Merge v2.6.37-rc8 into powerpc/next

29 files changed, 337 insertions, 442 deletions

diff --git a/arch/tile/include/asm/atomic_32.h b/arch/tile/include/asm/atomic_32.h
index 40a5a3a876d9..ed359aee8837 100644
--- a/arch/tile/include/asm/atomic_32.h
+++ b/arch/tile/include/asm/atomic_32.h
@@ -255,43 +255,6 @@ static inline void atomic64_set(atomic64_t *v, u64 n)
 #define smp_mb__after_atomic_dec()	do { } while (0)
 #define smp_mb__after_atomic_inc()	do { } while (0)
 
-
-/*
- * Support "tns" atomic integers.  These are atomic integers that can
- * hold any value but "1".  They are more efficient than regular atomic
- * operations because the "lock" (aka acquire) step is a single "tns"
- * in the uncontended case, and the "unlock" (aka release) step is a
- * single "store" without an mf.  (However, note that on tilepro the
- * "tns" will evict the local cache line, so it's not all upside.)
- *
- * Note that you can ONLY observe the value stored in the pointer
- * using these operations; a direct read of the value may confusingly
- * return the special value "1".
- */
-
-int __tns_atomic_acquire(atomic_t *);
-void __tns_atomic_release(atomic_t *p, int v);
-
-static inline void tns_atomic_set(atomic_t *v, int i)
-{
-	__tns_atomic_acquire(v);
-	__tns_atomic_release(v, i);
-}
-
-static inline int tns_atomic_cmpxchg(atomic_t *v, int o, int n)
-{
-	int ret = __tns_atomic_acquire(v);
-	__tns_atomic_release(v, (ret == o) ? n : ret);
-	return ret;
-}
-
-static inline int tns_atomic_xchg(atomic_t *v, int n)
-{
-	int ret = __tns_atomic_acquire(v);
-	__tns_atomic_release(v, n);
-	return ret;
-}
-
 #endif /* !__ASSEMBLY__ */
 
 /*
diff --git a/arch/tile/include/asm/backtrace.h b/arch/tile/include/asm/backtrace.h
index 6970bfcad549..f18887d82399 100644
--- a/arch/tile/include/asm/backtrace.h
+++ b/arch/tile/include/asm/backtrace.h
@@ -21,7 +21,9 @@
 
 #include <arch/chip.h>
 
-#if CHIP_VA_WIDTH() > 32
+#if defined(__tile__)
+typedef unsigned long VirtualAddress;
+#elif CHIP_VA_WIDTH() > 32
 typedef unsigned long long VirtualAddress;
 #else
 typedef unsigned int VirtualAddress;
@@ -144,7 +146,10 @@ enum {
 
 	CALLER_SP_IN_R52_BASE = 4,
 
-	CALLER_SP_OFFSET_BASE = 8
+	CALLER_SP_OFFSET_BASE = 8,
+
+	/* Marks the entry point of certain functions. */
+	ENTRY_POINT_INFO_OP = 16
 };
 
 
diff --git a/arch/tile/include/asm/bitops.h b/arch/tile/include/asm/bitops.h
index 84600f3514da..6d4f0ff2c68c 100644
--- a/arch/tile/include/asm/bitops.h
+++ b/arch/tile/include/asm/bitops.h
@@ -98,27 +98,29 @@ static inline int fls64(__u64 w)
 	return (sizeof(__u64) * 8) - __builtin_clzll(w);
 }
 
-static inline unsigned int hweight32(unsigned int w)
+static inline unsigned int __arch_hweight32(unsigned int w)
 {
 	return __builtin_popcount(w);
 }
 
-static inline unsigned int hweight16(unsigned int w)
+static inline unsigned int __arch_hweight16(unsigned int w)
 {
 	return __builtin_popcount(w & 0xffff);
 }
 
-static inline unsigned int hweight8(unsigned int w)
+static inline unsigned int __arch_hweight8(unsigned int w)
 {
 	return __builtin_popcount(w & 0xff);
 }
 
-static inline unsigned long hweight64(__u64 w)
+static inline unsigned long __arch_hweight64(__u64 w)
 {
 	return __builtin_popcountll(w);
 }
 
+#include <asm-generic/bitops/const_hweight.h>
 #include <asm-generic/bitops/lock.h>
+#include <asm-generic/bitops/find.h>
 #include <asm-generic/bitops/sched.h>
 #include <asm-generic/bitops/ext2-non-atomic.h>
 #include <asm-generic/bitops/minix.h>
diff --git a/arch/tile/include/asm/cache.h b/arch/tile/include/asm/cache.h
index f6101840c9e7..08a2815b5e4e 100644
--- a/arch/tile/include/asm/cache.h
+++ b/arch/tile/include/asm/cache.h
@@ -27,11 +27,10 @@
 #define L2_CACHE_ALIGN(x)	(((x)+(L2_CACHE_BYTES-1)) & -L2_CACHE_BYTES)
 
 /*
- * TILE-Gx is fully coherents so we don't need to define
- * ARCH_KMALLOC_MINALIGN.
+ * TILE-Gx is fully coherent so we don't need to define ARCH_DMA_MINALIGN.
  */
 #ifndef __tilegx__
-#define ARCH_KMALLOC_MINALIGN	L2_CACHE_BYTES
+#define ARCH_DMA_MINALIGN	L2_CACHE_BYTES
 #endif
 
 /* use the cache line size for the L2, which is where it counts */
diff --git a/arch/tile/include/asm/cacheflush.h b/arch/tile/include/asm/cacheflush.h
index c5741da4eeac..14a3f8556ace 100644
--- a/arch/tile/include/asm/cacheflush.h
+++ b/arch/tile/include/asm/cacheflush.h
@@ -137,4 +137,56 @@ static inline void finv_buffer(void *buffer, size_t size)
 	mb_incoherent();
 }
 
+/*
+ * Flush & invalidate a VA range that is homed remotely on a single core,
+ * waiting until the memory controller holds the flushed values.
+ */
+static inline void finv_buffer_remote(void *buffer, size_t size)
+{
+	char *p;
+	int i;
+
+	/*
+	 * Flush and invalidate the buffer out of the local L1/L2
+	 * and request the home cache to flush and invalidate as well.
+	 */
+	__finv_buffer(buffer, size);
+
+	/*
+	 * Wait for the home cache to acknowledge that it has processed
+	 * all the flush-and-invalidate requests.  This does not mean
+	 * that the flushed data has reached the memory controller yet,
+	 * but it does mean the home cache is processing the flushes.
+	 */
+	__insn_mf();
+
+	/*
+	 * Issue a load to the last cache line, which can't complete
+	 * until all the previously-issued flushes to the same memory
+	 * controller have also completed.  If we weren't striping
+	 * memory, that one load would be sufficient, but since we may
+	 * be, we also need to back up to the last load issued to
+	 * another memory controller, which would be the point where
+	 * we crossed an 8KB boundary (the granularity of striping
+	 * across memory controllers).  Keep backing up and doing this
+	 * until we are before the beginning of the buffer, or have
+	 * hit all the controllers.
+	 */
+	for (i = 0, p = (char *)buffer + size - 1;
+	     i < (1 << CHIP_LOG_NUM_MSHIMS()) && p >= (char *)buffer;
+	     ++i) {
+		const unsigned long STRIPE_WIDTH = 8192;
+
+		/* Force a load instruction to issue. */
+		*(volatile char *)p;
+
+		/* Jump to end of previous stripe. */
+		p -= STRIPE_WIDTH;
+		p = (char *)((unsigned long)p | (STRIPE_WIDTH - 1));
+	}
+
+	/* Wait for the loads (and thus flushes) to have completed. */
+	__insn_mf();
+}
+
 #endif /* _ASM_TILE_CACHEFLUSH_H */
diff --git a/arch/tile/include/asm/compat.h b/arch/tile/include/asm/compat.h
index 5a34da6cdd79..c3ae570c0a5d 100644
--- a/arch/tile/include/asm/compat.h
+++ b/arch/tile/include/asm/compat.h
@@ -195,7 +195,7 @@ static inline unsigned long ptr_to_compat_reg(void __user *uptr)
 	return (long)(int)(long __force)uptr;
 }
 
