Merge branch 'alpha' (alpha architecture patches)

Merge alpha architecture update from Michael Cree:
 "The Alpha Maintainer, Matt Turner, is currently unavailable, so I have
  collected up patches that have been posted to the linux-alpha mailing
  list over the last couple of months, and are forwarding them to you in
  the hope that you are prepared to accept them via me.

  The patches by Al Viro and myself I have been running against kernels
  for two months now so have had quite a bit of testing.  All except one
  patch were intended for the 3.5 kernel but because of Matt's
  unavailability never got forwarded to you."

* emailed patches from Michael Cree <mcree@orcon.net.nz>: (9 commits)
  alpha: Fix fall-out from disintegrating asm/system.h
  Redefine ATOMIC_INIT and ATOMIC64_INIT to drop the casts
  alpha: fix fpu.h usage in userspace
  alpha/mm/fault.c: Port OOM changes to do_page_fault
  alpha: take kernel_execve() out of entry.S
  alpha: take a bunch of syscalls into osf_sys.c
  alpha: Use new generic strncpy_from_user() and strnlen_user()
  alpha: Wire up cross memory attach syscalls
  alpha: Don't export SOCK_NONBLOCK to user space.

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/atomic.h

@@ -14,8 +14,8 @@
  */
 
 
-#define ATOMIC_INIT(i)		( (atomic_t) { (i) } )
-#define ATOMIC64_INIT(i)	( (atomic64_t) { (i) } )
+#define ATOMIC_INIT(i)		{ (i) }
+#define ATOMIC64_INIT(i)	{ (i) }
 
 #define atomic_read(v)		(*(volatile int *)&(v)->counter)
 #define atomic64_read(v)	(*(volatile long *)&(v)->counter)

arch/alpha/include/asm/fpu.h

 #ifndef __ASM_ALPHA_FPU_H
 #define __ASM_ALPHA_FPU_H
 
+#ifdef __KERNEL__
 #include <asm/special_insns.h>
+#endif
 
 /*
  * Alpha floating-point control register defines:

arch/alpha/include/asm/ptrace.h

@@ -76,7 +76,10 @@ struct switch_stack {
 #define task_pt_regs(task) \
   ((struct pt_regs *) (task_stack_page(task) + 2*PAGE_SIZE) - 1)
 
-#define force_successful_syscall_return() (task_pt_regs(current)->r0 = 0)
+#define current_pt_regs() \
+  ((struct pt_regs *) ((char *)current_thread_info() + 2*PAGE_SIZE) - 1)
+
+#define force_successful_syscall_return() (current_pt_regs()->r0 = 0)
 
 #endif
 

arch/alpha/include/asm/socket.h

@@ -76,9 +76,11 @@
 /* Instruct lower device to use last 4-bytes of skb data as FCS */
 #define SO_NOFCS		43
 
+#ifdef __KERNEL__
 /* O_NONBLOCK clashes with the bits used for socket types.  Therefore we
  * have to define SOCK_NONBLOCK to a different value here.
  */
 #define SOCK_NONBLOCK	0x40000000
+#endif /* __KERNEL__ */
 
 #endif /* _ASM_SOCKET_H */

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/unistd.h

@@ -465,10 +465,12 @@
 #define __NR_setns			501
 #define __NR_accept4			502
 #define __NR_sendmmsg			503
+#define __NR_process_vm_readv		504
+#define __NR_process_vm_writev		505
 
 #ifdef __KERNEL__
 
-#define NR_SYSCALLS			504
+#define NR_SYSCALLS			506
 
 #define __ARCH_WANT_OLD_READDIR
 #define __ARCH_WANT_STAT64

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

@@ -52,7 +52,6 @@ EXPORT_SYMBOL(alpha_write_fp_reg_s);
 
 /* entry.S */
 EXPORT_SYMBOL(kernel_thread);
-EXPORT_SYMBOL(kernel_execve);
 
 /* Networking helper routines. */
 EXPORT_SYMBOL(csum_tcpudp_magic);
@@ -74,8 +73,6 @@ 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/kernel/entry.S

@@ -663,58 +663,6 @@ kernel_thread:
 	br	ret_to_kernel
 .end kernel_thread
 
-/*
- * kernel_execve(path, argv, envp)
- */
-	.align	4
-	.globl	kernel_execve
-	.ent	kernel_execve
-kernel_execve:
-	/* We can be called from a module.  */
-	ldgp	$gp, 0($27)
-	lda	$sp, -(32+SIZEOF_PT_REGS+8)($sp)
-	.frame	$sp, 32+SIZEOF_PT_REGS+8, $26, 0
-	stq	$26, 0($sp)
-	stq	$16, 8($sp)
-	stq	$17, 16($sp)
-	stq	$18, 24($sp)
-	.prologue 1
-
-	lda	$16, 32($sp)
-	lda	$17, 0
-	lda	$18, SIZEOF_PT_REGS
-	bsr	$26, memset		!samegp
-
-	/* Avoid the HAE being gratuitously wrong, which would cause us
-	   to do the whole turn off interrupts thing and restore it.  */
-	ldq	$2, alpha_mv+HAE_CACHE
-	stq	$2, 152+32($sp)
-
-	ldq	$16, 8($sp)
-	ldq	$17, 16($sp)
-	ldq	$18, 24($sp)
-	lda	$19, 32($sp)
-	bsr	$26, do_execve		!samegp
-
-	ldq	$26, 0($sp)
-	bne	$0, 1f			/* error! */
-
-	/* Move the temporary pt_regs struct from its current location
-	   to the top of the kernel stack frame.  See copy_thread for
-	   details for a normal process.  */
-	lda	$16, 0x4000 - SIZEOF_PT_REGS($8)
-	lda	$17, 32($sp)
-	lda	$18, SIZEOF_PT_REGS
-	bsr	$26, memmove		!samegp
-
-	/* Take that over as our new stack frame and visit userland!  */
-	lda	$sp, 0x4000 - SIZEOF_PT_REGS($8)
-	br	$31, ret_from_sys_call
-
-1:	lda	$sp, 32+SIZEOF_PT_REGS+8($sp)
-	ret
-.end kernel_execve
-
 
