Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Cobalt: Fix UART I/O type
  [MIPS] Fixup copy_from_user_inatomic
  [MIPS] Fix struct sigcontext for N32 userland
  [MIPS] Make some __setup functions static
  [MIPS] Declare highstart_pfn, highend_pfn only if CONFIG_HIGHMEM=y
  [MIPS] Allow selection of KGDB only on platforms where it's supported.

10 files changed, 532 insertions, 33 deletions

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 46fa5ca9f5a..c6f74f1c639 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -144,6 +144,7 @@ config BASLER_EXCITE
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	help
 	  The eXcite is a smart camera platform manufactured by
 	  Basler Vision Technologies AG
@@ -210,6 +211,7 @@ config MIPS_EV64120
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	help
 	  This is an evaluation board based on the Galileo GT-64120
 	  single-chip system controller that contains a MIPS R5000 compatible
@@ -374,6 +376,7 @@ config MOMENCO_JAGUAR_ATX
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	help
 	  The Jaguar ATX is a MIPS-based Single Board Computer (SBC) made by
 	  Momentum Computer <http://www.momenco.com/>.
@@ -391,6 +394,7 @@ config MOMENCO_OCELOT
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	help
 	  The Ocelot is a MIPS-based Single Board Computer (SBC) made by
 	  Momentum Computer <http://www.momenco.com/>.
@@ -480,6 +484,8 @@ config DDB5477
 	select SYS_HAS_CPU_R5432
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_64BIT_KERNEL if EXPERIMENTAL
+	select SYS_SUPPORTS_KGDB
+	select SYS_SUPPORTS_KGDB
 	select SYS_SUPPORTS_LITTLE_ENDIAN
 	help
 	  This enables support for the R5432-based NEC DDB Vrc-5477,
@@ -508,6 +514,7 @@ config PMC_YOSEMITE
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
 	select SYS_SUPPORTS_HIGHMEM
+	select SYS_SUPPORTS_KGDB
 	select SYS_SUPPORTS_SMP
 	help
 	  Yosemite is an evaluation board for the RM9000x2 processor
@@ -584,6 +591,7 @@ config SGI_IP27
 	select SYS_HAS_CPU_R10000
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	select SYS_SUPPORTS_NUMA
 	select SYS_SUPPORTS_SMP
 	select GENERIC_HARDIRQS_NO__DO_IRQ
@@ -633,6 +641,7 @@ config SIBYTE_SWARM
 	select SYS_HAS_CPU_SB1
 	select SYS_SUPPORTS_BIG_ENDIAN
 	select SYS_SUPPORTS_HIGHMEM
+	select SYS_SUPPORTS_KGDB
 	select SYS_SUPPORTS_LITTLE_ENDIAN
 
 config SIBYTE_SENTOSA
@@ -773,6 +782,7 @@ config TOSHIBA_RBTX4927
 	select SYS_SUPPORTS_64BIT_KERNEL
 	select SYS_SUPPORTS_LITTLE_ENDIAN
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	select TOSHIBA_BOARDS
 	select GENERIC_HARDIRQS_NO__DO_IRQ
 	help
@@ -793,6 +803,7 @@ config TOSHIBA_RBTX4938
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_LITTLE_ENDIAN
 	select SYS_SUPPORTS_BIG_ENDIAN
+	select SYS_SUPPORTS_KGDB
 	select TOSHIBA_BOARDS
 	select GENERIC_HARDIRQS_NO__DO_IRQ
 	help
@@ -959,6 +970,7 @@ config IRQ_MV64340
 
 config DDB5XXX_COMMON
 	bool
+	select SYS_SUPPORTS_KGDB
 
 config MIPS_BOARDS_GEN
 	bool
@@ -1007,6 +1019,7 @@ config SOC_AU1X00
 	select SYS_HAS_CPU_MIPS32_R1
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select SYS_SUPPORTS_APM_EMULATION
+	select SYS_SUPPORTS_KGDB
 
