rtai-source 3.9.1-4 / usr / src / kernel-patches / diffs / rtai / hal-linux-2.6.29.4-powerpc-2.6-02.patch

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 69425b2..dccfb35 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -124,6 +124,7 @@ config PPC
 	select USE_GENERIC_SMP_HELPERS if SMP
 	select HAVE_OPROFILE
 	select HAVE_SYSCALL_WRAPPERS if PPC64
+	select HAVE_FUNCTION_TRACE_MCOUNT_TEST
 
 config EARLY_PRINTK
 	bool
@@ -139,6 +140,10 @@ config SYSVIPC_COMPAT
 	depends on COMPAT && SYSVIPC
 	default y
 
+config SOFTDISABLE
+	bool
+	default (PPC64 && !IPIPE)
+
 # All PPC32s use generic nvram driver through ppc_md
 config GENERIC_NVRAM
 	bool
@@ -236,6 +241,24 @@ source "arch/powerpc/platforms/Kconfig"
 
 menu "Kernel options"
 
+source "kernel/ipipe/Kconfig"
+
+if IPIPE
+config RUNLATCH
+	bool "Enable RUNLATCH support"
+	depends on PPC64
+	default n if IPIPE
+	---help---
+	This option is costly latency-wise, so default is to keep
+	it off when the interrupt pipeline is enabled.
+endif
+if !IPIPE
+config RUNLATCH
+	bool
+	depends on PPC64
+	default y
+endif
+
 config HIGHMEM
 	bool "High memory support"
 	depends on PPC32
diff --git a/arch/powerpc/boot/Makefile b/arch/powerpc/boot/Makefile
index e84df33..7fc0dfb 100644
--- a/arch/powerpc/boot/Makefile
+++ b/arch/powerpc/boot/Makefile
@@ -29,6 +29,14 @@ ifdef CONFIG_DEBUG_INFO
 BOOTCFLAGS	+= -g
 endif
 
+ifdef CONFIG_IPIPE_TRACE
+# do not trace the boot loader
+nullstring :=
+space      := $(nullstring) # end of the line
+pg_flag     = $(nullstring) -pg # end of the line
+BOOTCFLAGS     := $(subst ${pg_flag},${space},${BOOTCFLAGS})
+endif
+
 ifeq ($(call cc-option-yn, -fstack-protector),y)
 BOOTCFLAGS	+= -fno-stack-protector
 endif
diff --git a/arch/powerpc/include/asm/exception.h b/arch/powerpc/include/asm/exception.h
index d3d4534..0e88328 100644
--- a/arch/powerpc/include/asm/exception.h
+++ b/arch/powerpc/include/asm/exception.h
@@ -47,6 +47,30 @@
 #define EX_R3		64
 #define EX_LR		72
 
+#ifdef CONFIG_SOFTDISABLE	
+#define COPY_SOFTISTATE(mreg)			\
+	lbz	mreg,PACASOFTIRQEN(r13);	\
+	std	mreg,SOFTE(r1);
+#define TEST_SOFTISTATE(mreg, dlabel)		\
+	lbz	mreg,PACASOFTIRQEN(r13);	\
+	cmpwi	mreg,0;				\
+	beq-	dlabel;
+#else
+#ifdef CONFIG_IPIPE
+/* Do NOT alter Rc(eq) in this code;  our caller uses it. */
+#define COPY_SOFTISTATE(mreg)			\
+	ld	mreg,PACAROOTPCPU(r13);		\
+	ld	mreg,0(mreg);			\
+	nor	mreg,mreg,mreg;			\
+	clrldi	mreg,mreg,63;			\
+	std	mreg,SOFTE(r1);
+#define TEST_SOFTISTATE(mreg, dlabel)
+#else
+#define COPY_SOFTISTATE(mreg)
+#define TEST_SOFTISTATE(mreg, dlabel)
+#endif
+#endif
+
 /*
  * We're short on space and time in the exception prolog, so we can't
  * use the normal SET_REG_IMMEDIATE macro. Normally we just need the
@@ -125,9 +149,8 @@
 	std	r9,_LINK(r1);						   \
 	mfctr	r10;			/* save CTR in stackframe	*/ \
 	std	r10,_CTR(r1);						   \
-	lbz	r10,PACASOFTIRQEN(r13);				   \
+	COPY_SOFTISTATE(r10);						   \
 	mfspr	r11,SPRN_XER;		/* save XER in stackframe	*/ \
-	std	r10,SOFTE(r1);						   \
 	std	r11,_XER(r1);						   \
 	li	r9,(n)+1;						   \
 	std	r9,_TRAP(r1);		/* set trap number		*/ \
@@ -171,10 +194,8 @@ label##_pSeries:							\
 	mfspr	r13,SPRN_SPRG3;		/* get paca address into r13 */	\
 	std	r9,PACA_EXGEN+EX_R9(r13);	/* save r9, r10 */	\
 	std	r10,PACA_EXGEN+EX_R10(r13);				\
-	lbz	r10,PACASOFTIRQEN(r13);					\
 	mfcr	r9;							\
-	cmpwi	r10,0;							\
-	beq	masked_interrupt;					\
+	TEST_SOFTISTATE(r10, masked_interrupt);				\
 	mfspr	r10,SPRN_SPRG1;						\
 	std	r10,PACA_EXGEN+EX_R13(r13);				\
 	std	r11,PACA_EXGEN+EX_R11(r13);				\
@@ -189,6 +210,28 @@ label##_pSeries:							\
 	rfid;								\
 	b	.	/* prevent speculative execution */
 
+#ifdef CONFIG_IPIPE
+/* IBM legacy I-Series are not supported. */
+#define ENABLE_INTS				\
+	ld	r12,_MSR(r1);			\
+	mfmsr	r11;				\
+	rlwimi	r11,r12,0,MSR_EE;		\
+	mtmsrd	r11,1
+#define DISABLE_INTS	/* We lie, mostly... */	\
+	ld	r11,PACAROOTPCPU(r13);		\
+	ld	r10,0(r11);			\
+	ori	r10,r10,1;			\
+	std	r10,0(r11);			\
+	mfmsr	r10;				\
+	ori	r10,r10,MSR_EE;			\
+	mtmsrd	r10,1;
+#define DISABLE_INTS_REALLY			\
+	mfmsr	r11;				\
+	rldicl	r11,r11,48,1;/* clear MSR_EE */	\
+	rotldi	r11,r11,16;			\
+	mtmsrd	r11,1;
+#else /* !CONFIG_IPIPE */
+
 #ifdef CONFIG_PPC_ISERIES
 #define DISABLE_INTS				\
 	li	r11,0;				\
@@ -209,13 +252,15 @@ END_FW_FTR_SECTION_IFSET(FW_FEATURE_ISERIES)
 	stb	r11,PACAHARDIRQEN(r13);		\
 	TRACE_DISABLE_INTS
 #endif /* CONFIG_PPC_ISERIES */
-
+		
 #define ENABLE_INTS				\
 	ld	r12,_MSR(r1);			\
 	mfmsr	r11;				\
 	rlwimi	r11,r12,0,MSR_EE;		\
 	mtmsrd	r11,1
 
+#endif /* !CONFIG_IPIPE */
+
 #define STD_EXCEPTION_COMMON(trap, label, hdlr)		\
 	.align	7;					\
 	.globl label##_common;				\
@@ -223,6 +268,7 @@ label##_common:						\
 	EXCEPTION_PROLOG_COMMON(trap, PACA_EXGEN);	\
 	DISABLE_INTS;					\
 	bl	.save_nvgprs;				\
+	TRACE_DISABLE_INTS;				\
 	addi	r3,r1,STACK_FRAME_OVERHEAD;		\
 	bl	hdlr;					\
 	b	.ret_from_except
@@ -239,6 +285,7 @@ label##_common:						\
 	FINISH_NAP;					\
 	DISABLE_INTS;					\
 	bl	.save_nvgprs;				\
+	TRACE_DISABLE_INTS;				\
 	addi	r3,r1,STACK_FRAME_OVERHEAD;		\
 	bl	hdlr;					\
 	b	.ret_from_except
@@ -253,10 +300,24 @@ label##_common:						\
 BEGIN_FTR_SECTION					\
 	bl	.ppc64_runlatch_on;			\
 END_FTR_SECTION_IFSET(CPU_FTR_CTRL)			\
+	TRACE_DISABLE_INTS;				\
 	addi	r3,r1,STACK_FRAME_OVERHEAD;		\
 	bl	hdlr;					\
 	b	.ret_from_except_lite
 
+#ifdef CONFIG_IPIPE
+#define IPIPE_EXCEPTION_COMMON_LITE(trap, label, hdlr)	\
+	.align	7;					\
+	.globl label##_common;				\
+label##_common:						\
+	EXCEPTION_PROLOG_COMMON(trap, PACA_EXGEN);	\
+	DISABLE_INTS_REALLY;				\
+	TRACE_DISABLE_INTS_REALLY;			\
+	addi	r3,r1,STACK_FRAME_OVERHEAD;		\
+	bl	hdlr;					\
+	b	.__ipipe_ret_from_except_lite
+#endif /* CONFIG_IPIPE */
+
 /*
  * When the idle code in power4_idle puts the CPU into NAP mode,
  * it has to do so in a loop, and relies on the external interrupt
diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index f75a5fc..32242c5 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h
@@ -14,6 +14,8 @@
 extern void timer_interrupt(struct pt_regs *);
 
 #ifdef CONFIG_PPC64
+
+#ifdef CONFIG_SOFTDISABLE
 #include <asm/paca.h>
 
 static inline unsigned long local_get_flags(void)
@@ -64,17 +66,173 @@ static inline int irqs_disabled_flags(unsigned long flags)
 	return flags == 0;
 }
 
-#else
+#define local_irq_save_hw(x)	raw_local_irq_save(x)
+#define local_irq_restore_hw(x)	raw_local_irq_restore(x)
+#define local_irq_enable_hw()	raw_local_irq_enable()
+#define local_irq_disable_hw()	raw_local_irq_disable()
+#define irqs_disabled_hw()	raw_irqs_disabled()
+
+#else /* !CONFIG_SOFTDISABLE */
+
+#ifdef CONFIG_IPIPE
+
+#include <linux/ipipe_base.h>
+#include <linux/ipipe_trace.h>
+
+#define raw_local_save_flags(x)		do {			\
+		(x) = (!__ipipe_test_root()) << MSR_EE_LG;	\
+		__asm__ __volatile__("": : :"memory");		\
+	} while(0)
+
+#define raw_local_irq_restore(x)	do {			\
+		__asm__ __volatile__("": : :"memory");		\
+		__ipipe_restore_root(!((x) & MSR_EE));		\
+	} while(0)
+
+static inline void raw_local_irq_save_ptr(unsigned long *x)
+{
+	*x = (!__ipipe_test_and_stall_root()) << MSR_EE_LG;
+	barrier();
+}
+
+#define raw_local_irq_save(x)			\
+do {						\
+	ipipe_check_context(ipipe_root_domain);	\
+	raw_local_irq_save_ptr(&(x));		\
+} while(0)
+
+#define hard_irq_enable()	do {				\
+		barrier();					\
+		__ipipe_unstall_root();				\
+	} while(0)
+
+#define hard_irq_disable()	do {				\
+		ipipe_check_context(ipipe_root_domain);		\
+		__ipipe_stall_root();				\
+		barrier();					\
+	} while(0)
+
+#define raw_local_irq_disable()	hard_irq_disable()
+#define raw_local_irq_enable()	hard_irq_enable()
+#define raw_irqs_disabled()	(__ipipe_test_root() != 0)
+
+static inline int raw_irqs_disabled_flags(unsigned long x)
+{
+	return !(x & MSR_EE);
+}
+
+static inline unsigned long raw_mangle_irq_bits(int virt, unsigned long real)
+{
+	/* Merge virtual and real interrupt mask bits into a single
+	   64bit word. We know MSR_EE will not conflict with 1L<<31. */
+	return (real & ~(1L << 31)) | ((long)virt << 31);
+}
+
+static inline int raw_demangle_irq_bits(unsigned long *x)
+{
+	int virt = (*x & (1L << 31)) != 0;
+	*x &= ~(1L << 31);
+	return virt;
+}
+
+#define local_irq_disable_hw_notrace()		__mtmsrd(mfmsr() & ~MSR_EE, 1)
+#define local_irq_enable_hw_notrace()		__mtmsrd(mfmsr() | MSR_EE, 1)
+#define local_irq_save_hw_notrace(x)		({ (x) = __local_irq_save_hw(); })
+#define local_irq_restore_hw_notrace(x)		__mtmsrd(x, 1)
+
+static inline unsigned long __local_irq_save_hw(void)
+{
+	unsigned long msr = mfmsr();
+	local_irq_disable_hw_notrace();
+	__asm__ __volatile__("": : :"memory");
+	return msr;
+}
+	
+#define local_save_flags_hw(x)		((x) = mfmsr())
+#define irqs_disabled_hw()		((mfmsr() & MSR_EE) == 0)
+
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+#define local_irq_disable_hw() do {			\
+		if (!irqs_disabled_hw()) {		\
+			local_irq_disable_hw_notrace();	\
+			ipipe_trace_begin(0x80000000);	\
+		}					\
+	} while (0)
+#define local_irq_enable_hw() do {			\
+		if (irqs_disabled_hw()) {		\
+			ipipe_trace_end(0x80000000);	\
+			local_irq_enable_hw_notrace();	\
+		}					\
+	} while (0)
+#define local_irq_save_hw(x) do {			\
+		local_save_flags_hw(x);			\
+		if ((x) & MSR_EE) {			\
+			local_irq_disable_hw_notrace();	\
+			ipipe_trace_begin(0x80000001);	\
+		}					\
+	} while (0)
+#define local_irq_restore_hw(x) do {			\
+		if ((x) & MSR_EE)			\
+			ipipe_trace_end(0x80000001);	\
+		local_irq_restore_hw_notrace(x);	\
+	} while (0)
+#else /* !CONFIG_IPIPE_TRACE_IRQSOFF */
+#define local_irq_save_hw(x)	local_irq_save_hw_notrace(x)
+#define local_irq_restore_hw(x)	local_irq_restore_hw_notrace(x)
+#define local_irq_enable_hw()	local_irq_enable_hw_notrace()
+#define local_irq_disable_hw()	local_irq_disable_hw_notrace()
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
+
+#else /* !CONFIG_IPIPE */
+
+#define hard_irq_enable()	__mtmsrd(mfmsr() | MSR_EE, 1)
+#define hard_irq_disable()	__mtmsrd(mfmsr() & ~MSR_EE, 1)
+
+#define raw_local_save_flags(x)		((x) = mfmsr())
+#define raw_local_irq_restore(x) 	__mtmsrd(x, 1)
+#define raw_irqs_disabled()		((mfmsr() & MSR_EE) == 0)
+
+#define local_irq_save_hw(x)	raw_local_irq_save(x)
+#define local_irq_restore_hw(x)	raw_local_irq_restore(x)
+#define local_irq_enable_hw()	hard_irq_enable()
+#define local_irq_disable_hw()	hard_irq_disable()
+#define irqs_disabled_hw()	raw_irqs_disabled()
+
+#endif /* !CONFIG_IPIPE */
+
+#endif /* !CONFIG_SOFTDISABLE */
+
+#else /* !CONFIG_PPC64 */
+
+static inline unsigned long raw_mangle_irq_bits(int virt, unsigned long real)
+{
+	/* Merge virtual and real interrupt mask bits into a single
+	   32bit word. */
+	return (real & ~(1 << 31)) | ((virt != 0) << 31);
+}
+
+static inline int raw_demangle_irq_bits(unsigned long *x)
+{
+	int virt = (*x & (1 << 31)) != 0;
+	*x &= ~(1L << 31);
+	return virt;
+}
+
+#define local_save_flags_hw(x)			((x) = mfmsr())
+#define local_test_iflag_hw(x)			((x) & MSR_EE)
+#define irqs_disabled_hw()			((mfmsr() & MSR_EE) == 0)
+#define local_irq_save_hw_notrace(x)		local_irq_save_ptr_hw(&(x))
+#define raw_irqs_disabled_flags(x)		(!local_test_iflag_hw(x))
 
 #if defined(CONFIG_BOOKE)
-#define SET_MSR_EE(x)	mtmsr(x)
-#define local_irq_restore(flags)	__asm__ __volatile__("wrtee %0" : : "r" (flags) : "memory")
+#define local_irq_restore_hw_notrace(x)	\
+	__asm__ __volatile__("wrtee %0" : : "r" (x) : "memory")
 #else
 #define SET_MSR_EE(x)	mtmsr(x)
-#define local_irq_restore(flags)	mtmsr(flags)
+#define local_irq_restore_hw_notrace(x)	mtmsr(x)
 #endif
 
-static inline void local_irq_disable(void)
+static inline void local_irq_disable_hw_notrace(void)
 {
 #ifdef CONFIG_BOOKE
 	__asm__ __volatile__("wrteei 0": : :"memory");
@@ -86,7 +244,7 @@ static inline void local_irq_disable(void)
 #endif
 }
 
-static inline void local_irq_enable(void)
+static inline void local_irq_enable_hw_notrace(void)
 {
 #ifdef CONFIG_BOOKE
 	__asm__ __volatile__("wrteei 1": : :"memory");
@@ -98,11 +256,11 @@ static inline void local_irq_enable(void)
 #endif
 }
 
-static inline void local_irq_save_ptr(unsigned long *flags)
+static inline void local_irq_save_ptr_hw(unsigned long *x)
 {
 	unsigned long msr;
 	msr = mfmsr();
-	*flags = msr;
+	*x = msr;
 #ifdef CONFIG_BOOKE
 	__asm__ __volatile__("wrteei 0": : :"memory");
 #else
@@ -111,9 +269,107 @@ static inline void local_irq_save_ptr(unsigned long *flags)
 	__asm__ __volatile__("": : :"memory");
 }
 
-#define local_save_flags(flags)	((flags) = mfmsr())
-#define local_irq_save(flags)	local_irq_save_ptr(&flags)
-#define irqs_disabled()		((mfmsr() & MSR_EE) == 0)
+#ifdef CONFIG_IPIPE
+
+#include <linux/ipipe_base.h>
+#include <linux/ipipe_trace.h>
+
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+
+static inline void local_irq_disable_hw(void)
+{
+	if (!irqs_disabled_hw()) {
+		local_irq_disable_hw_notrace();
+		ipipe_trace_begin(0x80000000);
+	}
+}
+
+static inline void local_irq_enable_hw(void)
+{
+	if (irqs_disabled_hw()) {
+		ipipe_trace_end(0x80000000);
+		local_irq_enable_hw_notrace();
+	}
+}
+
+#define local_irq_save_hw(x) \
+do {						       \
+	local_irq_save_ptr_hw(&(x));		       \
+	if (local_test_iflag_hw(x))		       \
+		ipipe_trace_begin(0x80000001);	       \
+} while(0)
+
+static inline void local_irq_restore_hw(unsigned long x)
+{
+	if (local_test_iflag_hw(x))
+		ipipe_trace_end(0x80000001);
+
+	local_irq_restore_hw_notrace(x);
+}
+
+#else /* !CONFIG_IPIPE_TRACE_IRQSOFF */
+
+#define local_irq_disable_hw    local_irq_disable_hw_notrace
+#define local_irq_enable_hw     local_irq_enable_hw_notrace
+#define local_irq_save_hw       local_irq_save_hw_notrace
+#define local_irq_restore_hw    local_irq_restore_hw_notrace
+
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
+
+static inline void local_irq_disable(void)
+{
+	ipipe_check_context(ipipe_root_domain);
+	__ipipe_stall_root();
+	barrier();
+}
+
+static inline void local_irq_enable(void)
+{
+	barrier();
+	__ipipe_unstall_root();
+}
+
+static inline void local_irq_save_ptr(unsigned long *x)
+{
+	*x = (!__ipipe_test_and_stall_root()) << MSR_EE_LG;
+	barrier();
+}
+
+static inline void local_irq_restore(unsigned long x)
+{
+	barrier();
+	__ipipe_restore_root(!(x & MSR_EE));
+}
+
+#define local_save_flags(x)			\
+do {						\
+	(x) = (!__ipipe_test_root()) << MSR_EE_LG;	\
+	barrier();				\
+} while(0)
+
+#define local_irq_save(x)			\
+do {						\
+	ipipe_check_context(ipipe_root_domain);	\
+	local_irq_save_ptr(&(x));		\
+} while(0)
+
+#define irqs_disabled()		__ipipe_test_root()
+
+#else /* !CONFIG_IPIPE */
+
+#define local_irq_disable_hw    	local_irq_disable_hw_notrace
+#define local_irq_enable_hw     	local_irq_enable_hw_notrace
+#define local_irq_save_hw       	local_irq_save_hw_notrace
+#define local_irq_restore_hw    	local_irq_restore_hw_notrace
+#define local_irq_restore(x)		local_irq_restore_hw(x)
+#define local_irq_disable()		local_irq_disable_hw()
+#define local_irq_enable()		local_irq_enable_hw()
+#define local_irq_save_ptr(x)		local_irq_save_ptr_hw(x)
+#define irqs_disabled()			irqs_disabled_hw()
+#define local_save_flags(x)		local_save_flags_hw(x)
+#define local_irq_save(x)  		local_irq_save_hw(x)
+
+#endif /* !CONFIG_IPIPE */
 
 #define hard_irq_enable()	local_irq_enable()
 #define hard_irq_disable()	local_irq_disable()
diff --git a/arch/powerpc/include/asm/ipipe.h b/arch/powerpc/include/asm/ipipe.h
new file mode 100644
index 0000000..fc39c55
--- /dev/null
+++ b/arch/powerpc/include/asm/ipipe.h
@@ -0,0 +1,253 @@
+/*
+ *   include/asm-powerpc/ipipe.h
+ *
+ *   I-pipe 32/64bit merge - Copyright (C) 2007 Philippe Gerum.
+ *   I-pipe PA6T support - Copyright (C) 2007 Philippe Gerum.
+ *   I-pipe 64-bit PowerPC port - Copyright (C) 2005 Heikki Lindholm.
+ *   I-pipe PowerPC support - Copyright (C) 2002-2005 Philippe Gerum.
+ *
+ *   This program is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ *   USA; either version 2 of the License, or (at your option) any later
+ *   version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __ASM_POWERPC_IPIPE_H
+#define __ASM_POWERPC_IPIPE_H
+
+#ifdef CONFIG_IPIPE
+
+#include <asm/ptrace.h>
+#include <asm/hw_irq.h>
+#include <asm/irq.h>
+#include <asm/bitops.h>
+#include <asm/time.h>
+#include <linux/ipipe_percpu.h>
+#include <linux/list.h>
+#include <linux/cpumask.h>
+#include <linux/cache.h>
+#include <linux/threads.h>
+
+#ifdef CONFIG_PPC64
+#ifdef CONFIG_PPC_ISERIES
+#error "I-pipe: IBM I-series not supported, sorry"
+#endif
+#include <asm/paca.h>
+#endif
+
+#define IPIPE_ARCH_STRING	"2.6-02"
+#define IPIPE_MAJOR_NUMBER	2
+#define IPIPE_MINOR_NUMBER	6
+#define IPIPE_PATCH_NUMBER	2
+
+#ifdef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+
+#define prepare_arch_switch(next)			\
+	do {						\
+		local_irq_enable_hw();			\
+		ipipe_schedule_notify(current ,next);	\
+	} while(0)
+
+#define task_hijacked(p)						\
+	({								\
+		unsigned long __flags__;				\
+		int __x__;						\
+		local_irq_save_hw_smp(__flags__);			\
+		__x__ = __ipipe_root_domain_p;				\
+		__clear_bit(IPIPE_SYNC_FLAG, &ipipe_root_cpudom_var(status)); \
+		local_irq_restore_hw_smp(__flags__);			\
+		!__x__;							\
+	})
+
+#else /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
+
+#define prepare_arch_switch(next)			\
+	do {						\
+		ipipe_schedule_notify(current ,next);	\
+		local_irq_disable_hw();			\
+	} while(0)
+
+#define task_hijacked(p)						\
+	({								\
+		int __x__ = __ipipe_root_domain_p;			\
+		__clear_bit(IPIPE_SYNC_FLAG, &ipipe_root_cpudom_var(status)); \
+		if (__x__) local_irq_enable_hw(); !__x__;		\
+	})
+
+#endif /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
+
+struct ipipe_domain;
+
+struct ipipe_sysinfo {
+
+	int ncpus;			/* Number of CPUs on board */
+	u64 cpufreq;			/* CPU frequency (in Hz) */
+
+	/* Arch-dependent block */
+
+	struct {
+		unsigned tmirq;		/* Decrementer virtual IRQ */
+		u64 tmfreq;		/* Timebase frequency */
+	} archdep;
+};
+
+#ifdef CONFIG_DEBUGGER
+extern cpumask_t __ipipe_dbrk_pending;
+#endif
+
+#ifdef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+struct mm;
+DECLARE_PER_CPU(struct mm_struct *, ipipe_active_mm);
+#endif
+
+#define ipipe_cpu_freq()	ppc_tb_freq
+#ifdef CONFIG_PPC64
+#define ipipe_read_tsc(t)	(t = mftb())
+#define ipipe_tsc2ns(t)		(((t) * 1000UL) / (ipipe_cpu_freq() / 1000000UL))
+#define ipipe_tsc2us(t)		((t) / (ipipe_cpu_freq() / 1000000UL))
+#else
+#define ipipe_read_tsc(t)					\
+	({							\
+		unsigned long __tbu;				\
+		__asm__ __volatile__ ("1: mftbu %0\n"		\
+				      "mftb %1\n"		\
+				      "mftbu %2\n"		\
+				      "cmpw %2,%0\n"		\
+				      "bne- 1b\n"		\
+				      :"=r" (((unsigned long *)&t)[0]),	\
+				       "=r" (((unsigned long *)&t)[1]),	\
+				       "=r" (__tbu));			\
+		t;							\
+	})
+#define ipipe_tsc2ns(t)		((((unsigned long)(t)) * 1000) / (ipipe_cpu_freq() / 1000000))
+#define ipipe_tsc2us(t)						\
+	({							\
+		unsigned long long delta = (t);			\
+		do_div(delta, ipipe_cpu_freq()/1000000+1);	\
+		(unsigned long)delta;				\
+	})
+#endif
+#define __ipipe_read_timebase()					\
+	({							\
+ 	unsigned long long t;					\
+ 	ipipe_read_tsc(t);					\
+ 	t;							\
+ 	})
+
+/* Private interface -- Internal use only */
+
+#define __ipipe_check_platform()		do { } while(0)
+#define __ipipe_enable_irq(irq)			enable_irq(irq)
+#define __ipipe_disable_irq(irq)		disable_irq(irq)
+#define __ipipe_disable_irqdesc(ipd, irq)	do { } while(0)
+
+void __ipipe_enable_irqdesc(struct ipipe_domain *ipd, unsigned irq);
+
+void __ipipe_init_platform(void);
+
+void __ipipe_enable_pipeline(void);
+
+void __ipipe_end_irq(unsigned irq);
+
+static inline int __ipipe_check_tickdev(const char *devname)
+{
+	return 1;
+}
+
+#ifdef CONFIG_SMP
+struct ipipe_ipi_struct {
+	volatile unsigned long value;
+} ____cacheline_aligned;
+
+void __ipipe_hook_critical_ipi(struct ipipe_domain *ipd);
+
+void __ipipe_register_ipi(unsigned int irq);
+#else
+#define __ipipe_hook_critical_ipi(ipd)	do { } while(0)
+#endif /* CONFIG_SMP */
+
+DECLARE_PER_CPU(struct pt_regs, __ipipe_tick_regs);
+
+void __ipipe_handle_irq(int irq, struct pt_regs *regs);
+
+static inline void ipipe_handle_chained_irq(unsigned int irq)
+{
+	struct pt_regs regs;	/* dummy */
+
+	ipipe_trace_irq_entry(irq);
+	__ipipe_handle_irq(irq, &regs);
+	ipipe_trace_irq_exit(irq);
+}
+
+struct irq_desc;
+void __ipipe_ack_level_irq(unsigned irq, struct irq_desc *desc);
+void __ipipe_end_level_irq(unsigned irq, struct irq_desc *desc);
+void __ipipe_ack_edge_irq(unsigned irq, struct irq_desc *desc);
+void __ipipe_end_edge_irq(unsigned irq, struct irq_desc *desc);
+
+void __ipipe_serial_debug(const char *fmt, ...);
+
+#define __ipipe_tick_irq	IPIPE_TIMER_VIRQ
+
+static inline unsigned long __ipipe_ffnz(unsigned long ul)
+{
+#ifdef CONFIG_PPC64
+	__asm__ __volatile__("cntlzd %0, %1":"=r"(ul):"r"(ul & (-ul)));
+	return 63 - ul;
+#else
+	__asm__ __volatile__("cntlzw %0, %1":"=r"(ul):"r"(ul & (-ul)));
+	return 31 - ul;
+#endif
+}
+
+/*
+ * When running handlers, enable hw interrupts for all domains but the
+ * one heading the pipeline, so that IRQs can never be significantly
+ * deferred for the latter.
+ */
+#define __ipipe_run_isr(ipd, irq)					\
+do {									\
+	if (!__ipipe_pipeline_head_p(ipd))				\
+		local_irq_enable_hw();					\
+	if (ipd == ipipe_root_domain)					\
+		if (likely(!ipipe_virtual_irq_p(irq)))			\
+			ipd->irqs[irq].handler(irq, NULL);		\
+		else {							\
+			irq_enter();					\
+			ipd->irqs[irq].handler(irq, ipd->irqs[irq].cookie);\
+			irq_exit();					\
+		}							\
+	else {								\
+		__clear_bit(IPIPE_SYNC_FLAG, &ipipe_cpudom_var(ipd, status)); \
+		ipd->irqs[irq].handler(irq, ipd->irqs[irq].cookie);	\
+		__set_bit(IPIPE_SYNC_FLAG, &ipipe_cpudom_var(ipd, status)); \
+	}								\
+	local_irq_disable_hw();						\
+} while(0)
+
+#define __ipipe_syscall_watched_p(p, sc)	\
+	(((p)->flags & PF_EVNOTIFY) || (unsigned long)sc >= NR_syscalls)
+
+#define __ipipe_root_tick_p(regs)	((regs)->msr & MSR_EE)
+
+#else /* !CONFIG_IPIPE */
+
+#define task_hijacked(p)	0
+
+#define ipipe_handle_chained_irq(irq)	generic_handle_irq(irq)
+
+#endif /* CONFIG_IPIPE */
+
+#define ipipe_update_tick_evtdev(evtdev)	do { } while (0)
+
+#endif /* !__ASM_POWERPC_IPIPE_H */
diff --git a/arch/powerpc/include/asm/ipipe_base.h b/arch/powerpc/include/asm/ipipe_base.h
new file mode 100644
index 0000000..9ccfc23
--- /dev/null
+++ b/arch/powerpc/include/asm/ipipe_base.h
@@ -0,0 +1,154 @@
+/* -*- linux-c -*-
+ * include/asm-powerpc/ipipe_base.h
+ *
+ * Copyright (C) 2007 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __ASM_POWERPC_IPIPE_BASE_H
+#define __ASM_POWERPC_IPIPE_BASE_H
+
+#ifdef CONFIG_IPIPE
+
+#define IPIPE_NR_XIRQS		NR_IRQS
+#ifdef CONFIG_PPC64
+#define IPIPE_IRQ_ISHIFT	6		/* 64-bit arch. */
+#else
+#define IPIPE_IRQ_ISHIFT	5		/* 32-bit arch. */
+#endif
+
+/* 
+ * The first virtual interrupt is reserved for the timer (see
+ * __ipipe_init_platform). 
+ */
+#define IPIPE_TIMER_VIRQ	IPIPE_VIRQ_BASE
+
+#ifdef CONFIG_SMP
+/* 
+ * These are virtual IPI numbers. The OpenPIC supports only 4 IPIs and
+ * all are already used by Linux. The virtualization layer is
+ * implemented by piggybacking the debugger break IPI 0x3,
+ * which is demultiplexed in __ipipe_ipi_demux().
+ */
+/* these are bit numbers in practice */
+#define IPIPE_MSG_CRITICAL_IPI	0
+#define IPIPE_MSG_SERVICE_IPI0	(IPIPE_MSG_CRITICAL_IPI + 1)
+#define IPIPE_MSG_SERVICE_IPI1	(IPIPE_MSG_CRITICAL_IPI + 2)
+#define IPIPE_MSG_SERVICE_IPI2	(IPIPE_MSG_CRITICAL_IPI + 3)
+#define IPIPE_MSG_SERVICE_IPI3	(IPIPE_MSG_CRITICAL_IPI + 4)
+#define IPIPE_MSG_SERVICE_IPI4	(IPIPE_MSG_CRITICAL_IPI + 5)
+
+#define IPIPE_MSG_IPI_MASK	((1UL << IPIPE_MSG_CRITICAL_IPI) |	\
+				 (1UL << IPIPE_MSG_SERVICE_IPI0) |	\
+				 (1UL << IPIPE_MSG_SERVICE_IPI1) |	\
+				 (1UL << IPIPE_MSG_SERVICE_IPI2) |	\
+				 (1UL << IPIPE_MSG_SERVICE_IPI3) |	\
+				 (1UL << IPIPE_MSG_SERVICE_IPI4))
+
+#define IPIPE_CRITICAL_IPI	(IPIPE_VIRQ_BASE + 1)
+#define IPIPE_SERVICE_IPI0	(IPIPE_CRITICAL_IPI + 1)
+#define IPIPE_SERVICE_IPI1	(IPIPE_CRITICAL_IPI + 2)
+#define IPIPE_SERVICE_IPI2	(IPIPE_CRITICAL_IPI + 3)
+#define IPIPE_SERVICE_IPI3	(IPIPE_CRITICAL_IPI + 4)
+#define IPIPE_SERVICE_IPI4	(IPIPE_CRITICAL_IPI + 5)
+
+#define IPIPE_MSG_IPI_OFFSET	(IPIPE_CRITICAL_IPI)
+
+#define ipipe_processor_id()	raw_smp_processor_id()
+#else  /* !CONFIG_SMP */
+#define ipipe_processor_id()	0
+#endif /* CONFIG_SMP */
+
+/* traps */
+#define IPIPE_TRAP_ACCESS	 0	/* Data or instruction access exception */
+#define IPIPE_TRAP_ALIGNMENT	 1	/* Alignment exception */
+#define IPIPE_TRAP_ALTUNAVAIL	 2	/* Altivec unavailable */
+#define IPIPE_TRAP_PCE		 3	/* Program check exception */
+#define IPIPE_TRAP_MCE		 4	/* Machine check exception */
+#define IPIPE_TRAP_UNKNOWN	 5	/* Unknown exception */
+#define IPIPE_TRAP_IABR		 6	/* Instruction breakpoint */
+#define IPIPE_TRAP_RM		 7	/* Run mode exception */
+#define IPIPE_TRAP_SSTEP	 8	/* Single-step exception */
+#define IPIPE_TRAP_NREC		 9	/* Non-recoverable exception */
+#define IPIPE_TRAP_SOFTEMU	10	/* Software emulation */
+#define IPIPE_TRAP_DEBUG	11	/* Debug exception */
+#define IPIPE_TRAP_SPE		12	/* SPE exception */
+#define IPIPE_TRAP_ALTASSIST	13	/* Altivec assist exception */
+#define IPIPE_TRAP_CACHE	14	/* Cache-locking exception (FSL) */
+#define IPIPE_TRAP_KFPUNAVAIL	15	/* FP unavailable exception */
+#define IPIPE_NR_FAULTS		16
+/* Pseudo-vectors used for kernel events */
+#define IPIPE_FIRST_EVENT		IPIPE_NR_FAULTS
+#define IPIPE_EVENT_SYSCALL		(IPIPE_FIRST_EVENT)
+#define IPIPE_EVENT_SCHEDULE		(IPIPE_FIRST_EVENT + 1)
+#define IPIPE_EVENT_SIGWAKE		(IPIPE_FIRST_EVENT + 2)
+#define IPIPE_EVENT_SETSCHED		(IPIPE_FIRST_EVENT + 3)
+#define IPIPE_EVENT_INIT		(IPIPE_FIRST_EVENT + 4)
+#define IPIPE_EVENT_EXIT		(IPIPE_FIRST_EVENT + 5)
+#define IPIPE_EVENT_CLEANUP		(IPIPE_FIRST_EVENT + 6)
+#define IPIPE_LAST_EVENT		IPIPE_EVENT_CLEANUP
+#define IPIPE_NR_EVENTS			(IPIPE_LAST_EVENT + 1)
+
+#ifndef __ASSEMBLY__
+
+#ifdef CONFIG_SMP
+
+void __ipipe_stall_root(void);
+
+unsigned long __ipipe_test_and_stall_root(void);
+
+unsigned long __ipipe_test_root(void);
+
+#else /* !CONFIG_SMP */
+
+#include <linux/bitops.h>
+
+#if __GNUC__ >= 4
+/* Alias to ipipe_root_cpudom_var(status) */
+extern unsigned long __ipipe_root_status;
+#else
+extern unsigned long *const __ipipe_root_status_addr;
+#define __ipipe_root_status	(*__ipipe_root_status_addr)
+#endif
+
+static __inline__ void __ipipe_stall_root(void)
+{
+	volatile unsigned long *p = &__ipipe_root_status;
+	set_bit(0, p);
+}
+
+static __inline__ unsigned long __ipipe_test_and_stall_root(void)
+{
+	volatile unsigned long *p = &__ipipe_root_status;
+	return test_and_set_bit(0, p);
+}
+
+static __inline__ unsigned long __ipipe_test_root(void)
+{
+	volatile unsigned long *p = &__ipipe_root_status;
+	return test_bit(0, p);
+}
+
+#endif /* !CONFIG_SMP */
+
+#endif /* !__ASSEMBLY__ */
+
+#define __IPIPE_FEATURE_UNMASKED_CONTEXT_SWITCH  1
+
+#endif /* CONFIG_IPIPE */
+
+#endif	/* !__ASM_POWERPC_IPIPE_BASE_H */
diff --git a/arch/powerpc/include/asm/irqflags.h b/arch/powerpc/include/asm/irqflags.h
index 5f68ecf..7707090 100644
--- a/arch/powerpc/include/asm/irqflags.h
+++ b/arch/powerpc/include/asm/irqflags.h
@@ -10,7 +10,17 @@
  */
 #include <asm/hw_irq.h>
 