-static inline void __user *compat_alloc_user_space(long len)
+static inline void __user *arch_compat_alloc_user_space(long len)
 {
 	struct pt_regs *regs = task_pt_regs(current);
 	return (void __user *)regs->sp - len;
@@ -214,16 +214,18 @@ extern int compat_setup_rt_frame(int sig, struct k_sigaction *ka,
 struct compat_sigaction;
 struct compat_siginfo;
 struct compat_sigaltstack;
-long compat_sys_execve(char __user *path, compat_uptr_t __user *argv,
-		       compat_uptr_t __user *envp);
+long compat_sys_execve(const char __user *path,
+		       const compat_uptr_t __user *argv,
+		       const compat_uptr_t __user *envp, struct pt_regs *);
 long compat_sys_rt_sigaction(int sig, struct compat_sigaction __user *act,
 			     struct compat_sigaction __user *oact,
 			     size_t sigsetsize);
 long compat_sys_rt_sigqueueinfo(int pid, int sig,
 				struct compat_siginfo __user *uinfo);
-long compat_sys_rt_sigreturn(void);
+long compat_sys_rt_sigreturn(struct pt_regs *);
 long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			    struct compat_sigaltstack __user *uoss_ptr);
+			    struct compat_sigaltstack __user *uoss_ptr,
+			    struct pt_regs *);
 long compat_sys_truncate64(char __user *filename, u32 dummy, u32 low, u32 high);
 long compat_sys_ftruncate64(unsigned int fd, u32 dummy, u32 low, u32 high);
 long compat_sys_pread64(unsigned int fd, char __user *ubuf, size_t count,
@@ -254,4 +256,12 @@ long tile_compat_sys_ptrace(compat_long_t request, compat_long_t pid,
 /* Tilera Linux syscalls that don't have "compat" versions. */
 #define compat_sys_flush_cache sys_flush_cache
 
+/* These are the intvec_64.S trampolines. */
+long _compat_sys_execve(const char __user *path,
+			const compat_uptr_t __user *argv,
+			const compat_uptr_t __user *envp);
+long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
+			    struct compat_sigaltstack __user *uoss_ptr);
+long _compat_sys_rt_sigreturn(void);
+
 #endif /* _ASM_TILE_COMPAT_H */
diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h
index cf466b39aa13..15e1dceecc64 100644
--- a/arch/tile/include/asm/dma-mapping.h
+++ b/arch/tile/include/asm/dma-mapping.h
@@ -90,13 +90,4 @@ dma_set_mask(struct device *dev, u64 mask)
 	return 0;
 }
 
-static inline int
-dma_get_cache_alignment(void)
-{
-	return L2_CACHE_BYTES;
-}
-
-#define dma_is_consistent(d, h)	(1)
-
-
 #endif /* _ASM_TILE_DMA_MAPPING_H */
diff --git a/arch/tile/include/asm/highmem.h b/arch/tile/include/asm/highmem.h
index efdd12e91020..b2a6c5de79ab 100644
--- a/arch/tile/include/asm/highmem.h
+++ b/arch/tile/include/asm/highmem.h
@@ -23,7 +23,6 @@
 
 #include <linux/interrupt.h>
 #include <linux/threads.h>
-#include <asm/kmap_types.h>
 #include <asm/tlbflush.h>
 #include <asm/homecache.h>
 
@@ -60,12 +59,12 @@ void *kmap_fix_kpte(struct page *page, int finished);
 /* This macro is used only in map_new_virtual() to map "page". */
 #define kmap_prot page_to_kpgprot(page)
 
-void kunmap_atomic(void *kvaddr, enum km_type type);
-void *kmap_atomic_pfn(unsigned long pfn, enum km_type type);
-void *kmap_atomic_prot_pfn(unsigned long pfn, enum km_type type, pgprot_t prot);
+void *__kmap_atomic(struct page *page);
+void __kunmap_atomic(void *kvaddr);
+void *kmap_atomic_pfn(unsigned long pfn);
+void *kmap_atomic_prot_pfn(unsigned long pfn, pgprot_t prot);
 struct page *kmap_atomic_to_page(void *ptr);
-void *kmap_atomic_prot(struct page *page, enum km_type type, pgprot_t prot);
-void *kmap_atomic(struct page *page, enum km_type type);
+void *kmap_atomic_prot(struct page *page, pgprot_t prot);
 void kmap_atomic_fix_kpte(struct page *page, int finished);
 
 #define flush_cache_kmaps()	do { } while (0)
diff --git a/arch/tile/include/asm/io.h b/arch/tile/include/asm/io.h
index 8c95bef3fa45..d3cbb9b14cbe 100644
--- a/arch/tile/include/asm/io.h
+++ b/arch/tile/include/asm/io.h
@@ -55,9 +55,6 @@ extern void iounmap(volatile void __iomem *addr);
 #define ioremap_writethrough(physaddr, size)	ioremap(physaddr, size)
 #define ioremap_fullcache(physaddr, size)	ioremap(physaddr, size)
 
-void __iomem *ioport_map(unsigned long port, unsigned int len);
-extern inline void ioport_unmap(void __iomem *addr) {}
-
 #define mmiowb()
 
 /* Conversion between virtual and physical mappings.  */
@@ -164,22 +161,22 @@ static inline void _tile_writeq(u64 val, unsigned long addr)
 #define iowrite32 writel
 #define iowrite64 writeq
 
-static inline void *memcpy_fromio(void *dst, void *src, int len)
+static inline void memcpy_fromio(void *dst, const volatile void __iomem *src,
+				 size_t len)
 {
 	int x;
 	BUG_ON((unsigned long)src & 0x3);
 	for (x = 0; x < len; x += 4)
 		*(u32 *)(dst + x) = readl(src + x);
-	return dst;
 }
 
-static inline void *memcpy_toio(void *dst, void *src, int len)
+static inline void memcpy_toio(volatile void __iomem *dst, const void *src,
+				size_t len)
 {
 	int x;
 	BUG_ON((unsigned long)dst & 0x3);
 	for (x = 0; x < len; x += 4)
 		writel(*(u32 *)(src + x), dst + x);
-	return dst;
 }
 
 /*
@@ -189,12 +186,22 @@ static inline void *memcpy_toio(void *dst, void *src, int len)
  * we never run, uses them unconditionally.
  */
 
-static inline int ioport_panic(void)
+static inline long ioport_panic(void)
 {
 	panic("inb/outb and friends do not exist on tile");
 	return 0;
 }
 
