19 files changed, 219 insertions, 264 deletions

diff --git a/include/asm-sparc64/auxvec.h b/include/asm-sparc64/auxvec.h
new file mode 100644
index 00000000000..436a2912982
--- /dev/null
+++ b/include/asm-sparc64/auxvec.h
@@ -0,0 +1,4 @@
+#ifndef __ASM_SPARC64_AUXVEC_H
+#define __ASM_SPARC64_AUXVEC_H
+
+#endif /* !(__ASM_SPARC64_AUXVEC_H) */
diff --git a/include/asm-sparc64/cacheflush.h b/include/asm-sparc64/cacheflush.h
index 51b26e81d82..ededd2659ea 100644
--- a/include/asm-sparc64/cacheflush.h
+++ b/include/asm-sparc64/cacheflush.h
@@ -4,13 +4,6 @@
 #include <linux/config.h>
 #include <asm/page.h>
 
-/* Flushing for D-cache alias handling is only needed if
- * the page size is smaller than 16K.
- */
-#if PAGE_SHIFT < 14
-#define DCACHE_ALIASING_POSSIBLE
-#endif
-
 #ifndef __ASSEMBLY__
 
 #include <linux/mm.h>
diff --git a/include/asm-sparc64/compat.h b/include/asm-sparc64/compat.h
index b59122dd176..c73935dc7ba 100644
--- a/include/asm-sparc64/compat.h
+++ b/include/asm-sparc64/compat.h
@@ -12,8 +12,10 @@ typedef s32		compat_ssize_t;
 typedef s32		compat_time_t;
 typedef s32		compat_clock_t;
 typedef s32		compat_pid_t;
-typedef u16		compat_uid_t;
-typedef u16		compat_gid_t;
+typedef u16		__compat_uid_t;
+typedef u16		__compat_gid_t;
+typedef u32		__compat_uid32_t;
+typedef u32		__compat_gid32_t;
 typedef u16		compat_mode_t;
 typedef u32		compat_ino_t;
 typedef u16		compat_dev_t;
@@ -47,8 +49,8 @@ struct compat_stat {
 	compat_ino_t	st_ino;
 	compat_mode_t	st_mode;
 	compat_nlink_t	st_nlink;
-	compat_uid_t	st_uid;
-	compat_gid_t	st_gid;
+	__compat_uid_t	st_uid;
+	__compat_gid_t	st_gid;
 	compat_dev_t	st_rdev;
 	compat_off_t	st_size;
 	compat_time_t	st_atime;
@@ -177,10 +179,10 @@ static __inline__ void __user *compat_alloc_user_space(long len)
 
 struct compat_ipc64_perm {
 	compat_key_t key;
-	__kernel_uid_t uid;
-	__kernel_gid_t gid;
-	__kernel_uid_t cuid;
-	__kernel_gid_t cgid;
+	__compat_uid32_t uid;
+	__compat_gid32_t gid;
+	__compat_uid32_t cuid;
+	__compat_gid32_t cgid;
 	unsigned short __pad1;
 	compat_mode_t mode;
 	unsigned short __pad2;
diff --git a/include/asm-sparc64/cpudata.h b/include/asm-sparc64/cpudata.h
index cc7198aaac5..9a3a81f1cc5 100644
--- a/include/asm-sparc64/cpudata.h
+++ b/include/asm-sparc64/cpudata.h
@@ -1,6 +1,6 @@
 /* cpudata.h: Per-cpu parameters.
  *
- * Copyright (C) 2003 David S. Miller (davem@redhat.com)
+ * Copyright (C) 2003, 2005 David S. Miller (davem@redhat.com)
  */
 
 #ifndef _SPARC64_CPUDATA_H
@@ -10,7 +10,7 @@
 
 typedef struct {
 	/* Dcache line 1 */
-	unsigned int	__pad0;		/* bh_count moved to irq_stat for consistency. KAO */
+	unsigned int	__softirq_pending; /* must be 1st, see rtrap.S */
 	unsigned int	multiplier;
 	unsigned int	counter;
 	unsigned int	idle_volume;
diff --git a/include/asm-sparc64/fcntl.h b/include/asm-sparc64/fcntl.h
index e36def0d0d8..b2aecf0054b 100644
--- a/include/asm-sparc64/fcntl.h
+++ b/include/asm-sparc64/fcntl.h
@@ -4,10 +4,6 @@
 
