Merge branch 'work.uaccess' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull uaccess unification updates from Al Viro:
 "This is the uaccess unification pile. It's _not_ the end of uaccess
  work, but the next batch of that will go into the next cycle. This one
  mostly takes copy_from_user() and friends out of arch/* and gets the
  zero-padding behaviour in sync for all architectures.

  Dealing with the nocache/writethrough mess is for the next cycle;
  fortunately, that's x86-only. Same for cleanups in iov_iter.c (I am
  sold on access_ok() in there, BTW; just not in this pile), same for
  reducing __copy_... callsites, strn*... stuff, etc. - there will be a
  pile about as large as this one in the next merge window.

  This one sat in -next for weeks. -3KLoC"

* 'work.uaccess' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (96 commits)
  HAVE_ARCH_HARDENED_USERCOPY is unconditional now
  CONFIG_ARCH_HAS_RAW_COPY_USER is unconditional now
  m32r: switch to RAW_COPY_USER
  hexagon: switch to RAW_COPY_USER
  microblaze: switch to RAW_COPY_USER
  get rid of padding, switch to RAW_COPY_USER
  ia64: get rid of copy_in_user()
  ia64: sanitize __access_ok()
  ia64: get rid of 'segment' argument of __do_{get,put}_user()
  ia64: get rid of 'segment' argument of __{get,put}_user_check()
  ia64: add extable.h
  powerpc: get rid of zeroing, switch to RAW_COPY_USER
  esas2r: don't open-code memdup_user()
  alpha: fix stack smashing in old_adjtimex(2)
  don't open-code kernel_setsockopt()
  mips: switch to RAW_COPY_USER
  mips: get rid of tail-zeroing in primitives
  mips: make copy_from_user() zero tail explicitly
  mips: clean and reorder the forest of macros...
  mips: consolidate __invoke_... wrappers
  ...

9 files changed, 316 insertions, 558 deletions

diff --git a/arch/alpha/include/asm/extable.h b/arch/alpha/include/asm/extable.h
new file mode 100644
index 000000000000..048e209e524c
--- /dev/null
+++ b/arch/alpha/include/asm/extable.h
@@ -0,0 +1,55 @@
+#ifndef _ASM_EXTABLE_H
+#define _ASM_EXTABLE_H
+
+/*
+ * About the exception table:
+ *
+ * - insn is a 32-bit pc-relative offset from the faulting insn.
+ * - nextinsn is a 16-bit offset off of the faulting instruction
+ *   (not off of the *next* instruction as branches are).
+ * - errreg is the register in which to place -EFAULT.
+ * - valreg is the final target register for the load sequence
+ *   and will be zeroed.
+ *
+ * Either errreg or valreg may be $31, in which case nothing happens.
+ *
+ * The exception fixup information "just so happens" to be arranged
+ * as in a MEM format instruction.  This lets us emit our three
+ * values like so:
+ *
+ *      lda valreg, nextinsn(errreg)
+ *
+ */
+
+struct exception_table_entry
+{
+	signed int insn;
+	union exception_fixup {
+		unsigned unit;
+		struct {
+			signed int nextinsn : 16;
+			unsigned int errreg : 5;
+			unsigned int valreg : 5;
+		} bits;
+	} fixup;
+};
+
+/* Returns the new pc */
+#define fixup_exception(map_reg, _fixup, pc)			\
+({								\
+	if ((_fixup)->fixup.bits.valreg != 31)			\
+		map_reg((_fixup)->fixup.bits.valreg) = 0;	\
+	if ((_fixup)->fixup.bits.errreg != 31)			\
+		map_reg((_fixup)->fixup.bits.errreg) = -EFAULT;	\
+	(pc) + (_fixup)->fixup.bits.nextinsn;			\
+})
+
+#define ARCH_HAS_RELATIVE_EXTABLE
+
+#define swap_ex_entry_fixup(a, b, tmp, delta)			\
+	do {							\
+		(a)->fixup.unit = (b)->fixup.unit;		\
+		(b)->fixup.unit = (tmp).fixup.unit;		\
+	} while (0)
+
+#endif
diff --git a/arch/alpha/include/asm/futex.h b/arch/alpha/include/asm/futex.h
index f939794363ac..fb01dfb760c2 100644
--- a/arch/alpha/include/asm/futex.h
+++ b/arch/alpha/include/asm/futex.h
@@ -19,12 +19,8 @@
 	"3:	.subsection 2\n"				\
 	"4:	br	1b\n"					\
 	"	.previous\n"					\
-	"	.section __ex_table,\"a\"\n"			\
-	"	.long	1b-.\n"					\
-	"	lda	$31,3b-1b(%1)\n"			\
-	"	.long	2b-.\n"					\
-	"	lda	$31,3b-2b(%1)\n"			\
-	"	.previous\n"					\
+	EXC(1b,3b,%1,$31)					\
+	EXC(2b,3b,%1,$31)					\
 	:	"=&r" (oldval), "=&r"(ret)			\
 	:	"r" (uaddr), "r"(oparg)				\
 	:	"memory")
@@ -101,12 +97,8 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
 	"3:	.subsection 2\n"
 	"4:	br	1b\n"
 	"	.previous\n"
-	"	.section __ex_table,\"a\"\n"
-	"	.long	1b-.\n"
-	"	lda	$31,3b-1b(%0)\n"
-	"	.long	2b-.\n"
-	"	lda	$31,3b-2b(%0)\n"
-	"	.previous\n"
+	EXC(1b,3b,%0,$31)
+	EXC(2b,3b,%0,$31)
 	:	"+r"(ret), "=&r"(prev), "=&r"(cmp)
 	:	"r"(uaddr), "r"((long)(int)oldval), "r"(newval)
 	:	"memory");
diff --git a/arch/alpha/include/asm/uaccess.h b/arch/alpha/include/asm/uaccess.h
index 94f587535dee..7b82dc9a8556 100644
--- a/arch/alpha/include/asm/uaccess.h
+++ b/arch/alpha/include/asm/uaccess.h
@@ -1,10 +1,6 @@
 #ifndef __ALPHA_UACCESS_H
 #define __ALPHA_UACCESS_H
 
-#include <linux/errno.h>
-#include <linux/sched.h>
-
-
 /*
  * The fs value determines whether argument validity checking should be
  * performed or not.  If get_fs() == USER_DS, checking is performed, with
@@ -20,9 +16,6 @@
 #define KERNEL_DS	((mm_segment_t) { 0UL })
 #define USER_DS		((mm_segment_t) { -0x40000000000UL })
 
-#define VERIFY_READ	0
-#define VERIFY_WRITE	1
-
 #define get_fs()  (current_thread_info()->addr_limit)
 #define get_ds()  (KERNEL_DS)
 #define set_fs(x) (current_thread_info()->addr_limit = (x))
@@ -39,13 +32,13 @@
  *  - AND "addr+size" doesn't have any high-bits set
  *  - OR we are in kernel mode.
  */
-#define __access_ok(addr, size, segment) \
-	(((segment).seg & (addr | size | (addr+size))) == 0)
+#define __access_ok(addr, size) \
+	((get_fs().seg & (addr | size | (addr+size))) == 0)
 
-#define access_ok(type, addr, size)				\
-({								\
-	__chk_user_ptr(addr);					\
-	__access_ok(((unsigned long)(addr)), (size), get_fs());	\
+#define access_ok(type, addr, size)			\
+({							\
+	__chk_user_ptr(addr);				\
+	__access_ok(((unsigned long)(addr)), (size));	\
 })
 