+static inline void __iomem *ioport_map(unsigned long port, unsigned int len)
+{
+	return (void __iomem *) ioport_panic();
+}
+
+static inline void ioport_unmap(void __iomem *addr)
+{
+	ioport_panic();
+}
+
 static inline u8 inb(unsigned long addr)
 {
 	return ioport_panic();
diff --git a/arch/tile/include/asm/irqflags.h b/arch/tile/include/asm/irqflags.h
index 45cf67c2f286..641e4ff3d805 100644
--- a/arch/tile/include/asm/irqflags.h
+++ b/arch/tile/include/asm/irqflags.h
@@ -47,53 +47,53 @@
 	int __n = (n); \
 	int __mask = 1 << (__n & 0x1f); \
 	if (__n < 32) \
-		__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_0, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_0, __mask); \
 	else \
-		__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_1, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_1, __mask); \
 } while (0)
 #define interrupt_mask_reset(n) do { \
 	int __n = (n); \
 	int __mask = 1 << (__n & 0x1f); \
 	if (__n < 32) \
-		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_0, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_0, __mask); \
 	else \
-		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_1, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_1, __mask); \
 } while (0)
 #define interrupt_mask_check(n) ({ \
 	int __n = (n); \
 	(((__n < 32) ? \
-	 __insn_mfspr(SPR_INTERRUPT_MASK_1_0) : \
-	 __insn_mfspr(SPR_INTERRUPT_MASK_1_1)) \
+	 __insn_mfspr(SPR_INTERRUPT_MASK_K_0) : \
+	 __insn_mfspr(SPR_INTERRUPT_MASK_K_1)) \
 	  >> (__n & 0x1f)) & 1; \
 })
 #define interrupt_mask_set_mask(mask) do { \
 	unsigned long long __m = (mask); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_0, (unsigned long)(__m)); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_1, (unsigned long)(__m>>32)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_0, (unsigned long)(__m)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_1, (unsigned long)(__m>>32)); \
 } while (0)
 #define interrupt_mask_reset_mask(mask) do { \
 	unsigned long long __m = (mask); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_0, (unsigned long)(__m)); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_1, (unsigned long)(__m>>32)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_0, (unsigned long)(__m)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_1, (unsigned long)(__m>>32)); \
 } while (0)
 #else
 #define interrupt_mask_set(n) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1, (1UL << (n)))
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (1UL << (n)))
 #define interrupt_mask_reset(n) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1, (1UL << (n)))
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K, (1UL << (n)))
 #define interrupt_mask_check(n) \
-	((__insn_mfspr(SPR_INTERRUPT_MASK_1) >> (n)) & 1)
+	((__insn_mfspr(SPR_INTERRUPT_MASK_K) >> (n)) & 1)
 #define interrupt_mask_set_mask(mask) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1, (mask))
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (mask))
 #define interrupt_mask_reset_mask(mask) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1, (mask))
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K, (mask))
 #endif
 
 /*
  * The set of interrupts we want active if irqs are enabled.
  * Note that in particular, the tile timer interrupt comes and goes
  * from this set, since we have no other way to turn off the timer.
- * Likewise, INTCTRL_1 is removed and re-added during device
+ * Likewise, INTCTRL_K is removed and re-added during device
  * interrupts, as is the the hardwall UDN_FIREWALL interrupt.
  * We use a low bit (MEM_ERROR) as our sentinel value and make sure it
  * is always claimed as an "active interrupt" so we can query that bit
@@ -103,55 +103,57 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 #define INITIAL_INTERRUPTS_ENABLED INT_MASK(INT_MEM_ERROR)
 
 /* Disable interrupts. */
-#define raw_local_irq_disable() \
+#define arch_local_irq_disable() \
 	interrupt_mask_set_mask(LINUX_MASKABLE_INTERRUPTS)
 
 /* Disable all interrupts, including NMIs. */
-#define raw_local_irq_disable_all() \
+#define arch_local_irq_disable_all() \
 	interrupt_mask_set_mask(-1UL)
 
 /* Re-enable all maskable interrupts. */
-#define raw_local_irq_enable() \
+#define arch_local_irq_enable() \
 	interrupt_mask_reset_mask(__get_cpu_var(interrupts_enabled_mask))
 
 /* Disable or enable interrupts based on flag argument. */
-#define raw_local_irq_restore(disabled) do { \
+#define arch_local_irq_restore(disabled) do { \
 	if (disabled) \
-		raw_local_irq_disable(); \
+		arch_local_irq_disable(); \
 	else \
-		raw_local_irq_enable(); \
+		arch_local_irq_enable(); \
 } while (0)
 
 /* Return true if "flags" argument means interrupts are disabled. */
-#define raw_irqs_disabled_flags(flags) ((flags) != 0)
+#define arch_irqs_disabled_flags(flags) ((flags) != 0)
 
 /* Return true if interrupts are currently disabled. */
-#define raw_irqs_disabled() interrupt_mask_check(INT_MEM_ERROR)
+#define arch_irqs_disabled() interrupt_mask_check(INT_MEM_ERROR)
 
 /* Save whether interrupts are currently disabled. */
-#define raw_local_save_flags(flags) ((flags) = raw_irqs_disabled())
+#define arch_local_save_flags() arch_irqs_disabled()
 
 /* Save whether interrupts are currently disabled, then disable them. */
-#define raw_local_irq_save(flags) \
-	do { raw_local_save_flags(flags); raw_local_irq_disable(); } while (0)
+#define arch_local_irq_save() ({ \
+	unsigned long __flags = arch_local_save_flags(); \
+	arch_local_irq_disable(); \
+	__flags; })
 
 /* Prevent the given interrupt from being enabled next time we enable irqs. */
-#define raw_local_irq_mask(interrupt) \
+#define arch_local_irq_mask(interrupt) \
 	(__get_cpu_var(interrupts_enabled_mask) &= ~INT_MASK(interrupt))
 
 /* Prevent the given interrupt from being enabled immediately. */
-#define raw_local_irq_mask_now(interrupt) do { \
-	raw_local_irq_mask(interrupt); \
+#define arch_local_irq_mask_now(interrupt) do { \
+	arch_local_irq_mask(interrupt); \
 	interrupt_mask_set(interrupt); \
 } while (0)
 
 /* Allow the given interrupt to be enabled next time we enable irqs. */
-#define raw_local_irq_unmask(interrupt) \
+#define arch_local_irq_unmask(interrupt) \
 	(__get_cpu_var(interrupts_enabled_mask) |= INT_MASK(interrupt))
 
 /* Allow the given interrupt to be enabled immediately, if !irqs_disabled. */
-#define raw_local_irq_unmask_now(interrupt) do { \
-	raw_local_irq_unmask(interrupt); \
+#define arch_local_irq_unmask_now(interrupt) do { \
+	arch_local_irq_unmask(interrupt); \
 	if (!irqs_disabled()) \
 		interrupt_mask_reset(interrupt); \
 } while (0)
@@ -168,14 +170,14 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 
 /* Return 0 or 1 to indicate whether interrupts are currently disabled. */
 #define IRQS_DISABLED(tmp)					\
-	mfspr   tmp, INTERRUPT_MASK_1;				\
+	mfspr   tmp, SPR_INTERRUPT_MASK_K;			\
 	andi    tmp, tmp, 1
 
 /* Load up a pointer to &interrupts_enabled_mask. */
 #define GET_INTERRUPTS_ENABLED_MASK_PTR(reg)			\
-	moveli reg, hw2_last(interrupts_enabled_mask); \
-	shl16insli reg, reg, hw1(interrupts_enabled_mask); \
-	shl16insli reg, reg, hw0(interrupts_enabled_mask); \
+	moveli reg, hw2_last(interrupts_enabled_mask);		\
+	shl16insli reg, reg, hw1(interrupts_enabled_mask);	\
+	shl16insli reg, reg, hw0(interrupts_enabled_mask);	\
 	add     reg, reg, tp
 
 /* Disable interrupts. */
@@ -183,18 +185,18 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	moveli  tmp0, hw2_last(LINUX_MASKABLE_INTERRUPTS);	\
 	shl16insli tmp0, tmp0, hw1(LINUX_MASKABLE_INTERRUPTS);	\
 	shl16insli tmp0, tmp0, hw0(LINUX_MASKABLE_INTERRUPTS);	\
-	mtspr   INTERRUPT_MASK_SET_1, tmp0
+	mtspr   SPR_INTERRUPT_MASK_SET_K, tmp0
 
 /* Disable ALL synchronous interrupts (used by NMI entry). */
 #define IRQ_DISABLE_ALL(tmp)					\
 	movei   tmp, -1;					\
-	mtspr   INTERRUPT_MASK_SET_1, tmp
+	mtspr   SPR_INTERRUPT_MASK_SET_K, tmp
 
 /* Enable interrupts. */
 #define IRQ_ENABLE(tmp0, tmp1)					\
 	GET_INTERRUPTS_ENABLED_MASK_PTR(tmp0);			\
 	ld      tmp0, tmp0;					\
-	mtspr   INTERRUPT_MASK_RESET_1, tmp0
+	mtspr   SPR_INTERRUPT_MASK_RESET_K, tmp0
 
