Merge branch 'parisc-4.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull parisc updates from Helge Deller:

 - Add native high-resolution timing code for sched_clock() and other
   timing functions based on the processor internal cr16 cycle counters

 - Add syscall tracepoint support

 - Add regset support

 - Speed up get_user() and put_user() functions

 - Updated futex.h to match generic implementation (John David Anglin)

 - A few smaller ftrace build fixes

 - Fixed thuge-gen kernel self test to utilize architectured MAP_HUGETLB
   value

 - Added parisc architecture to seccomp_bpf kernel self test

 - Various typo fixes (Andrea Gelmini)

* 'parisc-4.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Whitespace cleanups in unistd.h
  parisc: Use long jump to reach ftrace_return_to_handler()
  parisc: Fix typo in fpudispatch.c
  parisc: Fix typos in eisa_eeprom.h
  parisc: Fix typo in ldcw.h
  parisc: Fix typo in pdc.h
  parisc: Update futex.h to match generic implementation
  parisc: Merge ftrace C-helper and assembler functions into .text.hot section
  selftests/thuge-gen: Use platform specific MAP_HUGETLB value
  parisc: Add native high-resolution sched_clock() implementation
  parisc: Add ARCH_TRACEHOOK and regset support
  parisc: Add 64bit get_user() and put_user() for 32bit kernel
  parisc: Simplify and speed up get_user() and put_user()
  parisc: Add syscall tracepoint support

19 files changed, 601 insertions, 137 deletions

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index 3d498a676551..dc117385ce2e 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -6,6 +6,7 @@ config PARISC
 	select HAVE_OPROFILE
 	select HAVE_FUNCTION_TRACER
 	select HAVE_FUNCTION_GRAPH_TRACER
+	select HAVE_SYSCALL_TRACEPOINTS
 	select ARCH_WANT_FRAME_POINTERS
 	select RTC_CLASS
 	select RTC_DRV_GENERIC
@@ -31,6 +32,8 @@ config PARISC
 	select HAVE_DEBUG_STACKOVERFLOW
 	select HAVE_ARCH_AUDITSYSCALL
 	select HAVE_ARCH_SECCOMP_FILTER
+	select HAVE_ARCH_TRACEHOOK
+	select HAVE_UNSTABLE_SCHED_CLOCK if (SMP || !64BIT)
 	select ARCH_NO_COHERENT_DMA_MMAP
 	select CPU_NO_EFFICIENT_FFS
 
diff --git a/arch/parisc/include/asm/cmpxchg.h b/arch/parisc/include/asm/cmpxchg.h
index 0a90b965cccb..7ada30900807 100644
--- a/arch/parisc/include/asm/cmpxchg.h
+++ b/arch/parisc/include/asm/cmpxchg.h
@@ -52,8 +52,7 @@ extern void __cmpxchg_called_with_bad_pointer(void);
 /* __cmpxchg_u32/u64 defined in arch/parisc/lib/bitops.c */
 extern unsigned long __cmpxchg_u32(volatile unsigned int *m, unsigned int old,
 				   unsigned int new_);
-extern unsigned long __cmpxchg_u64(volatile unsigned long *ptr,
-				   unsigned long old, unsigned long new_);
+extern u64 __cmpxchg_u64(volatile u64 *ptr, u64 old, u64 new_);
 
 /* don't worry...optimizer will get rid of most of this */
 static inline unsigned long
@@ -61,7 +60,7 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new_, int size)
 {
 	switch (size) {
 #ifdef CONFIG_64BIT
-	case 8: return __cmpxchg_u64((unsigned long *)ptr, old, new_);
+	case 8: return __cmpxchg_u64((u64 *)ptr, old, new_);
 #endif
 	case 4: return __cmpxchg_u32((unsigned int *)ptr,
 				     (unsigned int)old, (unsigned int)new_);
@@ -86,7 +85,7 @@ static inline unsigned long __cmpxchg_local(volatile void *ptr,
 {
 	switch (size) {
 #ifdef CONFIG_64BIT
-	case 8:	return __cmpxchg_u64((unsigned long *)ptr, old, new_);
+	case 8:	return __cmpxchg_u64((u64 *)ptr, old, new_);
 #endif
 	case 4:	return __cmpxchg_u32(ptr, old, new_);
 	default:
@@ -111,4 +110,6 @@ static inline unsigned long __cmpxchg_local(volatile void *ptr,
 #define cmpxchg64_local(ptr, o, n) __cmpxchg64_local_generic((ptr), (o), (n))
 #endif
 
+#define cmpxchg64(ptr, o, n) __cmpxchg_u64(ptr, o, n)
+
 #endif /* _ASM_PARISC_CMPXCHG_H_ */
diff --git a/arch/parisc/include/asm/eisa_eeprom.h b/arch/parisc/include/asm/eisa_eeprom.h
index 8ce8b85ca588..5637ac962f8e 100644
--- a/arch/parisc/include/asm/eisa_eeprom.h
+++ b/arch/parisc/include/asm/eisa_eeprom.h
@@ -99,7 +99,7 @@ struct eeprom_eisa_slot_info
 #define HPEE_MEMORY_DECODE_24BITS 0x04
 #define HPEE_MEMORY_DECODE_32BITS 0x08
 /* byte 2 and 3 are a 16bit LE value
- * containging the memory size in kilobytes */
+ * containing the memory size in kilobytes */
 /* byte 4,5,6 are a 24bit LE value
  * containing the memory base address */
 
@@ -135,7 +135,7 @@ struct eeprom_eisa_slot_info
 #define HPEE_PORT_SHARED    0x40
 #define HPEE_PORT_MORE      0x80
 /* byte 1 and 2 is a 16bit LE value
- * conating the start port number */
+ * containing the start port number */
 
 #define HPEE_PORT_INIT_MAX_LEN     60 /* in bytes here */
 /* port init entry byte 0 */
diff --git a/arch/parisc/include/asm/ftrace.h b/arch/parisc/include/asm/ftrace.h
index 24cd81d58d70..d635c6b0269d 100644
--- a/arch/parisc/include/asm/ftrace.h
+++ b/arch/parisc/include/asm/ftrace.h
@@ -6,6 +6,8 @@ extern void mcount(void);
 
 #define MCOUNT_INSN_SIZE 4
 
+extern unsigned long sys_call_table[];
+
 extern unsigned long return_address(unsigned int);
 
 #define ftrace_return_address(n) return_address(n)
diff --git a/arch/parisc/include/asm/futex.h b/arch/parisc/include/asm/futex.h
index 49df14805a9b..ac8bd586ace8 100644
--- a/arch/parisc/include/asm/futex.h
+++ b/arch/parisc/include/asm/futex.h
@@ -35,70 +35,57 @@ static inline int
 futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	unsigned long int flags;
-	u32 val;
 	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;
+	int oldval, ret;
+	u32 tmp;
+
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
 	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(*uaddr)))
 		return -EFAULT;
 
+	_futex_spin_lock_irqsave(uaddr, &flags);
 	pagefault_disable();
 