 /*
@@ -61,9 +54,9 @@
  * (b) require any knowledge of processes at this stage
  */
 #define put_user(x, ptr) \
-  __put_user_check((__typeof__(*(ptr)))(x), (ptr), sizeof(*(ptr)), get_fs())
+  __put_user_check((__typeof__(*(ptr)))(x), (ptr), sizeof(*(ptr)))
 #define get_user(x, ptr) \
-  __get_user_check((x), (ptr), sizeof(*(ptr)), get_fs())
+  __get_user_check((x), (ptr), sizeof(*(ptr)))
 
 /*
  * The "__xxx" versions do not do address space checking, useful when
@@ -81,6 +74,11 @@
  * more extensive comments with fixup_inline_exception below for
  * more information.
  */
+#define EXC(label,cont,res,err)				\
+	".section __ex_table,\"a\"\n"			\
+	"	.long "#label"-.\n"			\
+	"	lda "#res","#cont"-"#label"("#err")\n"	\
+	".previous\n"
 
 extern void __get_user_unknown(void);
 
@@ -100,23 +98,23 @@ extern void __get_user_unknown(void);
 	__gu_err;						\
 })
 
-#define __get_user_check(x, ptr, size, segment)				\
-({									\
-	long __gu_err = -EFAULT;					\
-	unsigned long __gu_val = 0;					\
-	const __typeof__(*(ptr)) __user *__gu_addr = (ptr);		\
-	if (__access_ok((unsigned long)__gu_addr, size, segment)) {	\
-		__gu_err = 0;						\
-		switch (size) {						\
-		  case 1: __get_user_8(__gu_addr); break;		\
-		  case 2: __get_user_16(__gu_addr); break;		\
-		  case 4: __get_user_32(__gu_addr); break;		\
-		  case 8: __get_user_64(__gu_addr); break;		\
-		  default: __get_user_unknown(); break;			\
-		}							\
-	}								\
-	(x) = (__force __typeof__(*(ptr))) __gu_val;			\
-	__gu_err;							\
+#define __get_user_check(x, ptr, size)				\
+({								\
+	long __gu_err = -EFAULT;				\
+	unsigned long __gu_val = 0;				\
+	const __typeof__(*(ptr)) __user *__gu_addr = (ptr);	\
+	if (__access_ok((unsigned long)__gu_addr, size)) {	\
+		__gu_err = 0;					\
+		switch (size) {					\
+		  case 1: __get_user_8(__gu_addr); break;	\
+		  case 2: __get_user_16(__gu_addr); break;	\
+		  case 4: __get_user_32(__gu_addr); break;	\
+		  case 8: __get_user_64(__gu_addr); break;	\
+		  default: __get_user_unknown(); break;		\
+		}						\
+	}							\
+	(x) = (__force __typeof__(*(ptr))) __gu_val;		\
+	__gu_err;						\
 })
 
 struct __large_struct { unsigned long buf[100]; };
@@ -125,20 +123,14 @@ struct __large_struct { unsigned long buf[100]; };
 #define __get_user_64(addr)				\
 	__asm__("1: ldq %0,%2\n"			\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - .\n"				\
-	"	lda %0, 2b-1b(%1)\n"			\
-	".previous"					\
+	EXC(1b,2b,%0,%1)				\
 		: "=r"(__gu_val), "=r"(__gu_err)	\
 		: "m"(__m(addr)), "1"(__gu_err))
 
 #define __get_user_32(addr)				\
 	__asm__("1: ldl %0,%2\n"			\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - .\n"				\
-	"	lda %0, 2b-1b(%1)\n"			\
-	".previous"					\
+	EXC(1b,2b,%0,%1)				\
 		: "=r"(__gu_val), "=r"(__gu_err)	\
 		: "m"(__m(addr)), "1"(__gu_err))
 
@@ -148,20 +140,14 @@ struct __large_struct { unsigned long buf[100]; };
 #define __get_user_16(addr)				\
 	__asm__("1: ldwu %0,%2\n"			\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - .\n"				\
-	"	lda %0, 2b-1b(%1)\n"			\
-	".previous"					\
+	EXC(1b,2b,%0,%1)				\
 		: "=r"(__gu_val), "=r"(__gu_err)	\
 		: "m"(__m(addr)), "1"(__gu_err))
 
 #define __get_user_8(addr)				\
 	__asm__("1: ldbu %0,%2\n"			\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - .\n"				\
-	"	lda %0, 2b-1b(%1)\n"			\
-	".previous"					\
+	EXC(1b,2b,%0,%1)				\
 		: "=r"(__gu_val), "=r"(__gu_err)	\
 		: "m"(__m(addr)), "1"(__gu_err))
 #else
@@ -177,12 +163,8 @@ struct __large_struct { unsigned long buf[100]; };
 	"	extwh %1,%3,%1\n"					\
 	"	or %0,%1,%0\n"						\
 	"3:\n"								\
-	".section __ex_table,\"a\"\n"					\
-	"	.long 1b - .\n"						\
-	"	lda %0, 3b-1b(%2)\n"					\
-	"	.long 2b - .\n"						\
-	"	lda %0, 3b-2b(%2)\n"					\
-	".previous"							\
+	EXC(1b,3b,%0,%2)						\
+	EXC(2b,3b,%0,%2)						\
 		: "=&r"(__gu_val), "=&r"(__gu_tmp), "=r"(__gu_err)	\
 		: "r"(addr), "2"(__gu_err));				\
 }
@@ -191,10 +173,7 @@ struct __large_struct { unsigned long buf[100]; };
 	__asm__("1: ldq_u %0,0(%2)\n"					\
 	"	extbl %0,%2,%0\n"					\
 	"2:\n"								\
-	".section __ex_table,\"a\"\n"					\
-	"	.long 1b - .\n"						\
-	"	lda %0, 2b-1b(%1)\n"					\
-	".previous"							\
+	EXC(1b,2b,%0,%1)						\
 		: "=&r"(__gu_val), "=r"(__gu_err)			\
 		: "r"(addr), "1"(__gu_err))
 #endif
@@ -215,21 +194,21 @@ extern void __put_user_unknown(void);
 	__pu_err;						\
 })
 
-#define __put_user_check(x, ptr, size, segment)				\
-({									\
-	long __pu_err = -EFAULT;					\
-	__typeof__(*(ptr)) __user *__pu_addr = (ptr);			\
-	if (__access_ok((unsigned long)__pu_addr, size, segment)) {	\
-		__pu_err = 0;						\
-		switch (size) {						\
-		  case 1: __put_user_8(x, __pu_addr); break;		\
-		  case 2: __put_user_16(x, __pu_addr); break;		\
-		  case 4: __put_user_32(x, __pu_addr); break;		\
-		  case 8: __put_user_64(x, __pu_addr); break;		\
-		  default: __put_user_unknown(); break;			\
-		}							\
-	}								\
-	__pu_err;							\
+#define __put_user_check(x, ptr, size)				\
+({								\
+	long __pu_err = -EFAULT;				\
+	__typeof__(*(ptr)) __user *__pu_addr = (ptr);		\
+	if (__access_ok((unsigned long)__pu_addr, size)) {	\
+		__pu_err = 0;					\
+		switch (size) {					\
+		  case 1: __put_user_8(x, __pu_addr); break;	\
+		  case 2: __put_user_16(x, __pu_addr); break;	\
+		  case 4: __put_user_32(x, __pu_addr); break;	\
+		  case 8: __put_user_64(x, __pu_addr); break;	\
+		  default: __put_user_unknown(); break;		\
+		}						\
+	}							\
+	__pu_err;						\
 })
 