-#else
+#elif CONFIG_IPIPE
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+#define TRACE_DISABLE_INTS		bl	__ipipe_trace_irqson
+#define TRACE_ENABLE_INTS		bl	__ipipe_trace_irqson
+#define TRACE_DISABLE_INTS_REALLY	bl	__ipipe_trace_irqsoff
+#else	
+#define TRACE_DISABLE_INTS
+#define TRACE_ENABLE_INTS
+#define TRACE_DISABLE_INTS_REALLY
+#endif
+#else /* !CONFIG_IPIPE */
 #ifdef CONFIG_TRACE_IRQFLAGS
 /*
  * Most of the CPU's IRQ-state tracing is done from assembly code; we
diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index ab4f192..1a2e8cd 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h
@@ -30,8 +30,14 @@ extern void set_context(unsigned long id, pgd_t *pgd);
 static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			     struct task_struct *tsk)
 {
+	int cpu = smp_processor_id();
+#ifndef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+#endif /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
 	/* Mark this context has been used on the new CPU */
-	cpu_set(smp_processor_id(), next->cpu_vm_mask);
+	cpu_set(cpu, next->cpu_vm_mask);
 
 	/* 32-bit keeps track of the current PGDIR in the thread struct */
 #ifdef CONFIG_PPC32
@@ -40,7 +46,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 
 	/* Nothing else to do if we aren't actually switching */
 	if (prev == next)
-		return;
+		goto done;
 
 	/* We must stop all altivec streams before changing the HW
 	 * context
@@ -53,6 +59,28 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	/* The actual HW switching method differs between the various
 	 * sub architectures.
 	 */
+#ifdef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+#ifdef CONFIG_PPC_STD_MMU_64
+	do {
+		per_cpu(ipipe_active_mm, cpu) = NULL; /* mm state is undefined. */
+		barrier();
+		if (cpu_has_feature(CPU_FTR_SLB))
+			switch_slb(tsk, next);
+		else
+			switch_stab(tsk, next);
+		barrier();
+		per_cpu(ipipe_active_mm, cpu) = next;
+	} while (test_and_clear_thread_flag(TIF_MMSWITCH_INT));
+#else
+	do {
+		per_cpu(ipipe_active_mm, cpu) = NULL; /* mm state is undefined. */
+		barrier();
+		switch_mmu_context(prev, next);
+		barrier();
+		per_cpu(ipipe_active_mm, cpu) = next;
+	} while (test_and_clear_thread_flag(TIF_MMSWITCH_INT));
+#endif
+#else /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
 #ifdef CONFIG_PPC_STD_MMU_64
 	if (cpu_has_feature(CPU_FTR_SLB))
 		switch_slb(tsk, next);
@@ -62,7 +90,12 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	/* Out of line for now */
 	switch_mmu_context(prev, next);
 #endif
-
+#endif /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
+done:
+#ifndef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+	local_irq_restore_hw(flags);
+#endif /* !CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH */
+	return;
 }
 
 #define deactivate_mm(tsk,mm)	do { } while (0)
diff --git a/arch/powerpc/include/asm/mpic.h b/arch/powerpc/include/asm/mpic.h
index 8fc3776..0e5e861 100644
--- a/arch/powerpc/include/asm/mpic.h
+++ b/arch/powerpc/include/asm/mpic.h
@@ -281,7 +281,7 @@ struct mpic
 #ifdef CONFIG_MPIC_U3_HT_IRQS
 	/* The fixup table */
 	struct mpic_irq_fixup	*fixups;
-	spinlock_t		fixup_lock;
+	ipipe_spinlock_t	fixup_lock;
 #endif
 
 	/* Register access method */
diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h
index 082b3ae..dfe2114 100644
--- a/arch/powerpc/include/asm/paca.h
+++ b/arch/powerpc/include/asm/paca.h
@@ -96,8 +96,12 @@ struct paca_struct {
 	u64 saved_r1;			/* r1 save for RTAS calls */
 	u64 saved_msr;			/* MSR saved here by enter_rtas */
 	u16 trap_save;			/* Used when bad stack is encountered */
+#ifdef CONFIG_SOFTDISABLE
 	u8 soft_enabled;		/* irq soft-enable flag */
 	u8 hard_enabled;		/* set if irqs are enabled in MSR */
+#elif CONFIG_IPIPE
+	u64 root_percpu;		/* Address of per_cpu data for the root domain */
+#endif
 	u8 io_sync;			/* writel() needs spin_unlock sync */
 
 	/* Stuff for accurate time accounting */
diff --git a/arch/powerpc/include/asm/ptrace.h b/arch/powerpc/include/asm/ptrace.h
index c9c678f..670f23a 100644
--- a/arch/powerpc/include/asm/ptrace.h
+++ b/arch/powerpc/include/asm/ptrace.h
@@ -36,7 +36,7 @@ struct pt_regs {
 	unsigned long xer;
 	unsigned long ccr;
 #ifdef __powerpc64__
-	unsigned long softe;		/* Soft enabled/disabled */
+	unsigned long softe;		/* Soft enabled/disabled (CONFIG_SOFTDISABLE || CONFIG_IPIPE) */
 #else
 	unsigned long mq;		/* 601 only (not used at present) */
 					/* Used on APUS to hold IPL value. */
diff --git a/arch/powerpc/include/asm/qe_ic.h b/arch/powerpc/include/asm/qe_ic.h
index cf51966..e7d287c 100644
--- a/arch/powerpc/include/asm/qe_ic.h
+++ b/arch/powerpc/include/asm/qe_ic.h
@@ -74,6 +74,13 @@ static inline unsigned int qe_ic_get_high_irq(struct qe_ic *qe_ic)
 { return 0; }
 #endif /* CONFIG_QUICC_ENGINE */
 
+#ifdef CONFIG_IPIPE
+void __ipipe_qe_ic_cascade_irq(struct qe_ic *qe_ic, unsigned int virq);
+#define qe_ic_cascade_irq(qe_ic, irq)	__ipipe_qe_ic_cascade_irq(qe_ic, irq)
+#else
+#define qe_ic_cascade_irq(qe_ic, irq)	generic_handle_irq(irq)
+#endif
+
 void qe_ic_set_highest_priority(unsigned int virq, int high);
 int qe_ic_set_priority(unsigned int virq, unsigned int priority);
 int qe_ic_set_high_priority(unsigned int virq, unsigned int priority, int high);
@@ -85,7 +92,7 @@ static inline void qe_ic_cascade_low_ipic(unsigned int irq,
 	unsigned int cascade_irq = qe_ic_get_low_irq(qe_ic);
 
 	if (cascade_irq != NO_IRQ)
-		generic_handle_irq(cascade_irq);
+		qe_ic_cascade_irq(qe_ic, cascade_irq);
 }
 
 static inline void qe_ic_cascade_high_ipic(unsigned int irq,
@@ -95,7 +102,7 @@ static inline void qe_ic_cascade_high_ipic(unsigned int irq,
 	unsigned int cascade_irq = qe_ic_get_high_irq(qe_ic);
 
 	if (cascade_irq != NO_IRQ)
-		generic_handle_irq(cascade_irq);
+		qe_ic_cascade_irq(qe_ic, cascade_irq);
 }
 
 static inline void qe_ic_cascade_low_mpic(unsigned int irq,
@@ -105,7 +112,7 @@ static inline void qe_ic_cascade_low_mpic(unsigned int irq,
 	unsigned int cascade_irq = qe_ic_get_low_irq(qe_ic);
 
 	if (cascade_irq != NO_IRQ)
-		generic_handle_irq(cascade_irq);
+		qe_ic_cascade_irq(qe_ic, cascade_irq);
 
 	desc->chip->eoi(irq);
 }
@@ -117,7 +124,7 @@ static inline void qe_ic_cascade_high_mpic(unsigned int irq,
 	unsigned int cascade_irq = qe_ic_get_high_irq(qe_ic);
 
 	if (cascade_irq != NO_IRQ)
-		generic_handle_irq(cascade_irq);
+		qe_ic_cascade_irq(qe_ic, cascade_irq);
 
 	desc->chip->eoi(irq);
 }
@@ -133,7 +140,7 @@ static inline void qe_ic_cascade_muxed_mpic(unsigned int irq,
 		cascade_irq = qe_ic_get_low_irq(qe_ic);
 
 	if (cascade_irq != NO_IRQ)
-		generic_handle_irq(cascade_irq);
+		qe_ic_cascade_irq(qe_ic, cascade_irq);
 
 	desc->chip->eoi(irq);
 }
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h
index 08d3a73..407d60f 100644
--- a/arch/powerpc/include/asm/reg.h
+++ b/arch/powerpc/include/asm/reg.h
@@ -769,7 +769,7 @@
 
 #define proc_trap()	asm volatile("trap")
 
-#ifdef CONFIG_PPC64
+#ifdef CONFIG_RUNLATCH
 
 extern void ppc64_runlatch_on(void);
 extern void ppc64_runlatch_off(void);
diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index c25f73d..2ba906e 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -54,8 +54,14 @@ void generic_mach_cpu_die(void);
 /* 32-bit */
 extern int smp_hw_index[];
 
+#ifdef CONFIG_IPIPE
+extern int smp_logical_index[];
+#define raw_smp_processor_id()		(smp_logical_index[mfspr(SPRN_PIR)])
+#define hard_smp_processor_id() 	(smp_hw_index[raw_smp_processor_id()])
+#else
 #define raw_smp_processor_id()	(current_thread_info()->cpu)
 #define hard_smp_processor_id() 	(smp_hw_index[smp_processor_id()])
+#endif
 
 static inline int get_hard_smp_processor_id(int cpu)
 {
@@ -65,6 +71,10 @@ static inline int get_hard_smp_processor_id(int cpu)
 static inline void set_hard_smp_processor_id(int cpu, int phys)
 {
 	smp_hw_index[cpu] = phys;
+#ifdef CONFIG_IPIPE
+	BUG_ON(phys >= NR_CPUS);
+	smp_logical_index[phys] = cpu;
+#endif
 }
 #endif
 
@@ -80,6 +90,7 @@ extern int cpu_to_core_id(int cpu);
 #define PPC_MSG_RESCHEDULE      1
 #define PPC_MSG_CALL_FUNC_SINGLE	2
 #define PPC_MSG_DEBUGGER_BREAK  3
+#define PPC_MSG_IPIPE_DEMUX     PPC_MSG_DEBUGGER_BREAK
 
 /*
  * irq controllers that have dedicated ipis per message and don't
diff --git a/arch/powerpc/include/asm/thread_info.h b/arch/powerpc/include/asm/thread_info.h
index e04286f..970863f 100644
--- a/arch/powerpc/include/asm/thread_info.h
+++ b/arch/powerpc/include/asm/thread_info.h
@@ -114,6 +114,7 @@ static inline struct thread_info *current_thread_info(void)
 #define TIF_FREEZE		14	/* Freezing for suspend */
 #define TIF_RUNLATCH		15	/* Is the runlatch enabled? */
 #define TIF_ABI_PENDING		16	/* 32/64 bit switch needed */
+#define TIF_MMSWITCH_INT	20	/* MMU context switch interrupted */
 
 /* as above, but as bit values */
 #define _TIF_SYSCALL_TRACE	(1<<TIF_SYSCALL_TRACE)
@@ -132,6 +133,7 @@ static inline struct thread_info *current_thread_info(void)
 #define _TIF_FREEZE		(1<<TIF_FREEZE)
 #define _TIF_RUNLATCH		(1<<TIF_RUNLATCH)
 #define _TIF_ABI_PENDING	(1<<TIF_ABI_PENDING)
+#define _TIF_MMSWITCH_INT	(1<<TIF_MMSWITCH_INT)
 #define _TIF_SYSCALL_T_OR_A	(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP)
 
 #define _TIF_USER_WORK_MASK	(_TIF_SIGPENDING | _TIF_NEED_RESCHED | \
diff --git a/arch/powerpc/include/asm/time.h b/arch/powerpc/include/asm/time.h
index d1bf957..f295275 100644
--- a/arch/powerpc/include/asm/time.h
+++ b/arch/powerpc/include/asm/time.h
@@ -30,6 +30,7 @@ extern unsigned long tb_ticks_per_usec;
 extern unsigned long tb_ticks_per_sec;
 extern u64 tb_to_xs;
 extern unsigned      tb_to_us;
+DECLARE_PER_CPU(int, disarm_decr);
 
 struct rtc_time;
 extern void to_tm(int tim, struct rtc_time * tm);
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 8d1a419..6775cf6 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -39,6 +39,7 @@ obj-$(CONFIG_PPC64)		+= setup_64.o sys_ppc32.o \
 obj64-$(CONFIG_RELOCATABLE)	+= reloc_64.o
 obj-$(CONFIG_PPC64)		+= vdso64/
 obj-$(CONFIG_ALTIVEC)		+= vecemu.o vector.o
+obj-$(CONFIG_IPIPE)		+= ipipe.o
 obj-$(CONFIG_PPC_970_NAP)	+= idle_power4.o
 obj-$(CONFIG_PPC_OF)		+= of_device.o of_platform.o prom_parse.o
 obj-$(CONFIG_PPC_CLOCK)		+= clock.o
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 19ee491..9d2175f 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -129,8 +129,12 @@ int main(void)
 	DEFINE(PACATOC, offsetof(struct paca_struct, kernel_toc));
 	DEFINE(PACAKBASE, offsetof(struct paca_struct, kernelbase));
 	DEFINE(PACAKMSR, offsetof(struct paca_struct, kernel_msr));
+#ifdef CONFIG_SOFTDISABLE
 	DEFINE(PACASOFTIRQEN, offsetof(struct paca_struct, soft_enabled));
 	DEFINE(PACAHARDIRQEN, offsetof(struct paca_struct, hard_enabled));
+#elif CONFIG_IPIPE
+	DEFINE(PACAROOTPCPU, offsetof(struct paca_struct, root_percpu));
+#endif
 	DEFINE(PACASLBCACHE, offsetof(struct paca_struct, slb_cache));
 	DEFINE(PACASLBCACHEPTR, offsetof(struct paca_struct, slb_cache_ptr));
 	DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
@@ -250,7 +254,9 @@ int main(void)
 	DEFINE(_DEAR, STACK_FRAME_OVERHEAD+offsetof(struct pt_regs, dar));
 	DEFINE(_ESR, STACK_FRAME_OVERHEAD+offsetof(struct pt_regs, dsisr));
 #else /* CONFIG_PPC64 */
+#if defined(CONFIG_SOFTDISABLE) || defined(CONFIG_IPIPE)
 	DEFINE(SOFTE, STACK_FRAME_OVERHEAD+offsetof(struct pt_regs, softe));
+#endif
 
 	/* These _only_ to be used with {PROM,RTAS}_FRAME_SIZE!!! */
 	DEFINE(_SRR0, STACK_FRAME_OVERHEAD+sizeof(struct pt_regs));
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 923f87a..dded6e4 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -1762,7 +1762,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
 
 static struct cpu_spec the_cpu_spec;
 
-struct cpu_spec * __init identify_cpu(unsigned long offset, unsigned int pvr)
+notrace struct cpu_spec * __init identify_cpu(unsigned long offset, unsigned int pvr)
 {
 	struct cpu_spec *s = cpu_specs;
 	struct cpu_spec *t = &the_cpu_spec;
diff --git a/arch/powerpc/kernel/entry_32.S b/arch/powerpc/kernel/entry_32.S
index 6f7eb7e..3f0b150 100644
--- a/arch/powerpc/kernel/entry_32.S
+++ b/arch/powerpc/kernel/entry_32.S
@@ -174,8 +174,25 @@ transfer_to_handler:
          * check for stack overflow
          */
 	lwz	r9,KSP_LIMIT(r12)
+#ifdef CONFIG_IPIPE
+	/*
+	 * Allow for private kernel-based stacks: those must not cause
+	 * the stack overflow detection to trigger when some activity has
+	 * been preempted over them. We just check if the kernel stack is
+	 * not treading on the memory area ranging from
+	 * &current->thread_info to &current->thread, which is coarser
+	 * than the vanilla implementation, but likely sensitive enough
+	 * to catch overflows soon enough though.
+	 */
+	addi	 r12,r9,THREAD
+	cmplw 0,r1,r9
+	cmplw 1,r1,r12
+	crand 1,1,4
+	bgt-  stack_ovf		/* if r9 < r1 < r9+THREAD */
+#else /* CONFIG_IPIPE */
 	cmplw	r1,r9			/* if r1 <= ksp_limit */
 	ble-	stack_ovf		/* then the kernel stack overflowed */
+#endif /* CONFIG_IPIPE */
 5:
 #if defined(CONFIG_6xx) || defined(CONFIG_E500)
 	rlwinm	r9,r1,0,0,31-THREAD_SHIFT
@@ -248,6 +265,21 @@ _GLOBAL(DoSyscall)
 	lwz	r11,_CCR(r1)	/* Clear SO bit in CR */
 	rlwinm	r11,r11,0,4,2
 	stw	r11,_CCR(r1)
+#ifdef CONFIG_IPIPE
+	addi	r3,r1,GPR0
+	bl	__ipipe_syscall_root
+	cmpwi	r3,0
+	lwz	r3,GPR3(r1)
+	lwz	r0,GPR0(r1)
+	lwz	r4,GPR4(r1)
+	lwz	r5,GPR5(r1)
+	lwz	r6,GPR6(r1)
+	lwz	r7,GPR7(r1)
+	lwz	r8,GPR8(r1)
+	lwz	r9,GPR9(r1)
+	bgt	.ipipe_end_syscall
+	blt	ret_from_syscall
+#endif /* CONFIG_IPIPE */
 #ifdef SHOW_SYSCALLS
 	bl	do_show_syscall
 #endif /* SHOW_SYSCALLS */
@@ -326,11 +358,34 @@ END_FTR_SECTION_IFSET(CPU_FTR_NEED_PAIRED_STWCX)
 	b	1b
 #endif  /* CONFIG_44x */
 
+#ifdef CONFIG_IPIPE
+.ipipe_end_syscall:
+	LOAD_MSR_KERNEL(r10,MSR_KERNEL)	/* doesn't include MSR_EE */
+	SYNC
+	MTMSRD(r10)
+	b syscall_exit_cont
+#endif /* CONFIG_IPIPE */
+
 66:	li	r3,-ENOSYS
 	b	ret_from_syscall
 
 	.globl	ret_from_fork
 ret_from_fork:
+#ifdef CONFIG_IPIPE
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	stwu	r1,-4(r1)
+	stw	r3,0(r1)
+	lis	r3,(0x80000000)@h
+	ori	r3,r3,(0x80000000)@l
+	bl	ipipe_trace_end
+	lwz	r3,0(r1)
+	addi	r1,r1,4
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
+	LOAD_MSR_KERNEL(r10,MSR_KERNEL)
+	ori	r10,r10,MSR_EE
+	SYNC
+	MTMSRD(r10)
+#endif /* CONFIG_IPIPE */		
 	REST_NVGPRS(r1)
 	bl	schedule_tail
 	li	r3,0
@@ -709,6 +764,12 @@ ret_from_except:
 	SYNC			/* Some chip revs have problems here... */
 	MTMSRD(r10)		/* disable interrupts */
 
+#ifdef CONFIG_IPIPE
+        bl __ipipe_check_root
+        cmpwi   r3, 0
+	mfmsr	r10		/* this is used later, might be messed */
+        beq- restore
+#endif /* CONFIG_IPIPE */
 	lwz	r3,_MSR(r1)	/* Returning to user mode? */
 	andi.	r0,r3,MSR_PR
 	beq	resume_kernel
@@ -732,6 +793,12 @@ restore_user:
 #ifdef CONFIG_PREEMPT
 	b	restore
 
+#ifdef CONFIG_IPIPE
+#define PREEMPT_SCHEDULE_IRQ	__ipipe_preempt_schedule_irq
+#else
+#define PREEMPT_SCHEDULE_IRQ	preempt_schedule_irq
+#endif
+
 /* N.B. the only way to get here is from the beq following ret_from_except. */
 resume_kernel:
 	/* check current_thread_info->preempt_count */
@@ -744,7 +811,7 @@ resume_kernel:
 	beq+	restore
 	andi.	r0,r3,MSR_EE	/* interrupts off? */
 	beq	restore		/* don't schedule if so */
-1:	bl	preempt_schedule_irq
+1:	bl	PREEMPT_SCHEDULE_IRQ
 	rlwinm	r9,r1,0,0,(31-THREAD_SHIFT)
 	lwz	r3,TI_FLAGS(r9)
 	andi.	r0,r3,_TIF_NEED_RESCHED
@@ -1108,6 +1175,13 @@ ee_restarts:
 	.space	4
 	.previous
 
+#ifdef CONFIG_IPIPE
+_GLOBAL(__ipipe_ret_from_except)
+        cmpwi   r3, 0
+        bne+ ret_from_except
+        b restore
+#endif /* CONFIG_IPIPE */
+
 /*
  * PROM code for specific machines follows.  Put it
  * here so it's easy to add arch-specific sections later.
@@ -1179,6 +1253,14 @@ _GLOBAL(ftrace_caller)
 	/* Based off of objdump optput from glibc */
 	stwu	r1,-48(r1)
 	stw	r3, 12(r1)
+        LOAD_REG_IMMEDIATE(r3, function_trace_stop)
+        lwz     r3, 0(r3)
+        cmpwi   r3, 0
+	lwz	r3, 12(r1)
+        beq     1f
+	addi	r1, r1, 48
+	blr
+1:
 	stw	r4, 16(r1)
 	stw	r5, 20(r1)
 	stw	r6, 24(r1)
@@ -1217,6 +1299,14 @@ _GLOBAL(mcount)
 _GLOBAL(_mcount)
 	stwu	r1,-48(r1)
 	stw	r3, 12(r1)
+        LOAD_REG_IMMEDIATE(r3, function_trace_stop)
+        lwz     r3, 0(r3)
+        cmpwi   r3, 0
+	lwz	r3, 12(r1)
+        beq     1f
+	addi	r1, r1, 48
+	blr
+1:
 	stw	r4, 16(r1)
 	stw	r5, 20(r1)
 	stw	r6, 24(r1)
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index ae7e445..8bb13f9 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -105,6 +105,7 @@ system_call_common:
 	addi	r9,r1,STACK_FRAME_OVERHEAD
 	ld	r12,_MSR(r1)
 #endif /* CONFIG_TRACE_IRQFLAGS */
+#ifdef CONFIG_SOFTDISABLE	
 	li	r10,1
 	stb	r10,PACASOFTIRQEN(r13)
 	stb	r10,PACAHARDIRQEN(r13)
@@ -120,9 +121,28 @@ BEGIN_FW_FTR_SECTION
 2:
 END_FW_FTR_SECTION_IFSET(FW_FEATURE_ISERIES)
 #endif /* CONFIG_PPC_ISERIES */
+#endif /* CONFIG_SOFTDISABLE */
+
+#ifdef CONFIG_IPIPE
+	addi	r3,r1,GPR0
+	bl	.__ipipe_syscall_root
+	cmpwi	r3,0
+	ld	r0,GPR0(r1)
+	ld	r3,GPR3(r1)
+	ld	r4,GPR4(r1)
+	ld	r5,GPR5(r1)
+	ld	r6,GPR6(r1)
+	ld	r7,GPR7(r1)
+	ld	r8,GPR8(r1)
+	ld	r9,GPR9(r1)
+	bgt	ipipe_end_syscall
+	blt	syscall_exit
+	addi	r9,r1,STACK_FRAME_OVERHEAD
+#else	/* !CONFIG_IPIPE */
 	mfmsr	r11
 	ori	r11,r11,MSR_EE
 	mtmsrd	r11,1
+#endif /* CONFIG_IPIPE */
 