 config PNX8550
 	bool
@@ -1019,6 +1032,7 @@ config SOC_PNX8550
 	select SYS_HAS_CPU_MIPS32_R1
 	select SYS_SUPPORTS_32BIT_KERNEL
 	select GENERIC_HARDIRQS_NO__DO_IRQ
+	select SYS_SUPPORTS_KGDB
 
 config SWAP_IO_SPACE
 	bool
diff --git a/arch/mips/Kconfig.debug b/arch/mips/Kconfig.debug
index 9351f1c04a9..d5e60a0c1de 100644
--- a/arch/mips/Kconfig.debug
+++ b/arch/mips/Kconfig.debug
@@ -39,7 +39,7 @@ config DEBUG_STACK_USAGE
 
 config KGDB
 	bool "Remote GDB kernel debugging"
-	depends on DEBUG_KERNEL
+	depends on DEBUG_KERNEL && SYS_SUPPORTS_KGDB
 	select DEBUG_INFO
 	help
 	  If you say Y here, it will be possible to remotely debug the MIPS
@@ -48,6 +48,9 @@ config KGDB
 	  better 32 MB RAM to avoid excessive linking time. This is only
 	  useful for kernel hackers. If unsure, say N.
 
+config SYS_SUPPORTS_KGDB
+	bool
+
 config GDB_CONSOLE
 	bool "Console output to GDB"
 	depends on KGDB
diff --git a/arch/mips/cobalt/setup.c b/arch/mips/cobalt/setup.c
index a4b69b543bd..415ff8710b5 100644
--- a/arch/mips/cobalt/setup.c
+++ b/arch/mips/cobalt/setup.c
@@ -130,7 +130,7 @@ void __init plat_mem_setup(void)
 
 	set_io_port_base(CKSEG1ADDR(GT_DEF_PCI0_IO_BASE));
 
-	/* I/O port resource must include UART and LCD/buttons */
+	/* I/O port resource must include LCD/buttons */
 	ioport_resource.end = 0x0fffffff;
 
 	/* request I/O space for devices used on all i[345]86 PCs */
@@ -149,24 +149,24 @@ void __init plat_mem_setup(void)
 	register_pci_controller(&cobalt_pci_controller);
 #endif
 
-#ifdef CONFIG_SERIAL_8250
 	if (cobalt_board_id > COBALT_BRD_ID_RAQ1) {
-
 #ifdef CONFIG_EARLY_PRINTK
 		cobalt_early_console();
 #endif
 
+#ifdef CONFIG_SERIAL_8250
 		uart.line	= 0;
 		uart.type	= PORT_UNKNOWN;
 		uart.uartclk	= 18432000;
 		uart.irq	= COBALT_SERIAL_IRQ;
-		uart.flags	= UPF_BOOT_AUTOCONF | UPF_SKIP_TEST;
-		uart.iobase	= 0xc800000;
-		uart.iotype	= UPIO_PORT;
+		uart.flags	= UPF_IOREMAP | UPF_BOOT_AUTOCONF |
+				  UPF_SKIP_TEST;
+		uart.iotype	= UPIO_MEM;
+		uart.mapbase	= 0x1c800000;
 
 		early_serial_setup(&uart);
-	}
 #endif
+	}
 }
 
 /*
diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c
index 50ed6d58ae7..ab755ea26c6 100644
--- a/arch/mips/kernel/cpu-probe.c
+++ b/arch/mips/kernel/cpu-probe.c
@@ -98,7 +98,7 @@ static void au1k_wait(void)
 
 static int __initdata nowait = 0;
 
-int __init wait_disable(char *s)
+static int __init wait_disable(char *s)
 {
 	nowait = 1;
 
diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c
index 394540fad76..11ab222a2e9 100644
--- a/arch/mips/kernel/setup.c
+++ b/arch/mips/kernel/setup.c
@@ -543,7 +543,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
 }
 