 /*
@@ -240,20 +219,14 @@ extern void __put_user_unknown(void);
 #define __put_user_64(x, addr)					\
 __asm__ __volatile__("1: stq %r2,%1\n"				\
 	"2:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31,2b-1b(%0)\n"				\
-	".previous"						\
+	EXC(1b,2b,$31,%0)					\
 		: "=r"(__pu_err)				\
 		: "m" (__m(addr)), "rJ" (x), "0"(__pu_err))
 
 #define __put_user_32(x, addr)					\
 __asm__ __volatile__("1: stl %r2,%1\n"				\
 	"2:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31,2b-1b(%0)\n"				\
-	".previous"						\
+	EXC(1b,2b,$31,%0)					\
 		: "=r"(__pu_err)				\
 		: "m"(__m(addr)), "rJ"(x), "0"(__pu_err))
 
@@ -263,20 +236,14 @@ __asm__ __volatile__("1: stl %r2,%1\n"				\
 #define __put_user_16(x, addr)					\
 __asm__ __volatile__("1: stw %r2,%1\n"				\
 	"2:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31,2b-1b(%0)\n"				\
-	".previous"						\
+	EXC(1b,2b,$31,%0)					\
 		: "=r"(__pu_err)				\
 		: "m"(__m(addr)), "rJ"(x), "0"(__pu_err))
 
 #define __put_user_8(x, addr)					\
 __asm__ __volatile__("1: stb %r2,%1\n"				\
 	"2:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31,2b-1b(%0)\n"				\
-	".previous"						\
+	EXC(1b,2b,$31,%0)					\
 		: "=r"(__pu_err)				\
 		: "m"(__m(addr)), "rJ"(x), "0"(__pu_err))
 #else
@@ -298,16 +265,10 @@ __asm__ __volatile__("1: stb %r2,%1\n"				\
 	"3:	stq_u %2,1(%5)\n"				\
 	"4:	stq_u %1,0(%5)\n"				\
 	"5:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31, 5b-1b(%0)\n"				\
-	"	.long 2b - .\n"					\
-	"	lda $31, 5b-2b(%0)\n"				\
-	"	.long 3b - .\n"					\
-	"	lda $31, 5b-3b(%0)\n"				\
-	"	.long 4b - .\n"					\
-	"	lda $31, 5b-4b(%0)\n"				\
-	".previous"						\
+	EXC(1b,5b,$31,%0)					\
+	EXC(2b,5b,$31,%0)					\
+	EXC(3b,5b,$31,%0)					\
+	EXC(4b,5b,$31,%0)					\
 		: "=r"(__pu_err), "=&r"(__pu_tmp1), 		\
 		  "=&r"(__pu_tmp2), "=&r"(__pu_tmp3), 		\
 		  "=&r"(__pu_tmp4)				\
@@ -324,12 +285,8 @@ __asm__ __volatile__("1: stb %r2,%1\n"				\
 	"	or %1,%2,%1\n"					\
 	"2:	stq_u %1,0(%4)\n"				\
 	"3:\n"							\
-	".section __ex_table,\"a\"\n"				\
-	"	.long 1b - .\n"					\
-	"	lda $31, 3b-1b(%0)\n"				\
-	"	.long 2b - .\n"					\
-	"	lda $31, 3b-2b(%0)\n"				\
-	".previous"						\
+	EXC(1b,3b,$31,%0)					\
+	EXC(2b,3b,$31,%0)					\
 		: "=r"(__pu_err), 				\
 	  	  "=&r"(__pu_tmp1), "=&r"(__pu_tmp2)		\
 		: "r"((unsigned long)(x)), "r"(addr), "0"(__pu_err)); \
@@ -341,153 +298,37 @@ __asm__ __volatile__("1: stb %r2,%1\n"				\
  * Complex access routines
  */
 
-/* This little bit of silliness is to get the GP loaded for a function
-   that ordinarily wouldn't.  Otherwise we could have it done by the macro
-   directly, which can be optimized the linker.  */
-#ifdef MODULE
-#define __module_address(sym)		"r"(sym),
-#define __module_call(ra, arg, sym)	"jsr $" #ra ",(%" #arg ")," #sym
-#else
-#define __module_address(sym)
-#define __module_call(ra, arg, sym)	"bsr $" #ra "," #sym " !samegp"
-#endif
-
-extern void __copy_user(void);
-
-extern inline long
-__copy_tofrom_user_nocheck(void *to, const void *from, long len)
-{
-	register void * __cu_to __asm__("$6") = to;
-	register const void * __cu_from __asm__("$7") = from;
-	register long __cu_len __asm__("$0") = len;
-
-	__asm__ __volatile__(
-		__module_call(28, 3, __copy_user)
-		: "=r" (__cu_len), "=r" (__cu_from), "=r" (__cu_to)
-		: __module_address(__copy_user)
-		  "0" (__cu_len), "1" (__cu_from), "2" (__cu_to)
-		: "$1", "$2", "$3", "$4", "$5", "$28", "memory");
-
-	return __cu_len;
-}
-
-#define __copy_to_user(to, from, n)					\
-({									\
-	__chk_user_ptr(to);						\
-	__copy_tofrom_user_nocheck((__force void *)(to), (from), (n));	\
-})
-#define __copy_from_user(to, from, n)					\
-({									\
-	__chk_user_ptr(from);						\
-	__copy_tofrom_user_nocheck((to), (__force void *)(from), (n));	\
-})
-
-#define __copy_to_user_inatomic __copy_to_user
-#define __copy_from_user_inatomic __copy_from_user
+extern long __copy_user(void *to, const void *from, long len);
 
-extern inline long
-copy_to_user(void __user *to, const void *from, long n)
+static inline unsigned long
+raw_copy_from_user(void *to, const void __user *from, unsigned long len)
 {
-	if (likely(__access_ok((unsigned long)to, n, get_fs())))
-		n = __copy_tofrom_user_nocheck((__force void *)to, from, n);
-	return n;
+	return __copy_user(to, (__force const void *)from, len);
 }
 
-extern inline long
-copy_from_user(void *to, const void __user *from, long n)
+static inline unsigned long
+raw_copy_to_user(void __user *to, const void *from, unsigned long len)
 {
-	long res = n;
-	if (likely(__access_ok((unsigned long)from, n, get_fs())))
-		res = __copy_from_user_inatomic(to, from, n);
-	if (unlikely(res))
-		memset(to + (n - res), 0, res);
-	return res;
+	return __copy_user((__force void *)to, from, len);
 }
 
-extern void __do_clear_user(void);
-
-extern inline long
-__clear_user(void __user *to, long len)
-{
-	register void __user * __cl_to __asm__("$6") = to;
-	register long __cl_len __asm__("$0") = len;
-	__asm__ __volatile__(
-		__module_call(28, 2, __do_clear_user)
-		: "=r"(__cl_len), "=r"(__cl_to)
-		: __module_address(__do_clear_user)
-		  "0"(__cl_len), "1"(__cl_to)
-		: "$1", "$2", "$3", "$4", "$5", "$28", "memory");
-	return __cl_len;
-}
+extern long __clear_user(void __user *to, long len);
 
 extern inline long
 clear_user(void __user *to, long len)
 {
-	if (__access_ok((unsigned long)to, len, get_fs()))
+	if (__access_ok((unsigned long)to, len))
 		len = __clear_user(to, len);
 	return len;
 }
 
-#undef __module_address
-#undef __module_call
-
 #define user_addr_max() \
-        (segment_eq(get_fs(), USER_DS) ? TASK_SIZE : ~0UL)
+        (uaccess_kernel() ? ~0UL : TASK_SIZE)
 
 extern long strncpy_from_user(char *dest, const char __user *src, long count);
 extern __must_check long strlen_user(const char __user *str);
 extern __must_check long strnlen_user(const char __user *str, long n);
 