 #ifdef SHOW_SYSCALLS
 	bl	.do_show_syscall
@@ -174,7 +194,7 @@ syscall_exit:
 	andi.	r10,r8,MSR_RI
 	beq-	unrecov_restore
 	mfmsr	r10
-	rldicl	r10,r10,48,1
+	rldicl	r10,r10,48,1		/* clear MSR_EE */
 	rotldi	r10,r10,16
 	mtmsrd	r10,1
 	ld	r9,TI_FLAGS(r12)
@@ -237,6 +257,35 @@ syscall_dotrace:
 	ld	r10,TI_FLAGS(r10)
 	b	syscall_dotrace_cont
 
+#ifdef CONFIG_IPIPE
+	.globl  ipipe_end_syscall
+ipipe_end_syscall:
+	mfmsr	r10
+	rldicl	r10,r10,48,1		/* clear MSR_EE - we don't trace this */
+	rotldi	r10,r10,16		/* short IRQs off section, since our caller */
+	mtmsrd	r10,1			/* did branch here with IRQs on. */
+	ld	r5,_CCR(r1)
+	ld	r8,_MSR(r1)
+	ld	r7,_NIP(r1)
+	stdcx.	r0,0,r1			/* to clear pending reservations */
+	andi.	r6,r8,MSR_PR
+	ld	r4,_LINK(r1)
+	beq-	1f			/* only restore r13 if */
+	ld	r13,GPR13(r1)		/* returning to usermode */
+1:	ld	r2,GPR2(r1)
+	li	r12,MSR_RI
+	mfmsr	r10
+	andc	r10,r10,r12
+	mtmsrd	r10,1			/* clear MSR.RI */
+	ld	r1,GPR1(r1)
+	mtlr	r4
+	mtcr	r5
+	mtspr	SPRN_SRR0,r7
+	mtspr	SPRN_SRR1,r8
+	rfid
+	b	.			/* prevent speculative execution */
+#endif /* CONFIG_IPIPE */
+
 syscall_enosys:
 	li	r3,-ENOSYS
 	b	syscall_exit
@@ -275,12 +324,21 @@ syscall_exit_work:
 	andi.	r0,r9,(_TIF_SYSCALL_T_OR_A|_TIF_SINGLESTEP)
 	beq	.ret_from_except_lite
 
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	bl	.save_nvgprs
+	bl	__ipipe_trace_irqson
+	/* Re-enable interrupts */
+	mfmsr	r10
+	ori	r10,r10,MSR_EE
+	mtmsrd	r10,1
+#else
 	/* Re-enable interrupts */
 	mfmsr	r10
 	ori	r10,r10,MSR_EE
 	mtmsrd	r10,1
-
 	bl	.save_nvgprs
+#endif		
+
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.do_syscall_trace_leave
 	b	.ret_from_except
@@ -331,6 +389,18 @@ _GLOBAL(ppc64_swapcontext)
 	b	syscall_exit
 
 _GLOBAL(ret_from_fork)
+#ifdef CONFIG_IPIPE
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	stdu	r1,-8(r1)
+	std	r3,0(r1)
+	bl	__ipipe_trace_irqson
+	ld	r3,0(r1)
+	addi	r1,r1,8
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
+	mfmsr	r10
+	ori	r10,r10,MSR_EE
+	mtmsrd	r10,1
+#endif /* CONFIG_IPIPE */		
 	bl	.schedule_tail
 	REST_NVGPRS(r1)
 	li	r3,0
@@ -482,6 +552,14 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 	blr
 
 	.align	7
+
+#ifdef CONFIG_IPIPE
+_GLOBAL(__ipipe_ret_from_except_lite)
+	cmpwi	r3,0
+	bne+	.ret_from_except_lite /* FIXME:	branching to __ipipe_check_root is useless here */
+	b	restore
+#endif /* CONFIG_IPIPE */
+
 _GLOBAL(ret_from_except)
 	ld	r11,_TRAP(r1)
 	andi.	r0,r11,1
@@ -494,11 +572,21 @@ _GLOBAL(ret_from_except_lite)
 	 * can't change between when we test it and when we return
 	 * from the interrupt.
 	 */
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	bl	__ipipe_trace_irqsoff
+#endif	
 	mfmsr	r10		/* Get current interrupt state */
 	rldicl	r9,r10,48,1	/* clear MSR_EE */
 	rotldi	r9,r9,16
 	mtmsrd	r9,1		/* Update machine state */
 
+#ifdef CONFIG_IPIPE
+	bl	.__ipipe_check_root
+	cmpwi	r3,0
+	mfmsr	r10		/* this is used later, might be messed */
+	beq-	restore
+#endif /* CONFIG_IPIPE */
+
 #ifdef CONFIG_PREEMPT
 	clrrdi	r9,r1,THREAD_SHIFT	/* current_thread_info() */
 	li	r0,_TIF_NEED_RESCHED	/* bits to check */
@@ -522,6 +610,7 @@ _GLOBAL(ret_from_except_lite)
 #endif
 
 restore:
+#ifdef CONFIG_SOFTDISABLE
 BEGIN_FW_FTR_SECTION
 	ld	r5,SOFTE(r1)
 FW_FTR_SECTION_ELSE
@@ -534,6 +623,22 @@ ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
 	ld	r3,_MSR(r1)
 	rldicl	r4,r3,49,63		/* r0 = (r3 >> 15) & 1 */
 	stb	r4,PACAHARDIRQEN(r13)
+#else	/* !CONFIG_SOFTDISABLE */
+#ifdef CONFIG_IPIPE
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	ld	r3,_MSR(r1)
+	rldicl	r3,r3,49,63		/* r0 = (r3 >> 15) & 1 */
+	bl	__ipipe_trace_irqsx
+#endif
+	ld	r3,SOFTE(r1)		/* currently hard-disabled, so this is safe */
+	nor	r3,r3,r3
+	ld	r4,PACAROOTPCPU(r13)
+	ld	r5,0(r4)
+	insrdi	r5,r3,1,63
+	std	r5,0(r4)
+	ld	r3,_MSR(r1)
+#endif	
+#endif	/* !CONFIG_SOFTDISABLE */
 
 	ld	r4,_CTR(r1)
 	ld	r0,_LINK(r1)
@@ -630,17 +735,33 @@ do_work:
 	 */
 	mfmsr	r10
 #endif /* CONFIG_TRACE_IRQFLAGS */
+#ifdef CONFIG_SOFTDISABLE
 	li	r0,1
 	stb	r0,PACASOFTIRQEN(r13)
 	stb	r0,PACAHARDIRQEN(r13)
+#else
+#ifdef CONFIG_IPIPE
+        ld      r10,PACAROOTPCPU(r13)
+        ld      r0,0(r10)
+        clrrdi  r0,r0,1
+        std     r0,0(r10)
+#endif /* CONFIG_IPIPE */
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	bl	__ipipe_trace_irqson
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
+	mfmsr	r10
+#endif /* !CONFIG_SOFTDISABLE */
 	ori	r10,r10,MSR_EE
 	mtmsrd	r10,1		/* reenable interrupts */
 	bl	.preempt_schedule
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	bl	__ipipe_trace_irqsoff
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
 	mfmsr	r10
-	clrrdi	r9,r1,THREAD_SHIFT
 	rldicl	r10,r10,48,1	/* disable interrupts again */
 	rotldi	r10,r10,16
 	mtmsrd	r10,1
+	clrrdi	r9,r1,THREAD_SHIFT
 	ld	r4,TI_FLAGS(r9)
 	andi.	r0,r4,_TIF_NEED_RESCHED
 	bne	1b
@@ -648,6 +769,12 @@ do_work:
 
 user_work:
 #endif
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+	bl	__ipipe_trace_irqson
+	clrrdi	r9,r1,THREAD_SHIFT
+	ld	r4,TI_FLAGS(r9)
+	mfmsr	r10
+#endif /* CONFIG_IPIPE_TRACE_IRQSOFF */
 	/* Enable interrupts */
 	ori	r10,r10,MSR_EE
 	mtmsrd	r10,1
@@ -707,7 +834,7 @@ _GLOBAL(enter_rtas)
 	li	r0,0
 	mtcr	r0
 
-#ifdef CONFIG_BUG	
+#ifdef CONFIG_BUG && CONFIG_SOFTDISABLE
 	/* There is no way it is acceptable to get here with interrupts enabled,
 	 * check it with the asm equivalent of WARN_ON
 	 */
@@ -901,6 +1028,10 @@ _GLOBAL(_mcount)
 	blr
 
 _GLOBAL(ftrace_caller)
+        LOAD_REG_IMMEDIATE(r3, function_trace_stop)
+        lwz     r3, 0(r3)
+        cmpwi   r3, 0
+        bne     ftrace_stub
 	/* Taken from output of objdump from lib64/glibc */
 	mflr	r3
 	ld	r11, 0(r1)
@@ -922,6 +1053,10 @@ _GLOBAL(mcount)
 	blr
 
 _GLOBAL(_mcount)
+        LOAD_REG_IMMEDIATE(r3, function_trace_stop)
+        lwz     r3, 0(r3)
+        cmpwi   r3, 0
+        bne     ftrace_stub
 	/* Taken from output of objdump from lib64/glibc */
 	mflr	r3
 	ld	r11, 0(r1)
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index d794a63..fe8c968 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -318,6 +318,12 @@ i##n:								\
 	EXC_XFER_TEMPLATE(n, hdlr, n, NOCOPY, transfer_to_handler_full,	\
 			  ret_from_except_full)
 
+#ifdef CONFIG_IPIPE
+#define EXC_XFER_IPIPE(n, hdlr)		\
+	EXC_XFER_TEMPLATE(n, hdlr, n+1, NOCOPY, transfer_to_handler, \
+			  __ipipe_ret_from_except)
+#endif /* CONFIG_IPIPE */
+
 #define EXC_XFER_LITE(n, hdlr)		\
 	EXC_XFER_TEMPLATE(n, hdlr, n+1, NOCOPY, transfer_to_handler, \
 			  ret_from_except)
@@ -402,7 +408,11 @@ InstructionAccess:
 	EXC_XFER_EE_LITE(0x400, handle_page_fault)
 
 /* External interrupt */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x500, HardwareInterrupt, __ipipe_grab_irq, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x500, HardwareInterrupt, do_IRQ, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 /* Alignment exception */
 	. = 0x600
@@ -436,7 +446,11 @@ END_FTR_SECTION_IFSET(CPU_FTR_FPU_UNAVAI
 	EXC_XFER_EE_LITE(0x800, kernel_fp_unavailable_exception)
 
 /* Decrementer */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x900, Decrementer, __ipipe_grab_timer, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x900, Decrementer, timer_interrupt, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 	EXCEPTION(0xa00, Trap_0a, unknown_exception, EXC_XFER_EE)
 	EXCEPTION(0xb00, Trap_0b, unknown_exception, EXC_XFER_EE)
@@ -1327,8 +1341,8 @@ swapper_pg_dir:
 
 	.globl intercept_table
 intercept_table:
-	.long 0, 0, i0x200, i0x300, i0x400, 0, i0x600, i0x700
-	.long i0x800, 0, 0, 0, 0, i0xd00, 0, 0
+	.long 0, 0, i0x200, i0x300, i0x400, i0x500, i0x600, i0x700
+	.long i0x800, i0x900, 0, 0, 0, i0xd00, 0, 0
 	.long 0, 0, 0, i0x1300, 0, 0, 0, 0
 	.long 0, 0, 0, 0, 0, 0, 0, 0
 	.long 0, 0, 0, 0, 0, 0, 0, 0
diff --git a/arch/powerpc/kernel/head_40x.S b/arch/powerpc/kernel/head_40x.S
index 56d8e5d..aee104e 100644
--- a/arch/powerpc/kernel/head_40x.S
+++ b/arch/powerpc/kernel/head_40x.S
@@ -233,6 +233,12 @@ label:
 	EXC_XFER_TEMPLATE(hdlr, n, MSR_KERNEL, NOCOPY, transfer_to_handler_full, \
 			  ret_from_except_full)
 
+#ifdef CONFIG_IPIPE
+#define EXC_XFER_IPIPE(n, hdlr)		\
+	EXC_XFER_TEMPLATE(hdlr, n+1, MSR_KERNEL, NOCOPY, transfer_to_handler, \
+			  __ipipe_ret_from_except)
+#endif /* CONFIG_IPIPE */
+
 #define EXC_XFER_LITE(n, hdlr)		\
 	EXC_XFER_TEMPLATE(hdlr, n+1, MSR_KERNEL, NOCOPY, transfer_to_handler, \
 			  ret_from_except)
@@ -401,7 +407,11 @@ label:
 	EXC_XFER_EE_LITE(0x400, handle_page_fault)
 
 /* 0x0500 - External Interrupt Exception */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x0500, HardwareInterrupt, __ipipe_grab_irq, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x0500, HardwareInterrupt, do_IRQ, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 /* 0x0600 - Alignment Exception */
 	START_EXCEPTION(0x0600, Alignment)
@@ -439,7 +449,11 @@ label:
 	lis	r0,TSR_PIS@h
 	mtspr	SPRN_TSR,r0		/* Clear the PIT exception */
 	addi	r3,r1,STACK_FRAME_OVERHEAD
+#ifdef CONFIG_IPIPE
+	EXC_XFER_IPIPE(0x1000, __ipipe_grab_timer)
+#else /* !CONFIG_IPIPE */
 	EXC_XFER_LITE(0x1000, timer_interrupt)
+#endif /* CONFIG_IPIPE */
 
 #if 0
 /* NOTE:
diff --git a/arch/powerpc/kernel/head_44x.S b/arch/powerpc/kernel/head_44x.S
index b56fecc..9eaca5b 100644
--- a/arch/powerpc/kernel/head_44x.S
+++ b/arch/powerpc/kernel/head_44x.S
@@ -309,8 +309,11 @@ interrupt_base:
 		/* Instruction Storage Interrupt */
 	INSTRUCTION_STORAGE_EXCEPTION
 
-	/* External Input Interrupt */
-	EXCEPTION(0x0500, ExternalInput, do_IRQ, EXC_XFER_LITE)
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x0500, ExternalInput, __ipipe_grab_irq, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
+ 	EXCEPTION(0x0500, ExternalInput, do_IRQ, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 	/* Alignment Interrupt */
 	ALIGNMENT_EXCEPTION
diff --git a/arch/powerpc/kernel/head_64.S b/arch/powerpc/kernel/head_64.S
index 82a9c37..4753eda 100644
--- a/arch/powerpc/kernel/head_64.S
+++ b/arch/powerpc/kernel/head_64.S
@@ -368,6 +368,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
 	STD_EXCEPTION_PSERIES(., altivec_unavailable)
 	STD_EXCEPTION_PSERIES(., vsx_unavailable)
 
+#ifdef CONFIG_SOFTDISABLE
 /*
  * An interrupt came in while soft-disabled; clear EE in SRR1,
  * clear paca->hard_enabled and return.
@@ -384,6 +385,7 @@ masked_interrupt:
 	mfspr	r13,SPRN_SPRG1
 	rfid
 	b	.
+#endif /* CONFIG_SOFTDISABLE */
 
 	.align	7
 do_stab_bolted_pSeries:
@@ -466,11 +468,16 @@ machine_check_common:
 	FINISH_NAP
 	DISABLE_INTS
 	bl	.save_nvgprs
+	TRACE_DISABLE_INTS
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.machine_check_exception
 	b	.ret_from_except
 
+#ifdef CONFIG_IPIPE
+	IPIPE_EXCEPTION_COMMON_LITE(0x900, decrementer, .__ipipe_grab_timer) 
+#else /* !CONFIG_IPIPE */
 	STD_EXCEPTION_COMMON_LITE(0x900, decrementer, .timer_interrupt)
+#endif /* CONFIG_IPIPE */
 	STD_EXCEPTION_COMMON(0xa00, trap_0a, .unknown_exception)
 	STD_EXCEPTION_COMMON(0xb00, trap_0b, .unknown_exception)
 	STD_EXCEPTION_COMMON(0xd00, single_step, .single_step_exception)
@@ -700,6 +707,7 @@ unrecov_slb:
 	EXCEPTION_PROLOG_COMMON(0x4100, PACA_EXSLB)
 	DISABLE_INTS
 	bl	.save_nvgprs
+ 	TRACE_DISABLE_INTS
 1:	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.unrecoverable_exception
 	b	1b
@@ -709,15 +717,23 @@ unrecov_slb:
 	.globl hardware_interrupt_entry
 hardware_interrupt_common:
 	EXCEPTION_PROLOG_COMMON(0x500, PACA_EXGEN)
-	FINISH_NAP
 hardware_interrupt_entry:
+#ifdef CONFIG_IPIPE
+	DISABLE_INTS_REALLY
+	TRACE_DISABLE_INTS_REALLY
+	addi	r3,r1,STACK_FRAME_OVERHEAD
+	bl	.__ipipe_grab_irq
+	b	.__ipipe_ret_from_except_lite
+#else /* !CONFIG_IPIPE */
 	DISABLE_INTS
+ 	FINISH_NAP
 BEGIN_FTR_SECTION
 	bl	.ppc64_runlatch_on
 END_FTR_SECTION_IFSET(CPU_FTR_CTRL)
 	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.do_IRQ
 	b	.ret_from_except_lite
+#endif /* CONFIG_IPIPE */
 
 #ifdef CONFIG_PPC_970_NAP
 power4_fixup_nap:
@@ -741,8 +757,9 @@ alignment_common:
 	std	r3,_DAR(r1)
 	std	r4,_DSISR(r1)
 	bl	.save_nvgprs
-	addi	r3,r1,STACK_FRAME_OVERHEAD
 	ENABLE_INTS
+	TRACE_ENABLE_INTS
+	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.alignment_exception
 	b	.ret_from_except
 
@@ -751,8 +768,9 @@ alignment_common:
 program_check_common:
 	EXCEPTION_PROLOG_COMMON(0x700, PACA_EXGEN)
 	bl	.save_nvgprs
-	addi	r3,r1,STACK_FRAME_OVERHEAD
 	ENABLE_INTS
+	TRACE_ENABLE_INTS
+	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.program_check_exception
 	b	.ret_from_except
 
@@ -762,8 +780,9 @@ fp_unavailable_common:
 	EXCEPTION_PROLOG_COMMON(0x800, PACA_EXGEN)
 	bne	1f			/* if from user, just load it up */
 	bl	.save_nvgprs
-	addi	r3,r1,STACK_FRAME_OVERHEAD
 	ENABLE_INTS
+	TRACE_ENABLE_INTS
+	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.kernel_fp_unavailable_exception
 	BUG_OPCODE
 1:	bl	.load_up_fpu
@@ -782,8 +801,9 @@ BEGIN_FTR_SECTION
 END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 #endif
 	bl	.save_nvgprs
-	addi	r3,r1,STACK_FRAME_OVERHEAD
 	ENABLE_INTS
+	TRACE_ENABLE_INTS
+	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.altivec_unavailable_exception
 	b	.ret_from_except
 
@@ -816,12 +836,23 @@ __end_handlers:
  * ret_from_except or ret_from_except_lite instead of this.
  */
 fast_exc_return_irq:			/* restores irq state too */
+#ifdef CONFIG_SOFTDISABLE	
 	ld	r3,SOFTE(r1)
 	TRACE_AND_RESTORE_IRQ(r3);
 	ld	r12,_MSR(r1)
 	rldicl	r4,r12,49,63		/* get MSR_EE to LSB */
 	stb	r4,PACAHARDIRQEN(r13)	/* restore paca->hard_enabled */
 	b	1f
+#else
+#ifdef CONFIG_IPIPE
+	ld	r3,SOFTE(r1)		/* currently hard-disabled, so this is safe */
+	nor	r3,r3,r3
+	ld	r4,PACAROOTPCPU(r13)
+	ld	r5,0(r4)
+	insrdi	r5,r3,1,63
+	std	r5,0(r4)
+#endif	/* CONFIG_IPIPE */
+#endif	/* CONFIG_SOFTDISABLE */
 
 	.globl	fast_exception_return
 fast_exception_return:
@@ -862,6 +893,7 @@ fast_exception_return:
 
 unrecov_fer:
 	bl	.save_nvgprs
+	TRACE_DISABLE_INTS
 1:	addi	r3,r1,STACK_FRAME_OVERHEAD
 	bl	.unrecoverable_exception
 	b	1b
@@ -1071,13 +1103,23 @@ END_FW_FTR_SECTION_IFCLR(FW_FEATURE_ISERIES)
 	 * to what it was before the trap.  Note that .raw_local_irq_restore
 	 * handles any interrupts pending at this point.
 	 */
+#ifdef CONFIG_SOFTDISABLE	
 	ld	r3,SOFTE(r1)
 	TRACE_AND_RESTORE_IRQ_PARTIAL(r3, 11f)
 	bl	.raw_local_irq_restore
+#else
+#ifdef CONFIG_IPIPE
+	ld	r3,SOFTE(r1)
+	nor	r3,r3,r3
+	andi.	r3,r3,1
+	bl	__ipipe_restore_if_root
+#endif			
+#endif
 	b	11f
 
 /* Here we have a page fault that hash_page can't handle. */
 handle_page_fault:
+	TRACE_ENABLE_INTS
 	ENABLE_INTS
 11:	ld	r4,_DAR(r1)
 	ld	r5,_DSISR(r1)
@@ -1642,10 +1684,12 @@ BEGIN_FW_FTR_SECTION
 	stb	r8,PACAHARDIRQEN(r13)
 END_FW_FTR_SECTION_IFSET(FW_FEATURE_ISERIES)
 #endif
+#ifdef CONFIG_SOFTDISABLE	
 BEGIN_FW_FTR_SECTION
 	stb	r7,PACAHARDIRQEN(r13)
 END_FW_FTR_SECTION_IFCLR(FW_FEATURE_ISERIES)
 	stb	r7,PACASOFTIRQEN(r13)
+#endif	
 
 	mtspr	SPRN_SRR0,r3
 	mtspr	SPRN_SRR1,r4
@@ -1768,8 +1812,10 @@ _INIT_GLOBAL(start_here_common)
 
 	/* Load up the kernel context */
 5:
+#ifdef CONFIG_SOFTDISABLE	
 	li	r5,0
 	stb	r5,PACASOFTIRQEN(r13)	/* Soft Disabled */
+#endif	
 #ifdef CONFIG_PPC_ISERIES
 BEGIN_FW_FTR_SECTION
 	mfmsr	r5
@@ -1778,7 +1824,9 @@ BEGIN_FW_FTR_SECTION
 	li	r5,1
 END_FW_FTR_SECTION_IFSET(FW_FEATURE_ISERIES)
 #endif
+#ifdef CONFIG_SOFTDISABLE	
 	stb	r5,PACAHARDIRQEN(r13)	/* Hard Disabled on others */
+#endif
 
 	bl	.start_kernel
 
diff --git a/arch/powerpc/kernel/head_8xx.S b/arch/powerpc/kernel/head_8xx.S
index 3c9452d..731677d 100644
--- a/arch/powerpc/kernel/head_8xx.S
+++ b/arch/powerpc/kernel/head_8xx.S
@@ -184,6 +184,12 @@ i##n:								\
 	EXC_XFER_TEMPLATE(n, hdlr, n, NOCOPY, transfer_to_handler_full,	\
 			  ret_from_except_full)
 
+#ifdef CONFIG_IPIPE
+#define EXC_XFER_IPIPE(n, hdlr)		\
+	EXC_XFER_TEMPLATE(n, hdlr, n+1, NOCOPY, transfer_to_handler, \
+			  __ipipe_ret_from_except)
+#endif /* CONFIG_IPIPE */
+ 
 #define EXC_XFER_LITE(n, hdlr)		\
 	EXC_XFER_TEMPLATE(n, hdlr, n+1, NOCOPY, transfer_to_handler, \
 			  ret_from_except)
@@ -235,7 +241,11 @@ InstructionAccess:
 	EXC_XFER_EE_LITE(0x400, handle_page_fault)
 
 /* External interrupt */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x500, HardwareInterrupt, __ipipe_grab_irq, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x500, HardwareInterrupt, do_IRQ, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 /* Alignment exception */
 	. = 0x600
@@ -256,7 +266,11 @@ Alignment:
 	EXCEPTION(0x800, FPUnavailable, unknown_exception, EXC_XFER_STD)
 
 /* Decrementer */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x900, Decrementer, __ipipe_grab_timer, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x900, Decrementer, timer_interrupt, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 	EXCEPTION(0xa00, Trap_0a, unknown_exception, EXC_XFER_EE)
 	EXCEPTION(0xb00, Trap_0b, unknown_exception, EXC_XFER_EE)
diff --git a/arch/powerpc/kernel/head_booke.h b/arch/powerpc/kernel/head_booke.h
index 17803ad..af978a6 100644
--- a/arch/powerpc/kernel/head_booke.h
+++ b/arch/powerpc/kernel/head_booke.h
@@ -218,6 +218,12 @@ label:
 	EXC_XFER_TEMPLATE(hdlr, n, MSR_KERNEL, NOCOPY, transfer_to_handler_full, \
 			  ret_from_except_full)
 
+#ifdef CONFIG_IPIPE
+#define EXC_XFER_IPIPE(n, hdlr)		\
+	EXC_XFER_TEMPLATE(hdlr, n+1, MSR_KERNEL, NOCOPY, transfer_to_handler, \
+			  __ipipe_ret_from_except)
+#endif /* CONFIG_IPIPE */
+
 #define EXC_XFER_LITE(n, hdlr)		\
 	EXC_XFER_TEMPLATE(hdlr, n+1, MSR_KERNEL, NOCOPY, transfer_to_handler, \
 			  ret_from_except)
@@ -382,6 +388,15 @@ label:
 	addi	r3,r1,STACK_FRAME_OVERHEAD;				      \
 	EXC_XFER_STD(0x0700, program_check_exception)
 
+#ifdef CONFIG_IPIPE
+#define DECREMENTER_EXCEPTION						      \
+	START_EXCEPTION(Decrementer)					      \
+	NORMAL_EXCEPTION_PROLOG;					      \
+	lis     r0,TSR_DIS@h;           /* Setup the DEC interrupt mask */    \
+	mtspr   SPRN_TSR,r0;		/* Clear the DEC interrupt */	      \
+	addi    r3,r1,STACK_FRAME_OVERHEAD;				      \
+	EXC_XFER_IPIPE(0x0900, __ipipe_grab_timer)
+#else /* !CONFIG_IPIPE */
 #define DECREMENTER_EXCEPTION						      \
 	START_EXCEPTION(Decrementer)					      \
 	NORMAL_EXCEPTION_PROLOG;					      \
@@ -389,6 +404,7 @@ label:
 	mtspr   SPRN_TSR,r0;		/* Clear the DEC interrupt */	      \
 	addi    r3,r1,STACK_FRAME_OVERHEAD;				      \
 	EXC_XFER_LITE(0x0900, timer_interrupt)
+#endif /* CONFIG_IPIPE */
 
 #define FP_UNAVAILABLE_EXCEPTION					      \
 	START_EXCEPTION(FloatingPointUnavailable)			      \
diff --git a/arch/powerpc/kernel/head_fsl_booke.S b/arch/powerpc/kernel/head_fsl_booke.S
index 36ffb35..9105bc8 100644
--- a/arch/powerpc/kernel/head_fsl_booke.S
+++ b/arch/powerpc/kernel/head_fsl_booke.S
@@ -495,7 +495,11 @@ interrupt_base:
 	INSTRUCTION_STORAGE_EXCEPTION
 
 	/* External Input Interrupt */
+#ifdef CONFIG_IPIPE
+	EXCEPTION(0x0500, ExternalInput, __ipipe_grab_irq, EXC_XFER_IPIPE)
+#else /* !CONFIG_IPIPE */
 	EXCEPTION(0x0500, ExternalInput, do_IRQ, EXC_XFER_LITE)
+#endif /* CONFIG_IPIPE */
 
 	/* Alignment Interrupt */
 	ALIGNMENT_EXCEPTION
diff --git a/arch/powerpc/kernel/idle.c b/arch/powerpc/kernel/idle.c
index 88d9c1d..e8f824a 100644
--- a/arch/powerpc/kernel/idle.c
+++ b/arch/powerpc/kernel/idle.c
@@ -59,6 +59,7 @@ void cpu_idle(void)
 		tick_nohz_stop_sched_tick(1);
 		while (!need_resched() && !cpu_should_die()) {
 			ppc64_runlatch_off();
+			ipipe_suspend_domain();
 
 			if (ppc_md.power_save) {
 				clear_thread_flag(TIF_POLLING_NRFLAG);
@@ -67,7 +68,7 @@ void cpu_idle(void)
 				 * is ordered w.r.t. need_resched() test.
 				 */
 				smp_mb();
-				local_irq_disable();
+				local_irq_disable_hw();
 
 				/* Don't trace irqs off for idle */
 				stop_critical_timings();
@@ -78,7 +79,7 @@ void cpu_idle(void)
 
 				start_critical_timings();
 
-				local_irq_enable();
+				local_irq_enable_hw();
 				set_thread_flag(TIF_POLLING_NRFLAG);
 