-	_futex_spin_lock_irqsave(uaddr, &flags);
+	ret = -EFAULT;
+	if (unlikely(get_user(oldval, uaddr) != 0))
+		goto out_pagefault_enable;
+
+	ret = 0;
+	tmp = oldval;
 
 	switch (op) {
 	case FUTEX_OP_SET:
-		/* *(int *)UADDR2 = OPARG; */
-		ret = get_user(oldval, uaddr);
-		if (!ret)
-			ret = put_user(oparg, uaddr);
+		tmp = oparg;
 		break;
 	case FUTEX_OP_ADD:
-		/* *(int *)UADDR2 += OPARG; */
-		ret = get_user(oldval, uaddr);
-		if (!ret) {
-			val = oldval + oparg;
-			ret = put_user(val, uaddr);
-		}
+		tmp += oparg;
 		break;
 	case FUTEX_OP_OR:
-		/* *(int *)UADDR2 |= OPARG; */
-		ret = get_user(oldval, uaddr);
-		if (!ret) {
-			val = oldval | oparg;
-			ret = put_user(val, uaddr);
-		}
+		tmp |= oparg;
 		break;
 	case FUTEX_OP_ANDN:
-		/* *(int *)UADDR2 &= ~OPARG; */
-		ret = get_user(oldval, uaddr);
-		if (!ret) {
-			val = oldval & ~oparg;
-			ret = put_user(val, uaddr);
-		}
+		tmp &= ~oparg;
 		break;
 	case FUTEX_OP_XOR:
-		/* *(int *)UADDR2 ^= OPARG; */
-		ret = get_user(oldval, uaddr);
-		if (!ret) {
-			val = oldval ^ oparg;
-			ret = put_user(val, uaddr);
-		}
+		tmp ^= oparg;
 		break;
 	default:
 		ret = -ENOSYS;
 	}
 
-	_futex_spin_unlock_irqrestore(uaddr, &flags);
+	if (ret == 0 && unlikely(put_user(tmp, uaddr) != 0))
+		ret = -EFAULT;
 
+out_pagefault_enable:
 	pagefault_enable();
+	_futex_spin_unlock_irqrestore(uaddr, &flags);
 
-	if (!ret) {
+	if (ret == 0) {
 		switch (cmp) {
 		case FUTEX_OP_CMP_EQ: ret = (oldval == cmparg); break;
 		case FUTEX_OP_CMP_NE: ret = (oldval != cmparg); break;
@@ -112,12 +99,10 @@ futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 	return ret;
 }
 
-/* Non-atomic version */
 static inline int
 futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
 			      u32 oldval, u32 newval)
 {
-	int ret;
 	u32 val;
 	unsigned long flags;
 
@@ -137,17 +122,20 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
 	 */
 
 	_futex_spin_lock_irqsave(uaddr, &flags);
+	if (unlikely(get_user(val, uaddr) != 0)) {
+		_futex_spin_unlock_irqrestore(uaddr, &flags);
+		return -EFAULT;
+	}
 
-	ret = get_user(val, uaddr);
-
-	if (!ret && val == oldval)
-		ret = put_user(newval, uaddr);
+	if (val == oldval && unlikely(put_user(newval, uaddr) != 0)) {
+		_futex_spin_unlock_irqrestore(uaddr, &flags);
+		return -EFAULT;
+	}
 
 	*uval = val;
-
 	_futex_spin_unlock_irqrestore(uaddr, &flags);
 
-	return ret;
+	return 0;
 }
 
 #endif /*__KERNEL__*/
diff --git a/arch/parisc/include/asm/ldcw.h b/arch/parisc/include/asm/ldcw.h
index 8121aa6db2ff..8be707e1b6c7 100644
--- a/arch/parisc/include/asm/ldcw.h
+++ b/arch/parisc/include/asm/ldcw.h
@@ -40,7 +40,7 @@
    memory to indicate to the compiler that the assembly code reads
    or writes to items other than those listed in the input and output
    operands.  This may pessimize the code somewhat but __ldcw is
-   usually used within code blocks surrounded by memory barriors.  */
+   usually used within code blocks surrounded by memory barriers.  */
 #define __ldcw(a) ({						\
 	unsigned __ret;						\
 	__asm__ __volatile__(__LDCW " 0(%1),%0"			\
diff --git a/arch/parisc/include/asm/syscall.h b/arch/parisc/include/asm/syscall.h
index 637ce8d6f375..5e0b4e6bd99d 100644
--- a/arch/parisc/include/asm/syscall.h
+++ b/arch/parisc/include/asm/syscall.h
@@ -8,6 +8,8 @@
 #include <linux/err.h>
 #include <asm/ptrace.h>
 
+#define NR_syscalls (__NR_Linux_syscalls)
+
 static inline long syscall_get_nr(struct task_struct *tsk,
 				  struct pt_regs *regs)
 {
@@ -33,12 +35,19 @@ static inline void syscall_get_arguments(struct task_struct *tsk,
 		args[1] = regs->gr[25];
 	case 1:
 		args[0] = regs->gr[26];
+	case 0:
 		break;
 	default:
 		BUG();
 	}
 }
 
+static inline long syscall_get_return_value(struct task_struct *task,
+						struct pt_regs *regs)
+{
+	return regs->gr[28];
+}
+
 static inline void syscall_set_return_value(struct task_struct *task,
 					    struct pt_regs *regs,
 					    int error, long val)
diff --git a/arch/parisc/include/asm/thread_info.h b/arch/parisc/include/asm/thread_info.h
index e96e693fd58c..7581330ea35b 100644
--- a/arch/parisc/include/asm/thread_info.h
+++ b/arch/parisc/include/asm/thread_info.h
@@ -55,6 +55,7 @@ struct thread_info {
 #define TIF_SINGLESTEP		9	/* single stepping? */
 #define TIF_BLOCKSTEP		10	/* branch stepping? */
 #define TIF_SECCOMP		11	/* secure computing */
+#define TIF_SYSCALL_TRACEPOINT	12	/* syscall tracepoint instrumentation */
 
 #define _TIF_SYSCALL_TRACE	(1 << TIF_SYSCALL_TRACE)
 #define _TIF_SIGPENDING		(1 << TIF_SIGPENDING)
@@ -66,12 +67,13 @@ struct thread_info {
 #define _TIF_SINGLESTEP		(1 << TIF_SINGLESTEP)
 #define _TIF_BLOCKSTEP		(1 << TIF_BLOCKSTEP)
 #define _TIF_SECCOMP		(1 << TIF_SECCOMP)
+#define _TIF_SYSCALL_TRACEPOINT	(1 << TIF_SYSCALL_TRACEPOINT)
 
 #define _TIF_USER_WORK_MASK     (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | \
                                  _TIF_NEED_RESCHED)
 #define _TIF_SYSCALL_TRACE_MASK (_TIF_SYSCALL_TRACE | _TIF_SINGLESTEP |	\
 				 _TIF_BLOCKSTEP | _TIF_SYSCALL_AUDIT | \
-				 _TIF_SECCOMP)
+				 _TIF_SECCOMP | _TIF_SYSCALL_TRACEPOINT)
 