 #else /* !__tilegx__ */
 
@@ -208,14 +210,14 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
  * (making the original code's write of the "high" mask word idempotent).
  */
 #define IRQS_DISABLED(tmp)					\
-	mfspr   tmp, INTERRUPT_MASK_1_0;			\
+	mfspr   tmp, SPR_INTERRUPT_MASK_K_0;			\
 	shri    tmp, tmp, INT_MEM_ERROR;			\
 	andi    tmp, tmp, 1
 
 /* Load up a pointer to &interrupts_enabled_mask. */
 #define GET_INTERRUPTS_ENABLED_MASK_PTR(reg)			\
-	moveli  reg, lo16(interrupts_enabled_mask);	\
-	auli    reg, reg, ha16(interrupts_enabled_mask);\
+	moveli  reg, lo16(interrupts_enabled_mask);		\
+	auli    reg, reg, ha16(interrupts_enabled_mask);	\
 	add     reg, reg, tp
 
 /* Disable interrupts. */
@@ -225,16 +227,16 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	 moveli tmp1, lo16(LINUX_MASKABLE_INTERRUPTS)		\
 	};							\
 	{							\
-	 mtspr  INTERRUPT_MASK_SET_1_0, tmp0;			\
+	 mtspr  SPR_INTERRUPT_MASK_SET_K_0, tmp0;		\
 	 auli   tmp1, tmp1, ha16(LINUX_MASKABLE_INTERRUPTS)	\
 	};							\
-	mtspr   INTERRUPT_MASK_SET_1_1, tmp1
+	mtspr   SPR_INTERRUPT_MASK_SET_K_1, tmp1
 
 /* Disable ALL synchronous interrupts (used by NMI entry). */
 #define IRQ_DISABLE_ALL(tmp)					\
 	movei   tmp, -1;					\
-	mtspr   INTERRUPT_MASK_SET_1_0, tmp;			\
-	mtspr   INTERRUPT_MASK_SET_1_1, tmp
+	mtspr   SPR_INTERRUPT_MASK_SET_K_0, tmp;		\
+	mtspr   SPR_INTERRUPT_MASK_SET_K_1, tmp
 
 /* Enable interrupts. */
 #define IRQ_ENABLE(tmp0, tmp1)					\
@@ -244,8 +246,8 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	 addi   tmp1, tmp0, 4					\
 	};							\
 	lw      tmp1, tmp1;					\
-	mtspr   INTERRUPT_MASK_RESET_1_0, tmp0;			\
-	mtspr   INTERRUPT_MASK_RESET_1_1, tmp1
+	mtspr   SPR_INTERRUPT_MASK_RESET_K_0, tmp0;		\
+	mtspr   SPR_INTERRUPT_MASK_RESET_K_1, tmp1
 #endif
 
 /*
diff --git a/arch/tile/include/asm/kmap_types.h b/arch/tile/include/asm/kmap_types.h
index 1480106d1c05..3d0f20246260 100644
--- a/arch/tile/include/asm/kmap_types.h
+++ b/arch/tile/include/asm/kmap_types.h
@@ -16,28 +16,42 @@
 #define _ASM_TILE_KMAP_TYPES_H
 
 /*
- * In TILE Linux each set of four of these uses another 16MB chunk of
- * address space, given 64 tiles and 64KB pages, so we only enable
- * ones that are required by the kernel configuration.
+ * In 32-bit TILE Linux we have to balance the desire to have a lot of
+ * nested atomic mappings with the fact that large page sizes and many
+ * processors chew up address space quickly.  In a typical
+ * 64-processor, 64KB-page layout build, making KM_TYPE_NR one larger
+ * adds 4MB of required address-space.  For now we leave KM_TYPE_NR
+ * set to depth 8.
  */
 enum km_type {
+	KM_TYPE_NR = 8
+};
+
+/*
+ * We provide dummy definitions of all the stray values that used to be
+ * required for kmap_atomic() and no longer are.
+ */
+enum {
 	KM_BOUNCE_READ,
 	KM_SKB_SUNRPC_DATA,
 	KM_SKB_DATA_SOFTIRQ,
 	KM_USER0,
 	KM_USER1,
 	KM_BIO_SRC_IRQ,
+	KM_BIO_DST_IRQ,
+	KM_PTE0,
+	KM_PTE1,
 	KM_IRQ0,
 	KM_IRQ1,
 	KM_SOFTIRQ0,
 	KM_SOFTIRQ1,
-	KM_MEMCPY0,
-	KM_MEMCPY1,
-#if defined(CONFIG_HIGHPTE)
-	KM_PTE0,
-	KM_PTE1,
-#endif
-	KM_TYPE_NR
+	KM_SYNC_ICACHE,
+	KM_SYNC_DCACHE,
+	KM_UML_USERCOPY,
+	KM_IRQ_PTE,
+	KM_NMI,
+	KM_NMI_PTE,
+	KM_KDB
 };
 
 #endif /* _ASM_TILE_KMAP_TYPES_H */
diff --git a/arch/tile/include/asm/mman.h b/arch/tile/include/asm/mman.h
index 4c6811e3e8dc..81b8fc348d63 100644
--- a/arch/tile/include/asm/mman.h
+++ b/arch/tile/include/asm/mman.h
@@ -23,6 +23,7 @@
 #define MAP_POPULATE	0x0040		/* populate (prefault) pagetables */
 #define MAP_NONBLOCK	0x0080		/* do not block on IO */
 #define MAP_GROWSDOWN	0x0100		/* stack-like segment */
+#define MAP_STACK	MAP_GROWSDOWN	/* provide convenience alias */
 #define MAP_LOCKED	0x0200		/* pages are locked */
 #define MAP_NORESERVE	0x0400		/* don't check for reservations */
 #define MAP_DENYWRITE	0x0800		/* ETXTBSY */
diff --git a/arch/tile/include/asm/page.h b/arch/tile/include/asm/page.h
index f894a9016da6..7979a45430d3 100644
--- a/arch/tile/include/asm/page.h
+++ b/arch/tile/include/asm/page.h
@@ -129,6 +129,11 @@ static inline u64 pmd_val(pmd_t pmd)
 
 #endif
 
+static inline __attribute_const__ int get_order(unsigned long size)
+{
+	return BITS_PER_LONG - __builtin_clzl((size - 1) >> PAGE_SHIFT);
+}
+
 #endif /* !__ASSEMBLY__ */
 
 #define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
@@ -194,17 +199,17 @@ static inline u64 pmd_val(pmd_t pmd)
  * If you want more physical memory than this then see the CONFIG_HIGHMEM
  * option in the kernel configuration.
  *
- * The top two 16MB chunks in the table below (VIRT and HV) are
- * unavailable to Linux.  Since the kernel interrupt vectors must live
- * at 0xfd000000, we map all of the bottom of RAM at this address with
- * a huge page table entry to minimize its ITLB footprint (as well as
- * at PAGE_OFFSET).  The last architected requirement is that user
- * interrupt vectors live at 0xfc000000, so we make that range of
- * memory available to user processes.  The remaining regions are sized
- * as shown; after the first four addresses, we show "typical" values,
- * since the actual addresses depend on kernel #defines.
+ * The top 16MB chunk in the table below is unavailable to Linux.  Since
+ * the kernel interrupt vectors must live at ether 0xfe000000 or 0xfd000000
+ * (depending on whether the kernel is at PL2 or Pl1), we map all of the
+ * bottom of RAM at this address with a huge page table entry to minimize
+ * its ITLB footprint (as well as at PAGE_OFFSET).  The last architected
+ * requirement is that user interrupt vectors live at 0xfc000000, so we
+ * make that range of memory available to user processes.  The remaining
+ * regions are sized as shown; the first four addresses use the PL 1
+ * values, and after that, we show "typical" values, since the actual
+ * addresses depend on kernel #defines.
  *
- * MEM_VIRT_INTRPT                 0xff000000
  * MEM_HV_INTRPT                   0xfe000000
  * MEM_SV_INTRPT (kernel code)     0xfd000000
  * MEM_USER_INTRPT (user vector)   0xfc000000
@@ -216,9 +221,14 @@ static inline u64 pmd_val(pmd_t pmd)
  */
 
