o Merge with Linux 2.1.116.

 o New Newport console code.
 o New G364 console code.

5 files changed, 188 insertions, 37 deletions

diff --git a/arch/alpha/lib/Makefile b/arch/alpha/lib/Makefile
index 7c1e37bac..4dd7d591d 100644
--- a/arch/alpha/lib/Makefile
+++ b/arch/alpha/lib/Makefile
@@ -7,7 +7,7 @@ OBJS  = __divqu.o __remqu.o __divlu.o __remlu.o memset.o memcpy.o io.o \
 	strcat.o strcpy.o strncat.o strncpy.o stxcpy.o stxncpy.o \
 	strchr.o strrchr.o \
 	copy_user.o clear_user.o strncpy_from_user.o strlen_user.o \
-	csum_ipv6_magic.o
+	csum_ipv6_magic.o strcasecmp.o
 
 lib.a: $(OBJS)
 	$(AR) rcs lib.a $(OBJS)
diff --git a/arch/alpha/lib/io.c b/arch/alpha/lib/io.c
index 405a3afa8..50cc1fe05 100644
--- a/arch/alpha/lib/io.c
+++ b/arch/alpha/lib/io.c
@@ -3,26 +3,9 @@
  * files..
  */
 #include <linux/kernel.h>
-
+#include <linux/types.h>
 #include <asm/io.h>
 
-/* 
- * Jensen has a separate "local" and "bus" IO space for
- * byte-wide IO.
- */
-#ifdef __is_local
-
-unsigned int _bus_inb(unsigned long addr)
-{
-	return __bus_inb(addr);
-}
-
-void _bus_outb(unsigned char b, unsigned long addr)
-{
-	__bus_outb(b, addr);
-}
-#endif
-
 unsigned int _inb(unsigned long addr)
 {
 	return __inb(addr);
@@ -70,6 +53,10 @@ unsigned long _readl(unsigned long addr)
 	return __readl(addr);
 }
 
+unsigned long _readq(unsigned long addr)
+{
+	return __readq(addr);
+}
 
 void _writeb(unsigned char b, unsigned long addr)
 {
@@ -86,6 +73,11 @@ void _writel(unsigned int b, unsigned long addr)
 	__writel(b, addr);
 }
 
+void _writeq(unsigned long b, unsigned long addr)
+{
+	__writeq(b, addr);
+}
+
 /*
  * Read COUNT 8-bit bytes from port PORT into memory starting at
  * SRC.
@@ -363,12 +355,48 @@ void outsl (unsigned long port, const void *src, unsigned long count)
  * Copy data from IO memory space to "real" memory space.
  * This needs to be optimized.
  */
-void _memcpy_fromio(void * to, unsigned long from, unsigned long count)
+void _memcpy_fromio(void * to, unsigned long from, long count)
 {
-	while (count) {
+	/* Optimize co-aligned transfers.  Everything else gets handled
+	   a byte at a time. */
+
+	if (count >= 8 && ((long)to & 7) == (from & 7)) {
+		count -= 8;
+		do {
+			*(u64 *)to = readq(from);
+			count -= 8;
+			to += 8;
+			from += 8;
+		} while (count >= 0);
+		count += 8;
+	}
+
+	if (count >= 4 && ((long)to & 3) == (from & 3)) {
+		count -= 4;
+		do {
+			*(u32 *)to = readl(from);
+			count -= 4;
+			to += 4;
+			from += 4;
+		} while (count >= 0);
+		count += 4;
+	}
+		
+	if (count >= 2 && ((long)to & 1) == (from & 1)) {
+		count -= 2;
+		do {
+			*(u16 *)to = readw(from);
+			count -= 2;
+			to += 2;
+			from += 2;
+		} while (count >= 0);
+		count += 2;
+	}
+
+	while (count > 0) {
+		*(u8 *) to = readb(from);
 		count--;
-		*(char *) to = readb(from);
-		((char *) to)++;
+		to++;
 		from++;
 	}
 }
@@ -377,25 +405,105 @@ void _memcpy_fromio(void * to, unsigned long from, unsigned long count)
  * Copy data from "real" memory space to IO memory space.
  * This needs to be optimized.
  */
-void _memcpy_toio(unsigned long to, void * from, unsigned long count)
+void _memcpy_toio(unsigned long to, void * from, long count)
 {
-	while (count) {
+	/* Optimize co-aligned transfers.  Everything else gets handled
+	   a byte at a time. */
+
+	if (count >= 8 && (to & 7) == ((long)from & 7)) {
+		count -= 8;
+		do {
+			writeq(*(u64 *)from, to);
+			count -= 8;
+			to += 8;
+			from += 8;
+		} while (count >= 0);
+		count += 8;
+	}
+
+	if (count >= 4 && (to & 3) == ((long)from & 3)) {
+		count -= 4;
+		do {
+			writel(*(u32 *)from, to);
+			count -= 4;
+			to += 4;
+			from += 4;
+		} while (count >= 0);
+		count += 4;
+	}
+		
+	if (count >= 2 && (to & 1) == ((long)from & 1)) {
+		count -= 2;
+		do {
+			writew(*(u16 *)from, to);
+			count -= 2;
+			to += 2;
+			from += 2;
+		} while (count >= 0);
+		count += 2;
+	}
+
+	while (count > 0) {
+		writeb(*(u8 *) from, to);
 		count--;
-		writeb(*(char *) from, to);
-		((char *) from)++;
 		to++;
+		from++;
 	}
 }
 