-int __init fpu_disable(char *s)
+static int __init fpu_disable(char *s)
 {
 	int i;
 
@@ -555,7 +555,7 @@ int __init fpu_disable(char *s)
 
 __setup("nofpu", fpu_disable);
 
-int __init dsp_disable(char *s)
+static int __init dsp_disable(char *s)
 {
 	cpu_data[0].ases &= ~MIPS_ASE_DSP;
 
diff --git a/arch/mips/lib/Makefile b/arch/mips/lib/Makefile
index 9e5d985936b..2453ea244cb 100644
--- a/arch/mips/lib/Makefile
+++ b/arch/mips/lib/Makefile
@@ -2,7 +2,7 @@
 # Makefile for MIPS-specific library files..
 #
 
-lib-y	+= csum_partial.o memcpy.o memset.o promlib.o \
+lib-y	+= csum_partial.o memcpy.o memcpy-inatomic.o memset.o promlib.o \
 	   strlen_user.o strncpy_user.o strnlen_user.o uncached.o
 
 obj-y			+= iomap.o
diff --git a/arch/mips/lib/memcpy-inatomic.S b/arch/mips/lib/memcpy-inatomic.S
new file mode 100644
index 00000000000..3a534b2baa0
--- /dev/null
+++ b/arch/mips/lib/memcpy-inatomic.S
@@ -0,0 +1,436 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Unified implementation of memcpy, memmove and the __copy_user backend.
+ *
+ * Copyright (C) 1998, 99, 2000, 01, 2002 Ralf Baechle (ralf@gnu.org)
+ * Copyright (C) 1999, 2000, 01, 2002 Silicon Graphics, Inc.
+ * Copyright (C) 2002 Broadcom, Inc.
+ *   memcpy/copy_user author: Mark Vandevoorde
+ *
+ * Mnemonic names for arguments to memcpy/__copy_user
+ */
+
+/*
+ * Hack to resolve longstanding prefetch issue
+ *
+ * Prefetching may be fatal on some systems if we're prefetching beyond the
+ * end of memory on some systems.  It's also a seriously bad idea on non
+ * dma-coherent systems.
+ */
+#if !defined(CONFIG_DMA_COHERENT) || !defined(CONFIG_DMA_IP27)
+#undef CONFIG_CPU_HAS_PREFETCH
+#endif
+#ifdef CONFIG_MIPS_MALTA
+#undef CONFIG_CPU_HAS_PREFETCH
+#endif
+
+#include <asm/asm.h>
+#include <asm/asm-offsets.h>
+#include <asm/regdef.h>
+
+#define dst a0
+#define src a1
+#define len a2
+
+/*
+ * Spec
+ *
+ * memcpy copies len bytes from src to dst and sets v0 to dst.
+ * It assumes that
+ *   - src and dst don't overlap
+ *   - src is readable
+ *   - dst is writable
+ * memcpy uses the standard calling convention
+ *
+ * __copy_user copies up to len bytes from src to dst and sets a2 (len) to
+ * the number of uncopied bytes due to an exception caused by a read or write.
+ * __copy_user assumes that src and dst don't overlap, and that the call is
+ * implementing one of the following:
+ *   copy_to_user
+ *     - src is readable  (no exceptions when reading src)