 #ifdef CONFIG_64BIT
 # ifdef CONFIG_COMPAT
diff --git a/arch/parisc/include/asm/uaccess.h b/arch/parisc/include/asm/uaccess.h
index 7955e43f3f3f..0f59fd9ca205 100644
--- a/arch/parisc/include/asm/uaccess.h
+++ b/arch/parisc/include/asm/uaccess.h
@@ -40,14 +40,10 @@ static inline long access_ok(int type, const void __user * addr,
 #define get_user __get_user
 
 #if !defined(CONFIG_64BIT)
-#define LDD_KERNEL(ptr)		BUILD_BUG()
-#define LDD_USER(ptr)		BUILD_BUG()
-#define STD_KERNEL(x, ptr)	__put_kernel_asm64(x, ptr)
+#define LDD_USER(ptr)		__get_user_asm64(ptr)
 #define STD_USER(x, ptr)	__put_user_asm64(x, ptr)
 #else
-#define LDD_KERNEL(ptr)		__get_kernel_asm("ldd", ptr)
 #define LDD_USER(ptr)		__get_user_asm("ldd", ptr)
-#define STD_KERNEL(x, ptr)	__put_kernel_asm("std", x, ptr)
 #define STD_USER(x, ptr)	__put_user_asm("std", x, ptr)
 #endif
 
@@ -80,70 +76,70 @@ struct exception_data {
 	unsigned long fault_addr;
 };
 
+/*
+ * load_sr2() preloads the space register %%sr2 - based on the value of
+ * get_fs() - with either a value of 0 to access kernel space (KERNEL_DS which
+ * is 0), or with the current value of %%sr3 to access user space (USER_DS)
+ * memory. The following __get_user_asm() and __put_user_asm() functions have
+ * %%sr2 hard-coded to access the requested memory.
+ */
+#define load_sr2() \
+	__asm__(" or,=  %0,%%r0,%%r0\n\t"	\
+		" mfsp %%sr3,%0\n\t"		\
+		" mtsp %0,%%sr2\n\t"		\
+		: : "r"(get_fs()) : )
+
 #define __get_user(x, ptr)                               \
 ({                                                       \
 	register long __gu_err __asm__ ("r8") = 0;       \
 	register long __gu_val __asm__ ("r9") = 0;       \
 							 \
-	if (segment_eq(get_fs(), KERNEL_DS)) {           \
-	    switch (sizeof(*(ptr))) {                    \
-	    case 1: __get_kernel_asm("ldb", ptr); break; \
-	    case 2: __get_kernel_asm("ldh", ptr); break; \
-	    case 4: __get_kernel_asm("ldw", ptr); break; \
-	    case 8: LDD_KERNEL(ptr); break;		 \
-	    default: BUILD_BUG(); break;		 \
-	    }                                            \
-	}                                                \
-	else {                                           \
-	    switch (sizeof(*(ptr))) {                    \
+	load_sr2();					 \
+	switch (sizeof(*(ptr))) {			 \
 	    case 1: __get_user_asm("ldb", ptr); break;   \
 	    case 2: __get_user_asm("ldh", ptr); break;   \
 	    case 4: __get_user_asm("ldw", ptr); break;   \
 	    case 8: LDD_USER(ptr);  break;		 \
 	    default: BUILD_BUG(); break;		 \
-	    }                                            \
 	}                                                \
 							 \
 	(x) = (__force __typeof__(*(ptr))) __gu_val;	 \
 	__gu_err;                                        \
 })
 
-#define __get_kernel_asm(ldx, ptr)                      \
-	__asm__("\n1:\t" ldx "\t0(%2),%0\n\t"		\
+#define __get_user_asm(ldx, ptr)                        \
+	__asm__("\n1:\t" ldx "\t0(%%sr2,%2),%0\n\t"	\
 		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_get_user_skip_1)\
 		: "=r"(__gu_val), "=r"(__gu_err)        \
 		: "r"(ptr), "1"(__gu_err)		\
 		: "r1");
 
-#define __get_user_asm(ldx, ptr)                        \
-	__asm__("\n1:\t" ldx "\t0(%%sr3,%2),%0\n\t"	\
-		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_get_user_skip_1)\
-		: "=r"(__gu_val), "=r"(__gu_err)        \
+#if !defined(CONFIG_64BIT)
+
+#define __get_user_asm64(ptr) 				\
+	__asm__("\n1:\tldw 0(%%sr2,%2),%0"		\
+		"\n2:\tldw 4(%%sr2,%2),%R0\n\t"		\
+		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_get_user_skip_2)\
+		ASM_EXCEPTIONTABLE_ENTRY(2b, fixup_get_user_skip_1)\
+		: "=r"(__gu_val), "=r"(__gu_err)	\
 		: "r"(ptr), "1"(__gu_err)		\
 		: "r1");
 
+#endif /* !defined(CONFIG_64BIT) */
+
+
 #define __put_user(x, ptr)                                      \
 ({								\
 	register long __pu_err __asm__ ("r8") = 0;      	\
         __typeof__(*(ptr)) __x = (__typeof__(*(ptr)))(x);	\
 								\
-	if (segment_eq(get_fs(), KERNEL_DS)) {                  \
-	    switch (sizeof(*(ptr))) {                           \
-	    case 1: __put_kernel_asm("stb", __x, ptr); break;   \
-	    case 2: __put_kernel_asm("sth", __x, ptr); break;   \
-	    case 4: __put_kernel_asm("stw", __x, ptr); break;   \
-	    case 8: STD_KERNEL(__x, ptr); break;		\
-	    default: BUILD_BUG(); break;			\
-	    }                                                   \
-	}                                                       \
-	else {                                                  \
-	    switch (sizeof(*(ptr))) {                           \
+	load_sr2();						\
+	switch (sizeof(*(ptr))) {				\
 	    case 1: __put_user_asm("stb", __x, ptr); break;     \
 	    case 2: __put_user_asm("sth", __x, ptr); break;     \
 	    case 4: __put_user_asm("stw", __x, ptr); break;     \
 	    case 8: STD_USER(__x, ptr); break;			\
 	    default: BUILD_BUG(); break;			\
-	    }                                                   \
 	}                                                       \
 								\
 	__pu_err;						\
@@ -159,17 +155,9 @@ struct exception_data {
  * r8/r9 are already listed as err/val.
  */
 
-#define __put_kernel_asm(stx, x, ptr)                       \
-	__asm__ __volatile__ (                              \
-		"\n1:\t" stx "\t%2,0(%1)\n\t"		    \
-		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_put_user_skip_1)\
-		: "=r"(__pu_err)                            \
-		: "r"(ptr), "r"(x), "0"(__pu_err)	    \
-	    	: "r1")
-
 #define __put_user_asm(stx, x, ptr)                         \
 	__asm__ __volatile__ (                              \
-		"\n1:\t" stx "\t%2,0(%%sr3,%1)\n\t"	    \
+		"\n1:\t" stx "\t%2,0(%%sr2,%1)\n\t"	    \
 		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_put_user_skip_1)\
 		: "=r"(__pu_err)                            \
 		: "r"(ptr), "r"(x), "0"(__pu_err)	    \
@@ -178,21 +166,10 @@ struct exception_data {
 
