch_end;
+
+#endif /* !(__ASSEMBLY__) */
+
+#define TRAP_PER_CPU_THREAD		0x00
+#define TRAP_PER_CPU_PGD_PADDR		0x08
+#define TRAP_PER_CPU_CPU_MONDO_PA	0x10
+#define TRAP_PER_CPU_DEV_MONDO_PA	0x18
+#define TRAP_PER_CPU_RESUM_MONDO_PA	0x20
+#define TRAP_PER_CPU_RESUM_KBUF_PA	0x28
+#define TRAP_PER_CPU_NONRESUM_MONDO_PA	0x30
+#define TRAP_PER_CPU_NONRESUM_KBUF_PA	0x38
+#define TRAP_PER_CPU_FAULT_INFO		0x40
+#define TRAP_PER_CPU_CPU_MONDO_BLOCK_PA	0xc0
+#define TRAP_PER_CPU_CPU_LIST_PA	0xc8
+
+#define TRAP_BLOCK_SZ_SHIFT		8
+
+#include <asm/scratchpad.h>
+
+#define __GET_CPUID(REG)				\
+	/* Spitfire implementation (default). */	\
+661:	ldxa		[%g0] ASI_UPA_CONFIG, REG;	\
+	srlx		REG, 17, REG;			\
+	 and		REG, 0x1f, REG;			\
+	nop;						\
+	.section	.cpuid_patch, "ax";		\
+	/* Instruction location. */			\
+	.word		661b;				\
+	/* Cheetah Safari implementation. */		\
+	ldxa		[%g0] ASI_SAFARI_CONFIG, REG;	\
+	srlx		REG, 17, REG;			\
+	and		REG, 0x3ff, REG;		\
+	nop;						\
+	/* Cheetah JBUS implementation. */		\
+	ldxa		[%g0] ASI_JBUS_CONFIG, REG;	\
+	srlx		REG, 17, REG;			\
+	and		REG, 0x1f, REG;			\
+	nop;						\
+	/* Starfire implementation. */			\
+	sethi		%hi(0x1fff40000d0 >> 9), REG;	\
+	sllx		REG, 9, REG;			\
+	or		REG, 0xd0, REG;			\
+	lduwa		[REG] ASI_PHYS_BYPASS_EC_E, REG;\
+	/* sun4v implementation. */			\
+	mov		SCRATCHPAD_CPUID, REG;		\
+	ldxa		[REG] ASI_SCRATCHPAD, REG;	\
+	nop;						\
+	nop;						\
+	.previous;
+
+#ifdef CONFIG_SMP
+
+#define TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	__GET_CPUID(TMP)			\
+	sethi	%hi(trap_block), DEST;		\
+	sllx	TMP, TRAP_BLOCK_SZ_SHIFT, TMP;	\
+	or	DEST, %lo(trap_block), DEST;	\
+	add	DEST, TMP, DEST;		\
+
+/* Clobbers TMP, current address space PGD phys address into DEST.  */
+#define TRAP_LOAD_PGD_PHYS(DEST, TMP)		\
+	TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	ldx	[DEST + TRAP_PER_CPU_PGD_PADDR], DEST;
+
+/* Clobbers TMP, loads local processor's IRQ work area into DEST.  */
+#define TRAP_LOAD_IRQ_WORK(DEST, TMP)		\
+	__GET_CPUID(TMP)			\
+	sethi	%hi(__irq_work), DEST;		\
+	sllx	TMP, 6, TMP;			\
+	or	DEST, %lo(__irq_work), DEST;	\
+	add	DEST, TMP, DEST;
+
+/* Clobbers TMP, loads DEST with current thread info pointer.  */
+#define TRAP_LOAD_THREAD_REG(DEST, TMP)		\
+	TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	ldx	[DEST + TRAP_PER_CPU_THREAD], DEST;
+
+/* Given the current thread info pointer in THR, load the per-cpu
+ * area base of the current processor into DEST.  REG1, REG2, and REG3 are
+ * clobbered.
+ *
+ * You absolutely cannot use DEST as a temporary in this code.  The
+ * reason is that traps can happen during execution, and return from
+ * trap will load the fully resolved DEST per-cpu base.  This can corrupt
+ * the calculations done by the macro mid-stream.
+ */
+#define LOAD_PER_CPU_BASE(DEST, THR, REG1, REG2, REG3)	\
+	ldub	[THR + TI_CPU], REG1;			\
+	sethi	%hi(__per_cpu_shift), REG3;		\
+	sethi	%hi(__per_cpu_base), REG2;		\
+	ldx	[REG3 + %lo(__per_cpu_shift)], REG3;	\
+	ldx	[REG2 + %lo(__per_cpu_base)], REG2;	\
+	sllx	REG1, REG3, REG3;			\
+	add	REG3, REG2, DEST;
+
+#else
+
+#define TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	sethi	%hi(trap_block), DEST;		\
+	or	DEST, %lo(trap_block), DEST;	\
+
+/* Uniprocessor versions, we know the cpuid is zero.  */
+#define TRAP_LOAD_PGD_PHYS(DEST, TMP)		\
+	TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	ldx	[DEST + TRAP_PER_CPU_PGD_PADDR], DEST;
+
+#define TRAP_LOAD_IRQ_WORK(DEST, TMP)		\
+	sethi	%hi(__irq_work), DEST;		\
+	or	DEST, %lo(__irq_work), DEST;
+
+#define TRAP_LOAD_THREAD_REG(DEST, TMP)		\
+	TRAP_LOAD_TRAP_BLOCK(DEST, TMP)		\
+	ldx	[DEST + TRAP_PER_CPU_THREAD], DEST;
+
+/* No per-cpu areas on uniprocessor, so no need to load DEST.  */
+#define LOAD_PER_CPU_BASE(DEST, THR, REG1, REG2, REG3)
+
+#endif /* !(CONFIG_SMP) */
+
 #endif /* _SPARC64_CPUDATA_H */
diff --git a/include/asm-sparc64/elf.h b/include/asm-sparc64/elf.h
index 69539a8ab833..303d85e2f82e 100644
--- a/include/asm-sparc64/elf.h
+++ b/include/asm-sparc64/elf.h
@@ -10,6 +10,7 @@
 #ifdef __KERNEL__
 #include <asm/processor.h>
 #include <asm/uaccess.h>
+#include <asm/spitfire.h>
 #endif
 