 /* open/fcntl - O_SYNC is only implemented on blocks devices and on files
    located on an ext2 file system */
-#define O_RDONLY	0x0000
-#define O_WRONLY	0x0001
-#define O_RDWR		0x0002
-#define O_ACCMODE	0x0003
 #define O_NDELAY	0x0004
 #define O_APPEND	0x0008
 #define FASYNC		0x0040	/* fcntl, for BSD compatibility */
@@ -17,62 +13,24 @@
 #define O_SYNC		0x2000
 #define O_NONBLOCK	0x4000
 #define O_NOCTTY	0x8000	/* not fcntl */
-#define O_DIRECTORY	0x10000	/* must be a directory */
-#define O_NOFOLLOW	0x20000	/* don't follow links */
 #define O_LARGEFILE	0x40000
 #define O_DIRECT        0x100000 /* direct disk access hint */
 #define O_NOATIME	0x200000
 
 
-#define F_DUPFD		0	/* dup */
-#define F_GETFD		1	/* get close_on_exec */
-#define F_SETFD		2	/* set/clear close_on_exec */
-#define F_GETFL		3	/* get file->f_flags */
-#define F_SETFL		4	/* set file->f_flags */
 #define F_GETOWN	5	/*  for sockets. */
 #define F_SETOWN	6	/*  for sockets. */
 #define F_GETLK		7
 #define F_SETLK		8
 #define F_SETLKW	9
-#define F_SETSIG	10	/*  for sockets. */
-#define F_GETSIG	11	/*  for sockets. */
-
-/* for F_[GET|SET]FL */
-#define FD_CLOEXEC	1	/* actually anything with low bit set goes */
 
 /* for posix fcntl() and lockf() */
 #define F_RDLCK		1
 #define F_WRLCK		2
 #define F_UNLCK		3
 
-/* for old implementation of bsd flock () */
-#define F_EXLCK		4	/* or 3 */
-#define F_SHLCK		8	/* or 4 */
-
-/* for leases */
-#define F_INPROGRESS	16
-
-/* operations for bsd flock(), also used by the kernel implementation */
-#define LOCK_SH		1	/* shared lock */
-#define LOCK_EX		2	/* exclusive lock */
-#define LOCK_NB		4	/* or'd with one of the above to prevent
-				   blocking */
-#define LOCK_UN		8	/* remove lock */
-
-#define LOCK_MAND	32	/* This is a mandatory flock */
-#define LOCK_READ	64	/* ... Which allows concurrent read operations */
-#define LOCK_WRITE	128	/* ... Which allows concurrent write operations */
-#define LOCK_RW		192	/* ... Which allows concurrent read & write ops */
-
-struct flock {
-	short l_type;
-	short l_whence;
-	off_t l_start;
-	off_t l_len;
-	pid_t l_pid;
-	short __unused;
-};
+#define __ARCH_FLOCK_PAD	short __unused;
 
-#define F_LINUX_SPECIFIC_BASE	1024
+#include <asm-generic/fcntl.h>
 
 #endif /* !(_SPARC64_FCNTL_H) */
diff --git a/include/asm-sparc64/futex.h b/include/asm-sparc64/futex.h
new file mode 100644
index 00000000000..9feff4ce142
--- /dev/null
+++ b/include/asm-sparc64/futex.h
@@ -0,0 +1,53 @@
+#ifndef _ASM_FUTEX_H
+#define _ASM_FUTEX_H
+
+#ifdef __KERNEL__
+
+#include <linux/futex.h>
+#include <asm/errno.h>
+#include <asm/uaccess.h>
+
+static inline int
+futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+{
+	int op = (encoded_op >> 28) & 7;
+	int cmp = (encoded_op >> 24) & 15;
+	int oparg = (encoded_op << 8) >> 20;
+	int cmparg = (encoded_op << 20) >> 20;
+	int oldval = 0, ret;
+	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
+		oparg = 1 << oparg;
+
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+		return -EFAULT;
+
+	inc_preempt_count();
+
+	switch (op) {
+	case FUTEX_OP_SET:
+	case FUTEX_OP_ADD:
+	case FUTEX_OP_OR:
+	case FUTEX_OP_ANDN:
+	case FUTEX_OP_XOR:
+	default:
+		ret = -ENOSYS;
+	}
+
+	dec_preempt_count();
+
+	if (!ret) {
+		switch (cmp) {
+		case FUTEX_OP_CMP_EQ: ret = (oldval == cmparg); break;
+		case FUTEX_OP_CMP_NE: ret = (oldval != cmparg); break;
+		case FUTEX_OP_CMP_LT: ret = (oldval < cmparg); break;
+		case FUTEX_OP_CMP_GE: ret = (oldval >= cmparg); break;
+		case FUTEX_OP_CMP_LE: ret = (oldval <= cmparg); break;
+		case FUTEX_OP_CMP_GT: ret = (oldval > cmparg); break;
+		default: ret = -ENOSYS;
+		}
+	}
+	return ret;
+}
+
+#endif
+#endif
diff --git a/include/asm-sparc64/hardirq.h b/include/asm-sparc64/hardirq.h
index d6db1aed764..f0cf71376ec 100644
--- a/include/asm-sparc64/hardirq.h
+++ b/include/asm-sparc64/hardirq.h
@@ -1,22 +1,16 @@
 /* hardirq.h: 64-bit Sparc hard IRQ support.
  *
- * Copyright (C) 1997, 1998 David S. Miller (davem@caip.rutgers.edu)
+ * Copyright (C) 1997, 1998, 2005 David S. Miller (davem@davemloft.net)
  */
 