 /*
  * Special system calls.  Most of these are special in that they either
@@ -796,115 +744,6 @@ sys_rt_sigreturn:
 	br	ret_from_sys_call
 .end sys_rt_sigreturn
 
-	.align	4
-	.globl	sys_sethae
-	.ent	sys_sethae
-sys_sethae:
-	.prologue 0
-	stq	$16, 152($sp)
-	ret
-.end sys_sethae
-
-	.align	4
-	.globl	osf_getpriority
-	.ent	osf_getpriority
-osf_getpriority:
-	lda	$sp, -16($sp)
-	stq	$26, 0($sp)
-	.prologue 0
-
-	jsr	$26, sys_getpriority
-
-	ldq	$26, 0($sp)
-	blt	$0, 1f
-
-	/* Return value is the unbiased priority, i.e. 20 - prio.
-	   This does result in negative return values, so signal
-	   no error by writing into the R0 slot.  */
-	lda	$1, 20
-	stq	$31, 16($sp)
-	subl	$1, $0, $0
-	unop
-
-1:	lda	$sp, 16($sp)
-	ret
-.end osf_getpriority
-
-	.align	4
-	.globl	sys_getxuid
-	.ent	sys_getxuid
-sys_getxuid:
-	.prologue 0
-	ldq	$2, TI_TASK($8)
-	ldq	$3, TASK_CRED($2)
-	ldl	$0, CRED_UID($3)
-	ldl	$1, CRED_EUID($3)
-	stq	$1, 80($sp)
-	ret
-.end sys_getxuid
-
-	.align	4
-	.globl	sys_getxgid
-	.ent	sys_getxgid
-sys_getxgid:
-	.prologue 0
-	ldq	$2, TI_TASK($8)
-	ldq	$3, TASK_CRED($2)
-	ldl	$0, CRED_GID($3)
-	ldl	$1, CRED_EGID($3)
-	stq	$1, 80($sp)
-	ret
-.end sys_getxgid
-
-	.align	4
-	.globl	sys_getxpid
-	.ent	sys_getxpid
-sys_getxpid:
-	.prologue 0
-	ldq	$2, TI_TASK($8)
-
-	/* See linux/kernel/timer.c sys_getppid for discussion
-	   about this loop.  */
-	ldq	$3, TASK_GROUP_LEADER($2)
-	ldq	$4, TASK_REAL_PARENT($3)
-	ldl	$0, TASK_TGID($2)
-1:	ldl	$1, TASK_TGID($4)
-#ifdef CONFIG_SMP
-	mov	$4, $5
-	mb
-	ldq	$3, TASK_GROUP_LEADER($2)
-	ldq	$4, TASK_REAL_PARENT($3)
-	cmpeq	$4, $5, $5
-	beq	$5, 1b
-#endif
-	stq	$1, 80($sp)
-	ret
-.end sys_getxpid
-
-	.align	4
-	.globl	sys_alpha_pipe
-	.ent	sys_alpha_pipe
-sys_alpha_pipe:
-	lda	$sp, -16($sp)
-	stq	$26, 0($sp)
-	.prologue 0
-
-	mov	$31, $17
-	lda	$16, 8($sp)
-	jsr	$26, do_pipe_flags
-
-	ldq	$26, 0($sp)
-	bne	$0, 1f
-
-	/* The return values are in $0 and $20.  */
-	ldl	$1, 12($sp)
-	ldl	$0, 8($sp)
-
-	stq	$1, 80+16($sp)
-1:	lda	$sp, 16($sp)
-	ret
-.end sys_alpha_pipe
-
 	.align	4
 	.globl	sys_execve
 	.ent	sys_execve

arch/alpha/kernel/osf_sys.c

@@ -1404,3 +1404,52 @@ SYSCALL_DEFINE3(osf_writev, unsigned long, fd,
 }
 
 #endif
+
+SYSCALL_DEFINE2(osf_getpriority, int, which, int, who)
+{
+	int prio = sys_getpriority(which, who);
+	if (prio >= 0) {
+		/* Return value is the unbiased priority, i.e. 20 - prio.
+		   This does result in negative return values, so signal
+		   no error */
+		force_successful_syscall_return();
+		prio = 20 - prio;
+	}
+	return prio;
+}
+
+SYSCALL_DEFINE0(getxuid)
+{
+	current_pt_regs()->r20 = sys_geteuid();
+	return sys_getuid();
+}
+
+SYSCALL_DEFINE0(getxgid)
+{
+	current_pt_regs()->r20 = sys_getegid();
+	return sys_getgid();
+}
+
+SYSCALL_DEFINE0(getxpid)
+{
+	current_pt_regs()->r20 = sys_getppid();
+	return sys_getpid();
+}
+
+SYSCALL_DEFINE0(alpha_pipe)
+{
+	int fd[2];
+	int res = do_pipe_flags(fd, 0);
+	if (!res) {
+		/* The return values are in $0 and $20.  */
+		current_pt_regs()->r20 = fd[1];
+		res = fd[0];
+	}
+	return res;
+}
+
+SYSCALL_DEFINE1(sethae, unsigned long, val)
+{
+	current_pt_regs()->hae = val;
+	return 0;
+}