 /*
@@ -68,6 +69,7 @@
 #define HWCAP_SPARC_MULDIV      8
 #define HWCAP_SPARC_V9		16
 #define HWCAP_SPARC_ULTRA3	32
+#define HWCAP_SPARC_BLKINIT	64
 
 /*
  * These are used to set parameters in the core dumps.
@@ -145,11 +147,21 @@ typedef struct {
    instruction set this cpu supports.  */
 
 /* On Ultra, we support all of the v8 capabilities. */
-#define ELF_HWCAP	((HWCAP_SPARC_FLUSH | HWCAP_SPARC_STBAR | \
-			  HWCAP_SPARC_SWAP | HWCAP_SPARC_MULDIV | \
-			  HWCAP_SPARC_V9) | \
-			 ((tlb_type == cheetah || tlb_type == cheetah_plus) ? \
-			  HWCAP_SPARC_ULTRA3 : 0))
+static inline unsigned int sparc64_elf_hwcap(void)
+{
+	unsigned int cap = (HWCAP_SPARC_FLUSH | HWCAP_SPARC_STBAR |
+			    HWCAP_SPARC_SWAP | HWCAP_SPARC_MULDIV |
+			    HWCAP_SPARC_V9);
+
+	if (tlb_type == cheetah || tlb_type == cheetah_plus)
+		cap |= HWCAP_SPARC_ULTRA3;
+	else if (tlb_type == hypervisor)
+		cap |= HWCAP_SPARC_BLKINIT;
+
+	return cap;
+}
+
+#define ELF_HWCAP	sparc64_elf_hwcap();
 
 /* This yields a string that ld.so will use to load implementation
    specific libraries for optimization.  This is more specific in
diff --git a/include/asm-sparc64/head.h b/include/asm-sparc64/head.h
index 0abd3a674e8f..67960a751f4d 100644
--- a/include/asm-sparc64/head.h
+++ b/include/asm-sparc64/head.h
@@ -4,12 +4,21 @@
 
 #include <asm/pstate.h>
 
+	/* wrpr	%g0, val, %gl */
+#define SET_GL(val)	\
+	.word	0xa1902000 | val
+
+	/* rdpr %gl, %gN */
+#define GET_GL_GLOBAL(N)	\
+	.word	0x81540000 | (N << 25)
+
 #define KERNBASE	0x400000
 
 #define	PTREGS_OFF	(STACK_BIAS + STACKFRAME_SZ)
 
 #define __CHEETAH_ID	0x003e0014
 #define __JALAPENO_ID	0x003e0016
+#define __SERRANO_ID	0x003e0022
 
 #define CHEETAH_MANUF		0x003e
 #define CHEETAH_IMPL		0x0014 /* Ultra-III   */
@@ -19,6 +28,12 @@
 #define PANTHER_IMPL		0x0019 /* Ultra-IV+   */
 #define SERRANO_IMPL		0x0022 /* Ultra-IIIi+ */
 