 #if !defined(CONFIG_64BIT)
 
-#define __put_kernel_asm64(__val, ptr) do {		    \
-	__asm__ __volatile__ (				    \
-		"\n1:\tstw %2,0(%1)"			    \
-		"\n2:\tstw %R2,4(%1)\n\t"		    \
-		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_put_user_skip_2)\
-		ASM_EXCEPTIONTABLE_ENTRY(2b, fixup_put_user_skip_1)\
-		: "=r"(__pu_err)                            \
-		: "r"(ptr), "r"(__val), "0"(__pu_err) \
-		: "r1");				    \
-} while (0)
-
 #define __put_user_asm64(__val, ptr) do {	    	    \
 	__asm__ __volatile__ (				    \
-		"\n1:\tstw %2,0(%%sr3,%1)"		    \
-		"\n2:\tstw %R2,4(%%sr3,%1)\n\t"		    \
+		"\n1:\tstw %2,0(%%sr2,%1)"		    \
+		"\n2:\tstw %R2,4(%%sr2,%1)\n\t"		    \
 		ASM_EXCEPTIONTABLE_ENTRY(1b, fixup_put_user_skip_2)\
 		ASM_EXCEPTIONTABLE_ENTRY(2b, fixup_put_user_skip_1)\
 		: "=r"(__pu_err)                            \
diff --git a/arch/parisc/include/uapi/asm/pdc.h b/arch/parisc/include/uapi/asm/pdc.h
index 702498f7705b..0609ff117f67 100644
--- a/arch/parisc/include/uapi/asm/pdc.h
+++ b/arch/parisc/include/uapi/asm/pdc.h
@@ -59,7 +59,7 @@
 #define PDC_MODEL_GET_BOOT__OP	8	/* returns boot test options	*/
 #define PDC_MODEL_SET_BOOT__OP	9	/* set boot test options	*/
 
-#define PA89_INSTRUCTION_SET	0x4	/* capatibilies returned	*/
+#define PA89_INSTRUCTION_SET	0x4	/* capabilities returned	*/
 #define PA90_INSTRUCTION_SET	0x8
 
 #define PDC_CACHE	5		/* return/set cache (& TLB) info*/
diff --git a/arch/parisc/include/uapi/asm/ptrace.h b/arch/parisc/include/uapi/asm/ptrace.h
index c4fa6c8b9ad9..02ce2eb99a7f 100644
--- a/arch/parisc/include/uapi/asm/ptrace.h
+++ b/arch/parisc/include/uapi/asm/ptrace.h
@@ -13,6 +13,11 @@
  * N.B. gdb/strace care about the size and offsets within this
  * structure. If you change things, you may break object compatibility
  * for those applications.
+ *
+ * Please do NOT use this structure for future programs, but use
+ * user_regs_struct (see below) instead.
+ *
+ * It can be accessed through PTRACE_PEEKUSR/PTRACE_POKEUSR only.
  */
 
 struct pt_regs {
@@ -33,6 +38,45 @@ struct pt_regs {
 	unsigned long ipsw;	/* CR22 */
 };
 
+/**
+ * struct user_regs_struct - User general purpose registers
+ *
+ * This is the user-visible general purpose register state structure
+ * which is used to define the elf_gregset_t.
+ *
+ * It can be accessed through PTRACE_GETREGSET with NT_PRSTATUS
+ * and through PTRACE_GETREGS.
+ */
+struct user_regs_struct {
+	unsigned long gr[32];	/* PSW is in gr[0] */
+	unsigned long sr[8];
+	unsigned long iaoq[2];
+	unsigned long iasq[2];
+	unsigned long sar;	/* CR11 */
+	unsigned long iir;	/* CR19 */
+	unsigned long isr;	/* CR20 */
+	unsigned long ior;	/* CR21 */
+	unsigned long ipsw;	/* CR22 */
+	unsigned long cr0;
+	unsigned long cr24, cr25, cr26, cr27, cr28, cr29, cr30, cr31;
+	unsigned long cr8, cr9, cr12, cr13, cr10, cr15;
+	unsigned long _pad[80-64];	/* pad to ELF_NGREG (80) */
+};
+
+/**
+ * struct user_fp_struct - User floating point registers
+ *
+ * This is the user-visible floating point register state structure.
+ * It uses the same layout and size as elf_fpregset_t.
+ *
+ * It can be accessed through PTRACE_GETREGSET with NT_PRFPREG
+ * and through PTRACE_GETFPREGS.
+ */
+struct user_fp_struct {
+	__u64 fr[32];
+};
+
+
 /*
  * The numbers chosen here are somewhat arbitrary but absolutely MUST
  * not overlap with any of the number assigned in <linux/ptrace.h>.
@@ -43,5 +87,9 @@ struct pt_regs {
  */
 #define PTRACE_SINGLEBLOCK	12	/* resume execution until next branch */
 
+#define PTRACE_GETREGS		18
+#define PTRACE_SETREGS		19
+#define PTRACE_GETFPREGS	14
+#define PTRACE_SETFPREGS	15
 
