Merge branch 'master' of /home/davem/src/GIT/linux-2.6/

Conflicts:
	drivers/firmware/iscsi_ibft.c

73 files changed, 1713 insertions, 1070 deletions

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index c802352..0d8cd9b 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -54,6 +54,9 @@ config GENERIC_BUG
 	depends on BUG
 	default y
 
+config GENERIC_BUG_RELATIVE_POINTERS
+	def_bool y
+
 config NO_IOMEM
 	def_bool y
 
@@ -87,6 +90,7 @@ config S390
 	select HAVE_SYSCALL_TRACEPOINTS
 	select HAVE_DYNAMIC_FTRACE
 	select HAVE_FUNCTION_GRAPH_TRACER
+	select HAVE_REGS_AND_STACK_ACCESS_API
 	select HAVE_DEFAULT_NO_SPIN_MUTEXES
 	select HAVE_OPROFILE
 	select HAVE_KPROBES
@@ -95,6 +99,9 @@ config S390
 	select HAVE_ARCH_TRACEHOOK
 	select INIT_ALL_POSSIBLE
 	select HAVE_PERF_EVENTS
+	select HAVE_KERNEL_GZIP
+	select HAVE_KERNEL_BZIP2
+	select HAVE_KERNEL_LZMA
 	select ARCH_INLINE_SPIN_TRYLOCK
 	select ARCH_INLINE_SPIN_TRYLOCK_BH
 	select ARCH_INLINE_SPIN_LOCK
diff --git a/arch/s390/Kconfig.debug b/arch/s390/Kconfig.debug
index 2283933..45e0c61 100644
--- a/arch/s390/Kconfig.debug
+++ b/arch/s390/Kconfig.debug
@@ -6,4 +6,17 @@ config TRACE_IRQFLAGS_SUPPORT
 
 source "lib/Kconfig.debug"
 
+config DEBUG_STRICT_USER_COPY_CHECKS
+	bool "Strict user copy size checks"
+	---help---
+	  Enabling this option turns a certain set of sanity checks for user
+	  copy operations into compile time warnings.
+
+	  The copy_from_user() etc checks are there to help test if there
+	  are sufficient security checks on the length argument of
+	  the copy operation, by having gcc prove that the argument is
+	  within bounds.
+
+	  If unsure, or if you run an older (pre 4.4) gcc, say N.
+
 endmenu
diff --git a/arch/s390/Makefile b/arch/s390/Makefile
index fc8fb20..0da1074 100644
--- a/arch/s390/Makefile
+++ b/arch/s390/Makefile
@@ -14,6 +14,7 @@
 #
 
 ifndef CONFIG_64BIT
+LD_BFD		:= elf32-s390
 LDFLAGS		:= -m elf_s390
 KBUILD_CFLAGS	+= -m31
 KBUILD_AFLAGS	+= -m31
@@ -21,6 +22,7 @@ UTS_MACHINE	:= s390
 STACK_SIZE	:= 8192
 CHECKFLAGS	+= -D__s390__ -msize-long
 else
+LD_BFD		:= elf64-s390
 LDFLAGS		:= -m elf64_s390
 MODFLAGS	+= -fpic -D__PIC__
 KBUILD_CFLAGS	+= -m64
@@ -30,6 +32,8 @@ STACK_SIZE	:= 16384
 CHECKFLAGS	+= -D__s390__ -D__s390x__
 endif
 
+export LD_BFD
+
 cflags-$(CONFIG_MARCH_G5)   += $(call cc-option,-march=g5)
 cflags-$(CONFIG_MARCH_Z900) += $(call cc-option,-march=z900)
 cflags-$(CONFIG_MARCH_Z990) += $(call cc-option,-march=z990)
@@ -85,7 +89,9 @@ KBUILD_AFLAGS	+= $(aflags-y)
 OBJCOPYFLAGS	:= -O binary
 LDFLAGS_vmlinux := -e start
 
-head-y		:= arch/s390/kernel/head.o arch/s390/kernel/init_task.o
+head-y		:= arch/s390/kernel/head.o
+head-y		+= arch/s390/kernel/$(if $(CONFIG_64BIT),head64.o,head31.o)
+head-y		+= arch/s390/kernel/init_task.o
 
 core-y		+= arch/s390/mm/ arch/s390/kernel/ arch/s390/crypto/ \
 		   arch/s390/appldata/ arch/s390/hypfs/ arch/s390/kvm/
@@ -99,12 +105,12 @@ drivers-$(CONFIG_OPROFILE)	+= arch/s390/oprofile/
 
 boot		:= arch/s390/boot
 
-all: image
+all: image bzImage
 
 install: vmlinux
 	$(Q)$(MAKE) $(build)=$(boot) $@
 
-image: vmlinux
+image bzImage: vmlinux
 	$(Q)$(MAKE) $(build)=$(boot) $(boot)/$@
 
 zfcpdump:
@@ -116,4 +122,5 @@ archclean:
 # Don't use tabs in echo arguments
 define archhelp
   echo  '* image           - Kernel image for IPL ($(boot)/image)'
+  echo	'* bzImage         - Compressed kernel image for IPL ($(boot)/bzImage)'
 endef
diff --git a/arch/s390/boot/Makefile b/arch/s390/boot/Makefile
index 4d97eef..8800cf09 100644
--- a/arch/s390/boot/Makefile
+++ b/arch/s390/boot/Makefile
@@ -9,10 +9,18 @@ COMPILE_VERSION := __linux_compile_version_id__`hostname |  \
 EXTRA_CFLAGS  := -DCOMPILE_VERSION=$(COMPILE_VERSION) -gstabs -I.
 
 targets := image
+targets += bzImage
+subdir- := compressed
 
 $(obj)/image: vmlinux FORCE
 	$(call if_changed,objcopy)
 
+$(obj)/bzImage: $(obj)/compressed/vmlinux FORCE
+	$(call if_changed,objcopy)
+
+$(obj)/compressed/vmlinux: FORCE
+	$(Q)$(MAKE) $(build)=$(obj)/compressed $@
+
 install: $(CONFIGURE) $(obj)/image
 	sh -x  $(srctree)/$(obj)/install.sh $(KERNELRELEASE) $(obj)/image \
 	      System.map Kerntypes "$(INSTALL_PATH)"
diff --git a/arch/s390/boot/compressed/Makefile b/arch/s390/boot/compressed/Makefile
new file mode 100644
index 0000000..6e4a67a
--- /dev/null
+++ b/arch/s390/boot/compressed/Makefile
@@ -0,0 +1,60 @@
+#
+# linux/arch/s390/boot/compressed/Makefile
+#
+# create a compressed vmlinux image from the original vmlinux
+#
+
+BITS := $(if $(CONFIG_64BIT),64,31)
+
+targets	:= vmlinux.lds vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 \
+	   vmlinux.bin.lzma misc.o piggy.o sizes.h head$(BITS).o
+
+KBUILD_CFLAGS := -m$(BITS) -D__KERNEL__ $(LINUX_INCLUDE) -O2
+KBUILD_CFLAGS += $(cflags-y)
+KBUILD_CFLAGS += $(call cc-option,-mpacked-stack)
+KBUILD_CFLAGS += $(call cc-option,-ffreestanding)
+
+GCOV_PROFILE := n
+
+OBJECTS := $(addprefix $(objtree)/arch/s390/kernel/, head.o sclp.o ebcdic.o)
+OBJECTS += $(obj)/head$(BITS).o $(obj)/misc.o $(obj)/piggy.o
+
+LDFLAGS_vmlinux := --oformat $(LD_BFD) -e startup -T
+$(obj)/vmlinux: $(obj)/vmlinux.lds $(OBJECTS)
+	$(call if_changed,ld)
+	@:
+
+sed-sizes := -e 's/^\([0-9a-fA-F]*\) . \(__bss_start\|_end\)$$/\#define SZ\2 0x\1/p'
+
+quiet_cmd_sizes = GEN $@
+      cmd_sizes = $(NM) $< | sed -n $(sed-sizes) > $@
+
+$(obj)/sizes.h: vmlinux
+	$(call if_changed,sizes)
+
+AFLAGS_head$(BITS).o += -I$(obj)
+$(obj)/head$(BITS).o: $(obj)/sizes.h
+
+CFLAGS_misc.o += -I$(obj)
+$(obj)/misc.o: $(obj)/sizes.h
+
+OBJCOPYFLAGS_vmlinux.bin :=  -R .comment -S
+$(obj)/vmlinux.bin: vmlinux
+	$(call if_changed,objcopy)
+
+vmlinux.bin.all-y := $(obj)/vmlinux.bin
+
+suffix-$(CONFIG_KERNEL_GZIP)  := gz
+suffix-$(CONFIG_KERNEL_BZIP2) := bz2
+suffix-$(CONFIG_KERNEL_LZMA)  := lzma
+
+$(obj)/vmlinux.bin.gz: $(vmlinux.bin.all-y)
+	$(call if_changed,gzip)
+$(obj)/vmlinux.bin.bz2: $(vmlinux.bin.all-y)
+	$(call if_changed,bzip2)
+$(obj)/vmlinux.bin.lzma: $(vmlinux.bin.all-y)
+	$(call if_changed,lzma)
+
+LDFLAGS_piggy.o := -r --format binary --oformat $(LD_BFD) -T
+$(obj)/piggy.o: $(obj)/vmlinux.scr $(obj)/vmlinux.bin.$(suffix-y)
+	$(call if_changed,ld)
diff --git a/arch/s390/boot/compressed/head31.S b/arch/s390/boot/compressed/head31.S
new file mode 100644
index 0000000..2a5523a
--- /dev/null
+++ b/arch/s390/boot/compressed/head31.S
@@ -0,0 +1,51 @@
+/*
+ * Startup glue code to uncompress the kernel
+ *
+ * Copyright IBM Corp. 2010
+ *
+ *   Author(s):	Martin Schwidefsky <schwidefsky@de.ibm.com>
+ */
+
+#include <linux/init.h>
+#include <asm/asm-offsets.h>
+#include <asm/thread_info.h>
+#include <asm/page.h>
+#include "sizes.h"
+
+__HEAD
+	.globl	startup_continue
+startup_continue:
+	basr	%r13,0			# get base
+.LPG1:
+	# setup stack
+	l	%r15,.Lstack-.LPG1(%r13)
+	ahi	%r15,-96
+	l	%r1,.Ldecompress-.LPG1(%r13)
+	basr	%r14,%r1
+	# setup registers for memory mover & branch to target
+	lr	%r4,%r2
+	l	%r2,.Loffset-.LPG1(%r13)
+	la	%r4,0(%r2,%r4)
+	l	%r3,.Lmvsize-.LPG1(%r13)
+	lr	%r5,%r3
+	# move the memory mover someplace safe
+	la	%r1,0x200
+	mvc	0(mover_end-mover,%r1),mover-.LPG1(%r13)
+	# decompress image is started at 0x11000
+	lr	%r6,%r2
+	br	%r1
+mover:
+	mvcle	%r2,%r4,0
+	jo	mover
+	br	%r6
+mover_end:
+
+	.align	8
+.Lstack:
+	.long	0x8000 + (1<<(PAGE_SHIFT+THREAD_ORDER))
+.Ldecompress:
+	.long	decompress_kernel
+.Loffset:
+	.long	0x11000
+.Lmvsize:
+	.long	SZ__bss_start
diff --git a/arch/s390/boot/compressed/head64.S b/arch/s390/boot/compressed/head64.S
new file mode 100644
index 0000000..2982cb1
--- /dev/null
+++ b/arch/s390/boot/compressed/head64.S
@@ -0,0 +1,48 @@
+/*
+ * Startup glue code to uncompress the kernel
+ *
+ * Copyright IBM Corp. 2010
+ *
+ *   Author(s):	Martin Schwidefsky <schwidefsky@de.ibm.com>
+ */
+
+#include <linux/init.h>
+#include <asm/asm-offsets.h>
+#include <asm/thread_info.h>
+#include <asm/page.h>
+#include "sizes.h"
+
+__HEAD
+	.globl	startup_continue
+startup_continue:
+	basr	%r13,0			# get base
+.LPG1:
+	# setup stack
+	lg	%r15,.Lstack-.LPG1(%r13)
+	aghi	%r15,-160
+	brasl	%r14,decompress_kernel
+	# setup registers for memory mover & branch to target
+	lgr	%r4,%r2
+	lg	%r2,.Loffset-.LPG1(%r13)
+	la	%r4,0(%r2,%r4)
+	lg	%r3,.Lmvsize-.LPG1(%r13)
+	lgr	%r5,%r3
+	# move the memory mover someplace safe
+	la	%r1,0x200
+	mvc	0(mover_end-mover,%r1),mover-.LPG1(%r13)
+	# decompress image is started at 0x11000
+	lgr	%r6,%r2
+	br	%r1
+mover:
+	mvcle	%r2,%r4,0
+	jo	mover
+	br	%r6
+mover_end:
+
+	.align	8
+.Lstack:
+	.quad	0x8000 + (1<<(PAGE_SHIFT+THREAD_ORDER))
+.Loffset:
+	.quad	0x11000
+.Lmvsize:
+	.quad	SZ__bss_start
diff --git a/arch/s390/boot/compressed/misc.c b/arch/s390/boot/compressed/misc.c
new file mode 100644
index 0000000..a97d695
--- /dev/null
+++ b/arch/s390/boot/compressed/misc.c
@@ -0,0 +1,158 @@
+/*
+ * Definitions and wrapper functions for kernel decompressor
+ *
+ * Copyright IBM Corp. 2010
+ *
+ * Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>
+ */
+
+#include <asm/uaccess.h>
+#include <asm/page.h>
+#include <asm/ipl.h>
+#include "sizes.h"
+
+/*
+ * gzip declarations
+ */
+#define STATIC static
+
+#undef memset
+#undef memcpy
+#undef memmove
+#define memzero(s, n) memset((s), 0, (n))
+
+/* Symbols defined by linker scripts */
+extern char input_data[];
+extern int input_len;
+extern int _text;
+extern int _end;
+
+static void error(char *m);
+
+static unsigned long free_mem_ptr;
+static unsigned long free_mem_end_ptr;
+
+#ifdef CONFIG_HAVE_KERNEL_BZIP2
+#define HEAP_SIZE	0x400000
+#else
+#define HEAP_SIZE	0x10000
+#endif
+
+#ifdef CONFIG_KERNEL_GZIP
+#include "../../../../lib/decompress_inflate.c"
+#endif
+
+#ifdef CONFIG_KERNEL_BZIP2
+#include "../../../../lib/decompress_bunzip2.c"
+#endif
+
+#ifdef CONFIG_KERNEL_LZMA
+#include "../../../../lib/decompress_unlzma.c"
+#endif
+
+extern _sclp_print_early(const char *);
+
+int puts(const char *s)
+{
+	_sclp_print_early(s);
+	return 0;
+}
+
+void *memset(void *s, int c, size_t n)
+{
+	char *xs;
+
+	if (c == 0)
+		return __builtin_memset(s, 0, n);
+
+	xs = (char *) s;
+	if (n > 0)
+		do {
+			*xs++ = c;
+		} while (--n > 0);
+	return s;
+}
+
+void *memcpy(void *__dest, __const void *__src, size_t __n)
+{
+	return __builtin_memcpy(__dest, __src, __n);
+}
+
+void *memmove(void *__dest, __const void *__src, size_t __n)
+{
+	char *d;
+	const char *s;
+
+	if (__dest <= __src)
+		return __builtin_memcpy(__dest, __src, __n);
+	d = __dest + __n;
+	s = __src + __n;
+	while (__n--)
+		*--d = *--s;
+	return __dest;
+}
+
+static void error(char *x)
+{
+	unsigned long long psw = 0x000a0000deadbeefULL;
+
+	puts("\n\n");
+	puts(x);
+	puts("\n\n -- System halted");
+
+	asm volatile("lpsw %0" : : "Q" (psw));
+}
+
+/*
+ * Safe guard the ipl parameter block against a memory area that will be
+ * overwritten. The validity check for the ipl parameter block is complex
+ * (see cio_get_iplinfo and ipl_save_parameters) but if the pointer to
+ * the ipl parameter block intersects with the passed memory area we can
+ * safely assume that we can read from that memory. In that case just copy
+ * the memory to IPL_PARMBLOCK_ORIGIN even if there is no ipl parameter
+ * block.
+ */
+static void check_ipl_parmblock(void *start, unsigned long size)
+{
+	void *src, *dst;
+
+	src = (void *)(unsigned long) S390_lowcore.ipl_parmblock_ptr;
+	if (src + PAGE_SIZE <= start || src >= start + size)
+		return;
+	dst = (void *) IPL_PARMBLOCK_ORIGIN;
+	memmove(dst, src, PAGE_SIZE);
+	S390_lowcore.ipl_parmblock_ptr = IPL_PARMBLOCK_ORIGIN;
+}
+
+unsigned long decompress_kernel(void)
+{
+	unsigned long output_addr;
+	unsigned char *output;
+
+	free_mem_ptr = (unsigned long)&_end;
+	free_mem_end_ptr = free_mem_ptr + HEAP_SIZE;
+	output = (unsigned char *) ((free_mem_end_ptr + 4095UL) & -4096UL);
+
+	check_ipl_parmblock((void *) 0, (unsigned long) output + SZ__bss_start);
+
+#ifdef CONFIG_BLK_DEV_INITRD
+	/*
+	 * Move the initrd right behind the end of the decompressed
+	 * kernel image.
+	 */
+	if (INITRD_START && INITRD_SIZE &&
+	    INITRD_START < (unsigned long) output + SZ__bss_start) {
+		check_ipl_parmblock(output + SZ__bss_start,
+				    INITRD_START + INITRD_SIZE);
+		memmove(output + SZ__bss_start,
+			(void *) INITRD_START, INITRD_SIZE);
+		INITRD_START = (unsigned long) output + SZ__bss_start;
+	}
+#endif
+
+	puts("Uncompressing Linux... ");
+	decompress(input_data, input_len, NULL, NULL, output, NULL, error);
+	puts("Ok, booting the kernel.\n");
+	return (unsigned long) output;
+}
+
diff --git a/arch/s390/boot/compressed/vmlinux.lds.S b/arch/s390/boot/compressed/vmlinux.lds.S
new file mode 100644
index 0000000..d80f79d
--- /dev/null
+++ b/arch/s390/boot/compressed/vmlinux.lds.S
@@ -0,0 +1,55 @@
+#include <asm-generic/vmlinux.lds.h>
+
+#ifdef CONFIG_64BIT
+OUTPUT_FORMAT("elf64-s390", "elf64-s390", "elf64-s390")
+OUTPUT_ARCH(s390:64-bit)
+#else
+OUTPUT_FORMAT("elf32-s390", "elf32-s390", "elf32-s390")
+OUTPUT_ARCH(s390)
+#endif
+
+ENTRY(startup)
+
+SECTIONS
+{
+	/* Be careful parts of head_64.S assume startup_32 is at
+	 * address 0.
+	 */
+	. = 0;
+	.head.text : {
+		_head = . ;
+		HEAD_TEXT
+		_ehead = . ;
+	}
+	.rodata.compressed : {
+		*(.rodata.compressed)
+	}
+	.text :	{
+		_text = .;	/* Text */
+		*(.text)
+		*(.text.*)
+		_etext = . ;
+	}
+	.rodata : {
+		_rodata = . ;
+		*(.rodata)	 /* read-only data */
+		*(.rodata.*)
+		_erodata = . ;
+	}
+	.data :	{
+		_data = . ;
+		*(.data)
+		*(.data.*)
+		_edata = . ;
+	}
+	. = ALIGN(256);
+	.bss : {
+		_bss = . ;
+		*(.bss)
+		*(.bss.*)
+		*(COMMON)
+		. = ALIGN(8);	/* For convenience during zeroing */
+		_ebss = .;
+	}
+	_end = .;
+}
diff --git a/arch/s390/boot/compressed/vmlinux.scr b/arch/s390/boot/compressed/vmlinux.scr
new file mode 100644
index 0000000..f02382a
--- /dev/null
+++ b/arch/s390/boot/compressed/vmlinux.scr
@@ -0,0 +1,10 @@
+SECTIONS
+{
+  .rodata.compressed : {
+	input_len = .;
+	LONG(input_data_end - input_data) input_data = .;
+	*(.data)
+	output_len = . - 4;
+	input_data_end = .;
+	}
+}
diff --git a/arch/s390/crypto/aes_s390.c b/arch/s390/crypto/aes_s390.c
index 6be4503..58f4673 100644
--- a/arch/s390/crypto/aes_s390.c
+++ b/arch/s390/crypto/aes_s390.c
@@ -78,14 +78,14 @@ static int setkey_fallback_cip(struct crypto_tfm *tfm, const u8 *in_key,
 	struct s390_aes_ctx *sctx = crypto_tfm_ctx(tfm);
 	int ret;
 
-	sctx->fallback.blk->base.crt_flags &= ~CRYPTO_TFM_REQ_MASK;
-	sctx->fallback.blk->base.crt_flags |= (tfm->crt_flags &
+	sctx->fallback.cip->base.crt_flags &= ~CRYPTO_TFM_REQ_MASK;
+	sctx->fallback.cip->base.crt_flags |= (tfm->crt_flags &
 			CRYPTO_TFM_REQ_MASK);
 
 	ret = crypto_cipher_setkey(sctx->fallback.cip, in_key, key_len);
 	if (ret) {
 		tfm->crt_flags &= ~CRYPTO_TFM_RES_MASK;
-		tfm->crt_flags |= (sctx->fallback.blk->base.crt_flags &
+		tfm->crt_flags |= (sctx->fallback.cip->base.crt_flags &
 				CRYPTO_TFM_RES_MASK);
 	}
 	return ret;
diff --git a/arch/s390/defconfig b/arch/s390/defconfig
index b416aa1..7ae71cc 100644
--- a/arch/s390/defconfig
+++ b/arch/s390/defconfig
@@ -36,6 +36,13 @@ CONFIG_LOCK_KERNEL=y
 CONFIG_INIT_ENV_ARG_LIMIT=32
 CONFIG_LOCALVERSION=""
 CONFIG_LOCALVERSION_AUTO=y
+CONFIG_HAVE_KERNEL_GZIP=y
+CONFIG_HAVE_KERNEL_BZIP2=y
+CONFIG_HAVE_KERNEL_LZMA=y
+CONFIG_KERNEL_GZIP=y
+# CONFIG_KERNEL_BZIP2 is not set
+# CONFIG_KERNEL_LZMA is not set
+# CONFIG_KERNEL_LZO is not set
 CONFIG_SWAP=y
 CONFIG_SYSVIPC=y
 CONFIG_SYSVIPC_SYSCTL=y
diff --git a/arch/s390/hypfs/hypfs_diag.c b/arch/s390/hypfs/hypfs_diag.c
index 2b92d50..87cf523 100644
--- a/arch/s390/hypfs/hypfs_diag.c
+++ b/arch/s390/hypfs/hypfs_diag.c
@@ -488,7 +488,7 @@ out:
 
 static int diag224(void *ptr)
 {
-	int rc = -ENOTSUPP;
+	int rc = -EOPNOTSUPP;
 
 	asm volatile(
 		"	diag	%1,%2,0x224\n"
@@ -507,7 +507,7 @@ static int diag224_get_name_table(void)
 		return -ENOMEM;
 	if (diag224(diag224_cpu_names)) {
 		kfree(diag224_cpu_names);
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	}
 	EBCASC(diag224_cpu_names + 16, (*diag224_cpu_names + 1) * 16);
 	return 0;
diff --git a/arch/s390/include/asm/atomic.h b/arch/s390/include/asm/atomic.h
index 2a113d6..451bfbb 100644
--- a/arch/s390/include/asm/atomic.h
+++ b/arch/s390/include/asm/atomic.h
@@ -18,8 +18,6 @@
 
 #define ATOMIC_INIT(i)  { (i) }
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 #define __CS_LOOP(ptr, op_val, op_string) ({				\
 	int old_val, new_val;						\
 	asm volatile(							\
@@ -35,26 +33,6 @@
 	new_val;							\
 })
 
-#else /* __GNUC__ */
-
-#define __CS_LOOP(ptr, op_val, op_string) ({				\
-	int old_val, new_val;						\
-	asm volatile(							\
-		"	l	%0,0(%3)\n"				\
-		"0:	lr	%1,%0\n"				\
-		op_string "	%1,%4\n"				\
-		"	cs	%0,%1,0(%3)\n"				\
-		"	jl	0b"					\
-		: "=&d" (old_val), "=&d" (new_val),			\
-		  "=m" (((atomic_t *)(ptr))->counter)			\
-		: "a" (ptr), "d" (op_val),				\
-		  "m" (((atomic_t *)(ptr))->counter)			\
-		: "cc", "memory");					\
-	new_val;							\
-})
-
-#endif /* __GNUC__ */
-
 static inline int atomic_read(const atomic_t *v)
 {
 	barrier();
@@ -101,19 +79,11 @@ static inline void atomic_set_mask(unsigned long mask, atomic_t *v)
 
 static inline int atomic_cmpxchg(atomic_t *v, int old, int new)
 {
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 	asm volatile(
 		"	cs	%0,%2,%1"
 		: "+d" (old), "=Q" (v->counter)
 		: "d" (new), "Q" (v->counter)
 		: "cc", "memory");
-#else /* __GNUC__ */
-	asm volatile(
-		"	cs	%0,%3,0(%2)"
-		: "+d" (old), "=m" (v->counter)
-		: "a" (v), "d" (new), "m" (v->counter)
-		: "cc", "memory");
-#endif /* __GNUC__ */
 	return old;
 }
 
@@ -140,8 +110,6 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 
 #ifdef CONFIG_64BIT
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 #define __CSG_LOOP(ptr, op_val, op_string) ({				\
 	long long old_val, new_val;					\
 	asm volatile(							\
@@ -157,26 +125,6 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 	new_val;							\
 })
 
-#else /* __GNUC__ */
-
-#define __CSG_LOOP(ptr, op_val, op_string) ({				\
-	long long old_val, new_val;					\
-	asm volatile(							\
-		"	lg	%0,0(%3)\n"				\
-		"0:	lgr	%1,%0\n"				\
-		op_string "	%1,%4\n"				\
-		"	csg	%0,%1,0(%3)\n"				\
-		"	jl	0b"					\
-		: "=&d" (old_val), "=&d" (new_val),			\
-		  "=m" (((atomic_t *)(ptr))->counter)			\
-		: "a" (ptr), "d" (op_val),				\
-		  "m" (((atomic_t *)(ptr))->counter)			\
-		: "cc", "memory");					\
-	new_val;							\
-})
-
-#endif /* __GNUC__ */
-
 static inline long long atomic64_read(const atomic64_t *v)
 {
 	barrier();
@@ -214,19 +162,11 @@ static inline void atomic64_set_mask(unsigned long mask, atomic64_t *v)
 static inline long long atomic64_cmpxchg(atomic64_t *v,
 					     long long old, long long new)
 {
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 	asm volatile(
 		"	csg	%0,%2,%1"
 		: "+d" (old), "=Q" (v->counter)
 		: "d" (new), "Q" (v->counter)
 		: "cc", "memory");
-#else /* __GNUC__ */
-	asm volatile(
-		"	csg	%0,%3,0(%2)"
-		: "+d" (old), "=m" (v->counter)
-		: "a" (v), "d" (new), "m" (v->counter)
-		: "cc", "memory");
-#endif /* __GNUC__ */
 	return old;
 }
 
@@ -243,10 +183,8 @@ static inline long long atomic64_read(const atomic64_t *v)
 	register_pair rp;
 
 	asm volatile(
-		"	lm	%0,%N0,0(%1)"
-		: "=&d" (rp)
-		: "a" (&v->counter), "m" (v->counter)
-		);
+		"	lm	%0,%N0,%1"
+		: "=&d" (rp) : "Q" (v->counter)	);
 	return rp.pair;
 }
 