+ *   copy_from_user
+ *     - dst is writable  (no exceptions when writing dst)
+ * __copy_user uses a non-standard calling convention; see
+ * include/asm-mips/uaccess.h
+ *
+ * When an exception happens on a load, the handler must
+ # ensure that all of the destination buffer is overwritten to prevent
+ * leaking information to user mode programs.
+ */
+
+/*
+ * Implementation
+ */
+
+/*
+ * The exception handler for loads requires that:
+ *  1- AT contain the address of the byte just past the end of the source
+ *     of the copy,
+ *  2- src_entry <= src < AT, and
+ *  3- (dst - src) == (dst_entry - src_entry),
+ * The _entry suffix denotes values when __copy_user was called.
+ *
+ * (1) is set up up by uaccess.h and maintained by not writing AT in copy_user
+ * (2) is met by incrementing src by the number of bytes copied
+ * (3) is met by not doing loads between a pair of increments of dst and src
+ *
+ * The exception handlers for stores adjust len (if necessary) and return.
+ * These handlers do not need to overwrite any data.
+ *
+ * For __rmemcpy and memmove an exception is always a kernel bug, therefore
+ * they're not protected.
+ */
+
+#define EXC(inst_reg,addr,handler)		\
+9:	inst_reg, addr;				\
+	.section __ex_table,"a";		\
+	PTR	9b, handler;			\
+	.previous
+
+/*
+ * Only on the 64-bit kernel we can made use of 64-bit registers.
+ */
+#ifdef CONFIG_64BIT
+#define USE_DOUBLE
+#endif
+
+#ifdef USE_DOUBLE
+
+#define LOAD   ld
+#define LOADL  ldl
+#define LOADR  ldr
+#define STOREL sdl
+#define STORER sdr
+#define STORE  sd
+#define ADD    daddu
+#define SUB    dsubu
+#define SRL    dsrl
+#define SRA    dsra
+#define SLL    dsll
+#define SLLV   dsllv
+#define SRLV   dsrlv
+#define NBYTES 8
+#define LOG_NBYTES 3
+
+/*
+ * As we are sharing code base with the mips32 tree (which use the o32 ABI
+ * register definitions). We need to redefine the register definitions from
+ * the n64 ABI register naming to the o32 ABI register naming.
+ */
+#undef t0
+#undef t1
+#undef t2
+#undef t3
+#define t0	$8
+#define t1	$9
+#define t2	$10
+#define t3	$11
+#define t4	$12
+#define t5	$13
+#define t6	$14
+#define t7	$15
+
+#else
+
+#define LOAD   lw
+#define LOADL  lwl
+#define LOADR  lwr
+#define STOREL swl
+#define STORER swr
+#define STORE  sw
+#define ADD    addu
+#define SUB    subu
+#define SRL    srl
+#define SLL    sll
+#define SRA    sra
+#define SLLV   sllv
+#define SRLV   srlv
+#define NBYTES 4
+#define LOG_NBYTES 2
+