 #ifndef __SPARC64_HARDIRQ_H
 #define __SPARC64_HARDIRQ_H
 
-#include <linux/config.h>
-#include <linux/threads.h>
-#include <linux/spinlock.h>
-#include <linux/cache.h>
+#include <asm/cpudata.h>
 
-/* rtrap.S is sensitive to the offsets of these fields */
-typedef struct {
-	unsigned int __softirq_pending;
-} ____cacheline_aligned irq_cpustat_t;
-
-#include <linux/irq_cpustat.h>	/* Standard mappings for irq_cpustat_t above */
+#define __ARCH_IRQ_STAT
+#define local_softirq_pending() \
+	(local_cpu_data().__softirq_pending)
 
 #define HARDIRQ_BITS	8
 
diff --git a/include/asm-sparc64/hdreg.h b/include/asm-sparc64/hdreg.h
deleted file mode 100644
index 7f7fd1af0af..00000000000
--- a/include/asm-sparc64/hdreg.h
+++ /dev/null
@@ -1 +0,0 @@
-#include <asm-generic/hdreg.h>
diff --git a/include/asm-sparc64/ide.h b/include/asm-sparc64/ide.h
index 4c1098474c7..c393f815b0b 100644
--- a/include/asm-sparc64/ide.h
+++ b/include/asm-sparc64/ide.h
@@ -15,6 +15,7 @@
 #include <asm/io.h>
 #include <asm/spitfire.h>
 #include <asm/cacheflush.h>
+#include <asm/page.h>
 
 #ifndef MAX_HWIFS
 # ifdef CONFIG_BLK_DEV_IDEPCI
diff --git a/include/asm-sparc64/io.h b/include/asm-sparc64/io.h
index afdcea90707..0056770e83a 100644
--- a/include/asm-sparc64/io.h
+++ b/include/asm-sparc64/io.h
@@ -100,18 +100,41 @@ static __inline__ void _outl(u32 l, unsigned long addr)
 #define inl_p(__addr)		inl(__addr)
 #define outl_p(__l, __addr)	outl(__l, __addr)
 