 #define MEM_USER_INTRPT		_AC(0xfc000000, UL)
+#if CONFIG_KERNEL_PL == 1
 #define MEM_SV_INTRPT		_AC(0xfd000000, UL)
 #define MEM_HV_INTRPT		_AC(0xfe000000, UL)
-#define MEM_VIRT_INTRPT		_AC(0xff000000, UL)
+#else
+#define MEM_GUEST_INTRPT	_AC(0xfd000000, UL)
+#define MEM_SV_INTRPT		_AC(0xfe000000, UL)
+#define MEM_HV_INTRPT		_AC(0xff000000, UL)
+#endif
 
 #define INTRPT_SIZE		0x4000
 
@@ -332,7 +342,6 @@ extern pte_t *virt_to_pte(struct mm_struct *mm, unsigned long addr);
 	(VM_READ | VM_WRITE | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
 
 #include <asm-generic/memory_model.h>
-#include <asm-generic/getorder.h>
 
 #endif /* __KERNEL__ */
 
diff --git a/arch/tile/include/asm/pci-bridge.h b/arch/tile/include/asm/pci-bridge.h
deleted file mode 100644
index e853b0e2793b..000000000000
--- a/arch/tile/include/asm/pci-bridge.h
+++ /dev/null
@@ -1,117 +0,0 @@
-/*
- * Copyright 2010 Tilera Corporation. All Rights Reserved.
- *
- *   This program is free software; you can redistribute it and/or
- *   modify it under the terms of the GNU General Public License
- *   as published by the Free Software Foundation, version 2.
- *
- *   This program is distributed in the hope that it will be useful, but
- *   WITHOUT ANY WARRANTY; without even the implied warranty of
- *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
- *   NON INFRINGEMENT.  See the GNU General Public License for
- *   more details.
- */
-
-#ifndef _ASM_TILE_PCI_BRIDGE_H
-#define _ASM_TILE_PCI_BRIDGE_H
-
-#include <linux/ioport.h>
-#include <linux/pci.h>
-
-struct device_node;
-struct pci_controller;
-
-/*
- * pci_io_base returns the memory address at which you can access
- * the I/O space for PCI bus number `bus' (or NULL on error).
- */
-extern void __iomem *pci_bus_io_base(unsigned int bus);
-extern unsigned long pci_bus_io_base_phys(unsigned int bus);
-extern unsigned long pci_bus_mem_base_phys(unsigned int bus);
-
-/* Allocate a new PCI host bridge structure */
-extern struct pci_controller *pcibios_alloc_controller(void);
-
-/* Helper function for setting up resources */
-extern void pci_init_resource(struct resource *res, unsigned long start,
-			      unsigned long end, int flags, char *name);
-
-/* Get the PCI host controller for a bus */
-extern struct pci_controller *pci_bus_to_hose(int bus);
-
-/*
- * Structure of a PCI controller (host bridge)
- */
-struct pci_controller {
-	int index;		/* PCI domain number */
-	struct pci_bus *root_bus;
-
-	int first_busno;
-	int last_busno;
-
-	int hv_cfg_fd[2];	/* config{0,1} fds for this PCIe controller */
-	int hv_mem_fd;		/* fd to Hypervisor for MMIO operations */
-
-	struct pci_ops *ops;
-
-	int irq_base;		/* Base IRQ from the Hypervisor	*/
-	int plx_gen1;		/* flag for PLX Gen 1 configuration */
-
-	/* Address ranges that are routed to this controller/bridge. */
-	struct resource mem_resources[3];
-};
-
-static inline struct pci_controller *pci_bus_to_host(struct pci_bus *bus)
-{
-	return bus->sysdata;
-}
-
-extern void setup_indirect_pci_nomap(struct pci_controller *hose,
-			       void __iomem *cfg_addr, void __iomem *cfg_data);
-extern void setup_indirect_pci(struct pci_controller *hose,
-			       u32 cfg_addr, u32 cfg_data);
-extern void setup_grackle(struct pci_controller *hose);
-
-extern unsigned char common_swizzle(struct pci_dev *, unsigned char *);
-
-/*
- *   The following code swizzles for exactly one bridge.  The routine
- *   common_swizzle below handles multiple bridges.  But there are a
- *   some boards that don't follow the PCI spec's suggestion so we
- *   break this piece out separately.
- */
-static inline unsigned char bridge_swizzle(unsigned char pin,
-		unsigned char idsel)
-{
-	return (((pin-1) + idsel) % 4) + 1;
-}
-
-/*
- * The following macro is used to lookup irqs in a standard table
- * format for those PPC systems that do not already have PCI
- * interrupts properly routed.
- */
-/* FIXME - double check this */
-#define PCI_IRQ_TABLE_LOOKUP ({ \
-	long _ctl_ = -1; \
-	if (idsel >= min_idsel && idsel <= max_idsel && pin <= irqs_per_slot) \
-		_ctl_ = pci_irq_table[idsel - min_idsel][pin-1]; \
-	_ctl_; \
-})
-
-/*
- * Scan the buses below a given PCI host bridge and assign suitable
- * resources to all devices found.
- */
-extern int pciauto_bus_scan(struct pci_controller *, int);
-
-#ifdef CONFIG_PCI
-extern unsigned long pci_address_to_pio(phys_addr_t address);
-#else
-static inline unsigned long pci_address_to_pio(phys_addr_t address)
-{
-	return (unsigned long)-1;
-}
-#endif
-
-#endif /* _ASM_TILE_PCI_BRIDGE_H */
diff --git a/arch/tile/include/asm/pci.h b/arch/tile/include/asm/pci.h
index b0c15da2d5d5..c3fc458a0d32 100644
--- a/arch/tile/include/asm/pci.h
+++ b/arch/tile/include/asm/pci.h
@@ -15,7 +15,29 @@
 #ifndef _ASM_TILE_PCI_H
 #define _ASM_TILE_PCI_H
 
-#include <asm/pci-bridge.h>
+#include <linux/pci.h>
+
+/*
+ * Structure of a PCI controller (host bridge)
+ */
+struct pci_controller {
+	int index;		/* PCI domain number */
+	struct pci_bus *root_bus;
+
+	int first_busno;
+	int last_busno;
+
+	int hv_cfg_fd[2];	/* config{0,1} fds for this PCIe controller */
+	int hv_mem_fd;		/* fd to Hypervisor for MMIO operations */
+
+	struct pci_ops *ops;
+
+	int irq_base;		/* Base IRQ from the Hypervisor	*/
+	int plx_gen1;		/* flag for PLX Gen 1 configuration */
+
+	/* Address ranges that are routed to this controller/bridge. */
+	struct resource mem_resources[3];
+};
 
 /*
  * The hypervisor maps the entirety of CPA-space as bus addresses, so
@@ -24,56 +46,12 @@
  */
 #define PCI_DMA_BUS_IS_PHYS     1
 
-struct pci_controller *pci_bus_to_hose(int bus);
-unsigned char __init common_swizzle(struct pci_dev *dev, unsigned char *pinp);
 int __init tile_pci_init(void);
-void pci_iounmap(struct pci_dev *dev, void __iomem *addr);
-void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max);
-void __devinit pcibios_fixup_bus(struct pci_bus *bus);
 
-int __devinit _tile_cfg_read(struct pci_controller *hose,
-				    int bus,
-				    int slot,
-				    int function,
-				    int offset,
-				    int size,
-				    u32 *val);
-int __devinit _tile_cfg_write(struct pci_controller *hose,
-				     int bus,
-				     int slot,
-				     int function,
-				     int offset,
-				     int size,
-				     u32 val);
+void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max);
+static inline void pci_iounmap(struct pci_dev *dev, void __iomem *addr) {}
 