arch/alpha/kernel/process.c

@@ -455,3 +455,22 @@ get_wchan(struct task_struct *p)
 	}
 	return pc;
 }
+
+int kernel_execve(const char *path, const char *const argv[], const char *const envp[])
+{
+	/* Avoid the HAE being gratuitously wrong, which would cause us
+	   to do the whole turn off interrupts thing and restore it.  */
+	struct pt_regs regs = {.hae = alpha_mv.hae_cache};
+	int err = do_execve(path, argv, envp, &regs);
+	if (!err) {
+		struct pt_regs *p = current_pt_regs();
+		/* copy regs to normal position and off to userland we go... */
+		*p = regs;
+		__asm__ __volatile__ (
+			"mov	%0, $sp;"
+			"br	$31, ret_from_sys_call"
+			: : "r"(p));
+	}
+	return err;
+}
+EXPORT_SYMBOL(kernel_execve);

arch/alpha/kernel/systbls.S

@@ -111,7 +111,7 @@ sys_call_table:
 	.quad sys_socket
 	.quad sys_connect
 	.quad sys_accept
-	.quad osf_getpriority			/* 100 */
+	.quad sys_osf_getpriority			/* 100 */
 	.quad sys_send
 	.quad sys_recv
 	.quad sys_sigreturn
@@ -522,6 +522,8 @@ sys_call_table:
 	.quad sys_setns
 	.quad sys_accept4
 	.quad sys_sendmmsg
+	.quad sys_process_vm_readv
+	.quad sys_process_vm_writev		/* 505 */
 