@@ -255,10 +193,8 @@ static inline void atomic64_set(atomic64_t *v, long long i)
 	register_pair rp = {.pair = i};
 
 	asm volatile(
-		"	stm	%1,%N1,0(%2)"
-		: "=m" (v->counter)
-		: "d" (rp), "a" (&v->counter)
-		);
+		"	stm	%1,%N1,%0"
+		: "=Q" (v->counter) : "d" (rp) );
 }
 
 static inline long long atomic64_xchg(atomic64_t *v, long long new)
@@ -267,11 +203,11 @@ static inline long long atomic64_xchg(atomic64_t *v, long long new)
 	register_pair rp_old;
 
 	asm volatile(
-		"	lm	%0,%N0,0(%2)\n"
-		"0:	cds	%0,%3,0(%2)\n"
+		"	lm	%0,%N0,%1\n"
+		"0:	cds	%0,%2,%1\n"
 		"	jl	0b\n"
-		: "=&d" (rp_old), "+m" (v->counter)
-		: "a" (&v->counter), "d" (rp_new)
+		: "=&d" (rp_old), "=Q" (v->counter)
+		: "d" (rp_new), "Q" (v->counter)
 		: "cc");
 	return rp_old.pair;
 }
@@ -283,9 +219,9 @@ static inline long long atomic64_cmpxchg(atomic64_t *v,
 	register_pair rp_new = {.pair = new};
 
 	asm volatile(
-		"	cds	%0,%3,0(%2)"
-		: "+&d" (rp_old), "+m" (v->counter)
-		: "a" (&v->counter), "d" (rp_new)
+		"	cds	%0,%2,%1"
+		: "+&d" (rp_old), "=Q" (v->counter)
+		: "d" (rp_new), "Q" (v->counter)
 		: "cc");
 	return rp_old.pair;
 }
diff --git a/arch/s390/include/asm/bitops.h b/arch/s390/include/asm/bitops.h
index b30606f..2e05972 100644
--- a/arch/s390/include/asm/bitops.h
+++ b/arch/s390/include/asm/bitops.h
@@ -71,8 +71,6 @@ extern const char _sb_findmap[];
 #define __BITOPS_AND		"nr"
 #define __BITOPS_XOR		"xr"
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 #define __BITOPS_LOOP(__old, __new, __addr, __val, __op_string)	\
 	asm volatile(						\
 		"	l	%0,%2\n"			\
@@ -85,22 +83,6 @@ extern const char _sb_findmap[];
 		: "d" (__val), "Q" (*(unsigned long *) __addr)	\
 		: "cc");
 
-#else /* __GNUC__ */
-
-#define __BITOPS_LOOP(__old, __new, __addr, __val, __op_string)	\
-	asm volatile(						\
-		"	l	%0,0(%4)\n"			\
-		"0:	lr	%1,%0\n"			\
-		__op_string "	%1,%3\n"			\
-		"	cs	%0,%1,0(%4)\n"			\
-		"	jl	0b"				\
-		: "=&d" (__old), "=&d" (__new),			\
-		  "=m" (*(unsigned long *) __addr)		\
-		: "d" (__val), "a" (__addr),			\
-		  "m" (*(unsigned long *) __addr) : "cc");
-
-#endif /* __GNUC__ */
-
 #else /* __s390x__ */
 
 #define __BITOPS_ALIGN		7
@@ -109,8 +91,6 @@ extern const char _sb_findmap[];
 #define __BITOPS_AND		"ngr"
 #define __BITOPS_XOR		"xgr"
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 #define __BITOPS_LOOP(__old, __new, __addr, __val, __op_string)	\
 	asm volatile(						\
 		"	lg	%0,%2\n"			\
@@ -123,23 +103,6 @@ extern const char _sb_findmap[];
 		: "d" (__val), "Q" (*(unsigned long *) __addr)	\
 		: "cc");
 
-#else /* __GNUC__ */
-
-#define __BITOPS_LOOP(__old, __new, __addr, __val, __op_string)	\
-	asm volatile(						\
-		"	lg	%0,0(%4)\n"			\
-		"0:	lgr	%1,%0\n"			\
-		__op_string "	%1,%3\n"			\
-		"	csg	%0,%1,0(%4)\n"			\
-		"	jl	0b"				\
-		: "=&d" (__old), "=&d" (__new),			\
-		  "=m" (*(unsigned long *) __addr)		\
-		: "d" (__val), "a" (__addr),			\
-		  "m" (*(unsigned long *) __addr) : "cc");
-
-
-#endif /* __GNUC__ */
-
 #endif /* __s390x__ */
 
 #define __BITOPS_WORDS(bits) (((bits)+__BITOPS_WORDSIZE-1)/__BITOPS_WORDSIZE)
@@ -261,9 +224,8 @@ static inline void __set_bit(unsigned long nr, volatile unsigned long *ptr)
 
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	asm volatile(
-		"	oc	0(1,%1),0(%2)"
-		: "=m" (*(char *) addr) : "a" (addr),
-		  "a" (_oi_bitmap + (nr & 7)), "m" (*(char *) addr) : "cc" );
+		"	oc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr) : "Q" (_oi_bitmap[nr & 7]) : "cc" );
 }
 
 static inline void 
@@ -290,9 +252,8 @@ __clear_bit(unsigned long nr, volatile unsigned long *ptr)
 
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	asm volatile(
-		"	nc	0(1,%1),0(%2)"
-		: "=m" (*(char *) addr)	: "a" (addr),
-		  "a" (_ni_bitmap + (nr & 7)), "m" (*(char *) addr) : "cc");
+		"	nc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr) : "Q" (_ni_bitmap[nr & 7]) : "cc" );
 }
 
 static inline void 
@@ -318,9 +279,8 @@ static inline void __change_bit(unsigned long nr, volatile unsigned long *ptr)
 
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	asm volatile(
-		"	xc	0(1,%1),0(%2)"
-		:  "=m" (*(char *) addr) : "a" (addr),
-		   "a" (_oi_bitmap + (nr & 7)), "m" (*(char *) addr) : "cc" );
+		"	xc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr) : "Q" (_oi_bitmap[nr & 7]) : "cc" );
 }
 
 static inline void 
@@ -349,10 +309,9 @@ test_and_set_bit_simple(unsigned long nr, volatile unsigned long *ptr)
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	ch = *(unsigned char *) addr;
 	asm volatile(
-		"	oc	0(1,%1),0(%2)"
-		: "=m" (*(char *) addr)
-		: "a" (addr), "a" (_oi_bitmap + (nr & 7)),
-		  "m" (*(char *) addr) : "cc", "memory");
+		"	oc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr)	: "Q" (_oi_bitmap[nr & 7])
+		: "cc", "memory");
 	return (ch >> (nr & 7)) & 1;
 }
 #define __test_and_set_bit(X,Y)		test_and_set_bit_simple(X,Y)
@@ -369,10 +328,9 @@ test_and_clear_bit_simple(unsigned long nr, volatile unsigned long *ptr)
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	ch = *(unsigned char *) addr;
 	asm volatile(
-		"	nc	0(1,%1),0(%2)"
-		: "=m" (*(char *) addr)
-		: "a" (addr), "a" (_ni_bitmap + (nr & 7)),
-		  "m" (*(char *) addr) : "cc", "memory");
+		"	nc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr)	: "Q" (_ni_bitmap[nr & 7])
+		: "cc", "memory");
 	return (ch >> (nr & 7)) & 1;
 }
 #define __test_and_clear_bit(X,Y)	test_and_clear_bit_simple(X,Y)
@@ -389,10 +347,9 @@ test_and_change_bit_simple(unsigned long nr, volatile unsigned long *ptr)
 	addr = (unsigned long) ptr + ((nr ^ (__BITOPS_WORDSIZE - 8)) >> 3);
 	ch = *(unsigned char *) addr;
 	asm volatile(
-		"	xc	0(1,%1),0(%2)"
-		: "=m" (*(char *) addr)
-		: "a" (addr), "a" (_oi_bitmap + (nr & 7)),
-		  "m" (*(char *) addr) : "cc", "memory");
+		"	xc	%O0(1,%R0),%1"
+		: "=Q" (*(char *) addr)	: "Q" (_oi_bitmap[nr & 7])
+		: "cc", "memory");
 	return (ch >> (nr & 7)) & 1;
 }
 #define __test_and_change_bit(X,Y)	test_and_change_bit_simple(X,Y)
@@ -591,11 +548,11 @@ static inline unsigned long __load_ulong_le(const unsigned long *p,
 	p = (unsigned long *)((unsigned long) p + offset);
 #ifndef __s390x__
 	asm volatile(
-		"	ic	%0,0(%1)\n"
-		"	icm	%0,2,1(%1)\n"
-		"	icm	%0,4,2(%1)\n"
-		"	icm	%0,8,3(%1)"
-		: "=&d" (word) : "a" (p), "m" (*p) : "cc");
+		"	ic	%0,%O1(%R1)\n"
+		"	icm	%0,2,%O1+1(%R1)\n"
+		"	icm	%0,4,%O1+2(%R1)\n"
+		"	icm	%0,8,%O1+3(%R1)"
+		: "=&d" (word) : "Q" (*p) : "cc");
 #else
 	asm volatile(
 		"	lrvg	%0,%1"
diff --git a/arch/s390/include/asm/bug.h b/arch/s390/include/asm/bug.h
index b1066b9..9beeb9d 100644
--- a/arch/s390/include/asm/bug.h
+++ b/arch/s390/include/asm/bug.h
@@ -5,12 +5,6 @@
 
 #ifdef CONFIG_BUG
 
-#ifdef CONFIG_64BIT
-#define S390_LONG ".quad"
-#else
-#define S390_LONG ".long"
-#endif
-
 #ifdef CONFIG_DEBUG_BUGVERBOSE
 
 #define __EMIT_BUG(x) do {					\
@@ -21,7 +15,7 @@
 		"2:	.asciz	\""__FILE__"\"\n"		\
 		".previous\n"					\
 		".section __bug_table,\"a\"\n"			\
-		"3:\t"	S390_LONG "\t1b,2b\n"			\
+		"3:	.long	1b-3b,2b-3b\n"			\
 		"	.short	%0,%1\n"			\
 		"	.org	3b+%2\n"			\
 		".previous\n"					\
@@ -37,7 +31,7 @@
 		"0:	j	0b+2\n"			\
 		"1:\n"					\
 		".section __bug_table,\"a\"\n"		\
-		"2:\t"	S390_LONG "\t1b\n"		\
+		"2:	.long	1b-2b\n"		\
 		"	.short	%0\n"			\
 		"	.org	2b+%1\n"		\
 		".previous\n"				\
diff --git a/arch/s390/include/asm/crw.h b/arch/s390/include/asm/crw.h
index 2185a6d..749a97e 100644
--- a/arch/s390/include/asm/crw.h
+++ b/arch/s390/include/asm/crw.h
@@ -32,6 +32,7 @@ typedef void (*crw_handler_t)(struct crw *, struct crw *, int);
 extern int crw_register_handler(int rsc, crw_handler_t handler);
 extern void crw_unregister_handler(int rsc);
 extern void crw_handle_channel_report(void);
+void crw_wait_for_channel_report(void);
 
 #define NR_RSCS 16
 
diff --git a/arch/s390/include/asm/etr.h b/arch/s390/include/asm/etr.h
index 80ef58c..538e1b3 100644
--- a/arch/s390/include/asm/etr.h
+++ b/arch/s390/include/asm/etr.h
@@ -145,11 +145,11 @@ static inline int etr_setr(struct etr_eacr *ctrl)
 	int rc = -ENOSYS;
 
 	asm volatile(
-		"	.insn	s,0xb2160000,0(%2)\n"
+		"	.insn	s,0xb2160000,%1\n"
 		"0:	la	%0,0\n"
 		"1:\n"
 		EX_TABLE(0b,1b)
-		: "+d" (rc) : "m" (*ctrl), "a" (ctrl));
+		: "+d" (rc) : "Q" (*ctrl));
 	return rc;
 }
 
@@ -159,11 +159,11 @@ static inline int etr_stetr(struct etr_aib *aib)
 	int rc = -ENOSYS;
 
 	asm volatile(
-		"	.insn	s,0xb2170000,0(%2)\n"
+		"	.insn	s,0xb2170000,%1\n"
 		"0:	la	%0,0\n"
 		"1:\n"
 		EX_TABLE(0b,1b)
-		: "+d" (rc) : "m" (*aib), "a" (aib));
+		: "+d" (rc) : "Q" (*aib));
 	return rc;
 }
 
@@ -174,11 +174,11 @@ static inline int etr_steai(struct etr_aib *aib, unsigned int func)
 	int rc = -ENOSYS;
 
 	asm volatile(
-		"	.insn	s,0xb2b30000,0(%2)\n"
+		"	.insn	s,0xb2b30000,%1\n"
 		"0:	la	%0,0\n"
 		"1:\n"
 		EX_TABLE(0b,1b)
-		: "+d" (rc) : "m" (*aib), "a" (aib), "d" (reg0));
+		: "+d" (rc) : "Q" (*aib), "d" (reg0));
 	return rc;
 }
 
diff --git a/arch/s390/include/asm/irqflags.h b/arch/s390/include/asm/irqflags.h
index c2fb432..15b3ac2 100644
--- a/arch/s390/include/asm/irqflags.h
+++ b/arch/s390/include/asm/irqflags.h
@@ -8,8 +8,6 @@
 
 #include <linux/types.h>
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 /* store then or system mask. */
 #define __raw_local_irq_stosm(__or)					\
 ({									\
@@ -36,40 +34,6 @@
 	asm volatile("ssm   %0" : : "Q" (__mask) : "memory");		\
 })
 
-#else /* __GNUC__ */
-
-/* store then or system mask. */
-#define __raw_local_irq_stosm(__or)					\
-({									\
-	unsigned long __mask;						\
-	asm volatile(							\
-		"	stosm	0(%1),%2"				\
-		: "=m" (__mask)						\
-		: "a" (&__mask), "i" (__or) : "memory");		\
-	__mask;								\
-})
-
-/* store then and system mask. */
-#define __raw_local_irq_stnsm(__and)					\
-({									\
-	unsigned long __mask;						\
-	asm volatile(							\
-		"	stnsm	0(%1),%2"				\
-		: "=m" (__mask)						\
-		: "a" (&__mask), "i" (__and) : "memory");		\
-	__mask;								\
-})
-
-/* set system mask. */
-#define __raw_local_irq_ssm(__mask)					\
-({									\
-	asm volatile(							\
-		"	ssm	0(%0)"					\
-		: : "a" (&__mask), "m" (__mask) : "memory");		\
-})
-
-#endif /* __GNUC__ */
-
 /* interrupt control.. */
 static inline unsigned long raw_local_irq_enable(void)
 {
diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h
index c25dfac..05527c0 100644
--- a/arch/s390/include/asm/lowcore.h
+++ b/arch/s390/include/asm/lowcore.h
@@ -1,141 +1,16 @@
 /*
- *  include/asm-s390/lowcore.h
- *
- *  S390 version
- *    Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
- *    Author(s): Hartmut Penner (hp@de.ibm.com),
- *               Martin Schwidefsky (schwidefsky@de.ibm.com),
- *               Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com)
+ *    Copyright IBM Corp. 1999,2010
+ *    Author(s): Hartmut Penner <hp@de.ibm.com>,
+ *		 Martin Schwidefsky <schwidefsky@de.ibm.com>,
+ *		 Denis Joseph Barrow,
  */
 
 #ifndef _ASM_S390_LOWCORE_H
 #define _ASM_S390_LOWCORE_H
 
-#define __LC_IPL_PARMBLOCK_PTR		0x0014
-#define __LC_EXT_PARAMS			0x0080
-#define __LC_CPU_ADDRESS		0x0084
-#define __LC_EXT_INT_CODE		0x0086
-
-#define __LC_SVC_ILC			0x0088
-#define __LC_SVC_INT_CODE		0x008a
-#define __LC_PGM_ILC			0x008c
-#define __LC_PGM_INT_CODE		0x008e
-
-#define __LC_PER_ATMID			0x0096
-#define __LC_PER_ADDRESS		0x0098
-#define __LC_PER_ACCESS_ID		0x00a1
-#define __LC_AR_MODE_ID			0x00a3
-
-#define __LC_SUBCHANNEL_ID		0x00b8
-#define __LC_SUBCHANNEL_NR		0x00ba
-#define __LC_IO_INT_PARM		0x00bc
-#define __LC_IO_INT_WORD		0x00c0
-#define __LC_STFL_FAC_LIST		0x00c8
-#define __LC_MCCK_CODE			0x00e8
-
-#define __LC_DUMP_REIPL			0x0e00
-
-#ifndef __s390x__
-#define __LC_EXT_OLD_PSW		0x0018
-#define __LC_SVC_OLD_PSW		0x0020
-#define __LC_PGM_OLD_PSW		0x0028
-#define __LC_MCK_OLD_PSW		0x0030
-#define __LC_IO_OLD_PSW			0x0038
-#define __LC_EXT_NEW_PSW		0x0058
-#define __LC_SVC_NEW_PSW		0x0060
-#define __LC_PGM_NEW_PSW		0x0068
-#define __LC_MCK_NEW_PSW		0x0070
-#define __LC_IO_NEW_PSW			0x0078
-#define __LC_SAVE_AREA			0x0200
-#define __LC_RETURN_PSW			0x0240
-#define __LC_RETURN_MCCK_PSW		0x0248
-#define __LC_SYNC_ENTER_TIMER		0x0250
-#define __LC_ASYNC_ENTER_TIMER		0x0258
-#define __LC_EXIT_TIMER			0x0260
-#define __LC_USER_TIMER			0x0268
-#define __LC_SYSTEM_TIMER		0x0270
-#define __LC_STEAL_TIMER		0x0278
-#define __LC_LAST_UPDATE_TIMER		0x0280
-#define __LC_LAST_UPDATE_CLOCK		0x0288
-#define __LC_CURRENT			0x0290
-#define __LC_THREAD_INFO		0x0294
-#define __LC_KERNEL_STACK		0x0298
-#define __LC_ASYNC_STACK		0x029c
-#define __LC_PANIC_STACK		0x02a0
-#define __LC_KERNEL_ASCE		0x02a4
-#define __LC_USER_ASCE			0x02a8
-#define __LC_USER_EXEC_ASCE		0x02ac
-#define __LC_CPUID			0x02b0
-#define __LC_INT_CLOCK			0x02c8
-#define __LC_MACHINE_FLAGS		0x02d8
-#define __LC_FTRACE_FUNC		0x02dc
-#define __LC_IRB			0x0300
-#define __LC_PFAULT_INTPARM		0x0080
-#define __LC_CPU_TIMER_SAVE_AREA	0x00d8
-#define __LC_CLOCK_COMP_SAVE_AREA	0x00e0
-#define __LC_PSW_SAVE_AREA		0x0100
-#define __LC_PREFIX_SAVE_AREA		0x0108
-#define __LC_AREGS_SAVE_AREA		0x0120
-#define __LC_FPREGS_SAVE_AREA		0x0160
-#define __LC_GPREGS_SAVE_AREA		0x0180
-#define __LC_CREGS_SAVE_AREA		0x01c0
-#else /* __s390x__ */
-#define __LC_LAST_BREAK			0x0110
-#define __LC_EXT_OLD_PSW		0x0130
-#define __LC_SVC_OLD_PSW		0x0140
-#define __LC_PGM_OLD_PSW		0x0150
-#define __LC_MCK_OLD_PSW		0x0160
-#define __LC_IO_OLD_PSW			0x0170
-#define __LC_RESTART_PSW		0x01a0
-#define __LC_EXT_NEW_PSW		0x01b0
-#define __LC_SVC_NEW_PSW		0x01c0
-#define __LC_PGM_NEW_PSW		0x01d0
-#define __LC_MCK_NEW_PSW		0x01e0
-#define __LC_IO_NEW_PSW			0x01f0
-#define __LC_SAVE_AREA			0x0200
-#define __LC_RETURN_PSW			0x0280
-#define __LC_RETURN_MCCK_PSW		0x0290
-#define __LC_SYNC_ENTER_TIMER		0x02a0
-#define __LC_ASYNC_ENTER_TIMER		0x02a8
-#define __LC_EXIT_TIMER			0x02b0
-#define __LC_USER_TIMER			0x02b8
-#define __LC_SYSTEM_TIMER		0x02c0
-#define __LC_STEAL_TIMER		0x02c8
-#define __LC_LAST_UPDATE_TIMER		0x02d0
-#define __LC_LAST_UPDATE_CLOCK		0x02d8
-#define __LC_CURRENT			0x02e0
-#define __LC_THREAD_INFO		0x02e8
-#define __LC_KERNEL_STACK		0x02f0
-#define __LC_ASYNC_STACK		0x02f8
-#define __LC_PANIC_STACK		0x0300
-#define __LC_KERNEL_ASCE		0x0308
-#define __LC_USER_ASCE			0x0310
-#define __LC_USER_EXEC_ASCE		0x0318
-#define __LC_CPUID			0x0320
-#define __LC_INT_CLOCK			0x0340
-#define __LC_VDSO_PER_CPU		0x0350
-#define __LC_MACHINE_FLAGS		0x0358
-#define __LC_FTRACE_FUNC		0x0360
-#define __LC_IRB			0x0380
-#define __LC_PASTE			0x03c0
-#define __LC_PFAULT_INTPARM		0x11b8
-#define __LC_FPREGS_SAVE_AREA		0x1200
-#define __LC_GPREGS_SAVE_AREA		0x1280
-#define __LC_PSW_SAVE_AREA		0x1300
-#define __LC_PREFIX_SAVE_AREA		0x1318
-#define __LC_FP_CREG_SAVE_AREA		0x131c
-#define __LC_TODREG_SAVE_AREA		0x1324
-#define __LC_CPU_TIMER_SAVE_AREA	0x1328
-#define __LC_CLOCK_COMP_SAVE_AREA	0x1331
-#define __LC_AREGS_SAVE_AREA		0x1340
-#define __LC_CREGS_SAVE_AREA		0x1380
-#endif /* __s390x__ */
-
-#ifndef __ASSEMBLY__
-
-#include <asm/cpu.h>
-#include <asm/ptrace.h>
 #include <linux/types.h>
+#include <asm/ptrace.h>
+#include <asm/cpu.h>
 
 void restart_int_handler(void);
 void ext_int_handler(void);
@@ -144,7 +19,12 @@ void pgm_check_handler(void);
 void mcck_int_handler(void);
 void io_int_handler(void);
 
-struct save_area_s390 {
+#ifdef CONFIG_32BIT
+
+#define LC_ORDER 0
+#define LC_PAGES 1
+
+struct save_area {
 	u32	ext_save;
 	u64	timer;
 	u64	clk_cmp;
@@ -156,54 +36,13 @@ struct save_area_s390 {
 	u64	fp_regs[4];
 	u32	gp_regs[16];
 	u32	ctrl_regs[16];
-}  __attribute__((packed));
+} __packed;
 
-struct save_area_s390x {
-	u64	fp_regs[16];
-	u64	gp_regs[16];
-	u8	psw[16];
-	u8	pad1[8];
-	u32	pref_reg;
-	u32	fp_ctrl_reg;
-	u8	pad2[4];
-	u32	tod_reg;
-	u64	timer;
-	u64	clk_cmp;
-	u8	pad3[8];
-	u32	acc_regs[16];
-	u64	ctrl_regs[16];
-}  __attribute__((packed));
-
-union save_area {
-	struct save_area_s390	s390;
-	struct save_area_s390x	s390x;
-};
-
-#define SAVE_AREA_BASE_S390	0xd4
-#define SAVE_AREA_BASE_S390X	0x1200
-
-#ifndef __s390x__
-#define SAVE_AREA_SIZE sizeof(struct save_area_s390)
-#define SAVE_AREA_BASE SAVE_AREA_BASE_S390
-#else
-#define SAVE_AREA_SIZE sizeof(struct save_area_s390x)
-#define SAVE_AREA_BASE SAVE_AREA_BASE_S390X
-#endif
-
-#ifndef __s390x__
-#define LC_ORDER 0
-#else
-#define LC_ORDER 1
-#endif
-
-#define LC_PAGES (1UL << LC_ORDER)
-
-struct _lowcore
-{
-#ifndef __s390x__
-	/* 0x0000 - 0x01ff: defined by architecture */
+struct _lowcore {
 	psw_t	restart_psw;			/* 0x0000 */
-	__u32	ccw2[4];			/* 0x0008 */
+	psw_t	restart_old_psw;		/* 0x0008 */
+	__u8	pad_0x0010[0x0014-0x0010];	/* 0x0010 */
+	__u32	ipl_parmblock_ptr;		/* 0x0014 */
 	psw_t	external_old_psw;		/* 0x0018 */
 	psw_t	svc_old_psw;			/* 0x0020 */
 	psw_t	program_old_psw;		/* 0x0028 */
@@ -229,7 +68,9 @@ struct _lowcore
 	__u32	monitor_code;			/* 0x009c */
 	__u8	exc_access_id;			/* 0x00a0 */
 	__u8	per_access_id;			/* 0x00a1 */
-	__u8	pad_0x00a2[0x00b8-0x00a2];	/* 0x00a2 */
+	__u8	op_access_id;			/* 0x00a2 */
+	__u8	ar_access_id;			/* 0x00a3 */
+	__u8	pad_0x00a4[0x00b8-0x00a4];	/* 0x00a4 */
 	__u16	subchannel_id;			/* 0x00b8 */
 	__u16	subchannel_nr;			/* 0x00ba */
 	__u32	io_int_parm;			/* 0x00bc */
@@ -245,8 +86,9 @@ struct _lowcore
 	__u32	external_damage_code;		/* 0x00f4 */
 	__u32	failing_storage_address;	/* 0x00f8 */
 	__u8	pad_0x00fc[0x0100-0x00fc];	/* 0x00fc */
-	__u32	st_status_fixed_logout[4];	/* 0x0100 */
-	__u8	pad_0x0110[0x0120-0x0110];	/* 0x0110 */
+	psw_t	psw_save_area;			/* 0x0100 */
+	__u32	prefixreg_save_area;		/* 0x0108 */
+	__u8	pad_0x010c[0x0120-0x010c];	/* 0x010c */
 
 	/* CPU register save area: defined by architecture */
 	__u32	access_regs_save_area[16];	/* 0x0120 */
@@ -310,10 +152,32 @@ struct _lowcore
 
 	/* Align to the top 1k of prefix area */
 	__u8	pad_0x0e08[0x1000-0x0e08];	/* 0x0e08 */
-#else /* !__s390x__ */
-	/* 0x0000 - 0x01ff: defined by architecture */
-	__u32	ccw1[2];			/* 0x0000 */
-	__u32	ccw2[4];			/* 0x0008 */
+} __packed;
+
+#else /* CONFIG_32BIT */
+
+#define LC_ORDER 1
+#define LC_PAGES 2
+
+struct save_area {
+	u64	fp_regs[16];
+	u64	gp_regs[16];
+	u8	psw[16];
+	u8	pad1[8];
+	u32	pref_reg;
+	u32	fp_ctrl_reg;
+	u8	pad2[4];
+	u32	tod_reg;
+	u64	timer;
+	u64	clk_cmp;
+	u8	pad3[8];
+	u32	acc_regs[16];
+	u64	ctrl_regs[16];
+} __packed;
+
+struct _lowcore {
+	__u8	pad_0x0000[0x0014-0x0000];	/* 0x0000 */
+	__u32	ipl_parmblock_ptr;		/* 0x0014 */
 	__u8	pad_0x0018[0x0080-0x0018];	/* 0x0018 */
 	__u32	ext_params;			/* 0x0080 */
 	__u16	cpu_addr;			/* 0x0084 */
@@ -344,7 +208,9 @@ struct _lowcore
 	__u8	pad_0x00f0[0x00f4-0x00f0];	/* 0x00f0 */
 	__u32	external_damage_code;		/* 0x00f4 */
 	addr_t	failing_storage_address;	/* 0x00f8 */
-	__u8	pad_0x0100[0x0120-0x0100];	/* 0x0100 */
+	__u8	pad_0x0100[0x0110-0x0100];	/* 0x0100 */
+	__u64	breaking_event_addr;		/* 0x0110 */
+	__u8	pad_0x0118[0x0120-0x0118];	/* 0x0118 */
 	psw_t	restart_old_psw;		/* 0x0120 */
 	psw_t	external_old_psw;		/* 0x0130 */
 	psw_t	svc_old_psw;			/* 0x0140 */
@@ -425,7 +291,7 @@ struct _lowcore
 	/* CPU register save area: defined by architecture */
 	__u64	floating_pt_save_area[16];	/* 0x1200 */
 	__u64	gpregs_save_area[16];		/* 0x1280 */
-	__u32	st_status_fixed_logout[4];	/* 0x1300 */
+	psw_t	psw_save_area;			/* 0x1300 */
 	__u8	pad_0x1310[0x1318-0x1310];	/* 0x1310 */
 	__u32	prefixreg_save_area;		/* 0x1318 */
 	__u32	fpt_creg_save_area;		/* 0x131c */
@@ -439,10 +305,12 @@ struct _lowcore
 