-/*
- * These are used to to config reads and writes in the early stages of
- * setup before the driver infrastructure has been set up enough to be
- * able to do config reads and writes.
- */
-#define early_cfg_read(where, size, value) \
-	_tile_cfg_read(controller, \
-		       current_bus, \
-		       pci_slot, \
-		       pci_fn, \
-		       where, \
-		       size, \
-		       value)
-
-#define early_cfg_write(where, size, value) \
-	_tile_cfg_write(controller, \
-		       current_bus, \
-		       pci_slot, \
-		       pci_fn, \
-		       where, \
-		       size, \
-		       value)
-
-
-
-#define PCICFG_BYTE	1
-#define PCICFG_WORD	2
-#define PCICFG_DWORD	4
+void __devinit pcibios_fixup_bus(struct pci_bus *bus);
 
 #define	TILE_NUM_PCIE	2
 
@@ -88,33 +66,33 @@ static inline int pci_proc_domain(struct pci_bus *bus)
 }
 
 /*
- * I/O space is currently not supported.
+ * pcibios_assign_all_busses() tells whether or not the bus numbers
+ * should be reassigned, in case the BIOS didn't do it correctly, or
+ * in case we don't have a BIOS and we want to let Linux do it.
  */
+static inline int pcibios_assign_all_busses(void)
+{
+	return 1;
+}
 
-#define TILE_PCIE_LOWER_IO		0x0
-#define TILE_PCIE_UPPER_IO		0x10000
-#define TILE_PCIE_PCIE_IO_SIZE		0x0000FFFF
-
-#define _PAGE_NO_CACHE		0
-#define _PAGE_GUARDED		0
-
-
-#define pcibios_assign_all_busses()    pci_assign_all_buses
-extern int pci_assign_all_buses;
-
+/*
+ * No special bus mastering setup handling.
+ */
 static inline void pcibios_set_master(struct pci_dev *dev)
 {
-	/* No special bus mastering setup handling */
 }
 
 #define PCIBIOS_MIN_MEM		0
-#define PCIBIOS_MIN_IO		TILE_PCIE_LOWER_IO
+#define PCIBIOS_MIN_IO		0
 
 /*
  * This flag tells if the platform is TILEmpower that needs
  * special configuration for the PLX switch chip.
  */
-extern int blade_pci;
+extern int tile_plx_gen1;
+
+/* Use any cpu for PCI. */
+#define cpumask_of_pcibus(bus) cpu_online_mask
 
 /* implement the pci_ DMA API in terms of the generic device dma_ one */
 #include <asm-generic/pci-dma-compat.h>
@@ -122,7 +100,4 @@ extern int blade_pci;
 /* generic pci stuff */
 #include <asm-generic/pci.h>
 
-/* Use any cpu for PCI. */
-#define cpumask_of_pcibus(bus) cpu_online_mask
-
 #endif /* _ASM_TILE_PCI_H */
diff --git a/arch/tile/include/asm/pgtable.h b/arch/tile/include/asm/pgtable.h
index b3367379d537..a6604e9485da 100644
--- a/arch/tile/include/asm/pgtable.h
+++ b/arch/tile/include/asm/pgtable.h
@@ -344,18 +344,11 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 
 #if defined(CONFIG_HIGHPTE)
-extern pte_t *_pte_offset_map(pmd_t *, unsigned long address, enum km_type);
-#define pte_offset_map(dir, address) \
-	_pte_offset_map(dir, address, KM_PTE0)
-#define pte_offset_map_nested(dir, address) \
-	_pte_offset_map(dir, address, KM_PTE1)
-#define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
-#define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
+extern pte_t *pte_offset_map(pmd_t *, unsigned long address);
+#define pte_unmap(pte) kunmap_atomic(pte)
 #else
 #define pte_offset_map(dir, address) pte_offset_kernel(dir, address)
-#define pte_offset_map_nested(dir, address) pte_offset_map(dir, address)
 #define pte_unmap(pte) do { } while (0)
-#define pte_unmap_nested(pte) do { } while (0)
 #endif
 
 /* Clear a non-executable kernel PTE and flush it from the TLB. */
diff --git a/arch/tile/include/asm/processor.h b/arch/tile/include/asm/processor.h
index d942d09b252e..a9e7c8760334 100644
--- a/arch/tile/include/asm/processor.h
+++ b/arch/tile/include/asm/processor.h
@@ -103,6 +103,18 @@ struct thread_struct {
 	/* Any other miscellaneous processor state bits */
 	unsigned long proc_status;
 #endif
+#if !CHIP_HAS_FIXED_INTVEC_BASE()
+	/* Interrupt base for PL0 interrupts */
+	unsigned long interrupt_vector_base;
+#endif
+#if CHIP_HAS_TILE_RTF_HWM()
+	/* Tile cache retry fifo high-water mark */
+	unsigned long tile_rtf_hwm;
+#endif
+#if CHIP_HAS_DSTREAM_PF()
+	/* Data stream prefetch control */
+	unsigned long dstream_pf;
+#endif
 #ifdef CONFIG_HARDWALL
 	/* Is this task tied to an activated hardwall? */
 	struct hardwall_info *hardwall;
@@ -280,8 +292,18 @@ extern int kstack_hash;
 /* Are we using huge pages in the TLB for kernel data? */
 extern int kdata_huge;
 
+/* Support standard Linux prefetching. */
+#define ARCH_HAS_PREFETCH
+#define prefetch(x) __builtin_prefetch(x)
 #define PREFETCH_STRIDE CHIP_L2_LINE_SIZE()
 
+/* Bring a value into the L1D, faulting the TLB if necessary. */
+#ifdef __tilegx__
+#define prefetch_L1(x) __insn_prefetch_l1_fault((void *)(x))
+#else
+#define prefetch_L1(x) __insn_prefetch_L1((void *)(x))
+#endif
+
 #else /* __ASSEMBLY__ */
 
 /* Do some slow action (e.g. read a slow SPR). */
@@ -316,18 +338,21 @@ extern int kdata_huge;
  * Note that assembly code assumes that USER_PL is zero.
  */
 #define USER_PL 0
-#define KERNEL_PL 1
+#if CONFIG_KERNEL_PL == 2
+#define GUEST_PL 1
+#endif
+#define KERNEL_PL CONFIG_KERNEL_PL
 
-/* SYSTEM_SAVE_1_0 holds the current cpu number ORed with ksp0. */
+/* SYSTEM_SAVE_K_0 holds the current cpu number ORed with ksp0. */
 #define CPU_LOG_MASK_VALUE 12
 #define CPU_MASK_VALUE ((1 << CPU_LOG_MASK_VALUE) - 1)
 #if CONFIG_NR_CPUS > CPU_MASK_VALUE
 # error Too many cpus!
 #endif
 #define raw_smp_processor_id() \
-	((int)__insn_mfspr(SPR_SYSTEM_SAVE_1_0) & CPU_MASK_VALUE)
+	((int)__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & CPU_MASK_VALUE)
 #define get_current_ksp0() \
-	(__insn_mfspr(SPR_SYSTEM_SAVE_1_0) & ~CPU_MASK_VALUE)
+	(__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & ~CPU_MASK_VALUE)
 #define next_current_ksp0(task) ({ \
 	unsigned long __ksp0 = task_ksp0(task); \
 	int __cpu = raw_smp_processor_id(); \
diff --git a/arch/tile/include/asm/ptrace.h b/arch/tile/include/asm/ptrace.h
index acdae814e016..ac6d343129d3 100644
--- a/arch/tile/include/asm/ptrace.h
+++ b/arch/tile/include/asm/ptrace.h
@@ -51,10 +51,7 @@ typedef uint_reg_t pt_reg_t;
 
 /*
  * This struct defines the way the registers are stored on the stack during a
- * system call/exception.  It should be a multiple of 8 bytes to preserve
- * normal stack alignment rules.
- *
- * Must track <sys/ucontext.h> and <sys/procfs.h>
+ * system call or exception.  "struct sigcontext" has the same shape.
  */
 struct pt_regs {
 	/* Saved main processor registers; 56..63 are special. */
@@ -65,8 +62,8 @@ struct pt_regs {
 	pt_reg_t lr;		/* aliases regs[TREG_LR] */
 