 	.size sys_call_table, . - sys_call_table
 	.type sys_call_table, @object

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/ev6-strncpy_from_user.S

-/*
- * arch/alpha/lib/ev6-strncpy_from_user.S
- * 21264 version contributed by Rick Gorton <rick.gorton@alpha-processor.com>
- *
- * Just like strncpy except in the return value:
- *
- * -EFAULT       if an exception occurs before the terminator is copied.
- * N             if the buffer filled.
- *
- * Otherwise the length of the string is returned.
- *
- * 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
- * A bunch of instructions got moved and temp registers were changed
- * to aid in scheduling.  Control flow was also re-arranged to eliminate
- * branches, and to provide longer code sequences to enable better scheduling.
- * A total rewrite (using byte load/stores for start & tail sequences)
- * is desirable, but very difficult to do without a from-scratch rewrite.
- * Save that for the future.
- */
-
-
-#include <asm/errno.h>
-#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 $31, $exception-99b($0); 	\
-	.previous
-
-
-	.set noat
-	.set noreorder
-	.text
-
-	.globl __strncpy_from_user
-	.ent __strncpy_from_user
-	.frame $30, 0, $26
-	.prologue 0
-
-	.align 4
-__strncpy_from_user:
-	and	a0, 7, t3	# E : find dest misalignment
-	beq	a2, $zerolength	# U :
-
-	/* Are source and destination co-aligned?  */
-	mov	a0, v0		# E : save the string start
-	xor	a0, a1, t4	# E :
-	EX( ldq_u t1, 0(a1) )	# L : Latency=3 load first quadword
-	ldq_u	t0, 0(a0)	# L : load first (partial) aligned dest quadword
-
-	addq	a2, t3, a2	# E : bias count by dest misalignment
-	subq	a2, 1, a3	# E :
-	addq	zero, 1, t10	# E :
-	and	t4, 7, t4	# E : misalignment between the two
-
-	and	a3, 7, t6	# E : number of tail bytes
-	sll	t10, t6, t10	# E : t10 = bitmask of last count byte
-	bne	t4, $unaligned	# U :
-	lda	t2, -1		# E : build a mask against false zero
-
-	/*
-	 * We are co-aligned; take care of a partial first word.
-	 * On entry to this basic block:
-	 * t0 == the first destination word for masking back in
-	 * t1 == the first source word.
-	 */
-
-	srl	a3, 3, a2	# E : a2 = loop counter = (count - 1)/8
-	addq	a1, 8, a1	# E :
-	mskqh	t2, a1, t2	# U :   detection in the src word
-	nop
-
-	/* Create the 1st output word and detect 0's in the 1st input word.  */
-	mskqh	t1, a1, t3	# U :
-	mskql	t0, a1, t0	# U : assemble the first output word
-	ornot	t1, t2, t2	# E :
-	nop
-
-	cmpbge	zero, t2, t8	# E : bits set iff null found
-	or	t0, t3, t0	# E :
-	beq	a2, $a_eoc	# U :
-	bne	t8, $a_eos	# U : 2nd branch in a quad.  Bad.
-
-	/* On entry to this basic block:
-	 * t0 == a source quad not containing a null.
-	 * a0 - current aligned destination address
-	 * a1 - current aligned source address
-	 * a2 - count of quadwords to move.
-	 * NOTE: Loop improvement - unrolling this is going to be
-	 *	a huge win, since we're going to stall otherwise.
-	 *	Fix this later.  For _really_ large copies, look
-	 *	at using wh64 on a look-ahead basis.  See the code
-	 *	in clear_user.S and copy_user.S.
-	 * Presumably, since (a0) and (a1) do not overlap (by C definition)
-	 * Lots of nops here:
-	 *	- Separate loads from stores
-	 *	- Keep it to 1 branch/quadpack so the branch predictor
-	 *	  can train.
-	 */
-$a_loop:
-	stq_u	t0, 0(a0)	# L :
-	addq	a0, 8, a0	# E :
-	nop
-	subq	a2, 1, a2	# E :
-
-	EX( ldq_u t0, 0(a1) )	# L :
-	addq	a1, 8, a1	# E :
-	cmpbge	zero, t0, t8	# E : Stall 2 cycles on t0
-	beq	a2, $a_eoc      # U :
-
-	beq	t8, $a_loop	# U :
-	nop
-	nop
-	nop
-
-	/* Take care of the final (partial) word store.  At this point
-	 * the end-of-count bit is set in t8 iff it applies.
-	 *
-	 * On entry to this basic block we have:
-	 * t0 == the source word containing the null
-	 * t8 == the cmpbge mask that found it.
-	 */
-$a_eos:
-	negq	t8, t12		# E : find low bit set
-	and	t8, t12, t12	# E : 
-
-	/* We're doing a partial word store and so need to combine
-	   our source and original destination words.  */
-	ldq_u	t1, 0(a0)	# L :
-	subq	t12, 1, t6	# E :
-
-	or	t12, t6, t8	# E :
-	zapnot	t0, t8, t0	# U : clear src bytes > null
-	zap	t1, t8, t1	# U : clear dst bytes <= null
-	or	t0, t1, t0	# E :
-
-	stq_u	t0, 0(a0)	# L :
-	br	$finish_up	# L0 :
-	nop
-	nop
-
-	/* Add the end-of-count bit to the eos detection bitmask.  */
-	.align 4
-$a_eoc:
-	or	t10, t8, t8
-	br	$a_eos
-	nop
-	nop
-
-
-/* The source and destination are not co-aligned.  Align the destination
-   and cope.  We have to be very careful about not reading too much and
-   causing a SEGV.  */
-
-	.align 4
-$u_head:
-	/* We know just enough now to be able to assemble the first
-	   full source word.  We can still find a zero at the end of it
-	   that prevents us from outputting the whole thing.
-
-	   On entry to this basic block:
-	   t0 == the first dest word, unmasked
-	   t1 == the shifted low bits of the first source word
-	   t6 == bytemask that is -1 in dest word bytes */
-
-	EX( ldq_u t2, 8(a1) )	# L : load second src word
-	addq	a1, 8, a1	# E :
-	mskql	t0, a0, t0	# U : mask trailing garbage in dst
-	extqh	t2, a1, t4	# U :
-
-	or	t1, t4, t1	# E : first aligned src word complete
-	mskqh	t1, a0, t1	# U : mask leading garbage in src
-	or	t0, t1, t0	# E : first output word complete
-	or	t0, t6, t6	# E : mask original data for zero test
-
-	cmpbge	zero, t6, t8	# E :
-	beq	a2, $u_eocfin	# U :
-	bne	t8, $u_final	# U : bad news - 2nd branch in a quad
-	lda	t6, -1		# E : mask out the bits we have
-
-	mskql	t6, a1, t6	# U :   already seen
-	stq_u	t0, 0(a0)	# L : store first output word
-	or      t6, t2, t2	# E :
-	cmpbge	zero, t2, t8	# E : find nulls in second partial
-
-	addq	a0, 8, a0		# E :
-	subq	a2, 1, a2		# E :
-	bne	t8, $u_late_head_exit	# U :
-	nop
-
-	/* Finally, we've got all the stupid leading edge cases taken care