-extern void outsb(void __iomem *addr, const void *src, unsigned long count);
-extern void outsw(void __iomem *addr, const void *src, unsigned long count);
-extern void outsl(void __iomem *addr, const void *src, unsigned long count);
-extern void insb(void __iomem *addr, void *dst, unsigned long count);
-extern void insw(void __iomem *addr, void *dst, unsigned long count);
-extern void insl(void __iomem *addr, void *dst, unsigned long count);
-#define ioread8_rep(a,d,c)	insb(a,d,c)
-#define ioread16_rep(a,d,c)	insw(a,d,c)
-#define ioread32_rep(a,d,c)	insl(a,d,c)
-#define iowrite8_rep(a,s,c)	outsb(a,s,c)
-#define iowrite16_rep(a,s,c)	outsw(a,s,c)
-#define iowrite32_rep(a,s,c)	outsl(a,s,c)
+extern void outsb(unsigned long, const void *, unsigned long);
+extern void outsw(unsigned long, const void *, unsigned long);
+extern void outsl(unsigned long, const void *, unsigned long);
+extern void insb(unsigned long, void *, unsigned long);
+extern void insw(unsigned long, void *, unsigned long);
+extern void insl(unsigned long, void *, unsigned long);
+
+static inline void ioread8_rep(void __iomem *port, void *buf, unsigned long count)
+{
+	insb((unsigned long __force)port, buf, count);
+}
+static inline void ioread16_rep(void __iomem *port, void *buf, unsigned long count)
+{
+	insw((unsigned long __force)port, buf, count);
+}
+
+static inline void ioread32_rep(void __iomem *port, void *buf, unsigned long count)
+{
+	insl((unsigned long __force)port, buf, count);
+}
+
+static inline void iowrite8_rep(void __iomem *port, const void *buf, unsigned long count)
+{
+	outsb((unsigned long __force)port, buf, count);
+}
+
+static inline void iowrite16_rep(void __iomem *port, const void *buf, unsigned long count)
+{
+	outsw((unsigned long __force)port, buf, count);
+}
+
+static inline void iowrite32_rep(void __iomem *port, const void *buf, unsigned long count)
+{
+	outsl((unsigned long __force)port, buf, count);
+}
 
 /* Memory functions, same as I/O accesses on Ultra. */
 static inline u8 _readb(const volatile void __iomem *addr)
diff --git a/include/asm-sparc64/page.h b/include/asm-sparc64/page.h
index b87dbbd64bc..7f8d764abc4 100644
--- a/include/asm-sparc64/page.h
+++ b/include/asm-sparc64/page.h
@@ -21,6 +21,13 @@
 #define PAGE_SIZE    (_AC(1,UL) << PAGE_SHIFT)
 #define PAGE_MASK    (~(PAGE_SIZE-1))
 
+/* Flushing for D-cache alias handling is only needed if
+ * the page size is smaller than 16K.
+ */
+#if PAGE_SHIFT < 14
+#define DCACHE_ALIASING_POSSIBLE
+#endif
+
 #ifdef __KERNEL__
 
 #ifndef __ASSEMBLY__
@@ -150,20 +157,6 @@ struct sparc_phys_banks {
 
 extern struct sparc_phys_banks sp_banks[SPARC_PHYS_BANKS];
 
-/* Pure 2^n version of get_order */
-static __inline__ int get_order(unsigned long size)
-{
-	int order;
-
-	size = (size-1) >> (PAGE_SHIFT-1);
-	order = -1;
-	do {
-		size >>= 1;
-		order++;
-	} while (size);
-	return order;
-}
-
 #endif /* !(__ASSEMBLY__) */
 
 #define VM_DATA_DEFAULT_FLAGS	(VM_READ | VM_WRITE | VM_EXEC | \
@@ -171,4 +164,6 @@ static __inline__ int get_order(unsigned long size)
 
 #endif /* !(__KERNEL__) */
 
+#include <asm-generic/page.h>
+
 #endif /* !(_SPARC64_PAGE_H) */
diff --git a/include/asm-sparc64/pci.h b/include/asm-sparc64/pci.h
index a4ab0ec7143..89bd71b1c0d 100644
--- a/include/asm-sparc64/pci.h
+++ b/include/asm-sparc64/pci.h
@@ -269,6 +269,8 @@ extern void
 pcibios_bus_to_resource(struct pci_dev *dev, struct resource *res,
 			struct pci_bus_region *region);
 
+extern struct resource *pcibios_select_root(struct pci_dev *, struct resource *);
+
 static inline void pcibios_add_platform_entries(struct pci_dev *dev)
 {
 }
diff --git a/include/asm-sparc64/pgalloc.h b/include/asm-sparc64/pgalloc.h
index b9b1914aae6..a96067cca96 100644
--- a/include/asm-sparc64/pgalloc.h
+++ b/include/asm-sparc64/pgalloc.h
@@ -10,6 +10,7 @@
 #include <asm/spitfire.h>
 #include <asm/cpudata.h>
 #include <asm/cacheflush.h>
+#include <asm/page.h>
 
 /* Page table allocation/freeing. */
 #ifdef CONFIG_SMP
diff --git a/include/asm-sparc64/pgtable.h b/include/asm-sparc64/pgtable.h
index 1ae00c5087f..a297f6144f0 100644
--- a/include/asm-sparc64/pgtable.h
+++ b/include/asm-sparc64/pgtable.h
@@ -24,21 +24,23 @@
 #include <asm/processor.h>
 #include <asm/const.h>
 