 			} else {
diff --git a/arch/powerpc/kernel/idle_power4.S b/arch/powerpc/kernel/idle_power4.S
index 5328709..8c3a2b7 100644
--- a/arch/powerpc/kernel/idle_power4.S
+++ b/arch/powerpc/kernel/idle_power4.S
@@ -34,9 +34,11 @@ END_FTR_SECTION_IFCLR(CPU_FTR_CAN_NAP)
 	rldicl	r0,r7,48,1
 	rotldi	r0,r0,16
 	mtmsrd	r0,1			/* hard-disable interrupts */
+#ifdef CONFIG_SOFTDISABLE	
 	li	r0,1
 	stb	r0,PACASOFTIRQEN(r13)	/* we'll hard-enable shortly */
 	stb	r0,PACAHARDIRQEN(r13)
+#endif CONFIG_SOFTDISABLE	
 BEGIN_FTR_SECTION
 	DSSALL
 	sync
@@ -59,10 +61,12 @@ _GLOBAL(power4_cpu_offline_powersave)
 	rldicl	r0,r7,48,1
 	rotldi	r0,r0,16
 	mtmsrd	r0,1			/* hard-disable interrupts */
+#ifdef CONFIG_SOFTDISABLE	
 	li	r0,1
 	li	r6,0
 	stb	r0,PACAHARDIRQEN(r13)	/* we'll hard-enable shortly */
 	stb	r6,PACASOFTIRQEN(r13)	/* soft-disable irqs */
+#endif
 BEGIN_FTR_SECTION
 	DSSALL
 	sync
diff --git a/arch/powerpc/kernel/ipipe.c b/arch/powerpc/kernel/ipipe.c
new file mode 100644
index 0000000..0c779a5
--- /dev/null
+++ b/arch/powerpc/kernel/ipipe.c
@@ -0,0 +1,867 @@
+/* -*- linux-c -*-
+ * linux/arch/powerpc/kernel/ipipe.c
+ *
+ * Copyright (C) 2005 Heikki Lindholm (PPC64 port).
+ * Copyright (C) 2004 Wolfgang Grandegger (Adeos/ppc port over 2.4).
+ * Copyright (C) 2002-2007 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Architecture-dependent I-PIPE core support for PowerPC 32/64bit.
+ */
+
+#include <linux/kernel.h>
+#include <linux/smp.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/bitops.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/module.h>
+#include <linux/kernel_stat.h>
+#include <asm/reg.h>
+#include <asm/system.h>
+#include <asm/mmu_context.h>
+#include <asm/unistd.h>
+#include <asm/machdep.h>
+#include <asm/atomic.h>
+#include <asm/hardirq.h>
+#include <asm/io.h>
+#include <asm/time.h>
+
+static void __ipipe_do_IRQ(unsigned irq, void *cookie);
+
+static void __ipipe_do_timer(unsigned irq, void *cookie);
+
+DEFINE_PER_CPU(struct pt_regs, __ipipe_tick_regs);
+#ifdef CONFIG_IPIPE_UNMASKED_CONTEXT_SWITCH
+DEFINE_PER_CPU(struct mm_struct *, ipipe_active_mm);
+EXPORT_PER_CPU_SYMBOL(ipipe_active_mm);
+#endif
+
+#define DECREMENTER_MAX	0x7fffffff
+
+#ifdef CONFIG_SMP
+
+static cpumask_t __ipipe_cpu_sync_map;
+
+static cpumask_t __ipipe_cpu_lock_map;
+
+static ipipe_spinlock_t __ipipe_cpu_barrier = IPIPE_SPIN_LOCK_UNLOCKED;
+
+static atomic_t __ipipe_critical_count = ATOMIC_INIT(0);
+
+static void (*__ipipe_cpu_sync) (void);
+
+static DEFINE_PER_CPU(struct ipipe_ipi_struct, ipipe_ipi_message);
+
+unsigned int __ipipe_ipi_irq = NR_IRQS + 1; /* dummy value */
+
+#ifdef CONFIG_DEBUGGER
+cpumask_t __ipipe_dbrk_pending;	/* pending debugger break IPIs */
+#endif
+
+/* Always called with hw interrupts off. */
+
+void __ipipe_do_critical_sync(unsigned irq, void *cookie)
+{
+	cpu_set(ipipe_processor_id(), __ipipe_cpu_sync_map);
+
+	/*
+	 * Now we are in sync with the lock requestor running on another
+	 * CPU. Enter a spinning wait until he releases the global
+	 * lock.
+	 */
+	spin_lock(&__ipipe_cpu_barrier);
+
+	/* Got it. Now get out. */
+
+	if (__ipipe_cpu_sync)
+		/* Call the sync routine if any. */
+		__ipipe_cpu_sync();
+
+	spin_unlock(&__ipipe_cpu_barrier);
+
+	cpu_clear(ipipe_processor_id(), __ipipe_cpu_sync_map);
+}
+
+void __ipipe_hook_critical_ipi(struct ipipe_domain *ipd)
+{
+	ipd->irqs[IPIPE_CRITICAL_IPI].acknowledge = NULL;
+	ipd->irqs[IPIPE_CRITICAL_IPI].handler = &__ipipe_do_critical_sync;
+	ipd->irqs[IPIPE_CRITICAL_IPI].cookie = NULL;
+	/* Immediately handle in the current domain but *never* pass */
+	ipd->irqs[IPIPE_CRITICAL_IPI].control =
+		IPIPE_HANDLE_MASK|IPIPE_STICKY_MASK|IPIPE_SYSTEM_MASK;
+}
+
+void __ipipe_register_ipi(unsigned int irq)
+{
+	__ipipe_ipi_irq = irq;
+	mb();
+} 
+
+static void __ipipe_ipi_demux(int irq, struct pt_regs *regs)
+{
+	int ipi, cpu = ipipe_processor_id();
+	irq_desc_t *desc = irq_desc + irq;
+
+	desc->ipipe_ack(irq, desc);
+
+	kstat_cpu(cpu).irqs[irq]++;
+
+	while (per_cpu(ipipe_ipi_message, cpu).value & IPIPE_MSG_IPI_MASK) {
+		for (ipi = IPIPE_MSG_CRITICAL_IPI; ipi <= IPIPE_MSG_SERVICE_IPI4; ++ipi) {
+			if (test_and_clear_bit(ipi, &per_cpu(ipipe_ipi_message, cpu).value)) {
+				mb();
+				__ipipe_handle_irq(ipi + IPIPE_MSG_IPI_OFFSET, NULL);
+			}
+		}
+	}
+
+#ifdef CONFIG_DEBUGGER
+	/*
+	 * The debugger IPI handler should be NMI-safe, so let's call
+	 * it immediately in case the IPI is pending.
+	 */
+	if (cpu_isset(cpu, __ipipe_dbrk_pending)) {
+		cpu_clear(cpu, __ipipe_dbrk_pending);
+		debugger_ipi(regs);
+	}
+#endif /* CONFIG_DEBUGGER */
+
+	__ipipe_end_irq(irq);
+}
+
+cpumask_t __ipipe_set_irq_affinity(unsigned irq, cpumask_t cpumask)
+{
+	cpumask_t oldmask = irq_to_desc(irq)->affinity;
+
+	if (irq_to_desc(irq)->chip->set_affinity == NULL)
+		return CPU_MASK_NONE;
+
+	if (cpus_empty(cpumask))
+		return oldmask; /* Return mask value -- no change. */
+
+	cpus_and(cpumask,cpumask,cpu_online_map);
+
+	if (cpus_empty(cpumask))
+		return CPU_MASK_NONE;	/* Error -- bad mask value or non-routable IRQ. */
+
+	irq_to_desc(irq)->chip->set_affinity(irq, &cpumask);
+
+	return oldmask;
+}
+
+int __ipipe_send_ipi(unsigned ipi, cpumask_t cpumask)
+{
+	extern void mpic_send_ipi(unsigned int ipi_no, unsigned int cpu_mask);
+	unsigned long flags;
+	cpumask_t testmask;
+	int cpu;
+
+	local_irq_save_hw(flags);
+
+	ipi -= IPIPE_MSG_IPI_OFFSET;
+	for_each_online_cpu(cpu) {
+		if (cpu_isset(cpu, cpumask))
+			set_bit(ipi, &per_cpu(ipipe_ipi_message, cpu).value);
+	}
+	mb();	 
+	
+	if (unlikely(cpus_empty(cpumask)))
+		goto out;
+
+	cpus_setall(testmask);
+	cpu_clear(ipipe_processor_id(), testmask);
+	if (likely(cpus_equal(cpumask, testmask)))
+		smp_ops->message_pass(MSG_ALL_BUT_SELF, PPC_MSG_IPIPE_DEMUX);
+	else {
+		/* Long path. */
+		for_each_cpu_mask_nr(cpu, cpumask)
+			smp_ops->message_pass(cpu, PPC_MSG_IPIPE_DEMUX);
+	}
+out:
+	local_irq_restore_hw(flags);
+
+	return 0;
+}
+
+void __ipipe_stall_root(void)
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+	set_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+	local_irq_restore_hw(flags);
+}
+
+unsigned long __ipipe_test_and_stall_root(void)
+{
+	unsigned long flags;
+	int x;
+
+	local_irq_save_hw(flags);
+	x = test_and_set_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+	local_irq_restore_hw(flags);
+
+	return x;
+}
+
+unsigned long __ipipe_test_root(void)
+{
+	unsigned long flags;
+	int x;
+
+	local_irq_save_hw(flags);
+	x = test_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+	local_irq_restore_hw(flags);
+
+	return x;
+}
+
+#endif	/* CONFIG_SMP */
+
+/*
+ * ipipe_critical_enter() -- Grab the superlock excluding all CPUs
+ * but the current one from a critical section. This lock is used when
+ * we must enforce a global critical section for a single CPU in a
+ * possibly SMP system whichever context the CPUs are running.
+ */
+unsigned long ipipe_critical_enter(void (*syncfn) (void))
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+
+#ifdef CONFIG_SMP
+	if (likely(num_online_cpus() > 1)) {
+		/* We might be running a SMP-kernel on a UP box... */
+		int cpu = ipipe_processor_id();
+		cpumask_t lock_map;
+		cpumask_t others;
+
+		if (!cpu_test_and_set(cpu, __ipipe_cpu_lock_map)) {
+			while (cpu_test_and_set(BITS_PER_LONG - 1, __ipipe_cpu_lock_map)) {
+				int n = 0;
+				do {
+					cpu_relax();
+				} while (++n < cpu);
+			}
+
+			spin_lock(&__ipipe_cpu_barrier);
+
+			__ipipe_cpu_sync = syncfn;
+
+			/* Send the sync IPI to all processors but the current one. */
+			cpus_setall(others);
+			cpu_clear(ipipe_processor_id(), others);
+			__ipipe_send_ipi(IPIPE_CRITICAL_IPI, others);
+
+			cpus_andnot(lock_map, cpu_online_map,
+				    __ipipe_cpu_lock_map);
+
+			while (!cpus_equal(__ipipe_cpu_sync_map, lock_map))
+				cpu_relax();
+		}
+
+		atomic_inc(&__ipipe_critical_count);
+	}
+#endif	/* CONFIG_SMP */
+
+	return flags;
+}
+
+/* ipipe_critical_exit() -- Release the superlock. */
+
+void ipipe_critical_exit(unsigned long flags)
+{
+#ifdef CONFIG_SMP
+	if (likely(num_online_cpus() > 1)) {
+		/* We might be running a SMP-kernel on a UP box... */
+		if (atomic_dec_and_test(&__ipipe_critical_count)) {
+			spin_unlock(&__ipipe_cpu_barrier);
+
+			while (!cpus_empty(__ipipe_cpu_sync_map))
+				cpu_relax();
+
+			cpu_clear(ipipe_processor_id(), __ipipe_cpu_lock_map);
+			cpu_clear(BITS_PER_LONG - 1, __ipipe_cpu_lock_map);
+		}
+	}
+#endif	/* CONFIG_SMP */
+
+	local_irq_restore_hw(flags);
+}
+
+void __ipipe_init_platform(void)
+{
+	unsigned int virq;
+
+	/*
+	 * Allocate a virtual IRQ for the decrementer trap early to
+	 * get it mapped to IPIPE_VIRQ_BASE
+	 */
+
+	virq = ipipe_alloc_virq();
+
+	if (virq != IPIPE_TIMER_VIRQ)
+		panic("I-pipe: cannot reserve timer virq #%d (got #%d)",
+		      IPIPE_TIMER_VIRQ, virq);
+
+#ifdef CONFIG_SMP
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_CRITICAL_IPI)
+		panic("I-pipe: cannot reserve critical IPI virq #%d (got #%d)",
+		      IPIPE_CRITICAL_IPI, virq);
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_SERVICE_IPI0)
+		panic("I-pipe: cannot reserve service IPI 0 virq #%d (got #%d)",
+		      IPIPE_SERVICE_IPI0, virq);
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_SERVICE_IPI1)
+		panic("I-pipe: cannot reserve service IPI 1 virq #%d (got #%d)",
+		      IPIPE_SERVICE_IPI1, virq);
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_SERVICE_IPI2)
+		panic("I-pipe: cannot reserve service IPI 2 virq #%d (got #%d)",
+		      IPIPE_SERVICE_IPI2, virq);
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_SERVICE_IPI3)
+		panic("I-pipe: cannot reserve service IPI 3 virq #%d (got #%d)",
+		      IPIPE_SERVICE_IPI3, virq);
+	virq = ipipe_alloc_virq();
+	if (virq != IPIPE_SERVICE_IPI4)
+		panic("I-pipe: cannot reserve service IPI 4 virq #%d (got #%d)",
+		      IPIPE_SERVICE_IPI4, virq);
+#endif
+}
+
+void __ipipe_end_irq(unsigned irq)
+{
+	irq_desc_t *desc = get_irq_desc(irq);
+	desc->ipipe_end(irq, desc);
+}
+
+void __ipipe_enable_irqdesc(struct ipipe_domain *ipd, unsigned irq)
+{
+	get_irq_desc(irq)->status &= ~IRQ_DISABLED;
+}
+
+static void __ipipe_ack_irq(unsigned irq, struct irq_desc *desc)
+{
+	desc->ipipe_ack(irq, desc);
+}
+
+/*
+ * __ipipe_enable_pipeline() -- We are running on the boot CPU, hw
+ * interrupts are off, and secondary CPUs are still lost in space.
+ */
+void __ipipe_enable_pipeline(void)
+{
+	unsigned long flags;
+	unsigned irq;
+
+	flags = ipipe_critical_enter(NULL);
+
+	/* First, virtualize all interrupts from the root domain. */
+
+	for (irq = 0; irq < NR_IRQS; irq++)
+		ipipe_virtualize_irq(ipipe_root_domain,
+				     irq,
+				     &__ipipe_do_IRQ, NULL,
+				     &__ipipe_ack_irq,
+				     IPIPE_HANDLE_MASK | IPIPE_PASS_MASK);
+	/*
+	 * We use a virtual IRQ to handle the timer irq (decrementer trap)
+	 * which has been allocated early in __ipipe_init_platform().
+	 */
+
+	ipipe_virtualize_irq(ipipe_root_domain,
+			     IPIPE_TIMER_VIRQ,
+			     &__ipipe_do_timer, NULL,
+			     NULL, IPIPE_HANDLE_MASK | IPIPE_PASS_MASK);
+
+	ipipe_critical_exit(flags);
+}
+
+int ipipe_get_sysinfo(struct ipipe_sysinfo *info)
+{
+	info->ncpus = num_online_cpus();
+	info->cpufreq = ipipe_cpu_freq();
+	info->archdep.tmirq = IPIPE_TIMER_VIRQ;
+	info->archdep.tmfreq = info->cpufreq;
+
+	return 0;
+}
+
+/*
+ * ipipe_trigger_irq() -- Push the interrupt at front of the pipeline
+ * just like if it has been actually received from a hw source. Also
+ * works for virtual interrupts.
+ */
+int ipipe_trigger_irq(unsigned irq)
+{
+	unsigned long flags;
+
+#ifdef CONFIG_IPIPE_DEBUG
+	if (irq >= IPIPE_NR_IRQS ||
+	    (ipipe_virtual_irq_p(irq)
+	     && !test_bit(irq - IPIPE_VIRQ_BASE, &__ipipe_virtual_irq_map)))
+		return -EINVAL;
+#endif
+	local_irq_save_hw(flags);
+	__ipipe_handle_irq(irq, NULL);
+	local_irq_restore_hw(flags);
+
+	return 1;
+}
+
+/*
+ * __ipipe_handle_irq() -- IPIPE's generic IRQ handler. An optimistic
+ * interrupt protection log is maintained here for each domain. Hw
+ * interrupts are off on entry.
+ */
+void __ipipe_handle_irq(int irq, struct pt_regs *regs)
+{
+	struct ipipe_domain *this_domain, *next_domain;
+	struct list_head *head, *pos;
+	int m_ack;
+
+	/* Software-triggered IRQs do not need any ack. */
+	m_ack = (regs == NULL);
+
+#ifdef CONFIG_IPIPE_DEBUG
+	if (unlikely(irq >= IPIPE_NR_IRQS)) {
+		printk(KERN_ERR "I-pipe: spurious interrupt %d\n", irq);
+		return;
+	}
+#endif
+	this_domain = __ipipe_current_domain;
+
+	if (unlikely(test_bit(IPIPE_STICKY_FLAG, &this_domain->irqs[irq].control)))
+		head = &this_domain->p_link;
+	else {
+		head = __ipipe_pipeline.next;
+		next_domain = list_entry(head, struct ipipe_domain, p_link);
+		if (likely(test_bit(IPIPE_WIRED_FLAG, &next_domain->irqs[irq].control))) {
+			if (!m_ack && next_domain->irqs[irq].acknowledge)
+				next_domain->irqs[irq].acknowledge(irq, irq_desc + irq);
+			__ipipe_dispatch_wired(next_domain, irq);
+			return;
+		}
+	}
+
+	/* Ack the interrupt. */
+
+	pos = head;
+
+	while (pos != &__ipipe_pipeline) {
+		next_domain = list_entry(pos, struct ipipe_domain, p_link);
+		prefetch(next_domain);
+		/*
+		 * For each domain handling the incoming IRQ, mark it as
+		 * pending in its log.
+		 */
+		if (test_bit(IPIPE_HANDLE_FLAG, &next_domain->irqs[irq].control)) {
+			/*
+			 * Domains that handle this IRQ are polled for
+			 * acknowledging it by decreasing priority order. The
+			 * interrupt must be made pending _first_ in the
+			 * domain's status flags before the PIC is unlocked.
+			 */
+			__ipipe_set_irq_pending(next_domain, irq);
+
+			if (!m_ack && next_domain->irqs[irq].acknowledge) {
+				next_domain->irqs[irq].acknowledge(irq, irq_desc + irq);
+				m_ack = 1;
+			}
+		}
+
+		/*
+		 * If the domain does not want the IRQ to be passed down the
+		 * interrupt pipe, exit the loop now.
+		 */
+		if (!test_bit(IPIPE_PASS_FLAG, &next_domain->irqs[irq].control))
+			break;
+
+		pos = next_domain->p_link.next;
+	}
+
+	/*
+	 * If the interrupt preempted the head domain, then do not
+	 * even try to walk the pipeline, unless an interrupt is
+	 * pending for it.
+	 */
+	if (test_bit(IPIPE_AHEAD_FLAG, &this_domain->flags) &&
+	    ipipe_head_cpudom_var(irqpend_himask) == 0)
+		return;
+
+	/*
+	 * Now walk the pipeline, yielding control to the highest
+	 * priority domain that has pending interrupt(s) or
+	 * immediately to the current domain if the interrupt has been
+	 * marked as 'sticky'. This search does not go beyond the
+	 * current domain in the pipeline.
+	 */
+
+	__ipipe_walk_pipeline(head);
+}
+
+int __ipipe_grab_irq(struct pt_regs *regs)
+{
+	extern int ppc_spurious_interrupts;
+	int irq;
+
+	irq = ppc_md.get_irq();
+	if (unlikely(irq == NO_IRQ)) {
+		ppc_spurious_interrupts++;
+		goto root_checks;
+	}
+
+	if (likely(irq != NO_IRQ_IGNORE)) {
+		ipipe_trace_irq_entry(irq);
+#ifdef CONFIG_SMP
+		/* Check for cascaded I-pipe IPIs */
+		if (irq == __ipipe_ipi_irq) {
+			__ipipe_ipi_demux(irq, regs);
+			ipipe_trace_irq_exit(irq);
+			goto root_checks;
+		}
+#endif /* CONFIG_SMP */
+		__ipipe_handle_irq(irq, regs);
+		ipipe_trace_irq_exit(irq);
+	}
+
+root_checks:
+
+	if (__ipipe_root_domain_p) {
+#ifdef CONFIG_PPC_970_NAP
+		struct thread_info *ti = current_thread_info();
+		/* Emulate the napping check when 100% sure we do run
+		 * over the root context. */
+		if (test_and_clear_bit(TLF_NAPPING, &ti->local_flags))
+			regs->nip = regs->link;
+#endif
+#ifdef CONFIG_PPC64
+		ppc64_runlatch_on();
+#endif
+		if (!test_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status)))
+			return 1;
+	}
+
+	return 0;
+}
+
+static void __ipipe_do_IRQ(unsigned irq, void *cookie)
+{
+	struct pt_regs *old_regs;
+#ifdef CONFIG_IRQSTACKS
+	struct thread_info *curtp, *irqtp;
+#endif
+
+	/* Provide a valid register frame, even if not the exact one. */
+	old_regs = set_irq_regs(&__raw_get_cpu_var(__ipipe_tick_regs));
+
+	irq_enter();
+
+#ifdef CONFIG_DEBUG_STACKOVERFLOW
+	/* Debugging check for stack overflow: is there less than 2KB free? */
+	{
+		long sp;
+
+		sp = __get_SP() & (THREAD_SIZE-1);
+
+		if (unlikely(sp < (sizeof(struct thread_info) + 2048))) {
+			printk("do_IRQ: stack overflow: %ld\n",
+				sp - sizeof(struct thread_info));
+			dump_stack();
+		}
+	}
+#endif
+
+#ifdef CONFIG_IRQSTACKS
+		/* Switch to the irq stack to handle this */
+		curtp = current_thread_info();
+		irqtp = hardirq_ctx[smp_processor_id()];
+		if (curtp != irqtp) {
+			struct irq_desc *desc = irq_desc + irq;
+			void *handler = desc->handle_irq;
+			if (handler == NULL)
+				handler = &__do_IRQ;
+			irqtp->task = curtp->task;
+			irqtp->flags = 0;
+			call_handle_irq(irq, desc, irqtp, handler);
+			irqtp->task = NULL;
+			if (irqtp->flags)
+				set_bits(irqtp->flags, &curtp->flags);
+		} else
+#endif
+			generic_handle_irq(irq);
+
+	irq_exit();
+
+	set_irq_regs(old_regs);
+}
+
+static void __ipipe_do_timer(unsigned irq, void *cookie)
+{
+	timer_interrupt(&__raw_get_cpu_var(__ipipe_tick_regs));
+}
+
+int __ipipe_grab_timer(struct pt_regs *regs)
+{
+	struct ipipe_domain *ipd, *head;
+
+	ipd = __ipipe_current_domain;
+	head = __ipipe_pipeline_head();
+
+	set_dec(DECREMENTER_MAX);
+
+	ipipe_trace_irq_entry(IPIPE_TIMER_VIRQ);
+
+	__raw_get_cpu_var(__ipipe_tick_regs).msr = regs->msr; /* for timer_interrupt() */
+	__raw_get_cpu_var(__ipipe_tick_regs).nip = regs->nip;
+
+	if (ipd != &ipipe_root)
+		__raw_get_cpu_var(__ipipe_tick_regs).msr &= ~MSR_EE;
+
+	if (test_bit(IPIPE_WIRED_FLAG, &head->irqs[IPIPE_TIMER_VIRQ].control))
+		/*
+		 * Finding a wired IRQ means that we do have a
+		 * registered head domain as well. The decrementer
+		 * interrupt requires no acknowledge, so we may branch
+		 * to the wired IRQ dispatcher directly. Additionally,
+		 * we may bypass checks for locked interrupts or
+		 * stalled stage (the decrementer cannot be locked and
+		 * the head domain is obviously not stalled since we
+		 * got there).
+		 */
+		__ipipe_dispatch_wired_nocheck(head, IPIPE_TIMER_VIRQ);
+	else
+		__ipipe_handle_irq(IPIPE_TIMER_VIRQ, NULL);
+
+	ipipe_trace_irq_exit(IPIPE_TIMER_VIRQ);
+
+	if (ipd == &ipipe_root) {
+#ifdef CONFIG_PPC_970_NAP
+		struct thread_info *ti = current_thread_info();
+		/* Emulate the napping check when 100% sure we do run
+		 * over the root context. */
+		if (test_and_clear_bit(TLF_NAPPING, &ti->local_flags))
+			regs->nip = regs->link;
+#endif
+#ifdef CONFIG_PPC64
+		ppc64_runlatch_on();
+#endif
+		if (!test_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status)))
+			return 1;
+	}
+
+	return 0;
+}
+
+notrace int __ipipe_check_root(void) /* hw IRQs off */
+{
+	return __ipipe_root_domain_p;
+}
+
+#ifdef CONFIG_PPC64
+
+#include <asm/firmware.h>
+#include <asm/lv1call.h>
+
+notrace void __ipipe_restore_if_root(unsigned long x) /* hw IRQs off */
+{
+	if (likely(!__ipipe_root_domain_p))
+		return;
+
+	if (x)
+		__set_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+	else
+		__clear_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+
+	if ((int)mfspr(SPRN_DEC) < 0)
+		mtspr(SPRN_DEC, 1);
+
+	/*
+	 * Force the delivery of pending soft-disabled interrupts on
+	 * PS3.  Any HV call will have this side effect.
+	 */
+	if (firmware_has_feature(FW_FEATURE_PS3_LV1)) {
+		u64 tmp;
+		lv1_get_version_info(&tmp);
+	}
+
+	local_irq_enable_hw();
+}
+
+#else
+
+#ifdef CONFIG_PREEMPT
+
+asmlinkage void __sched preempt_schedule_irq(void);
+
+void __sched  __ipipe_preempt_schedule_irq(void)
+{
+	struct ipipe_percpu_domain_data *p; 
+	unsigned long flags;  
+	/*
+	 * We have no IRQ state fixup on entry to exceptions, so we
+	 * have to stall the root stage before rescheduling.
+	 */  
+#ifdef CONFIG_IPIPE_DEBUG
+	BUG_ON(!irqs_disabled_hw());
+#endif
+	local_irq_save(flags);	
+	local_irq_enable_hw();	
+	preempt_schedule_irq(); /* Ok, may reschedule now. */  
+	local_irq_disable_hw(); 
+	/*
+	 * Flush any pending interrupt that may have been logged after
+	 * preempt_schedule_irq() stalled the root stage before
+	 * returning to us, and now.
+	 */
+	p = ipipe_root_cpudom_ptr(); 
+	if (unlikely(p->irqpend_himask != 0)) { 
+		add_preempt_count(PREEMPT_ACTIVE);
+		clear_bit(IPIPE_STALL_FLAG, &p->status); 
+		__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY); 
+		sub_preempt_count(PREEMPT_ACTIVE);
+	} 
+
+	__local_irq_restore_nosync(flags);  
+}
+
+#endif /* CONFIG_PREEMPT */
+
+#endif
+
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+
+notrace void __ipipe_trace_irqsoff(void)
+{
+	ipipe_trace_irqsoff();
+}
+
+notrace void __ipipe_trace_irqson(void)
+{
+	ipipe_trace_irqson();
+}
+
+notrace void __ipipe_trace_irqsx(unsigned long msr_ee)
+{
+	if (msr_ee)
+		ipipe_trace_irqson();
+	else
+		ipipe_trace_irqsoff();
+}
+
+#endif
+
+int __ipipe_syscall_root(struct pt_regs *regs)
+{
+	struct ipipe_percpu_domain_data *p;
+	unsigned long flags;
+        int ret;
+
+#ifdef CONFIG_PPC64
+	/*
+	 * Unlike ppc32, hw interrupts are off on entry here.  We did
+	 * not copy the stall state on entry yet, so do it now.
+	 */
+	p = ipipe_root_cpudom_ptr();
+	regs->softe = !test_bit(IPIPE_STALL_FLAG, &p->status);
+
+	/* We ran DISABLE_INTS before being sent to the syscall
+	 * dispatcher, so we need to unstall the root stage, unless
+	 * the root domain is not current. */
+	if (__ipipe_root_domain_p)
+		__clear_bit(IPIPE_STALL_FLAG, &p->status);
+
+	local_irq_enable_hw();
+#endif
+        /*
+         * This routine either returns:
+         * 0 -- if the syscall is to be passed to Linux;
+         * >0 -- if the syscall should not be passed to Linux, and no
+         * tail work should be performed;
+         * <0 -- if the syscall should not be passed to Linux but the
+         * tail work has to be performed (for handling signals etc).
+         */
+
+        if (!__ipipe_syscall_watched_p(current, regs->gpr[0]) ||
+            !__ipipe_event_monitored_p(IPIPE_EVENT_SYSCALL))
+                return 0;
+
+        ret = __ipipe_dispatch_event(IPIPE_EVENT_SYSCALL, regs);
+
+	local_irq_save_hw(flags);
+
+        if (!__ipipe_root_domain_p) {
+		local_irq_restore_hw(flags);
+		return 1;
+	}
+
+	p = ipipe_root_cpudom_ptr();
+	if ((p->irqpend_himask & IPIPE_IRQMASK_VIRT) != 0)
+		__ipipe_sync_pipeline(IPIPE_IRQMASK_VIRT);
+
+	local_irq_restore_hw(flags);
+
+	return -ret;
+}
+
+void __ipipe_pin_range_globally(unsigned long start, unsigned long end)
+{
+	/* We don't support this. */
+}
+
+EXPORT_PER_CPU_SYMBOL(__ipipe_tick_regs);
+extern void *sys_call_table;
+EXPORT_SYMBOL(sys_call_table);
+
+#ifdef CONFIG_SMP
+EXPORT_SYMBOL(__ipipe_stall_root);
+EXPORT_SYMBOL(__ipipe_test_root);
+EXPORT_SYMBOL(__ipipe_test_and_stall_root);
+#endif
+
+EXPORT_SYMBOL_GPL(__switch_to);
+EXPORT_SYMBOL_GPL(show_stack);
+EXPORT_SYMBOL_GPL(_switch);
+EXPORT_SYMBOL_GPL(tasklist_lock);
+#ifdef CONFIG_PPC64
+EXPORT_PER_CPU_SYMBOL(ppc64_tlb_batch);
+EXPORT_SYMBOL_GPL(switch_slb);
+EXPORT_SYMBOL_GPL(switch_stab);
+EXPORT_SYMBOL_GPL(__flush_tlb_pending);
+EXPORT_SYMBOL_GPL(mmu_linear_psize);
+EXPORT_SYMBOL_GPL(mmu_psize_defs);
+#else  /* !CONFIG_PPC64 */
+void atomic_set_mask(unsigned long mask, unsigned long *ptr);
+void atomic_clear_mask(unsigned long mask, unsigned long *ptr);
+#ifdef FEW_CONTEXTS
+EXPORT_SYMBOL_GPL(nr_free_contexts);
+EXPORT_SYMBOL_GPL(context_mm);
+EXPORT_SYMBOL_GPL(steal_context);
+#endif	/* !FEW_CONTEXTS */
+EXPORT_SYMBOL_GPL(atomic_set_mask);
+EXPORT_SYMBOL_GPL(atomic_clear_mask);
+#ifndef CONFIG_SMP
+EXPORT_SYMBOL_GPL(last_task_used_math);
+#endif
+#endif	/* !CONFIG_PPC64 */
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 23b8b5e..8a325ea 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -71,7 +71,7 @@
 #endif
 
 int __irq_offset_value;
-static int ppc_spurious_interrupts;
+int ppc_spurious_interrupts;
 