-	   of and we can set up to enter the main loop.  */
-
-	extql	t2, a1, t1	# U : position hi-bits of lo word
-	EX( ldq_u t2, 8(a1) )	# L : read next high-order source word
-	addq	a1, 8, a1	# E :
-	cmpbge	zero, t2, t8	# E :
-
-	beq	a2, $u_eoc	# U :
-	bne	t8, $u_eos	# U :
-	nop
-	nop
-
-	/* Unaligned copy main loop.  In order to avoid reading too much,
-	   the loop is structured to detect zeros in aligned source words.
-	   This has, unfortunately, effectively pulled half of a loop
-	   iteration out into the head and half into the tail, but it does
-	   prevent nastiness from accumulating in the very thing we want
-	   to run as fast as possible.
-
-	   On entry to this basic block:
-	   t1 == the shifted high-order bits from the previous source word
-	   t2 == the unshifted current source word
-
-	   We further know that t2 does not contain a null terminator.  */
-
-	/*
-	 * Extra nops here:
-	 *	separate load quads from store quads
-	 *	only one branch/quad to permit predictor training
-	 */
-
-	.align 4
-$u_loop:
-	extqh	t2, a1, t0	# U : extract high bits for current word
-	addq	a1, 8, a1	# E :
-	extql	t2, a1, t3	# U : extract low bits for next time
-	addq	a0, 8, a0	# E :
-
-	or	t0, t1, t0	# E : current dst word now complete
-	EX( ldq_u t2, 0(a1) )	# L : load high word for next time
-	subq	a2, 1, a2	# E :
-	nop
-
-	stq_u	t0, -8(a0)	# L : save the current word
-	mov	t3, t1		# E :
-	cmpbge	zero, t2, t8	# E : test new word for eos
-	beq	a2, $u_eoc	# U :
-
-	beq	t8, $u_loop	# U :
-	nop
-	nop
-	nop
-
-	/* We've found a zero somewhere in the source word we just read.
-	   If it resides in the lower half, we have one (probably partial)
-	   word to write out, and if it resides in the upper half, we
-	   have one full and one partial word left to write out.
-
-	   On entry to this basic block:
-	   t1 == the shifted high-order bits from the previous source word
-	   t2 == the unshifted current source word.  */
-	.align 4
-$u_eos:
-	extqh	t2, a1, t0	# U :
-	or	t0, t1, t0	# E : first (partial) source word complete
-	cmpbge	zero, t0, t8	# E : is the null in this first bit?
-	nop
-
-	bne	t8, $u_final	# U :
-	stq_u	t0, 0(a0)	# L : the null was in the high-order bits
-	addq	a0, 8, a0	# E :
-	subq	a2, 1, a2	# E :
-
-	.align 4
-$u_late_head_exit:
-	extql	t2, a1, t0	# U :
-	cmpbge	zero, t0, t8	# E :
-	or	t8, t10, t6	# E :
-	cmoveq	a2, t6, t8	# E :
-
-	/* Take care of a final (probably partial) result word.
-	   On entry to this basic block:
-	   t0 == assembled source word
-	   t8 == cmpbge mask that found the null.  */
-	.align 4
-$u_final:
-	negq	t8, t6		# E : isolate low bit set
-	and	t6, t8, t12	# E :
-	ldq_u	t1, 0(a0)	# L :
-	subq	t12, 1, t6	# E :
-
-	or	t6, t12, t8	# E :
-	zapnot	t0, t8, t0	# U : kill source bytes > null
-	zap	t1, t8, t1	# U : kill dest bytes <= null
-	or	t0, t1, t0	# E :
-
-	stq_u	t0, 0(a0)	# E :
-	br	$finish_up	# U :
-	nop
-	nop
-
-	.align 4
-$u_eoc:				# end-of-count
-	extqh	t2, a1, t0	# U :
-	or	t0, t1, t0	# E :
-	cmpbge	zero, t0, t8	# E :
-	nop
-
-	.align 4
-$u_eocfin:			# end-of-count, final word
-	or	t10, t8, t8	# E :
-	br	$u_final	# U :
-	nop
-	nop
-
-	/* Unaligned copy entry point.  */
-	.align 4
-$unaligned:
-
-	srl	a3, 3, a2	# U : a2 = loop counter = (count - 1)/8
-	and	a0, 7, t4	# E : find dest misalignment
-	and	a1, 7, t5	# E : find src misalignment
-	mov	zero, t0	# E :
-
-	/* Conditionally load the first destination word and a bytemask
-	   with 0xff indicating that the destination byte is sacrosanct.  */
-
-	mov	zero, t6	# E :
-	beq	t4, 1f		# U :
-	ldq_u	t0, 0(a0)	# L :
-	lda	t6, -1		# E :
-
-	mskql	t6, a0, t6	# E :
-	nop
-	nop
-	nop
-
-	.align 4
-1:
-	subq	a1, t4, a1	# E : sub dest misalignment from src addr
-	/* If source misalignment is larger than dest misalignment, we need
-	   extra startup checks to avoid SEGV.  */
-	cmplt	t4, t5, t12	# E :
-	extql	t1, a1, t1	# U : shift src into place
-	lda	t2, -1		# E : for creating masks later
-
-	beq	t12, $u_head	# U :
-	mskqh	t2, t5, t2	# U : begin src byte validity mask
-	cmpbge	zero, t1, t8	# E : is there a zero?
-	nop
-
-	extql	t2, a1, t2	# U :
-	or	t8, t10, t5	# E : test for end-of-count too
-	cmpbge	zero, t2, t3	# E :
-	cmoveq	a2, t5, t8	# E : Latency=2, extra map slot
-
-	nop			# E : goes with cmov
-	andnot	t8, t3, t8	# E :
-	beq	t8, $u_head	# U :
-	nop
-
-	/* At this point we've found a zero in the first partial word of
-	   the source.  We need to isolate the valid source data and mask
-	   it into the original destination data.  (Incidentally, we know
-	   that we'll need at least one byte of that original dest word.) */
-
-	ldq_u	t0, 0(a0)	# L :
-	negq	t8, t6		# E : build bitmask of bytes <= zero
-	mskqh	t1, t4, t1	# U :
-	and	t6, t8, t12	# E :
-
-	subq	t12, 1, t6	# E :
-	or	t6, t12, t8	# E :
-	zapnot	t2, t8, t2	# U : prepare source word; mirror changes
-	zapnot	t1, t8, t1	# U : to source validity mask
-
-	andnot	t0, t2, t0	# E : zero place for source to reside
-	or	t0, t1, t0	# E : and put it there
-	stq_u	t0, 0(a0)	# L :
-	nop
-
-	.align 4
-$finish_up:
-	zapnot	t0, t12, t4	# U : was last byte written null?
-	and	t12, 0xf0, t3	# E : binary search for the address of the
-	cmovne	t4, 1, t4	# E : Latency=2, extra map slot
-	nop			# E : with cmovne
-
-	and	t12, 0xcc, t2	# E : last byte written
-	and	t12, 0xaa, t1	# E :
-	cmovne	t3, 4, t3	# E : Latency=2, extra map slot
-	nop			# E : with cmovne
-
-	bic	a0, 7, t0
-	cmovne	t2, 2, t2	# E : Latency=2, extra map slot
-	nop			# E : with cmovne
-	nop
-
-	cmovne	t1, 1, t1	# E : Latency=2, extra map slot
-	nop			# E : with cmovne
-	addq	t0, t3, t0	# E :
-	addq	t1, t2, t1	# E :
-
-	addq	t0, t1, t0	# E :
-	addq	t0, t4, t0	# add one if we filled the buffer
-	subq	t0, v0, v0	# find string length
-	ret			# L0 :
-
-	.align 4
-$zerolength:
-	nop
-	nop
-	nop
-	clr	v0
-
-$exception:
-	nop
-	nop
-	nop
-	ret
-
-	.end __strncpy_from_user

