alpha: Use new generic strncpy_from_user() and strnlen_user()

Similar to x86/sparc/powerpc implementations except:
1) we implement an extremely efficient has_zero()/find_zero()
   sequence with both prep_zero_mask() and create_zero_mask()
   no-operations.
2) Our output from prep_zero_mask() differs in that only the
   lowest eight bits are used to represent the zero bytes
   nevertheless it can be safely ORed with other similar masks
   from prep_zero_mask() and forms input to create_zero_mask(),
   the two fundamental properties prep_zero_mask() must satisfy.

Tests on EV67 and EV68 CPUs revealed that the generic code is
essentially as fast (to within 0.5% of CPU cycles) of the old
Alpha specific code for large quadword-aligned strings, despite
the 30% extra CPU instructions executed.  In contrast, the
generic code for unaligned strings is substantially slower (by
more than a factor of 3) than the old Alpha specific code.
Signed-off-by: Michael Cree <mcree@orcon.net.nz>
Acked-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arch/alpha/Kconfig

@@ -18,6 +18,8 @@ config ALPHA
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG
 	select GENERIC_SMP_IDLE_THREAD
 	select GENERIC_CMOS_UPDATE
+	select GENERIC_STRNCPY_FROM_USER
+	select GENERIC_STRNLEN_USER
 	help
 	  The Alpha is a 64-bit general-purpose processor designed and
 	  marketed by the Digital Equipment Corporation of blessed memory,

arch/alpha/include/asm/uaccess.h

@@ -433,36 +433,12 @@ clear_user(void __user *to, long len)
 #undef __module_address
 #undef __module_call
 
-/* Returns: -EFAULT if exception before terminator, N if the entire
-   buffer filled, else strlen.  */
+#define user_addr_max() \
+        (segment_eq(get_fs(), USER_DS) ? TASK_SIZE : ~0UL)
 
-extern long __strncpy_from_user(char *__to, const char __user *__from, long __to_len);
-
-extern inline long
-strncpy_from_user(char *to, const char __user *from, long n)
-{
-	long ret = -EFAULT;
-	if (__access_ok((unsigned long)from, 0, get_fs()))
-		ret = __strncpy_from_user(to, from, n);
-	return ret;
-}
-
-/* Returns: 0 if bad, string length+1 (memory size) of string if ok */
-extern long __strlen_user(const char __user *);
-
-extern inline long strlen_user(const char __user *str)
-{
-	return access_ok(VERIFY_READ,str,0) ? __strlen_user(str) : 0;
-}
-
-/* Returns: 0 if exception before NUL or reaching the supplied limit (N),
- * a value greater than N if the limit would be exceeded, else strlen.  */
-extern long __strnlen_user(const char __user *, long);
-
-extern inline long strnlen_user(const char __user *str, long n)
-{
-	return access_ok(VERIFY_READ,str,0) ? __strnlen_user(str, n) : 0;
-}
+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:

arch/alpha/include/asm/word-at-a-time.h

+#ifndef _ASM_WORD_AT_A_TIME_H
+#define _ASM_WORD_AT_A_TIME_H
+
+#include <asm/compiler.h>
+
+/*
+ * word-at-a-time interface for Alpha.
+ */
+
+/*
+ * We do not use the word_at_a_time struct on Alpha, but it needs to be
+ * implemented to humour the generic code.
+ */
+struct word_at_a_time {
+	const unsigned long unused;
+};
+
+#define WORD_AT_A_TIME_CONSTANTS { 0 }
+
+/* Return nonzero if val has a zero */
+static inline unsigned long has_zero(unsigned long val, unsigned long *bits, const struct word_at_a_time *c)
+{
+	unsigned long zero_locations = __kernel_cmpbge(0, val);
+	*bits = zero_locations;
+	return zero_locations;
+}
+
+static inline unsigned long prep_zero_mask(unsigned long val, unsigned long bits, const struct word_at_a_time *c)
+{
+	return bits;
+}
+
+#define create_zero_mask(bits) (bits)
+
+static inline unsigned long find_zero(unsigned long bits)
+{
+#if defined(CONFIG_ALPHA_EV6) && defined(CONFIG_ALPHA_EV67)
+	/* Simple if have CIX instructions */
+	return __kernel_cttz(bits);
+#else
+	unsigned long t1, t2, t3;
+	/* Retain lowest set bit only */
+	bits &= -bits;
+	/* Binary search for lowest set bit */
+	t1 = bits & 0xf0;
+	t2 = bits & 0xcc;
+	t3 = bits & 0xaa;
+	if (t1) t1 = 4;
+	if (t2) t2 = 2;
+	if (t3) t3 = 1;
+	return t1 + t2 + t3;
+#endif
+}
+
+#endif /* _ASM_WORD_AT_A_TIME_H */