 #endif /* _UAPI_PARISC_PTRACE_H */
diff --git a/arch/parisc/include/uapi/asm/unistd.h b/arch/parisc/include/uapi/asm/unistd.h
index cc0ce92c93c7..a9b9407f38f7 100644
--- a/arch/parisc/include/uapi/asm/unistd.h
+++ b/arch/parisc/include/uapi/asm/unistd.h
@@ -102,7 +102,7 @@
 #define __NR_uselib              (__NR_Linux + 86)
 #define __NR_swapon              (__NR_Linux + 87)
 #define __NR_reboot              (__NR_Linux + 88)
-#define __NR_mmap2             (__NR_Linux + 89)
+#define __NR_mmap2               (__NR_Linux + 89)
 #define __NR_mmap                (__NR_Linux + 90)
 #define __NR_munmap              (__NR_Linux + 91)
 #define __NR_truncate            (__NR_Linux + 92)
@@ -114,7 +114,7 @@
 #define __NR_recv                (__NR_Linux + 98)
 #define __NR_statfs              (__NR_Linux + 99)
 #define __NR_fstatfs            (__NR_Linux + 100)
-#define __NR_stat64           (__NR_Linux + 101)
+#define __NR_stat64             (__NR_Linux + 101)
 /* #define __NR_socketcall         (__NR_Linux + 102) */
 #define __NR_syslog             (__NR_Linux + 103)
 #define __NR_setitimer          (__NR_Linux + 104)
@@ -140,17 +140,17 @@
 #define __NR_adjtimex           (__NR_Linux + 124)
 #define __NR_mprotect           (__NR_Linux + 125)
 #define __NR_sigprocmask        (__NR_Linux + 126)
-#define __NR_create_module      (__NR_Linux + 127)
+#define __NR_create_module      (__NR_Linux + 127) /* not used */
 #define __NR_init_module        (__NR_Linux + 128)
 #define __NR_delete_module      (__NR_Linux + 129)
-#define __NR_get_kernel_syms    (__NR_Linux + 130)
+#define __NR_get_kernel_syms    (__NR_Linux + 130) /* not used */
 #define __NR_quotactl           (__NR_Linux + 131)
 #define __NR_getpgid            (__NR_Linux + 132)
 #define __NR_fchdir             (__NR_Linux + 133)
 #define __NR_bdflush            (__NR_Linux + 134)
 #define __NR_sysfs              (__NR_Linux + 135)
 #define __NR_personality        (__NR_Linux + 136)
-#define __NR_afs_syscall        (__NR_Linux + 137) /* Syscall for Andrew File System */
+#define __NR_afs_syscall        (__NR_Linux + 137) /* not used */
 #define __NR_setfsuid           (__NR_Linux + 138)
 #define __NR_setfsgid           (__NR_Linux + 139)
 #define __NR__llseek            (__NR_Linux + 140)
@@ -180,9 +180,9 @@
 #define __NR_setresuid          (__NR_Linux + 164)
 #define __NR_getresuid          (__NR_Linux + 165)
 #define __NR_sigaltstack        (__NR_Linux + 166)
-#define __NR_query_module       (__NR_Linux + 167)
+#define __NR_query_module       (__NR_Linux + 167) /* not used */
 #define __NR_poll               (__NR_Linux + 168)
-#define __NR_nfsservctl         (__NR_Linux + 169)
+#define __NR_nfsservctl         (__NR_Linux + 169) /* not used */
 #define __NR_setresgid          (__NR_Linux + 170)
 #define __NR_getresgid          (__NR_Linux + 171)
 #define __NR_prctl              (__NR_Linux + 172)
@@ -209,18 +209,16 @@
 #define __NR_shmdt              (__NR_Linux + 193)
 #define __NR_shmget             (__NR_Linux + 194)
 #define __NR_shmctl             (__NR_Linux + 195)
-
-#define __NR_getpmsg		(__NR_Linux + 196) /* Somebody *wants* streams? */
-#define __NR_putpmsg		(__NR_Linux + 197)
-
+#define __NR_getpmsg            (__NR_Linux + 196) /* not used */
+#define __NR_putpmsg            (__NR_Linux + 197) /* not used */
 #define __NR_lstat64            (__NR_Linux + 198)
 #define __NR_truncate64         (__NR_Linux + 199)
 #define __NR_ftruncate64        (__NR_Linux + 200)
 #define __NR_getdents64         (__NR_Linux + 201)
 #define __NR_fcntl64            (__NR_Linux + 202)
-#define __NR_attrctl            (__NR_Linux + 203)
-#define __NR_acl_get            (__NR_Linux + 204)
-#define __NR_acl_set            (__NR_Linux + 205)
+#define __NR_attrctl            (__NR_Linux + 203) /* not used */
+#define __NR_acl_get            (__NR_Linux + 204) /* not used */
+#define __NR_acl_set            (__NR_Linux + 205) /* not used */
 #define __NR_gettid             (__NR_Linux + 206)
 #define __NR_readahead          (__NR_Linux + 207)
 #define __NR_tkill              (__NR_Linux + 208)
@@ -228,8 +226,8 @@
 #define __NR_futex              (__NR_Linux + 210)
 #define __NR_sched_setaffinity  (__NR_Linux + 211)
 #define __NR_sched_getaffinity  (__NR_Linux + 212)
-#define __NR_set_thread_area    (__NR_Linux + 213)
-#define __NR_get_thread_area    (__NR_Linux + 214)
+#define __NR_set_thread_area    (__NR_Linux + 213) /* not used */
+#define __NR_get_thread_area    (__NR_Linux + 214) /* not used */
 #define __NR_io_setup           (__NR_Linux + 215)
 #define __NR_io_destroy         (__NR_Linux + 216)
 #define __NR_io_getevents       (__NR_Linux + 217)
@@ -278,7 +276,7 @@
 #define __NR_mbind		(__NR_Linux + 260)
 #define __NR_get_mempolicy	(__NR_Linux + 261)
 #define __NR_set_mempolicy	(__NR_Linux + 262)
-#define __NR_vserver		(__NR_Linux + 263)
+#define __NR_vserver		(__NR_Linux + 263) /* not used */
 #define __NR_add_key		(__NR_Linux + 264)
 #define __NR_request_key	(__NR_Linux + 265)
 #define __NR_keyctl		(__NR_Linux + 266)
@@ -318,7 +316,7 @@
 #define __NR_kexec_load		(__NR_Linux + 300)
 #define __NR_utimensat		(__NR_Linux + 301)
 #define __NR_signalfd		(__NR_Linux + 302)
-#define __NR_timerfd		(__NR_Linux + 303)
+#define __NR_timerfd		(__NR_Linux + 303) /* not used */
 #define __NR_eventfd		(__NR_Linux + 304)
 #define __NR_fallocate		(__NR_Linux + 305)
 #define __NR_timerfd_create	(__NR_Linux + 306)
diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S
index 39127d3e70e5..baa3d9d6e971 100644
--- a/arch/parisc/kernel/entry.S
+++ b/arch/parisc/kernel/entry.S
@@ -667,7 +667,7 @@
 	 * boundary
 	 */
 
-	.text
+	.section .text.hot
 	.align 2048
 
 ENTRY(fault_vector_20)
@@ -2019,6 +2019,7 @@ ftrace_stub:
 	.procend
 ENDPROC(mcount)
 
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 	.align 8
 	.globl return_to_handler
 	.type  return_to_handler, @function
@@ -2040,11 +2041,17 @@ parisc_return_to_handler:
 #endif
 
 	/* call ftrace_return_to_handler(0) */
+	.import ftrace_return_to_handler,code
+	load32 ftrace_return_to_handler,%ret0
+	load32 .Lftrace_ret,%r2
 #ifdef CONFIG_64BIT
 	ldo -16(%sp),%ret1		/* Reference param save area */
+	bve	(%ret0)
+#else
+	bv	%r0(%ret0)
 #endif
-	BL ftrace_return_to_handler,%r2
 	ldi 0,%r26
+.Lftrace_ret:
 	copy %ret0,%rp
 
 	/* restore original return values */
@@ -2062,6 +2069,8 @@ parisc_return_to_handler:
 	.procend
 ENDPROC(return_to_handler)
 
+#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
 #endif	/* CONFIG_FUNCTION_TRACER */
 
 #ifdef CONFIG_IRQSTACKS
diff --git a/arch/parisc/kernel/ftrace.c b/arch/parisc/kernel/ftrace.c
index b13f9ec6f294..a828a0adf52c 100644
--- a/arch/parisc/kernel/ftrace.c
+++ b/arch/parisc/kernel/ftrace.c
@@ -18,12 +18,15 @@
 #include <asm/ftrace.h>
 