-/*
- * About the exception table:
- *
- * - insn is a 32-bit pc-relative offset from the faulting insn.
- * - nextinsn is a 16-bit offset off of the faulting instruction
- *   (not off of the *next* instruction as branches are).
- * - errreg is the register in which to place -EFAULT.
- * - valreg is the final target register for the load sequence
- *   and will be zeroed.
- *
- * Either errreg or valreg may be $31, in which case nothing happens.
- *
- * The exception fixup information "just so happens" to be arranged
- * as in a MEM format instruction.  This lets us emit our three
- * values like so:
- *
- *      lda valreg, nextinsn(errreg)
- *
- */
-
-struct exception_table_entry
-{
-	signed int insn;
-	union exception_fixup {
-		unsigned unit;
-		struct {
-			signed int nextinsn : 16;
-			unsigned int errreg : 5;
-			unsigned int valreg : 5;
-		} bits;
-	} fixup;
-};
-
-/* Returns the new pc */
-#define fixup_exception(map_reg, _fixup, pc)			\
-({								\
-	if ((_fixup)->fixup.bits.valreg != 31)			\
-		map_reg((_fixup)->fixup.bits.valreg) = 0;	\
-	if ((_fixup)->fixup.bits.errreg != 31)			\
-		map_reg((_fixup)->fixup.bits.errreg) = -EFAULT;	\
-	(pc) + (_fixup)->fixup.bits.nextinsn;			\
-})
-
-#define ARCH_HAS_RELATIVE_EXTABLE
-
-#define swap_ex_entry_fixup(a, b, tmp, delta)			\
-	do {							\
-		(a)->fixup.unit = (b)->fixup.unit;		\
-		(b)->fixup.unit = (tmp).fixup.unit;		\
-	} while (0)
-
+#include <asm/extable.h>
 
 #endif /* __ALPHA_UACCESS_H */
diff --git a/arch/alpha/kernel/traps.c b/arch/alpha/kernel/traps.c
index b137390e87e7..65bb102d985b 100644
--- a/arch/alpha/kernel/traps.c
+++ b/arch/alpha/kernel/traps.c
@@ -482,12 +482,8 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"	extwl %1,%3,%1\n"
 		"	extwh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -502,12 +498,8 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"	extll %1,%3,%1\n"
 		"	extlh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -522,12 +514,8 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"	extql %1,%3,%1\n"
 		"	extqh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -551,16 +539,10 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"3:	stq_u %2,1(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(una_reg(reg)), "0"(0));
@@ -581,16 +563,10 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"3:	stq_u %2,3(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(una_reg(reg)), "0"(0));
@@ -611,16 +587,10 @@ do_entUna(void * va, unsigned long opcode, unsigned long reg,
 		"3:	stq_u %2,7(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n\t"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(una_reg(reg)), "0"(0));
@@ -802,7 +772,7 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 	/* Don't bother reading ds in the access check since we already
 	   know that this came from the user.  Also rely on the fact that
 	   the page at TASK_SIZE is unmapped and so can't be touched anyway. */
-	if (!__access_ok((unsigned long)va, 0, USER_DS))
+	if ((unsigned long)va >= TASK_SIZE)
 		goto give_sigsegv;
 
 	++unaligned[1].count;
@@ -835,12 +805,8 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"	extwl %1,%3,%1\n"
 		"	extwh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -855,12 +821,8 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"	extll %1,%3,%1\n"
 		"	extlh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -875,12 +837,8 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"	extql %1,%3,%1\n"
 		"	extqh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -895,12 +853,8 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"	extll %1,%3,%1\n"
 		"	extlh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -915,12 +869,8 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"	extql %1,%3,%1\n"
 		"	extqh %2,%3,%2\n"
 		"3:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %1,3b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %2,3b-2b(%0)\n"
-		".previous"
+		EXC(1b,3b,%1,%0)
+		EXC(2b,3b,%2,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2)
 			: "r"(va), "0"(0));
 		if (error)
@@ -944,16 +894,10 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"3:	stq_u %2,1(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(*reg_addr), "0"(0));
@@ -978,16 +922,10 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"3:	stq_u %2,3(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(*reg_addr), "0"(0));
@@ -1012,16 +950,10 @@ do_entUnaUser(void __user * va, unsigned long opcode,
 		"3:	stq_u %2,7(%5)\n"
 		"4:	stq_u %1,0(%5)\n"
 		"5:\n"
-		".section __ex_table,\"a\"\n\t"
-		"	.long 1b - .\n"
-		"	lda %2,5b-1b(%0)\n"
-		"	.long 2b - .\n"
-		"	lda %1,5b-2b(%0)\n"
-		"	.long 3b - .\n"
-		"	lda $31,5b-3b(%0)\n"
-		"	.long 4b - .\n"
-		"	lda $31,5b-4b(%0)\n"
-		".previous"
+		EXC(1b,5b,%2,%0)
+		EXC(2b,5b,%1,%0)
+		EXC(3b,5b,$31,%0)
+		EXC(4b,5b,$31,%0)
 			: "=r"(error), "=&r"(tmp1), "=&r"(tmp2),
 			  "=&r"(tmp3), "=&r"(tmp4)
 			: "r"(va), "r"(*reg_addr), "0"(0));
@@ -1047,7 +979,7 @@ give_sigsegv:
 	/* We need to replicate some of the logic in mm/fault.c,
 	   since we don't have access to the fault code in the
 	   exception handling return path.  */
-	if (!__access_ok((unsigned long)va, 0, USER_DS))
+	if ((unsigned long)va >= TASK_SIZE)
 		info.si_code = SEGV_ACCERR;
 	else {
 		struct mm_struct *mm = current->mm;
diff --git a/arch/alpha/lib/clear_user.S b/arch/alpha/lib/clear_user.S
index bf5b931866ba..006f469fef73 100644
--- a/arch/alpha/lib/clear_user.S
+++ b/arch/alpha/lib/clear_user.S
@@ -8,21 +8,6 @@
  * right "bytes left to zero" value (and that it is updated only _after_
  * a successful copy).  There is also some rather minor exception setup
  * stuff.
- *
- * NOTE! This is not directly C-callable, because the calling semantics
- * are different:
- *
- * Inputs:
- *	length in $0
- *	destination address in $6
- *	exception pointer in $7
- *	return address in $28 (exceptions expect it there)
- *
- * Outputs:
- *	bytes left to copy in $0
- *
- * Clobbers:
- *	$1,$2,$3,$4,$5,$6
  */
 #include <asm/export.h>
 
@@ -38,62 +23,63 @@
 	.set noreorder
 	.align 4
 
-	.globl __do_clear_user
-	.ent __do_clear_user
-	.frame	$30, 0, $28
+	.globl __clear_user
+	.ent __clear_user
+	.frame	$30, 0, $26
 	.prologue 0
 
 $loop:
 	and	$1, 3, $4	# e0    :
 	beq	$4, 1f		# .. e1 :
 
-0:	EX( stq_u $31, 0($6) )	# e0    : zero one word
+0:	EX( stq_u $31, 0($16) )	# e0    : zero one word
 	subq	$0, 8, $0	# .. e1 :
 	subq	$4, 1, $4	# e0    :
-	addq	$6, 8, $6	# .. e1 :
+	addq	$16, 8, $16	# .. e1 :
 	bne	$4, 0b		# e1    :
 	unop			#       :
 
 1:	bic	$1, 3, $1	# e0    :
 	beq	$1, $tail	# .. e1 :
 