+#define BRANCH_IF_SUN4V(tmp1,label)		\
+	sethi	%hi(is_sun4v), %tmp1;		\
+	lduw	[%tmp1 + %lo(is_sun4v)], %tmp1; \
+	brnz,pn	%tmp1, label;			\
+	 nop
+
 #define BRANCH_IF_CHEETAH_BASE(tmp1,tmp2,label)	\
 	rdpr	%ver, %tmp1;			\
 	sethi	%hi(__CHEETAH_ID), %tmp2;	\
diff --git a/include/asm-sparc64/hypervisor.h b/include/asm-sparc64/hypervisor.h
new file mode 100644
index 000000000000..612bf319753f
--- /dev/null
+++ b/include/asm-sparc64/hypervisor.h
@@ -0,0 +1,2128 @@
+#ifndef _SPARC64_HYPERVISOR_H
+#define _SPARC64_HYPERVISOR_H
+
+/* Sun4v hypervisor interfaces and defines.
+ *
+ * Hypervisor calls are made via traps to software traps number 0x80
+ * and above.  Registers %o0 to %o5 serve as argument, status, and
+ * return value registers.
+ *
+ * There are two kinds of these traps.  First there are the normal
+ * "fast traps" which use software trap 0x80 and encode the function
+ * to invoke by number in register %o5.  Argument and return value
+ * handling is as follows:
+ *
+ * -----------------------------------------------
+ * |  %o5  | function number |     undefined     |
+ * |  %o0  |   argument 0    |   return status   |
+ * |  %o1  |   argument 1    |   return value 1  |
+ * |  %o2  |   argument 2    |   return value 2  |
+ * |  %o3  |   argument 3    |   return value 3  |
+ * |  %o4  |   argument 4    |   return value 4  |
+ * -----------------------------------------------
+ *
+ * The second type are "hyper-fast traps" which encode the function
+ * number in the software trap number itself.  So these use trap
+ * numbers > 0x80.  The register usage for hyper-fast traps is as
+ * follows:
+ *
+ * -----------------------------------------------
+ * |  %o0  |   argument 0    |   return status   |
+ * |  %o1  |   argument 1    |   return value 1  |
+ * |  %o2  |   argument 2    |   return value 2  |
+ * |  %o3  |   argument 3    |   return value 3  |
+ * |  %o4  |   argument 4    |   return value 4  |
+ * -----------------------------------------------
+ *
+ * Registers providing explicit arguments to the hypervisor calls
+ * are volatile across the call.  Upon return their values are
+ * undefined unless explicitly specified as containing a particular
+ * return value by the specific call.  The return status is always
+ * returned in register %o0, zero indicates a successful execution of
+ * the hypervisor call and other values indicate an error status as
+ * defined below.  So, for example, if a hyper-fast trap takes
+ * arguments 0, 1, and 2, then %o0, %o1, and %o2 are volatile across
+ * the call and %o3, %o4, and %o5 would be preserved.
+ *
+ * If the hypervisor trap is invalid, or the fast trap function number
+ * is invalid, HV_EBADTRAP will be returned in %o0.  Also, all 64-bits
+ * of the argument and return values are significant.
+ */
+
+/* Trap numbers.  */
+#define HV_FAST_TRAP		0x80
+#define HV_MMU_MAP_ADDR_TRAP	0x83
+#define HV_MMU_UNMAP_ADDR_TRAP	0x84
+#define HV_TTRACE_ADDENTRY_TRAP	0x85
+#define HV_CORE_TRAP		0xff
+
+/* Error codes.  */
+#define HV_EOK				0  /* Successful return            */
+#define HV_ENOCPU			1  /* Invalid CPU id               */
+#define HV_ENORADDR			2  /* Invalid real address         */
+#define HV_ENOINTR			3  /* Invalid interrupt id         */
+#define HV_EBADPGSZ			4  /* Invalid pagesize encoding    */
+#define HV_EBADTSB			5  /* Invalid TSB description      */
+#define HV_EINVAL			6  /* Invalid argument             */
+#define HV_EBADTRAP			7  /* Invalid function number      */
+#define HV_EBADALIGN			8  /* Invalid address alignment    */
+#define HV_EWOULDBLOCK			9  /* Cannot complete w/o blocking */
+#define HV_ENOACCESS			10 /* No access to resource        */
+#define HV_EIO				11 /* I/O error                    */
+#define HV_ECPUERROR			12 /* CPU in error state           */
+#define HV_ENOTSUPPORTED		13 /* Function not supported       */
+#define HV_ENOMAP			14 /* No mapping found             */
+#define HV_ETOOMANY			15 /* Too many items specified     */
+
+/* mach_exit()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_EXIT
+ * ARG0:	exit code
+ * ERRORS:	This service does not return.
+ *
+ * Stop all CPUs in the virtual domain and place them into the stopped