 
+#define __hot __attribute__ ((__section__ (".text.hot")))
+
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 /*
  * Hook the return address and push it in the stack of return addrs
  * in current thread info.
  */
-static void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
+static void __hot prepare_ftrace_return(unsigned long *parent,
+					unsigned long self_addr)
 {
 	unsigned long old;
 	struct ftrace_graph_ent trace;
@@ -53,7 +56,7 @@ static void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr
 }
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
 
-void notrace ftrace_function_trampoline(unsigned long parent,
+void notrace __hot ftrace_function_trampoline(unsigned long parent,
 				unsigned long self_addr,
 				unsigned long org_sp_gr3)
 {
diff --git a/arch/parisc/kernel/ptrace.c b/arch/parisc/kernel/ptrace.c
index 8fb81a391599..b5458b37fc5b 100644
--- a/arch/parisc/kernel/ptrace.c
+++ b/arch/parisc/kernel/ptrace.c
@@ -4,18 +4,20 @@
  * Copyright (C) 2000 Hewlett-Packard Co, Linuxcare Inc.
  * Copyright (C) 2000 Matthew Wilcox <matthew@wil.cx>
  * Copyright (C) 2000 David Huggins-Daines <dhd@debian.org>
- * Copyright (C) 2008 Helge Deller <deller@gmx.de>
+ * Copyright (C) 2008-2016 Helge Deller <deller@gmx.de>
  */
 
 #include <linux/kernel.h>
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
+#include <linux/elf.h>
 #include <linux/errno.h>
 #include <linux/ptrace.h>
 #include <linux/tracehook.h>
 #include <linux/user.h>
 #include <linux/personality.h>
+#include <linux/regset.h>
 #include <linux/security.h>
 #include <linux/seccomp.h>
 #include <linux/compat.h>
@@ -30,6 +32,17 @@
 /* PSW bits we allow the debugger to modify */
 #define USER_PSW_BITS	(PSW_N | PSW_B | PSW_V | PSW_CB)
 
+#define CREATE_TRACE_POINTS
+#include <trace/events/syscalls.h>
+
+/*
+ * These are our native regset flavors.
+ */
+enum parisc_regset {
+	REGSET_GENERAL,
+	REGSET_FP
+};
+
 /*
  * Called by kernel/ptrace.c when detaching..
  *
@@ -114,6 +127,7 @@ void user_enable_block_step(struct task_struct *task)
 long arch_ptrace(struct task_struct *child, long request,
 		 unsigned long addr, unsigned long data)
 {
+	unsigned long __user *datap = (unsigned long __user *)data;
 	unsigned long tmp;
 	long ret = -EIO;
 
@@ -126,7 +140,7 @@ long arch_ptrace(struct task_struct *child, long request,
 		     addr >= sizeof(struct pt_regs))
 			break;
 		tmp = *(unsigned long *) ((char *) task_regs(child) + addr);
-		ret = put_user(tmp, (unsigned long __user *) data);
+		ret = put_user(tmp, datap);
 		break;
 
 	/* Write the word at location addr in the USER area.  This will need
@@ -165,6 +179,34 @@ long arch_ptrace(struct task_struct *child, long request,
 		}
 		break;
 
+	case PTRACE_GETREGS:	/* Get all gp regs from the child. */
+		return copy_regset_to_user(child,
+					   task_user_regset_view(current),
+					   REGSET_GENERAL,
+					   0, sizeof(struct user_regs_struct),
+					   datap);
+
+	case PTRACE_SETREGS:	/* Set all gp regs in the child. */
+		return copy_regset_from_user(child,
+					     task_user_regset_view(current),
+					     REGSET_GENERAL,
+					     0, sizeof(struct user_regs_struct),
+					     datap);
+
+	case PTRACE_GETFPREGS:	/* Get the child FPU state. */
+		return copy_regset_to_user(child,
+					   task_user_regset_view(current),
+					   REGSET_FP,
+					   0, sizeof(struct user_fp_struct),
+					   datap);
+
+	case PTRACE_SETFPREGS:	/* Set the child FPU state. */
+		return copy_regset_from_user(child,
+					     task_user_regset_view(current),
+					     REGSET_FP,
+					     0, sizeof(struct user_fp_struct),
+					     datap);
+
 	default:
 		ret = ptrace_request(child, request, addr, data);
 		break;
@@ -283,6 +325,10 @@ long do_syscall_trace_enter(struct pt_regs *regs)
 		regs->gr[20] = -1UL;
 		goto out;
 	}
+#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
+	if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
+		trace_sys_enter(regs, regs->gr[20]);
+#endif
 
 #ifdef CONFIG_64BIT
 	if (!is_compat_task())
@@ -311,6 +357,324 @@ void do_syscall_trace_exit(struct pt_regs *regs)
 
 	audit_syscall_exit(regs);
 