-2:	EX( stq_u $31, 0($6) )	# e0    : zero four words
+2:	EX( stq_u $31, 0($16) )	# e0    : zero four words
 	subq	$0, 8, $0	# .. e1 :
-	EX( stq_u $31, 8($6) )	# e0    :
+	EX( stq_u $31, 8($16) )	# e0    :
 	subq	$0, 8, $0	# .. e1 :
-	EX( stq_u $31, 16($6) )	# e0    :
+	EX( stq_u $31, 16($16) )	# e0    :
 	subq	$0, 8, $0	# .. e1 :
-	EX( stq_u $31, 24($6) )	# e0    :
+	EX( stq_u $31, 24($16) )	# e0    :
 	subq	$0, 8, $0	# .. e1 :
 	subq	$1, 4, $1	# e0    :
-	addq	$6, 32, $6	# .. e1 :
+	addq	$16, 32, $16	# .. e1 :
 	bne	$1, 2b		# e1    :
 
 $tail:
 	bne	$2, 1f		# e1    : is there a tail to do?
-	ret	$31, ($28), 1	# .. e1 :
+	ret	$31, ($26), 1	# .. e1 :
 
-1:	EX( ldq_u $5, 0($6) )	# e0    :
+1:	EX( ldq_u $5, 0($16) )	# e0    :
 	clr	$0		# .. e1 :
 	nop			# e1    :
 	mskqh	$5, $0, $5	# e0    :
-	EX( stq_u $5, 0($6) )	# e0    :
-	ret	$31, ($28), 1	# .. e1 :
+	EX( stq_u $5, 0($16) )	# e0    :
+	ret	$31, ($26), 1	# .. e1 :
 
-__do_clear_user:
-	and	$6, 7, $4	# e0    : find dest misalignment
+__clear_user:
+	and	$17, $17, $0
+	and	$16, 7, $4	# e0    : find dest misalignment
 	beq	$0, $zerolength # .. e1 :
 	addq	$0, $4, $1	# e0    : bias counter
 	and	$1, 7, $2	# e1    : number of bytes in tail
 	srl	$1, 3, $1	# e0    :
 	beq	$4, $loop	# .. e1 :
 
-	EX( ldq_u $5, 0($6) )	# e0    : load dst word to mask back in
+	EX( ldq_u $5, 0($16) )	# e0    : load dst word to mask back in
 	beq	$1, $oneword	# .. e1 : sub-word store?
 
-	mskql	$5, $6, $5	# e0    : take care of misaligned head
-	addq	$6, 8, $6	# .. e1 :
-	EX( stq_u $5, -8($6) )	# e0    :
+	mskql	$5, $16, $5	# e0    : take care of misaligned head
+	addq	$16, 8, $16	# .. e1 :
+	EX( stq_u $5, -8($16) )	# e0    :
 	addq	$0, $4, $0	# .. e1 : bytes left -= 8 - misalignment
 	subq	$1, 1, $1	# e0    :
 	subq	$0, 8, $0	# .. e1 :
@@ -101,15 +87,15 @@ __do_clear_user:
 	unop			#       :
 
 $oneword:
-	mskql	$5, $6, $4	# e0    :
+	mskql	$5, $16, $4	# e0    :
 	mskqh	$5, $2, $5	# e0    :
 	or	$5, $4, $5	# e1    :
-	EX( stq_u $5, 0($6) )	# e0    :
+	EX( stq_u $5, 0($16) )	# e0    :
 	clr	$0		# .. e1 :
 
 $zerolength:
 $exception:
-	ret	$31, ($28), 1	# .. e1 :
+	ret	$31, ($26), 1	# .. e1 :
 
-	.end __do_clear_user
-	EXPORT_SYMBOL(__do_clear_user)
+	.end __clear_user
+	EXPORT_SYMBOL(__clear_user)
diff --git a/arch/alpha/lib/copy_user.S b/arch/alpha/lib/copy_user.S
index 509f62b65311..159f1b7e6e49 100644
--- a/arch/alpha/lib/copy_user.S
+++ b/arch/alpha/lib/copy_user.S
@@ -9,21 +9,6 @@
  * contains the right "bytes left to copy" value (and that it is updated
  * only _after_ a successful copy). There is also some rather minor
  * exception setup stuff..
- *
- * NOTE! This is not directly C-callable, because the calling semantics are
- * different:
- *
- * Inputs:
- *	length in $0
- *	destination address in $6
- *	source address in $7
- *	return address in $28
- *
- * Outputs:
- *	bytes left to copy in $0
- *
- * Clobbers:
- *	$1,$2,$3,$4,$5,$6,$7
  */
 
 #include <asm/export.h>
@@ -49,58 +34,59 @@
 	.ent __copy_user
 __copy_user:
 	.prologue 0
-	and $6,7,$3
+	and $18,$18,$0
+	and $16,7,$3
 	beq $0,$35
 	beq $3,$36
 	subq $3,8,$3
 	.align 4
 $37:
-	EXI( ldq_u $1,0($7) )
-	EXO( ldq_u $2,0($6) )
-	extbl $1,$7,$1
-	mskbl $2,$6,$2
-	insbl $1,$6,$1
+	EXI( ldq_u $1,0($17) )
+	EXO( ldq_u $2,0($16) )
+	extbl $1,$17,$1
+	mskbl $2,$16,$2
+	insbl $1,$16,$1
 	addq $3,1,$3
 	bis $1,$2,$1
-	EXO( stq_u $1,0($6) )
+	EXO( stq_u $1,0($16) )
 	subq $0,1,$0
-	addq $6,1,$6
-	addq $7,1,$7
+	addq $16,1,$16
+	addq $17,1,$17
 	beq $0,$41
 	bne $3,$37
 $36:
-	and $7,7,$1
+	and $17,7,$1
 	bic $0,7,$4
 	beq $1,$43
 	beq $4,$48
-	EXI( ldq_u $3,0($7) )
+	EXI( ldq_u $3,0($17) )
 	.align 4
 $50:
-	EXI( ldq_u $2,8($7) )
+	EXI( ldq_u $2,8($17) )
 	subq $4,8,$4
-	extql $3,$7,$3
-	extqh $2,$7,$1
+	extql $3,$17,$3
+	extqh $2,$17,$1
 	bis $3,$1,$1
-	EXO( stq $1,0($6) )
-	addq $7,8,$7
+	EXO( stq $1,0($16) )
+	addq $17,8,$17
 	subq $0,8,$0
-	addq $6,8,$6
+	addq $16,8,$16
 	bis $2,$2,$3
 	bne $4,$50
 $48:
 	beq $0,$41
 	.align 4
 $57:
-	EXI( ldq_u $1,0($7) )
-	EXO( ldq_u $2,0($6) )
-	extbl $1,$7,$1
-	mskbl $2,$6,$2
-	insbl $1,$6,$1
+	EXI( ldq_u $1,0($17) )
+	EXO( ldq_u $2,0($16) )
+	extbl $1,$17,$1
+	mskbl $2,$16,$2
+	insbl $1,$16,$1
 	bis $1,$2,$1
-	EXO( stq_u $1,0($6) )
+	EXO( stq_u $1,0($16) )
 	subq $0,1,$0
-	addq $6,1,$6
-	addq $7,1,$7
+	addq $16,1,$16
+	addq $17,1,$17
 	bne $0,$57
 	br $31,$41
 	.align 4
@@ -108,27 +94,27 @@ $43:
 	beq $4,$65
 	.align 4
 $66:
-	EXI( ldq $1,0($7) )
+	EXI( ldq $1,0($17) )
 	subq $4,8,$4
-	EXO( stq $1,0($6) )
-	addq $7,8,$7
+	EXO( stq $1,0($16) )
+	addq $17,8,$17
 	subq $0,8,$0
-	addq $6,8,$6
+	addq $16,8,$16
 	bne $4,$66
 $65:
 	beq $0,$41
-	EXI( ldq $2,0($7) )
-	EXO( ldq $1,0($6) )
+	EXI( ldq $2,0($17) )
+	EXO( ldq $1,0($16) )
 	mskql $2,$0,$2
 	mskqh $1,$0,$1
 	bis $2,$1,$2
-	EXO( stq $2,0($6) )
+	EXO( stq $2,0($16) )
 	bis $31,$31,$0
 $41:
 $35:
 $exitin:
 $exitout:
-	ret $31,($28),1
+	ret $31,($26),1
 