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

arch/alpha/lib/strncpy_from_user.S

-/*
- * arch/alpha/lib/strncpy_from_user.S
- * Contributed by Richard Henderson (rth@tamu.edu)
- *
- * Just like strncpy except in the return value:
- *
- * -EFAULT       if an exception occurs before the terminator is copied.
- * N             if the buffer filled.
- *
- * Otherwise the length of the string is returned.
- */
-
-
-#include <asm/errno.h>
-#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 $31, $exception-99b($0); 	\
-	.previous
-
-
-	.set noat
-	.set noreorder
-	.text
-
-	.globl __strncpy_from_user
-	.ent __strncpy_from_user
-	.frame $30, 0, $26
-	.prologue 0
-
-	.align 3
-$aligned:
-	/* On entry to this basic block:
-	   t0 == the first destination word for masking back in
-	   t1 == the first source word.  */
-
-	/* Create the 1st output word and detect 0's in the 1st input word.  */
-	lda	t2, -1		# e1    : build a mask against false zero
-	mskqh	t2, a1, t2	# e0    :   detection in the src word
-	mskqh	t1, a1, t3	# e0    :
-	ornot	t1, t2, t2	# .. e1 :
-	mskql	t0, a1, t0	# e0    : assemble the first output word
-	cmpbge	zero, t2, t8	# .. e1 : bits set iff null found
-	or	t0, t3, t0	# e0    :
-	beq	a2, $a_eoc	# .. e1 :
-	bne	t8, $a_eos	# .. e1 :
-
-	/* On entry to this basic block:
-	   t0 == a source word not containing a null.  */
-
-$a_loop:
-	stq_u	t0, 0(a0)	# e0    :
-	addq	a0, 8, a0	# .. e1 :
-	EX( ldq_u t0, 0(a1) )	# e0    :
-	addq	a1, 8, a1	# .. e1 :
-	subq	a2, 1, a2	# e0    :
-	cmpbge	zero, t0, t8	# .. e1 (stall)
-	beq	a2, $a_eoc      # e1    :
-	beq	t8, $a_loop	# e1    :
-
-	/* Take care of the final (partial) word store.  At this point
-	   the end-of-count bit is set in t8 iff it applies.
-
-	   On entry to this basic block we have:
-	   t0 == the source word containing the null
-	   t8 == the cmpbge mask that found it.  */
-
-$a_eos:
-	negq	t8, t12		# e0    : find low bit set
-	and	t8, t12, t12	# e1 (stall)
-
-	/* For the sake of the cache, don't read a destination word
-	   if we're not going to need it.  */
-	and	t12, 0x80, t6	# e0    :
-	bne	t6, 1f		# .. e1 (zdb)
-
-	/* We're doing a partial word store and so need to combine
-	   our source and original destination words.  */
-	ldq_u	t1, 0(a0)	# e0    :
-	subq	t12, 1, t6	# .. e1 :
-	or	t12, t6, t8	# e0    :
-	unop			#
-	zapnot	t0, t8, t0	# e0    : clear src bytes > null
-	zap	t1, t8, t1	# .. e1 : clear dst bytes <= null
-	or	t0, t1, t0	# e1    :
-
-1:	stq_u	t0, 0(a0)
-	br	$finish_up
-
-	/* Add the end-of-count bit to the eos detection bitmask.  */
-$a_eoc:
-	or	t10, t8, t8
-	br	$a_eos
-
-	/*** The Function Entry Point ***/
-	.align 3
-__strncpy_from_user:
-	mov	a0, v0		# save the string start
-	beq	a2, $zerolength
-
-	/* Are source and destination co-aligned?  */
-	xor	a0, a1, t1	# e0    :
-	and	a0, 7, t0	# .. e1 : find dest misalignment
-	and	t1, 7, t1	# e0    :
-	addq	a2, t0, a2	# .. e1 : bias count by dest misalignment
-	subq	a2, 1, a2	# e0    :
-	and	a2, 7, t2	# e1    :
-	srl	a2, 3, a2	# e0    : a2 = loop counter = (count - 1)/8
-	addq	zero, 1, t10	# .. e1 :
-	sll	t10, t2, t10	# e0    : t10 = bitmask of last count byte
-	bne	t1, $unaligned	# .. e1 :
-
-	/* We are co-aligned; take care of a partial first word.  */
-
-	EX( ldq_u t1, 0(a1) )	# e0    : load first src word
-	addq	a1, 8, a1	# .. e1 :
-
-	beq	t0, $aligned	# avoid loading dest word if not needed
-	ldq_u	t0, 0(a0)	# e0    :
-	br	$aligned	# .. e1 :
-
-
-/* The source and destination are not co-aligned.  Align the destination
-   and cope.  We have to be very careful about not reading too much and
-   causing a SEGV.  */
-
-	.align 3
-$u_head:
-	/* We know just enough now to be able to assemble the first
-	   full source word.  We can still find a zero at the end of it
-	   that prevents us from outputting the whole thing.
-
-	   On entry to this basic block:
-	   t0 == the first dest word, unmasked
-	   t1 == the shifted low bits of the first source word
-	   t6 == bytemask that is -1 in dest word bytes */
-
-	EX( ldq_u t2, 8(a1) )	# e0    : load second src word
-	addq	a1, 8, a1	# .. e1 :
-	mskql	t0, a0, t0	# e0    : mask trailing garbage in dst
-	extqh	t2, a1, t4	# e0    :
-	or	t1, t4, t1	# e1    : first aligned src word complete
-	mskqh	t1, a0, t1	# e0    : mask leading garbage in src
-	or	t0, t1, t0	# e0    : first output word complete
-	or	t0, t6, t6	# e1    : mask original data for zero test
-	cmpbge	zero, t6, t8	# e0    :
-	beq	a2, $u_eocfin	# .. e1 :
-	bne	t8, $u_final	# e1    :
-
-	lda	t6, -1			# e1    : mask out the bits we have
-	mskql	t6, a1, t6		# e0    :   already seen
-	stq_u	t0, 0(a0)		# e0    : store first output word
-	or      t6, t2, t2		# .. e1 :
-	cmpbge	zero, t2, t8		# e0    : find nulls in second partial
-	addq	a0, 8, a0		# .. e1 :
-	subq	a2, 1, a2		# e0    :
-	bne	t8, $u_late_head_exit	# .. e1 :
-
-	/* Finally, we've got all the stupid leading edge cases taken care
-	   of and we can set up to enter the main loop.  */
-
-	extql	t2, a1, t1	# e0    : position hi-bits of lo word
-	EX( ldq_u t2, 8(a1) )	# .. e1 : read next high-order source word
-	addq	a1, 8, a1	# e0    :
-	cmpbge	zero, t2, t8	# e1 (stall)
-	beq	a2, $u_eoc	# e1    :
-	bne	t8, $u_eos	# e1    :