 	/* align to the top of the prefix area */
 	__u8	pad_0x1400[0x2000-0x1400];	/* 0x1400 */
-#endif /* !__s390x__ */
-} __attribute__((packed)); /* End structure*/
+} __packed;
+
+#endif /* CONFIG_32BIT */
 
 #define S390_lowcore (*((struct _lowcore *) 0))
+
 extern struct _lowcore *lowcore_ptr[];
 
 static inline void set_prefix(__u32 address)
@@ -458,6 +326,4 @@ static inline __u32 store_prefix(void)
 	return address;
 }
 
-#endif
-
-#endif
+#endif /* _ASM_S390_LOWCORE_H */
diff --git a/arch/s390/include/asm/page.h b/arch/s390/include/asm/page.h
index 5e9daf5..af650fb 100644
--- a/arch/s390/include/asm/page.h
+++ b/arch/s390/include/asm/page.h
@@ -107,9 +107,6 @@ typedef pte_t *pgtable_t;
 #define __pgd(x)        ((pgd_t) { (x) } )
 #define __pgprot(x)     ((pgprot_t) { (x) } )
 
-/* default storage key used for all pages */
-extern unsigned int default_storage_key;
-
 static inline void
 page_set_storage_key(unsigned long addr, unsigned int skey)
 {
diff --git a/arch/s390/include/asm/processor.h b/arch/s390/include/asm/processor.h
index b427154..73e2598 100644
--- a/arch/s390/include/asm/processor.h
+++ b/arch/s390/include/asm/processor.h
@@ -28,7 +28,7 @@
 
 static inline void get_cpu_id(struct cpuid *ptr)
 {
-	asm volatile("stidp 0(%1)" : "=m" (*ptr) : "a" (ptr));
+	asm volatile("stidp %0" : "=Q" (*ptr));
 }
 
 extern void s390_adjust_jiffies(void);
@@ -184,9 +184,9 @@ static inline void psw_set_key(unsigned int key)
 static inline void __load_psw(psw_t psw)
 {
 #ifndef __s390x__
-	asm volatile("lpsw  0(%0)" : : "a" (&psw), "m" (psw) : "cc");
+	asm volatile("lpsw  %0" : : "Q" (psw) : "cc");
 #else
-	asm volatile("lpswe 0(%0)" : : "a" (&psw), "m" (psw) : "cc");
+	asm volatile("lpswe %0" : : "Q" (psw) : "cc");
 #endif
 }
 
@@ -206,17 +206,17 @@ static inline void __load_psw_mask (unsigned long mask)
 	asm volatile(
 		"	basr	%0,0\n"
 		"0:	ahi	%0,1f-0b\n"
-		"	st	%0,4(%1)\n"
-		"	lpsw	0(%1)\n"
+		"	st	%0,%O1+4(%R1)\n"
+		"	lpsw	%1\n"
 		"1:"
-		: "=&d" (addr) : "a" (&psw), "m" (psw) : "memory", "cc");
+		: "=&d" (addr), "=Q" (psw) : "Q" (psw) : "memory", "cc");
 #else /* __s390x__ */
 	asm volatile(
 		"	larl	%0,1f\n"
-		"	stg	%0,8(%1)\n"
-		"	lpswe	0(%1)\n"
+		"	stg	%0,%O1+8(%R1)\n"
+		"	lpswe	%1\n"
 		"1:"
-		: "=&d" (addr) : "a" (&psw), "m" (psw) : "memory", "cc");
+		: "=&d" (addr), "=Q" (psw) : "Q" (psw) : "memory", "cc");
 #endif /* __s390x__ */
 }
  
diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h
index 95dcf18..dd2d913 100644
--- a/arch/s390/include/asm/ptrace.h
+++ b/arch/s390/include/asm/ptrace.h
@@ -492,13 +492,24 @@ struct user_regs_struct
 struct task_struct;
 extern void user_enable_single_step(struct task_struct *);
 extern void user_disable_single_step(struct task_struct *);
+extern void show_regs(struct pt_regs * regs);
 
 #define user_mode(regs) (((regs)->psw.mask & PSW_MASK_PSTATE) != 0)
 #define instruction_pointer(regs) ((regs)->psw.addr & PSW_ADDR_INSN)
 #define user_stack_pointer(regs)((regs)->gprs[15])
 #define regs_return_value(regs)((regs)->gprs[2])
 #define profile_pc(regs) instruction_pointer(regs)
-extern void show_regs(struct pt_regs * regs);
+
+int regs_query_register_offset(const char *name);
+const char *regs_query_register_name(unsigned int offset);
+unsigned long regs_get_register(struct pt_regs *regs, unsigned int offset);
+unsigned long regs_get_kernel_stack_nth(struct pt_regs *regs, unsigned int n);
+
+static inline unsigned long kernel_stack_pointer(struct pt_regs *regs)
+{
+	return regs->gprs[15] & PSW_ADDR_INSN;
+}
+
 #endif /* __KERNEL__ */
 #endif /* __ASSEMBLY__ */
 
diff --git a/arch/s390/include/asm/qdio.h b/arch/s390/include/asm/qdio.h
index 79d849f..c666bfe 100644
--- a/arch/s390/include/asm/qdio.h
+++ b/arch/s390/include/asm/qdio.h
@@ -13,7 +13,8 @@
 #include <asm/cio.h>
 #include <asm/ccwdev.h>
 
-#define QDIO_MAX_QUEUES_PER_IRQ		32
+/* only use 4 queues to save some cachelines */
+#define QDIO_MAX_QUEUES_PER_IRQ		4
 #define QDIO_MAX_BUFFERS_PER_Q		128
 #define QDIO_MAX_BUFFERS_MASK		(QDIO_MAX_BUFFERS_PER_Q - 1)
 #define QDIO_MAX_ELEMENTS_PER_BUFFER	16
diff --git a/arch/s390/include/asm/rwsem.h b/arch/s390/include/asm/rwsem.h
index 9d2a179..423fdda 100644
--- a/arch/s390/include/asm/rwsem.h
+++ b/arch/s390/include/asm/rwsem.h
@@ -124,21 +124,21 @@ static inline void __down_read(struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	ahi	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	ahi	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	aghi	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	aghi	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count),
-		  "i" (RWSEM_ACTIVE_READ_BIAS) : "cc", "memory");
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "i" (RWSEM_ACTIVE_READ_BIAS)
+		: "cc", "memory");
 	if (old < 0)
 		rwsem_down_read_failed(sem);
 }
@@ -152,25 +152,25 @@ static inline int __down_read_trylock(struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	ltr	%1,%0\n"
 		"	jm	1f\n"
-		"	ahi	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	ahi	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b\n"
 		"1:"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	ltgr	%1,%0\n"
 		"	jm	1f\n"
-		"	aghi	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	aghi	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b\n"
 		"1:"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count),
-		  "i" (RWSEM_ACTIVE_READ_BIAS) : "cc", "memory");
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "i" (RWSEM_ACTIVE_READ_BIAS)
+		: "cc", "memory");
 	return old >= 0 ? 1 : 0;
 }
 
@@ -184,20 +184,20 @@ static inline void __down_write_nested(struct rw_semaphore *sem, int subclass)
 	tmp = RWSEM_ACTIVE_WRITE_BIAS;
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	a	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	a	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	ag	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	ag	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count), "m" (tmp)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "m" (tmp)
 		: "cc", "memory");
 	if (old != 0)
 		rwsem_down_write_failed(sem);
@@ -217,22 +217,22 @@ static inline int __down_write_trylock(struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%2)\n"
+		"	l	%0,%1\n"
 		"0:	ltr	%0,%0\n"
 		"	jnz	1f\n"
-		"	cs	%0,%4,0(%2)\n"
+		"	cs	%0,%3,%1\n"
 		"	jl	0b\n"
 #else /* __s390x__ */
-		"	lg	%0,0(%2)\n"
+		"	lg	%0,%1\n"
 		"0:	ltgr	%0,%0\n"
 		"	jnz	1f\n"
-		"	csg	%0,%4,0(%2)\n"
+		"	csg	%0,%3,%1\n"
 		"	jl	0b\n"
 #endif /* __s390x__ */
 		"1:"
-		: "=&d" (old), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count),
-		  "d" (RWSEM_ACTIVE_WRITE_BIAS) : "cc", "memory");
+		: "=&d" (old), "=Q" (sem->count)
+		: "Q" (sem->count), "d" (RWSEM_ACTIVE_WRITE_BIAS)
+		: "cc", "memory");
 	return (old == RWSEM_UNLOCKED_VALUE) ? 1 : 0;
 }
 
@@ -245,21 +245,20 @@ static inline void __up_read(struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	ahi	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	ahi	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	aghi	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	aghi	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count),
-		  "i" (-RWSEM_ACTIVE_READ_BIAS)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "i" (-RWSEM_ACTIVE_READ_BIAS)
 		: "cc", "memory");
 	if (new < 0)
 		if ((new & RWSEM_ACTIVE_MASK) == 0)
@@ -276,20 +275,20 @@ static inline void __up_write(struct rw_semaphore *sem)
 	tmp = -RWSEM_ACTIVE_WRITE_BIAS;
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	a	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	a	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	ag	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	ag	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count), "m" (tmp)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "m" (tmp)
 		: "cc", "memory");
 	if (new < 0)
 		if ((new & RWSEM_ACTIVE_MASK) == 0)
@@ -306,20 +305,20 @@ static inline void __downgrade_write(struct rw_semaphore *sem)
 	tmp = -RWSEM_WAITING_BIAS;
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	a	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	a	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	ag	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	ag	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count), "m" (tmp)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "m" (tmp)
 		: "cc", "memory");
 	if (new > 1)
 		rwsem_downgrade_wake(sem);
@@ -334,20 +333,20 @@ static inline void rwsem_atomic_add(long delta, struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	ar	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	ar	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	agr	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	agr	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count), "d" (delta)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "d" (delta)
 		: "cc", "memory");
 }
 
@@ -360,20 +359,20 @@ static inline long rwsem_atomic_update(long delta, struct rw_semaphore *sem)
 
 	asm volatile(
 #ifndef __s390x__
-		"	l	%0,0(%3)\n"
+		"	l	%0,%2\n"
 		"0:	lr	%1,%0\n"
-		"	ar	%1,%5\n"
-		"	cs	%0,%1,0(%3)\n"
+		"	ar	%1,%4\n"
+		"	cs	%0,%1,%2\n"
 		"	jl	0b"
 #else /* __s390x__ */
-		"	lg	%0,0(%3)\n"
+		"	lg	%0,%2\n"
 		"0:	lgr	%1,%0\n"
-		"	agr	%1,%5\n"
-		"	csg	%0,%1,0(%3)\n"
+		"	agr	%1,%4\n"
+		"	csg	%0,%1,%2\n"
 		"	jl	0b"
 #endif /* __s390x__ */
-		: "=&d" (old), "=&d" (new), "=m" (sem->count)
-		: "a" (&sem->count), "m" (sem->count), "d" (delta)
+		: "=&d" (old), "=&d" (new), "=Q" (sem->count)
+		: "Q" (sem->count), "d" (delta)
 		: "cc", "memory");
 	return new;
 }
diff --git a/arch/s390/include/asm/setup.h b/arch/s390/include/asm/setup.h
index 52a779c..9ab6bd3 100644
--- a/arch/s390/include/asm/setup.h
+++ b/arch/s390/include/asm/setup.h
@@ -14,14 +14,14 @@
 
 #ifdef __KERNEL__
 
-#include <asm/lowcore.h>
-#include <asm/types.h>
-
 #define PARMAREA		0x10400
 #define MEMORY_CHUNKS		256
 
 #ifndef __ASSEMBLY__
 
+#include <asm/lowcore.h>
+#include <asm/types.h>
+
 #ifndef __s390x__
 #define IPL_DEVICE        (*(unsigned long *)  (0x10404))
 #define INITRD_START      (*(unsigned long *)  (0x1040C))
@@ -71,9 +71,12 @@ extern unsigned int user_mode;
 #define MACHINE_FLAG_KVM	(1UL << 9)
 #define MACHINE_FLAG_HPAGE	(1UL << 10)
 #define MACHINE_FLAG_PFMF	(1UL << 11)
+#define MACHINE_FLAG_LPAR	(1UL << 12)
 
 #define MACHINE_IS_VM		(S390_lowcore.machine_flags & MACHINE_FLAG_VM)
 #define MACHINE_IS_KVM		(S390_lowcore.machine_flags & MACHINE_FLAG_KVM)
+#define MACHINE_IS_LPAR		(S390_lowcore.machine_flags & MACHINE_FLAG_LPAR)
+
 #define MACHINE_HAS_DIAG9C	(S390_lowcore.machine_flags & MACHINE_FLAG_DIAG9C)
 
 #ifndef __s390x__
diff --git a/arch/s390/include/asm/sigp.h b/arch/s390/include/asm/sigp.h
index f72d611..e3bffd4 100644
--- a/arch/s390/include/asm/sigp.h
+++ b/arch/s390/include/asm/sigp.h
@@ -1,24 +1,19 @@
 /*
- *  include/asm-s390/sigp.h
+ *  Routines and structures for signalling other processors.
  *
- *  S390 version
- *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
- *    Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com),
- *               Martin Schwidefsky (schwidefsky@de.ibm.com)
- *               Heiko Carstens (heiko.carstens@de.ibm.com)
- *
- *  sigp.h by D.J. Barrow (c) IBM 1999
- *  contains routines / structures for signalling other S/390 processors in an
- *  SMP configuration.
+ *    Copyright IBM Corp. 1999,2010
+ *    Author(s): Denis Joseph Barrow,
+ *		 Martin Schwidefsky <schwidefsky@de.ibm.com>,
+ *		 Heiko Carstens <heiko.carstens@de.ibm.com>,
  */
 
-#ifndef __SIGP__
-#define __SIGP__
+#ifndef __ASM_SIGP_H
+#define __ASM_SIGP_H
 
 #include <asm/system.h>
 
-/* get real cpu address from logical cpu number */
-extern int __cpu_logical_map[];
+/* Get real cpu address from logical cpu number. */
+extern unsigned short __cpu_logical_map[];
 
 static inline int cpu_logical_map(int cpu)
 {
@@ -29,107 +24,108 @@ static inline int cpu_logical_map(int cpu)
 #endif
 }
 
-typedef enum
-{
-	sigp_unassigned=0x0,
-	sigp_sense,
-	sigp_external_call,
-	sigp_emergency_signal,
-	sigp_start,
-	sigp_stop,
-	sigp_restart,
-	sigp_unassigned1,
-	sigp_unassigned2,
-	sigp_stop_and_store_status,
-	sigp_unassigned3,
-	sigp_initial_cpu_reset,
-	sigp_cpu_reset,
-	sigp_set_prefix,
-	sigp_store_status_at_address,
-	sigp_store_extended_status_at_address
-} sigp_order_code;
-
-typedef __u32 sigp_status_word;
-
-typedef enum
-{
-        sigp_order_code_accepted=0,
-	sigp_status_stored,
-	sigp_busy,
-	sigp_not_operational
-} sigp_ccode;
-
+enum {
+	sigp_sense = 1,
+	sigp_external_call = 2,
+	sigp_emergency_signal = 3,
+	sigp_start = 4,
+	sigp_stop = 5,
+	sigp_restart = 6,
+	sigp_stop_and_store_status = 9,
+	sigp_initial_cpu_reset = 11,
+	sigp_cpu_reset = 12,
+	sigp_set_prefix = 13,
+	sigp_store_status_at_address = 14,
+	sigp_store_extended_status_at_address = 15,
+	sigp_set_architecture = 18,
+	sigp_conditional_emergency_signal = 19,
+	sigp_sense_running = 21,
+};
+
+enum {
+	sigp_order_code_accepted = 0,
+	sigp_status_stored = 1,
+	sigp_busy = 2,
+	sigp_not_operational = 3,
+};
 
 /*
- * Definitions for the external call
+ * Definitions for external call.
  */
-
-/* 'Bit' signals, asynchronous */
-typedef enum
-{
-	ec_schedule=0,
+enum {
+	ec_schedule = 0,
 	ec_call_function,
 	ec_call_function_single,
-	ec_bit_last
-} ec_bit_sig;
+};
 
 /*
- * Signal processor
+ * Signal processor.
  */
-static inline sigp_ccode
-signal_processor(__u16 cpu_addr, sigp_order_code order_code)
+static inline int raw_sigp(u16 cpu, int order)
 {
 	register unsigned long reg1 asm ("1") = 0;
-	sigp_ccode ccode;
+	int ccode;
 
 	asm volatile(
 		"	sigp	%1,%2,0(%3)\n"
 		"	ipm	%0\n"
 		"	srl	%0,28\n"
 		:	"=d"	(ccode)
-		: "d" (reg1), "d" (cpu_logical_map(cpu_addr)),
-		  "a" (order_code) : "cc" , "memory");
+		: "d" (reg1), "d" (cpu),
+		  "a" (order) : "cc" , "memory");
 	return ccode;
 }
 
 /*
- * Signal processor with parameter
+ * Signal processor with parameter.
  */
-static inline sigp_ccode
-signal_processor_p(__u32 parameter, __u16 cpu_addr, sigp_order_code order_code)
+static inline int raw_sigp_p(u32 parameter, u16 cpu, int order)
 {
 	register unsigned int reg1 asm ("1") = parameter;
-	sigp_ccode ccode;
+	int ccode;
 
 	asm volatile(
 		"	sigp	%1,%2,0(%3)\n"
 		"	ipm	%0\n"
 		"	srl	%0,28\n"
 		: "=d" (ccode)
-		: "d" (reg1), "d" (cpu_logical_map(cpu_addr)),
-		  "a" (order_code) : "cc" , "memory");
+		: "d" (reg1), "d" (cpu),
+		  "a" (order) : "cc" , "memory");
 	return ccode;
 }
 
 /*
- * Signal processor with parameter and return status
+ * Signal processor with parameter and return status.
  */
-static inline sigp_ccode
-signal_processor_ps(__u32 *statusptr, __u32 parameter, __u16 cpu_addr,
-		    sigp_order_code order_code)
+static inline int raw_sigp_ps(u32 *status, u32 parm, u16 cpu, int order)
 {
-	register unsigned int reg1 asm ("1") = parameter;
-	sigp_ccode ccode;
+	register unsigned int reg1 asm ("1") = parm;
+	int ccode;
 
 	asm volatile(
 		"	sigp	%1,%2,0(%3)\n"
 		"	ipm	%0\n"
 		"	srl	%0,28\n"
 		: "=d" (ccode), "+d" (reg1)
-		: "d" (cpu_logical_map(cpu_addr)), "a" (order_code)
+		: "d" (cpu), "a" (order)
 		: "cc" , "memory");
-	*statusptr = reg1;
+	*status = reg1;
 	return ccode;
 }
 
-#endif /* __SIGP__ */
+static inline int sigp(int cpu, int order)
+{
+	return raw_sigp(cpu_logical_map(cpu), order);
+}
+
+static inline int sigp_p(u32 parameter, int cpu, int order)
+{
+	return raw_sigp_p(parameter, cpu_logical_map(cpu), order);
+}
+
+static inline int sigp_ps(u32 *status, u32 parm, int cpu, int order)
+{
+	return raw_sigp_ps(status, parm, cpu_logical_map(cpu), order);
+}
+
+#endif /* __ASM_SIGP_H */
diff --git a/arch/s390/include/asm/smp.h b/arch/s390/include/asm/smp.h
index 2ab1141..edc03cb 100644
--- a/arch/s390/include/asm/smp.h
+++ b/arch/s390/include/asm/smp.h
@@ -29,7 +29,43 @@ extern int smp_cpu_polarization[];
 extern void arch_send_call_function_single_ipi(int cpu);
 extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
 
-extern union save_area *zfcpdump_save_areas[NR_CPUS + 1];
+extern struct save_area *zfcpdump_save_areas[NR_CPUS + 1];
+
+extern void smp_switch_to_ipl_cpu(void (*func)(void *), void *);
+extern void smp_switch_to_cpu(void (*)(void *), void *, unsigned long sp,
+			      int from, int to);
+extern void smp_restart_cpu(void);
+
+/*
+ * returns 1 if (virtual) cpu is scheduled
+ * returns 0 otherwise
+ */
+static inline int smp_vcpu_scheduled(int cpu)
+{
+	u32 status;
+
+	switch (sigp_ps(&status, 0, cpu, sigp_sense_running)) {
+	case sigp_status_stored:
+		/* Check for running status */
+		if (status & 0x400)
+			return 0;
+		break;
+	case sigp_not_operational:
+		return 0;
+	default:
+		break;
+	}
+	return 1;
+}
+
+#else /* CONFIG_SMP */
+
+static inline void smp_switch_to_ipl_cpu(void (*func)(void *), void *data)
+{
+	func(data);
+}
+
+#define smp_vcpu_scheduled	(1)
 
 #endif /* CONFIG_SMP */
 
diff --git a/arch/s390/include/asm/spinlock.h b/arch/s390/include/asm/spinlock.h
index a587907..56612fc 100644
--- a/arch/s390/include/asm/spinlock.h
+++ b/arch/s390/include/asm/spinlock.h
@@ -13,8 +13,6 @@
 
 #include <linux/smp.h>
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
-
 static inline int
 _raw_compare_and_swap(volatile unsigned int *lock,
 		      unsigned int old, unsigned int new)
@@ -27,22 +25,6 @@ _raw_compare_and_swap(volatile unsigned int *lock,
 	return old;
 }
 
-#else /* __GNUC__ */
-
-static inline int
-_raw_compare_and_swap(volatile unsigned int *lock,
-		      unsigned int old, unsigned int new)
-{
-	asm volatile(
-		"	cs	%0,%3,0(%4)"
-		: "=d" (old), "=m" (*lock)
-		: "0" (old), "d" (new), "a" (lock), "m" (*lock)
-		: "cc", "memory" );
-	return old;
-}
-
-#endif /* __GNUC__ */
-
 /*
  * Simple spin lock operations.  There are two variants, one clears IRQ's
  * on the local processor, one does not.
diff --git a/arch/s390/include/asm/swab.h b/arch/s390/include/asm/swab.h
index eb18dc1..6bdee21 100644
--- a/arch/s390/include/asm/swab.h
+++ b/arch/s390/include/asm/swab.h
@@ -47,11 +47,11 @@ static inline __u32 __arch_swab32p(const __u32 *x)
 	
 	asm volatile(
 #ifndef __s390x__
-		"	icm	%0,8,3(%1)\n"
-		"	icm	%0,4,2(%1)\n"
-		"	icm	%0,2,1(%1)\n"
-		"	ic	%0,0(%1)"
-		: "=&d" (result) : "a" (x), "m" (*x) : "cc");
+		"	icm	%0,8,%O1+3(%R1)\n"
+		"	icm	%0,4,%O1+2(%R1)\n"
+		"	icm	%0,2,%O1+1(%R1)\n"
+		"	ic	%0,%1"
+		: "=&d" (result) : "Q" (*x) : "cc");
 #else /* __s390x__ */
 		"	lrv	%0,%1"
 		: "=d" (result) : "m" (*x));
@@ -77,9 +77,9 @@ static inline __u16 __arch_swab16p(const __u16 *x)
 	
 	asm volatile(
 #ifndef __s390x__
-		"	icm	%0,2,1(%1)\n"
-		"	ic	%0,0(%1)\n"
-		: "=&d" (result) : "a" (x), "m" (*x) : "cc");
+		"	icm	%0,2,%O+1(%R1)\n"
+		"	ic	%0,%1\n"
+		: "=&d" (result) : "Q" (*x) : "cc");
 #else /* __s390x__ */
 		"	lrvh	%0,%1"
 		: "=d" (result) : "m" (*x));
diff --git a/arch/s390/include/asm/syscall.h b/arch/s390/include/asm/syscall.h
index e0a73d3..8429686 100644
--- a/arch/s390/include/asm/syscall.h
+++ b/arch/s390/include/asm/syscall.h
@@ -15,6 +15,13 @@
 #include <linux/sched.h>
 #include <asm/ptrace.h>
 