 	/* Saved special registers. */
-	pt_reg_t pc;		/* stored in EX_CONTEXT_1_0 */
-	pt_reg_t ex1;		/* stored in EX_CONTEXT_1_1 (PL and ICS bit) */
+	pt_reg_t pc;		/* stored in EX_CONTEXT_K_0 */
+	pt_reg_t ex1;		/* stored in EX_CONTEXT_K_1 (PL and ICS bit) */
 	pt_reg_t faultnum;	/* fault number (INT_SWINT_1 for syscall) */
 	pt_reg_t orig_r0;	/* r0 at syscall entry, else zero */
 	pt_reg_t flags;		/* flags (see below) */
@@ -80,11 +77,6 @@ struct pt_regs {
 
 #endif /* __ASSEMBLY__ */
 
-/* Flag bits in pt_regs.flags */
-#define PT_FLAGS_DISABLE_IRQ    1  /* on return to kernel, disable irqs */
-#define PT_FLAGS_CALLER_SAVES   2  /* caller-save registers are valid */
-#define PT_FLAGS_RESTORE_REGS   4  /* restore callee-save regs on return */
-
 #define PTRACE_GETREGS		12
 #define PTRACE_SETREGS		13
 #define PTRACE_GETFPREGS	14
@@ -101,6 +93,11 @@ struct pt_regs {
 
 #ifdef __KERNEL__
 
+/* Flag bits in pt_regs.flags */
+#define PT_FLAGS_DISABLE_IRQ    1  /* on return to kernel, disable irqs */
+#define PT_FLAGS_CALLER_SAVES   2  /* caller-save registers are valid */
+#define PT_FLAGS_RESTORE_REGS   4  /* restore callee-save regs on return */
+
 #ifndef __ASSEMBLY__
 
 #define instruction_pointer(regs) ((regs)->pc)
diff --git a/arch/tile/include/asm/scatterlist.h b/arch/tile/include/asm/scatterlist.h
index c5604242c0d5..35d786fe93ae 100644
--- a/arch/tile/include/asm/scatterlist.h
+++ b/arch/tile/include/asm/scatterlist.h
@@ -1,22 +1 @@
-/*
- * Copyright 2010 Tilera Corporation. All Rights Reserved.
- *
- *   This program is free software; you can redistribute it and/or
- *   modify it under the terms of the GNU General Public License
- *   as published by the Free Software Foundation, version 2.
- *
- *   This program is distributed in the hope that it will be useful, but
- *   WITHOUT ANY WARRANTY; without even the implied warranty of
- *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
- *   NON INFRINGEMENT.  See the GNU General Public License for
- *   more details.
- */
-
-#ifndef _ASM_TILE_SCATTERLIST_H
-#define _ASM_TILE_SCATTERLIST_H
-
-#define ISA_DMA_THRESHOLD	(~0UL)
-
 #include <asm-generic/scatterlist.h>
-
-#endif /* _ASM_TILE_SCATTERLIST_H */
diff --git a/arch/tile/include/asm/setup.h b/arch/tile/include/asm/setup.h
index 823ddd47ff6e..7caf0f36b030 100644
--- a/arch/tile/include/asm/setup.h
+++ b/arch/tile/include/asm/setup.h
@@ -15,6 +15,10 @@
 #ifndef _ASM_TILE_SETUP_H
 #define _ASM_TILE_SETUP_H
 
+#define COMMAND_LINE_SIZE	2048
+
+#ifdef __KERNEL__
+
 #include <linux/pfn.h>
 #include <linux/init.h>
 
@@ -23,10 +27,10 @@
  */
 #define MAXMEM_PFN	PFN_DOWN(MAXMEM)
 
-#define COMMAND_LINE_SIZE	2048
-
 void early_panic(const char *fmt, ...);
 void warn_early_printk(void);
 void __init disable_early_printk(void);
 
+#endif /* __KERNEL__ */
+
 #endif /* _ASM_TILE_SETUP_H */
diff --git a/arch/tile/include/asm/sigcontext.h b/arch/tile/include/asm/sigcontext.h
index 7cd7672e3ad4..5e2d03336f53 100644
--- a/arch/tile/include/asm/sigcontext.h
+++ b/arch/tile/include/asm/sigcontext.h
@@ -15,13 +15,21 @@
 #ifndef _ASM_TILE_SIGCONTEXT_H
 #define _ASM_TILE_SIGCONTEXT_H
 
-/* NOTE: we can't include <linux/ptrace.h> due to #include dependencies. */
-#include <asm/ptrace.h>
-
-/* Must track <sys/ucontext.h> */
+#include <arch/abi.h>
 
+/*
+ * struct sigcontext has the same shape as struct pt_regs,
+ * but is simplified since we know the fault is from userspace.
+ */
 struct sigcontext {
-	struct pt_regs regs;
+	uint_reg_t gregs[53];	/* General-purpose registers.  */
+	uint_reg_t tp;		/* Aliases gregs[TREG_TP].  */
+	uint_reg_t sp;		/* Aliases gregs[TREG_SP].  */
+	uint_reg_t lr;		/* Aliases gregs[TREG_LR].  */
+	uint_reg_t pc;		/* Program counter.  */
+	uint_reg_t ics;		/* In Interrupt Critical Section?  */
+	uint_reg_t faultnum;	/* Fault number.  */
+	uint_reg_t pad[5];
 };
 
 #endif /* _ASM_TILE_SIGCONTEXT_H */
diff --git a/arch/tile/include/asm/siginfo.h b/arch/tile/include/asm/siginfo.h
index 0c12d1b9ddf2..56d661bb010b 100644
--- a/arch/tile/include/asm/siginfo.h
+++ b/arch/tile/include/asm/siginfo.h
@@ -17,6 +17,10 @@
 
 #define __ARCH_SI_TRAPNO
 
+#ifdef __LP64__
+# define __ARCH_SI_PREAMBLE_SIZE	(4 * sizeof(int))
+#endif
+
 #include <asm-generic/siginfo.h>
 
 /*
diff --git a/arch/tile/include/asm/signal.h b/arch/tile/include/asm/signal.h
index eb0253f32202..81d92a45cd4b 100644
--- a/arch/tile/include/asm/signal.h
+++ b/arch/tile/include/asm/signal.h
@@ -24,7 +24,8 @@
 #include <asm-generic/signal.h>
 
 #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
-int restore_sigcontext(struct pt_regs *, struct sigcontext __user *, long *);
+struct pt_regs;
+int restore_sigcontext(struct pt_regs *, struct sigcontext __user *);
 int setup_sigcontext(struct sigcontext __user *, struct pt_regs *);
 void do_signal(struct pt_regs *regs);
 #endif
diff --git a/arch/tile/include/asm/stat.h b/arch/tile/include/asm/stat.h
index 3dc90fa92c70..b16e5db8f0e7 100644
--- a/arch/tile/include/asm/stat.h
+++ b/arch/tile/include/asm/stat.h
@@ -1 +1,4 @@
+#ifdef CONFIG_COMPAT
+#define __ARCH_WANT_STAT64	/* Used for compat_sys_stat64() etc. */
+#endif
 #include <asm-generic/stat.h>
diff --git a/arch/tile/include/asm/syscalls.h b/arch/tile/include/asm/syscalls.h
index af165a74537f..3b5507c31eae 100644
--- a/arch/tile/include/asm/syscalls.h
+++ b/arch/tile/include/asm/syscalls.h
@@ -32,8 +32,9 @@ extern void *compat_sys_call_table[];
 
 /*
  * Note that by convention, any syscall which requires the current
- * register set takes an additional "struct pt_regs *" pointer; the
- * sys_xxx() function just adds the pointer and tail-calls to _sys_xxx().
+ * register set takes an additional "struct pt_regs *" pointer; a
+ * _sys_xxx() trampoline in intvec*.S just sets up the pointer and
+ * jumps to sys_xxx().
  */
 