+ * state.  The 64-bit exit code may be passed to a service entity as
+ * the domain's exit status.  On systems without a service entity, the
+ * domain will undergo a reset, and the boot firmware will be
+ * reloaded.
+ *
+ * This function will never return to the guest that invokes it.
+ *
+ * Note: By convention an exit code of zero denotes a successful exit by
+ *       the guest code.  A non-zero exit code denotes a guest specific
+ *       error indication.
+ *
+ */
+#define HV_FAST_MACH_EXIT		0x00
+
+/* Domain services.  */
+
+/* mach_desc()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_DESC
+ * ARG0:	buffer
+ * ARG1:	length
+ * RET0:	status
+ * RET1:	length
+ * ERRORS:	HV_EBADALIGN	Buffer is badly aligned
+ *		HV_ENORADDR	Buffer is to an illegal real address.
+ *		HV_EINVAL	Buffer length is too small for complete
+ *				machine description.
+ *
+ * Copy the most current machine description into the buffer indicated
+ * by the real address in ARG0.  The buffer provided must be 16 byte
+ * aligned.  Upon success or HV_EINVAL, this service returns the
+ * actual size of the machine description in the RET1 return value.
+ *
+ * Note: A method of determining the appropriate buffer size for the
+ *       machine description is to first call this service with a buffer
+ *       length of 0 bytes.
+ */
+#define HV_FAST_MACH_DESC		0x01
+
+/* mach_exit()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_SIR
+ * ERRORS:	This service does not return.
+ *
+ * Perform a software initiated reset of the virtual machine domain.
+ * All CPUs are captured as soon as possible, all hardware devices are
+ * returned to the entry default state, and the domain is restarted at
+ * the SIR (trap type 0x04) real trap table (RTBA) entry point on one
+ * of the CPUs.  The single CPU restarted is selected as determined by
+ * platform specific policy.  Memory is preserved across this
+ * operation.
+ */
+#define HV_FAST_MACH_SIR		0x02
+
+/* mach_set_soft_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_SET_SOFT_STATE
+ * ARG0:	software state
+ * ARG1:	software state description pointer
+ * RET0:	status
+ * ERRORS:	EINVAL		software state not valid or software state
+ *				description is not NULL terminated
+ *		ENORADDR	software state description pointer is not a
+ *				valid real address
+ *		EBADALIGNED	software state description is not correctly
+ *				aligned
+ *
+ * This allows the guest to report it's soft state to the hypervisor.  There
+ * are two primary components to this state.  The first part states whether
+ * the guest software is running or not.  The second containts optional
+ * details specific to the software.
+ *
+ * The software state argument is defined below in HV_SOFT_STATE_*, and
+ * indicates whether the guest is operating normally or in a transitional
+ * state.
+ *
+ * The software state description argument is a real address of a data buffer
+ * of size 32-bytes aligned on a 32-byte boundary.  It is treated as a NULL
+ * terminated 7-bit ASCII string of up to 31 characters not including the
+ * NULL termination.
+ */
+#define HV_FAST_MACH_SET_SOFT_STATE	0x03
+#define  HV_SOFT_STATE_NORMAL		 0x01
+#define  HV_SOFT_STATE_TRANSITION	 0x02
+
+/* mach_get_soft_state()
+ * TRAP:	HV_FAST_TRAP
+ * FUNCTION:	HV_FAST_MACH_GET_SOFT_STATE
+ * ARG0:	software state description pointer
+ * RET0:	status
+ * RET1:	software state
+ * ERRORS:	ENORADDR	software state description pointer is not a
+ *				valid real address
+ *		EBADALIGNED	software state description is not correctly
+ *				aligned
+ *
+ * Retrieve the current value of the guest's software state.  The rules
+ * for the software state pointer are the same as for mach_set_soft_state()
+ * above.
+ */
+#define HV_FAST_MACH_GET_SOFT_STATE	0x04
+
+/* CPU services.
+ *
+ * CPUs represent devices that can execute software threads.  A single
+ * chip that contains multiple cores or strands is represented as
+ * multiple CPUs with unique CPU identifiers.  CPUs are exported to
+ * OBP via the machine description (and to the OS via the OBP device
+ * tree).  CPUs are always in one of three states: stopped, running,
+ * or error.
+ *
+ * A CPU ID is a pre-assi