 #ifdef CONFIG_PPC32
 EXPORT_SYMBOL(__irq_offset_value);
@@ -88,6 +88,8 @@ EXPORT_SYMBOL(irq_desc);
 
 int distribute_irqs = 1;
 
+#ifdef CONFIG_SOFTDISABLE
+
 static inline notrace unsigned long get_hard_enabled(void)
 {
 	unsigned long enabled;
@@ -165,6 +167,9 @@ notrace void raw_local_irq_restore(unsigned long en)
 	__hard_irq_enable();
 }
 EXPORT_SYMBOL(raw_local_irq_restore);
+
+#endif /* !CONFIG_SOFTDISABLE */
+
 #endif /* CONFIG_PPC64 */
 
 int show_interrupts(struct seq_file *p, void *v)
diff --git a/arch/powerpc/kernel/ppc_ksyms.c b/arch/powerpc/kernel/ppc_ksyms.c
index 69956f7..e1c9db2 100644
--- a/arch/powerpc/kernel/ppc_ksyms.c
+++ b/arch/powerpc/kernel/ppc_ksyms.c
@@ -123,6 +123,9 @@ EXPORT_SYMBOL(flush_dcache_range);
 #ifdef CONFIG_SMP
 #ifdef CONFIG_PPC32
 EXPORT_SYMBOL(smp_hw_index);
+#ifdef CONFIG_IPIPE
+EXPORT_SYMBOL(smp_logical_index);
+#endif
 #endif
 #endif
 
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 67df0bb..7b7122f 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -95,8 +95,12 @@ void flush_fp_to_thread(struct task_struct *tsk)
 
 void enable_kernel_fp(void)
 {
+	unsigned long flags;
+
 	WARN_ON(preemptible());
 
+	local_irq_save_hw_cond(flags);
+
 #ifdef CONFIG_SMP
 	if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
 		giveup_fpu(current);
@@ -105,6 +109,7 @@ void enable_kernel_fp(void)
 #else
 	giveup_fpu(last_task_used_math);
 #endif /* CONFIG_SMP */
+	local_irq_restore_hw_cond(flags);
 }
 EXPORT_SYMBOL(enable_kernel_fp);
 
@@ -399,7 +404,7 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	}
 #endif
 
-	local_irq_save(flags);
+	local_irq_save_hw(flags);
 
 	account_system_vtime(current);
 	account_process_vtime(current);
@@ -418,7 +423,7 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	vsid = get_vsid(current->mm->context.id, 0, ssize);
 
 	/* current is still really us, just a different us :-) */
-	if (current->mm) {
+	if (__ipipe_root_domain_p && current->mm) {
 #ifdef CONFIG_PPC_64K_PAGES
 		__hash_page_64K(0, _PAGE_USER|_PAGE_RW, vsid, &current->zero_pte.pte, 0x300, 1, ssize);
 #else
@@ -427,7 +432,7 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	}
 #endif
 
-	local_irq_restore(flags);
+	local_irq_restore_hw(flags);
 
 	return last;
 }
@@ -1078,7 +1083,7 @@ void dump_stack(void)
 }
 EXPORT_SYMBOL(dump_stack);
 
-#ifdef CONFIG_PPC64
+#ifdef CONFIG_RUNLATCH
 void ppc64_runlatch_on(void)
 {
 	unsigned long ctrl;
diff --git a/arch/powerpc/kernel/prom_init_check.sh b/arch/powerpc/kernel/prom_init_check.sh
index ea3a2ec..0c233d3 100644
--- a/arch/powerpc/kernel/prom_init_check.sh
+++ b/arch/powerpc/kernel/prom_init_check.sh
@@ -20,7 +20,7 @@ WHITELIST="add_reloc_offset __bss_start __bss_stop copy_and_flush
 _end enter_prom memcpy memset reloc_offset __secondary_hold
 __secondary_hold_acknowledge __secondary_hold_spinloop __start
 strcmp strcpy strlcpy strlen strncmp strstr logo_linux_clut224
-reloc_got2 kernstart_addr memstart_addr"
+reloc_got2 kernstart_addr memstart_addr _mcount"
 
 NM="$1"
 OBJ="$2"
diff --git a/arch/powerpc/kernel/setup_32.c b/arch/powerpc/kernel/setup_32.c
index 9e1ca74..84da26c 100644
--- a/arch/powerpc/kernel/setup_32.c
+++ b/arch/powerpc/kernel/setup_32.c
@@ -51,6 +51,9 @@ EXPORT_SYMBOL_GPL(boot_cpuid);
 int boot_cpuid_phys;
 
 int smp_hw_index[NR_CPUS];
+#ifdef CONFIG_IPIPE
+int smp_logical_index[NR_CPUS];
+#endif
 
 unsigned long ISA_DMA_THRESHOLD;
 unsigned int DMA_MODE_READ;
diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index 2d34196..3b72de3 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -235,8 +235,10 @@ void early_setup_secondary(void)
 {
 	struct paca_struct *lpaca = get_paca();
 
+#ifdef CONFIG_SOFTDISABLE
 	/* Mark interrupts enabled in PACA */
 	lpaca->soft_enabled = 0;
+#endif
 
 	/* Initialize hash table for that CPU */
 	htab_initialize_secondary();
@@ -354,6 +356,12 @@ static void __init initialize_cache_info(void)
  */
 void __init setup_system(void)
 {
+#ifdef CONFIG_IPIPE
+	/* Early temporary init, before per-cpu areas are moved to
+	 * their final location. */
+	get_paca()->root_percpu = (u64)&ipipe_percpudom(&ipipe_root, status, 0);
+#endif	
+
 	DBG(" -> setup_system()\n");
 
 	/* Apply the CPUs-specific and firmware specific fixups to kernel
@@ -611,6 +619,10 @@ void __init setup_per_cpu_areas(void)
 		paca[i].data_offset = ptr - __per_cpu_start;
 		memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 	}
+#ifdef CONFIG_IPIPE
+	/* Reset pointer to the relocated per-cpu root domain data. */
+	get_paca()->root_percpu = (u64)&ipipe_percpudom(&ipipe_root, status, 0);
+#endif	
 }
 #endif
 
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 65484b2..8968b24 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -153,7 +153,7 @@ const char *smp_ipi_name[] = {
 	[PPC_MSG_CALL_FUNCTION] =  "ipi call function",
 	[PPC_MSG_RESCHEDULE] = "ipi reschedule",
 	[PPC_MSG_CALL_FUNC_SINGLE] = "ipi call function single",
-	[PPC_MSG_DEBUGGER_BREAK] = "ipi debugger",
+	[PPC_MSG_DEBUGGER_BREAK] = "ipi I-pipe/debugger",
 };
 
 /* optional function to request ipi, for controllers with >= 4 ipis */
@@ -169,6 +169,12 @@ int smp_request_message_ipi(int virq, int msg)
 		return 1;
 	}
 #endif
+#ifdef CONFIG_IPIPE
+	if (msg == PPC_MSG_DEBUGGER_BREAK)
+		/* Piggyback the debugger IPI for the I-pipe. */
+		__ipipe_register_ipi(virq);
+#endif
+
 	err = request_irq(virq, smp_ipi_action[msg], IRQF_DISABLED|IRQF_PERCPU,
 			  smp_ipi_name[msg], 0);
 	WARN(err < 0, "unable to request_irq %d for %s (rc %d)\n",
@@ -199,8 +205,12 @@ void arch_send_call_function_ipi(cpumask_t mask)
 #ifdef CONFIG_DEBUGGER
 void smp_send_debugger_break(int cpu)
 {
-	if (likely(smp_ops))
+	if (likely(smp_ops)) {
+#ifdef CONFIG_IPIPE
+		cpu_set(cpu, __ipipe_dbrk_pending);
+#endif
 		smp_ops->message_pass(cpu, PPC_MSG_DEBUGGER_BREAK);
+	}
 }
 #endif
 
@@ -209,6 +219,10 @@ void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
 {
 	crash_ipi_function_ptr = crash_ipi_callback;
 	if (crash_ipi_callback && smp_ops) {
+#ifdef CONFIG_IPIPE
+		cpus_setall(__ipipe_dbrk_pending);
+		cpu_clear(ipipe_processor_id(), __ipipe_dbrk_pending);
+#endif
 		mb();
 		smp_ops->message_pass(MSG_ALL_BUT_SELF, PPC_MSG_DEBUGGER_BREAK);
 	}
@@ -485,6 +499,9 @@ int __devinit start_secondary(void *unused)
 	struct device_node *l2_cache;
 	int i, base;
 
+#if defined(CONFIG_IPIPE) && defined(CONFIG_PPC64)
+	get_paca()->root_percpu = (u64)&ipipe_percpudom(&ipipe_root, status, cpu);
+#endif
 	atomic_inc(&init_mm.mm_count);
 	current->active_mm = &init_mm;
 
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 9da1b2c..8ec0813 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -124,6 +124,9 @@ struct decrementer_clock {
 
 static DEFINE_PER_CPU(struct decrementer_clock, decrementers);
 
+DEFINE_PER_CPU(int, disarm_decr);
+EXPORT_PER_CPU_SYMBOL(disarm_decr);
+
 #ifdef CONFIG_PPC_ISERIES
 static unsigned long __initdata iSeries_recal_titan;
 static signed long __initdata iSeries_recal_tb;
@@ -543,14 +545,16 @@ void timer_interrupt(struct pt_regs * regs)
 	struct pt_regs *old_regs;
 	struct decrementer_clock *decrementer =  &__get_cpu_var(decrementers);
 	struct clock_event_device *evt = &decrementer->event;
+	int cpu = smp_processor_id();
 	u64 now;
 
 	/* Ensure a positive value is written to the decrementer, or else
 	 * some CPUs will continuue to take decrementer exceptions */
-	set_dec(DECREMENTER_MAX);
+	if (!per_cpu(disarm_decr, cpu))
+		set_dec(DECREMENTER_MAX);
 
 #ifdef CONFIG_PPC_PASEMI_A2_WORKAROUNDS
-	extern spinlock_t native_tlbie_lock;
+	extern ipipe_spinlock_t native_tlbie_lock;
 
 	spin_lock(&native_tlbie_lock);
 	asm("ptesync");
@@ -562,16 +566,25 @@ void timer_interrupt(struct pt_regs * regs)
 		do_IRQ(regs);
 #endif
 
-	now = get_tb_or_rtc();
-	if (now < decrementer->next_tb) {
-		/* not time for this event yet */
-		now = decrementer->next_tb - now;
-		if (now <= DECREMENTER_MAX)
-			set_dec((int)now);
-		return;
+ 	if (!per_cpu(disarm_decr, cpu)) {
+		now = get_tb_or_rtc();
+		if (now < decrementer->next_tb) {
+			/* not time for this event yet */
+			now = decrementer->next_tb - now;
+			if (now <= DECREMENTER_MAX)
+				set_dec((int)now);
+			return;
+		}
 	}
 	old_regs = set_irq_regs(regs);
+#ifndef CONFIG_IPIPE
+	/*
+	 * The timer interrupt is a virtual one when the I-pipe is
+	 * active, therefore we already called irq_enter() for it (see
+	 * __ipipe_run_isr).
+	 */
 	irq_enter();
+#endif
 
 	calculate_steal_time();
 
@@ -596,7 +609,9 @@ void timer_interrupt(struct pt_regs * regs)
 	}
 #endif
 
+#ifndef CONFIG_IPIPE
 	irq_exit();
+#endif
 	set_irq_regs(old_regs);
 }
 
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index 5457e95..80e86b5 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -485,6 +485,9 @@ void machine_check_exception(struct pt_regs *regs)
 {
 	int recover = 0;
 
+	if (ipipe_trap_notify(IPIPE_TRAP_MCE,regs))
+	    	return;
+
 	/* See if any machine dependent calls. In theory, we would want
 	 * to call the CPU first, and call the ppc_md. one if the CPU
 	 * one returns a positive number. However there is existing code
@@ -543,11 +546,17 @@ void unknown_exception(struct pt_regs *regs)
 	printk("Bad trap at PC: %lx, SR: %lx, vector=%lx\n",
 	       regs->nip, regs->msr, regs->trap);
 
+	if (ipipe_trap_notify(IPIPE_TRAP_UNKNOWN,regs))
+	    	return;
+
 	_exception(SIGTRAP, regs, 0, 0);
 }
 
 void instruction_breakpoint_exception(struct pt_regs *regs)
 {
+	if (ipipe_trap_notify(IPIPE_TRAP_IABR,regs))
+	    	return;
+
 	if (notify_die(DIE_IABR_MATCH, "iabr_match", regs, 5,
 					5, SIGTRAP) == NOTIFY_STOP)
 		return;
@@ -558,6 +567,8 @@ void instruction_breakpoint_exception(struct pt_regs *regs)
 
 void RunModeException(struct pt_regs *regs)
 {
+	if (ipipe_trap_notify(IPIPE_TRAP_RM,regs))
+	    	return;
 	_exception(SIGTRAP, regs, 0, 0);
 }
 
@@ -565,6 +576,9 @@ void __kprobes single_step_exception(struct pt_regs *regs)
 {
 	regs->msr &= ~(MSR_SE | MSR_BE);  /* Turn off 'trace' bits */
 
+	if (ipipe_trap_notify(IPIPE_TRAP_SSTEP,regs))
+	    return;
+
 	if (notify_die(DIE_SSTEP, "single_step", regs, 5,
 					5, SIGTRAP) == NOTIFY_STOP)
 		return;
@@ -584,6 +598,8 @@ static void emulate_single_step(struct pt_regs *regs)
 {
 	if (single_stepping(regs)) {
 		clear_single_step(regs);
+		if (ipipe_trap_notify(IPIPE_TRAP_SSTEP,regs))
+		    return;
 		_exception(SIGTRAP, regs, TRAP_TRACE, 0);
 	}
 }
@@ -825,6 +841,9 @@ void __kprobes program_check_exception(struct pt_regs *regs)
 	/* We can now get here via a FP Unavailable exception if the core
 	 * has no FPU, in that case the reason flags will be 0 */
 
+	if (ipipe_trap_notify(IPIPE_TRAP_PCE,regs))
+	    	return;
+
 	if (reason & REASON_FP) {
 		/* IEEE FP exception */
 		parse_fpe(regs);
@@ -897,6 +916,9 @@ void alignment_exception(struct pt_regs *regs)
 {
 	int sig, code, fixed = 0;
 
+	if (ipipe_trap_notify(IPIPE_TRAP_ALIGNMENT,regs))
+	    	return;
+
 	/* we don't implement logging of alignment exceptions */
 	if (!(current->thread.align_ctl & PR_UNALIGN_SIGBUS))
 		fixed = fix_alignment(regs);
@@ -934,6 +956,8 @@ void nonrecoverable_exception(struct pt_regs *regs)
 {
 	printk(KERN_ERR "Non-recoverable exception at PC=%lx MSR=%lx\n",
 	       regs->nip, regs->msr);
+	if (ipipe_trap_notify(IPIPE_TRAP_NREC,regs))
+	    	return;
 	debugger(regs);
 	die("nonrecoverable exception", regs, SIGKILL);
 }
@@ -949,11 +973,16 @@ void kernel_fp_unavailable_exception(struct pt_regs *regs)
 {
 	printk(KERN_EMERG "Unrecoverable FP Unavailable Exception "
 			  "%lx at %lx\n", regs->trap, regs->nip);
+	if (ipipe_trap_notify(IPIPE_TRAP_KFPUNAVAIL,regs))
+	    	return;
 	die("Unrecoverable FP Unavailable Exception", regs, SIGABRT);
 }
 
 void altivec_unavailable_exception(struct pt_regs *regs)
 {
+	if (ipipe_trap_notify(IPIPE_TRAP_ALTUNAVAIL,regs))
+	    	return;
+
 	if (user_mode(regs)) {
 		/* A user program has executed an altivec instruction,
 		   but this kernel doesn't support altivec. */
@@ -994,6 +1023,9 @@ void SoftwareEmulation(struct pt_regs *regs)
 	int errcode;
 #endif
 
+	if (ipipe_trap_notify(IPIPE_TRAP_SOFTEMU,regs))
+	    	return;
+
 	CHECK_FULL_REGS(regs);
 
 	if (!user_mode(regs)) {
@@ -1045,6 +1077,9 @@ void SoftwareEmulation(struct pt_regs *regs)
 
 void __kprobes DebugException(struct pt_regs *regs, unsigned long debug_status)
 {
+	if (ipipe_trap_notify(IPIPE_TRAP_DEBUG,regs))
+	    	return;
+
 	if (debug_status & DBSR_IC) {	/* instruction completion */
 		regs->msr &= ~MSR_DE;
 
@@ -1099,6 +1134,9 @@ void altivec_assist_exception(struct pt_regs *regs)
 {
 	int err;
 
+	if (ipipe_trap_notify(IPIPE_TRAP_ALTASSIST,regs))
+	    	return;
+
 	if (!user_mode(regs)) {
 		printk(KERN_EMERG "VMX/Altivec assist exception in kernel mode"
 		       " at %lx\n", regs->nip);
@@ -1151,8 +1189,11 @@ void CacheLockingException(struct pt_regs *regs, unsigned long address,
 	 * as priv ops, in the future we could try to do
 	 * something smarter
 	 */
-	if (error_code & (ESR_DLK|ESR_ILK))
+	if (error_code & (ESR_DLK|ESR_ILK)) {
+		if (ipipe_trap_notify(IPIPE_TRAP_CACHE,regs))
+			return;
 		_exception(SIGILL, regs, ILL_PRVOPC, regs->nip);
+	}
 	return;
 }
 #endif /* CONFIG_FSL_BOOKE */
@@ -1171,6 +1212,9 @@ void SPEFloatingPointException(struct pt_regs *regs)
 		giveup_spe(current);
 	preempt_enable();
 
+	if (ipipe_trap_notify(IPIPE_TRAP_SPE,regs))
+	    	return;
+
 	spefscr = current->thread.spefscr;
 	fpexc_mode = current->thread.fpexc_mode;
 
@@ -1251,6 +1295,8 @@ void unrecoverable_exception(struct pt_regs *regs)
 {
 	printk(KERN_EMERG "Unrecoverable exception %lx at %lx\n",
 	       regs->trap, regs->nip);
+	if (ipipe_trap_notify(IPIPE_TRAP_NREC,regs))
+	    	return;
 	die("Unrecoverable exception", regs, SIGABRT);
 }
 
diff --git a/arch/powerpc/lib/code-patching.c b/arch/powerpc/lib/code-patching.c
index 7c975d4..0c163f2 100644
--- a/arch/powerpc/lib/code-patching.c
+++ b/arch/powerpc/lib/code-patching.c
@@ -15,17 +15,20 @@
 #include <asm/code-patching.h>
 
 
+notrace
 void patch_instruction(unsigned int *addr, unsigned int instr)
 {
 	*addr = instr;
 	asm ("dcbst 0, %0; sync; icbi 0,%0; sync; isync" : : "r" (addr));
 }
 
+notrace
 void patch_branch(unsigned int *addr, unsigned long target, int flags)
 {
 	patch_instruction(addr, create_branch(addr, target, flags));
 }
 
+notrace
 unsigned int create_branch(const unsigned int *addr,
 			   unsigned long target, int flags)
 {
@@ -46,6 +49,7 @@ unsigned int create_branch(const unsigned int *addr,
 	return instruction;
 }
 
+notrace
 unsigned int create_cond_branch(const unsigned int *addr,
 				unsigned long target, int flags)
 {
@@ -66,21 +70,25 @@ unsigned int create_cond_branch(const unsigned int *addr,
 	return instruction;
 }
 
+notrace
 static unsigned int branch_opcode(unsigned int instr)
 {
 	return (instr >> 26) & 0x3F;
 }
 
+notrace
 static int instr_is_branch_iform(unsigned int instr)
 {
 	return branch_opcode(instr) == 18;
 }
 
+notrace
 static int instr_is_branch_bform(unsigned int instr)
 {
 	return branch_opcode(instr) == 16;
 }
 
+notrace
 int instr_is_relative_branch(unsigned int instr)
 {
 	if (instr & BRANCH_ABSOLUTE)
@@ -89,6 +97,7 @@ int instr_is_relative_branch(unsigned int instr)
 	return instr_is_branch_iform(instr) || instr_is_branch_bform(instr);
 }
 
+notrace
 static unsigned long branch_iform_target(const unsigned int *instr)
 {
 	signed long imm;
@@ -105,6 +114,7 @@ static unsigned long branch_iform_target(const unsigned int *instr)
 	return (unsigned long)imm;
 }
 
+notrace
 static unsigned long branch_bform_target(const unsigned int *instr)
 {
 	signed long imm;
@@ -121,6 +131,7 @@ static unsigned long branch_bform_target(const unsigned int *instr)
 	return (unsigned long)imm;
 }
 
+notrace
 unsigned long branch_target(const unsigned int *instr)
 {
 	if (instr_is_branch_iform(*instr))
@@ -131,6 +142,7 @@ unsigned long branch_target(const unsigned int *instr)
 	return 0;
 }
 
+notrace
 int instr_is_branch_to_addr(const unsigned int *instr, unsigned long addr)
 {
 	if (instr_is_branch_iform(*instr) || instr_is_branch_bform(*instr))
@@ -139,6 +151,7 @@ int instr_is_branch_to_addr(const unsigned int *instr, unsigned long addr)
 	return 0;
 }
 
+notrace
 unsigned int translate_branch(const unsigned int *dest, const unsigned int *src)
 {
 	unsigned long target;
diff --git a/arch/powerpc/lib/feature-fixups.c b/arch/powerpc/lib/feature-fixups.c
index 8c5a03b..84cb78c 100644
--- a/arch/powerpc/lib/feature-fixups.c
+++ b/arch/powerpc/lib/feature-fixups.c
@@ -28,6 +28,7 @@ struct fixup_entry {
 	long		alt_end_off;
 };
 
+notrace
 static unsigned int *calc_addr(struct fixup_entry *fcur, long offset)
 {
 	/*
@@ -38,6 +39,7 @@ static unsigned int *calc_addr(struct fixup_entry *fcur, long offset)
 	return (unsigned int *)((unsigned long)fcur + offset);
 }
 
+notrace
 static int patch_alt_instruction(unsigned int *src, unsigned int *dest,
 				 unsigned int *alt_start, unsigned int *alt_end)
 {
@@ -61,6 +63,7 @@ static int patch_alt_instruction(unsigned int *src, unsigned int *dest,
 	return 0;
 }
 
+notrace
 static int patch_feature_section(unsigned long value, struct fixup_entry *fcur)
 {
 	unsigned int *start, *end, *alt_start, *alt_end, *src, *dest;
@@ -90,6 +93,7 @@ static int patch_feature_section(unsigned long value, struct fixup_entry *fcur)
 	return 0;
 }
 
+notrace
 void do_feature_fixups(unsigned long value, void *fixup_start, void *fixup_end)
 {
 	struct fixup_entry *fcur, *fend;
@@ -110,6 +114,7 @@ void do_feature_fixups(unsigned long value, void *fixup_start, void *fixup_end)
 	}
 }
 
+notrace
 void do_lwsync_fixups(unsigned long value, void *fixup_start, void *fixup_end)
 {
 	unsigned int *start, *end, *dest;
diff --git a/arch/powerpc/mm/fault.c b/arch/powerpc/mm/fault.c
index 91c7b86..3d9c670 100644
--- a/arch/powerpc/mm/fault.c
+++ b/arch/powerpc/mm/fault.c
@@ -118,13 +118,18 @@ int __kprobes do_page_fault(struct pt_regs *regs, unsigned long address,
 			    unsigned long error_code)
 {
 	struct vm_area_struct * vma;
-	struct mm_struct *mm = current->mm;
+	struct mm_struct *mm;
 	siginfo_t info;
 	int code = SEGV_MAPERR;
 	int is_write = 0, ret;
 	int trap = TRAP(regs);
  	int is_exec = trap == 0x400;
 
+	if (ipipe_trap_notify(IPIPE_TRAP_ACCESS,regs))
+	    	return 0;
+
+	mm = current->mm;
+
 #if !(defined(CONFIG_4xx) || defined(CONFIG_BOOKE))
 	/*
 	 * Fortunately the bit assignments in SRR1 for an instruction
diff --git a/arch/powerpc/mm/hash_low_32.S b/arch/powerpc/mm/hash_low_32.S
index 14af8ce..25373be 100644
--- a/arch/powerpc/mm/hash_low_32.S
+++ b/arch/powerpc/mm/hash_low_32.S
@@ -496,7 +496,11 @@ htab_hash_searches:
  *
  * We assume that there is a hash table in use (Hash != 0).
  */
+#ifdef CONFIG_IPIPE
+_GLOBAL(__flush_hash_pages)
+#else
 _GLOBAL(flush_hash_pages)
+#endif
 	tophys(r7,0)
 
 	/*
@@ -531,18 +535,9 @@ _GLOBAL(flush_hash_pages)
 	addi	r6,r6,-1
 	b	1b
 
-	/* Convert context and va to VSID */
-2:	mulli	r3,r3,897*16		/* multiply context by context skew */
-	rlwinm	r0,r4,4,28,31		/* get ESID (top 4 bits of va) */
-	mulli	r0,r0,0x111		/* multiply by ESID skew */
-	add	r3,r3,r0		/* note code below trims to 24 bits */
-
-	/* Construct the high word of the PPC-style PTE (r11) */
-	rlwinm	r11,r3,7,1,24		/* put VSID in 0x7fffff80 bits */
-	rlwimi	r11,r4,10,26,31		/* put in API (abbrev page index) */
-	SET_V(r11)			/* set V (valid) bit */
-
+2:
 #ifdef CONFIG_SMP
+	li	r11,0
 	addis	r9,r7,mmu_hash_lock@ha
 	addi	r9,r9,mmu_hash_lock@l
 	rlwinm	r8,r1,0,0,(31-THREAD_SHIFT)
@@ -557,10 +552,36 @@ _GLOBAL(flush_hash_pages)
 11:	lwz	r0,0(r9)
 	cmpi	0,r0,0
 	beq	10b
+	mtmsr	r10
+	SYNC_601
+	isync
+	li	r11,1
+	rlwinm	r0,r10,0,17,15		/* clear bit 16 (MSR_EE) */
+	rlwinm	r0,r0,0,28,26		/* clear MSR_DR */
+	mtmsr	r0
+	SYNC_601
+	isync
 	b	11b
 12:	isync
+	cmpwi	r11,0
+	beq	13f
+	li	r0,0
+	stw	r0,0(r9)		/* clear mmu_hash_lock */
+	b	1b
+13:
 #endif
 
+	/* Convert context and va to VSID */
+	mulli	r3,r3,897*16		/* multiply context by context skew */
+	rlwinm	r0,r4,4,28,31		/* get ESID (top 4 bits of va) */
+	mulli	r0,r0,0x111		/* multiply by ESID skew */
+	add	r3,r3,r0		/* note code below trims to 24 bits */
+
+	/* Construct the high word of the PPC-style PTE (r11) */
+	rlwinm	r11,r3,7,1,24		/* put VSID in 0x7fffff80 bits */
+	rlwimi	r11,r4,10,26,31		/* put in API (abbrev page index) */
+	SET_V(r11)			/* set V (valid) bit */
+
 	/*
 	 * Check the _PAGE_HASHPTE bit in the linux PTE.  If it is
 	 * already clear, we're done (for this pte).  If not,
@@ -631,7 +652,7 @@ _GLOBAL(flush_hash_patch_B)
 
 19:	mtmsr	r10
 	SYNC_601
-	isync
+	sync
 	blr
 
 /*
diff --git a/arch/powerpc/mm/hash_native_64.c b/arch/powerpc/mm/hash_native_64.c
index c22caff..ebeb85e 100644
--- a/arch/powerpc/mm/hash_native_64.c
+++ b/arch/powerpc/mm/hash_native_64.c
@@ -37,9 +37,9 @@
 #define HPTE_LOCK_BIT 3
 
 #ifdef CONFIG_PPC_PASEMI_A2_WORKAROUNDS
-DEFINE_SPINLOCK(native_tlbie_lock);
+IPIPE_DEFINE_SPINLOCK(native_tlbie_lock);
 #else
-static DEFINE_SPINLOCK(native_tlbie_lock);
+static IPIPE_DEFINE_SPINLOCK(native_tlbie_lock);
 #endif
 
 static inline void __tlbie(unsigned long va, int psize, int ssize)
@@ -137,7 +137,7 @@ static long native_hpte_insert(unsigned long hpte_group, unsigned long va,
 			unsigned long vflags, int psize, int ssize)
 {
 	struct hash_pte *hptep = htab_address + hpte_group;
-	unsigned long hpte_v, hpte_r;
+	unsigned long hpte_v, hpte_r, flags;
 	int i;
 
 	if (!(vflags & HPTE_V_BOLTED)) {
@@ -146,6 +146,8 @@ static long native_hpte_insert(unsigned long hpte_group, unsigned long va,
 			hpte_group, va, pa, rflags, vflags, psize);
 	}
 