+#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
+	if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
+		trace_sys_exit(regs, regs->gr[20]);
+#endif
+
 	if (stepping || test_thread_flag(TIF_SYSCALL_TRACE))
 		tracehook_report_syscall_exit(regs, stepping);
 }
+
+
+/*
+ * regset functions.
+ */
+
+static int fpr_get(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     void *kbuf, void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	__u64 *k = kbuf;
+	__u64 __user *u = ubuf;
+	__u64 reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NFPREG; --count)
+			*k++ = regs->fr[pos++];
+	else
+		for (; count > 0 && pos < ELF_NFPREG; --count)
+			if (__put_user(regs->fr[pos++], u++))
+				return -EFAULT;
+
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyout_zero(&pos, &count, &kbuf, &ubuf,
+					ELF_NFPREG * sizeof(reg), -1);
+}
+
+static int fpr_set(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     const void *kbuf, const void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	const __u64 *k = kbuf;
+	const __u64 __user *u = ubuf;
+	__u64 reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NFPREG; --count)
+			regs->fr[pos++] = *k++;
+	else
+		for (; count > 0 && pos < ELF_NFPREG; --count) {
+			if (__get_user(reg, u++))
+				return -EFAULT;
+			regs->fr[pos++] = reg;
+		}
+
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyin_ignore(&pos, &count, &kbuf, &ubuf,
+					 ELF_NFPREG * sizeof(reg), -1);
+}
+
+#define RI(reg) (offsetof(struct user_regs_struct,reg) / sizeof(long))
+
+static unsigned long get_reg(struct pt_regs *regs, int num)
+{
+	switch (num) {
+	case RI(gr[0]) ... RI(gr[31]):	return regs->gr[num - RI(gr[0])];
+	case RI(sr[0]) ... RI(sr[7]):	return regs->sr[num - RI(sr[0])];
+	case RI(iasq[0]):		return regs->iasq[0];
+	case RI(iasq[1]):		return regs->iasq[1];
+	case RI(iaoq[0]):		return regs->iaoq[0];
+	case RI(iaoq[1]):		return regs->iaoq[1];
+	case RI(sar):			return regs->sar;
+	case RI(iir):			return regs->iir;
+	case RI(isr):			return regs->isr;
+	case RI(ior):			return regs->ior;
+	case RI(ipsw):			return regs->ipsw;
+	case RI(cr27):			return regs->cr27;
+	case RI(cr0):			return mfctl(0);
+	case RI(cr24):			return mfctl(24);
+	case RI(cr25):			return mfctl(25);
+	case RI(cr26):			return mfctl(26);
+	case RI(cr28):			return mfctl(28);
+	case RI(cr29):			return mfctl(29);
+	case RI(cr30):			return mfctl(30);
+	case RI(cr31):			return mfctl(31);
+	case RI(cr8):			return mfctl(8);
+	case RI(cr9):			return mfctl(9);
+	case RI(cr12):			return mfctl(12);
+	case RI(cr13):			return mfctl(13);
+	case RI(cr10):			return mfctl(10);
+	case RI(cr15):			return mfctl(15);
+	default:			return 0;
+	}
+}
+
+static void set_reg(struct pt_regs *regs, int num, unsigned long val)
+{
+	switch (num) {
+	case RI(gr[0]): /*
+			 * PSW is in gr[0].
+			 * Allow writing to Nullify, Divide-step-correction,
+			 * and carry/borrow bits.
+			 * BEWARE, if you set N, and then single step, it won't
+			 * stop on the nullified instruction.
+			 */
+			val &= USER_PSW_BITS;
+			regs->gr[0] &= ~USER_PSW_BITS;
+			regs->gr[0] |= val;
+			return;
+	case RI(gr[1]) ... RI(gr[31]):
+			regs->gr[num - RI(gr[0])] = val;
+			return;
+	case RI(iaoq[0]):
+	case RI(iaoq[1]):
+			regs->iaoq[num - RI(iaoq[0])] = val;
+			return;
+	case RI(sar):	regs->sar = val;
+			return;
+	default:	return;
+#if 0
+	/* do not allow to change any of the following registers (yet) */
+	case RI(sr[0]) ... RI(sr[7]):	return regs->sr[num - RI(sr[0])];
+	case RI(iasq[0]):		return regs->iasq[0];
+	case RI(iasq[1]):		return regs->iasq[1];
+	case RI(iir):			return regs->iir;
+	case RI(isr):			return regs->isr;
+	case RI(ior):			return regs->ior;
+	case RI(ipsw):			return regs->ipsw;
+	case RI(cr27):			return regs->cr27;
+        case cr0, cr24, cr25, cr26, cr27, cr28, cr29, cr30, cr31;
+        case cr8, cr9, cr12, cr13, cr10, cr15;
+#endif
+	}
+}
+
+static int gpr_get(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     void *kbuf, void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	unsigned long *k = kbuf;
+	unsigned long __user *u = ubuf;
+	unsigned long reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			*k++ = get_reg(regs, pos++);
+	else
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			if (__put_user(get_reg(regs, pos++), u++))
+				return -EFAULT;
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyout_zero(&pos, &count, &kbuf, &ubuf,
+					ELF_NGREG * sizeof(reg), -1);
+}
+
+static int gpr_set(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     const void *kbuf, const void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	const unsigned long *k = kbuf;
+	const unsigned long __user *u = ubuf;
+	unsigned long reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			set_reg(regs, pos++, *k++);
+	else
+		for (; count > 0 && pos < ELF_NGREG; --count) {
+			if (__get_user(reg, u++))
+				return -EFAULT;
+			set_reg(regs, pos++, reg);
+		}
+
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyin_ignore(&pos, &count, &kbuf, &ubuf,
+					 ELF_NGREG * sizeof(reg), -1);
+}
+
+static const struct user_regset native_regsets[] = {
+	[REGSET_GENERAL] = {
+		.core_note_type = NT_PRSTATUS, .n = ELF_NGREG,
+		.size = sizeof(long), .align = sizeof(long),
+		.get = gpr_get, .set = gpr_set
+	},
+	[REGSET_FP] = {
+		.core_note_type = NT_PRFPREG, .n = ELF_NFPREG,
+		.size = sizeof(__u64), .align = sizeof(__u64),
+		.get = fpr_get, .set = fpr_set
+	}
+};
+
+static const struct user_regset_view user_parisc_native_view = {
+	.name = "parisc", .e_machine = ELF_ARCH, .ei_osabi = ELFOSABI_LINUX,
+	.regsets = native_regsets, .n = ARRAY_SIZE(native_regsets)
+};
+
+#ifdef CONFIG_64BIT
+#include <linux/compat.h>
+
+static int gpr32_get(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     void *kbuf, void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	compat_ulong_t *k = kbuf;
+	compat_ulong_t __user *u = ubuf;
+	compat_ulong_t reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			*k++ = get_reg(regs, pos++);
+	else
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			if (__put_user((compat_ulong_t) get_reg(regs, pos++), u++))
+				return -EFAULT;
+
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyout_zero(&pos, &count, &kbuf, &ubuf,
+					ELF_NGREG * sizeof(reg), -1);
+}
+
+static int gpr32_set(struct task_struct *target,
+		     const struct user_regset *regset,
+		     unsigned int pos, unsigned int count,
+		     const void *kbuf, const void __user *ubuf)
+{
+	struct pt_regs *regs = task_regs(target);
+	const compat_ulong_t *k = kbuf;
+	const compat_ulong_t __user *u = ubuf;
+	compat_ulong_t reg;
+
+	pos /= sizeof(reg);
+	count /= sizeof(reg);
+
+	if (kbuf)
+		for (; count > 0 && pos < ELF_NGREG; --count)
+			set_reg(regs, pos++, *k++);
+	else
+		for (; count > 0 && pos < ELF_NGREG; --count) {
+			if (__get_user(reg, u++))
+				return -EFAULT;
+			set_reg(regs, pos++, reg);
+		}
+
+	kbuf = k;
+	ubuf = u;
+	pos *= sizeof(reg);
+	count *= sizeof(reg);
+	return user_regset_copyin_ignore(&pos, &count, &kbuf, &ubuf,
+					 ELF_NGREG * sizeof(reg), -1);
+}
+
+/*
+ * These are the regset flavors matching the 32bit native set.
+ */
+static const struct user_regset compat_regsets[] = {
+	[REGSET_GENERAL] = {
+		.core_note_type = NT_PRSTATUS, .n = ELF_NGREG,
+		.size = sizeof(compat_long_t), .align = sizeof(compat_long_t),
+		.get = gpr32_get, .set = gpr32_set
+	},
+	[REGSET_FP] = {
+		.core_note_type = NT_PRFPREG, .n = ELF_NFPREG,
+		.size = sizeof(__u64), .align = sizeof(__u64),
+		.get = fpr_get, .set = fpr_set
+	}
+};
+
+static const struct user_regset_view user_parisc_compat_view = {
+	.name = "parisc", .e_machine = EM_PARISC, .ei_osabi = ELFOSABI_LINUX,
+	.regsets = compat_regsets, .n = ARRAY_SIZE(compat_regsets)
+};
+#endif	/* CONFIG_64BIT */
+
+const struct user_regset_view *task_user_regset_view(struct task_struct *task)
+{
+	BUILD_BUG_ON(sizeof(struct user_regs_struct)/sizeof(long) != ELF_NGREG);
+	BUILD_BUG_ON(sizeof(struct user_fp_struct)/sizeof(__u64) != ELF_NFPREG);
+#ifdef CONFIG_64BIT
+	if (is_compat_task())
+		return &user_parisc_compat_view;
+#endif
+	return &user_parisc_native_view;
+}
diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S
index 57b4836b7ecd..d03422e5f188 100644
--- a/arch/parisc/kernel/syscall.S
+++ b/arch/parisc/kernel/syscall.S
@@ -912,6 +912,7 @@ END(lws_table)
 