+#endif /* USE_DOUBLE */
+
+#ifdef CONFIG_CPU_LITTLE_ENDIAN
+#define LDFIRST LOADR
+#define LDREST  LOADL
+#define STFIRST STORER
+#define STREST  STOREL
+#define SHIFT_DISCARD SLLV
+#else
+#define LDFIRST LOADL
+#define LDREST  LOADR
+#define STFIRST STOREL
+#define STREST  STORER
+#define SHIFT_DISCARD SRLV
+#endif
+
+#define FIRST(unit) ((unit)*NBYTES)
+#define REST(unit)  (FIRST(unit)+NBYTES-1)
+#define UNIT(unit)  FIRST(unit)
+
+#define ADDRMASK (NBYTES-1)
+
+	.text
+	.set	noreorder
+	.set	noat
+
+/*
+ * A combined memcpy/__copy_user
+ * __copy_user sets len to 0 for success; else to an upper bound of
+ * the number of uncopied bytes.
+ * memcpy sets v0 to dst.
+ */
+	.align	5
+LEAF(__copy_user_inatomic)
+	/*
+	 * Note: dst & src may be unaligned, len may be 0
+	 * Temps
+	 */
+#define rem t8
+
+	/*
+	 * The "issue break"s below are very approximate.
+	 * Issue delays for dcache fills will perturb the schedule, as will
+	 * load queue full replay traps, etc.
+	 *
+	 * If len < NBYTES use byte operations.
+	 */
+	PREF(	0, 0(src) )
+	PREF(	1, 0(dst) )
+	sltu	t2, len, NBYTES
+	and	t1, dst, ADDRMASK
+	PREF(	0, 1*32(src) )
+	PREF(	1, 1*32(dst) )
+	bnez	t2, copy_bytes_checklen
+	 and	t0, src, ADDRMASK
+	PREF(	0, 2*32(src) )
+	PREF(	1, 2*32(dst) )
+	bnez	t1, dst_unaligned
+	 nop
+	bnez	t0, src_unaligned_dst_aligned
+	/*
+	 * use delay slot for fall-through
+	 * src and dst are aligned; need to compute rem
+	 */
+both_aligned:
+	 SRL	t0, len, LOG_NBYTES+3    # +3 for 8 units/iter
+	beqz	t0, cleanup_both_aligned # len < 8*NBYTES
+	 and	rem, len, (8*NBYTES-1)	 # rem = len % (8*NBYTES)
+	PREF(	0, 3*32(src) )
+	PREF(	1, 3*32(dst) )
+	.align	4
+1:
+EXC(	LOAD	t0, UNIT(0)(src),	l_exc)
+EXC(	LOAD	t1, UNIT(1)(src),	l_exc_copy)
+EXC(	LOAD	t2, UNIT(2)(src),	l_exc_copy)
+EXC(	LOAD	t3, UNIT(3)(src),	l_exc_copy)
+	SUB	len, len, 8*NBYTES
+EXC(	LOAD	t4, UNIT(4)(src),	l_exc_copy)
+EXC(	LOAD	t7, UNIT(5)(src),	l_exc_copy)
+	STORE	t0, UNIT(0)(dst)
+	STORE	t1, UNIT(1)(dst)
+EXC(	LOAD	t0, UNIT(6)(src),	l_exc_copy)
+EXC(	LOAD	t1, UNIT(7)(src),	l_exc_copy)
+	ADD	src, src, 8*NBYTES
+	ADD	dst, dst, 8*NBYTES
+	STORE	t2, UNIT(-6)(dst)
+	STORE	t3, UNIT(-5)(dst)
+	STORE	t4, UNIT(-4)(dst)
+	STORE	t7, UNIT(-3)(dst)
+	STORE	t0, UNIT(-2)(dst)
+	STORE	t1, UNIT(-1)(dst)
+	PREF(	0, 8*32(src) )
+	PREF(	1, 8*32(dst) )
+	bne	len, rem, 1b
+	 nop
+
+	/*
+	 * len == rem == the number of bytes left to copy < 8*NBYTES
+	 */
+cleanup_both_aligned:
+	beqz	len, done