-/* The kernel image occupies 0x4000000 to 0x1000000 (4MB --> 16MB).
- * The page copy blockops use 0x1000000 to 0x18000000 (16MB --> 24MB).
+/* The kernel image occupies 0x4000000 to 0x1000000 (4MB --> 32MB).
+ * The page copy blockops can use 0x2000000 to 0x10000000.
  * The PROM resides in an area spanning 0xf0000000 to 0x100000000.
- * The vmalloc area spans 0x140000000 to 0x200000000.
+ * The vmalloc area spans 0x100000000 to 0x200000000.
+ * Since modules need to be in the lowest 32-bits of the address space,
+ * we place them right before the OBP area from 0x10000000 to 0xf0000000.
  * There is a single static kernel PMD which maps from 0x0 to address
  * 0x400000000.
  */
-#define	TLBTEMP_BASE		_AC(0x0000000001000000,UL)
-#define MODULES_VADDR		_AC(0x0000000002000000,UL)
-#define MODULES_LEN		_AC(0x000000007e000000,UL)
-#define MODULES_END		_AC(0x0000000080000000,UL)
-#define VMALLOC_START		_AC(0x0000000140000000,UL)
-#define VMALLOC_END		_AC(0x0000000200000000,UL)
+#define	TLBTEMP_BASE		_AC(0x0000000002000000,UL)
+#define MODULES_VADDR		_AC(0x0000000010000000,UL)
+#define MODULES_LEN		_AC(0x00000000e0000000,UL)
+#define MODULES_END		_AC(0x00000000f0000000,UL)
 #define LOW_OBP_ADDRESS		_AC(0x00000000f0000000,UL)
 #define HI_OBP_ADDRESS		_AC(0x0000000100000000,UL)
+#define VMALLOC_START		_AC(0x0000000100000000,UL)
+#define VMALLOC_END		_AC(0x0000000200000000,UL)
 
 /* XXX All of this needs to be rethought so we can take advantage
  * XXX cheetah's full 64-bit virtual address space, ie. no more hole
@@ -410,9 +412,6 @@ extern unsigned long *sparc64_valid_addr_bitmap;
 #define kern_addr_valid(addr)	\
 	(test_bit(__pa((unsigned long)(addr))>>22, sparc64_valid_addr_bitmap))
 
-extern int io_remap_page_range(struct vm_area_struct *vma, unsigned long from,
-			       unsigned long offset,
-			       unsigned long size, pgprot_t prot, int space);
 extern int io_remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
 			       unsigned long pfn,
 			       unsigned long size, pgprot_t prot);
diff --git a/include/asm-sparc64/spinlock.h b/include/asm-sparc64/spinlock.h
index a02c4370eb4..ec85d12d73b 100644
--- a/include/asm-sparc64/spinlock.h
+++ b/include/asm-sparc64/spinlock.h
@@ -29,24 +29,13 @@
  * must be pre-V9 branches.
  */
 
-#ifndef CONFIG_DEBUG_SPINLOCK
+#define __raw_spin_is_locked(lp)	((lp)->lock != 0)
 
-typedef struct {
-	volatile unsigned char lock;
-#ifdef CONFIG_PREEMPT
-	unsigned int break_lock;
-#endif
-} spinlock_t;
-#define SPIN_LOCK_UNLOCKED	(spinlock_t) {0,}
+#define __raw_spin_unlock_wait(lp)	\
+	do {	rmb();			\
+	} while((lp)->lock)
 
-#define spin_lock_init(lp)	do { *(lp)= SPIN_LOCK_UNLOCKED; } while(0)
-#define spin_is_locked(lp)  ((lp)->lock != 0)
-
-#define spin_unlock_wait(lp)	\
-do {	rmb();			\
-} while((lp)->lock)
-
-static inline void _raw_spin_lock(spinlock_t *lock)
+static inline void __raw_spin_lock(raw_spinlock_t *lock)
 {
 	unsigned long tmp;
 
@@ -67,7 +56,7 @@ static inline void _raw_spin_lock(spinlock_t *lock)
 	: "memory");
 }
 
-static inline int _raw_spin_trylock(spinlock_t *lock)
+static inline int __raw_spin_trylock(raw_spinlock_t *lock)
 {
 	unsigned long result;
 