 	.align 8
 ENTRY(sys_call_table)
+	.export sys_call_table,data
 #include "syscall_table.S"
 END(sys_call_table)
 
diff --git a/arch/parisc/kernel/time.c b/arch/parisc/kernel/time.c
index 400acac0a304..58dd6801f5be 100644
--- a/arch/parisc/kernel/time.c
+++ b/arch/parisc/kernel/time.c
@@ -38,6 +38,18 @@
 
 static unsigned long clocktick __read_mostly;	/* timer cycles per tick */
 
+#ifndef CONFIG_64BIT
+/*
+ * The processor-internal cycle counter (Control Register 16) is used as time
+ * source for the sched_clock() function.  This register is 64bit wide on a
+ * 64-bit kernel and 32bit on a 32-bit kernel. Since sched_clock() always
+ * requires a 64bit counter we emulate on the 32-bit kernel the higher 32bits
+ * with a per-cpu variable which we increase every time the counter
+ * wraps-around (which happens every ~4 secounds).
+ */
+static DEFINE_PER_CPU(unsigned long, cr16_high_32_bits);
+#endif
+
 /*
  * We keep time on PA-RISC Linux by using the Interval Timer which is
  * a pair of registers; one is read-only and one is write-only; both
@@ -108,6 +120,12 @@ irqreturn_t __irq_entry timer_interrupt(int irq, void *dev_id)
 	 */
 	mtctl(next_tick, 16);
 
+#if !defined(CONFIG_64BIT)
+	/* check for overflow on a 32bit kernel (every ~4 seconds). */
+	if (unlikely(next_tick < now))
+		this_cpu_inc(cr16_high_32_bits);
+#endif
+
 	/* Skip one clocktick on purpose if we missed next_tick.
 	 * The new CR16 must be "later" than current CR16 otherwise
 	 * itimer would not fire until CR16 wrapped - e.g 4 seconds
@@ -219,6 +237,12 @@ void __init start_cpu_itimer(void)
 	unsigned int cpu = smp_processor_id();
 	unsigned long next_tick = mfctl(16) + clocktick;
 
+#if defined(CONFIG_HAVE_UNSTABLE_SCHED_CLOCK) && defined(CONFIG_64BIT)
+	/* With multiple 64bit CPUs online, the cr16's are not syncronized. */
+	if (cpu != 0)
+		clear_sched_clock_stable();
+#endif
+
 	mtctl(next_tick, 16);		/* kick off Interval Timer (CR16) */
 
 	per_cpu(cpu_data, cpu).it_value = next_tick;
@@ -246,15 +270,52 @@ void read_persistent_clock(struct timespec *ts)
 	}
 }
 
+
+/*
+ * sched_clock() framework
+ */
+
+static u32 cyc2ns_mul __read_mostly;
+static u32 cyc2ns_shift __read_mostly;
+
+u64 sched_clock(void)
+{
+	u64 now;
+
+	/* Get current cycle counter (Control Register 16). */
+#ifdef CONFIG_64BIT
+	now = mfctl(16);
+#else
+	now = mfctl(16) + (((u64) this_cpu_read(cr16_high_32_bits)) << 32);
+#endif
+
+	/* return the value in ns (cycles_2_ns) */
+	return mul_u64_u32_shr(now, cyc2ns_mul, cyc2ns_shift);
+}
+
+
+/*
+ * timer interrupt and sched_clock() initialization
+ */
+
 void __init time_init(void)
 {
 	unsigned long current_cr16_khz;
 
+	current_cr16_khz = PAGE0->mem_10msec/10;  /* kHz */
 	clocktick = (100 * PAGE0->mem_10msec) / HZ;
 
+	/* calculate mult/shift values for cr16 */
+	clocks_calc_mult_shift(&cyc2ns_mul, &cyc2ns_shift, current_cr16_khz,
+				NSEC_PER_MSEC, 0);
+
+#if defined(CONFIG_HAVE_UNSTABLE_SCHED_CLOCK) && defined(CONFIG_64BIT)
+	/* At bootup only one 64bit CPU is online and cr16 is "stable" */
+	set_sched_clock_stable();
+#endif
+
 	start_cpu_itimer();	/* get CPU 0 started */
 
 	/* register at clocksource framework */
-	current_cr16_khz = PAGE0->mem_10msec/10;  /* kHz */
 	clocksource_register_khz(&clocksource_cr16, current_cr16_khz);
 }
diff --git a/arch/parisc/lib/bitops.c b/arch/parisc/lib/bitops.c
index 187118841af1..8e45b0a97abf 100644
--- a/arch/parisc/lib/bitops.c
+++ b/arch/parisc/lib/bitops.c
@@ -55,11 +55,10 @@ unsigned long __xchg8(char x, char *ptr)
 }
 
 
-#ifdef CONFIG_64BIT
-unsigned long __cmpxchg_u64(volatile unsigned long *ptr, unsigned long old, unsigned long new)
+u64 __cmpxchg_u64(volatile u64 *ptr, u64 old, u64 new)
 {
 	unsigned long flags;
-	unsigned long prev;
+	u64 prev;
 
 	_atomic_spin_lock_irqsave(ptr, flags);
 	if ((prev = *ptr) == old)
@@ -67,7 +66,6 @@ unsigned long __cmpxchg_u64(volatile unsigned long *ptr, unsigned long old, unsi
 	_atomic_spin_unlock_irqrestore(ptr, flags);
 	return prev;
 }
-#endif
 
 unsigned long __cmpxchg_u32(volatile unsigned int *ptr, unsigned int old, unsigned int new)
 {
diff --git a/arch/parisc/math-emu/fpudispatch.c b/arch/parisc/math-emu/fpudispatch.c
index 673b73e8420d..18df1237c93c 100644
--- a/arch/parisc/math-emu/fpudispatch.c
+++ b/arch/parisc/math-emu/fpudispatch.c
@@ -184,7 +184,7 @@ static void parisc_linux_get_fpu_type(u_int fpregs[])
 
 /*
  * this routine will decode the excepting floating point instruction and
- * call the approiate emulation routine.
+ * call the appropriate emulation routine.
  * It is called by decode_fpu with the following parameters:
  * fpudispatch(current_ir, unimplemented_code, 0, &Fpu_register)
  * where current_ir is the instruction to be emulated,