+/*
+ * The syscall table always contains 32 bit pointers since we know that the
+ * address of the function to be called is (way) below 4GB.  So the "int"
+ * type here is what we want [need] for both 32 bit and 64 bit systems.
+ */
+extern const unsigned int sys_call_table[];
+
 static inline long syscall_get_nr(struct task_struct *task,
 				  struct pt_regs *regs)
 {
diff --git a/arch/s390/include/asm/sysinfo.h b/arch/s390/include/asm/sysinfo.h
index 9d70057..22bdb2a 100644
--- a/arch/s390/include/asm/sysinfo.h
+++ b/arch/s390/include/asm/sysinfo.h
@@ -87,7 +87,8 @@ struct sysinfo_2_2_2 {
 
 struct sysinfo_3_2_2 {
 	char reserved_0[31];
-	unsigned char count;
+	unsigned char :4;
+	unsigned char count:4;
 	struct {
 		char reserved_0[4];
 		unsigned short cpus_total;
diff --git a/arch/s390/include/asm/system.h b/arch/s390/include/asm/system.h
index 379661d..67ee6c3 100644
--- a/arch/s390/include/asm/system.h
+++ b/arch/s390/include/asm/system.h
@@ -24,65 +24,65 @@ extern struct task_struct *__switch_to(void *, void *);
 static inline void save_fp_regs(s390_fp_regs *fpregs)
 {
 	asm volatile(
-		"	std	0,8(%1)\n"
-		"	std	2,24(%1)\n"
-		"	std	4,40(%1)\n"
-		"	std	6,56(%1)"
-		: "=m" (*fpregs) : "a" (fpregs), "m" (*fpregs) : "memory");
+		"	std	0,%O0+8(%R0)\n"
+		"	std	2,%O0+24(%R0)\n"
+		"	std	4,%O0+40(%R0)\n"
+		"	std	6,%O0+56(%R0)"
+		: "=Q" (*fpregs) : "Q" (*fpregs));
 	if (!MACHINE_HAS_IEEE)
 		return;
 	asm volatile(
-		"	stfpc	0(%1)\n"
-		"	std	1,16(%1)\n"
-		"	std	3,32(%1)\n"
-		"	std	5,48(%1)\n"
-		"	std	7,64(%1)\n"
-		"	std	8,72(%1)\n"
-		"	std	9,80(%1)\n"
-		"	std	10,88(%1)\n"
-		"	std	11,96(%1)\n"
-		"	std	12,104(%1)\n"
-		"	std	13,112(%1)\n"
-		"	std	14,120(%1)\n"
-		"	std	15,128(%1)\n"
-		: "=m" (*fpregs) : "a" (fpregs), "m" (*fpregs) : "memory");
+		"	stfpc	%0\n"
+		"	std	1,%O0+16(%R0)\n"
+		"	std	3,%O0+32(%R0)\n"
+		"	std	5,%O0+48(%R0)\n"
+		"	std	7,%O0+64(%R0)\n"
+		"	std	8,%O0+72(%R0)\n"
+		"	std	9,%O0+80(%R0)\n"
+		"	std	10,%O0+88(%R0)\n"
+		"	std	11,%O0+96(%R0)\n"
+		"	std	12,%O0+104(%R0)\n"
+		"	std	13,%O0+112(%R0)\n"
+		"	std	14,%O0+120(%R0)\n"
+		"	std	15,%O0+128(%R0)\n"
+		: "=Q" (*fpregs) : "Q" (*fpregs));
 }
 
 static inline void restore_fp_regs(s390_fp_regs *fpregs)
 {
 	asm volatile(
-		"	ld	0,8(%0)\n"
-		"	ld	2,24(%0)\n"
-		"	ld	4,40(%0)\n"
-		"	ld	6,56(%0)"
-		: : "a" (fpregs), "m" (*fpregs));
+		"	ld	0,%O0+8(%R0)\n"
+		"	ld	2,%O0+24(%R0)\n"
+		"	ld	4,%O0+40(%R0)\n"
+		"	ld	6,%O0+56(%R0)"
+		: : "Q" (*fpregs));
 	if (!MACHINE_HAS_IEEE)
 		return;
 	asm volatile(
-		"	lfpc	0(%0)\n"
-		"	ld	1,16(%0)\n"
-		"	ld	3,32(%0)\n"
-		"	ld	5,48(%0)\n"
-		"	ld	7,64(%0)\n"
-		"	ld	8,72(%0)\n"
-		"	ld	9,80(%0)\n"
-		"	ld	10,88(%0)\n"
-		"	ld	11,96(%0)\n"
-		"	ld	12,104(%0)\n"
-		"	ld	13,112(%0)\n"
-		"	ld	14,120(%0)\n"
-		"	ld	15,128(%0)\n"
-		: : "a" (fpregs), "m" (*fpregs));
+		"	lfpc	%0\n"
+		"	ld	1,%O0+16(%R0)\n"
+		"	ld	3,%O0+32(%R0)\n"
+		"	ld	5,%O0+48(%R0)\n"
+		"	ld	7,%O0+64(%R0)\n"
+		"	ld	8,%O0+72(%R0)\n"
+		"	ld	9,%O0+80(%R0)\n"
+		"	ld	10,%O0+88(%R0)\n"
+		"	ld	11,%O0+96(%R0)\n"
+		"	ld	12,%O0+104(%R0)\n"
+		"	ld	13,%O0+112(%R0)\n"
+		"	ld	14,%O0+120(%R0)\n"
+		"	ld	15,%O0+128(%R0)\n"
+		: : "Q" (*fpregs));
 }
 
 static inline void save_access_regs(unsigned int *acrs)
 {
-	asm volatile("stam 0,15,0(%0)" : : "a" (acrs) : "memory");
+	asm volatile("stam 0,15,%0" : "=Q" (*acrs));
 }
 
 static inline void restore_access_regs(unsigned int *acrs)
 {
-	asm volatile("lam 0,15,0(%0)" : : "a" (acrs));
+	asm volatile("lam 0,15,%0" : : "Q" (*acrs));
 }
 
 #define switch_to(prev,next,last) do {					     \
@@ -139,48 +139,48 @@ static inline unsigned long __xchg(unsigned long x, void * ptr, int size)
 		shift = (3 ^ (addr & 3)) << 3;
 		addr ^= addr & 3;
 		asm volatile(
-			"	l	%0,0(%4)\n"
+			"	l	%0,%4\n"
 			"0:	lr	0,%0\n"
 			"	nr	0,%3\n"
 			"	or	0,%2\n"
-			"	cs	%0,0,0(%4)\n"
+			"	cs	%0,0,%4\n"
 			"	jl	0b\n"
-			: "=&d" (old), "=m" (*(int *) addr)
-			: "d" (x << shift), "d" (~(255 << shift)), "a" (addr),
-			  "m" (*(int *) addr) : "memory", "cc", "0");
+			: "=&d" (old), "=Q" (*(int *) addr)
+			: "d" (x << shift), "d" (~(255 << shift)),
+			  "Q" (*(int *) addr) : "memory", "cc", "0");
 		return old >> shift;
 	case 2:
 		addr = (unsigned long) ptr;
 		shift = (2 ^ (addr & 2)) << 3;
 		addr ^= addr & 2;
 		asm volatile(
-			"	l	%0,0(%4)\n"
+			"	l	%0,%4\n"
 			"0:	lr	0,%0\n"
 			"	nr	0,%3\n"
 			"	or	0,%2\n"
-			"	cs	%0,0,0(%4)\n"
+			"	cs	%0,0,%4\n"
 			"	jl	0b\n"
-			: "=&d" (old), "=m" (*(int *) addr)
-			: "d" (x << shift), "d" (~(65535 << shift)), "a" (addr),
-			  "m" (*(int *) addr) : "memory", "cc", "0");
+			: "=&d" (old), "=Q" (*(int *) addr)
+			: "d" (x << shift), "d" (~(65535 << shift)),
+			  "Q" (*(int *) addr) : "memory", "cc", "0");
 		return old >> shift;
 	case 4:
 		asm volatile(
-			"	l	%0,0(%3)\n"
-			"0:	cs	%0,%2,0(%3)\n"
+			"	l	%0,%3\n"
+			"0:	cs	%0,%2,%3\n"
 			"	jl	0b\n"
-			: "=&d" (old), "=m" (*(int *) ptr)
-			: "d" (x), "a" (ptr), "m" (*(int *) ptr)
+			: "=&d" (old), "=Q" (*(int *) ptr)
+			: "d" (x), "Q" (*(int *) ptr)
 			: "memory", "cc");
 		return old;
 #ifdef __s390x__
 	case 8:
 		asm volatile(
-			"	lg	%0,0(%3)\n"
-			"0:	csg	%0,%2,0(%3)\n"
+			"	lg	%0,%3\n"
+			"0:	csg	%0,%2,%3\n"
 			"	jl	0b\n"
 			: "=&d" (old), "=m" (*(long *) ptr)
-			: "d" (x), "a" (ptr), "m" (*(long *) ptr)
+			: "d" (x), "Q" (*(long *) ptr)
 			: "memory", "cc");
 		return old;
 #endif /* __s390x__ */
@@ -215,20 +215,20 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
 		shift = (3 ^ (addr & 3)) << 3;
 		addr ^= addr & 3;
 		asm volatile(
-			"	l	%0,0(%4)\n"
+			"	l	%0,%2\n"
 			"0:	nr	%0,%5\n"
 			"	lr	%1,%0\n"
 			"	or	%0,%2\n"
 			"	or	%1,%3\n"
-			"	cs	%0,%1,0(%4)\n"
+			"	cs	%0,%1,%2\n"
 			"	jnl	1f\n"
 			"	xr	%1,%0\n"
 			"	nr	%1,%5\n"
 			"	jnz	0b\n"
 			"1:"
-			: "=&d" (prev), "=&d" (tmp)
-			: "d" (old << shift), "d" (new << shift), "a" (ptr),
-			  "d" (~(255 << shift))
+			: "=&d" (prev), "=&d" (tmp), "=Q" (*(int *) ptr)
+			: "d" (old << shift), "d" (new << shift),
+			  "d" (~(255 << shift)), "Q" (*(int *) ptr)
 			: "memory", "cc");
 		return prev >> shift;
 	case 2:
@@ -236,33 +236,35 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
 		shift = (2 ^ (addr & 2)) << 3;
 		addr ^= addr & 2;
 		asm volatile(
-			"	l	%0,0(%4)\n"
+			"	l	%0,%2\n"
 			"0:	nr	%0,%5\n"
 			"	lr	%1,%0\n"
 			"	or	%0,%2\n"
 			"	or	%1,%3\n"
-			"	cs	%0,%1,0(%4)\n"
+			"	cs	%0,%1,%2\n"
 			"	jnl	1f\n"
 			"	xr	%1,%0\n"
 			"	nr	%1,%5\n"
 			"	jnz	0b\n"
 			"1:"
-			: "=&d" (prev), "=&d" (tmp)
-			: "d" (old << shift), "d" (new << shift), "a" (ptr),
-			  "d" (~(65535 << shift))
+			: "=&d" (prev), "=&d" (tmp), "=Q" (*(int *) ptr)
+			: "d" (old << shift), "d" (new << shift),
+			  "d" (~(65535 << shift)), "Q" (*(int *) ptr)
 			: "memory", "cc");
 		return prev >> shift;
 	case 4:
 		asm volatile(
-			"	cs	%0,%2,0(%3)\n"
-			: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
+			"	cs	%0,%3,%1\n"
+			: "=&d" (prev), "=Q" (*(int *) ptr)
+			: "0" (old), "d" (new), "Q" (*(int *) ptr)
 			: "memory", "cc");
 		return prev;
 #ifdef __s390x__
 	case 8:
 		asm volatile(
-			"	csg	%0,%2,0(%3)\n"
-			: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
+			"	csg	%0,%3,%1\n"
+			: "=&d" (prev), "=Q" (*(long *) ptr)
+			: "0" (old), "d" (new), "Q" (*(long *) ptr)
 			: "memory", "cc");
 		return prev;
 #endif /* __s390x__ */
@@ -302,17 +304,17 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
 #define __ctl_load(array, low, high) ({				\
 	typedef struct { char _[sizeof(array)]; } addrtype;	\
 	asm volatile(						\
-		"	lctlg	%1,%2,0(%0)\n"			\
-		: : "a" (&array), "i" (low), "i" (high),	\
-		    "m" (*(addrtype *)(&array)));		\
+		"	lctlg	%1,%2,%0\n"			\
+		: : "Q" (*(addrtype *)(&array)),		\
+		    "i" (low), "i" (high));			\
 	})
 
 #define __ctl_store(array, low, high) ({			\
 	typedef struct { char _[sizeof(array)]; } addrtype;	\
 	asm volatile(						\
-		"	stctg	%2,%3,0(%1)\n"			\
-		: "=m" (*(addrtype *)(&array))			\
-		: "a" (&array), "i" (low), "i" (high));		\
+		"	stctg	%1,%2,%0\n"			\
+		: "=Q" (*(addrtype *)(&array))			\
+		: "i" (low), "i" (high));			\
 	})
 
 #else /* __s390x__ */
@@ -320,17 +322,17 @@ __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
 #define __ctl_load(array, low, high) ({				\
 	typedef struct { char _[sizeof(array)]; } addrtype;	\
 	asm volatile(						\
-		"	lctl	%1,%2,0(%0)\n"			\
-		: : "a" (&array), "i" (low), "i" (high),	\
-		    "m" (*(addrtype *)(&array)));		\
+		"	lctl	%1,%2,%0\n"			\
+		: : "Q" (*(addrtype *)(&array)),		\
+		    "i" (low), "i" (high));			\
 })
 
 #define __ctl_store(array, low, high) ({			\
 	typedef struct { char _[sizeof(array)]; } addrtype;	\
 	asm volatile(						\
-		"	stctl	%2,%3,0(%1)\n"			\
-		: "=m" (*(addrtype *)(&array))			\
-		: "a" (&array), "i" (low), "i" (high));		\
+		"	stctl	%1,%2,%0\n"			\
+		: "=Q" (*(addrtype *)(&array))			\
+		: "i" (low), "i" (high));			\
 	})
 
 #endif /* __s390x__ */
diff --git a/arch/s390/include/asm/thread_info.h b/arch/s390/include/asm/thread_info.h
index 66069e7..34f0873 100644
--- a/arch/s390/include/asm/thread_info.h
+++ b/arch/s390/include/asm/thread_info.h
@@ -73,7 +73,7 @@ struct thread_info {
 /* how to get the thread information struct from C */
 static inline struct thread_info *current_thread_info(void)
 {
-	return (struct thread_info *)((*(unsigned long *) __LC_KERNEL_STACK)-THREAD_SIZE);
+	return (struct thread_info *)(S390_lowcore.kernel_stack - THREAD_SIZE);
 }
 
 #define THREAD_SIZE_ORDER THREAD_ORDER
diff --git a/arch/s390/include/asm/timex.h b/arch/s390/include/asm/timex.h
index 68d9fea..f174bda 100644
--- a/arch/s390/include/asm/timex.h
+++ b/arch/s390/include/asm/timex.h
@@ -20,10 +20,10 @@ static inline int set_clock(__u64 time)
 	int cc;
 
 	asm volatile(
-		"   sck   0(%2)\n"
+		"   sck   %1\n"
 		"   ipm   %0\n"
 		"   srl   %0,28\n"
-		: "=d" (cc) : "m" (time), "a" (&time) : "cc");
+		: "=d" (cc) : "Q" (time) : "cc");
 	return cc;
 }
 
@@ -32,21 +32,21 @@ static inline int store_clock(__u64 *time)
 	int cc;
 
 	asm volatile(
-		"   stck  0(%2)\n"
+		"   stck  %1\n"
 		"   ipm   %0\n"
 		"   srl   %0,28\n"
-		: "=d" (cc), "=m" (*time) : "a" (time) : "cc");
+		: "=d" (cc), "=Q" (*time) : : "cc");
 	return cc;
 }
 
 static inline void set_clock_comparator(__u64 time)
 {
-	asm volatile("sckc 0(%1)" : : "m" (time), "a" (&time));
+	asm volatile("sckc %0" : : "Q" (time));
 }
 
 static inline void store_clock_comparator(__u64 *time)
 {
-	asm volatile("stckc 0(%1)" : "=m" (*time) : "a" (time));
+	asm volatile("stckc %0" : "=Q" (*time));
 }
 
 #define CLOCK_TICK_RATE	1193180 /* Underlying HZ */
@@ -57,11 +57,7 @@ static inline unsigned long long get_clock (void)
 {
 	unsigned long long clk;
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 	asm volatile("stck %0" : "=Q" (clk) : : "cc");
-#else /* __GNUC__ */
-	asm volatile("stck 0(%1)" : "=m" (clk) : "a" (&clk) : "cc");
-#endif /* __GNUC__ */
 	return clk;
 }
 
@@ -69,13 +65,7 @@ static inline unsigned long long get_clock_xt(void)
 {
 	unsigned char clk[16];
 
-#if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 	asm volatile("stcke %0" : "=Q" (clk) : : "cc");
-#else /* __GNUC__ */
-	asm volatile("stcke 0(%1)" : "=m" (clk)
-				   : "a" (clk) : "cc");
-#endif /* __GNUC__ */
-
 	return *((unsigned long long *)&clk[1]);
 }
 
diff --git a/arch/s390/include/asm/uaccess.h b/arch/s390/include/asm/uaccess.h
index cbf0a87..d6b1ed0 100644
--- a/arch/s390/include/asm/uaccess.h
+++ b/arch/s390/include/asm/uaccess.h
@@ -265,6 +265,12 @@ __copy_from_user(void *to, const void __user *from, unsigned long n)
 		return uaccess.copy_from_user(n, from, to);
 }
 
+extern void copy_from_user_overflow(void)
+#ifdef CONFIG_DEBUG_STRICT_USER_COPY_CHECKS
+__compiletime_warning("copy_from_user() buffer size is not provably correct")
+#endif
+;
+
 /**
  * copy_from_user: - Copy a block of data from user space.
  * @to:   Destination address, in kernel space.
@@ -284,7 +290,13 @@ __copy_from_user(void *to, const void __user *from, unsigned long n)
 static inline unsigned long __must_check
 copy_from_user(void *to, const void __user *from, unsigned long n)
 {
+	unsigned int sz = __compiletime_object_size(to);
+
 	might_fault();
+	if (unlikely(sz != -1 && sz < n)) {
+		copy_from_user_overflow();
+		return n;
+	}
 	if (access_ok(VERIFY_READ, from, n))
 		n = __copy_from_user(to, from, n);
 	else
diff --git a/arch/s390/include/asm/vdso.h b/arch/s390/include/asm/vdso.h
index 7bdd7c8..4a76d94 100644
--- a/arch/s390/include/asm/vdso.h
+++ b/arch/s390/include/asm/vdso.h
@@ -7,7 +7,7 @@
 #define VDSO32_LBASE	0
 #define VDSO64_LBASE	0
 
-#define VDSO_VERSION_STRING	LINUX_2.6.26
+#define VDSO_VERSION_STRING	LINUX_2.6.29
 
 #ifndef __ASSEMBLY__
 
diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile
index 683f638..64230bc 100644
--- a/arch/s390/kernel/Makefile
+++ b/arch/s390/kernel/Makefile
@@ -29,9 +29,12 @@ obj-y	+= $(if $(CONFIG_64BIT),entry64.o,entry.o)
 obj-y	+= $(if $(CONFIG_64BIT),reipl64.o,reipl.o)
 
 extra-y				+= head.o init_task.o vmlinux.lds
+extra-y				+= $(if $(CONFIG_64BIT),head64.o,head31.o)
 
 obj-$(CONFIG_MODULES)		+= s390_ksyms.o module.o
 obj-$(CONFIG_SMP)		+= smp.o topology.o
+obj-$(CONFIG_SMP)		+= $(if $(CONFIG_64BIT),switch_cpu64.o, \
+							switch_cpu.o)
 obj-$(CONFIG_HIBERNATION)	+= suspend.o swsusp_asm64.o
 obj-$(CONFIG_AUDIT)		+= audit.o
 compat-obj-$(CONFIG_AUDIT)	+= compat_audit.o
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index 63e4643..08db736 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -4,18 +4,27 @@
  * and format the required data.
  */
 
-#include <linux/sched.h>
+#define ASM_OFFSETS_C
+
 #include <linux/kbuild.h>
+#include <linux/sched.h>
 #include <asm/vdso.h>
 #include <asm/sigp.h>
 
+/*
+ * Make sure that the compiler is new enough. We want a compiler that
+ * is known to work with the "Q" assembler constraint.
+ */
+#if __GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 3)
+#error Your compiler is too old; please use version 3.3.3 or newer
+#endif
+
 int main(void)
 {
 	DEFINE(__THREAD_info, offsetof(struct task_struct, stack));
 	DEFINE(__THREAD_ksp, offsetof(struct task_struct, thread.ksp));
 	DEFINE(__THREAD_per, offsetof(struct task_struct, thread.per_info));
-	DEFINE(__THREAD_mm_segment,
-	       offsetof(struct task_struct, thread.mm_segment));
+	DEFINE(__THREAD_mm_segment, offsetof(struct task_struct, thread.mm_segment));
 	BLANK();
 	DEFINE(__TASK_pid, offsetof(struct task_struct, pid));
 	BLANK();
@@ -52,18 +61,94 @@ int main(void)
 	DEFINE(__VDSO_WTOM_NSEC, offsetof(struct vdso_data, wtom_clock_nsec));
 	DEFINE(__VDSO_TIMEZONE, offsetof(struct vdso_data, tz_minuteswest));
 	DEFINE(__VDSO_ECTG_OK, offsetof(struct vdso_data, ectg_available));
-	DEFINE(__VDSO_ECTG_BASE,
-	       offsetof(struct vdso_per_cpu_data, ectg_timer_base));
-	DEFINE(__VDSO_ECTG_USER,
-	       offsetof(struct vdso_per_cpu_data, ectg_user_time));
+	DEFINE(__VDSO_ECTG_BASE, offsetof(struct vdso_per_cpu_data, ectg_timer_base));
+	DEFINE(__VDSO_ECTG_USER, offsetof(struct vdso_per_cpu_data, ectg_user_time));
 	/* constants used by the vdso */
 	DEFINE(CLOCK_REALTIME, CLOCK_REALTIME);
 	DEFINE(CLOCK_MONOTONIC, CLOCK_MONOTONIC);
 	DEFINE(CLOCK_REALTIME_RES, MONOTONIC_RES_NSEC);
+	BLANK();
 	/* constants for SIGP */
 	DEFINE(__SIGP_STOP, sigp_stop);
 	DEFINE(__SIGP_RESTART, sigp_restart);
 	DEFINE(__SIGP_SENSE, sigp_sense);
 	DEFINE(__SIGP_INITIAL_CPU_RESET, sigp_initial_cpu_reset);
+	BLANK();
+	/* lowcore offsets */
+	DEFINE(__LC_EXT_PARAMS, offsetof(struct _lowcore, ext_params));
+	DEFINE(__LC_CPU_ADDRESS, offsetof(struct _lowcore, cpu_addr));
+	DEFINE(__LC_EXT_INT_CODE, offsetof(struct _lowcore, ext_int_code));
+	DEFINE(__LC_SVC_ILC, offsetof(struct _lowcore, svc_ilc));
+	DEFINE(__LC_SVC_INT_CODE, offsetof(struct _lowcore, svc_code));
+	DEFINE(__LC_PGM_ILC, offsetof(struct _lowcore, pgm_ilc));
+	DEFINE(__LC_PGM_INT_CODE, offsetof(struct _lowcore, pgm_code));
+	DEFINE(__LC_PER_ATMID, offsetof(struct _lowcore, per_perc_atmid));
+	DEFINE(__LC_PER_ADDRESS, offsetof(struct _lowcore, per_address));
+	DEFINE(__LC_PER_ACCESS_ID, offsetof(struct _lowcore, per_access_id));
+	DEFINE(__LC_AR_MODE_ID, offsetof(struct _lowcore, ar_access_id));
+	DEFINE(__LC_SUBCHANNEL_ID, offsetof(struct _lowcore, subchannel_id));
+	DEFINE(__LC_SUBCHANNEL_NR, offsetof(struct _lowcore, subchannel_nr));
+	DEFINE(__LC_IO_INT_PARM, offsetof(struct _lowcore, io_int_parm));
+	DEFINE(__LC_IO_INT_WORD, offsetof(struct _lowcore, io_int_word));
+	DEFINE(__LC_STFL_FAC_LIST, offsetof(struct _lowcore, stfl_fac_list));
+	DEFINE(__LC_MCCK_CODE, offsetof(struct _lowcore, mcck_interruption_code));
+	DEFINE(__LC_DUMP_REIPL, offsetof(struct _lowcore, ipib));
+	BLANK();
+	DEFINE(__LC_RST_NEW_PSW, offsetof(struct _lowcore, restart_psw));
+	DEFINE(__LC_RST_OLD_PSW, offsetof(struct _lowcore, restart_old_psw));
+	DEFINE(__LC_EXT_OLD_PSW, offsetof(struct _lowcore, external_old_psw));
+	DEFINE(__LC_SVC_OLD_PSW, offsetof(struct _lowcore, svc_old_psw));
+	DEFINE(__LC_PGM_OLD_PSW, offsetof(struct _lowcore, program_old_psw));
+	DEFINE(__LC_MCK_OLD_PSW, offsetof(struct _lowcore, mcck_old_psw));
+	DEFINE(__LC_IO_OLD_PSW, offsetof(struct _lowcore, io_old_psw));
+	DEFINE(__LC_EXT_NEW_PSW, offsetof(struct _lowcore, external_new_psw));
+	DEFINE(__LC_SVC_NEW_PSW, offsetof(struct _lowcore, svc_new_psw));
+	DEFINE(__LC_PGM_NEW_PSW, offsetof(struct _lowcore, program_new_psw));
+	DEFINE(__LC_MCK_NEW_PSW, offsetof(struct _lowcore, mcck_new_psw));
+	DEFINE(__LC_IO_NEW_PSW, offsetof(struct _lowcore, io_new_psw));
+	DEFINE(__LC_SAVE_AREA, offsetof(struct _lowcore, save_area));
+	DEFINE(__LC_RETURN_PSW, offsetof(struct _lowcore, return_psw));
+	DEFINE(__LC_RETURN_MCCK_PSW, offsetof(struct _lowcore, return_mcck_psw));
+	DEFINE(__LC_SYNC_ENTER_TIMER, offsetof(struct _lowcore, sync_enter_timer));
+	DEFINE(__LC_ASYNC_ENTER_TIMER, offsetof(struct _lowcore, async_enter_timer));
+	DEFINE(__LC_EXIT_TIMER, offsetof(struct _lowcore, exit_timer));
+	DEFINE(__LC_USER_TIMER, offsetof(struct _lowcore, user_timer));
+	DEFINE(__LC_SYSTEM_TIMER, offsetof(struct _lowcore, system_timer));
+	DEFINE(__LC_STEAL_TIMER, offsetof(struct _lowcore, steal_timer));
+	DEFINE(__LC_LAST_UPDATE_TIMER, offsetof(struct _lowcore, last_update_timer));
+	DEFINE(__LC_LAST_UPDATE_CLOCK, offsetof(struct _lowcore, last_update_clock));
+	DEFINE(__LC_CURRENT, offsetof(struct _lowcore, current_task));
+	DEFINE(__LC_THREAD_INFO, offsetof(struct _lowcore, thread_info));
+	DEFINE(__LC_KERNEL_STACK, offsetof(struct _lowcore, kernel_stack));
+	DEFINE(__LC_ASYNC_STACK, offsetof(struct _lowcore, async_stack));
+	DEFINE(__LC_PANIC_STACK, offsetof(struct _lowcore, panic_stack));
+	DEFINE(__LC_KERNEL_ASCE, offsetof(struct _lowcore, kernel_asce));
+	DEFINE(__LC_USER_ASCE, offsetof(struct _lowcore, user_asce));
+	DEFINE(__LC_USER_EXEC_ASCE, offsetof(struct _lowcore, user_exec_asce));
+	DEFINE(__LC_CPUID, offsetof(struct _lowcore, cpu_id));
+	DEFINE(__LC_INT_CLOCK, offsetof(struct _lowcore, int_clock));
+	DEFINE(__LC_MACHINE_FLAGS, offsetof(struct _lowcore, machine_flags));
+	DEFINE(__LC_FTRACE_FUNC, offsetof(struct _lowcore, ftrace_func));
+	DEFINE(__LC_IRB, offsetof(struct _lowcore, irb));
+	DEFINE(__LC_CPU_TIMER_SAVE_AREA, offsetof(struct _lowcore, cpu_timer_save_area));
+	DEFINE(__LC_CLOCK_COMP_SAVE_AREA, offsetof(struct _lowcore, clock_comp_save_area));
+	DEFINE(__LC_PSW_SAVE_AREA, offsetof(struct _lowcore, psw_save_area));
+	DEFINE(__LC_PREFIX_SAVE_AREA, offsetof(struct _lowcore, prefixreg_save_area));
+	DEFINE(__LC_AREGS_SAVE_AREA, offsetof(struct _lowcore, access_regs_save_area));
+	DEFINE(__LC_FPREGS_SAVE_AREA, offsetof(struct _lowcore, floating_pt_save_area));
+	DEFINE(__LC_GPREGS_SAVE_AREA, offsetof(struct _lowcore, gpregs_save_area));
+	DEFINE(__LC_CREGS_SAVE_AREA, offsetof(struct _lowcore, cregs_save_area));
+#ifdef CONFIG_32BIT
+	DEFINE(__LC_PFAULT_INTPARM, offsetof(struct _lowcore, ext_params));
+	DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, extended_save_area_addr));
+#else /* CONFIG_32BIT */
+	DEFINE(__LC_PFAULT_INTPARM, offsetof(struct _lowcore, ext_params2));
+	DEFINE(__LC_EXT_PARAMS2, offsetof(struct _lowcore, ext_params2));
+	DEFINE(SAVE_AREA_BASE, offsetof(struct _lowcore, floating_pt_save_area));
+	DEFINE(__LC_PASTE, offsetof(struct _lowcore, paste));
+	DEFINE(__LC_FP_CREG_SAVE_AREA, offsetof(struct _lowcore, fpt_creg_save_area));
+	DEFINE(__LC_LAST_BREAK, offsetof(struct _lowcore, breaking_event_addr));
+	DEFINE(__LC_VDSO_PER_CPU, offsetof(struct _lowcore, vdso_per_cpu_data));
+#endif /* CONFIG_32BIT */
 	return 0;
 }
diff --git a/arch/s390/kernel/base.S b/arch/s390/kernel/base.S
index dc7e525..15e46ca 100644
--- a/arch/s390/kernel/base.S
+++ b/arch/s390/kernel/base.S
@@ -6,8 +6,8 @@
  *		 Michael Holzheu <holzheu@de.ibm.com>
  */
 