arch/alpha/kernel/alpha_ksyms.c

@@ -74,8 +74,6 @@ EXPORT_SYMBOL(alpha_fp_emul);
  */
 EXPORT_SYMBOL(__copy_user);
 EXPORT_SYMBOL(__do_clear_user);
-EXPORT_SYMBOL(__strncpy_from_user);
-EXPORT_SYMBOL(__strnlen_user);
 
 /* 
  * SMP-specific symbols.

arch/alpha/lib/Makefile

@@ -31,8 +31,6 @@ lib-y =	__divqu.o __remqu.o __divlu.o __remlu.o \
 	$(ev6-y)memchr.o \
 	$(ev6-y)copy_user.o \
 	$(ev6-y)clear_user.o \
-	$(ev6-y)strncpy_from_user.o \
-	$(ev67-y)strlen_user.o \
 	$(ev6-y)csum_ipv6_magic.o \
 	$(ev6-y)clear_page.o \
 	$(ev6-y)copy_page.o \

arch/alpha/lib/ev67-strlen_user.S

-/*
- * arch/alpha/lib/ev67-strlen_user.S
- * 21264 version contributed by Rick Gorton <rick.gorton@api-networks.com>
- *
- * Return the length of the string including the NULL terminator
- * (strlen+1) or zero if an error occurred.
- *
- * In places where it is critical to limit the processing time,
- * and the data is not trusted, strnlen_user() should be used.
- * It will return a value greater than its second argument if
- * that limit would be exceeded. This implementation is allowed
- * to access memory beyond the limit, but will not cross a page
- * boundary when doing so.
- *
- * 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
- *      ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html
- * Scheduling notation:
- *      E       - either cluster
- *      U       - upper subcluster; U0 - subcluster U0; U1 - subcluster U1
- *      L       - lower subcluster; L0 - subcluster L0; L1 - subcluster L1
- * Try not to change the actual algorithm if possible for consistency.
- */
-
-#include <asm/regdef.h>
-
-
-/* Allow an exception for an insn; exit if we get one.  */
-#define EX(x,y...)			\
-	99: x,##y;			\
-	.section __ex_table,"a";	\
-	.long 99b - .;			\
-	lda v0, $exception-99b(zero);	\
-	.previous
-
-
-	.set noreorder
-	.set noat
-	.text
-
-	.globl __strlen_user
-	.ent __strlen_user
-	.frame sp, 0, ra
-
-	.align 4
-__strlen_user:
-	ldah	a1, 32767(zero)	# do not use plain strlen_user() for strings
-				# that might be almost 2 GB long; you should
-				# be using strnlen_user() instead
-	nop
-	nop
-	nop
-
-	.globl __strnlen_user
-
-	.align 4
-__strnlen_user:
-	.prologue 0
-	EX( ldq_u t0, 0(a0) )	# L : load first quadword (a0 may be misaligned)
-	lda     t1, -1(zero)	# E :
-
-	insqh   t1, a0, t1	# U :
-	andnot  a0, 7, v0	# E :
-	or      t1, t0, t0	# E :
-	subq	a0, 1, a0	# E : get our +1 for the return 
-
-	cmpbge  zero, t0, t1	# E : t1 <- bitmask: bit i == 1 <==> i-th byte == 0
-	subq	a1, 7, t2	# E :
-	subq	a0, v0, t0	# E :
-	bne     t1, $found	# U :
-
-	addq	t2, t0, t2	# E :
-	addq	a1, 1, a1	# E :
-	nop			# E :
-	nop			# E :
-
-	.align 4
-$loop:	ble	t2, $limit	# U :
-	EX( ldq t0, 8(v0) )	# L :
-	nop			# E :
-	nop			# E :
-
-	cmpbge  zero, t0, t1	# E :
-	subq	t2, 8, t2	# E :
-	addq    v0, 8, v0	# E : addr += 8
-	beq     t1, $loop	# U :
-
-$found: cttz	t1, t2		# U0 :
-	addq	v0, t2, v0	# E :
-	subq    v0, a0, v0	# E :
-	ret			# L0 :
-
-$exception:
-	nop
-	nop
-	nop
-	ret
-
-	.align 4		# currently redundant
-$limit:
-	nop
-	nop
-	subq	a1, t2, v0
-	ret
-
-	.end __strlen_user