 	.end __copy_user
 EXPORT_SYMBOL(__copy_user)
diff --git a/arch/alpha/lib/csum_partial_copy.c b/arch/alpha/lib/csum_partial_copy.c
index 5dfb7975895f..ab42afba1720 100644
--- a/arch/alpha/lib/csum_partial_copy.c
+++ b/arch/alpha/lib/csum_partial_copy.c
@@ -45,10 +45,7 @@ __asm__ __volatile__("insqh %1,%2,%0":"=r" (z):"r" (x),"r" (y))
 	__asm__ __volatile__(				\
 	"1:	ldq_u %0,%2\n"				\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - .\n"				\
-	"	lda %0,2b-1b(%1)\n"			\
-	".previous"					\
+	EXC(1b,2b,%0,%1)				\
 		: "=r"(x), "=r"(__guu_err)		\
 		: "m"(__m(ptr)), "1"(0));		\
 	__guu_err;					\
@@ -60,10 +57,7 @@ __asm__ __volatile__("insqh %1,%2,%0":"=r" (z):"r" (x),"r" (y))
 	__asm__ __volatile__(				\
 	"1:	stq_u %2,%1\n"				\
 	"2:\n"						\
-	".section __ex_table,\"a\"\n"			\
-	"	.long 1b - ."				\
-	"	lda $31,2b-1b(%0)\n"			\
-	".previous"					\
+	EXC(1b,2b,$31,%0)				\
 		: "=r"(__puu_err)			\
 		: "m"(__m(addr)), "rJ"(x), "0"(0));	\
 	__puu_err;					\
diff --git a/arch/alpha/lib/ev6-clear_user.S b/arch/alpha/lib/ev6-clear_user.S
index 05bef6b50598..e179e4757ef8 100644
--- a/arch/alpha/lib/ev6-clear_user.S
+++ b/arch/alpha/lib/ev6-clear_user.S
@@ -9,21 +9,6 @@
  * a successful copy).  There is also some rather minor exception setup
  * stuff.
  *
- * NOTE! This is not directly C-callable, because the calling semantics
- * are different:
- *
- * Inputs:
- *	length in $0
- *	destination address in $6
- *	exception pointer in $7
- *	return address in $28 (exceptions expect it there)
- *
- * Outputs:
- *	bytes left to copy in $0
- *
- * Clobbers:
- *	$1,$2,$3,$4,$5,$6
- *
  * Much of the information about 21264 scheduling/coding comes from:
  *	Compiler Writer's Guide for the Alpha 21264
  *	abbreviated as 'CWG' in other comments here
@@ -56,14 +41,15 @@
 	.set noreorder
 	.align 4
 
-	.globl __do_clear_user
-	.ent __do_clear_user
-	.frame	$30, 0, $28
+	.globl __clear_user
+	.ent __clear_user
+	.frame	$30, 0, $26
 	.prologue 0
 
 				# Pipeline info : Slotting & Comments
-__do_clear_user:
-	and	$6, 7, $4	# .. E  .. ..	: find dest head misalignment
+__clear_user:
+	and	$17, $17, $0
+	and	$16, 7, $4	# .. E  .. ..	: find dest head misalignment
 	beq	$0, $zerolength # U  .. .. ..	:  U L U L
 
 	addq	$0, $4, $1	# .. .. .. E	: bias counter
@@ -75,14 +61,14 @@ __do_clear_user:
 
 /*
  * Head is not aligned.  Write (8 - $4) bytes to head of destination
- * This means $6 is known to be misaligned
+ * This means $16 is known to be misaligned
  */
-	EX( ldq_u $5, 0($6) )	# .. .. .. L	: load dst word to mask back in
+	EX( ldq_u $5, 0($16) )	# .. .. .. L	: load dst word to mask back in
 	beq	$1, $onebyte	# .. .. U  ..	: sub-word store?
-	mskql	$5, $6, $5	# .. U  .. ..	: take care of misaligned head
-	addq	$6, 8, $6	# E  .. .. .. 	: L U U L
+	mskql	$5, $16, $5	# .. U  .. ..	: take care of misaligned head
+	addq	$16, 8, $16	# E  .. .. .. 	: L U U L
 
-	EX( stq_u $5, -8($6) )	# .. .. .. L	:
+	EX( stq_u $5, -8($16) )	# .. .. .. L	:
 	subq	$1, 1, $1	# .. .. E  ..	:
 	addq	$0, $4, $0	# .. E  .. ..	: bytes left -= 8 - misalignment
 	subq	$0, 8, $0	# E  .. .. ..	: U L U L
@@ -93,11 +79,11 @@ __do_clear_user:
  * values upon initial entry to the loop
  * $1 is number of quadwords to clear (zero is a valid value)
  * $2 is number of trailing bytes (0..7) ($2 never used...)
- * $6 is known to be aligned 0mod8
+ * $16 is known to be aligned 0mod8
  */
 $headalign:
 	subq	$1, 16, $4	# .. .. .. E	: If < 16, we can not use the huge loop
-	and	$6, 0x3f, $2	# .. .. E  ..	: Forward work for huge loop
+	and	$16, 0x3f, $2	# .. .. E  ..	: Forward work for huge loop
 	subq	$2, 0x40, $3	# .. E  .. ..	: bias counter (huge loop)
 	blt	$4, $trailquad	# U  .. .. ..	: U L U L
 
@@ -114,21 +100,21 @@ $headalign:
 	beq	$3, $bigalign	# U  .. .. ..	: U L U L : Aligned 0mod64
 
 $alignmod64:
-	EX( stq_u $31, 0($6) )	# .. .. .. L
+	EX( stq_u $31, 0($16) )	# .. .. .. L
 	addq	$3, 8, $3	# .. .. E  ..
 	subq	$0, 8, $0	# .. E  .. ..
 	nop			# E  .. .. ..	: U L U L
 
 	nop			# .. .. .. E
 	subq	$1, 1, $1	# .. .. E  ..
-	addq	$6, 8, $6	# .. E  .. ..
+	addq	$16, 8, $16	# .. E  .. ..
 	blt	$3, $alignmod64	# U  .. .. ..	: U L U L
 
 $bigalign:
 /*
  * $0 is the number of bytes left
  * $1 is the number of quads left
- * $6 is aligned 0mod64
+ * $16 is aligned 0mod64
  * we know that we'll be taking a minimum of one trip through
  * CWG Section 3.7.6: do not expect a sustained store rate of > 1/cycle
  * We are _not_ going to update $0 after every single store.  That
@@ -145,39 +131,39 @@ $bigalign:
 	nop			# E :
 	nop			# E :
 	nop			# E :
-	bis	$6,$6,$3	# E : U L U L : Initial wh64 address is dest
+	bis	$16,$16,$3	# E : U L U L : Initial wh64 address is dest
 	/* This might actually help for the current trip... */
 