@@ -81,7 +70,7 @@ static inline int _raw_spin_trylock(spinlock_t *lock)
 	return (result == 0UL);
 }
 
-static inline void _raw_spin_unlock(spinlock_t *lock)
+static inline void __raw_spin_unlock(raw_spinlock_t *lock)
 {
 	__asm__ __volatile__(
 "	membar		#StoreStore | #LoadStore\n"
@@ -91,7 +80,7 @@ static inline void _raw_spin_unlock(spinlock_t *lock)
 	: "memory");
 }
 
-static inline void _raw_spin_lock_flags(spinlock_t *lock, unsigned long flags)
+static inline void __raw_spin_lock_flags(raw_spinlock_t *lock, unsigned long flags)
 {
 	unsigned long tmp1, tmp2;
 
@@ -115,51 +104,9 @@ static inline void _raw_spin_lock_flags(spinlock_t *lock, unsigned long flags)
 	: "memory");
 }
 
-#else /* !(CONFIG_DEBUG_SPINLOCK) */
-
-typedef struct {
-	volatile unsigned char lock;
-	unsigned int owner_pc, owner_cpu;
-#ifdef CONFIG_PREEMPT
-	unsigned int break_lock;
-#endif
-} spinlock_t;
-#define SPIN_LOCK_UNLOCKED (spinlock_t) { 0, 0, 0xff }
-#define spin_lock_init(lp)	do { *(lp)= SPIN_LOCK_UNLOCKED; } while(0)
-#define spin_is_locked(__lock)	((__lock)->lock != 0)
-#define spin_unlock_wait(__lock)	\
-do { \
-	rmb(); \
-} while((__lock)->lock)
-
-extern void _do_spin_lock(spinlock_t *lock, char *str, unsigned long caller);
-extern void _do_spin_unlock(spinlock_t *lock);
-extern int _do_spin_trylock(spinlock_t *lock, unsigned long caller);
-
-#define _raw_spin_trylock(lp)	\
-	_do_spin_trylock(lp, (unsigned long) __builtin_return_address(0))
-#define _raw_spin_lock(lock)	\
-	_do_spin_lock(lock, "spin_lock", \
-		      (unsigned long) __builtin_return_address(0))
-#define _raw_spin_unlock(lock)	_do_spin_unlock(lock)
-#define _raw_spin_lock_flags(lock, flags) _raw_spin_lock(lock)
-
-#endif /* CONFIG_DEBUG_SPINLOCK */
-
 /* Multi-reader locks, these are much saner than the 32-bit Sparc ones... */
 
-#ifndef CONFIG_DEBUG_SPINLOCK
-
-typedef struct {
-	volatile unsigned int lock;
-#ifdef CONFIG_PREEMPT
-	unsigned int break_lock;
-#endif
-} rwlock_t;
-#define RW_LOCK_UNLOCKED	(rwlock_t) {0,}
-#define rwlock_init(lp) do { *(lp) = RW_LOCK_UNLOCKED; } while(0)
-
-static void inline __read_lock(rwlock_t *lock)
+static void inline __read_lock(raw_rwlock_t *lock)
 {
 	unsigned long tmp1, tmp2;
 
@@ -184,7 +131,7 @@ static void inline __read_lock(rwlock_t *lock)
 	: "memory");
 }
 
-static void inline __read_unlock(rwlock_t *lock)
+static void inline __read_unlock(raw_rwlock_t *lock)
 {
 	unsigned long tmp1, tmp2;
 
@@ -201,7 +148,7 @@ static void inline __read_unlock(rwlock_t *lock)
 	: "memory");
 }
 
-static void inline __write_lock(rwlock_t *lock)
+static void inline __write_lock(raw_rwlock_t *lock)
 {
 	unsigned long mask, tmp1, tmp2;
 
@@ -228,7 +175,7 @@ static void inline __write_lock(rwlock_t *lock)
 	: "memory");
 }
 
-static void inline __write_unlock(rwlock_t *lock)
+static void inline __write_unlock(raw_rwlock_t *lock)
 {
 	__asm__ __volatile__(
 "	membar		#LoadStore | #StoreStore\n"
@@ -238,7 +185,7 @@ static void inline __write_unlock(rwlock_t *lock)
 	: "memory");
 }
 