arch/alpha/lib/strlen_user.S

-/*
- * arch/alpha/lib/strlen_user.S
- *
- * Return the length of the string including the NUL terminator
- * (strlen+1) or zero if an error occurred.
- *
- * In places where it is critical to limit the processing time,
- * and the data is not trusted, strnlen_user() should be used.
- * It will return a value greater than its second argument if
- * that limit would be exceeded. This implementation is allowed
- * to access memory beyond the limit, but will not cross a page
- * boundary when doing so.
- */
-
-#include <asm/regdef.h>
-
-
-/* Allow an exception for an insn; exit if we get one.  */
-#define EX(x,y...)			\
-	99: x,##y;			\
-	.section __ex_table,"a";	\
-	.long 99b - .;			\
-	lda v0, $exception-99b(zero);	\
-	.previous
-
-
-	.set noreorder
-	.set noat
-	.text
-
-	.globl __strlen_user
-	.ent __strlen_user
-	.frame sp, 0, ra
-
-	.align 3
-__strlen_user:
-	ldah	a1, 32767(zero)	# do not use plain strlen_user() for strings
-				# that might be almost 2 GB long; you should
-				# be using strnlen_user() instead
-
-	.globl __strnlen_user
-
-	.align 3
-__strnlen_user:
-	.prologue 0
-
-	EX( ldq_u t0, 0(a0) )	# load first quadword (a0 may be misaligned)
-	lda     t1, -1(zero)
-	insqh   t1, a0, t1
-	andnot  a0, 7, v0
-	or      t1, t0, t0
-	subq	a0, 1, a0	# get our +1 for the return 
-	cmpbge  zero, t0, t1	# t1 <- bitmask: bit i == 1 <==> i-th byte == 0
-	subq	a1, 7, t2
-	subq	a0, v0, t0
-	bne     t1, $found
-
-	addq	t2, t0, t2
-	addq	a1, 1, a1
-
-	.align 3
-$loop:	ble	t2, $limit
-	EX( ldq t0, 8(v0) )
-	subq	t2, 8, t2
-	addq    v0, 8, v0	# addr += 8
-	cmpbge  zero, t0, t1
-	beq     t1, $loop
-
-$found:	negq    t1, t2		# clear all but least set bit
-	and     t1, t2, t1
-
-	and     t1, 0xf0, t2	# binary search for that set bit
-	and	t1, 0xcc, t3
-	and	t1, 0xaa, t4
-	cmovne	t2, 4, t2
-	cmovne	t3, 2, t3
-	cmovne	t4, 1, t4
-	addq	t2, t3, t2
-	addq	v0, t4, v0
-	addq	v0, t2, v0
-	nop			# dual issue next two on ev4 and ev5
-	subq    v0, a0, v0
-$exception:
-	ret
-
-	.align 3		# currently redundant
-$limit:
-	subq	a1, t2, v0
-	ret
-
-	.end __strlen_user