+	 sltu	t0, len, 4*NBYTES
+	bnez	t0, less_than_4units
+	 and	rem, len, (NBYTES-1)	# rem = len % NBYTES
+	/*
+	 * len >= 4*NBYTES
+	 */
+EXC(	LOAD	t0, UNIT(0)(src),	l_exc)
+EXC(	LOAD	t1, UNIT(1)(src),	l_exc_copy)
+EXC(	LOAD	t2, UNIT(2)(src),	l_exc_copy)
+EXC(	LOAD	t3, UNIT(3)(src),	l_exc_copy)
+	SUB	len, len, 4*NBYTES
+	ADD	src, src, 4*NBYTES
+	STORE	t0, UNIT(0)(dst)
+	STORE	t1, UNIT(1)(dst)
+	STORE	t2, UNIT(2)(dst)
+	STORE	t3, UNIT(3)(dst)
+	beqz	len, done
+	 ADD	dst, dst, 4*NBYTES
+less_than_4units:
+	/*
+	 * rem = len % NBYTES
+	 */
+	beq	rem, len, copy_bytes
+	 nop
+1:
+EXC(	LOAD	t0, 0(src),		l_exc)
+	ADD	src, src, NBYTES
+	SUB	len, len, NBYTES
+	STORE	t0, 0(dst)
+	bne	rem, len, 1b
+	 ADD	dst, dst, NBYTES
+
+	/*
+	 * src and dst are aligned, need to copy rem bytes (rem < NBYTES)
+	 * A loop would do only a byte at a time with possible branch
+	 * mispredicts.  Can't do an explicit LOAD dst,mask,or,STORE
+	 * because can't assume read-access to dst.  Instead, use
+	 * STREST dst, which doesn't require read access to dst.
+	 *
+	 * This code should perform better than a simple loop on modern,
+	 * wide-issue mips processors because the code has fewer branches and
+	 * more instruction-level parallelism.
+	 */
+#define bits t2
+	beqz	len, done
+	 ADD	t1, dst, len	# t1 is just past last byte of dst
+	li	bits, 8*NBYTES
+	SLL	rem, len, 3	# rem = number of bits to keep
+EXC(	LOAD	t0, 0(src),		l_exc)
+	SUB	bits, bits, rem	# bits = number of bits to discard
+	SHIFT_DISCARD t0, t0, bits
+	STREST	t0, -1(t1)
+	jr	ra
+	 move	len, zero
+dst_unaligned:
+	/*
+	 * dst is unaligned
+	 * t0 = src & ADDRMASK
+	 * t1 = dst & ADDRMASK; T1 > 0
+	 * len >= NBYTES
+	 *
+	 * Copy enough bytes to align dst
+	 * Set match = (src and dst have same alignment)
+	 */
+#define match rem
+EXC(	LDFIRST	t3, FIRST(0)(src),	l_exc)
+	ADD	t2, zero, NBYTES
+EXC(	LDREST	t3, REST(0)(src),	l_exc_copy)
+	SUB	t2, t2, t1	# t2 = number of bytes copied
+	xor	match, t0, t1
+	STFIRST t3, FIRST(0)(dst)
+	beq	len, t2, done
+	 SUB	len, len, t2
+	ADD	dst, dst, t2
+	beqz	match, both_aligned
+	 ADD	src, src, t2
+
+src_unaligned_dst_aligned:
+	SRL	t0, len, LOG_NBYTES+2    # +2 for 4 units/iter
+	PREF(	0, 3*32(src) )
+	beqz	t0, cleanup_src_unaligned
+	 and	rem, len, (4*NBYTES-1)   # rem = len % 4*NBYTES
+	PREF(	1, 3*32(dst) )
+1:
+/*
+ * Avoid consecutive LD*'s to the same register since some mips
+ * implementations can't issue them in the same cycle.