-static int inline __write_trylock(rwlock_t *lock)
+static int inline __write_trylock(raw_rwlock_t *lock)
 {
 	unsigned long mask, tmp1, tmp2, result;
 
@@ -263,78 +210,15 @@ static int inline __write_trylock(rwlock_t *lock)
 	return result;
 }
 
-#define _raw_read_lock(p)	__read_lock(p)
-#define _raw_read_unlock(p)	__read_unlock(p)
-#define _raw_write_lock(p)	__write_lock(p)
-#define _raw_write_unlock(p)	__write_unlock(p)
-#define _raw_write_trylock(p)	__write_trylock(p)
-
-#else /* !(CONFIG_DEBUG_SPINLOCK) */
-
-typedef struct {
-	volatile unsigned long lock;
-	unsigned int writer_pc, writer_cpu;
-	unsigned int reader_pc[NR_CPUS];
-#ifdef CONFIG_PREEMPT
-	unsigned int break_lock;
-#endif
-} rwlock_t;
-#define RW_LOCK_UNLOCKED	(rwlock_t) { 0, 0, 0xff, { } }
-#define rwlock_init(lp) do { *(lp) = RW_LOCK_UNLOCKED; } while(0)
-
-extern void _do_read_lock(rwlock_t *rw, char *str, unsigned long caller);
-extern void _do_read_unlock(rwlock_t *rw, char *str, unsigned long caller);
-extern void _do_write_lock(rwlock_t *rw, char *str, unsigned long caller);
-extern void _do_write_unlock(rwlock_t *rw, unsigned long caller);
-extern int _do_write_trylock(rwlock_t *rw, char *str, unsigned long caller);
-
-#define _raw_read_lock(lock) \
-do {	unsigned long flags; \
-	local_irq_save(flags); \
-	_do_read_lock(lock, "read_lock", \
-		      (unsigned long) __builtin_return_address(0)); \
-	local_irq_restore(flags); \
-} while(0)
-
-#define _raw_read_unlock(lock) \
-do {	unsigned long flags; \
-	local_irq_save(flags); \
-	_do_read_unlock(lock, "read_unlock", \
-		      (unsigned long) __builtin_return_address(0)); \
-	local_irq_restore(flags); \
-} while(0)
-
-#define _raw_write_lock(lock) \
-do {	unsigned long flags; \
-	local_irq_save(flags); \
-	_do_write_lock(lock, "write_lock", \
-		      (unsigned long) __builtin_return_address(0)); \
-	local_irq_restore(flags); \
-} while(0)
-
-#define _raw_write_unlock(lock) \
-do {	unsigned long flags; \
-	local_irq_save(flags); \
-	_do_write_unlock(lock, \
-		      (unsigned long) __builtin_return_address(0)); \
-	local_irq_restore(flags); \
-} while(0)
-
-#define _raw_write_trylock(lock) \
-({	unsigned long flags; \
-	int val; \
-	local_irq_save(flags); \
-	val = _do_write_trylock(lock, "write_trylock", \
-				(unsigned long) __builtin_return_address(0)); \
-	local_irq_restore(flags); \
-	val; \
-})
-
-#endif /* CONFIG_DEBUG_SPINLOCK */
-
-#define _raw_read_trylock(lock) generic_raw_read_trylock(lock)
-#define read_can_lock(rw)	(!((rw)->lock & 0x80000000UL))
-#define write_can_lock(rw)	(!(rw)->lock)
+#define __raw_read_lock(p)	__read_lock(p)
+#define __raw_read_unlock(p)	__read_unlock(p)
+#define __raw_write_lock(p)	__write_lock(p)
+#define __raw_write_unlock(p)	__write_unlock(p)
+#define __raw_write_trylock(p)	__write_trylock(p)
+
+#define __raw_read_trylock(lock)	generic__raw_read_trylock(lock)
+#define __raw_read_can_lock(rw)		(!((rw)->lock & 0x80000000UL))
+#define __raw_write_can_lock(rw)	(!(rw)->lock)
 