+#include <asm/asm-offsets.h>
 #include <asm/ptrace.h>
-#include <asm/lowcore.h>
 
 #ifdef CONFIG_64BIT
 
diff --git a/arch/s390/kernel/dis.c b/arch/s390/kernel/dis.c
index db943a7..b39b27d 100644
--- a/arch/s390/kernel/dis.c
+++ b/arch/s390/kernel/dis.c
@@ -86,10 +86,17 @@ enum {
 	U4_12,	/* 4 bit unsigned value starting at 12 */
 	U4_16,	/* 4 bit unsigned value starting at 16 */
 	U4_20,	/* 4 bit unsigned value starting at 20 */
+	U4_32,	/* 4 bit unsigned value starting at 32 */
 	U8_8,	/* 8 bit unsigned value starting at 8 */
 	U8_16,	/* 8 bit unsigned value starting at 16 */
+	U8_24,	/* 8 bit unsigned value starting at 24 */
+	U8_32,	/* 8 bit unsigned value starting at 32 */
+	I8_8,	/* 8 bit signed value starting at 8 */
+	I8_32,	/* 8 bit signed value starting at 32 */
 	I16_16,	/* 16 bit signed value starting at 16 */
+	I16_32,	/* 32 bit signed value starting at 16 */
 	U16_16,	/* 16 bit unsigned value starting at 16 */
+	U16_32,	/* 32 bit unsigned value starting at 16 */
 	J16_16,	/* PC relative jump offset at 16 */
 	J32_16,	/* PC relative long offset at 16 */
 	I32_16,	/* 32 bit signed value starting at 16 */
@@ -104,21 +111,37 @@ enum {
  */
 enum {
 	INSTR_INVALID,
-	INSTR_E, INSTR_RIE_RRP, INSTR_RIL_RI, INSTR_RIL_RP, INSTR_RIL_RU,
-	INSTR_RIL_UP, INSTR_RI_RI, INSTR_RI_RP, INSTR_RI_RU, INSTR_RI_UP,
+	INSTR_E,
+	INSTR_RIE_R0IU, INSTR_RIE_R0UU, INSTR_RIE_RRP, INSTR_RIE_RRPU,
+	INSTR_RIE_RRUUU, INSTR_RIE_RUPI, INSTR_RIE_RUPU,
+	INSTR_RIL_RI, INSTR_RIL_RP, INSTR_RIL_RU, INSTR_RIL_UP,
+	INSTR_RIS_R0RDU, INSTR_RIS_R0UU, INSTR_RIS_RURDI, INSTR_RIS_RURDU,
+	INSTR_RI_RI, INSTR_RI_RP, INSTR_RI_RU, INSTR_RI_UP,
 	INSTR_RRE_00, INSTR_RRE_0R, INSTR_RRE_AA, INSTR_RRE_AR, INSTR_RRE_F0,
-	INSTR_RRE_FF, INSTR_RRE_R0, INSTR_RRE_RA, INSTR_RRE_RF, INSTR_RRE_RR,
-	INSTR_RRE_RR_OPT, INSTR_RRF_F0FF, INSTR_RRF_FUFF, INSTR_RRF_M0RR,
-	INSTR_RRF_R0RR, INSTR_RRF_RURR, INSTR_RRF_U0FF, INSTR_RRF_U0RF,
+	INSTR_RRE_FF, INSTR_RRE_FR, INSTR_RRE_R0, INSTR_RRE_RA, INSTR_RRE_RF,
+	INSTR_RRE_RR, INSTR_RRE_RR_OPT,
+	INSTR_RRF_0UFF, INSTR_RRF_F0FF, INSTR_RRF_F0FF2, INSTR_RRF_F0FR,
+	INSTR_RRF_FFRU, INSTR_RRF_FUFF, INSTR_RRF_M0RR, INSTR_RRF_R0RR,
+	INSTR_RRF_RURR, INSTR_RRF_U0FF, INSTR_RRF_U0RF, INSTR_RRF_U0RR,
+	INSTR_RRF_UUFF, INSTR_RRR_F0FF, INSTR_RRS_RRRDU,
 	INSTR_RR_FF, INSTR_RR_R0, INSTR_RR_RR, INSTR_RR_U0, INSTR_RR_UR,
-	INSTR_RSE_CCRD, INSTR_RSE_RRRD, INSTR_RSE_RURD, INSTR_RSI_RRP,
-	INSTR_RSL_R0RD, INSTR_RSY_AARD, INSTR_RSY_CCRD, INSTR_RSY_RRRD,
-	INSTR_RSY_RURD, INSTR_RS_AARD, INSTR_RS_CCRD, INSTR_RS_R0RD,
-	INSTR_RS_RRRD, INSTR_RS_RURD, INSTR_RXE_FRRD, INSTR_RXE_RRRD,
-	INSTR_RXF_FRRDF, INSTR_RXY_FRRD, INSTR_RXY_RRRD, INSTR_RX_FRRD,
-	INSTR_RX_RRRD, INSTR_RX_URRD, INSTR_SIY_URD, INSTR_SI_URD,
-	INSTR_SSE_RDRD, INSTR_SSF_RRDRD, INSTR_SS_L0RDRD, INSTR_SS_LIRDRD,
-	INSTR_SS_LLRDRD, INSTR_SS_RRRDRD, INSTR_SS_RRRDRD2, INSTR_SS_RRRDRD3,
+	INSTR_RSE_CCRD, INSTR_RSE_RRRD, INSTR_RSE_RURD,
+	INSTR_RSI_RRP,
+	INSTR_RSL_R0RD,
+	INSTR_RSY_AARD, INSTR_RSY_CCRD, INSTR_RSY_RRRD, INSTR_RSY_RURD,
+	INSTR_RS_AARD, INSTR_RS_CCRD, INSTR_RS_R0RD, INSTR_RS_RRRD,
+	INSTR_RS_RURD,
+	INSTR_RXE_FRRD, INSTR_RXE_RRRD,
+	INSTR_RXF_FRRDF,
+	INSTR_RXY_FRRD, INSTR_RXY_RRRD, INSTR_RXY_URRD,
+	INSTR_RX_FRRD, INSTR_RX_RRRD, INSTR_RX_URRD,
+	INSTR_SIL_RDI, INSTR_SIL_RDU,
+	INSTR_SIY_IRD, INSTR_SIY_URD,
+	INSTR_SI_URD,
+	INSTR_SSE_RDRD,
+	INSTR_SSF_RRDRD,
+	INSTR_SS_L0RDRD, INSTR_SS_LIRDRD, INSTR_SS_LLRDRD, INSTR_SS_RRRDRD,
+	INSTR_SS_RRRDRD2, INSTR_SS_RRRDRD3,
 	INSTR_S_00, INSTR_S_RD,
 };
 
@@ -129,7 +152,7 @@ struct operand {
 };
 
 struct insn {
-	const char name[5];
+	const char name[6];
 	unsigned char opfrag;
 	unsigned char format;
 };
@@ -170,11 +193,16 @@ static const struct operand operands[] =
 	[U4_12]  = {  4, 12, 0 },
 	[U4_16]  = {  4, 16, 0 },
 	[U4_20]  = {  4, 20, 0 },
+	[U4_32]  = {  4, 32, 0 },
 	[U8_8]	 = {  8,  8, 0 },
 	[U8_16]  = {  8, 16, 0 },
+	[U8_24]  = {  8, 24, 0 },
+	[U8_32]  = {  8, 32, 0 },
 	[I16_16] = { 16, 16, OPERAND_SIGNED },
 	[U16_16] = { 16, 16, 0 },
+	[U16_32] = { 16, 32, 0 },
 	[J16_16] = { 16, 16, OPERAND_PCREL },
+	[I16_32] = { 16, 32, OPERAND_SIGNED },
 	[J32_16] = { 32, 16, OPERAND_PCREL },
 	[I32_16] = { 32, 16, OPERAND_SIGNED },
 	[U32_16] = { 32, 16, 0 },
@@ -183,82 +211,93 @@ static const struct operand operands[] =
 };
 
 static const unsigned char formats[][7] = {
-	[INSTR_E]	  = { 0xff, 0,0,0,0,0,0 },	       /* e.g. pr    */
-	[INSTR_RIE_RRP]	  = { 0xff, R_8,R_12,J16_16,0,0,0 },   /* e.g. brxhg */
-	[INSTR_RIL_RP]	  = { 0x0f, R_8,J32_16,0,0,0,0 },      /* e.g. brasl */
-	[INSTR_RIL_UP]	  = { 0x0f, U4_8,J32_16,0,0,0,0 },     /* e.g. brcl  */
-	[INSTR_RIL_RI]	  = { 0x0f, R_8,I32_16,0,0,0,0 },      /* e.g. afi   */
-	[INSTR_RIL_RU]	  = { 0x0f, R_8,U32_16,0,0,0,0 },      /* e.g. alfi  */
-	[INSTR_RI_RI]	  = { 0x0f, R_8,I16_16,0,0,0,0 },      /* e.g. ahi   */
-	[INSTR_RI_RP]	  = { 0x0f, R_8,J16_16,0,0,0,0 },      /* e.g. brct  */
-	[INSTR_RI_RU]	  = { 0x0f, R_8,U16_16,0,0,0,0 },      /* e.g. tml   */
-	[INSTR_RI_UP]	  = { 0x0f, U4_8,J16_16,0,0,0,0 },     /* e.g. brc   */
-	[INSTR_RRE_00]	  = { 0xff, 0,0,0,0,0,0 },	       /* e.g. palb  */
-	[INSTR_RRE_0R]	  = { 0xff, R_28,0,0,0,0,0 },	       /* e.g. tb    */
-	[INSTR_RRE_AA]	  = { 0xff, A_24,A_28,0,0,0,0 },       /* e.g. cpya  */
-	[INSTR_RRE_AR]	  = { 0xff, A_24,R_28,0,0,0,0 },       /* e.g. sar   */
-	[INSTR_RRE_F0]	  = { 0xff, F_24,0,0,0,0,0 },	       /* e.g. sqer  */
-	[INSTR_RRE_FF]	  = { 0xff, F_24,F_28,0,0,0,0 },       /* e.g. debr  */
-	[INSTR_RRE_R0]	  = { 0xff, R_24,0,0,0,0,0 },	       /* e.g. ipm   */
-	[INSTR_RRE_RA]	  = { 0xff, R_24,A_28,0,0,0,0 },       /* e.g. ear   */
-	[INSTR_RRE_RF]	  = { 0xff, R_24,F_28,0,0,0,0 },       /* e.g. cefbr */
-	[INSTR_RRE_RR]	  = { 0xff, R_24,R_28,0,0,0,0 },       /* e.g. lura  */
-	[INSTR_RRE_RR_OPT]= { 0xff, R_24,RO_28,0,0,0,0 },      /* efpc, sfpc */
-	[INSTR_RRF_F0FF]  = { 0xff, F_16,F_24,F_28,0,0,0 },    /* e.g. madbr */
-	[INSTR_RRF_FUFF]  = { 0xff, F_24,F_16,F_28,U4_20,0,0 },/* e.g. didbr */
-	[INSTR_RRF_RURR]  = { 0xff, R_24,R_28,R_16,U4_20,0,0 },/* e.g. .insn */
-	[INSTR_RRF_R0RR]  = { 0xff, R_24,R_16,R_28,0,0,0 },    /* e.g. idte  */
-	[INSTR_RRF_U0FF]  = { 0xff, F_24,U4_16,F_28,0,0,0 },   /* e.g. fixr  */
-	[INSTR_RRF_U0RF]  = { 0xff, R_24,U4_16,F_28,0,0,0 },   /* e.g. cfebr */
-	[INSTR_RRF_M0RR]  = { 0xff, R_24,R_28,M_16,0,0,0 },    /* e.g. sske  */
-	[INSTR_RR_FF]	  = { 0xff, F_8,F_12,0,0,0,0 },        /* e.g. adr   */
-	[INSTR_RR_R0]	  = { 0xff, R_8, 0,0,0,0,0 },	       /* e.g. spm   */
-	[INSTR_RR_RR]	  = { 0xff, R_8,R_12,0,0,0,0 },        /* e.g. lr    */
-	[INSTR_RR_U0]	  = { 0xff, U8_8, 0,0,0,0,0 },	       /* e.g. svc   */
-	[INSTR_RR_UR]	  = { 0xff, U4_8,R_12,0,0,0,0 },       /* e.g. bcr   */
-	[INSTR_RSE_RRRD]  = { 0xff, R_8,R_12,D_20,B_16,0,0 },  /* e.g. lmh   */
-	[INSTR_RSE_CCRD]  = { 0xff, C_8,C_12,D_20,B_16,0,0 },  /* e.g. lmh   */
-	[INSTR_RSE_RURD]  = { 0xff, R_8,U4_12,D_20,B_16,0,0 }, /* e.g. icmh  */
-	[INSTR_RSL_R0RD]  = { 0xff, R_8,D_20,B_16,0,0,0 },     /* e.g. tp    */
-	[INSTR_RSI_RRP]	  = { 0xff, R_8,R_12,J16_16,0,0,0 },   /* e.g. brxh  */
-	[INSTR_RSY_RRRD]  = { 0xff, R_8,R_12,D20_20,B_16,0,0 },/* e.g. stmy  */
+	[INSTR_E]	  = { 0xff, 0,0,0,0,0,0 },
+	[INSTR_RIE_R0UU]  = { 0xff, R_8,U16_16,U4_32,0,0,0 },
+	[INSTR_RIE_RRPU]  = { 0xff, R_8,R_12,U4_32,J16_16,0,0 },
+	[INSTR_RIE_RRP]	  = { 0xff, R_8,R_12,J16_16,0,0,0 },
+	[INSTR_RIE_RRUUU] = { 0xff, R_8,R_12,U8_16,U8_24,U8_32,0 },
+	[INSTR_RIE_RUPI]  = { 0xff, R_8,I8_32,U4_12,J16_16,0,0 },
+	[INSTR_RIL_RI]	  = { 0x0f, R_8,I32_16,0,0,0,0 },
+	[INSTR_RIL_RP]	  = { 0x0f, R_8,J32_16,0,0,0,0 },
+	[INSTR_RIL_RU]	  = { 0x0f, R_8,U32_16,0,0,0,0 },
+	[INSTR_RIL_UP]	  = { 0x0f, U4_8,J32_16,0,0,0,0 },
+	[INSTR_RIS_R0RDU] = { 0xff, R_8,U8_32,D_20,B_16,0,0 },
+	[INSTR_RIS_RURDI] = { 0xff, R_8,I8_32,U4_12,D_20,B_16,0 },
+	[INSTR_RIS_RURDU] = { 0xff, R_8,U8_32,U4_12,D_20,B_16,0 },
+	[INSTR_RI_RI]	  = { 0x0f, R_8,I16_16,0,0,0,0 },
+	[INSTR_RI_RP]	  = { 0x0f, R_8,J16_16,0,0,0,0 },
+	[INSTR_RI_RU]	  = { 0x0f, R_8,U16_16,0,0,0,0 },
+	[INSTR_RI_UP]	  = { 0x0f, U4_8,J16_16,0,0,0,0 },
+	[INSTR_RRE_00]	  = { 0xff, 0,0,0,0,0,0 },
+	[INSTR_RRE_0R]	  = { 0xff, R_28,0,0,0,0,0 },
+	[INSTR_RRE_AA]	  = { 0xff, A_24,A_28,0,0,0,0 },
+	[INSTR_RRE_AR]	  = { 0xff, A_24,R_28,0,0,0,0 },
+	[INSTR_RRE_F0]	  = { 0xff, F_24,0,0,0,0,0 },
+	[INSTR_RRE_FF]	  = { 0xff, F_24,F_28,0,0,0,0 },
+	[INSTR_RRE_FR]	  = { 0xff, F_24,R_28,0,0,0,0 },
+	[INSTR_RRE_R0]	  = { 0xff, R_24,0,0,0,0,0 },
+	[INSTR_RRE_RA]	  = { 0xff, R_24,A_28,0,0,0,0 },
+	[INSTR_RRE_RF]	  = { 0xff, R_24,F_28,0,0,0,0 },
+	[INSTR_RRE_RR]	  = { 0xff, R_24,R_28,0,0,0,0 },
+	[INSTR_RRE_RR_OPT]= { 0xff, R_24,RO_28,0,0,0,0 },
+	[INSTR_RRF_0UFF]  = { 0xff, F_24,F_28,U4_20,0,0,0 },
+	[INSTR_RRF_F0FF2] = { 0xff, F_24,F_16,F_28,0,0,0 },
+	[INSTR_RRF_F0FF]  = { 0xff, F_16,F_24,F_28,0,0,0 },
+	[INSTR_RRF_F0FR]  = { 0xff, F_24,F_16,R_28,0,0,0 },
+	[INSTR_RRF_FFRU]  = { 0xff, F_24,F_16,R_28,U4_20,0,0 },
+	[INSTR_RRF_FUFF]  = { 0xff, F_24,F_16,F_28,U4_20,0,0 },
+	[INSTR_RRF_M0RR]  = { 0xff, R_24,R_28,M_16,0,0,0 },
+	[INSTR_RRF_R0RR]  = { 0xff, R_24,R_16,R_28,0,0,0 },
+	[INSTR_RRF_RURR]  = { 0xff, R_24,R_28,R_16,U4_20,0,0 },
+	[INSTR_RRF_U0FF]  = { 0xff, F_24,U4_16,F_28,0,0,0 },
+	[INSTR_RRF_U0RF]  = { 0xff, R_24,U4_16,F_28,0,0,0 },
+	[INSTR_RRF_U0RR]  = { 0xff, R_24,R_28,U4_16,0,0,0 },
+	[INSTR_RRF_UUFF]  = { 0xff, F_24,U4_16,F_28,U4_20,0,0 },
+	[INSTR_RRR_F0FF]  = { 0xff, F_24,F_28,F_16,0,0,0 },
+	[INSTR_RRS_RRRDU] = { 0xff, R_8,R_12,U4_32,D_20,B_16,0 },
+	[INSTR_RR_FF]	  = { 0xff, F_8,F_12,0,0,0,0 },
+	[INSTR_RR_R0]	  = { 0xff, R_8, 0,0,0,0,0 },
+	[INSTR_RR_RR]	  = { 0xff, R_8,R_12,0,0,0,0 },
+	[INSTR_RR_U0]	  = { 0xff, U8_8, 0,0,0,0,0 },
+	[INSTR_RR_UR]	  = { 0xff, U4_8,R_12,0,0,0,0 },
+	[INSTR_RSE_CCRD]  = { 0xff, C_8,C_12,D_20,B_16,0,0 },
+	[INSTR_RSE_RRRD]  = { 0xff, R_8,R_12,D_20,B_16,0,0 },
+	[INSTR_RSE_RURD]  = { 0xff, R_8,U4_12,D_20,B_16,0,0 },
+	[INSTR_RSI_RRP]	  = { 0xff, R_8,R_12,J16_16,0,0,0 },
+	[INSTR_RSL_R0RD]  = { 0xff, D_20,L4_8,B_16,0,0,0 },
+	[INSTR_RSY_AARD]  = { 0xff, A_8,A_12,D20_20,B_16,0,0 },
+	[INSTR_RSY_CCRD]  = { 0xff, C_8,C_12,D20_20,B_16,0,0 },
+	[INSTR_RSY_RRRD]  = { 0xff, R_8,R_12,D20_20,B_16,0,0 },
 	[INSTR_RSY_RURD]  = { 0xff, R_8,U4_12,D20_20,B_16,0,0 },
-							       /* e.g. icmh  */
-	[INSTR_RSY_AARD]  = { 0xff, A_8,A_12,D20_20,B_16,0,0 },/* e.g. lamy  */
-	[INSTR_RSY_CCRD]  = { 0xff, C_8,C_12,D20_20,B_16,0,0 },/* e.g. lamy  */
-	[INSTR_RS_AARD]	  = { 0xff, A_8,A_12,D_20,B_16,0,0 },  /* e.g. lam   */
-	[INSTR_RS_CCRD]	  = { 0xff, C_8,C_12,D_20,B_16,0,0 },  /* e.g. lctl  */
-	[INSTR_RS_R0RD]	  = { 0xff, R_8,D_20,B_16,0,0,0 },     /* e.g. sll   */
-	[INSTR_RS_RRRD]	  = { 0xff, R_8,R_12,D_20,B_16,0,0 },  /* e.g. cs    */
-	[INSTR_RS_RURD]	  = { 0xff, R_8,U4_12,D_20,B_16,0,0 }, /* e.g. icm   */
-	[INSTR_RXE_FRRD]  = { 0xff, F_8,D_20,X_12,B_16,0,0 },  /* e.g. axbr  */
-	[INSTR_RXE_RRRD]  = { 0xff, R_8,D_20,X_12,B_16,0,0 },  /* e.g. lg    */
+	[INSTR_RS_AARD]	  = { 0xff, A_8,A_12,D_20,B_16,0,0 },
+	[INSTR_RS_CCRD]	  = { 0xff, C_8,C_12,D_20,B_16,0,0 },
+	[INSTR_RS_R0RD]	  = { 0xff, R_8,D_20,B_16,0,0,0 },
+	[INSTR_RS_RRRD]	  = { 0xff, R_8,R_12,D_20,B_16,0,0 },
+	[INSTR_RS_RURD]	  = { 0xff, R_8,U4_12,D_20,B_16,0,0 },
+	[INSTR_RXE_FRRD]  = { 0xff, F_8,D_20,X_12,B_16,0,0 },
+	[INSTR_RXE_RRRD]  = { 0xff, R_8,D_20,X_12,B_16,0,0 },
 	[INSTR_RXF_FRRDF] = { 0xff, F_32,F_8,D_20,X_12,B_16,0 },
-							       /* e.g. madb  */
-	[INSTR_RXY_RRRD]  = { 0xff, R_8,D20_20,X_12,B_16,0,0 },/* e.g. ly    */
-	[INSTR_RXY_FRRD]  = { 0xff, F_8,D20_20,X_12,B_16,0,0 },/* e.g. ley   */
-	[INSTR_RX_FRRD]	  = { 0xff, F_8,D_20,X_12,B_16,0,0 },  /* e.g. ae    */
-	[INSTR_RX_RRRD]	  = { 0xff, R_8,D_20,X_12,B_16,0,0 },  /* e.g. l     */
-	[INSTR_RX_URRD]	  = { 0xff, U4_8,D_20,X_12,B_16,0,0 }, /* e.g. bc    */
-	[INSTR_SI_URD]	  = { 0xff, D_20,B_16,U8_8,0,0,0 },    /* e.g. cli   */
-	[INSTR_SIY_URD]	  = { 0xff, D20_20,B_16,U8_8,0,0,0 },  /* e.g. tmy   */
-	[INSTR_SSE_RDRD]  = { 0xff, D_20,B_16,D_36,B_32,0,0 }, /* e.g. mvsdk */
+	[INSTR_RXY_FRRD]  = { 0xff, F_8,D20_20,X_12,B_16,0,0 },
+	[INSTR_RXY_RRRD]  = { 0xff, R_8,D20_20,X_12,B_16,0,0 },
+	[INSTR_RXY_URRD]  = { 0xff, U4_8,D20_20,X_12,B_16,0,0 },
+	[INSTR_RX_FRRD]	  = { 0xff, F_8,D_20,X_12,B_16,0,0 },
+	[INSTR_RX_RRRD]	  = { 0xff, R_8,D_20,X_12,B_16,0,0 },
+	[INSTR_RX_URRD]	  = { 0xff, U4_8,D_20,X_12,B_16,0,0 },
+	[INSTR_SIL_RDI]   = { 0xff, D_20,B_16,I16_32,0,0,0 },
+	[INSTR_SIL_RDU]   = { 0xff, D_20,B_16,U16_32,0,0,0 },
+	[INSTR_SIY_IRD]   = { 0xff, D20_20,B_16,I8_8,0,0,0 },
+	[INSTR_SIY_URD]	  = { 0xff, D20_20,B_16,U8_8,0,0,0 },
+	[INSTR_SI_URD]	  = { 0xff, D_20,B_16,U8_8,0,0,0 },
+	[INSTR_SSE_RDRD]  = { 0xff, D_20,B_16,D_36,B_32,0,0 },
+	[INSTR_SSF_RRDRD] = { 0x00, D_20,B_16,D_36,B_32,R_8,0 },
 	[INSTR_SS_L0RDRD] = { 0xff, D_20,L8_8,B_16,D_36,B_32,0 },
-							       /* e.g. mvc   */
 	[INSTR_SS_LIRDRD] = { 0xff, D_20,L4_8,B_16,D_36,B_32,U4_12 },
-							       /* e.g. srp   */
 	[INSTR_SS_LLRDRD] = { 0xff, D_20,L4_8,B_16,D_36,L4_12,B_32 },
-							       /* e.g. pack  */
-	[INSTR_SS_RRRDRD] = { 0xff, D_20,R_8,B_16,D_36,B_32,R_12 },
-							       /* e.g. mvck  */
 	[INSTR_SS_RRRDRD2]= { 0xff, R_8,D_20,B_16,R_12,D_36,B_32 },
-							       /* e.g. plo   */
 	[INSTR_SS_RRRDRD3]= { 0xff, R_8,R_12,D_20,B_16,D_36,B_32 },
-							       /* e.g. lmd   */
-	[INSTR_S_00]	  = { 0xff, 0,0,0,0,0,0 },	       /* e.g. hsch  */
-	[INSTR_S_RD]	  = { 0xff, D_20,B_16,0,0,0,0 },       /* e.g. lpsw  */
-	[INSTR_SSF_RRDRD] = { 0x00, D_20,B_16,D_36,B_32,R_8,0 },
-							       /* e.g. mvcos */
+	[INSTR_SS_RRRDRD] = { 0xff, D_20,R_8,B_16,D_36,B_32,R_12 },
+	[INSTR_S_00]	  = { 0xff, 0,0,0,0,0,0 },
+	[INSTR_S_RD]	  = { 0xff, D_20,B_16,0,0,0,0 },
 };
 