 /*
  * "memset" on IO memory space.
- * This needs to be optimized.
  */
-void _memset_io(unsigned long dst, int c, unsigned long count)
+void _memset_c_io(unsigned long to, unsigned long c, long count)
 {
-	while (count) {
+	/* Handle any initial odd byte */
+	if (count > 0 && (to & 1)) {
+		writeb(c, to);
+		to++;
 		count--;
-		writeb(c, dst);
-		dst++;
+	}
+
+	/* Handle any initial odd halfword */
+	if (count >= 2 && (to & 2)) {
+		writew(c, to);
+		to += 2;
+		count -= 2;
+	}
+
+	/* Handle any initial odd word */
+	if (count >= 4 && (to & 4)) {
+		writel(c, to);
+		to += 4;
+		count -= 4;
+	}
+
+	/* Handle all full-sized quadwords: we're aligned (or have a small count) */
+	count -= 8;
+	if (count >= 0) {
+		do {
+			writeq(c, to);
+			to += 8;
+			count -= 8;
+		} while (count >= 0);
+	}
+	count += 8;
+
+	/* The tail is word-aligned if we still have count >= 4 */
+	if (count >= 4) {
+		writel(c, to);
+		to += 4;
+		count -= 4;
+	}
+
+	/* The tail is half-word aligned if we have count >= 2 */
+	if (count >= 2) {
+		writew(c, to);
+		to += 2;
+		count -= 2;
+	}
+
+	/* And finally, one last byte.. */
+	if (count) {
+		writeb(c, to);
 	}
 }
diff --git a/arch/alpha/lib/memset.S b/arch/alpha/lib/memset.S
index dc2075e1f..f39197dee 100644
--- a/arch/alpha/lib/memset.S
+++ b/arch/alpha/lib/memset.S
@@ -18,6 +18,7 @@
 	.set noreorder
 .text
 	.globl __memset
+	.globl __memsetw
 	.globl __constant_c_memset
 	.ent __memset
 .align 5
@@ -25,8 +26,8 @@ __memset:
 	.frame $30,0,$26,0
 	.prologue 0
 
-	zapnot $17,1,$17	/* E0 */
-	sll $17,8,$1		/* E1 (p-c latency, next cycle) */
+	and $17,255,$1		/* E1 */
+	insbl $17,1,$17		/* .. E0 */
 	bis $17,$1,$17		/* E0 (p-c latency, next cycle) */
 	sll $17,16,$1		/* E1 (p-c latency, next cycle) */
 
@@ -52,9 +53,9 @@ __constant_c_memset:
 	insql $17,$16,$2	/* E0 */
 	subq $3,8,$3		/* .. E1 */
 
-	addq $18,$3,$18		/* E0		$18 is new count ($3 is negative) */
+	addq $18,$3,$18		/* E0	$18 is new count ($3 is negative) */
 	mskql $4,$16,$4		/* .. E1 (and possible load stall) */
-	subq $16,$3,$16		/* E0 		$16 is new aligned destination */
+	subq $16,$3,$16		/* E0 	$16 is new aligned destination */
 	bis $2,$4,$1		/* .. E1 */
 
 	bis $31,$31,$31		/* E0 */
@@ -103,6 +104,22 @@ end:
 	ret $31,($26),1		/* E1 */
 	.end __memset
 
+	.align 5
+	.ent __memsetw
+__memsetw:
+	.prologue 0
+
+	inswl $17,0,$1		/* E0 */
+	inswl $17,2,$2		/* E0 */
+	inswl $17,4,$3		/* E0 */
+	or $1,$2,$1		/* .. E1 */
+	inswl $17,6,$4		/* E0 */
+	or $1,$3,$1		/* .. E1 */
+	or $1,$4,$17		/* E0 */
+	br __constant_c_memset	/* .. E1 */
+
+	.end __memsetw
+
 #ifdef __ELF__
  .weak memset; memset = __memset
 #else
diff --git a/arch/alpha/lib/strcasecmp.c b/arch/alpha/lib/strcasecmp.c
new file mode 100644
index 000000000..4e57a216f
--- /dev/null
+++ b/arch/alpha/lib/strcasecmp.c
@@ -0,0 +1,26 @@
+/*
+ *  linux/arch/alpha/lib/strcasecmp.c
+ */
+
+#include <linux/string.h>
+
+
+/* We handle nothing here except the C locale.  Since this is used in
+   only one place, on strings known to contain only 7 bit ASCII, this
+   is ok.  */
+
+int strcasecmp(const char *a, const char *b)
+{
+	int ca, cb;
+
+	do {
+		ca = *a++ & 0xff;
+		cb = *b++ & 0xff;
+		if (ca >= 'A' && ca <= 'Z')
+			ca += 'a' - 'A';
+		if (cb >= 'A' && cb <= 'Z')
+			cb += 'a' - 'A';
+	} while (ca == cb && ca != '\0');
+
+	return ca - cb;
+}
diff --git a/arch/alpha/lib/strlen_user.S b/arch/alpha/lib/strlen_user.S
index b84d9a4c3..cdf71158f 100644
--- a/arch/alpha/lib/strlen_user.S
+++ b/arch/alpha/lib/strlen_user.S
@@ -2,7 +2,7 @@
  * arch/alpha/lib/strlen_user.S
  *
  * Return the length of the string including the NUL terminator
- * (strlen+1) or zero if an error occured.
+ * (strlen+1) or zero if an error occurred.
  */
 
 #include <alpha/regdef.h>