+	local_irq_save_hw(flags);
+
 	for (i = 0; i < HPTES_PER_GROUP; i++) {
 		if (! (hptep->v & HPTE_V_VALID)) {
 			/* retry with lock held */
@@ -158,8 +160,10 @@ static long native_hpte_insert(unsigned long hpte_group, unsigned long va,
 		hptep++;
 	}
 
-	if (i == HPTES_PER_GROUP)
+	if (i == HPTES_PER_GROUP) {
+		local_irq_restore_hw(flags);
 		return -1;
+	}
 
 #ifdef CONFIG_PPC_PASEMI_A2_WORKAROUNDS
 	/* Workaround for bug 4910: No non-guarded access over IOB */
@@ -184,6 +188,8 @@ static long native_hpte_insert(unsigned long hpte_group, unsigned long va,
 	 */
 	hptep->v = hpte_v;
 
+	local_irq_restore_hw(flags);
+
 	__asm__ __volatile__ ("ptesync" : : : "memory");
 
 	return i | (!!(vflags & HPTE_V_SECONDARY) << 3);
@@ -194,13 +200,15 @@ static long native_hpte_remove(unsigned long hpte_group)
 	struct hash_pte *hptep;
 	int i;
 	int slot_offset;
-	unsigned long hpte_v;
+	unsigned long hpte_v, flags;
 
 	DBG_LOW("    remove(group=%lx)\n", hpte_group);
 
 	/* pick a random entry to start at */
 	slot_offset = mftb() & 0x7;
 
+	local_irq_save_hw(flags);
+
 	for (i = 0; i < HPTES_PER_GROUP; i++) {
 		hptep = htab_address + hpte_group + slot_offset;
 		hpte_v = hptep->v;
@@ -219,12 +227,16 @@ static long native_hpte_remove(unsigned long hpte_group)
 		slot_offset &= 0x7;
 	}
 
-	if (i == HPTES_PER_GROUP)
+	if (i == HPTES_PER_GROUP) {
+		local_irq_restore_hw(flags);
 		return -1;
+	}
 
 	/* Invalidate the hpte. NOTE: this also unlocks it */
 	hptep->v = 0;
 
+	local_irq_restore_hw(flags);
+
 	return i;
 }
 
@@ -233,7 +245,7 @@ static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
 				 int local)
 {
 	struct hash_pte *hptep = htab_address + slot;
-	unsigned long hpte_v, want_v;
+	unsigned long hpte_v, want_v, flags;
 	int ret = 0;
 
 	want_v = hpte_encode_v(va, psize, ssize);
@@ -241,6 +253,8 @@ static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
 	DBG_LOW("    update(va=%016lx, avpnv=%016lx, hash=%016lx, newpp=%x)",
 		va, want_v & HPTE_V_AVPN, slot, newpp);
 
+	local_irq_save_hw(flags);
+
 	native_lock_hpte(hptep);
 
 	hpte_v = hptep->v;
@@ -257,6 +271,8 @@ static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
 	}
 	native_unlock_hpte(hptep);
 
+	local_irq_restore_hw(flags);
+
 	/* Ensure it is out of the tlb too. */
 	tlbie(va, psize, ssize, local);
 
@@ -327,10 +343,10 @@ static void native_hpte_invalidate(unsigned long slot, unsigned long va,
 	unsigned long want_v;
 	unsigned long flags;
 
-	local_irq_save(flags);
-
 	DBG_LOW("    invalidate(va=%016lx, hash: %x)\n", va, slot);
 
+	local_irq_save(flags);
+
 	want_v = hpte_encode_v(va, psize, ssize);
 	native_lock_hpte(hptep);
 	hpte_v = hptep->v;
diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
index 8d5b475..976d988 100644
--- a/arch/powerpc/mm/hash_utils_64.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -111,7 +111,7 @@ int mmu_ci_restrictions;
 #ifdef CONFIG_DEBUG_PAGEALLOC
 static u8 *linear_map_hash_slots;
 static unsigned long linear_map_hash_count;
-static DEFINE_SPINLOCK(linear_map_hash_lock);
+static IPIPE_DEFINE_SPINLOCK(linear_map_hash_lock);
 #endif /* CONFIG_DEBUG_PAGEALLOC */
 
 /* There are definitions of page sizes arrays to be used when none
@@ -861,6 +861,7 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 	cpumask_t tmp;
 	int rc, user_region = 0, local = 0;
 	int psize, ssize;
+	unsigned long flags;
 
 	DBG_LOW("hash_page(ea=%016lx, access=%lx, trap=%lx\n",
 		ea, access, trap);
@@ -979,6 +980,9 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 #endif
 		}
 	}
+
+	local_irq_save_hw(flags);
+
 	if (user_region) {
 		if (psize != get_paca_psize(ea)) {
 			get_paca()->context = mm->context;
@@ -990,6 +994,10 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 			mmu_psize_defs[mmu_vmalloc_psize].sllp;
 		slb_vmalloc_update();
 	}
+
+	local_irq_restore_hw(flags);
+#else
+	(void)flags;
 #endif /* CONFIG_PPC_64K_PAGES */
 
 #ifdef CONFIG_PPC_HAS_HASH_64K
@@ -1124,6 +1132,10 @@ void flush_hash_range(unsigned long number, int local)
  */
 void low_hash_fault(struct pt_regs *regs, unsigned long address, int rc)
 {
+	if (ipipe_trap_notify(IPIPE_TRAP_ACCESS, regs))
+		/* Not all access faults go through do_page_fault(). */
+	    	return;
+
 	if (user_mode(regs)) {
 #ifdef CONFIG_PPC_SUBPAGE_PROT
 		if (rc == -2)
diff --git a/arch/powerpc/mm/mmu_context_nohash.c b/arch/powerpc/mm/mmu_context_nohash.c
index 52a0cfc..9c341bb 100644
--- a/arch/powerpc/mm/mmu_context_nohash.c
+++ b/arch/powerpc/mm/mmu_context_nohash.c
@@ -46,7 +46,7 @@ static unsigned int next_context, nr_free_contexts;
 static unsigned long *context_map;
 static unsigned long *stale_map[NR_CPUS];
 static struct mm_struct **context_mm;
-static spinlock_t context_lock = SPIN_LOCK_UNLOCKED;
+static IPIPE_DEFINE_SPINLOCK(context_lock);
 
 #define CTX_MAP_SIZE	\
 	(sizeof(unsigned long) * (last_context / BITS_PER_LONG + 1))
@@ -91,7 +91,7 @@ static unsigned int steal_context_smp(unsigned int id)
 			continue;
 		}
 		pr_debug("[%d] steal context %d from mm @%p\n",
-			 smp_processor_id(), id, mm);
+			 ipipe_processor_id(), id, mm);
 
 		/* Mark this mm has having no context anymore */
 		mm->context.id = MMU_NO_CONTEXT;
@@ -120,7 +120,7 @@ static unsigned int steal_context_smp(unsigned int id)
 static unsigned int steal_context_up(unsigned int id)
 {
 	struct mm_struct *mm;
-	int cpu = smp_processor_id();
+	int cpu = ipipe_processor_id();
 
 	/* Pick up the victim mm */
 	mm = context_mm[id];
@@ -172,9 +172,10 @@ static void context_check_map(void) { }
 
 void switch_mmu_context(struct mm_struct *prev, struct mm_struct *next)
 {
-	unsigned int id, cpu = smp_processor_id();
-	unsigned long *map;
+	unsigned int id, cpu = ipipe_processor_id();
+	unsigned long *map, flags;
 
+	local_irq_save_hw_cond(flags);
 	/* No lockless fast path .. yet */
 	spin_lock(&context_lock);
 
@@ -254,6 +255,7 @@ void switch_mmu_context(struct mm_struct *prev, struct mm_struct *next)
 	/* Flick the MMU and release lock */
 	set_context(id, next->pgd);
 	spin_unlock(&context_lock);
+	local_irq_restore_hw_cond(flags);
 }
 
 /*
diff --git a/arch/powerpc/mm/slb.c b/arch/powerpc/mm/slb.c
index 267b2b0..b3399c7 100644
--- a/arch/powerpc/mm/slb.c
+++ b/arch/powerpc/mm/slb.c
@@ -105,8 +105,11 @@ void slb_flush_and_rebolt(void)
 	 * appropriately too. */
 	unsigned long linear_llp, vmalloc_llp, lflags, vflags;
 	unsigned long ksp_esid_data, ksp_vsid_data;
+	unsigned long flags;
 
+#ifndef CONFIG_IPIPE
 	WARN_ON(!irqs_disabled());
+#endif
 
 	linear_llp = mmu_psize_defs[mmu_linear_psize].sllp;
 	vmalloc_llp = mmu_psize_defs[mmu_vmalloc_psize].sllp;
@@ -128,7 +131,12 @@ void slb_flush_and_rebolt(void)
 	 * We can't take a PMU exception in the following code, so hard
 	 * disable interrupts.
 	 */
+#ifdef CONFIG_IPIPE
+	local_irq_save_hw(flags);
+#else
 	hard_irq_disable();
+	(void)flags;
+#endif
 
 	/* We need to do this all in asm, so we're sure we don't touch
 	 * the stack between the slbia and rebolting it. */
@@ -154,6 +162,9 @@ void slb_flush_and_rebolt(void)
 		        "r"(ksp_vsid_data),
 		        "r"(ksp_esid_data)
 		     : "memory");
+#ifdef CONFIG_IPIPE
+	local_irq_restore_hw(flags);
+#endif
 }
 
 void slb_vmalloc_update(void)
diff --git a/arch/powerpc/mm/tlb_hash32.c b/arch/powerpc/mm/tlb_hash32.c
index 6519058..d7bef4b 100644
--- a/arch/powerpc/mm/tlb_hash32.c
+++ b/arch/powerpc/mm/tlb_hash32.c
@@ -97,6 +97,37 @@ void tlb_flush(struct mmu_gather *tlb)
 #define FINISH_FLUSH	do { } while (0)
 #endif
 
+#ifdef CONFIG_IPIPE
+
+int __flush_hash_pages(unsigned context, unsigned long va,
+		       unsigned long pmdval, int count);
+
+int flush_hash_pages(unsigned context, unsigned long va,
+		     unsigned long pmdval, int count)
+{
+	int bulk, ret = 0;
+	/*
+	 * Submitting flush requests on insanely large PTE counts
+	 * (e.g. HIGHMEM) may cause severe latency penalty on high
+	 * priority domains since this must be done with hw interrupts
+	 * off (typically, peaks over 400 us have been observed on
+	 * 864xD). We split flush requests in bulks of 64 PTEs to
+	 * prevent that; the modified assembly helper which performs
+	 * the actual flush (__flush_hash_pages()) will spin on the
+	 * mmu_lock with interrupts enabled to further reduce latency.
+	 */
+	while (count > 0) {
+		bulk = count > 64 ? 64 : count;
+		ret |= __flush_hash_pages(context, va, pmdval, bulk);
+		va += (bulk << PAGE_SHIFT);
+		count -= bulk;
+	}
+
+	return ret;
+}
+
+#endif /* CONFIG_IPIPE */
+
 static void flush_range(struct mm_struct *mm, unsigned long start,
 			unsigned long end)
 {
diff --git a/arch/powerpc/platforms/82xx/pq2ads-pci-pic.c b/arch/powerpc/platforms/82xx/pq2ads-pci-pic.c
index ddf0bdc..1bbf650 100644
--- a/arch/powerpc/platforms/82xx/pq2ads-pci-pic.c
+++ b/arch/powerpc/platforms/82xx/pq2ads-pci-pic.c
@@ -24,7 +24,7 @@
 
 #include "pq2.h"
 
-static DEFINE_SPINLOCK(pci_pic_lock);
+static IPIPE_DEFINE_SPINLOCK(pci_pic_lock);
 
 struct pq2ads_pci_pic {
 	struct device_node *node;
@@ -38,18 +38,42 @@ struct pq2ads_pci_pic {
 
 #define NUM_IRQS 32
 
+static inline void __pq2ads_pci_mask_irq(struct pq2ads_pci_pic *priv,
+					 unsigned int irq)
+{
+	setbits32(&priv->regs->mask, 1 << irq);
+	mb();
+}
+
+static inline void __pq2ads_pci_unmask_irq(struct pq2ads_pci_pic *priv,
+					   unsigned int irq)
+{
+	clrbits32(&priv->regs->mask, 1 << irq);
+}
+
 static void pq2ads_pci_mask_irq(unsigned int virq)
 {
 	struct pq2ads_pci_pic *priv = get_irq_chip_data(virq);
 	int irq = NUM_IRQS - virq_to_hw(virq) - 1;
+	unsigned long flags;
 
 	if (irq != -1) {
-		unsigned long flags;
 		spin_lock_irqsave(&pci_pic_lock, flags);
+		__pq2ads_pci_mask_irq(priv, irq);
+		ipipe_irq_lock(virq);
+		spin_unlock_irqrestore(&pci_pic_lock, flags);
+	}
+}
 
-		setbits32(&priv->regs->mask, 1 << irq);
-		mb();
+static void pq2ads_pci_mask_ack_irq(unsigned int virq)
+{
+	struct pq2ads_pci_pic *priv = get_irq_chip_data(virq);
+	int irq = NUM_IRQS - virq_to_hw(virq) - 1;
 
+	if (irq != -1) {
+		unsigned long flags;
+		spin_lock_irqsave(&pci_pic_lock, flags);
+		__pq2ads_pci_mask_irq(priv, irq);
 		spin_unlock_irqrestore(&pci_pic_lock, flags);
 	}
 }
@@ -58,12 +82,12 @@ static void pq2ads_pci_unmask_irq(unsigned int virq)
 {
 	struct pq2ads_pci_pic *priv = get_irq_chip_data(virq);
 	int irq = NUM_IRQS - virq_to_hw(virq) - 1;
+	unsigned long flags;
 
 	if (irq != -1) {
-		unsigned long flags;
-
 		spin_lock_irqsave(&pci_pic_lock, flags);
-		clrbits32(&priv->regs->mask, 1 << irq);
+		__pq2ads_pci_unmask_irq(priv, irq);
+		ipipe_irq_unlock(virq);
 		spin_unlock_irqrestore(&pci_pic_lock, flags);
 	}
 }
@@ -73,7 +97,7 @@ static struct irq_chip pq2ads_pci_ic = {
 	.name = "PQ2 ADS PCI",
 	.end = pq2ads_pci_unmask_irq,
 	.mask = pq2ads_pci_mask_irq,
-	.mask_ack = pq2ads_pci_mask_irq,
+	.mask_ack = pq2ads_pci_mask_ack_irq,
 	.ack = pq2ads_pci_mask_irq,
 	.unmask = pq2ads_pci_unmask_irq,
 	.enable = pq2ads_pci_unmask_irq,
@@ -98,7 +122,7 @@ static void pq2ads_pci_irq_demux(unsigned int irq, struct irq_desc *desc)
 		for (bit = 0; pend != 0; ++bit, pend <<= 1) {
 			if (pend & 0x80000000) {
 				int virq = irq_linear_revmap(priv->host, bit);
-				generic_handle_irq(virq);
+				ipipe_handle_chained_irq(virq);
 			}
 		}
 	}
diff --git a/arch/powerpc/platforms/85xx/tqm85xx.c b/arch/powerpc/platforms/85xx/tqm85xx.c
index 2933a8e..efbbde4 100644
--- a/arch/powerpc/platforms/85xx/tqm85xx.c
+++ b/arch/powerpc/platforms/85xx/tqm85xx.c
@@ -46,10 +46,10 @@ static void cpm2_cascade(unsigned int irq, struct irq_desc *desc)
 {
 	int cascade_irq;
 
-	while ((cascade_irq = cpm2_get_irq()) >= 0)
-		generic_handle_irq(cascade_irq);
-
 	desc->chip->eoi(irq);
+
+	while ((cascade_irq = cpm2_get_irq()) >= 0)
+		ipipe_handle_chained_irq(cascade_irq);
 }
 #endif /* CONFIG_CPM2 */
 
diff --git a/arch/powerpc/platforms/cell/spu_base.c b/arch/powerpc/platforms/cell/spu_base.c
index e487ad6..a681d87 100644
--- a/arch/powerpc/platforms/cell/spu_base.c
+++ b/arch/powerpc/platforms/cell/spu_base.c
@@ -57,7 +57,7 @@ EXPORT_SYMBOL_GPL(force_sig_info);
 /*
  * Protects cbe_spu_info and spu->number.
  */
-static DEFINE_SPINLOCK(spu_lock);
+static IPIPE_DEFINE_SPINLOCK(spu_lock);
 
 /*
  * List of all spus in the system.
diff --git a/arch/powerpc/platforms/iseries/irq.c b/arch/powerpc/platforms/iseries/irq.c
index 701d929..331c422 100644
--- a/arch/powerpc/platforms/iseries/irq.c
+++ b/arch/powerpc/platforms/iseries/irq.c
@@ -80,7 +80,7 @@ struct pci_event {
 	} data;
 };
 
-static DEFINE_SPINLOCK(pending_irqs_lock);
+static IPIPE_DEFINE_SPINLOCK(pending_irqs_lock);
 static int num_pending_irqs;
 static int pending_irqs[NR_IRQS];
 
diff --git a/arch/powerpc/platforms/powermac/pic.c b/arch/powerpc/platforms/powermac/pic.c
index 6d149ae..c46a3f7 100644
--- a/arch/powerpc/platforms/powermac/pic.c
+++ b/arch/powerpc/platforms/powermac/pic.c
@@ -57,7 +57,7 @@ static int max_irqs;
 static int max_real_irqs;
 static u32 level_mask[4];
 
-static DEFINE_SPINLOCK(pmac_pic_lock);
+static IPIPE_DEFINE_SPINLOCK(pmac_pic_lock);
 
 #define NR_MASK_WORDS	((NR_IRQS + 31) / 32)
 static unsigned long ppc_lost_interrupts[NR_MASK_WORDS];
diff --git a/arch/powerpc/platforms/ps3/htab.c b/arch/powerpc/platforms/ps3/htab.c
index 1e8a1e3..9e46980 100644
--- a/arch/powerpc/platforms/ps3/htab.c
+++ b/arch/powerpc/platforms/ps3/htab.c
@@ -41,7 +41,7 @@ enum ps3_lpar_vas_id {
 };
 
 
-static DEFINE_SPINLOCK(ps3_htab_lock);
+static IPIPE_DEFINE_SPINLOCK(ps3_htab_lock);
 
 static long ps3_hpte_insert(unsigned long hpte_group, unsigned long va,
 	unsigned long pa, unsigned long rflags, unsigned long vflags,
diff --git a/arch/powerpc/platforms/ps3/interrupt.c b/arch/powerpc/platforms/ps3/interrupt.c
index 8ec5ccf..13a5755 100644
--- a/arch/powerpc/platforms/ps3/interrupt.c
+++ b/arch/powerpc/platforms/ps3/interrupt.c
@@ -74,7 +74,7 @@ struct ps3_bmp {
 		u64 unused_2[3];
 	};
 	u64 ipi_debug_brk_mask;
-	spinlock_t lock;
+	ipipe_spinlock_t lock;
 };
 
 /**
diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c
index 52a80e5..1145792 100644
--- a/arch/powerpc/platforms/pseries/lpar.c
+++ b/arch/powerpc/platforms/pseries/lpar.c
@@ -334,7 +334,7 @@ static long pSeries_lpar_hpte_insert(unsigned long hpte_group,
 	return (slot & 7) | (!!(vflags & HPTE_V_SECONDARY) << 3);
 }
 
-static DEFINE_SPINLOCK(pSeries_lpar_tlbie_lock);
+static IPIPE_DEFINE_SPINLOCK(pSeries_lpar_tlbie_lock);
 
 static long pSeries_lpar_hpte_remove(unsigned long hpte_group)
 {
diff --git a/arch/powerpc/sysdev/cpm2_pic.c b/arch/powerpc/sysdev/cpm2_pic.c
index 78f1f7c..91b47fd 100644
--- a/arch/powerpc/sysdev/cpm2_pic.c
+++ b/arch/powerpc/sysdev/cpm2_pic.c
@@ -82,44 +82,61 @@ static void cpm2_mask_irq(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = virq_to_hw(virq);
+	unsigned long flags;
 
 	bit = irq_to_siubit[irq_nr];
 	word = irq_to_siureg[irq_nr];
 
+	local_irq_save_hw_cond(flags);
+	ipipe_irq_lock(virq);
 	ppc_cached_irq_mask[word] &= ~(1 << bit);
 	out_be32(&cpm2_intctl->ic_simrh + word, ppc_cached_irq_mask[word]);
+	local_irq_restore_hw_cond(flags);
 }
 
 static void cpm2_unmask_irq(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = virq_to_hw(virq);
+	unsigned long flags;
 
 	bit = irq_to_siubit[irq_nr];
 	word = irq_to_siureg[irq_nr];
 
+	local_irq_save_hw_cond(flags);
 	ppc_cached_irq_mask[word] |= 1 << bit;
 	out_be32(&cpm2_intctl->ic_simrh + word, ppc_cached_irq_mask[word]);
+	ipipe_irq_unlock(virq);
+	local_irq_restore_hw_cond(flags);
 }
 
-static void cpm2_ack(unsigned int virq)
+static void cpm2_mask_ack(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = virq_to_hw(virq);
+	unsigned long flags;
 
 	bit = irq_to_siubit[irq_nr];
 	word = irq_to_siureg[irq_nr];
 
+	local_irq_save_hw_cond(flags);
+	ppc_cached_irq_mask[word] &= ~(1 << bit);
+	out_be32(&cpm2_intctl->ic_simrh + word, ppc_cached_irq_mask[word]);
 	out_be32(&cpm2_intctl->ic_sipnrh + word, 1 << bit);
+	local_irq_restore_hw_cond(flags);
 }
 
 static void cpm2_end_irq(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = virq_to_hw(virq);
+	unsigned long flags;
 
-	if (!(irq_desc[irq_nr].status & (IRQ_DISABLED|IRQ_INPROGRESS))
-			&& irq_desc[irq_nr].action) {
+	local_irq_save_hw_cond(flags);
+	
+	if (!__ipipe_root_domain_p ||
+	    (!(irq_desc[irq_nr].status & (IRQ_DISABLED|IRQ_INPROGRESS))
+	     && irq_desc[irq_nr].action)) {
 
 		bit = irq_to_siubit[irq_nr];
 		word = irq_to_siureg[irq_nr];
@@ -133,6 +150,8 @@ static void cpm2_end_irq(unsigned int virq)
 		 */
 		mb();
 	}
+
+	local_irq_restore_hw_cond(flags);
 }
 
 static int cpm2_set_irq_type(unsigned int virq, unsigned int flow_type)
@@ -185,7 +204,7 @@ static struct irq_chip cpm2_pic = {
 	.typename = " CPM2 SIU ",
 	.mask = cpm2_mask_irq,
 	.unmask = cpm2_unmask_irq,
-	.ack = cpm2_ack,
+	.mask_ack = cpm2_mask_ack,
 	.eoi = cpm2_end_irq,
 	.set_type = cpm2_set_irq_type,
 };
diff --git a/arch/powerpc/sysdev/i8259.c b/arch/powerpc/sysdev/i8259.c
index a96584a..c037c88 100644
--- a/arch/powerpc/sysdev/i8259.c
+++ b/arch/powerpc/sysdev/i8259.c
@@ -23,7 +23,7 @@ static unsigned char cached_8259[2] = { 0xff, 0xff };
 #define cached_A1 (cached_8259[0])
 #define cached_21 (cached_8259[1])
 
-static DEFINE_SPINLOCK(i8259_lock);
+static IPIPE_DEFINE_SPINLOCK(i8259_lock);
 
 static struct irq_host *i8259_host;
 
diff --git a/arch/powerpc/sysdev/ipic.c b/arch/powerpc/sysdev/ipic.c
index 9a89cd3..8f61414 100644
--- a/arch/powerpc/sysdev/ipic.c
+++ b/arch/powerpc/sysdev/ipic.c
@@ -32,7 +32,7 @@
 
 static struct ipic * primary_ipic;
 static struct irq_chip ipic_level_irq_chip, ipic_edge_irq_chip;
-static DEFINE_SPINLOCK(ipic_lock);
+static IPIPE_DEFINE_SPINLOCK(ipic_lock);
 
 static struct ipic_info ipic_info[] = {
 	[1] = {
diff --git a/arch/powerpc/sysdev/mpc8xx_pic.c b/arch/powerpc/sysdev/mpc8xx_pic.c
index 5d2d552..5903a16 100644
--- a/arch/powerpc/sysdev/mpc8xx_pic.c
+++ b/arch/powerpc/sysdev/mpc8xx_pic.c
@@ -29,24 +29,30 @@ static void mpc8xx_unmask_irq(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = (unsigned int)irq_map[virq].hwirq;
+	unsigned long flags;
 
 	bit = irq_nr & 0x1f;
 	word = irq_nr >> 5;
 
+	local_irq_save_hw_cond(flags);
 	ppc_cached_irq_mask[word] |= (1 << (31-bit));
 	out_be32(&siu_reg->sc_simask, ppc_cached_irq_mask[word]);
+	local_irq_restore_hw_cond(flags);
 }
 
 static void mpc8xx_mask_irq(unsigned int virq)
 {
 	int	bit, word;
 	unsigned int irq_nr = (unsigned int)irq_map[virq].hwirq;
+	unsigned long flags;
 
 	bit = irq_nr & 0x1f;
 	word = irq_nr >> 5;
 
+	local_irq_save_hw_cond(flags);
 	ppc_cached_irq_mask[word] &= ~(1 << (31-bit));
 	out_be32(&siu_reg->sc_simask, ppc_cached_irq_mask[word]);
+	local_irq_save_hw_cond(flags);
 }
 
 static void mpc8xx_ack(unsigned int virq)
diff --git a/arch/powerpc/sysdev/mpic.c b/arch/powerpc/sysdev/mpic.c
index a35297d..778feea 100644
--- a/arch/powerpc/sysdev/mpic.c
+++ b/arch/powerpc/sysdev/mpic.c
@@ -46,7 +46,7 @@
 
 static struct mpic *mpics;
 static struct mpic *mpic_primary;
-static DEFINE_SPINLOCK(mpic_lock);
+static IPIPE_DEFINE_SPINLOCK(mpic_lock);
 
 #ifdef CONFIG_PPC32	/* XXX for now */
 #ifdef CONFIG_IRQ_ALL_CPUS
@@ -666,33 +666,44 @@ static inline void mpic_eoi(struct mpic *mpic)
  */
 
 
-void mpic_unmask_irq(unsigned int irq)
+void __mpic_unmask_irq(unsigned int irq)
 {
 	unsigned int loops = 100000;
 	struct mpic *mpic = mpic_from_irq(irq);
 	unsigned int src = mpic_irq_to_hw(irq);
 
-	DBG("%p: %s: enable_irq: %d (src %d)\n", mpic, mpic->name, irq, src);
-
 	mpic_irq_write(src, MPIC_INFO(IRQ_VECTOR_PRI),
 		       mpic_irq_read(src, MPIC_INFO(IRQ_VECTOR_PRI)) &
 		       ~MPIC_VECPRI_MASK);
 	/* make sure mask gets to controller before we return to user */
 	do {
 		if (!loops--) {
-			printk(KERN_ERR "mpic_enable_irq timeout\n");
+			printk(KERN_ERR "mpic_unmask_irq timeout\n");
 			break;
 		}
 	} while(mpic_irq_read(src, MPIC_INFO(IRQ_VECTOR_PRI)) & MPIC_VECPRI_MASK);
 }
 
-void mpic_mask_irq(unsigned int irq)
+void mpic_unmask_irq(unsigned int irq)
 {
-	unsigned int loops = 100000;
+#ifdef DEBUG
 	struct mpic *mpic = mpic_from_irq(irq);
-	unsigned int src = mpic_irq_to_hw(irq);
+#endif
+	unsigned long flags;
 
-	DBG("%s: disable_irq: %d (src %d)\n", mpic->name, irq, src);
+	DBG("%p: %s: unmask_irq: %d (src %d)\n", mpic, mpic->name, irq, src);
+
+	spin_lock_irqsave(&mpic_lock, flags);
+	__mpic_unmask_irq(irq);
+	ipipe_irq_unlock(irq);
+	spin_unlock_irqrestore(&mpic_lock, flags);
+}
+
+static inline void __mpic_mask_irq(unsigned int irq)
+{
+	struct mpic *mpic = mpic_from_irq(irq);
+	unsigned int src = mpic_irq_to_hw(irq);
+	unsigned int loops = 100000;
 
 	mpic_irq_write(src, MPIC_INFO(IRQ_VECTOR_PRI),
 		       mpic_irq_read(src, MPIC_INFO(IRQ_VECTOR_PRI)) |
@@ -701,15 +712,31 @@ void mpic_mask_irq(unsigned int irq)
 	/* make sure mask gets to controller before we return to user */
 	do {
 		if (!loops--) {
-			printk(KERN_ERR "mpic_enable_irq timeout\n");
+			printk(KERN_ERR "mpic_mask_irq timeout, irq %u\n", irq);
 			break;
 		}
 	} while(!(mpic_irq_read(src, MPIC_INFO(IRQ_VECTOR_PRI)) & MPIC_VECPRI_MASK));
 }
 
+void mpic_mask_irq(unsigned int irq)
+{
+#ifdef DEBUG
+	struct mpic *mpic = mpic_from_irq(irq);
+#endif
+	unsigned long flags;
+
+	DBG("%s: mask_irq: irq %u (src %d)\n", mpic->name, irq, mpic_irq_to_hw(irq));
+
+	spin_lock_irqsave(&mpic_lock, flags);
+	__mpic_mask_irq(irq);
+	ipipe_irq_lock(irq);
+	spin_unlock_irqrestore(&mpic_lock, flags);
+}
+
 void mpic_end_irq(unsigned int irq)
 {
 	struct mpic *mpic = mpic_from_irq(irq);
+	unsigned long flags;
 
 #ifdef DEBUG_IRQ
 	DBG("%s: end_irq: %d\n", mpic->name, irq);
@@ -719,6 +746,14 @@ void mpic_end_irq(unsigned int irq)
 	 * latched another edge interrupt coming in anyway
 	 */
 