 static struct insn opcode[] = {
@@ -454,6 +493,8 @@ static struct insn opcode[] = {
 static struct insn opcode_01[] = {
 #ifdef CONFIG_64BIT
 	{ "sam64", 0x0e, INSTR_E },
+	{ "pfpo", 0x0a, INSTR_E },
+	{ "ptff", 0x04, INSTR_E },
 #endif
 	{ "pr", 0x01, INSTR_E },
 	{ "upt", 0x02, INSTR_E },
@@ -519,6 +560,8 @@ static struct insn opcode_b2[] = {
 	{ "cutfu", 0xa7, INSTR_RRF_M0RR },
 	{ "stfle", 0xb0, INSTR_S_RD },
 	{ "lpswe", 0xb2, INSTR_S_RD },
+	{ "srnmt", 0xb9, INSTR_S_RD },
+	{ "lfas", 0xbd, INSTR_S_RD },
 #endif
 	{ "stidp", 0x02, INSTR_S_RD },
 	{ "sck", 0x04, INSTR_S_RD },
@@ -589,7 +632,6 @@ static struct insn opcode_b2[] = {
 	{ "clst", 0x5d, INSTR_RRE_RR },
 	{ "srst", 0x5e, INSTR_RRE_RR },
 	{ "cmpsc", 0x63, INSTR_RRE_RR },
-	{ "cmpsc", 0x63, INSTR_RRE_RR },
 	{ "siga", 0x74, INSTR_S_RD },
 	{ "xsch", 0x76, INSTR_S_00 },
 	{ "rp", 0x77, INSTR_S_RD },
@@ -630,6 +672,57 @@ static struct insn opcode_b3[] = {
 	{ "cger", 0xc8, INSTR_RRF_U0RF },
 	{ "cgdr", 0xc9, INSTR_RRF_U0RF },
 	{ "cgxr", 0xca, INSTR_RRF_U0RF },
+	{ "lpdfr", 0x70, INSTR_RRE_FF },
+	{ "lndfr", 0x71, INSTR_RRE_FF },
+	{ "cpsdr", 0x72, INSTR_RRF_F0FF2 },
+	{ "lcdfr", 0x73, INSTR_RRE_FF },
+	{ "ldgr", 0xc1, INSTR_RRE_FR },
+	{ "lgdr", 0xcd, INSTR_RRE_RF },
+	{ "adtr", 0xd2, INSTR_RRR_F0FF },
+	{ "axtr", 0xda, INSTR_RRR_F0FF },
+	{ "cdtr", 0xe4, INSTR_RRE_FF },
+	{ "cxtr", 0xec, INSTR_RRE_FF },
+	{ "kdtr", 0xe0, INSTR_RRE_FF },
+	{ "kxtr", 0xe8, INSTR_RRE_FF },
+	{ "cedtr", 0xf4, INSTR_RRE_FF },
+	{ "cextr", 0xfc, INSTR_RRE_FF },
+	{ "cdgtr", 0xf1, INSTR_RRE_FR },
+	{ "cxgtr", 0xf9, INSTR_RRE_FR },
+	{ "cdstr", 0xf3, INSTR_RRE_FR },
+	{ "cxstr", 0xfb, INSTR_RRE_FR },
+	{ "cdutr", 0xf2, INSTR_RRE_FR },
+	{ "cxutr", 0xfa, INSTR_RRE_FR },
+	{ "cgdtr", 0xe1, INSTR_RRF_U0RF },
+	{ "cgxtr", 0xe9, INSTR_RRF_U0RF },
+	{ "csdtr", 0xe3, INSTR_RRE_RF },
+	{ "csxtr", 0xeb, INSTR_RRE_RF },
+	{ "cudtr", 0xe2, INSTR_RRE_RF },
+	{ "cuxtr", 0xea, INSTR_RRE_RF },
+	{ "ddtr", 0xd1, INSTR_RRR_F0FF },
+	{ "dxtr", 0xd9, INSTR_RRR_F0FF },
+	{ "eedtr", 0xe5, INSTR_RRE_RF },
+	{ "eextr", 0xed, INSTR_RRE_RF },
+	{ "esdtr", 0xe7, INSTR_RRE_RF },
+	{ "esxtr", 0xef, INSTR_RRE_RF },
+	{ "iedtr", 0xf6, INSTR_RRF_F0FR },
+	{ "iextr", 0xfe, INSTR_RRF_F0FR },
+	{ "ltdtr", 0xd6, INSTR_RRE_FF },
+	{ "ltxtr", 0xde, INSTR_RRE_FF },
+	{ "fidtr", 0xd7, INSTR_RRF_UUFF },
+	{ "fixtr", 0xdf, INSTR_RRF_UUFF },
+	{ "ldetr", 0xd4, INSTR_RRF_0UFF },
+	{ "lxdtr", 0xdc, INSTR_RRF_0UFF },
+	{ "ledtr", 0xd5, INSTR_RRF_UUFF },
+	{ "ldxtr", 0xdd, INSTR_RRF_UUFF },
+	{ "mdtr", 0xd0, INSTR_RRR_F0FF },
+	{ "mxtr", 0xd8, INSTR_RRR_F0FF },
+	{ "qadtr", 0xf5, INSTR_RRF_FUFF },
+	{ "qaxtr", 0xfd, INSTR_RRF_FUFF },
+	{ "rrdtr", 0xf7, INSTR_RRF_FFRU },
+	{ "rrxtr", 0xff, INSTR_RRF_FFRU },
+	{ "sfasr", 0x85, INSTR_RRE_R0 },
+	{ "sdtr", 0xd3, INSTR_RRR_F0FF },
+	{ "sxtr", 0xdb, INSTR_RRR_F0FF },
 #endif
 	{ "lpebr", 0x00, INSTR_RRE_FF },
 	{ "lnebr", 0x01, INSTR_RRE_FF },
@@ -780,6 +873,14 @@ static struct insn opcode_b9[] = {
 	{ "cu24", 0xb1, INSTR_RRF_M0RR },
 	{ "cu41", 0xb2, INSTR_RRF_M0RR },
 	{ "cu42", 0xb3, INSTR_RRF_M0RR },
+	{ "crt", 0x72, INSTR_RRF_U0RR },
+	{ "cgrt", 0x60, INSTR_RRF_U0RR },
+	{ "clrt", 0x73, INSTR_RRF_U0RR },
+	{ "clgrt", 0x61, INSTR_RRF_U0RR },
+	{ "ptf", 0xa2, INSTR_RRE_R0 },
+	{ "pfmf", 0xaf, INSTR_RRE_RR },
+	{ "trte", 0xbf, INSTR_RRF_M0RR },
+	{ "trtre", 0xbd, INSTR_RRF_M0RR },
 #endif
 	{ "kmac", 0x1e, INSTR_RRE_RR },
 	{ "lrvr", 0x1f, INSTR_RRE_RR },
@@ -835,6 +936,43 @@ static struct insn opcode_c2[] = {
 	{ "cfi", 0x0d, INSTR_RIL_RI },
 	{ "clgfi", 0x0e, INSTR_RIL_RU },
 	{ "clfi", 0x0f, INSTR_RIL_RU },
+	{ "msfi", 0x01, INSTR_RIL_RI },
+	{ "msgfi", 0x00, INSTR_RIL_RI },
+#endif
+	{ "", 0, INSTR_INVALID }
+};
+
+static struct insn opcode_c4[] = {
+#ifdef CONFIG_64BIT
+	{ "lrl", 0x0d, INSTR_RIL_RP },
+	{ "lgrl", 0x08, INSTR_RIL_RP },
+	{ "lgfrl", 0x0c, INSTR_RIL_RP },
+	{ "lhrl", 0x05, INSTR_RIL_RP },
+	{ "lghrl", 0x04, INSTR_RIL_RP },
+	{ "llgfrl", 0x0e, INSTR_RIL_RP },
+	{ "llhrl", 0x02, INSTR_RIL_RP },
+	{ "llghrl", 0x06, INSTR_RIL_RP },
+	{ "strl", 0x0f, INSTR_RIL_RP },
+	{ "stgrl", 0x0b, INSTR_RIL_RP },
+	{ "sthrl", 0x07, INSTR_RIL_RP },
+#endif
+	{ "", 0, INSTR_INVALID }
+};
+
+static struct insn opcode_c6[] = {
+#ifdef CONFIG_64BIT
+	{ "crl", 0x0d, INSTR_RIL_RP },
+	{ "cgrl", 0x08, INSTR_RIL_RP },
+	{ "cgfrl", 0x0c, INSTR_RIL_RP },
+	{ "chrl", 0x05, INSTR_RIL_RP },
+	{ "cghrl", 0x04, INSTR_RIL_RP },
+	{ "clrl", 0x0f, INSTR_RIL_RP },
+	{ "clgrl", 0x0a, INSTR_RIL_RP },
+	{ "clgfrl", 0x0e, INSTR_RIL_RP },
+	{ "clhrl", 0x07, INSTR_RIL_RP },
+	{ "clghrl", 0x06, INSTR_RIL_RP },
+	{ "pfdrl", 0x02, INSTR_RIL_UP },
+	{ "exrl", 0x00, INSTR_RIL_RP },
 #endif
 	{ "", 0, INSTR_INVALID }
 };
@@ -842,6 +980,8 @@ static struct insn opcode_c2[] = {
 static struct insn opcode_c8[] = {
 #ifdef CONFIG_64BIT
 	{ "mvcos", 0x00, INSTR_SSF_RRDRD },
+	{ "ectg", 0x01, INSTR_SSF_RRDRD },
+	{ "csst", 0x02, INSTR_SSF_RRDRD },
 #endif
 	{ "", 0, INSTR_INVALID }
 };
@@ -917,6 +1057,12 @@ static struct insn opcode_e3[] = {
 	{ "llgh", 0x91, INSTR_RXY_RRRD },
 	{ "llc", 0x94, INSTR_RXY_RRRD },
 	{ "llh", 0x95, INSTR_RXY_RRRD },
+	{ "cgh", 0x34, INSTR_RXY_RRRD },
+	{ "laey", 0x75, INSTR_RXY_RRRD },
+	{ "ltgf", 0x32, INSTR_RXY_RRRD },
+	{ "mfy", 0x5c, INSTR_RXY_RRRD },
+	{ "mhy", 0x7c, INSTR_RXY_RRRD },
+	{ "pfd", 0x36, INSTR_RXY_URRD },
 #endif
 	{ "lrv", 0x1e, INSTR_RXY_RRRD },
 	{ "lrvh", 0x1f, INSTR_RXY_RRRD },
@@ -931,6 +1077,15 @@ static struct insn opcode_e3[] = {
 static struct insn opcode_e5[] = {
 #ifdef CONFIG_64BIT
 	{ "strag", 0x02, INSTR_SSE_RDRD },
+	{ "chhsi", 0x54, INSTR_SIL_RDI },
+	{ "chsi", 0x5c, INSTR_SIL_RDI },
+	{ "cghsi", 0x58, INSTR_SIL_RDI },
+	{ "clhhsi", 0x55, INSTR_SIL_RDU },
+	{ "clfhsi", 0x5d, INSTR_SIL_RDU },
+	{ "clghsi", 0x59, INSTR_SIL_RDU },
+	{ "mvhhi", 0x44, INSTR_SIL_RDI },
+	{ "mvhi", 0x4c, INSTR_SIL_RDI },
+	{ "mvghi", 0x48, INSTR_SIL_RDI },
 #endif
 	{ "lasp", 0x00, INSTR_SSE_RDRD },
 	{ "tprot", 0x01, INSTR_SSE_RDRD },
@@ -977,6 +1132,11 @@ static struct insn opcode_eb[] = {
 	{ "lmy", 0x98, INSTR_RSY_RRRD },
 	{ "lamy", 0x9a, INSTR_RSY_AARD },
 	{ "stamy", 0x9b, INSTR_RSY_AARD },
+	{ "asi", 0x6a, INSTR_SIY_IRD },
+	{ "agsi", 0x7a, INSTR_SIY_IRD },
+	{ "alsi", 0x6e, INSTR_SIY_IRD },
+	{ "algsi", 0x7e, INSTR_SIY_IRD },
+	{ "ecag", 0x4c, INSTR_RSY_RRRD },
 #endif
 	{ "rll", 0x1d, INSTR_RSY_RRRD },
 	{ "mvclu", 0x8e, INSTR_RSY_RRRD },
@@ -988,6 +1148,30 @@ static struct insn opcode_ec[] = {
 #ifdef CONFIG_64BIT
 	{ "brxhg", 0x44, INSTR_RIE_RRP },
 	{ "brxlg", 0x45, INSTR_RIE_RRP },
+	{ "crb", 0xf6, INSTR_RRS_RRRDU },
+	{ "cgrb", 0xe4, INSTR_RRS_RRRDU },
+	{ "crj", 0x76, INSTR_RIE_RRPU },
+	{ "cgrj", 0x64, INSTR_RIE_RRPU },
+	{ "cib", 0xfe, INSTR_RIS_RURDI },
+	{ "cgib", 0xfc, INSTR_RIS_RURDI },
+	{ "cij", 0x7e, INSTR_RIE_RUPI },
+	{ "cgij", 0x7c, INSTR_RIE_RUPI },
+	{ "cit", 0x72, INSTR_RIE_R0IU },
+	{ "cgit", 0x70, INSTR_RIE_R0IU },
+	{ "clrb", 0xf7, INSTR_RRS_RRRDU },
+	{ "clgrb", 0xe5, INSTR_RRS_RRRDU },
+	{ "clrj", 0x77, INSTR_RIE_RRPU },
+	{ "clgrj", 0x65, INSTR_RIE_RRPU },
+	{ "clib", 0xff, INSTR_RIS_RURDU },
+	{ "clgib", 0xfd, INSTR_RIS_RURDU },
+	{ "clij", 0x7f, INSTR_RIE_RUPU },
+	{ "clgij", 0x7d, INSTR_RIE_RUPU },
+	{ "clfit", 0x73, INSTR_RIE_R0UU },
+	{ "clgit", 0x71, INSTR_RIE_R0UU },
+	{ "rnsbg", 0x54, INSTR_RIE_RRUUU },
+	{ "rxsbg", 0x57, INSTR_RIE_RRUUU },
+	{ "rosbg", 0x56, INSTR_RIE_RRUUU },
+	{ "risbg", 0x55, INSTR_RIE_RRUUU },
 #endif
 	{ "", 0, INSTR_INVALID }
 };
@@ -1004,6 +1188,16 @@ static struct insn opcode_ed[] = {
 	{ "ldy", 0x65, INSTR_RXY_FRRD },
 	{ "stey", 0x66, INSTR_RXY_FRRD },
 	{ "stdy", 0x67, INSTR_RXY_FRRD },
+	{ "sldt", 0x40, INSTR_RXF_FRRDF },
+	{ "slxt", 0x48, INSTR_RXF_FRRDF },
+	{ "srdt", 0x41, INSTR_RXF_FRRDF },
+	{ "srxt", 0x49, INSTR_RXF_FRRDF },
+	{ "tdcet", 0x50, INSTR_RXE_FRRD },
+	{ "tdcdt", 0x54, INSTR_RXE_FRRD },
+	{ "tdcxt", 0x58, INSTR_RXE_FRRD },
+	{ "tdget", 0x51, INSTR_RXE_FRRD },
+	{ "tdgdt", 0x55, INSTR_RXE_FRRD },
+	{ "tdgxt", 0x59, INSTR_RXE_FRRD },
 #endif
 	{ "ldeb", 0x04, INSTR_RXE_FRRD },
 	{ "lxdb", 0x05, INSTR_RXE_FRRD },
@@ -1037,6 +1231,7 @@ static struct insn opcode_ed[] = {
 	{ "mae", 0x2e, INSTR_RXF_FRRDF },
 	{ "mse", 0x2f, INSTR_RXF_FRRDF },
 	{ "sqe", 0x34, INSTR_RXE_FRRD },
+	{ "sqd", 0x35, INSTR_RXE_FRRD },
 	{ "mee", 0x37, INSTR_RXE_FRRD },
 	{ "mad", 0x3e, INSTR_RXF_FRRDF },
 	{ "msd", 0x3f, INSTR_RXF_FRRDF },
@@ -1117,6 +1312,12 @@ static struct insn *find_insn(unsigned char *code)
 	case 0xc2:
 		table = opcode_c2;
 		break;
+	case 0xc4:
+		table = opcode_c4;
+		break;
+	case 0xc6:
+		table = opcode_c6;
+		break;
 	case 0xc8:
 		table = opcode_c8;
 		break;
diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c
index e49e9e0..31d618a 100644
--- a/arch/s390/kernel/early.c
+++ b/arch/s390/kernel/early.c
@@ -214,10 +214,13 @@ static __initdata struct sysinfo_3_2_2 vmms __aligned(PAGE_SIZE);
 
 static noinline __init void detect_machine_type(void)
 {
-	/* No VM information? Looks like LPAR */
-	if (stsi(&vmms, 3, 2, 2) == -ENOSYS)
+	/* Check current-configuration-level */
+	if ((stsi(NULL, 0, 0, 0) >> 28) <= 2) {
+		S390_lowcore.machine_flags |= MACHINE_FLAG_LPAR;
 		return;
-	if (!vmms.count)
+	}
+	/* Get virtual-machine cpu information. */
+	if (stsi(&vmms, 3, 2, 2) == -ENOSYS || !vmms.count)
 		return;
 
 	/* Running under KVM? If not we assume z/VM */
@@ -402,8 +405,19 @@ static void __init append_to_cmdline(size_t (*ipl_data)(char *, size_t))
 
 static void __init setup_boot_command_line(void)
 {
+	int i;
+
+	/* convert arch command line to ascii */
+	for (i = 0; i < ARCH_COMMAND_LINE_SIZE; i++)
+		if (COMMAND_LINE[i] & 0x80)
+			break;
+	if (i < ARCH_COMMAND_LINE_SIZE)
+		EBCASC(COMMAND_LINE, ARCH_COMMAND_LINE_SIZE);
+	COMMAND_LINE[ARCH_COMMAND_LINE_SIZE-1] = 0;
+
 	/* copy arch command line */
-	strlcpy(boot_command_line, COMMAND_LINE, ARCH_COMMAND_LINE_SIZE);
+	strlcpy(boot_command_line, strstrip(COMMAND_LINE),
+		ARCH_COMMAND_LINE_SIZE);
 
 	/* append IPL PARM data to the boot command line */
 	if (MACHINE_IS_VM)
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index e8ef21c..4348f9b 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -13,7 +13,6 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/cache.h>
-#include <asm/lowcore.h>
 #include <asm/errno.h>
 #include <asm/ptrace.h>
 #include <asm/thread_info.h>
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index f33658f..29fd0f1 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -9,11 +9,9 @@
  *		 Heiko Carstens <heiko.carstens@de.ibm.com>
  */
 
-#include <linux/sys.h>
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/cache.h>
-#include <asm/lowcore.h>
 #include <asm/errno.h>
 #include <asm/ptrace.h>
 #include <asm/thread_info.h>
diff --git a/arch/s390/kernel/ftrace.c b/arch/s390/kernel/ftrace.c
index 5a82bc6..6a83d05 100644
--- a/arch/s390/kernel/ftrace.c
+++ b/arch/s390/kernel/ftrace.c
@@ -13,7 +13,7 @@
 #include <linux/kernel.h>
 #include <linux/types.h>
 #include <trace/syscall.h>
-#include <asm/lowcore.h>
+#include <asm/asm-offsets.h>
 
 #ifdef CONFIG_DYNAMIC_FTRACE
 
@@ -200,13 +200,3 @@ out:
 	return parent;
 }
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
-
-#ifdef CONFIG_FTRACE_SYSCALLS
-
-extern unsigned int sys_call_table[];
-
-unsigned long __init arch_syscall_addr(int nr)
-{
-	return (unsigned long)sys_call_table[nr];
-}
-#endif
diff --git a/arch/s390/kernel/head.S b/arch/s390/kernel/head.S
index c52b4f7..ca4a62b 100644
--- a/arch/s390/kernel/head.S
+++ b/arch/s390/kernel/head.S
@@ -1,5 +1,5 @@
 /*
- * Copyright IBM Corp. 1999,2009
+ * Copyright IBM Corp. 1999,2010
  *
  *    Author(s): Hartmut Penner <hp@de.ibm.com>
  *		 Martin Schwidefsky <schwidefsky@de.ibm.com>
@@ -22,12 +22,9 @@
  */
 
 #include <linux/init.h>
-#include <asm/setup.h>
-#include <asm/lowcore.h>
 #include <asm/asm-offsets.h>
 #include <asm/thread_info.h>
 #include <asm/page.h>
-#include <asm/cpu.h>
 
 #ifdef CONFIG_64BIT
 #define ARCH_OFFSET	4
@@ -288,19 +285,7 @@ iplstart:
 	bz	.Lagain1		# skip dateset trailer
 	la	%r5,0(%r4,%r2)
 	lr	%r3,%r2
-.Lidebc:
-	tm	0(%r5),0x80		# high order bit set ?
-	bo	.Ldocv			#  yes -> convert from EBCDIC
-	ahi	%r5,-1
-	bct	%r3,.Lidebc
-	b	.Lnocv
-.Ldocv:
-	l	%r3,.Lcvtab
-	tr	0(256,%r4),0(%r3)	# convert parameters to ascii
-	tr	256(256,%r4),0(%r3)
-	tr	512(256,%r4),0(%r3)
-	tr	768(122,%r4),0(%r3)
-.Lnocv: la	%r3,COMMAND_LINE-PARMAREA(%r12) # load adr. of command line
+	la	%r3,COMMAND_LINE-PARMAREA(%r12) # load adr. of command line
 	mvc	0(256,%r3),0(%r4)
 	mvc	256(256,%r3),256(%r4)
 	mvc	512(256,%r3),512(%r4)
@@ -384,7 +369,6 @@ iplstart:
 .Linitrd:.long _end + 0x400000		# default address of initrd
 .Lparm:	.long  PARMAREA
 .Lstartup: .long startup
-.Lcvtab:.long	_ebcasc 		# ebcdic to ascii table
 .Lreset:.byte	0xc3,0xc8,0xc1,0xd5,0xc7,0xc5,0x40,0xd9,0xc4,0xd9,0x40
 	.byte	0xc1,0xd3,0xd3,0x40,0xd2,0xc5,0xc5,0xd7,0x40,0xd5,0xd6
 	.byte	0xc8,0xd6,0xd3,0xc4	# "change rdr all keep nohold"
@@ -417,13 +401,10 @@ start:
 .sk8x8:
 	mvc	0(240,%r8),0(%r9)	# copy iplparms into buffer
 .gotr:
-	l	%r10,.tbl		# EBCDIC to ASCII table
-	tr	0(240,%r8),0(%r10)
 	slr	%r0,%r0
 	st	%r0,INITRD_SIZE+ARCH_OFFSET-PARMAREA(%r11)
 	st	%r0,INITRD_START+ARCH_OFFSET-PARMAREA(%r11)
 	j	startup 		# continue with startup
-.tbl:	.long	_ebcasc			# translate table
 .cmd:	.long	COMMAND_LINE		# address of command line buffer
 .parm:	.long	PARMAREA
 .lowcase:
@@ -467,16 +448,15 @@ start:
 # or linload or SALIPL
 #
 	.org	0x10000
-startup:basr	%r13,0			# get base
+	.globl	startup
+startup:
+	basr	%r13,0			# get base
 .LPG0:
 	xc	0x200(256),0x200	# partially clear lowcore
 	xc	0x300(256),0x300
-	l	%r1,5f-.LPG0(%r13)
-	stck	0(%r1)
-	spt	6f-.LPG0(%r13)
-	mvc	__LC_LAST_UPDATE_CLOCK(8),0(%r1)
-	mvc	__LC_LAST_UPDATE_TIMER(8),6f-.LPG0(%r13)
-	mvc	__LC_EXIT_TIMER(8),5f-.LPG0(%r13)
+	stck	__LC_LAST_UPDATE_CLOCK
+	spt	5f-.LPG0(%r13)
+	mvc	__LC_LAST_UPDATE_TIMER(8),5f-.LPG0(%r13)
 #ifndef CONFIG_MARCH_G5
 	# check capabilities against MARCH_{G5,Z900,Z990,Z9_109,Z10}
 	xc	__LC_STFL_FAC_LIST(8),__LC_STFL_FAC_LIST
@@ -494,7 +474,6 @@ startup:basr	%r13,0			# get base
 	cl	%r0,2f+12-.LPG0(%r13)
 	je	3f
 1:	l	%r15,.Lstack-.LPG0(%r13)
-	ahi	%r15,1<<(PAGE_SHIFT+THREAD_ORDER) # init_task_union+THREAD_SIZE
 	ahi	%r15,-96
 	la	%r2,.Lals_string-.LPG0(%r13)
 	l	%r3,.Lsclp_print-.LPG0(%r13)
@@ -505,7 +484,7 @@ startup:basr	%r13,0			# get base
 .Lsclp_print:
 	.long	_sclp_print_early
 .Lstack:
-	.long	init_thread_union
+	.long	0x8000 + (1<<(PAGE_SHIFT+THREAD_ORDER))
 	.align 16
 2:	.long	0x000a0000,0x8badcccc
 #if defined(CONFIG_64BIT)
@@ -532,13 +511,22 @@ startup:basr	%r13,0			# get base
 3:
 #endif
 
+#ifdef CONFIG_64BIT
+	mvi	__LC_AR_MODE_ID,1	# set esame flag
+	slr	%r0,%r0 		# set cpuid to zero
+	lhi	%r1,2			# mode 2 = esame (dump)
+	sigp	%r1,%r0,0x12		# switch to esame mode
+	sam64				# switch to 64 bit mode
+	jg	startup_continue
+#else
+	mvi	__LC_AR_MODE_ID,0	# set ESA flag (mode 0)
 	l	%r13,4f-.LPG0(%r13)
 	b	0(%r13)
-	.align	4
+	.align	8
 4:	.long	startup_continue
-5:	.long	sched_clock_base_cc
+#endif
 	.align	8
-6:	.long	0x7fffffff,0xffffffff
+5:	.long	0x7fffffff,0xffffffff
 
 #
 # params at 10400 (setup.h)
@@ -552,8 +540,4 @@ startup:basr	%r13,0			# get base
 	.byte	"root=/dev/ram0 ro"
 	.byte	0
 
-#ifdef CONFIG_64BIT
-#include "head64.S"
-#else
-#include "head31.S"
-#endif
+	.org	0x11000
diff --git a/arch/s390/kernel/head31.S b/arch/s390/kernel/head31.S
index 602b508..1bbcc49 100644
--- a/arch/s390/kernel/head31.S
+++ b/arch/s390/kernel/head31.S
@@ -1,7 +1,7 @@
 /*
  * arch/s390/kernel/head31.S
  *
- * Copyright (C) IBM Corp. 2005,2006
+ * Copyright (C) IBM Corp. 2005,2010
  *
  *   Author(s):	Hartmut Penner <hp@de.ibm.com>
  *		Martin Schwidefsky <schwidefsky@de.ibm.com>
@@ -10,13 +10,19 @@
  *
  */
 
-	.org	0x11000
+#include <linux/init.h>
+#include <asm/asm-offsets.h>
+#include <asm/thread_info.h>
+#include <asm/page.h>
 
+__HEAD
+	.globl	startup_continue
 startup_continue:
 	basr	%r13,0			# get base
 .LPG1:
 
-	mvi	__LC_AR_MODE_ID,0	# set ESA flag (mode 0)
+	l	%r1,.Lbase_cc-.LPG1(%r13)
+	mvc	0(8,%r1),__LC_LAST_UPDATE_CLOCK
 	lctl	%c0,%c15,.Lctl-.LPG1(%r13) # load control registers
 	l	%r12,.Lparmaddr-.LPG1(%r13) # pointer to parameter area
 					# move IPL device to lowcore
@@ -69,10 +75,12 @@ startup_continue:
 .Lduald:.rept	8
 	.long	0x80000000,0,0,0	# invalid access-list entries
 	.endr
+.Lbase_cc:
+	.long	sched_clock_base_cc
 
-	.org	0x12000
 	.globl	_ehead
 _ehead:
+
 #ifdef CONFIG_SHARED_KERNEL
 	.org	0x100000
 #endif
diff --git a/arch/s390/kernel/head64.S b/arch/s390/kernel/head64.S
index d984a2a..39580e7 100644
--- a/arch/s390/kernel/head64.S
+++ b/arch/s390/kernel/head64.S
@@ -1,7 +1,7 @@
 /*
  * arch/s390/kernel/head64.S
  *
- * Copyright (C) IBM Corp. 1999,2006
+ * Copyright (C) IBM Corp. 1999,2010
  *
  *   Author(s):	Hartmut Penner <hp@de.ibm.com>
  *		Martin Schwidefsky <schwidefsky@de.ibm.com>
@@ -10,80 +10,17 @@
  *
  */
 