 $do_wh64:
 	wh64	($3)		# .. .. .. L1	: memory subsystem hint
 	subq	$1, 16, $4	# .. .. E  ..	: Forward calculation - repeat the loop?
-	EX( stq_u $31, 0($6) )	# .. L  .. ..
+	EX( stq_u $31, 0($16) )	# .. L  .. ..
 	subq	$0, 8, $0	# E  .. .. ..	: U L U L
 
-	addq	$6, 128, $3	# E : Target address of wh64
-	EX( stq_u $31, 8($6) )	# L :
-	EX( stq_u $31, 16($6) )	# L :
+	addq	$16, 128, $3	# E : Target address of wh64
+	EX( stq_u $31, 8($16) )	# L :
+	EX( stq_u $31, 16($16) )	# L :
 	subq	$0, 16, $0	# E : U L L U
 
 	nop			# E :
-	EX( stq_u $31, 24($6) )	# L :
-	EX( stq_u $31, 32($6) )	# L :
+	EX( stq_u $31, 24($16) )	# L :
+	EX( stq_u $31, 32($16) )	# L :
 	subq	$0, 168, $5	# E : U L L U : two trips through the loop left?
 	/* 168 = 192 - 24, since we've already completed some stores */
 
 	subq	$0, 16, $0	# E :
-	EX( stq_u $31, 40($6) )	# L :
-	EX( stq_u $31, 48($6) )	# L :
-	cmovlt	$5, $6, $3	# E : U L L U : Latency 2, extra mapping cycle
+	EX( stq_u $31, 40($16) )	# L :
+	EX( stq_u $31, 48($16) )	# L :
+	cmovlt	$5, $16, $3	# E : U L L U : Latency 2, extra mapping cycle
 
 	subq	$1, 8, $1	# E :
 	subq	$0, 16, $0	# E :
-	EX( stq_u $31, 56($6) )	# L :
+	EX( stq_u $31, 56($16) )	# L :
 	nop			# E : U L U L
 
 	nop			# E :
 	subq	$0, 8, $0	# E :
-	addq	$6, 64, $6	# E :
+	addq	$16, 64, $16	# E :
 	bge	$4, $do_wh64	# U : U L U L
 
 $trailquad:
@@ -190,14 +176,14 @@ $trailquad:
 	beq	$1, $trailbytes	# U  .. .. ..	: U L U L : Only 0..7 bytes to go
 
 $onequad:
-	EX( stq_u $31, 0($6) )	# .. .. .. L
+	EX( stq_u $31, 0($16) )	# .. .. .. L
 	subq	$1, 1, $1	# .. .. E  ..
 	subq	$0, 8, $0	# .. E  .. ..
 	nop			# E  .. .. ..	: U L U L
 
 	nop			# .. .. .. E
 	nop			# .. .. E  ..
-	addq	$6, 8, $6	# .. E  .. ..
+	addq	$16, 8, $16	# .. E  .. ..
 	bgt	$1, $onequad	# U  .. .. ..	: U L U L
 
 	# We have an unknown number of bytes left to go.
@@ -211,9 +197,9 @@ $trailbytes:
 	# so we will use $0 as the loop counter
 	# We know for a fact that $0 > 0 zero due to previous context
 $onebyte:
-	EX( stb $31, 0($6) )	# .. .. .. L
+	EX( stb $31, 0($16) )	# .. .. .. L
 	subq	$0, 1, $0	# .. .. E  ..	:
-	addq	$6, 1, $6	# .. E  .. ..	:
+	addq	$16, 1, $16	# .. E  .. ..	:
 	bgt	$0, $onebyte	# U  .. .. ..	: U L U L
 
 $zerolength:
@@ -221,6 +207,6 @@ $exception:			# Destination for exception recovery(?)
 	nop			# .. .. .. E	:
 	nop			# .. .. E  ..	:
 	nop			# .. E  .. ..	:
-	ret	$31, ($28), 1	# L0 .. .. ..	: L U L U
-	.end __do_clear_user
-	EXPORT_SYMBOL(__do_clear_user)
+	ret	$31, ($26), 1	# L0 .. .. ..	: L U L U
+	.end __clear_user
+	EXPORT_SYMBOL(__clear_user)
diff --git a/arch/alpha/lib/ev6-copy_user.S b/arch/alpha/lib/ev6-copy_user.S
index be720b518af9..35e6710d0700 100644
--- a/arch/alpha/lib/ev6-copy_user.S
+++ b/arch/alpha/lib/ev6-copy_user.S
@@ -12,21 +12,6 @@
  * only _after_ a successful copy). There is also some rather minor
  * exception setup stuff..
  *
- * NOTE! This is not directly C-callable, because the calling semantics are
- * different:
- *
- * Inputs:
- *	length in $0
- *	destination address in $6
- *	source address in $7
- *	return address in $28
- *
- * Outputs:
- *	bytes left to copy in $0
- *
- * Clobbers:
- *	$1,$2,$3,$4,$5,$6,$7
- *
  * Much of the information about 21264 scheduling/coding comes from:
  *	Compiler Writer's Guide for the Alpha 21264
  *	abbreviated as 'CWG' in other comments here
@@ -60,10 +45,11 @@
 				# Pipeline info: Slotting & Comments
 __copy_user:
 	.prologue 0
-	subq $0, 32, $1		# .. E  .. ..	: Is this going to be a small copy?
+	andq $18, $18, $0
+	subq $18, 32, $1	# .. E  .. ..	: Is this going to be a small copy?
 	beq $0, $zerolength	# U  .. .. ..	: U L U L
 
-	and $6,7,$3		# .. .. .. E	: is leading dest misalignment
+	and $16,7,$3		# .. .. .. E	: is leading dest misalignment
 	ble $1, $onebyteloop	# .. .. U  ..	: 1st branch : small amount of data
 	beq $3, $destaligned	# .. U  .. ..	: 2nd (one cycle fetcher stall)
 	subq $3, 8, $3		# E  .. .. ..	: L U U L : trip counter
@@ -73,17 +59,17 @@ __copy_user:
  * We know we have at least one trip through this loop
  */
 $aligndest:
-	EXI( ldbu $1,0($7) )	# .. .. .. L	: Keep loads separate from stores
-	addq $6,1,$6		# .. .. E  ..	: Section 3.8 in the CWG
+	EXI( ldbu $1,0($17) )	# .. .. .. L	: Keep loads separate from stores
+	addq $16,1,$16		# .. .. E  ..	: Section 3.8 in the CWG
 	addq $3,1,$3		# .. E  .. ..	:
 	nop			# E  .. .. ..	: U L U L
 
 /*
- * the -1 is to compensate for the inc($6) done in a previous quadpack
+ * the -1 is to compensate for the inc($16) done in a previous quadpack
  * which allows us zero dependencies within either quadpack in the loop
  */
-	EXO( stb $1,-1($6) )	# .. .. .. L	:
-	addq $7,1,$7		# .. .. E  ..	: Section 3.8 in the CWG
+	EXO( stb $1,-1($16) )	# .. .. .. L	:
+	addq $17,1,$17		# .. .. E  ..	: Section 3.8 in the CWG
 	subq $0,1,$0		# .. E  .. ..	:
 	bne $3, $aligndest	# U  .. .. ..	: U L U L
 
@@ -92,29 +78,29 @@ $aligndest:
  * If we arrived via branch, we have a minimum of 32 bytes
  */
 $destaligned:
-	and $7,7,$1		# .. .. .. E	: Check _current_ source alignment
+	and $17,7,$1		# .. .. .. E	: Check _current_ source alignment
 	bic $0,7,$4		# .. .. E  ..	: number bytes as a quadword loop
-	EXI( ldq_u $3,0($7) )	# .. L  .. ..	: Forward fetch for fallthrough code
+	EXI( ldq_u $3,0($17) )	# .. L  .. ..	: Forward fetch for fallthrough code
 	beq $1,$quadaligned	# U  .. .. ..	: U L U L
 