+#ifdef CONFIG_IPIPE
+	spin_lock_irqsave(&mpic_lock, flags);
+	if (!(irq_desc[irq].status & IRQ_NOREQUEST))
+		__mpic_mask_irq(irq);
+	spin_unlock_irqrestore(&mpic_lock, flags);
+#else
+	(void)flags;
+#endif
 	mpic_eoi(mpic);
 }
 
@@ -728,8 +763,11 @@ static void mpic_unmask_ht_irq(unsigned int irq)
 {
 	struct mpic *mpic = mpic_from_irq(irq);
 	unsigned int src = mpic_irq_to_hw(irq);
+	unsigned long flags;
 
-	mpic_unmask_irq(irq);
+	spin_lock_irqsave(&mpic_lock, flags);
+	__mpic_unmask_irq(irq);
+	spin_unlock_irqrestore(&mpic_lock, flags);
 
 	if (irq_desc[irq].status & IRQ_LEVEL)
 		mpic_ht_end_irq(mpic, src);
@@ -759,9 +797,18 @@ static void mpic_end_ht_irq(unsigned int irq)
 {
 	struct mpic *mpic = mpic_from_irq(irq);
 	unsigned int src = mpic_irq_to_hw(irq);
+	unsigned long flags;
 
 #ifdef DEBUG_IRQ
-	DBG("%s: end_irq: %d\n", mpic->name, irq);
+	DBG("%s: end_ht_irq: %d\n", mpic->name, irq);
+#endif
+
+#ifdef CONFIG_IPIPE
+	spin_lock_irqsave(&mpic_lock, flags);
+	__mpic_mask_irq(irq);
+	spin_unlock_irqrestore(&mpic_lock, flags);
+#else
+	(void)flags;
 #endif
 	/* We always EOI on end_irq() even for edge interrupts since that
 	 * should only lower the priority, the MPIC should have properly
@@ -780,9 +827,12 @@ static void mpic_unmask_ipi(unsigned int irq)
 {
 	struct mpic *mpic = mpic_from_ipi(irq);
 	unsigned int src = mpic_irq_to_hw(irq) - mpic->ipi_vecs[0];
+	unsigned long flags;
 
-	DBG("%s: enable_ipi: %d (ipi %d)\n", mpic->name, irq, src);
+	DBG("%s: unmask_ipi: %d (ipi %d)\n", mpic->name, irq, src);
+	spin_lock_irqsave(&mpic_lock, flags);
 	mpic_ipi_write(src, mpic_ipi_read(src) & ~MPIC_VECPRI_MASK);
+	spin_unlock_irqrestore(&mpic_lock, flags);
 }
 
 static void mpic_mask_ipi(unsigned int irq)
@@ -852,6 +902,7 @@ int mpic_set_irq_type(unsigned int virq, unsigned int flow_type)
 	unsigned int src = mpic_irq_to_hw(virq);
 	struct irq_desc *desc = get_irq_desc(virq);
 	unsigned int vecpri, vold, vnew;
+	unsigned long flags;
 
 	DBG("mpic: set_irq_type(mpic:@%p,virq:%d,src:0x%x,type:0x%x)\n",
 	    mpic, virq, src, flow_type);
@@ -876,6 +927,8 @@ int mpic_set_irq_type(unsigned int virq, unsigned int flow_type)
 	else
 		vecpri = mpic_type_to_vecpri(mpic, flow_type);
 
+	local_irq_save_hw_cond(flags);
+
 	vold = mpic_irq_read(src, MPIC_INFO(IRQ_VECTOR_PRI));
 	vnew = vold & ~(MPIC_INFO(VECPRI_POLARITY_MASK) |
 			MPIC_INFO(VECPRI_SENSE_MASK));
@@ -883,6 +936,8 @@ int mpic_set_irq_type(unsigned int virq, unsigned int flow_type)
 	if (vold != vnew)
 		mpic_irq_write(src, MPIC_INFO(IRQ_VECTOR_PRI), vnew);
 
+	local_irq_restore_hw_cond(flags);
+
 	return 0;
 }
 
@@ -1534,6 +1589,7 @@ unsigned int mpic_get_mcirq(void)
 }
 
 #ifdef CONFIG_SMP
+
 void mpic_request_ipis(void)
 {
 	struct mpic *mpic = mpic_primary;
diff --git a/arch/powerpc/sysdev/qe_lib/qe_ic.c b/arch/powerpc/sysdev/qe_lib/qe_ic.c
index 63cdf98..e5a35fb 100644
--- a/arch/powerpc/sysdev/qe_lib/qe_ic.c
+++ b/arch/powerpc/sysdev/qe_lib/qe_ic.c
@@ -33,7 +33,7 @@
 
 #include "qe_ic.h"
 
-static DEFINE_SPINLOCK(qe_ic_lock);
+static IPIPE_DEFINE_SPINLOCK(qe_ic_lock);
 
 static struct qe_ic_info qe_ic_info[] = {
 	[1] = {
@@ -236,6 +236,20 @@ static void qe_ic_mask_irq(unsigned int virq)
 	spin_unlock_irqrestore(&qe_ic_lock, flags);
 }
 
+#ifdef CONFIG_IPIPE
+
+void __ipipe_qe_ic_cascade_irq(struct qe_ic *qe_ic, unsigned int virq)
+{
+	
+	struct pt_regs regs;    /* Contents not used. */
+
+	ipipe_trace_irq_entry(virq);
+        __ipipe_handle_irq(virq, &regs);
+        ipipe_trace_irq_exit(virq);
+}
+
+#endif
+
 static struct irq_chip qe_ic_irq_chip = {
 	.typename = " QEIC  ",
 	.unmask = qe_ic_unmask_irq,
diff --git a/arch/powerpc/sysdev/tsi108_pci.c b/arch/powerpc/sysdev/tsi108_pci.c
index 24e1f5a..cb8ffa8 100644
--- a/arch/powerpc/sysdev/tsi108_pci.c
+++ b/arch/powerpc/sysdev/tsi108_pci.c
@@ -250,7 +250,9 @@ static void tsi108_pci_int_mask(u_int irq)
 {
 	u_int irp_cfg;
 	int int_line = (irq - IRQ_PCI_INTAD_BASE);
+	unsigned long flags;
 
+	local_irq_save_hw_cond(flags);
 	irp_cfg = tsi108_read_reg(TSI108_PCI_OFFSET + TSI108_PCI_IRP_CFG_CTL);
 	mb();
 	irp_cfg |= (1 << int_line);	/* INTx_DIR = output */
@@ -258,19 +260,23 @@ static void tsi108_pci_int_mask(u_int irq)
 	tsi108_write_reg(TSI108_PCI_OFFSET + TSI108_PCI_IRP_CFG_CTL, irp_cfg);
 	mb();
 	irp_cfg = tsi108_read_reg(TSI108_PCI_OFFSET + TSI108_PCI_IRP_CFG_CTL);
+	local_irq_restore_hw_cond(flags);
 }
 
 static void tsi108_pci_int_unmask(u_int irq)
 {
 	u_int irp_cfg;
 	int int_line = (irq - IRQ_PCI_INTAD_BASE);
+	unsigned long flags;
 
+	local_irq_save_hw_cond(flags);
 	irp_cfg = tsi108_read_reg(TSI108_PCI_OFFSET + TSI108_PCI_IRP_CFG_CTL);
 	mb();
 	irp_cfg &= ~(1 << int_line);
 	irp_cfg |= (3 << (8 + (int_line * 2)));
 	tsi108_write_reg(TSI108_PCI_OFFSET + TSI108_PCI_IRP_CFG_CTL, irp_cfg);
 	mb();
+	local_irq_restore_hw_cond(flags);
 }
 
 static void init_pci_source(void)
@@ -361,6 +367,9 @@ static void tsi108_pci_irq_ack(u_int irq)
 
 static void tsi108_pci_irq_end(u_int irq)
 {
+	unsigned long flags;
+
+	local_irq_save_hw_cond(flags);
 	tsi108_pci_int_unmask(irq);
 
 	/* Enable interrupts from PCI block */
@@ -368,6 +377,7 @@ static void tsi108_pci_irq_end(u_int irq)
 			 tsi108_read_reg(TSI108_PCI_OFFSET +
 					 TSI108_PCI_IRP_ENABLE) |
 			 TSI108_PCI_IRP_ENABLE_P_INT);
+	local_irq_restore_hw_cond(flags);
 	mb();
 }
 
diff --git a/arch/powerpc/sysdev/uic.c b/arch/powerpc/sysdev/uic.c
index d35405c..f2ffcb4 100644
--- a/arch/powerpc/sysdev/uic.c
+++ b/arch/powerpc/sysdev/uic.c
@@ -49,7 +49,7 @@ struct uic {
 	int index;
 	int dcrbase;
 
-	spinlock_t lock;
+	ipipe_spinlock_t lock;
 
 	/* The remapper for this UIC */
 	struct irq_host	*irqhost;
@@ -71,6 +71,7 @@ static void uic_unmask_irq(unsigned int virq)
 	er = mfdcr(uic->dcrbase + UIC_ER);
 	er |= sr;
 	mtdcr(uic->dcrbase + UIC_ER, er);
+	ipipe_irq_unlock(virq);
 	spin_unlock_irqrestore(&uic->lock, flags);
 }
 
@@ -82,6 +83,7 @@ static void uic_mask_irq(unsigned int virq)
 	u32 er;
 
 	spin_lock_irqsave(&uic->lock, flags);
+	ipipe_irq_lock(virq);
 	er = mfdcr(uic->dcrbase + UIC_ER);
 	er &= ~(1 << (31 - src));
 	mtdcr(uic->dcrbase + UIC_ER, er);
@@ -239,7 +241,16 @@ void uic_irq_cascade(unsigned int virq, struct irq_desc *desc)
 	src = 32 - ffs(msr);
 
 	subvirq = irq_linear_revmap(uic->irqhost, src);
+#ifdef CONFIG_IPIPE
+	{
+		struct pt_regs regs;    /* Contents not used. */
+		ipipe_trace_irq_entry(subvirq);
+		__ipipe_handle_irq(subvirq, &regs);
+		ipipe_trace_irq_exit(subvirq);
+	}
+#else
 	generic_handle_irq(subvirq);
+#endif
 
 uic_irq_ret:
 	spin_lock(&desc->lock);
diff --git a/drivers/pci/htirq.c b/drivers/pci/htirq.c
index bf7d6ce..0ffac4e 100644
--- a/drivers/pci/htirq.c
+++ b/drivers/pci/htirq.c
@@ -21,7 +21,7 @@
  * With multiple simultaneous hypertransport irq devices it might pay
  * to make this more fine grained.  But start with simple, stupid, and correct.
  */
-static DEFINE_SPINLOCK(ht_irq_lock);
+static IPIPE_DEFINE_SPINLOCK(ht_irq_lock);
 
 struct ht_irq_cfg {
 	struct pci_dev *dev;
diff --git a/drivers/serial/8250.c b/drivers/serial/8250.c
index 3f597cb..a7b844c 100644
--- a/drivers/serial/8250.c
+++ b/drivers/serial/8250.c
@@ -3139,6 +3139,34 @@ static int serial8250_resume(struct platform_device *dev)
 	return 0;
 }
 
+#if defined(CONFIG_IPIPE_DEBUG) && defined(CONFIG_SERIAL_8250_CONSOLE)
+
+#include <stdarg.h>
+
+static IPIPE_DEFINE_SPINLOCK(serial_lock);
+
+void __ipipe_serial_debug(const char *fmt, ...)
+{
+        struct uart_8250_port *up = &serial8250_ports[0];
+        unsigned long flags;
+        unsigned int count;
+        char buf[128];
+        va_list ap;
+
+        va_start(ap, fmt);
+        vsnprintf(buf, sizeof(buf), fmt, ap);
+        va_end(ap);
+        count = strlen(buf);
+
+        spin_lock_irqsave(&serial_lock, flags);
+	uart_console_write(&up->port, buf, count, serial8250_console_putchar);
+        spin_unlock_irqrestore(&serial_lock, flags);
+}
+
+EXPORT_SYMBOL_GPL(__ipipe_serial_debug);
+
+#endif
+
 static struct platform_driver serial8250_isa_driver = {
 	.probe		= serial8250_probe,
 	.remove		= __devexit_p(serial8250_remove),
diff --git a/include/asm-generic/cmpxchg-local.h b/include/asm-generic/cmpxchg-local.h
index b2ba2fc..ed01ab9 100644
--- a/include/asm-generic/cmpxchg-local.h
+++ b/include/asm-generic/cmpxchg-local.h
@@ -20,7 +20,7 @@ static inline unsigned long __cmpxchg_local_generic(volatile void *ptr,
 	if (size == 8 && sizeof(unsigned long) != 8)
 		wrong_size_cmpxchg(ptr);
 
-	local_irq_save(flags);
+	local_irq_save_hw(flags);
 	switch (size) {
 	case 1: prev = *(u8 *)ptr;
 		if (prev == old)
@@ -41,7 +41,7 @@ static inline unsigned long __cmpxchg_local_generic(volatile void *ptr,
 	default:
 		wrong_size_cmpxchg(ptr);
 	}
-	local_irq_restore(flags);
+	local_irq_restore_hw(flags);
 	return prev;
 }
 
@@ -54,11 +54,11 @@ static inline u64 __cmpxchg64_local_generic(volatile void *ptr,
 	u64 prev;
 	unsigned long flags;
 
-	local_irq_save(flags);
+	local_irq_save_hw(flags);
 	prev = *(u64 *)ptr;
 	if (prev == old)
 		*(u64 *)ptr = new;
-	local_irq_restore(flags);
+	local_irq_restore_hw(flags);
 	return prev;
 }
 
diff --git a/include/asm-generic/percpu.h b/include/asm-generic/percpu.h
index b0e63c6..3ecd12f 100644
--- a/include/asm-generic/percpu.h
+++ b/include/asm-generic/percpu.h
@@ -60,6 +60,20 @@ extern unsigned long __per_cpu_offset[NR_CPUS];
 #define __raw_get_cpu_var(var) \
 	(*SHIFT_PERCPU_PTR(&per_cpu_var(var), __my_cpu_offset))
 
+#ifdef CONFIG_IPIPE
+#if defined(CONFIG_IPIPE_DEBUG) && defined(CONFIG_SMP)
+extern int __ipipe_check_percpu_access(void);
+#define __ipipe_local_cpu_offset				\
+	({							\
+		WARN_ON_ONCE(__ipipe_check_percpu_access());	\
+		__my_cpu_offset;				\
+	})
+#else
+#define __ipipe_local_cpu_offset  __my_cpu_offset
+#endif
+#define __ipipe_get_cpu_var(var) \
+	(*SHIFT_PERCPU_PTR(&per_cpu_var(var), __ipipe_local_cpu_offset))
+#endif /* CONFIG_IPIPE */
 
 #ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA
 extern void setup_per_cpu_areas(void);
@@ -70,6 +84,7 @@ extern void setup_per_cpu_areas(void);
 #define per_cpu(var, cpu)			(*((void)(cpu), &per_cpu_var(var)))
 #define __get_cpu_var(var)			per_cpu_var(var)
 #define __raw_get_cpu_var(var)			per_cpu_var(var)
+#define __ipipe_get_cpu_var(var)		__raw_get_cpu_var(var)
 
 #endif	/* SMP */
 
diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
index f832883..02a9391 100644
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -166,18 +166,24 @@ extern void irq_exit(void);
 
 #define nmi_enter()				\
 	do {					\
-		ftrace_nmi_enter();		\
-		lockdep_off();			\
-		rcu_nmi_enter();		\
-		__irq_enter();			\
+ 		ipipe_nmi_enter();		\
+ 		if (__ipipe_root_domain_p) {	\
+			ftrace_nmi_enter();	\
+			lockdep_off();		\
+			rcu_nmi_enter();	\
+			__irq_enter();		\
+		}				\
 	} while (0)
 
 #define nmi_exit()				\
 	do {					\
-		__irq_exit();			\
-		rcu_nmi_exit();			\
-		lockdep_on();			\
-		ftrace_nmi_exit();		\
+ 		if (__ipipe_root_domain_p) {	\
+			__irq_exit();		\
+			rcu_nmi_exit();		\
+			lockdep_on();		\
+			ftrace_nmi_exit();	\
+		}				\
+ 		ipipe_nmi_exit();		\
 	} while (0)
 