-	.org	0x11000
+#include <linux/init.h>
+#include <asm/asm-offsets.h>
+#include <asm/thread_info.h>
+#include <asm/page.h>
 
+__HEAD
+	.globl	startup_continue
 startup_continue:
-	basr	%r13,0			# get base
-.LPG1:	sll	%r13,1			# remove high order bit
-	srl	%r13,1
-
-#ifdef CONFIG_ZFCPDUMP
-
-	# check if we have been ipled using zfcp dump:
-
-	tm	0xb9,0x01		# test if subchannel is enabled
-	jno	.nodump			# subchannel disabled
-	l	%r1,0xb8
-	la	%r5,.Lipl_schib-.LPG1(%r13)
-	stsch	0(%r5)			# get schib of subchannel
-	jne	.nodump			# schib not available
-	tm	5(%r5),0x01		# devno valid?
-	jno	.nodump
-	tm	4(%r5),0x80		# qdio capable device?
-	jno	.nodump
-	l	%r2,20(%r0)		# address of ipl parameter block
-	lhi	%r3,0
-	ic	%r3,0x148(%r2)		# get opt field
-	chi	%r3,0x20		# load with dump?
-	jne	.nodump
-
-	# store all prefix registers in case of load with dump:
-
-	la	%r7,0			# base register for 0 page
-	la	%r8,0			# first cpu
-	l	%r11,.Lpref_arr_ptr-.LPG1(%r13)	# address of prefix array
-	ahi	%r11,4			# skip boot cpu
-	lr	%r12,%r11
-	ahi	%r12,(CONFIG_NR_CPUS*4)	# end of prefix array
-	stap	.Lcurrent_cpu+2-.LPG1(%r13)	# store current cpu addr
-1:
-	cl	%r8,.Lcurrent_cpu-.LPG1(%r13)	# is ipl cpu ?
-	je	4f				# if yes get next cpu
-2:
-	lr	%r9,%r7
-	sigp	%r9,%r8,0x9		# stop & store status of cpu
-	brc	8,3f			# accepted
-	brc	4,4f			# status stored: next cpu
-	brc	2,2b			# busy: 	 try again
-	brc	1,4f			# not op:	 next cpu
-3:
-	mvc	0(4,%r11),264(%r7)	# copy prefix register to prefix array
-	ahi	%r11,4			# next element in prefix array
-	clr	%r11,%r12
-	je	5f			# no more space in prefix array
-4:
-	ahi	%r8,1			# next cpu (r8 += 1)
-	chi	%r8,MAX_CPU_ADDRESS	# is last possible cpu ?
-	jle	1b			# jump if not last cpu
-5:
-	lhi	%r1,2			# mode 2 = esame (dump)
-	j	6f
-	.align 4
-.Lipl_schib:
-	.rept 13
-	.long 0
-	.endr
-.nodump:
-	lhi	%r1,1			# mode 1 = esame (normal ipl)
-6:
-#else
-	lhi	%r1,1			# mode 1 = esame (normal ipl)
-#endif /* CONFIG_ZFCPDUMP */
-	mvi	__LC_AR_MODE_ID,1	# set esame flag
-	slr	%r0,%r0 		# set cpuid to zero
-	sigp	%r1,%r0,0x12		# switch to esame mode
-	sam64				# switch to 64 bit mode
-	llgfr	%r13,%r13		# clear high-order half of base reg
+	larl	%r1,sched_clock_base_cc
+	mvc	0(8,%r1),__LC_LAST_UPDATE_CLOCK
+	larl	%r13,.LPG1		# get base
 	lmh	%r0,%r15,.Lzero64-.LPG1(%r13)	# clear high-order half
 	lctlg	%c0,%c15,.Lctl-.LPG1(%r13)	# load control registers
 	lg	%r12,.Lparmaddr-.LPG1(%r13)	# pointer to parameter area
@@ -108,6 +45,7 @@ startup_continue:
 	lpswe	.Lentry-.LPG1(13)	# jump to _stext in primary-space,
 					# virtual and never return ...
 	.align	16
+.LPG1:
 .Lentry:.quad	0x0000000180000000,_stext
 .Lctl:	.quad	0x04350002		# cr0: various things
 	.quad	0			# cr1: primary space segment table
@@ -130,12 +68,6 @@ startup_continue:
 .Lscan2g:.quad	0x80000000 + 0x20000 - 8	# 2GB + 128K - 8
 .Lnop:	.long	0x07000700
 .Lzero64:.fill	16,4,0x0
-#ifdef CONFIG_ZFCPDUMP
-.Lcurrent_cpu:
-	.long 0x0
-.Lpref_arr_ptr:
-	.long zfcpdump_prefix_array
-#endif /* CONFIG_ZFCPDUMP */
 .Lparmaddr:
 	.quad	PARMAREA
 	.align	64
@@ -146,9 +78,9 @@ startup_continue:
 	.long	0x80000000,0,0,0	# invalid access-list entries
 	.endr
 
-	.org	0x12000
 	.globl	_ehead
 _ehead:
+
 #ifdef CONFIG_SHARED_KERNEL
 	.org	0x100000
 #endif
diff --git a/arch/s390/kernel/ipl.c b/arch/s390/kernel/ipl.c
index 4d73296..7eedbbc 100644
--- a/arch/s390/kernel/ipl.c
+++ b/arch/s390/kernel/ipl.c
@@ -553,7 +553,7 @@ out:
 	return rc;
 }
 
-static void ipl_run(struct shutdown_trigger *trigger)
+static void __ipl_run(void *unused)
 {
 	diag308(DIAG308_IPL, NULL);
 	if (MACHINE_IS_VM)
@@ -562,6 +562,11 @@ static void ipl_run(struct shutdown_trigger *trigger)
 		reipl_ccw_dev(&ipl_info.data.ccw.dev_id);
 }
 
+static void ipl_run(struct shutdown_trigger *trigger)
+{
+	smp_switch_to_ipl_cpu(__ipl_run, NULL);
+}
+
 static int __init ipl_init(void)
 {
 	int rc;
@@ -1039,7 +1044,7 @@ static void get_ipl_string(char *dst, struct ipl_parameter_block *ipb,
 		sprintf(dst + pos, " PARM %s", vmparm);
 }
 
-static void reipl_run(struct shutdown_trigger *trigger)
+static void __reipl_run(void *unused)
 {
 	struct ccw_dev_id devid;
 	static char buf[128];
@@ -1087,6 +1092,11 @@ static void reipl_run(struct shutdown_trigger *trigger)
 	disabled_wait((unsigned long) __builtin_return_address(0));
 }
 
+static void reipl_run(struct shutdown_trigger *trigger)
+{
+	smp_switch_to_ipl_cpu(__reipl_run, NULL);
+}
+
 static void reipl_block_ccw_init(struct ipl_parameter_block *ipb)
 {
 	ipb->hdr.len = IPL_PARM_BLK_CCW_LEN;
@@ -1369,20 +1379,18 @@ static struct kobj_attribute dump_type_attr =
 
 static struct kset *dump_kset;
 
-static void dump_run(struct shutdown_trigger *trigger)
+static void __dump_run(void *unused)
 {
 	struct ccw_dev_id devid;
 	static char buf[100];
 
 	switch (dump_method) {
 	case DUMP_METHOD_CCW_CIO:
-		smp_send_stop();
 		devid.devno = dump_block_ccw->ipl_info.ccw.devno;
 		devid.ssid  = 0;
 		reipl_ccw_dev(&devid);
 		break;
 	case DUMP_METHOD_CCW_VM:
-		smp_send_stop();
 		sprintf(buf, "STORE STATUS");
 		__cpcmd(buf, NULL, 0, NULL);
 		sprintf(buf, "IPL %X", dump_block_ccw->ipl_info.ccw.devno);
@@ -1396,10 +1404,17 @@ static void dump_run(struct shutdown_trigger *trigger)
 		diag308(DIAG308_SET, dump_block_fcp);
 		diag308(DIAG308_DUMP, NULL);
 		break;
-	case DUMP_METHOD_NONE:
-		return;
+	default:
+		break;
 	}
-	printk(KERN_EMERG "Dump failed!\n");
+}
+
+static void dump_run(struct shutdown_trigger *trigger)
+{
+	if (dump_method == DUMP_METHOD_NONE)
+		return;
+	smp_send_stop();
+	smp_switch_to_ipl_cpu(__dump_run, NULL);
 }
 
 static int __init dump_ccw_init(void)
@@ -1577,7 +1592,7 @@ static void vmcmd_run(struct shutdown_trigger *trigger)
 static int vmcmd_init(void)
 {
 	if (!MACHINE_IS_VM)
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	vmcmd_kset = kset_create_and_add("vmcmd", NULL, firmware_kobj);
 	if (!vmcmd_kset)
 		return -ENOMEM;
@@ -1595,7 +1610,7 @@ static void stop_run(struct shutdown_trigger *trigger)
 {
 	if (strcmp(trigger->name, ON_PANIC_STR) == 0)
 		disabled_wait((unsigned long) __builtin_return_address(0));
-	while (signal_processor(smp_processor_id(), sigp_stop) == sigp_busy)
+	while (sigp(smp_processor_id(), sigp_stop) == sigp_busy)
 		cpu_relax();
 	for (;;);
 }
@@ -1902,7 +1917,6 @@ void __init ipl_update_parameters(void)
 void __init ipl_save_parameters(void)
 {
 	struct cio_iplinfo iplinfo;
-	unsigned int *ipl_ptr;
 	void *src, *dst;
 
 	if (cio_get_iplinfo(&iplinfo))
@@ -1913,11 +1927,10 @@ void __init ipl_save_parameters(void)
 	if (!iplinfo.is_qdio)
 		return;
 	ipl_flags |= IPL_PARMBLOCK_VALID;
-	ipl_ptr = (unsigned int *)__LC_IPL_PARMBLOCK_PTR;
-	src = (void *)(unsigned long)*ipl_ptr;
+	src = (void *)(unsigned long)S390_lowcore.ipl_parmblock_ptr;
 	dst = (void *)IPL_PARMBLOCK_ORIGIN;
 	memmove(dst, src, PAGE_SIZE);
-	*ipl_ptr = IPL_PARMBLOCK_ORIGIN;
+	S390_lowcore.ipl_parmblock_ptr = IPL_PARMBLOCK_ORIGIN;
 }
 
 static LIST_HEAD(rcall);
diff --git a/arch/s390/kernel/machine_kexec.c b/arch/s390/kernel/machine_kexec.c
index 131d7ee..a922d51 100644
--- a/arch/s390/kernel/machine_kexec.c
+++ b/arch/s390/kernel/machine_kexec.c
@@ -54,11 +54,11 @@ void machine_shutdown(void)
 {
 }
 
-void machine_kexec(struct kimage *image)
+static void __machine_kexec(void *data)
 {
 	relocate_kernel_t data_mover;
+	struct kimage *image = data;
 
-	smp_send_stop();
 	pfault_fini();
 	s390_reset_system();
 
@@ -68,3 +68,9 @@ void machine_kexec(struct kimage *image)
 	(*data_mover)(&image->head, image->start);
 	for (;;);
 }
+
+void machine_kexec(struct kimage *image)
+{
+	smp_send_stop();
+	smp_switch_to_ipl_cpu(__machine_kexec, image);
+}
diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c
index 7cf4642..33fdc5a 100644
--- a/arch/s390/kernel/ptrace.c
+++ b/arch/s390/kernel/ptrace.c
@@ -992,3 +992,61 @@ const struct user_regset_view *task_user_regset_view(struct task_struct *task)
 #endif
 	return &user_s390_view;
 }
+
+static const char *gpr_names[NUM_GPRS] = {
+	"r0", "r1",  "r2",  "r3",  "r4",  "r5",  "r6",  "r7",
+	"r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
+};
+
+unsigned long regs_get_register(struct pt_regs *regs, unsigned int offset)
+{
+	if (offset >= NUM_GPRS)
+		return 0;
+	return regs->gprs[offset];
+}
+
+int regs_query_register_offset(const char *name)
+{
+	unsigned long offset;
+
+	if (!name || *name != 'r')
+		return -EINVAL;
+	if (strict_strtoul(name + 1, 10, &offset))
+		return -EINVAL;
+	if (offset >= NUM_GPRS)
+		return -EINVAL;
+	return offset;
+}
+
+const char *regs_query_register_name(unsigned int offset)
+{
+	if (offset >= NUM_GPRS)
+		return NULL;
+	return gpr_names[offset];
+}
+
+static int regs_within_kernel_stack(struct pt_regs *regs, unsigned long addr)
+{
+	unsigned long ksp = kernel_stack_pointer(regs);
+
+	return (addr & ~(THREAD_SIZE - 1)) == (ksp & ~(THREAD_SIZE - 1));
+}
+
+/**
+ * regs_get_kernel_stack_nth() - get Nth entry of the stack
+ * @regs:pt_regs which contains kernel stack pointer.
+ * @n:stack entry number.
+ *
+ * regs_get_kernel_stack_nth() returns @n th entry of the kernel stack which
+ * is specifined by @regs. If the @n th entry is NOT in the kernel stack,
+ * this returns 0.
+ */
+unsigned long regs_get_kernel_stack_nth(struct pt_regs *regs, unsigned int n)
+{
+	unsigned long addr;
+
+	addr = kernel_stack_pointer(regs) + n * sizeof(long);
+	if (!regs_within_kernel_stack(regs, addr))
+		return 0;
+	return *(unsigned long *)addr;
+}
diff --git a/arch/s390/kernel/reipl.S b/arch/s390/kernel/reipl.S
index 2f481cc..cb899d9 100644
--- a/arch/s390/kernel/reipl.S
+++ b/arch/s390/kernel/reipl.S
@@ -6,7 +6,7 @@
  *    Author(s): Holger Smolinski (Holger.Smolinski@de.ibm.com)
  */
 
-#include <asm/lowcore.h>
+#include <asm/asm-offsets.h>
 
 #
 # do_reipl_asm
diff --git a/arch/s390/kernel/reipl64.S b/arch/s390/kernel/reipl64.S
index 7741478..5e73dee 100644
--- a/arch/s390/kernel/reipl64.S
+++ b/arch/s390/kernel/reipl64.S
@@ -4,7 +4,7 @@
  *		 Denis Joseph Barrow,
  */
 
-#include <asm/lowcore.h>
+#include <asm/asm-offsets.h>
 
 #
 # do_reipl_asm
diff --git a/arch/s390/kernel/sclp.S b/arch/s390/kernel/sclp.S
index e27ca63..27af3bf 100644
--- a/arch/s390/kernel/sclp.S
+++ b/arch/s390/kernel/sclp.S
@@ -9,8 +9,10 @@
  */
 
 LC_EXT_NEW_PSW		= 0x58			# addr of ext int handler
+LC_EXT_NEW_PSW_64	= 0x1b0			# addr of ext int handler 64 bit
 LC_EXT_INT_PARAM	= 0x80			# addr of ext int parameter
 LC_EXT_INT_CODE		= 0x86			# addr of ext int code
+LC_AR_MODE_ID		= 0xa3
 
 #
 # Subroutine which waits synchronously until either an external interruption
@@ -30,8 +32,16 @@ _sclp_wait_int:
 .LbaseS1:
 	ahi	%r15,-96			# create stack frame
 	la	%r8,LC_EXT_NEW_PSW		# register int handler
-	mvc	.LoldpswS1-.LbaseS1(8,%r13),0(%r8)
-	mvc	0(8,%r8),.LextpswS1-.LbaseS1(%r13)
+	la	%r9,.LextpswS1-.LbaseS1(%r13)
+#ifdef CONFIG_64BIT
+	tm	LC_AR_MODE_ID,1
+	jno	.Lesa1
+	la	%r8,LC_EXT_NEW_PSW_64		# register int handler 64 bit
+	la	%r9,.LextpswS1_64-.LbaseS1(%r13)
+.Lesa1:
+#endif
+	mvc	.LoldpswS1-.LbaseS1(16,%r13),0(%r8)
+	mvc	0(16,%r8),0(%r9)
 	lhi	%r6,0x0200			# cr mask for ext int (cr0.54)
 	ltr	%r2,%r2
 	jz	.LsetctS1
@@ -64,15 +74,19 @@ _sclp_wait_int:
 .LtimeoutS1:
 	lctl	%c0,%c0,.LctlS1-.LbaseS1(%r13)	# restore interrupt setting
 	# restore old handler
-	mvc	0(8,%r8),.LoldpswS1-.LbaseS1(%r13)
+	mvc	0(16,%r8),.LoldpswS1-.LbaseS1(%r13)
 	lm	%r6,%r15,120(%r15)		# restore registers
 	br	%r14				# return to caller
 
 	.align	8
 .LoldpswS1:
-	.long	0, 0				# old ext int PSW
+	.long	0, 0, 0, 0			# old ext int PSW
 .LextpswS1:
 	.long	0x00080000, 0x80000000+.LwaitS1	# PSW to handle ext int
+#ifdef CONFIG_64BIT
+.LextpswS1_64:
+	.quad	0x0000000180000000, .LwaitS1	# PSW to handle ext int, 64 bit
+#endif
 .LwaitpswS1:
 	.long	0x010a0000, 0x00000000+.LloopS1	# PSW to wait for ext int
 .LtimeS1:
@@ -250,6 +264,13 @@ _sclp_print:
 _sclp_print_early:
 	stm	%r6,%r15,24(%r15)		# save registers
 	ahi	%r15,-96			# create stack frame
+#ifdef CONFIG_64BIT
+	tm	LC_AR_MODE_ID,1
+	jno	.Lesa2
+	ahi	%r15,-80
+	stmh	%r6,%r15,96(%r15)		# store upper register halves
+.Lesa2:
+#endif
 	lr	%r10,%r2			# save string pointer
 	lhi	%r2,0
 	bras	%r14,_sclp_setup		# enable console
@@ -262,6 +283,13 @@ _sclp_print_early:
 	lhi	%r2,1
 	bras	%r14,_sclp_setup		# disable console
 .LendS5:
+#ifdef CONFIG_64BIT
+	tm	LC_AR_MODE_ID,1
+	jno	.Lesa3
+	lmh	%r6,%r15,96(%r15)		# store upper register halves
+	ahi	%r15,80
+.Lesa3:
+#endif
 	lm	%r6,%r15,120(%r15)		# restore registers
 	br	%r14
 
diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c
index 8d8957b..77a63ae 100644
--- a/arch/s390/kernel/setup.c
+++ b/arch/s390/kernel/setup.c
@@ -396,15 +396,12 @@ static void __init
 setup_lowcore(void)
 {
 	struct _lowcore *lc;
-	int lc_pages;
 
 	/*
 	 * Setup lowcore for boot cpu
 	 */
-	lc_pages = sizeof(void *) == 8 ? 2 : 1;
-	lc = (struct _lowcore *)
-		__alloc_bootmem(lc_pages * PAGE_SIZE, lc_pages * PAGE_SIZE, 0);
-	memset(lc, 0, lc_pages * PAGE_SIZE);
+	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
+	lc = __alloc_bootmem(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 	lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
 	lc->restart_psw.addr =
 		PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
@@ -804,7 +801,7 @@ setup_arch(char **cmdline_p)
 	if (MACHINE_IS_VM)
 		pr_info("Linux is running as a z/VM "
 			"guest operating system in 31-bit mode\n");
-	else
+	else if (MACHINE_IS_LPAR)
 		pr_info("Linux is running natively in 31-bit mode\n");
 	if (MACHINE_HAS_IEEE)
 		pr_info("The hardware system has IEEE compatible "
@@ -818,7 +815,7 @@ setup_arch(char **cmdline_p)
 			"guest operating system in 64-bit mode\n");
 	else if (MACHINE_IS_KVM)
 		pr_info("Linux is running under KVM in 64-bit mode\n");
-	else
+	else if (MACHINE_IS_LPAR)
 		pr_info("Linux is running natively in 64-bit mode\n");
 #endif /* CONFIG_64BIT */
 
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index 76a6fdd..8b10127 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -36,6 +36,7 @@
 #include <linux/cpu.h>
 #include <linux/timex.h>
 #include <linux/bootmem.h>
+#include <asm/asm-offsets.h>
 #include <asm/ipl.h>
 #include <asm/setup.h>
 #include <asm/sigp.h>
@@ -53,7 +54,7 @@
 #include "entry.h"
 
 /* logical cpu to cpu address */
-int __cpu_logical_map[NR_CPUS];
+unsigned short __cpu_logical_map[NR_CPUS];
 
 static struct task_struct *current_set[NR_CPUS];
 
@@ -72,13 +73,13 @@ static int cpu_management;
 
 static DEFINE_PER_CPU(struct cpu, cpu_devices);
 
-static void smp_ext_bitcall(int, ec_bit_sig);
+static void smp_ext_bitcall(int, int);
 
-static int cpu_stopped(int cpu)
+static int raw_cpu_stopped(int cpu)
 {
-	__u32 status;
+	u32 status;
 
-	switch (signal_processor_ps(&status, 0, cpu, sigp_sense)) {
+	switch (raw_sigp_ps(&status, 0, cpu, sigp_sense)) {
 	case sigp_status_stored:
 		/* Check for stopped and check stop state */
 		if (status & 0x50)
@@ -90,6 +91,44 @@ static int cpu_stopped(int cpu)
 	return 0;
 }
 
+static inline int cpu_stopped(int cpu)
+{
+	return raw_cpu_stopped(cpu_logical_map(cpu));
+}
+
+void smp_switch_to_ipl_cpu(void (*func)(void *), void *data)
+{
+	struct _lowcore *lc, *current_lc;
+	struct stack_frame *sf;
+	struct pt_regs *regs;
+	unsigned long sp;
+
+	if (smp_processor_id() == 0)
+		func(data);
+	__load_psw_mask(PSW_BASE_BITS | PSW_DEFAULT_KEY);
+	/* Disable lowcore protection */
+	__ctl_clear_bit(0, 28);
+	current_lc = lowcore_ptr[smp_processor_id()];
+	lc = lowcore_ptr[0];
+	if (!lc)
+		lc = current_lc;
+	lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
+	lc->restart_psw.addr = PSW_ADDR_AMODE | (unsigned long) smp_restart_cpu;
+	if (!cpu_online(0))
+		smp_switch_to_cpu(func, data, 0, stap(), __cpu_logical_map[0]);
+	while (sigp(0, sigp_stop_and_store_status) == sigp_busy)
+		cpu_relax();
+	sp = lc->panic_stack;
+	sp -= sizeof(struct pt_regs);
+	regs = (struct pt_regs *) sp;
+	memcpy(&regs->gprs, &current_lc->gpregs_save_area, sizeof(regs->gprs));
+	regs->psw = lc->psw_save_area;
+	sp -= STACK_FRAME_OVERHEAD;
+	sf = (struct stack_frame *) sp;
+	sf->back_chain = regs->gprs[15];
+	smp_switch_to_cpu(func, data, sp, stap(), __cpu_logical_map[0]);
+}
+
 void smp_send_stop(void)
 {
 	int cpu, rc;
@@ -103,7 +142,7 @@ void smp_send_stop(void)
 		if (cpu == smp_processor_id())
 			continue;
 		do {
-			rc = signal_processor(cpu, sigp_stop);
+			rc = sigp(cpu, sigp_stop);
 		} while (rc == sigp_busy);
 
 		while (!cpu_stopped(cpu))
@@ -139,13 +178,13 @@ static void do_ext_call_interrupt(__u16 code)
  * Send an external call sigp to another cpu and return without waiting
  * for its completion.
  */
-static void smp_ext_bitcall(int cpu, ec_bit_sig sig)
+static void smp_ext_bitcall(int cpu, int sig)
 {
 	/*
 	 * Set signaling bit in lowcore of target cpu and kick it
 	 */
 	set_bit(sig, (unsigned long *) &lowcore_ptr[cpu]->ext_call_fast);
-	while (signal_processor(cpu, sigp_emergency_signal) == sigp_busy)
+	while (sigp(cpu, sigp_emergency_signal) == sigp_busy)
 		udelay(10);
 }
 
@@ -239,24 +278,8 @@ void smp_ctl_clear_bit(int cr, int bit)
 }
 EXPORT_SYMBOL(smp_ctl_clear_bit);
 
-/*
- * In early ipl state a temp. logically cpu number is needed, so the sigp
- * functions can be used to sense other cpus. Since NR_CPUS is >= 2 on
- * CONFIG_SMP and the ipl cpu is logical cpu 0, it must be 1.
- */
-#define CPU_INIT_NO	1
-
 #ifdef CONFIG_ZFCPDUMP
 
-/*
- * zfcpdump_prefix_array holds prefix registers for the following scenario:
- * 64 bit zfcpdump kernel and 31 bit kernel which is to be dumped. We have to
- * save its prefix registers, since they get lost, when switching from 31 bit
- * to 64 bit.
- */
-unsigned int zfcpdump_prefix_array[NR_CPUS + 1] \
-	__attribute__((__section__(".data")));
-
 static void __init smp_get_save_area(unsigned int cpu, unsigned int phy_cpu)
 {
 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
@@ -266,21 +289,15 @@ static void __init smp_get_save_area(unsigned int cpu, unsigned int phy_cpu)
 			   "the dump\n", cpu, NR_CPUS - 1);
 		return;
 	}
-	zfcpdump_save_areas[cpu] = kmalloc(sizeof(union save_area), GFP_KERNEL);
-	__cpu_logical_map[CPU_INIT_NO] = (__u16) phy_cpu;
-	while (signal_processor(CPU_INIT_NO, sigp_stop_and_store_status) ==
-	       sigp_busy)
+	zfcpdump_save_areas[cpu] = kmalloc(sizeof(struct save_area), GFP_KERNEL);
+	while (raw_sigp(phy_cpu, sigp_stop_and_store_status) == sigp_busy)
 		cpu_relax();
 	memcpy(zfcpdump_save_areas[cpu],
 	       (void *)(unsigned long) store_prefix() + SAVE_AREA_BASE,
-	       SAVE_AREA_SIZE);
-#ifdef CONFIG_64BIT
-	/* copy original prefix register */
-	zfcpdump_save_areas[cpu]->s390x.pref_reg = zfcpdump_prefix_array[cpu];
-#endif
+	       sizeof(struct save_area));
 }
 
-union save_area *zfcpdump_save_areas[NR_CPUS + 1];
+struct save_area *zfcpdump_save_areas[NR_CPUS + 1];
 EXPORT_SYMBOL_GPL(zfcpdump_save_areas);
 
 #else
@@ -389,8 +406,7 @@ static void __init smp_detect_cpus(void)
 		for (cpu = 0; cpu <= MAX_CPU_ADDRESS; cpu++) {
 			if (cpu == boot_cpu_addr)
 				continue;
-			__cpu_logical_map[CPU_INIT_NO] = cpu;
-			if (!cpu_stopped(CPU_INIT_NO))
+			if (!raw_cpu_stopped(cpu))
 				continue;
 			smp_get_save_area(c_cpus, cpu);
 			c_cpus++;
@@ -413,8 +429,7 @@ static void __init smp_detect_cpus(void)
 		cpu_addr = info->cpu[cpu].address;
 		if (cpu_addr == boot_cpu_addr)
 			continue;
-		__cpu_logical_map[CPU_INIT_NO] = cpu_addr;
-		if (!cpu_stopped(CPU_INIT_NO)) {
+		if (!raw_cpu_stopped(cpu_addr)) {
 			s_cpus++;
 			continue;
 		}
@@ -533,18 +548,18 @@ static void smp_free_lowcore(int cpu)
 /* Upping and downing of CPUs */
 int __cpuinit __cpu_up(unsigned int cpu)
 {
-	struct task_struct *idle;
 	struct _lowcore *cpu_lowcore;
+	struct task_struct *idle;
 	struct stack_frame *sf;
-	sigp_ccode ccode;
 	u32 lowcore;
+	int ccode;
 
 	if (smp_cpu_state[cpu] != CPU_STATE_CONFIGURED)
 		return -EIO;
 	if (smp_alloc_lowcore(cpu))
 		return -ENOMEM;
 	do {
-		ccode = signal_processor(cpu, sigp_initial_cpu_reset);
+		ccode = sigp(cpu, sigp_initial_cpu_reset);
 		if (ccode == sigp_busy)
 			udelay(10);
 		if (ccode == sigp_not_operational)
@@ -552,7 +567,7 @@ int __cpuinit __cpu_up(unsigned int cpu)
 	} while (ccode == sigp_busy);
 
 	lowcore = (u32)(unsigned long)lowcore_ptr[cpu];
-	while (signal_processor_p(lowcore, cpu, sigp_set_prefix) == sigp_busy)
+	while (sigp_p(lowcore, cpu, sigp_set_prefix) == sigp_busy)
 		udelay(10);
 
 	idle = current_set[cpu];
@@ -578,7 +593,7 @@ int __cpuinit __cpu_up(unsigned int cpu)
 	cpu_lowcore->ftrace_func = S390_lowcore.ftrace_func;
 	eieio();
 