 /*
- * In the worst case, we've just executed an ldq_u here from 0($7)
+ * In the worst case, we've just executed an ldq_u here from 0($17)
  * and we'll repeat it once if we take the branch
  */
 
 /* Misaligned quadword loop - not unrolled.  Leave it that way. */
 $misquad:
-	EXI( ldq_u $2,8($7) )	# .. .. .. L	:
+	EXI( ldq_u $2,8($17) )	# .. .. .. L	:
 	subq $4,8,$4		# .. .. E  ..	:
-	extql $3,$7,$3		# .. U  .. ..	:
-	extqh $2,$7,$1		# U  .. .. ..	: U U L L
+	extql $3,$17,$3		# .. U  .. ..	:
+	extqh $2,$17,$1		# U  .. .. ..	: U U L L
 
 	bis $3,$1,$1		# .. .. .. E	:
-	EXO( stq $1,0($6) )	# .. .. L  ..	:
-	addq $7,8,$7		# .. E  .. ..	:
+	EXO( stq $1,0($16) )	# .. .. L  ..	:
+	addq $17,8,$17		# .. E  .. ..	:
 	subq $0,8,$0		# E  .. .. ..	: U L L U
 
-	addq $6,8,$6		# .. .. .. E	:
+	addq $16,8,$16		# .. .. .. E	:
 	bis $2,$2,$3		# .. .. E  ..	:
 	nop			# .. E  .. ..	:
 	bne $4,$misquad		# U  .. .. ..	: U L U L
@@ -125,8 +111,8 @@ $misquad:
 	beq $0,$zerolength	# U  .. .. ..	: U L U L
 
 /* We know we have at least one trip through the byte loop */
-	EXI ( ldbu $2,0($7) )	# .. .. .. L	: No loads in the same quad
-	addq $6,1,$6		# .. .. E  ..	: as the store (Section 3.8 in CWG)
+	EXI ( ldbu $2,0($17) )	# .. .. .. L	: No loads in the same quad
+	addq $16,1,$16		# .. .. E  ..	: as the store (Section 3.8 in CWG)
 	nop			# .. E  .. ..	:
 	br $31, $dirtyentry	# L0 .. .. ..	: L U U L
 /* Do the trailing byte loop load, then hop into the store part of the loop */
@@ -136,8 +122,8 @@ $misquad:
  * Based upon the usage context, it's worth the effort to unroll this loop
  * $0 - number of bytes to be moved
  * $4 - number of bytes to move as quadwords
- * $6 is current destination address
- * $7 is current source address
+ * $16 is current destination address
+ * $17 is current source address
  */
 $quadaligned:
 	subq	$4, 32, $2	# .. .. .. E	: do not unroll for small stuff
@@ -155,29 +141,29 @@ $quadaligned:
  * instruction memory hint instruction).
  */
 $unroll4:
-	EXI( ldq $1,0($7) )	# .. .. .. L
-	EXI( ldq $2,8($7) )	# .. .. L  ..
+	EXI( ldq $1,0($17) )	# .. .. .. L
+	EXI( ldq $2,8($17) )	# .. .. L  ..
 	subq	$4,32,$4	# .. E  .. ..
 	nop			# E  .. .. ..	: U U L L
 
-	addq	$7,16,$7	# .. .. .. E
-	EXO( stq $1,0($6) )	# .. .. L  ..
-	EXO( stq $2,8($6) )	# .. L  .. ..
+	addq	$17,16,$17	# .. .. .. E
+	EXO( stq $1,0($16) )	# .. .. L  ..
+	EXO( stq $2,8($16) )	# .. L  .. ..
 	subq	$0,16,$0	# E  .. .. ..	: U L L U
 
-	addq	$6,16,$6	# .. .. .. E
-	EXI( ldq $1,0($7) )	# .. .. L  ..
-	EXI( ldq $2,8($7) )	# .. L  .. ..
+	addq	$16,16,$16	# .. .. .. E
+	EXI( ldq $1,0($17) )	# .. .. L  ..
+	EXI( ldq $2,8($17) )	# .. L  .. ..
 	subq	$4, 32, $3	# E  .. .. ..	: U U L L : is there enough for another trip?
 
-	EXO( stq $1,0($6) )	# .. .. .. L
-	EXO( stq $2,8($6) )	# .. .. L  ..
+	EXO( stq $1,0($16) )	# .. .. .. L
+	EXO( stq $2,8($16) )	# .. .. L  ..
 	subq	$0,16,$0	# .. E  .. ..
-	addq	$7,16,$7	# E  .. .. ..	: U L L U
+	addq	$17,16,$17	# E  .. .. ..	: U L L U
 
 	nop			# .. .. .. E
 	nop			# .. .. E  ..
-	addq	$6,16,$6	# .. E  .. ..
+	addq	$16,16,$16	# .. E  .. ..
 	bgt	$3,$unroll4	# U  .. .. ..	: U L U L
 
 	nop
@@ -186,14 +172,14 @@ $unroll4:
 	beq	$4, $noquads
 
 $onequad:
-	EXI( ldq $1,0($7) )
+	EXI( ldq $1,0($17) )
 	subq	$4,8,$4
-	addq	$7,8,$7
+	addq	$17,8,$17
 	nop
 
-	EXO( stq $1,0($6) )
+	EXO( stq $1,0($16) )
 	subq	$0,8,$0
-	addq	$6,8,$6
+	addq	$16,8,$16
 	bne	$4,$onequad
 
 $noquads:
@@ -207,23 +193,23 @@ $noquads:
  * There's no point in doing a lot of complex alignment calculations to try to
  * to quadword stuff for a small amount of data.
  *	$0 - remaining number of bytes left to copy
- *	$6 - current dest addr
- *	$7 - current source addr
+ *	$16 - current dest addr
+ *	$17 - current source addr
  */
 
 $onebyteloop:
-	EXI ( ldbu $2,0($7) )	# .. .. .. L	: No loads in the same quad
-	addq $6,1,$6		# .. .. E  ..	: as the store (Section 3.8 in CWG)
+	EXI ( ldbu $2,0($17) )	# .. .. .. L	: No loads in the same quad
+	addq $16,1,$16		# .. .. E  ..	: as the store (Section 3.8 in CWG)
 	nop			# .. E  .. ..	:
 	nop			# E  .. .. ..	: U L U L
 
 $dirtyentry:
 /*
- * the -1 is to compensate for the inc($6) done in a previous quadpack
+ * the -1 is to compensate for the inc($16) done in a previous quadpack
  * which allows us zero dependencies within either quadpack in the loop
  */
-	EXO ( stb $2,-1($6) )	# .. .. .. L	:
-	addq $7,1,$7		# .. .. E  ..	: quadpack as the load
+	EXO ( stb $2,-1($16) )	# .. .. .. L	:
+	addq $17,1,$17		# .. .. E  ..	: quadpack as the load
 	subq $0,1,$0		# .. E  .. ..	: change count _after_ copy
 	bgt $0,$onebyteloop	# U  .. .. ..	: U L U L
 
@@ -233,7 +219,7 @@ $exitout:			# Destination for exception recovery(?)
 	nop			# .. .. .. E
 	nop			# .. .. E  ..
 	nop			# .. E  .. ..
-	ret $31,($28),1		# L0 .. .. ..	: L U L U
+	ret $31,($26),1		# L0 .. .. ..	: L U L U
 
 	.end __copy_user
 	EXPORT_SYMBOL(__copy_user)