+ * It's OK to load FIRST(N+1) before REST(N) because the two addresses
+ * are to the same unit (unless src is aligned, but it's not).
+ */
+EXC(	LDFIRST	t0, FIRST(0)(src),	l_exc)
+EXC(	LDFIRST	t1, FIRST(1)(src),	l_exc_copy)
+	SUB     len, len, 4*NBYTES
+EXC(	LDREST	t0, REST(0)(src),	l_exc_copy)
+EXC(	LDREST	t1, REST(1)(src),	l_exc_copy)
+EXC(	LDFIRST	t2, FIRST(2)(src),	l_exc_copy)
+EXC(	LDFIRST	t3, FIRST(3)(src),	l_exc_copy)
+EXC(	LDREST	t2, REST(2)(src),	l_exc_copy)
+EXC(	LDREST	t3, REST(3)(src),	l_exc_copy)
+	PREF(	0, 9*32(src) )		# 0 is PREF_LOAD  (not streamed)
+	ADD	src, src, 4*NBYTES
+#ifdef CONFIG_CPU_SB1
+	nop				# improves slotting
+#endif
+	STORE	t0, UNIT(0)(dst)
+	STORE	t1, UNIT(1)(dst)
+	STORE	t2, UNIT(2)(dst)
+	STORE	t3, UNIT(3)(dst)
+	PREF(	1, 9*32(dst) )     	# 1 is PREF_STORE (not streamed)
+	bne	len, rem, 1b
+	 ADD	dst, dst, 4*NBYTES
+
+cleanup_src_unaligned:
+	beqz	len, done
+	 and	rem, len, NBYTES-1  # rem = len % NBYTES
+	beq	rem, len, copy_bytes
+	 nop
+1:
+EXC(	LDFIRST t0, FIRST(0)(src),	l_exc)
+EXC(	LDREST	t0, REST(0)(src),	l_exc_copy)
+	ADD	src, src, NBYTES
+	SUB	len, len, NBYTES
+	STORE	t0, 0(dst)
+	bne	len, rem, 1b
+	 ADD	dst, dst, NBYTES
+
+copy_bytes_checklen:
+	beqz	len, done
+	 nop
+copy_bytes:
+	/* 0 < len < NBYTES  */
+#define COPY_BYTE(N)			\
+EXC(	lb	t0, N(src), l_exc);	\
+	SUB	len, len, 1;		\
+	beqz	len, done;		\
+	 sb	t0, N(dst)
+
+	COPY_BYTE(0)
+	COPY_BYTE(1)
+#ifdef USE_DOUBLE
+	COPY_BYTE(2)
+	COPY_BYTE(3)
+	COPY_BYTE(4)
+	COPY_BYTE(5)
+#endif
+EXC(	lb	t0, NBYTES-2(src), l_exc)
+	SUB	len, len, 1
+	jr	ra
+	 sb	t0, NBYTES-2(dst)
+done:
+	jr	ra
+	 nop
+	END(__copy_user_inatomic)
+
+l_exc_copy:
+	/*
+	 * Copy bytes from src until faulting load address (or until a
+	 * lb faults)
+	 *
+	 * When reached by a faulting LDFIRST/LDREST, THREAD_BUADDR($28)
+	 * may be more than a byte beyond the last address.
+	 * Hence, the lb below may get an exception.
+	 *
+	 * Assumes src < THREAD_BUADDR($28)
+	 */
+	LOAD	t0, TI_TASK($28)
+	 nop
+	LOAD	t0, THREAD_BUADDR(t0)
+1:
+EXC(	lb	t1, 0(src),	l_exc)
+	ADD	src, src, 1
+	sb	t1, 0(dst)	# can't fault -- we're copy_from_user
+	bne	src, t0, 1b
+	 ADD	dst, dst, 1
+l_exc:
+	LOAD	t0, TI_TASK($28)
+	 nop
+	LOAD	t0, THREAD_BUADDR(t0)	# t0 is just past last good address
+	 nop
+	SUB	len, AT, t0		# len number of uncopied bytes
+	jr	ra
+	 nop
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 13a42084443..f08ae71c46f 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -61,8 +61,6 @@
 