-	while (signal_processor(cpu, sigp_restart) == sigp_busy)
+	while (sigp(cpu, sigp_restart) == sigp_busy)
 		udelay(10);
 
 	while (!cpu_online(cpu))
@@ -640,7 +655,7 @@ void __cpu_die(unsigned int cpu)
 	/* Wait until target cpu is down */
 	while (!cpu_stopped(cpu))
 		cpu_relax();
-	while (signal_processor_p(0, cpu, sigp_set_prefix) == sigp_busy)
+	while (sigp_p(0, cpu, sigp_set_prefix) == sigp_busy)
 		udelay(10);
 	smp_free_lowcore(cpu);
 	pr_info("Processor %d stopped\n", cpu);
@@ -649,7 +664,7 @@ void __cpu_die(unsigned int cpu)
 void cpu_die(void)
 {
 	idle_task_exit();
-	while (signal_processor(smp_processor_id(), sigp_stop) == sigp_busy)
+	while (sigp(smp_processor_id(), sigp_stop) == sigp_busy)
 		cpu_relax();
 	for (;;);
 }
@@ -765,7 +780,8 @@ static ssize_t cpu_configure_store(struct sys_device *dev,
 	get_online_cpus();
 	mutex_lock(&smp_cpu_state_mutex);
 	rc = -EBUSY;
-	if (cpu_online(cpu))
+	/* disallow configuration changes of online cpus and cpu 0 */
+	if (cpu_online(cpu) || cpu == 0)
 		goto out;
 	rc = 0;
 	switch (val) {
diff --git a/arch/s390/kernel/switch_cpu.S b/arch/s390/kernel/switch_cpu.S
new file mode 100644
index 0000000..469f11b
--- /dev/null
+++ b/arch/s390/kernel/switch_cpu.S
@@ -0,0 +1,58 @@
+/*
+ * 31-bit switch cpu code
+ *
+ * Copyright IBM Corp. 2009
+ *
+ */
+
+#include <asm/asm-offsets.h>
+#include <asm/ptrace.h>
+
+# smp_switch_to_cpu switches to destination cpu and executes the passed function
+# Parameter: %r2 - function to call
+#	     %r3 - function parameter
+#	     %r4 - stack poiner
+#	     %r5 - current cpu
+#	     %r6 - destination cpu
+
+	.section .text
+	.align 4
+	.globl smp_switch_to_cpu
+smp_switch_to_cpu:
+	stm	%r6,%r15,__SF_GPRS(%r15)
+	lr	%r1,%r15
+	ahi	%r15,-STACK_FRAME_OVERHEAD
+	st	%r1,__SF_BACKCHAIN(%r15)
+	basr	%r13,0
+0:	la	%r1,.gprregs_addr-0b(%r13)
+	l	%r1,0(%r1)
+	stm	%r0,%r15,0(%r1)
+1:	sigp	%r0,%r6,__SIGP_RESTART	/* start destination CPU */
+	brc	2,1b			/* busy, try again */
+2:	sigp	%r0,%r5,__SIGP_STOP	/* stop current CPU */
+	brc	2,2b			/* busy, try again */
+3:	j	3b
+
+	.globl	smp_restart_cpu
+smp_restart_cpu:
+	basr	%r13,0
+0:	la	%r1,.gprregs_addr-0b(%r13)
+	l	%r1,0(%r1)
+	lm	%r0,%r15,0(%r1)
+1:	sigp	%r0,%r5,__SIGP_SENSE	/* Wait for calling CPU */
+	brc	10,1b			/* busy, accepted (status 0), running */
+	tmll	%r0,0x40		/* Test if calling CPU is stopped */
+	jz	1b
+	ltr	%r4,%r4			/* New stack ? */
+	jz	1f
+	lr	%r15,%r4
+1:	basr	%r14,%r2
+
+.gprregs_addr:
+	.long	.gprregs
+
+	.section .data,"aw",@progbits
+.gprregs:
+	.rept	16
+	.long	0
+	.endr
diff --git a/arch/s390/kernel/switch_cpu64.S b/arch/s390/kernel/switch_cpu64.S
new file mode 100644
index 0000000..d94aacc
--- /dev/null
+++ b/arch/s390/kernel/switch_cpu64.S
@@ -0,0 +1,51 @@
+/*
+ * 64-bit switch cpu code
+ *
+ * Copyright IBM Corp. 2009
+ *
+ */
+
+#include <asm/asm-offsets.h>
+#include <asm/ptrace.h>
+
+# smp_switch_to_cpu switches to destination cpu and executes the passed function
+# Parameter: %r2 - function to call
+#	     %r3 - function parameter
+#	     %r4 - stack poiner
+#	     %r5 - current cpu
+#	     %r6 - destination cpu
+
+	.section .text
+	.align 4
+	.globl smp_switch_to_cpu
+smp_switch_to_cpu:
+	stmg	%r6,%r15,__SF_GPRS(%r15)
+	lgr	%r1,%r15
+	aghi	%r15,-STACK_FRAME_OVERHEAD
+	stg	%r1,__SF_BACKCHAIN(%r15)
+	larl	%r1,.gprregs
+	stmg	%r0,%r15,0(%r1)
+1:	sigp	%r0,%r6,__SIGP_RESTART	/* start destination CPU */
+	brc	2,1b			/* busy, try again */
+2:	sigp	%r0,%r5,__SIGP_STOP	/* stop current CPU */
+	brc	2,2b			/* busy, try again */
+3:	j	3b
+
+	.globl	smp_restart_cpu
+smp_restart_cpu:
+	larl	%r1,.gprregs
+	lmg	%r0,%r15,0(%r1)
+1:	sigp	%r0,%r5,__SIGP_SENSE	/* Wait for calling CPU */
+	brc	10,1b			/* busy, accepted (status 0), running */
+	tmll	%r0,0x40		/* Test if calling CPU is stopped */
+	jz	1b
+	ltgr	%r4,%r4			/* New stack ? */
+	jz	1f
+	lgr	%r15,%r4
+1:	basr	%r14,%r2
+
+	.section .data,"aw",@progbits
+.gprregs:
+	.rept	16
+	.quad	0
+	.endr
diff --git a/arch/s390/kernel/swsusp_asm64.S b/arch/s390/kernel/swsusp_asm64.S
index 0c26cc1..b354427 100644
--- a/arch/s390/kernel/swsusp_asm64.S
+++ b/arch/s390/kernel/swsusp_asm64.S
@@ -176,7 +176,7 @@ pgm_check_entry:
 	cgr	%r1,%r2
 	je	restore_registers		/* r1 = r2 -> nothing to do */
 	larl	%r4,.Lrestart_suspend_psw	/* Set new restart PSW */
-	mvc	__LC_RESTART_PSW(16,%r0),0(%r4)
+	mvc	__LC_RST_NEW_PSW(16,%r0),0(%r4)
 3:
 	sigp	%r9,%r1,__SIGP_INITIAL_CPU_RESET
 	brc	8,4f	/* accepted */
diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c
index 65065ac..a8f93f1 100644
--- a/arch/s390/kernel/time.c
+++ b/arch/s390/kernel/time.c
@@ -51,14 +51,6 @@
 #define USECS_PER_JIFFY     ((unsigned long) 1000000/HZ)
 #define CLK_TICKS_PER_JIFFY ((unsigned long) USECS_PER_JIFFY << 12)
 
-/*
- * Create a small time difference between the timer interrupts
- * on the different cpus to avoid lock contention.
- */
-#define CPU_DEVIATION       (smp_processor_id() << 12)
-
-#define TICK_SIZE tick
-
 u64 sched_clock_base_cc = -1;	/* Force to data section. */
 EXPORT_SYMBOL_GPL(sched_clock_base_cc);
 
diff --git a/arch/s390/kernel/vdso.c b/arch/s390/kernel/vdso.c
index 5f99e66..6bc9c19 100644
--- a/arch/s390/kernel/vdso.c
+++ b/arch/s390/kernel/vdso.c
@@ -23,6 +23,7 @@
 #include <linux/security.h>
 #include <linux/bootmem.h>
 #include <linux/compat.h>
+#include <asm/asm-offsets.h>
 #include <asm/pgtable.h>
 #include <asm/system.h>
 #include <asm/processor.h>
diff --git a/arch/s390/kvm/diag.c b/arch/s390/kvm/diag.c
index 8300309..9e4c841 100644
--- a/arch/s390/kvm/diag.c
+++ b/arch/s390/kvm/diag.c
@@ -39,7 +39,7 @@ static int __diag_ipl_functions(struct kvm_vcpu *vcpu)
 		vcpu->run->s390_reset_flags = 0;
 		break;
 	default:
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	}
 
 	atomic_clear_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags);
@@ -62,6 +62,6 @@ int kvm_s390_handle_diag(struct kvm_vcpu *vcpu)
 	case 0x308:
 		return __diag_ipl_functions(vcpu);
 	default:
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	}
 }
diff --git a/arch/s390/kvm/intercept.c b/arch/s390/kvm/intercept.c
index b400964..3ddc308 100644
--- a/arch/s390/kvm/intercept.c
+++ b/arch/s390/kvm/intercept.c
@@ -32,7 +32,7 @@ static int handle_lctlg(struct kvm_vcpu *vcpu)
 
 	vcpu->stat.instruction_lctlg++;
 	if ((vcpu->arch.sie_block->ipb & 0xff) != 0x2f)
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 
 	useraddr = disp2;
 	if (base2)
@@ -138,7 +138,7 @@ static int handle_stop(struct kvm_vcpu *vcpu)
 		rc = __kvm_s390_vcpu_store_status(vcpu,
 						  KVM_S390_STORE_STATUS_NOADDR);
 		if (rc >= 0)
-			rc = -ENOTSUPP;
+			rc = -EOPNOTSUPP;
 	}
 
 	if (vcpu->arch.local_int.action_bits & ACTION_RELOADVCPU_ON_STOP) {
@@ -150,7 +150,7 @@ static int handle_stop(struct kvm_vcpu *vcpu)
 	if (vcpu->arch.local_int.action_bits & ACTION_STOP_ON_STOP) {
 		vcpu->arch.local_int.action_bits &= ~ACTION_STOP_ON_STOP;
 		VCPU_EVENT(vcpu, 3, "%s", "cpu stopped");
-		rc = -ENOTSUPP;
+		rc = -EOPNOTSUPP;
 	}
 
 	spin_unlock_bh(&vcpu->arch.local_int.lock);
@@ -171,9 +171,9 @@ static int handle_validity(struct kvm_vcpu *vcpu)
 			 2*PAGE_SIZE);
 		if (rc)
 			/* user will receive sigsegv, exit to user */
-			rc = -ENOTSUPP;
+			rc = -EOPNOTSUPP;
 	} else
-		rc = -ENOTSUPP;
+		rc = -EOPNOTSUPP;
 
 	if (rc)
 		VCPU_EVENT(vcpu, 2, "unhandled validity intercept code %d",
@@ -189,7 +189,7 @@ static int handle_instruction(struct kvm_vcpu *vcpu)
 	handler = instruction_handlers[vcpu->arch.sie_block->ipa >> 8];
 	if (handler)
 		return handler(vcpu);
-	return -ENOTSUPP;
+	return -EOPNOTSUPP;
 }
 
 static int handle_prog(struct kvm_vcpu *vcpu)
@@ -206,7 +206,7 @@ static int handle_instruction_and_prog(struct kvm_vcpu *vcpu)
 	rc = handle_instruction(vcpu);
 	rc2 = handle_prog(vcpu);
 
-	if (rc == -ENOTSUPP)
+	if (rc == -EOPNOTSUPP)
 		vcpu->arch.sie_block->icptcode = 0x04;
 	if (rc)
 		return rc;
@@ -231,9 +231,9 @@ int kvm_handle_sie_intercept(struct kvm_vcpu *vcpu)
 	u8 code = vcpu->arch.sie_block->icptcode;
 
 	if (code & 3 || (code >> 2) >= ARRAY_SIZE(intercept_funcs))
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	func = intercept_funcs[code >> 2];
 	if (func)
 		return func(vcpu);
-	return -ENOTSUPP;
+	return -EOPNOTSUPP;
 }
diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c
index 43486c2..834774d 100644
--- a/arch/s390/kvm/interrupt.c
+++ b/arch/s390/kvm/interrupt.c
@@ -10,12 +10,12 @@
  *    Author(s): Carsten Otte <cotte@de.ibm.com>
  */
 
-#include <asm/lowcore.h>
-#include <asm/uaccess.h>
-#include <linux/hrtimer.h>
 #include <linux/interrupt.h>
 #include <linux/kvm_host.h>
+#include <linux/hrtimer.h>
 #include <linux/signal.h>
+#include <asm/asm-offsets.h>
+#include <asm/uaccess.h>
 #include "kvm-s390.h"
 #include "gaccess.h"
 
@@ -187,8 +187,8 @@ static void __do_deliver_interrupt(struct kvm_vcpu *vcpu,
 		if (rc == -EFAULT)
 			exception = 1;
 
-		rc = put_guest_u64(vcpu, __LC_PFAULT_INTPARM,
-			inti->ext.ext_params2);
+		rc = put_guest_u64(vcpu, __LC_EXT_PARAMS2,
+				   inti->ext.ext_params2);
 		if (rc == -EFAULT)
 			exception = 1;
 		break;
@@ -342,7 +342,7 @@ int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
 	if (psw_interrupts_disabled(vcpu)) {
 		VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
 		__unset_cpu_idle(vcpu);
-		return -ENOTSUPP; /* disabled wait */
+		return -EOPNOTSUPP; /* disabled wait */
 	}
 
 	if (psw_extint_disabled(vcpu) ||
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index f8bcaef..3fa0a10 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -23,6 +23,7 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/timer.h>
+#include <asm/asm-offsets.h>
 #include <asm/lowcore.h>
 #include <asm/pgtable.h>
 #include <asm/nmi.h>
@@ -543,7 +544,7 @@ rerun_vcpu:
 		rc = -EINTR;
 	}
 
-	if (rc == -ENOTSUPP) {
+	if (rc == -EOPNOTSUPP) {
 		/* intercept cannot be handled in-kernel, prepare kvm-run */
 		kvm_run->exit_reason         = KVM_EXIT_S390_SIEIC;
 		kvm_run->s390_sieic.icptcode = vcpu->arch.sie_block->icptcode;
@@ -603,45 +604,45 @@ int __kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
 	} else
 		prefix = 0;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, fp_regs),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, fp_regs),
 			vcpu->arch.guest_fpregs.fprs, 128, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, gp_regs),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, gp_regs),
 			vcpu->arch.guest_gprs, 128, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, psw),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, psw),
 			&vcpu->arch.sie_block->gpsw, 16, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, pref_reg),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, pref_reg),
 			&vcpu->arch.sie_block->prefix, 4, prefix))
 		return -EFAULT;
 
 	if (__guestcopy(vcpu,
-			addr + offsetof(struct save_area_s390x, fp_ctrl_reg),
+			addr + offsetof(struct save_area, fp_ctrl_reg),
 			&vcpu->arch.guest_fpregs.fpc, 4, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, tod_reg),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, tod_reg),
 			&vcpu->arch.sie_block->todpr, 4, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, timer),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, timer),
 			&vcpu->arch.sie_block->cputm, 8, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, clk_cmp),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, clk_cmp),
 			&vcpu->arch.sie_block->ckc, 8, prefix))
 		return -EFAULT;
 
-	if (__guestcopy(vcpu, addr + offsetof(struct save_area_s390x, acc_regs),
+	if (__guestcopy(vcpu, addr + offsetof(struct save_area, acc_regs),
 			&vcpu->arch.guest_acrs, 64, prefix))
 		return -EFAULT;
 
 	if (__guestcopy(vcpu,
-			addr + offsetof(struct save_area_s390x, ctrl_regs),
+			addr + offsetof(struct save_area, ctrl_regs),
 			&vcpu->arch.sie_block->gcr, 128, prefix))
 		return -EFAULT;
 	return 0;
diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c
index d426aac..28c5567 100644
--- a/arch/s390/kvm/priv.c
+++ b/arch/s390/kvm/priv.c
@@ -323,5 +323,5 @@ int kvm_s390_handle_b2(struct kvm_vcpu *vcpu)
 		else
 			return handler(vcpu);
 	}
-	return -ENOTSUPP;
+	return -EOPNOTSUPP;
 }
diff --git a/arch/s390/kvm/sigp.c b/arch/s390/kvm/sigp.c
index 15ee111..241a484 100644
--- a/arch/s390/kvm/sigp.c
+++ b/arch/s390/kvm/sigp.c
@@ -172,7 +172,7 @@ static int __sigp_set_arch(struct kvm_vcpu *vcpu, u32 parameter)
 		rc = 0; /* order accepted */
 		break;
 	default:
-		rc = -ENOTSUPP;
+		rc = -EOPNOTSUPP;
 	}
 	return rc;
 }
@@ -293,7 +293,7 @@ int kvm_s390_handle_sigp(struct kvm_vcpu *vcpu)
 		vcpu->stat.instruction_sigp_restart++;
 		/* user space must know about restart */
 	default:
-		return -ENOTSUPP;
+		return -EOPNOTSUPP;
 	}
 
 	if (rc < 0)
diff --git a/arch/s390/lib/Makefile b/arch/s390/lib/Makefile
index 97975ec..cd54a1c 100644
--- a/arch/s390/lib/Makefile
+++ b/arch/s390/lib/Makefile
@@ -2,7 +2,7 @@
 # Makefile for s390-specific library files..
 #
 
-lib-y += delay.o string.o uaccess_std.o uaccess_pt.o
+lib-y += delay.o string.o uaccess_std.o uaccess_pt.o usercopy.o
 obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o
 lib-$(CONFIG_64BIT) += uaccess_mvcos.o
 lib-$(CONFIG_SMP) += spinlock.o
diff --git a/arch/s390/lib/spinlock.c b/arch/s390/lib/spinlock.c
index cff327f..91754ff 100644
--- a/arch/s390/lib/spinlock.c
+++ b/arch/s390/lib/spinlock.c
@@ -43,16 +43,24 @@ void arch_spin_lock_wait(arch_spinlock_t *lp)
 {
 	int count = spin_retry;
 	unsigned int cpu = ~smp_processor_id();
+	unsigned int owner;
 
 	while (1) {
-		if (count-- <= 0) {
-			unsigned int owner = lp->owner_cpu;
-			if (owner != 0)
-				_raw_yield_cpu(~owner);
-			count = spin_retry;
+		owner = lp->owner_cpu;
+		if (!owner || smp_vcpu_scheduled(~owner)) {
+			for (count = spin_retry; count > 0; count--) {
+				if (arch_spin_is_locked(lp))
+					continue;
+				if (_raw_compare_and_swap(&lp->owner_cpu, 0,
+							  cpu) == 0)
+					return;
+			}
+			if (MACHINE_IS_LPAR)
+				continue;
 		}
-		if (arch_spin_is_locked(lp))
-			continue;
+		owner = lp->owner_cpu;
+		if (owner)
+			_raw_yield_cpu(~owner);
 		if (_raw_compare_and_swap(&lp->owner_cpu, 0, cpu) == 0)
 			return;
 	}
@@ -63,17 +71,27 @@ void arch_spin_lock_wait_flags(arch_spinlock_t *lp, unsigned long flags)
 {
 	int count = spin_retry;
 	unsigned int cpu = ~smp_processor_id();
+	unsigned int owner;
 
 	local_irq_restore(flags);
 	while (1) {
-		if (count-- <= 0) {
-			unsigned int owner = lp->owner_cpu;
-			if (owner != 0)
-				_raw_yield_cpu(~owner);
-			count = spin_retry;
+		owner = lp->owner_cpu;
+		if (!owner || smp_vcpu_scheduled(~owner)) {
+			for (count = spin_retry; count > 0; count--) {
+				if (arch_spin_is_locked(lp))
+					continue;
+				local_irq_disable();
+				if (_raw_compare_and_swap(&lp->owner_cpu, 0,
+							  cpu) == 0)
+					return;
+				local_irq_restore(flags);
+			}
+			if (MACHINE_IS_LPAR)
+				continue;
 		}
-		if (arch_spin_is_locked(lp))
-			continue;
+		owner = lp->owner_cpu;
+		if (owner)
+			_raw_yield_cpu(~owner);
 		local_irq_disable();
 		if (_raw_compare_and_swap(&lp->owner_cpu, 0, cpu) == 0)
 			return;
@@ -100,8 +118,11 @@ EXPORT_SYMBOL(arch_spin_trylock_retry);
 void arch_spin_relax(arch_spinlock_t *lock)
 {
 	unsigned int cpu = lock->owner_cpu;
-	if (cpu != 0)
-		_raw_yield_cpu(~cpu);
+	if (cpu != 0) {
+		if (MACHINE_IS_VM || MACHINE_IS_KVM ||
+		    !smp_vcpu_scheduled(~cpu))
+			_raw_yield_cpu(~cpu);
+	}
 }
 EXPORT_SYMBOL(arch_spin_relax);
 
diff --git a/arch/s390/lib/usercopy.c b/arch/s390/lib/usercopy.c
new file mode 100644
index 0000000..14b363f
--- /dev/null
+++ b/arch/s390/lib/usercopy.c
@@ -0,0 +1,8 @@
+#include <linux/module.h>
+#include <linux/bug.h>
+
+void copy_from_user_overflow(void)
+{
+	WARN(1, "Buffer overflow detected!\n");
+}
+EXPORT_SYMBOL(copy_from_user_overflow);
diff --git a/arch/s390/mm/extmem.c b/arch/s390/mm/extmem.c
index 5c84571..6409fd5 100644
--- a/arch/s390/mm/extmem.c
+++ b/arch/s390/mm/extmem.c
@@ -309,7 +309,7 @@ query_segment_type (struct dcss_segment *seg)
 	}
 #endif
 	if (qout->segcnt > 6) {
-		rc = -ENOTSUPP;
+		rc = -EOPNOTSUPP;
 		goto out_free;
 	}
 
@@ -324,11 +324,11 @@ query_segment_type (struct dcss_segment *seg)
 		for (i=0; i<qout->segcnt; i++) {
 			if (((qout->range[i].start & 0xff) != SEG_TYPE_EW) &&
 			    ((qout->range[i].start & 0xff) != SEG_TYPE_EN)) {
-				rc = -ENOTSUPP;
+				rc = -EOPNOTSUPP;
 				goto out_free;
 			}
 			if (start != qout->range[i].start >> PAGE_SHIFT) {
-				rc = -ENOTSUPP;
+				rc = -EOPNOTSUPP;
 				goto out_free;
 			}
 			start = (qout->range[i].end >> PAGE_SHIFT) + 1;
@@ -357,7 +357,7 @@ query_segment_type (struct dcss_segment *seg)
  * -ENOSYS  : we are not running on VM
  * -EIO     : could not perform query diagnose
  * -ENOENT  : no such segment
- * -ENOTSUPP: multi-part segment cannot be used with linux
+ * -EOPNOTSUPP: multi-part segment cannot be used with linux
  * -ENOMEM  : out of memory
  * 0 .. 6   : type of segment as defined in include/asm-s390/extmem.h
  */
@@ -515,7 +515,7 @@ __segment_load (char *name, int do_nonshared, unsigned long *addr, unsigned long
  * -ENOSYS  : we are not running on VM
  * -EIO     : could not perform query or load diagnose
  * -ENOENT  : no such segment
- * -ENOTSUPP: multi-part segment cannot be used with linux
+ * -EOPNOTSUPP: multi-part segment cannot be used with linux
  * -ENOSPC  : segment cannot be used (overlaps with storage)
  * -EBUSY   : segment can temporarily not be used (overlaps with dcss)
  * -ERANGE  : segment cannot be used (exceeds kernel mapping range)
@@ -742,7 +742,7 @@ void segment_warning(int rc, char *seg_name)
 		pr_err("Loading or querying DCSS %s resulted in a "
 		       "hardware error\n", seg_name);
 		break;
-	case -ENOTSUPP:
+	case -EOPNOTSUPP:
 		pr_err("DCSS %s has multiple page ranges and cannot be "
 		       "loaded or queried\n", seg_name);
 		break;
diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
index fc102e7..3040d7c 100644
--- a/arch/s390/mm/fault.c
+++ b/arch/s390/mm/fault.c
@@ -30,6 +30,7 @@
 #include <linux/kprobes.h>
 #include <linux/uaccess.h>
 #include <linux/hugetlb.h>
+#include <asm/asm-offsets.h>
 #include <asm/system.h>
 #include <asm/pgtable.h>
 #include <asm/s390_ext.h>
@@ -59,15 +60,13 @@ static inline int notify_page_fault(struct pt_regs *regs)
 {
 	int ret = 0;
 
-#ifdef CONFIG_KPROBES
 	/* kprobe_running() needs smp_processor_id() */
-	if (!user_mode(regs)) {
+	if (kprobes_built_in() && !user_mode(regs)) {
 		preempt_disable();
 		if (kprobe_running() && kprobe_fault_handler(regs, 14))
 			ret = 1;
 		preempt_enable();
 	}
-#endif
 	return ret;
 }
 
diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
index 7656479..d5865e4 100644
--- a/arch/s390/mm/init.c
+++ b/arch/s390/mm/init.c
@@ -143,33 +143,34 @@ void kernel_map_pages(struct page *page, int numpages, int enable)
 }
 #endif
 
-void free_initmem(void)
+void free_init_pages(char *what, unsigned long begin, unsigned long end)
 {
-        unsigned long addr;
+	unsigned long addr = begin;
 
-        addr = (unsigned long)(&__init_begin);
-        for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) {
+	if (begin >= end)
+		return;
+	for (; addr < end; addr += PAGE_SIZE) {
 		ClearPageReserved(virt_to_page(addr));
 		init_page_count(virt_to_page(addr));
-		memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE);
+		memset((void *)(addr & PAGE_MASK), POISON_FREE_INITMEM,
+		       PAGE_SIZE);
 		free_page(addr);
 		totalram_pages++;
-        }
-        printk ("Freeing unused kernel memory: %ldk freed\n",
-		((unsigned long)&__init_end - (unsigned long)&__init_begin) >> 10);
+	}
+	printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10);
+}
+
+void free_initmem(void)
+{
+	free_init_pages("unused kernel memory",
+			(unsigned long)&__init_begin,
+			(unsigned long)&__init_end);
 }
 
 #ifdef CONFIG_BLK_DEV_INITRD
 void free_initrd_mem(unsigned long start, unsigned long end)
 {
-        if (start < end)
-                printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
-        for (; start < end; start += PAGE_SIZE) {
-                ClearPageReserved(virt_to_page(start));
-                init_page_count(virt_to_page(start));
-                free_page(start);
-                totalram_pages++;
-        }
+	free_init_pages("initrd memory", start, end);
 }
 #endif