-
-	/* Unaligned copy main loop.  In order to avoid reading too much,
-	   the loop is structured to detect zeros in aligned source words.
-	   This has, unfortunately, effectively pulled half of a loop
-	   iteration out into the head and half into the tail, but it does
-	   prevent nastiness from accumulating in the very thing we want
-	   to run as fast as possible.
-
-	   On entry to this basic block:
-	   t1 == the shifted high-order bits from the previous source word
-	   t2 == the unshifted current source word
-
-	   We further know that t2 does not contain a null terminator.  */
-
-	.align 3
-$u_loop:
-	extqh	t2, a1, t0	# e0    : extract high bits for current word
-	addq	a1, 8, a1	# .. e1 :
-	extql	t2, a1, t3	# e0    : extract low bits for next time
-	addq	a0, 8, a0	# .. e1 :
-	or	t0, t1, t0	# e0    : current dst word now complete
-	EX( ldq_u t2, 0(a1) )	# .. e1 : load high word for next time
-	stq_u	t0, -8(a0)	# e0    : save the current word
-	mov	t3, t1		# .. e1 :
-	subq	a2, 1, a2	# e0    :
-	cmpbge	zero, t2, t8	# .. e1 : test new word for eos
-	beq	a2, $u_eoc	# e1    :
-	beq	t8, $u_loop	# e1    :
-
-	/* We've found a zero somewhere in the source word we just read.
-	   If it resides in the lower half, we have one (probably partial)
-	   word to write out, and if it resides in the upper half, we
-	   have one full and one partial word left to write out.
-
-	   On entry to this basic block:
-	   t1 == the shifted high-order bits from the previous source word
-	   t2 == the unshifted current source word.  */
-$u_eos:
-	extqh	t2, a1, t0	# e0    :
-	or	t0, t1, t0	# e1    : first (partial) source word complete
-
-	cmpbge	zero, t0, t8	# e0    : is the null in this first bit?
-	bne	t8, $u_final	# .. e1 (zdb)
-
-	stq_u	t0, 0(a0)	# e0    : the null was in the high-order bits
-	addq	a0, 8, a0	# .. e1 :
-	subq	a2, 1, a2	# e1    :
-
-$u_late_head_exit:
-	extql	t2, a1, t0	# .. e0 :
-	cmpbge	zero, t0, t8	# e0    :
-	or	t8, t10, t6	# e1    :
-	cmoveq	a2, t6, t8	# e0    :
-	nop			# .. e1 :
-
-	/* Take care of a final (probably partial) result word.
-	   On entry to this basic block:
-	   t0 == assembled source word
-	   t8 == cmpbge mask that found the null.  */
-$u_final:
-	negq	t8, t6		# e0    : isolate low bit set
-	and	t6, t8, t12	# e1    :
-
-	and	t12, 0x80, t6	# e0    : avoid dest word load if we can
-	bne	t6, 1f		# .. e1 (zdb)
-
-	ldq_u	t1, 0(a0)	# e0    :
-	subq	t12, 1, t6	# .. e1 :
-	or	t6, t12, t8	# e0    :
-	zapnot	t0, t8, t0	# .. e1 : kill source bytes > null
-	zap	t1, t8, t1	# e0    : kill dest bytes <= null
-	or	t0, t1, t0	# e1    :
-
-1:	stq_u	t0, 0(a0)	# e0    :
-	br	$finish_up
-
-$u_eoc:				# end-of-count
-	extqh	t2, a1, t0
-	or	t0, t1, t0
-	cmpbge	zero, t0, t8
-
-$u_eocfin:			# end-of-count, final word
-	or	t10, t8, t8
-	br	$u_final
-
-	/* Unaligned copy entry point.  */
-	.align 3
-$unaligned:
-
-	EX( ldq_u t1, 0(a1) )	# e0    : load first source word
-
-	and	a0, 7, t4	# .. e1 : find dest misalignment
-	and	a1, 7, t5	# e0    : find src misalignment
-
-	/* Conditionally load the first destination word and a bytemask
-	   with 0xff indicating that the destination byte is sacrosanct.  */
-
-	mov	zero, t0	# .. e1 :
-	mov	zero, t6	# e0    :
-	beq	t4, 1f		# .. e1 :
-	ldq_u	t0, 0(a0)	# e0    :
-	lda	t6, -1		# .. e1 :
-	mskql	t6, a0, t6	# e0    :
-1:
-	subq	a1, t4, a1	# .. e1 : sub dest misalignment from src addr
-
-	/* If source misalignment is larger than dest misalignment, we need
-	   extra startup checks to avoid SEGV.  */
-
-	cmplt	t4, t5, t12	# e1    :
-	extql	t1, a1, t1	# .. e0 : shift src into place
-	lda	t2, -1		# e0    : for creating masks later
-	beq	t12, $u_head	# e1    :
-
-	mskqh	t2, t5, t2	# e0    : begin src byte validity mask
-	cmpbge	zero, t1, t8	# .. e1 : is there a zero?
-	extql	t2, a1, t2	# e0    :
-	or	t8, t10, t5	# .. e1 : test for end-of-count too
-	cmpbge	zero, t2, t3	# e0    :
-	cmoveq	a2, t5, t8	# .. e1 :
-	andnot	t8, t3, t8	# e0    :
-	beq	t8, $u_head	# .. e1 (zdb)
-
-	/* At this point we've found a zero in the first partial word of
-	   the source.  We need to isolate the valid source data and mask
-	   it into the original destination data.  (Incidentally, we know
-	   that we'll need at least one byte of that original dest word.) */
-
-	ldq_u	t0, 0(a0)	# e0    :
-	negq	t8, t6		# .. e1 : build bitmask of bytes <= zero
-	mskqh	t1, t4, t1	# e0    :
-	and	t6, t8, t12	# .. e1 :
-	subq	t12, 1, t6	# e0    :
-	or	t6, t12, t8	# e1    :
-
-	zapnot	t2, t8, t2	# e0    : prepare source word; mirror changes
-	zapnot	t1, t8, t1	# .. e1 : to source validity mask
-
-	andnot	t0, t2, t0	# e0    : zero place for source to reside
-	or	t0, t1, t0	# e1    : and put it there
-	stq_u	t0, 0(a0)	# e0    :
-
-$finish_up:
-	zapnot	t0, t12, t4	# was last byte written null?
-	cmovne	t4, 1, t4
-
-	and	t12, 0xf0, t3	# binary search for the address of the
-	and	t12, 0xcc, t2	# last byte written
-	and	t12, 0xaa, t1
-	bic	a0, 7, t0
-	cmovne	t3, 4, t3
-	cmovne	t2, 2, t2
-	cmovne	t1, 1, t1
-	addq	t0, t3, t0
-	addq	t1, t2, t1
-	addq	t0, t1, t0
-	addq	t0, t4, t0	# add one if we filled the buffer
-
-	subq	t0, v0, v0	# find string length
-	ret
-
-$zerolength:
-	clr	v0
-$exception:
-	ret
-
-	.end __strncpy_from_user