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 
-unsigned long highstart_pfn, highend_pfn;
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
@@ -261,6 +259,8 @@ EXPORT_SYMBOL(copy_from_user_page);
 
 
 #ifdef CONFIG_HIGHMEM
+unsigned long highstart_pfn, highend_pfn;
+
 pte_t *kmap_pte;
 pgprot_t kmap_prot;
 
diff --git a/include/asm-mips/sigcontext.h b/include/asm-mips/sigcontext.h
index 3c175a7e434..972947474eb 100644
--- a/include/asm-mips/sigcontext.h
+++ b/include/asm-mips/sigcontext.h
@@ -42,6 +42,7 @@ struct sigcontext {
 
 #if _MIPS_SIM == _MIPS_SIM_ABI64 || _MIPS_SIM == _MIPS_SIM_NABI32
 
+#include <linux/posix_types.h>
 /*
  * Keep this struct definition in sync with the sigcontext fragment
  * in arch/mips/tools/offset.c
@@ -53,27 +54,25 @@ struct sigcontext {
  * entries, add sc_dsp and sc_reserved for padding.  No prisoners.
  */
 struct sigcontext {
-	unsigned long	sc_regs[32];
-	unsigned long	sc_fpregs[32];
-	unsigned long	sc_mdhi;
-	unsigned long	sc_hi1;
-	unsigned long	sc_hi2;
-	unsigned long	sc_hi3;
-	unsigned long	sc_mdlo;
-	unsigned long	sc_lo1;
-	unsigned long	sc_lo2;
-	unsigned long	sc_lo3;
-	unsigned long	sc_pc;
-	unsigned int	sc_fpc_csr;
-	unsigned int	sc_used_math;
-	unsigned int	sc_dsp;
-	unsigned int	sc_reserved;
+	__u64	sc_regs[32];
+	__u64	sc_fpregs[32];
+	__u64	sc_mdhi;
+	__u64	sc_hi1;
+	__u64	sc_hi2;
+	__u64	sc_hi3;
+	__u64	sc_mdlo;
+	__u64	sc_lo1;
+	__u64	sc_lo2;
+	__u64	sc_lo3;
+	__u64	sc_pc;
+	__u32	sc_fpc_csr;
+	__u32	sc_used_math;
+	__u32	sc_dsp;
+	__u32	sc_reserved;
 };
 
 #ifdef __KERNEL__
 
-#include <linux/posix_types.h>
-
 struct sigcontext32 {
 	__u32		sc_regmask;	/* Unused */
 	__u32		sc_status;	/* Unused */
diff --git a/include/asm-mips/uaccess.h b/include/asm-mips/uaccess.h
index 3eff8d8fe28..c62c20e7b5c 100644
--- a/include/asm-mips/uaccess.h
+++ b/include/asm-mips/uaccess.h
@@ -435,8 +435,32 @@ extern size_t __copy_user(void *__to, const void *__from, size_t __n);
 	__cu_len;							\
 })
 
-#define __copy_to_user_inatomic __copy_to_user
-#define __copy_from_user_inatomic __copy_from_user
+#define __copy_to_user_inatomic(to,from,n)				\
+({									\
+	void __user *__cu_to;						\
+	const void *__cu_from;						\
+	long __cu_len;							\
+									\
+	__cu_to = (to);							\
+	__cu_from = (from);						\
+	__cu_len = (n);							\
+	__cu_len = __invoke_copy_to_user(__cu_to, __cu_from, __cu_len);	\
+	__cu_len;							\
+})
+
+#define __copy_from_user_inatomic(to,from,n)				\
+({									\
+	void *__cu_to;							\
+	const void __user *__cu_from;					\
+	long __cu_len;							\
+									\
+	__cu_to = (to);							\
+	__cu_from = (from);						\
+	__cu_len = (n);							\
+	__cu_len = __invoke_copy_from_user_inatomic(__cu_to, __cu_from,	\
+	                                            __cu_len);		\
+	__cu_len;							\
+})
 
 /*
  * copy_to_user: - Copy a block of data into user space.
@@ -490,6 +514,29 @@ extern size_t __copy_user(void *__to, const void *__from, size_t __n);
 	__cu_len_r;							\
 })
 
+#define __invoke_copy_from_user_inatomic(to,from,n)			\
+({									\
+	register void *__cu_to_r __asm__ ("$4");			\
+	register const void __user *__cu_from_r __asm__ ("$5");		\
+	register long __cu_len_r __asm__ ("$6");			\
+									\
+	__cu_to_r = (to);						\
+	__cu_from_r = (from);						\
+	__cu_len_r = (n);						\
+	__asm__ __volatile__(						\
+	".set\tnoreorder\n\t"						\
+	__MODULE_JAL(__copy_user_inatomic)				\
+	".set\tnoat\n\t"						\
+	__UA_ADDU "\t$1, %1, %2\n\t"					\
+	".set\tat\n\t"							\
+	".set\treorder"							\
+	: "+r" (__cu_to_r), "+r" (__cu_from_r), "+r" (__cu_len_r)	\
+	:								\
+	: "$8", "$9", "$10", "$11", "$12", "$15", "$24", "$31",		\
+	  "memory");							\
+	__cu_len_r;							\
+})
+
 /*
  * __copy_from_user: - Copy a block of data from user space, with less checking.
  * @to:   Destination address, in kernel space.