 #endif /* !(__ASSEMBLY__) */
 
diff --git a/include/asm-sparc64/spinlock_types.h b/include/asm-sparc64/spinlock_types.h
new file mode 100644
index 00000000000..e128112a0d7
--- /dev/null
+++ b/include/asm-sparc64/spinlock_types.h
@@ -0,0 +1,20 @@
+#ifndef __SPARC64_SPINLOCK_TYPES_H
+#define __SPARC64_SPINLOCK_TYPES_H
+
+#ifndef __LINUX_SPINLOCK_TYPES_H
+# error "please don't include this file directly"
+#endif
+
+typedef struct {
+	volatile unsigned char lock;
+} raw_spinlock_t;
+
+#define __RAW_SPIN_LOCK_UNLOCKED	{ 0 }
+
+typedef struct {
+	volatile unsigned int lock;
+} raw_rwlock_t;
+
+#define __RAW_RW_LOCK_UNLOCKED		{ 0 }
+
+#endif
diff --git a/include/asm-sparc64/system.h b/include/asm-sparc64/system.h
index 5e94c05dc2f..b5417529f6f 100644
--- a/include/asm-sparc64/system.h
+++ b/include/asm-sparc64/system.h
@@ -28,13 +28,48 @@ enum sparc_cpu {
 #define ARCH_SUN4C_SUN4 0
 #define ARCH_SUN4 0
 
-extern void mb(void);
-extern void rmb(void);
-extern void wmb(void);
-extern void membar_storeload(void);
-extern void membar_storeload_storestore(void);
-extern void membar_storeload_loadload(void);
-extern void membar_storestore_loadstore(void);
+/* These are here in an effort to more fully work around Spitfire Errata
+ * #51.  Essentially, if a memory barrier occurs soon after a mispredicted
+ * branch, the chip can stop executing instructions until a trap occurs.
+ * Therefore, if interrupts are disabled, the chip can hang forever.
+ *
+ * It used to be believed that the memory barrier had to be right in the
+ * delay slot, but a case has been traced recently wherein the memory barrier
+ * was one instruction after the branch delay slot and the chip still hung.
+ * The offending sequence was the following in sym_wakeup_done() of the
+ * sym53c8xx_2 driver:
+ *
+ *	call	sym_ccb_from_dsa, 0
+ *	 movge	%icc, 0, %l0
+ *	brz,pn	%o0, .LL1303
+ *	 mov	%o0, %l2
+ *	membar	#LoadLoad
+ *
+ * The branch has to be mispredicted for the bug to occur.  Therefore, we put
+ * the memory barrier explicitly into a "branch always, predicted taken"
+ * delay slot to avoid the problem case.
+ */
+#define membar_safe(type) \
+do {	__asm__ __volatile__("ba,pt	%%xcc, 1f\n\t" \
+			     " membar	" type "\n" \
+			     "1:\n" \
+			     : : : "memory"); \
+} while (0)
+
+#define mb()	\
+	membar_safe("#LoadLoad | #LoadStore | #StoreStore | #StoreLoad")
+#define rmb()	\
+	membar_safe("#LoadLoad")
+#define wmb()	\
+	membar_safe("#StoreStore")
+#define membar_storeload() \
+	membar_safe("#StoreLoad")
+#define membar_storeload_storestore() \
+	membar_safe("#StoreLoad | #StoreStore")
+#define membar_storeload_loadload() \
+	membar_safe("#StoreLoad | #LoadLoad")
+#define membar_storestore_loadstore() \
+	membar_safe("#StoreStore | #LoadStore")
 
 #endif
 
diff --git a/include/asm-sparc64/types.h b/include/asm-sparc64/types.h
index 6248ed1a9a7..d0ee7f10583 100644
--- a/include/asm-sparc64/types.h
+++ b/include/asm-sparc64/types.h
@@ -56,8 +56,6 @@ typedef unsigned long u64;
 typedef u32 dma_addr_t;
 typedef u64 dma64_addr_t;
 
-typedef unsigned short kmem_bufctl_t;
-
 #endif /* __ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
diff --git a/include/asm-sparc64/uaccess.h b/include/asm-sparc64/uaccess.h
index 5690142f82d..80a65d7e3db 100644
--- a/include/asm-sparc64/uaccess.h
+++ b/include/asm-sparc64/uaccess.h
@@ -59,12 +59,6 @@ static inline int access_ok(int type, const void __user * addr, unsigned long si
 	return 1;
 }
 
-/* this function will go away soon - use access_ok() instead */
-static inline int __deprecated verify_area(int type, const void __user * addr, unsigned long size)
-{
-	return 0;
-}
-
 /*
  * The exception table consists of pairs of addresses: the first is the
  * address of an instruction that is allowed to fault, and the second is