arch/alpha/mm/fault.c

@@ -89,6 +89,8 @@ do_page_fault(unsigned long address, unsigned long mmcsr,
 	const struct exception_table_entry *fixup;
 	int fault, si_code = SEGV_MAPERR;
 	siginfo_t info;
+	unsigned int flags = (FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
+			      (cause > 0 ? FAULT_FLAG_WRITE : 0));
 
 	/* As of EV6, a load into $31/$f31 is a prefetch, and never faults
 	   (or is suppressed by the PALcode).  Support that for older CPUs
@@ -114,6 +116,7 @@ do_page_fault(unsigned long address, unsigned long mmcsr,
 		goto vmalloc_fault;
 #endif
 
+retry:
 	down_read(&mm->mmap_sem);
 	vma = find_vma(mm, address);
 	if (!vma)
@@ -144,8 +147,11 @@ do_page_fault(unsigned long address, unsigned long mmcsr,
 	/* If for any reason at all we couldn't handle the fault,
 	   make sure we exit gracefully rather than endlessly redo
 	   the fault.  */
-	fault = handle_mm_fault(mm, vma, address, cause > 0 ? FAULT_FLAG_WRITE : 0);
-	up_read(&mm->mmap_sem);
+	fault = handle_mm_fault(mm, vma, address, flags);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
 	if (unlikely(fault & VM_FAULT_ERROR)) {
 		if (fault & VM_FAULT_OOM)
 			goto out_of_memory;
@@ -153,10 +159,26 @@ do_page_fault(unsigned long address, unsigned long mmcsr,
 			goto do_sigbus;
 		BUG();
 	}
-	if (fault & VM_FAULT_MAJOR)
-		current->maj_flt++;
-	else
-		current->min_flt++;
+
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR)
+			current->maj_flt++;
+		else
+			current->min_flt++;
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+
+			 /* No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
+	}
+
+	up_read(&mm->mmap_sem);
+
 	return;
 
 	/* Something tried to access memory that isn't in our memory map.
@@ -186,12 +208,14 @@ do_page_fault(unsigned long address, unsigned long mmcsr,
 	/* We ran out of memory, or some other thing happened to us that
 	   made us unable to handle the page fault gracefully.  */
  out_of_memory:
+	up_read(&mm->mmap_sem);
 	if (!user_mode(regs))
 		goto no_context;
 	pagefault_out_of_memory();
 	return;
 
  do_sigbus:
+	up_read(&mm->mmap_sem);
 	/* Send a sigbus, regardless of whether we were in kernel
 	   or user mode.  */
 	info.si_signo = SIGBUS;

arch/alpha/oprofile/common.c

@@ -12,6 +12,7 @@
 #include <linux/smp.h>
 #include <linux/errno.h>
 #include <asm/ptrace.h>
+#include <asm/special_insns.h>
 
 #include "op_impl.h"
 

kernel/timer.c

@@ -1407,13 +1407,6 @@ SYSCALL_DEFINE1(alarm, unsigned int, seconds)
 
 #endif
 
-#ifndef __alpha__
-
-/*
- * The Alpha uses getxpid, getxuid, and getxgid instead.  Maybe this
- * should be moved into arch/i386 instead?
- */
-
 /**
  * sys_getpid - return the thread group id of the current process
  *
@@ -1469,8 +1462,6 @@ SYSCALL_DEFINE0(getegid)
 	return from_kgid_munged(current_user_ns(), current_egid());
 }
 
-#endif
-
 static void process_timeout(unsigned long __data)
 {
 	wake_up_process((struct task_struct *)__data);