 /* kernel/sys.c */
@@ -43,61 +44,17 @@ long sys32_fadvise64(int fd, u32 offset_lo, u32 offset_hi,
 int sys32_fadvise64_64(int fd, u32 offset_lo, u32 offset_hi,
 		       u32 len_lo, u32 len_hi, int advice);
 long sys_flush_cache(void);
-long sys_mmap2(unsigned long addr, unsigned long len,
-	       unsigned long prot, unsigned long flags,
-	       unsigned long fd, unsigned long pgoff);
-#ifdef __tilegx__
-long sys_mmap(unsigned long addr, unsigned long len,
-	      unsigned long prot, unsigned long flags,
-	      unsigned long fd, off_t pgoff);
+#ifndef __tilegx__  /* No mmap() in the 32-bit kernel. */
+#define sys_mmap sys_mmap
 #endif
 
-/* kernel/process.c */
-long sys_clone(unsigned long clone_flags, unsigned long newsp,
-	       void __user *parent_tid, void __user *child_tid);
-long _sys_clone(unsigned long clone_flags, unsigned long newsp,
-		void __user *parent_tid, void __user *child_tid,
-		struct pt_regs *regs);
-long sys_fork(void);
-long _sys_fork(struct pt_regs *regs);
-long sys_vfork(void);
-long _sys_vfork(struct pt_regs *regs);
-long sys_execve(char __user *filename, char __user * __user *argv,
-		char __user * __user *envp);
-long _sys_execve(char __user *filename, char __user * __user *argv,
-		 char __user * __user *envp, struct pt_regs *regs);
-
-/* kernel/signal.c */
-long sys_sigaltstack(const stack_t __user *, stack_t __user *);
-long _sys_sigaltstack(const stack_t __user *, stack_t __user *,
-		      struct pt_regs *);
-long sys_rt_sigreturn(void);
-long _sys_rt_sigreturn(struct pt_regs *regs);
-
-/* platform-independent functions */
-long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize);
-long sys_rt_sigaction(int sig, const struct sigaction __user *act,
-		      struct sigaction __user *oact, size_t sigsetsize);
-
 #ifndef __tilegx__
 /* mm/fault.c */
-int sys_cmpxchg_badaddr(unsigned long address);
-int _sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *);
+long sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *);
+long _sys_cmpxchg_badaddr(unsigned long address);
 #endif
 
 #ifdef CONFIG_COMPAT
-long compat_sys_execve(char __user *path, compat_uptr_t __user *argv,
-		       compat_uptr_t __user *envp);
-long _compat_sys_execve(char __user *path, compat_uptr_t __user *argv,
-			compat_uptr_t __user *envp, struct pt_regs *regs);
-long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			    struct compat_sigaltstack __user *uoss_ptr);
-long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			     struct compat_sigaltstack __user *uoss_ptr,
-			     struct pt_regs *regs);
-long compat_sys_rt_sigreturn(void);
-long _compat_sys_rt_sigreturn(struct pt_regs *regs);
-
 /* These four are not defined for 64-bit, but serve as "compat" syscalls. */
 long sys_fcntl64(unsigned int fd, unsigned int cmd, unsigned long arg);
 long sys_fstat64(unsigned long fd, struct stat64 __user *statbuf);
@@ -105,4 +62,15 @@ long sys_truncate64(const char __user *path, loff_t length);
 long sys_ftruncate64(unsigned int fd, loff_t length);
 #endif
 
+/* These are the intvec*.S trampolines. */
+long _sys_sigaltstack(const stack_t __user *, stack_t __user *);
+long _sys_rt_sigreturn(void);
+long _sys_clone(unsigned long clone_flags, unsigned long newsp,
+		void __user *parent_tid, void __user *child_tid);
+long _sys_execve(const char __user *filename,
+		 const char __user *const __user *argv,
+		 const char __user *const __user *envp);
+
+#include <asm-generic/syscalls.h>
+
 #endif /* _ASM_TILE_SYSCALLS_H */
diff --git a/arch/tile/include/asm/system.h b/arch/tile/include/asm/system.h
index f749be327ce0..5388850deeb2 100644
--- a/arch/tile/include/asm/system.h
+++ b/arch/tile/include/asm/system.h
@@ -89,6 +89,10 @@
 #define get_cycles_low() __insn_mfspr(SPR_CYCLE)   /* just get all 64 bits */
 #endif
 
+#if !CHIP_HAS_MF_WAITS_FOR_VICTIMS()
+int __mb_incoherent(void);  /* Helper routine for mb_incoherent(). */
+#endif
+
 /* Fence to guarantee visibility of stores to incoherent memory. */
 static inline void
 mb_incoherent(void)
@@ -97,7 +101,6 @@ mb_incoherent(void)
 
 #if !CHIP_HAS_MF_WAITS_FOR_VICTIMS()
 	{
-		int __mb_incoherent(void);
 #if CHIP_HAS_TILE_WRITE_PENDING()
 		const unsigned long WRITE_TIMEOUT_CYCLES = 400;
 		unsigned long start = get_cycles_low();
@@ -161,7 +164,7 @@ extern struct task_struct *_switch_to(struct task_struct *prev,
 /* Helper function for _switch_to(). */
 extern struct task_struct *__switch_to(struct task_struct *prev,
 				       struct task_struct *next,
-				       unsigned long new_system_save_1_0);
+				       unsigned long new_system_save_k_0);
 
 /* Address that switched-away from tasks are at. */
 extern unsigned long get_switch_to_pc(void);
@@ -214,13 +217,6 @@ int hardwall_deactivate(struct task_struct *task);
 } while (0)
 #endif
 
-/* Invoke the simulator "syscall" mechanism (see arch/tile/kernel/entry.S). */
-extern int _sim_syscall(int syscall_num, ...);
-#define sim_syscall(syscall_num, ...) \
-	_sim_syscall(SIM_CONTROL_SYSCALL + \
-		((syscall_num) << _SIM_CONTROL_OPERATOR_BITS), \
-		## __VA_ARGS__)
-
 /*
  * Kernel threads can check to see if they need to migrate their
  * stack whenever they return from a context switch; for user
diff --git a/arch/tile/include/asm/traps.h b/arch/tile/include/asm/traps.h
index 432a9c15c8a2..d06e35f57201 100644
--- a/arch/tile/include/asm/traps.h
+++ b/arch/tile/include/asm/traps.h
@@ -59,4 +59,8 @@ void do_hardwall_trap(struct pt_regs *, int fault_num);
 void do_breakpoint(struct pt_regs *, int fault_num);
 
 
+#ifdef __tilegx__
+void gx_singlestep_handle(struct pt_regs *, int fault_num);
+#endif
+
 #endif /* _ASM_TILE_SYSCALLS_H */
diff --git a/arch/tile/include/asm/uaccess.h b/arch/tile/include/asm/uaccess.h
index ed17a80ec0ed..ef34d2caa5b1 100644
--- a/arch/tile/include/asm/uaccess.h
+++ b/arch/tile/include/asm/uaccess.h
@@ -389,14 +389,14 @@ static inline unsigned long __must_check copy_from_user(void *to,
  * Returns number of bytes that could not be copied.
  * On success, this will be zero.
  */
-extern unsigned long __copy_in_user_asm(
+extern unsigned long __copy_in_user_inatomic(
 	void __user *to, const void __user *from, unsigned long n);
 
 static inline unsigned long __must_check
 __copy_in_user(void __user *to, const void __user *from, unsigned long n)
 {
 	might_sleep();
-	return __copy_in_user_asm(to, from, n);
+	return __copy_in_user_inatomic(to, from, n);
 }
 
 static inline unsigned long __must_check
diff --git a/arch/tile/include/asm/unistd.h b/arch/tile/include/asm/unistd.h
index f2e3ff485333..b35c2db71199 100644
--- a/arch/tile/include/asm/unistd.h
+++ b/arch/tile/include/asm/unistd.h
@@ -41,6 +41,7 @@ __SYSCALL(__NR_cmpxchg_badaddr, sys_cmpxchg_badaddr)
 #ifdef CONFIG_COMPAT
 #define __ARCH_WANT_SYS_LLSEEK
 #endif
+#define __ARCH_WANT_SYS_NEWFSTATAT
 #endif
 
 #endif /* _ASM_TILE_UNISTD_H */