 #endif /* LINUX_HARDIRQ_H */
diff --git a/include/linux/ipipe.h b/include/linux/ipipe.h
new file mode 100644
index 0000000..bee1953
--- /dev/null
+++ b/include/linux/ipipe.h
@@ -0,0 +1,678 @@
+/* -*- linux-c -*-
+ * include/linux/ipipe.h
+ *
+ * Copyright (C) 2002-2007 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __LINUX_IPIPE_H
+#define __LINUX_IPIPE_H
+
+#include <linux/spinlock.h>
+#include <linux/cache.h>
+#include <linux/percpu.h>
+#include <linux/mutex.h>
+#include <linux/linkage.h>
+#include <linux/ipipe_base.h>
+#include <linux/ipipe_compat.h>
+#include <asm/ipipe.h>
+
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+
+#include <linux/cpumask.h>
+#include <asm/system.h>
+
+static inline int ipipe_disable_context_check(int cpu)
+{
+	return xchg(&per_cpu(ipipe_percpu_context_check, cpu), 0);
+}
+
+static inline void ipipe_restore_context_check(int cpu, int old_state)
+{
+	per_cpu(ipipe_percpu_context_check, cpu) = old_state;
+}
+
+static inline void ipipe_context_check_off(void)
+{
+	int cpu;
+	for_each_online_cpu(cpu)
+		per_cpu(ipipe_percpu_context_check, cpu) = 0;
+}
+
+#else	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
+
+static inline int ipipe_disable_context_check(int cpu)
+{
+	return 0;
+}
+
+static inline void ipipe_restore_context_check(int cpu, int old_state) { }
+
+static inline void ipipe_context_check_off(void) { }
+
+#endif	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
+
+#ifdef CONFIG_IPIPE
+
+/*
+ * Sanity check: IPIPE_VIRQ_BASE depends on CONFIG_NR_CPUS, and if the
+ * latter gets too large, we fail to map the virtual interrupts.
+ */
+#if IPIPE_VIRQ_BASE / BITS_PER_LONG > BITS_PER_LONG
+#error "CONFIG_NR_CPUS is too large, please lower it."
+#endif
+
+#define IPIPE_VERSION_STRING	IPIPE_ARCH_STRING
+#define IPIPE_RELEASE_NUMBER	((IPIPE_MAJOR_NUMBER << 16) | \
+				 (IPIPE_MINOR_NUMBER <<  8) | \
+				 (IPIPE_PATCH_NUMBER))
+
+#ifndef BROKEN_BUILTIN_RETURN_ADDRESS
+#define __BUILTIN_RETURN_ADDRESS0 ((unsigned long)__builtin_return_address(0))
+#define __BUILTIN_RETURN_ADDRESS1 ((unsigned long)__builtin_return_address(1))
+#endif /* !BUILTIN_RETURN_ADDRESS */
+
+#define IPIPE_ROOT_PRIO		100
+#define IPIPE_ROOT_ID		0
+#define IPIPE_ROOT_NPTDKEYS	4	/* Must be <= BITS_PER_LONG */
+
+#define IPIPE_RESET_TIMER	0x1
+#define IPIPE_GRAB_TIMER	0x2
+
+/* Global domain flags */
+#define IPIPE_SPRINTK_FLAG	0	/* Synchronous printk() allowed */
+#define IPIPE_AHEAD_FLAG	1	/* Domain always heads the pipeline */
+
+/* Interrupt control bits */
+#define IPIPE_HANDLE_FLAG	0
+#define IPIPE_PASS_FLAG		1
+#define IPIPE_ENABLE_FLAG	2
+#define IPIPE_DYNAMIC_FLAG	IPIPE_HANDLE_FLAG
+#define IPIPE_STICKY_FLAG	3
+#define IPIPE_SYSTEM_FLAG	4
+#define IPIPE_LOCK_FLAG		5
+#define IPIPE_WIRED_FLAG	6
+#define IPIPE_EXCLUSIVE_FLAG	7
+
+#define IPIPE_HANDLE_MASK	(1 << IPIPE_HANDLE_FLAG)
+#define IPIPE_PASS_MASK		(1 << IPIPE_PASS_FLAG)
+#define IPIPE_ENABLE_MASK	(1 << IPIPE_ENABLE_FLAG)
+#define IPIPE_DYNAMIC_MASK	IPIPE_HANDLE_MASK
+#define IPIPE_STICKY_MASK	(1 << IPIPE_STICKY_FLAG)
+#define IPIPE_SYSTEM_MASK	(1 << IPIPE_SYSTEM_FLAG)
+#define IPIPE_LOCK_MASK		(1 << IPIPE_LOCK_FLAG)
+#define IPIPE_WIRED_MASK	(1 << IPIPE_WIRED_FLAG)
+#define IPIPE_EXCLUSIVE_MASK	(1 << IPIPE_EXCLUSIVE_FLAG)
+
+#define IPIPE_DEFAULT_MASK	(IPIPE_HANDLE_MASK|IPIPE_PASS_MASK)
+#define IPIPE_STDROOT_MASK	(IPIPE_HANDLE_MASK|IPIPE_PASS_MASK|IPIPE_SYSTEM_MASK)
+
+#define IPIPE_EVENT_SELF        0x80000000
+
+#define IPIPE_NR_CPUS		NR_CPUS
+
+/* This accessor assumes hw IRQs are off on SMP; allows assignment. */
+#define __ipipe_current_domain	__ipipe_get_cpu_var(ipipe_percpu_domain)
+/* This read-only accessor makes sure that hw IRQs are off on SMP. */
+#define ipipe_current_domain				\
+	({						\
+		struct ipipe_domain *__ipd__;		\
+		unsigned long __flags__;		\
+		local_irq_save_hw_smp(__flags__);	\
+		__ipd__ = __ipipe_current_domain;	\
+		local_irq_restore_hw_smp(__flags__);	\
+		__ipd__;				\
+	})
+
+#define ipipe_virtual_irq_p(irq)	((irq) >= IPIPE_VIRQ_BASE && \
+					 (irq) < IPIPE_NR_IRQS)
+
+#define IPIPE_SAME_HANDLER	((ipipe_irq_handler_t)(-1))
+
+struct irq_desc;
+
+typedef void (*ipipe_irq_ackfn_t)(unsigned irq, struct irq_desc *desc);
+
+typedef int (*ipipe_event_handler_t)(unsigned event,
+				     struct ipipe_domain *from,
+				     void *data);
+struct ipipe_domain {
+
+	int slot;			/* Slot number in percpu domain data array. */
+	struct list_head p_link;	/* Link in pipeline */
+	ipipe_event_handler_t evhand[IPIPE_NR_EVENTS]; /* Event handlers. */
+	unsigned long long evself;	/* Self-monitored event bits. */
+
+	struct {
+		unsigned long control;
+		ipipe_irq_ackfn_t acknowledge;
+		ipipe_irq_handler_t handler;
+		void *cookie;
+	} ____cacheline_aligned irqs[IPIPE_NR_IRQS];
+
+	int priority;
+	void *pdd;
+	unsigned long flags;
+	unsigned domid;
+	const char *name;
+	struct mutex mutex;
+};
+
+#define IPIPE_HEAD_PRIORITY	(-1) /* For domains always heading the pipeline */
+
+struct ipipe_domain_attr {
+
+	unsigned domid;		/* Domain identifier -- Magic value set by caller */
+	const char *name;	/* Domain name -- Warning: won't be dup'ed! */
+	int priority;		/* Priority in interrupt pipeline */
+	void (*entry) (void);	/* Domain entry point */
+	void *pdd;		/* Per-domain (opaque) data pointer */
+};
+
+#define __ipipe_irq_cookie(ipd, irq)		(ipd)->irqs[irq].cookie
+#define __ipipe_irq_handler(ipd, irq)		(ipd)->irqs[irq].handler
+#define __ipipe_cpudata_irq_hits(ipd, cpu, irq)	ipipe_percpudom(ipd, irqall, cpu)[irq]
+
+extern unsigned __ipipe_printk_virq;
+
+extern unsigned long __ipipe_virtual_irq_map;
+
+extern struct list_head __ipipe_pipeline;
+
+extern int __ipipe_event_monitors[];
+
+/* Private interface */
+
+void ipipe_init(void);
+
+#ifdef CONFIG_PROC_FS
+void ipipe_init_proc(void);
+
+#ifdef CONFIG_IPIPE_TRACE
+void __ipipe_init_tracer(void);
+#else /* !CONFIG_IPIPE_TRACE */
+#define __ipipe_init_tracer()       do { } while(0)
+#endif /* CONFIG_IPIPE_TRACE */
+
+#else	/* !CONFIG_PROC_FS */
+#define ipipe_init_proc()	do { } while(0)
+#endif	/* CONFIG_PROC_FS */
+
+void __ipipe_init_stage(struct ipipe_domain *ipd);
+
+void __ipipe_cleanup_domain(struct ipipe_domain *ipd);
+
+void __ipipe_add_domain_proc(struct ipipe_domain *ipd);
+
+void __ipipe_remove_domain_proc(struct ipipe_domain *ipd);
+
+void __ipipe_flush_printk(unsigned irq, void *cookie);
+
+void __ipipe_walk_pipeline(struct list_head *pos);
+
+void __ipipe_pend_irq(unsigned irq, struct list_head *head);
+
+int __ipipe_dispatch_event(unsigned event, void *data);
+
+void __ipipe_dispatch_wired_nocheck(struct ipipe_domain *head, unsigned irq);
+
+void __ipipe_dispatch_wired(struct ipipe_domain *head, unsigned irq);
+
+void __ipipe_sync_stage(unsigned long syncmask);
+
+void __ipipe_set_irq_pending(struct ipipe_domain *ipd, unsigned irq);
+
+void __ipipe_lock_irq(struct ipipe_domain *ipd, int cpu, unsigned irq);
+
+void __ipipe_unlock_irq(struct ipipe_domain *ipd, unsigned irq);
+
+void __ipipe_pin_range_globally(unsigned long start, unsigned long end);
+
+/* Must be called hw IRQs off. */
+static inline void ipipe_irq_lock(unsigned irq)
+{
+	__ipipe_lock_irq(__ipipe_current_domain, ipipe_processor_id(), irq);
+}
+
+/* Must be called hw IRQs off. */
+static inline void ipipe_irq_unlock(unsigned irq)
+{
+	__ipipe_unlock_irq(__ipipe_current_domain, irq);
+}
+
+#ifndef __ipipe_sync_pipeline
+#define __ipipe_sync_pipeline(syncmask) __ipipe_sync_stage(syncmask)
+#endif
+
+#ifndef __ipipe_run_irqtail
+#define __ipipe_run_irqtail() do { } while(0)
+#endif
+
+#define __ipipe_pipeline_head_p(ipd) (&(ipd)->p_link == __ipipe_pipeline.next)
+
+/*
+ * Keep the following as a macro, so that client code could check for
+ * the support of the invariant pipeline head optimization.
+ */
+#define __ipipe_pipeline_head() \
+	list_entry(__ipipe_pipeline.next, struct ipipe_domain, p_link)
+
+#define local_irq_enable_hw_cond()		local_irq_enable_hw()
+#define local_irq_disable_hw_cond()		local_irq_disable_hw()
+#define local_irq_save_hw_cond(flags)		local_irq_save_hw(flags)
+#define local_irq_restore_hw_cond(flags)	local_irq_restore_hw(flags)
+
+#ifdef CONFIG_SMP
+cpumask_t __ipipe_set_irq_affinity(unsigned irq, cpumask_t cpumask);
+int __ipipe_send_ipi(unsigned ipi, cpumask_t cpumask);
+#define local_irq_save_hw_smp(flags)		local_irq_save_hw(flags)
+#define local_irq_restore_hw_smp(flags)		local_irq_restore_hw(flags)
+#else /* !CONFIG_SMP */
+#define local_irq_save_hw_smp(flags)		do { (void)(flags); } while(0)
+#define local_irq_restore_hw_smp(flags)		do { } while(0)
+#endif /* CONFIG_SMP */
+
+#define local_irq_save_full(vflags, rflags)		\
+	do {						\
+		local_irq_save(vflags);			\
+		local_irq_save_hw(rflags);		\
+	} while(0)
+
+#define local_irq_restore_full(vflags, rflags)		\
+	do {						\
+		local_irq_restore_hw(rflags);		\
+		local_irq_restore(vflags);		\
+	} while(0)
+
+static inline void __local_irq_restore_nosync(unsigned long x)
+{
+	struct ipipe_percpu_domain_data *p = ipipe_root_cpudom_ptr();
+
+	if (raw_irqs_disabled_flags(x))
+		set_bit(IPIPE_STALL_FLAG, &p->status);
+	else
+		clear_bit(IPIPE_STALL_FLAG, &p->status);
+}
+
+static inline void local_irq_restore_nosync(unsigned long x)
+{
+	unsigned long flags;
+	local_irq_save_hw_smp(flags);
+	__local_irq_restore_nosync(x);
+	local_irq_restore_hw_smp(flags);
+}
+
+#define __ipipe_root_domain_p	(__ipipe_current_domain == ipipe_root_domain)
+#define ipipe_root_domain_p	(ipipe_current_domain == ipipe_root_domain)
+
+static inline int __ipipe_event_monitored_p(int ev)
+{
+	if (__ipipe_event_monitors[ev] > 0)
+		return 1;
+
+	return (ipipe_current_domain->evself & (1LL << ev)) != 0;
+}
+
+#define ipipe_sigwake_notify(p)	\
+do {					\
+	if (((p)->flags & PF_EVNOTIFY) && __ipipe_event_monitored_p(IPIPE_EVENT_SIGWAKE)) \
+		__ipipe_dispatch_event(IPIPE_EVENT_SIGWAKE, p);		\
+} while(0)
+
+#define ipipe_exit_notify(p)	\
+do {				\
+	if (((p)->flags & PF_EVNOTIFY) && __ipipe_event_monitored_p(IPIPE_EVENT_EXIT)) \
+		__ipipe_dispatch_event(IPIPE_EVENT_EXIT, p);		\
+} while(0)
+
+#define ipipe_setsched_notify(p)	\
+do {					\
+	if (((p)->flags & PF_EVNOTIFY) && __ipipe_event_monitored_p(IPIPE_EVENT_SETSCHED)) \
+		__ipipe_dispatch_event(IPIPE_EVENT_SETSCHED, p);	\
+} while(0)
+
+#define ipipe_schedule_notify(prev, next)				\
+do {									\
+	if ((((prev)->flags|(next)->flags) & PF_EVNOTIFY) &&		\
+	    __ipipe_event_monitored_p(IPIPE_EVENT_SCHEDULE))		\
+		__ipipe_dispatch_event(IPIPE_EVENT_SCHEDULE,next);	\
+} while(0)
+
+#define ipipe_trap_notify(ex, regs)					\
+({									\
+	unsigned long __flags__;					\
+	int __ret__ = 0;						\
+	local_irq_save_hw_smp(__flags__);				\
+	if ((test_bit(IPIPE_NOSTACK_FLAG, &ipipe_this_cpudom_var(status)) || \
+	     ((current)->flags & PF_EVNOTIFY)) &&			\
+	    __ipipe_event_monitored_p(ex)) {				\
+		local_irq_restore_hw_smp(__flags__);			\
+		__ret__ = __ipipe_dispatch_event(ex, regs);		\
+	} else								\
+		local_irq_restore_hw_smp(__flags__);			\
+	__ret__;							\
+})
+
+static inline void ipipe_init_notify(struct task_struct *p)
+{
+	if (__ipipe_event_monitored_p(IPIPE_EVENT_INIT))
+		__ipipe_dispatch_event(IPIPE_EVENT_INIT, p);
+}
+
+struct mm_struct;
+
+static inline void ipipe_cleanup_notify(struct mm_struct *mm)
+{
+	if (__ipipe_event_monitored_p(IPIPE_EVENT_CLEANUP))
+		__ipipe_dispatch_event(IPIPE_EVENT_CLEANUP, mm);
+}
+
+/* Public interface */
+
+int ipipe_register_domain(struct ipipe_domain *ipd,
+			  struct ipipe_domain_attr *attr);
+
+int ipipe_unregister_domain(struct ipipe_domain *ipd);
+
+void ipipe_suspend_domain(void);
+
+int ipipe_virtualize_irq(struct ipipe_domain *ipd,
+			 unsigned irq,
+			 ipipe_irq_handler_t handler,
+			 void *cookie,
+			 ipipe_irq_ackfn_t acknowledge,
+			 unsigned modemask);
+
+int ipipe_control_irq(unsigned irq,
+		      unsigned clrmask,
+		      unsigned setmask);
+
+unsigned ipipe_alloc_virq(void);
+
+int ipipe_free_virq(unsigned virq);
+
+int ipipe_trigger_irq(unsigned irq);
+
+static inline void __ipipe_propagate_irq(unsigned irq)
+{
+	struct list_head *next = __ipipe_current_domain->p_link.next;
+	if (next == &ipipe_root.p_link) {
+		/* Fast path: root must handle all interrupts. */
+		__ipipe_set_irq_pending(&ipipe_root, irq);
+		return;
+	}
+	__ipipe_pend_irq(irq, next);
+}
+
+static inline void __ipipe_schedule_irq(unsigned irq)
+{
+	__ipipe_pend_irq(irq, &__ipipe_current_domain->p_link);
+}
+
+static inline void __ipipe_schedule_irq_head(unsigned irq)
+{
+	__ipipe_set_irq_pending(__ipipe_pipeline_head(), irq);
+}
+
+static inline void __ipipe_schedule_irq_root(unsigned irq)
+{
+	__ipipe_set_irq_pending(&ipipe_root, irq);
+}
+
+static inline void ipipe_propagate_irq(unsigned irq)
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+	__ipipe_propagate_irq(irq);
+	local_irq_restore_hw(flags);
+}
+
+static inline void ipipe_schedule_irq(unsigned irq)
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+	__ipipe_schedule_irq(irq);
+	local_irq_restore_hw(flags);
+}
+
+static inline void ipipe_schedule_irq_head(unsigned irq)
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+	__ipipe_schedule_irq_head(irq);
+	local_irq_restore_hw(flags);
+}
+
+static inline void ipipe_schedule_irq_root(unsigned irq)
+{
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+	__ipipe_schedule_irq_root(irq);
+	local_irq_restore_hw(flags);
+}
+
+void ipipe_stall_pipeline_from(struct ipipe_domain *ipd);
+
+unsigned long ipipe_test_and_stall_pipeline_from(struct ipipe_domain *ipd);
+
+unsigned long ipipe_test_and_unstall_pipeline_from(struct ipipe_domain *ipd);
+
+static inline void ipipe_unstall_pipeline_from(struct ipipe_domain *ipd)
+{
+	ipipe_test_and_unstall_pipeline_from(ipd);
+}
+
+void ipipe_restore_pipeline_from(struct ipipe_domain *ipd,
+					  unsigned long x);
+
+static inline unsigned long ipipe_test_pipeline_from(struct ipipe_domain *ipd)
+{
+	return test_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(ipd, status));
+}
+
+static inline void ipipe_stall_pipeline_head(void)
+{
+	local_irq_disable_hw();
+	__set_bit(IPIPE_STALL_FLAG, &ipipe_head_cpudom_var(status));
+}
+
+static inline unsigned long ipipe_test_and_stall_pipeline_head(void)
+{
+	local_irq_disable_hw();
+	return __test_and_set_bit(IPIPE_STALL_FLAG, &ipipe_head_cpudom_var(status));
+}
+
+void ipipe_unstall_pipeline_head(void);
+
+void __ipipe_restore_pipeline_head(unsigned long x);
+
+static inline void ipipe_restore_pipeline_head(unsigned long x)
+{
+	/* On some archs, __test_and_set_bit() might return different
+	 * truth value than test_bit(), so we test the exclusive OR of
+	 * both statuses, assuming that the lowest bit is always set in
+	 * the truth value (if this is wrong, the failed optimization will
+	 * be caught in __ipipe_restore_pipeline_head() if
+	 * CONFIG_DEBUG_KERNEL is set). */
+	if ((x ^ test_bit(IPIPE_STALL_FLAG, &ipipe_head_cpudom_var(status))) & 1)
+		__ipipe_restore_pipeline_head(x);
+}
+
+#define ipipe_unstall_pipeline() \
+	ipipe_unstall_pipeline_from(ipipe_current_domain)
+
+#define ipipe_test_and_unstall_pipeline() \
+	ipipe_test_and_unstall_pipeline_from(ipipe_current_domain)
+
+#define ipipe_test_pipeline() \
+	ipipe_test_pipeline_from(ipipe_current_domain)
+
+#define ipipe_test_and_stall_pipeline() \
+	ipipe_test_and_stall_pipeline_from(ipipe_current_domain)
+
+#define ipipe_stall_pipeline() \
+	ipipe_stall_pipeline_from(ipipe_current_domain)
+
+#define ipipe_restore_pipeline(x) \
+	ipipe_restore_pipeline_from(ipipe_current_domain, (x))
+
+void ipipe_init_attr(struct ipipe_domain_attr *attr);
+
+int ipipe_get_sysinfo(struct ipipe_sysinfo *sysinfo);
+
+unsigned long ipipe_critical_enter(void (*syncfn) (void));
+
+void ipipe_critical_exit(unsigned long flags);
+
+static inline void ipipe_set_printk_sync(struct ipipe_domain *ipd)
+{
+	set_bit(IPIPE_SPRINTK_FLAG, &ipd->flags);
+}
+
+static inline void ipipe_set_printk_async(struct ipipe_domain *ipd)
+{
+	clear_bit(IPIPE_SPRINTK_FLAG, &ipd->flags);
+}
+
+static inline void ipipe_set_foreign_stack(struct ipipe_domain *ipd)
+{
+	/* Must be called hw interrupts off. */
+	__set_bit(IPIPE_NOSTACK_FLAG, &ipipe_cpudom_var(ipd, status));
+}
+
+static inline void ipipe_clear_foreign_stack(struct ipipe_domain *ipd)
+{
+	/* Must be called hw interrupts off. */
+	__clear_bit(IPIPE_NOSTACK_FLAG, &ipipe_cpudom_var(ipd, status));
+}
+
+#ifndef ipipe_safe_current
+#define ipipe_safe_current()					\
+({								\
+	struct task_struct *p;					\
+	unsigned long flags;					\
+	local_irq_save_hw_smp(flags);				\
+	p = test_bit(IPIPE_NOSTACK_FLAG,			\
+		     &ipipe_this_cpudom_var(status)) ? &init_task : current; \
+	local_irq_restore_hw_smp(flags);			\
+	p; \
+})
+#endif
+
+ipipe_event_handler_t ipipe_catch_event(struct ipipe_domain *ipd,
+					unsigned event,
+					ipipe_event_handler_t handler);
+
+cpumask_t ipipe_set_irq_affinity(unsigned irq,
+				 cpumask_t cpumask);
+
+int ipipe_send_ipi(unsigned ipi,
+		   cpumask_t cpumask);
+
+int ipipe_setscheduler_root(struct task_struct *p,
+			    int policy,
+			    int prio);
+
+int ipipe_reenter_root(struct task_struct *prev,
+		       int policy,
+		       int prio);
+
+int ipipe_alloc_ptdkey(void);
+
+int ipipe_free_ptdkey(int key);
+
+int ipipe_set_ptd(int key,
+		  void *value);
+
+void *ipipe_get_ptd(int key);
+
+int ipipe_disable_ondemand_mappings(struct task_struct *tsk);
+
+static inline void ipipe_nmi_enter(void)
+{
+	int cpu = ipipe_processor_id();
+
+	per_cpu(ipipe_nmi_saved_root, cpu) = ipipe_root_cpudom_var(status);
+	__set_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+	per_cpu(ipipe_saved_context_check_state, cpu) =
+		ipipe_disable_context_check(cpu);
+#endif /* CONFIG_IPIPE_DEBUG_CONTEXT */
+}
+
+static inline void ipipe_nmi_exit(void)
+{
+	int cpu = ipipe_processor_id();
+
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+	ipipe_restore_context_check
+		(cpu, per_cpu(ipipe_saved_context_check_state, cpu));
+#endif /* CONFIG_IPIPE_DEBUG_CONTEXT */
+
+	if (!test_bit(IPIPE_STALL_FLAG, &per_cpu(ipipe_nmi_saved_root, cpu)))
+		__clear_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status));
+}
+
+#else	/* !CONFIG_IPIPE */
+
+#define ipipe_init()			do { } while(0)
+#define ipipe_suspend_domain()		do { } while(0)
+#define ipipe_sigwake_notify(p)		do { } while(0)
+#define ipipe_setsched_notify(p)	do { } while(0)
+#define ipipe_init_notify(p)		do { } while(0)
+#define ipipe_exit_notify(p)		do { } while(0)
+#define ipipe_cleanup_notify(mm)	do { } while(0)
+#define ipipe_trap_notify(t,r)		0
+#define ipipe_init_proc()		do { } while(0)
+
+static inline void __ipipe_pin_range_globally(unsigned long start,
+					      unsigned long end)
+{
+}
+
+#define local_irq_enable_hw_cond()		do { } while(0)
+#define local_irq_disable_hw_cond()		do { } while(0)
+#define local_irq_save_hw_cond(flags)		do { (void)(flags); } while(0)
+#define local_irq_restore_hw_cond(flags)	do { } while(0)
+#define local_irq_save_hw_smp(flags)		do { (void)(flags); } while(0)
+#define local_irq_restore_hw_smp(flags)		do { } while(0)
+
+#define ipipe_irq_lock(irq)		do { } while(0)
+#define ipipe_irq_unlock(irq)		do { } while(0)
+
+#define __ipipe_root_domain_p		1
+#define ipipe_root_domain_p		1
+#define ipipe_safe_current		current
+#define ipipe_processor_id()		smp_processor_id()
+
+#define ipipe_nmi_enter()		do { } while (0)
+#define ipipe_nmi_exit()		do { } while (0)
+
+#define local_irq_disable_head()	local_irq_disable()
+
+#define local_irq_save_full(vflags, rflags)	do { (void)(vflags); local_irq_save(rflags); } while(0)
+#define local_irq_restore_full(vflags, rflags)	do { (void)(vflags); local_irq_restore(rflags); } while(0)
+#define local_irq_restore_nosync(vflags)	local_irq_restore(vflags)
+
+#endif	/* CONFIG_IPIPE */
+
+#endif	/* !__LINUX_IPIPE_H */
diff --git a/include/linux/ipipe_base.h b/include/linux/ipipe_base.h
new file mode 100644
index 0000000..c49fe14
--- /dev/null
+++ b/include/linux/ipipe_base.h
@@ -0,0 +1,90 @@
+/* -*- linux-c -*-
+ * include/linux/ipipe_base.h
+ *
+ * Copyright (C) 2002-2007 Philippe Gerum.
+ *               2007 Jan Kiszka.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __LINUX_IPIPE_BASE_H
+#define __LINUX_IPIPE_BASE_H
+
+#ifdef CONFIG_IPIPE
+
+#include <linux/bitops.h>
+#include <asm/ipipe_base.h>
+
+/* Number of virtual IRQs */
+#define IPIPE_NR_VIRQS		BITS_PER_LONG
+/* First virtual IRQ # */
+#define IPIPE_VIRQ_BASE		(((IPIPE_NR_XIRQS + BITS_PER_LONG - 1) / BITS_PER_LONG) * BITS_PER_LONG)
+/* Total number of IRQ slots */
+#define IPIPE_NR_IRQS		(IPIPE_VIRQ_BASE + IPIPE_NR_VIRQS)
+/* Number of indirect words needed to map the whole IRQ space. */
+#define IPIPE_IRQ_IWORDS	((IPIPE_NR_IRQS + BITS_PER_LONG - 1) / BITS_PER_LONG)
+#define IPIPE_IRQ_IMASK		(BITS_PER_LONG - 1)
+#define IPIPE_IRQMASK_ANY	(~0L)
+#define IPIPE_IRQMASK_VIRT	(IPIPE_IRQMASK_ANY << (IPIPE_VIRQ_BASE / BITS_PER_LONG))
+
+/* Per-cpu pipeline status */
+#define IPIPE_STALL_FLAG	0	/* Stalls a pipeline stage -- guaranteed at bit #0 */
+#define IPIPE_SYNC_FLAG		1	/* The interrupt syncer is running for the domain */
+#define IPIPE_NOSTACK_FLAG	2	/* Domain currently runs on a foreign stack */
+
+#define IPIPE_STALL_MASK	(1L << IPIPE_STALL_FLAG)
+#define IPIPE_SYNC_MASK		(1L << IPIPE_SYNC_FLAG)
+#define IPIPE_NOSTACK_MASK	(1L << IPIPE_NOSTACK_FLAG)
+
+typedef void (*ipipe_irq_handler_t)(unsigned irq,
+				    void *cookie);
+
+extern struct ipipe_domain ipipe_root;
+
+#define ipipe_root_domain (&ipipe_root)
+
+void __ipipe_unstall_root(void);
+
+void __ipipe_restore_root(unsigned long x);
+
+#define ipipe_preempt_disable(flags)	local_irq_save_hw(flags)
+#define ipipe_preempt_enable(flags)	local_irq_restore_hw(flags)
+ 
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+void ipipe_check_context(struct ipipe_domain *border_ipd);
+#else /* !CONFIG_IPIPE_DEBUG_CONTEXT */
+static inline void ipipe_check_context(struct ipipe_domain *border_ipd) { }
+#endif /* !CONFIG_IPIPE_DEBUG_CONTEXT */
+
+/* Generic features */
+
+#ifdef CONFIG_GENERIC_CLOCKEVENTS
+#define __IPIPE_FEATURE_REQUEST_TICKDEV    1
+#endif
+#define __IPIPE_FEATURE_DELAYED_ATOMICSW   1
+#define __IPIPE_FEATURE_FASTPEND_IRQ       1
+#define __IPIPE_FEATURE_TRACE_EVENT	   1
+
+#else /* !CONFIG_IPIPE */
+#define ipipe_preempt_disable(flags)	do { \
+						preempt_disable(); \
+						(void)(flags); \
+					} while (0)
+#define ipipe_preempt_enable(flags)	preempt_enable()
+#define ipipe_check_context(ipd)	do { } while(0)
+#endif	/* CONFIG_IPIPE */
+
+#endif	/* !__LINUX_IPIPE_BASE_H */
diff --git a/include/linux/ipipe_compat.h b/include/linux/ipipe_compat.h
new file mode 100644
index 0000000..50a245c
--- /dev/null
+++ b/include/linux/ipipe_compat.h
@@ -0,0 +1,54 @@
+/* -*- linux-c -*-
+ * include/linux/ipipe_compat.h
+ *
+ * Copyright (C) 2007 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __LINUX_IPIPE_COMPAT_H
+#define __LINUX_IPIPE_COMPAT_H
+
+#ifdef CONFIG_IPIPE_COMPAT
+/*
+ * OBSOLETE: defined only for backward compatibility. Will be removed
+ * in future releases, please update client code accordingly.
+ */
+
+#ifdef CONFIG_SMP
+#define ipipe_declare_cpuid	int cpuid
+#define ipipe_load_cpuid()	do { \
+					cpuid = ipipe_processor_id();	\
+				} while(0)
+#define ipipe_lock_cpu(flags)	do { \
+					local_irq_save_hw(flags); \
+					cpuid = ipipe_processor_id(); \
+				} while(0)
+#define ipipe_unlock_cpu(flags)	local_irq_restore_hw(flags)
+#define ipipe_get_cpu(flags)	ipipe_lock_cpu(flags)
+#define ipipe_put_cpu(flags)	ipipe_unlock_cpu(flags)
+#else /* !CONFIG_SMP */
+#define ipipe_declare_cpuid	const int cpuid = 0
+#define ipipe_load_cpuid()	do { } while(0)
+#define ipipe_lock_cpu(flags)	local_irq_save_hw(flags)
+#define ipipe_unlock_cpu(flags)	local_irq_restore_hw(flags)
+#define ipipe_get_cpu(flags)	do { (void)(flags); } while(0)
+#define ipipe_put_cpu(flags)	do { } while(0)
+#endif /* CONFIG_SMP */
+
+#endif /* CONFIG_IPIPE_COMPAT */
+
+#endif	/* !__LINUX_IPIPE_COMPAT_H */
diff --git a/include/linux/ipipe_percpu.h b/include/linux/ipipe_percpu.h
new file mode 100644
index 0000000..4d83119
--- /dev/null
+++ b/include/linux/ipipe_percpu.h
@@ -0,0 +1,86 @@
+/*   -*- linux-c -*-
+ *   include/linux/ipipe_percpu.h
+ *
+ *   Copyright (C) 2007 Philippe Gerum.
+ *
+ *   This program is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ *   USA; either version 2 of the License, or (at your option) any later
+ *   version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program; if not, write to the Free Software
+ *   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __LINUX_IPIPE_PERCPU_H
+#define __LINUX_IPIPE_PERCPU_H
+
+#include <asm/percpu.h>
+#include <asm/ptrace.h>
+
+struct ipipe_domain;
+
+struct ipipe_percpu_domain_data {
+	unsigned long status;	/* <= Must be first in struct. */
+	unsigned long irqpend_himask;
+	unsigned long irqpend_lomask[IPIPE_IRQ_IWORDS];
+	unsigned long irqheld_mask[IPIPE_IRQ_IWORDS];
+	unsigned long irqall[IPIPE_NR_IRQS];
+	u64 evsync;
+};
+
+/*
+ * CAREFUL: all accessors based on __raw_get_cpu_var() you may find in
+ * this file should be used only while hw interrupts are off, to
+ * prevent from CPU migration regardless of the running domain.
+ */
+#ifdef CONFIG_SMP
+#define ipipe_percpudom_ptr(ipd, cpu)	\
+	(&per_cpu(ipipe_percpu_darray, cpu)[(ipd)->slot])
+#define ipipe_cpudom_ptr(ipd)	\
+	(&__ipipe_get_cpu_var(ipipe_percpu_darray)[(ipd)->slot])
+#else
+DECLARE_PER_CPU(struct ipipe_percpu_domain_data *, ipipe_percpu_daddr[CONFIG_IPIPE_DOMAINS]);
+#define ipipe_percpudom_ptr(ipd, cpu)	\
+	(per_cpu(ipipe_percpu_daddr, cpu)[(ipd)->slot])
+#define ipipe_cpudom_ptr(ipd)	\
+	(__ipipe_get_cpu_var(ipipe_percpu_daddr)[(ipd)->slot])
+#endif
+#define ipipe_percpudom(ipd, var, cpu)	(ipipe_percpudom_ptr(ipd, cpu)->var)
+#define ipipe_cpudom_var(ipd, var)	(ipipe_cpudom_ptr(ipd)->var)
+
+#define IPIPE_ROOT_SLOT			0
+#define IPIPE_HEAD_SLOT			(CONFIG_IPIPE_DOMAINS - 1)
+
+DECLARE_PER_CPU(struct ipipe_percpu_domain_data, ipipe_percpu_darray[CONFIG_IPIPE_DOMAINS]);
+
+DECLARE_PER_CPU(struct ipipe_domain *, ipipe_percpu_domain);
+
+DECLARE_PER_CPU(unsigned long, ipipe_nmi_saved_root);
+
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+DECLARE_PER_CPU(int, ipipe_percpu_context_check);
+DECLARE_PER_CPU(int, ipipe_saved_context_check_state);
+#endif
+
+#define ipipe_root_cpudom_ptr(var)	\
+	(&__ipipe_get_cpu_var(ipipe_percpu_darray)[IPIPE_ROOT_SLOT])
+
+#define ipipe_root_cpudom_var(var)	ipipe_root_cpudom_ptr()->var
+
+#define ipipe_this_cpudom_var(var)	\
+	ipipe_cpudom_var(__ipipe_current_domain, var)
+
+#define ipipe_head_cpudom_ptr()		\
+	(&__ipipe_get_cpu_var(ipipe_percpu_darray)[IPIPE_HEAD_SLOT])
+
+#define ipipe_head_cpudom_var(var)	ipipe_head_cpudom_ptr()->var
+
+#endif	/* !__LINUX_IPIPE_PERCPU_H */
diff --git a/include/linux/ipipe_tickdev.h b/include/linux/ipipe_tickdev.h
new file mode 100644
index 0000000..4a1cb1b
--- /dev/null
+++ b/include/linux/ipipe_tickdev.h
@@ -0,0 +1,58 @@
+/* -*- linux-c -*-
+ * include/linux/ipipe_tickdev.h
+ *
+ * Copyright (C) 2007 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef __LINUX_IPIPE_TICKDEV_H
+#define __LINUX_IPIPE_TICKDEV_H
+
+#if defined(CONFIG_IPIPE) && defined(CONFIG_GENERIC_CLOCKEVENTS)
+
+#include <linux/clockchips.h>
+
+struct tick_device;
+
+struct ipipe_tick_device {
+
+	void (*emul_set_mode)(enum clock_event_mode,
+			      struct clock_event_device *cdev);
+	int (*emul_set_tick)(unsigned long delta,
+			     struct clock_event_device *cdev);
+	void (*real_set_mode)(enum clock_event_mode mode,
+			      struct clock_event_device *cdev);
+	int (*real_set_tick)(unsigned long delta,
+			     struct clock_event_device *cdev);
+	struct tick_device *slave;
+	unsigned long real_max_delta_ns;
+	unsigned long real_mult;
+	int real_shift;
+};
+
+int ipipe_request_tickdev(const char *devname,
+			  void (*emumode)(enum clock_event_mode mode,
+					  struct clock_event_device *cdev),
+			  int (*emutick)(unsigned long evt,
+					 struct clock_event_device *cdev),
+			  int cpu, unsigned long *tmfreq);
+
+void ipipe_release_tickdev(int cpu);
+
+#endif /* CONFIG_IPIPE && CONFIG_GENERIC_CLOCKEVENTS */
+
+#endif /* !__LINUX_IPIPE_TICKDEV_H */
diff --git a/include/linux/ipipe_trace.h b/include/linux/ipipe_trace.h
new file mode 100644
index 0000000..627b354
--- /dev/null
+++ b/include/linux/ipipe_trace.h
@@ -0,0 +1,72 @@
+/* -*- linux-c -*-
+ * include/linux/ipipe_trace.h
+ *
+ * Copyright (C) 2005 Luotao Fu.
+ *               2005-2007 Jan Kiszka.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#ifndef _LINUX_IPIPE_TRACE_H
+#define _LINUX_IPIPE_TRACE_H
+
+#ifdef CONFIG_IPIPE_TRACE
+
+#include <linux/types.h>
+
+void ipipe_trace_begin(unsigned long v);
+void ipipe_trace_end(unsigned long v);
+void ipipe_trace_freeze(unsigned long v);
+void ipipe_trace_special(unsigned char special_id, unsigned long v);
+void ipipe_trace_pid(pid_t pid, short prio);
+void ipipe_trace_event(unsigned char id, unsigned long delay_tsc);
+int ipipe_trace_max_reset(void);
+int ipipe_trace_frozen_reset(void);
+
+#else /* !CONFIG_IPIPE_TRACE */
+
+#define ipipe_trace_begin(v)			do { (void)(v); } while(0)
+#define ipipe_trace_end(v)			do { (void)(v); } while(0)
+#define ipipe_trace_freeze(v)			do { (void)(v); } while(0)
+#define ipipe_trace_special(id, v)		do { (void)(id); (void)(v); } while(0)
+#define ipipe_trace_pid(pid, prio)		do { (void)(pid); (void)(prio); } while(0)
+#define ipipe_trace_event(id, delay_tsc)	do { (void)(id); (void)(delay_tsc); } while(0)
+#define ipipe_trace_max_reset()			do { } while(0)
+#define ipipe_trace_froze_reset()		do { } while(0)
+
+#endif /* !CONFIG_IPIPE_TRACE */
+
+#ifdef CONFIG_IPIPE_TRACE_PANIC
+void ipipe_trace_panic_freeze(void);
+void ipipe_trace_panic_dump(void);
+#else
+static inline void ipipe_trace_panic_freeze(void) { }
+static inline void ipipe_trace_panic_dump(void) { }
+#endif
+
+#ifdef CONFIG_IPIPE_TRACE_IRQSOFF
+#define ipipe_trace_irq_entry(irq)	ipipe_trace_begin(irq)
+#define ipipe_trace_irq_exit(irq)	ipipe_trace_end(irq)
+#define ipipe_trace_irqsoff()		ipipe_trace_begin(0x80000000UL)
+#define ipipe_trace_irqson()		ipipe_trace_end(0x80000000UL)
+#else
+#define ipipe_trace_irq_entry(irq)	do { (void)(irq);} while(0)
+#define ipipe_trace_irq_exit(irq)	do { (void)(irq);} while(0)
+#define ipipe_trace_irqsoff()		do { } while(0)
+#define ipipe_trace_irqson()		do { } while(0)
+#endif
+
+#endif	/* !__LINUX_IPIPE_TRACE_H */
diff --git a/include/linux/irq.h b/include/linux/irq.h
index f899b50..f6a0c96 100644
--- a/include/linux/irq.h
+++ b/include/linux/irq.h
@@ -115,6 +115,9 @@ struct irq_chip {
 	void		(*end)(unsigned int irq);
 	void		(*set_affinity)(unsigned int irq,
 					const struct cpumask *dest);
+#ifdef CONFIG_IPIPE
+	void		(*move)(unsigned int irq);
+#endif /* CONFIG_IPIPE */
 	int		(*retrigger)(unsigned int irq);
 	int		(*set_type)(unsigned int irq, unsigned int flow_type);
 	int		(*set_wake)(unsigned int irq, unsigned int on);
@@ -159,6 +162,12 @@ struct irq_2_iommu;
  * @name:		flow handler name for /proc/interrupts output
  */
 struct irq_desc {
+#ifdef CONFIG_IPIPE
+	void			(*ipipe_ack)(unsigned int irq,
+					     struct irq_desc *desc);
+	void			(*ipipe_end)(unsigned int irq,
+					     struct irq_desc *desc);
+#endif /* CONFIG_IPIPE */
 	unsigned int		irq;
 #ifdef CONFIG_SPARSE_IRQ
 	struct timer_rand_state *timer_rand_state;
diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index 7fa3718..7262a76 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -14,6 +14,7 @@
 #include <linux/compiler.h>
 #include <linux/bitops.h>
 #include <linux/log2.h>
+#include <linux/ipipe_base.h>
 #include <linux/typecheck.h>
 #include <linux/ratelimit.h>
 #include <linux/dynamic_printk.h>
@@ -117,9 +118,12 @@ struct user;
 
 #ifdef CONFIG_PREEMPT_VOLUNTARY
 extern int _cond_resched(void);
-# define might_resched() _cond_resched()
+# define might_resched() do { \
+		ipipe_check_context(ipipe_root_domain); \
+		_cond_resched(); \
+	} while (0)
 #else
-# define might_resched() do { } while (0)
+# define might_resched() ipipe_check_context(ipipe_root_domain)
 #endif
 
 #ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 93d0a69..4466bea 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -105,6 +105,8 @@ extern unsigned int kobjsize(const void *objp);
 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
 #define VM_SAO		0x20000000	/* Strong Access Ordering (powerpc) */
 
+#define VM_PINNED	0x40000000	/* Disable faults for the vma */
+
 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 #endif
diff --git a/include/linux/preempt.h b/include/linux/preempt.h
index 72b1a10..80553be 100644
--- a/include/linux/preempt.h
+++ b/include/linux/preempt.h
@@ -9,13 +9,20 @@
 #include <linux/thread_info.h>
 #include <linux/linkage.h>
 #include <linux/list.h>
+#include <linux/ipipe_base.h>
 
 #if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_PREEMPT_TRACER)
   extern void add_preempt_count(int val);
   extern void sub_preempt_count(int val);
 #else
-# define add_preempt_count(val)	do { preempt_count() += (val); } while (0)
-# define sub_preempt_count(val)	do { preempt_count() -= (val); } while (0)
+# define add_preempt_count(val)	do {		\
+    ipipe_check_context(ipipe_root_domain);	\
+    preempt_count() += (val);			\
+  } while (0)
+# define sub_preempt_count(val)	do {		\
+    ipipe_check_context(ipipe_root_domain);	\
+    preempt_count() -= (val);			\
+  } while (0)
 #endif
 
 #define inc_preempt_count() add_preempt_count(1)
diff --git a/include/linux/sched.h b/include/linux/sched.h
index addcf81..a31be7d 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -59,6 +59,7 @@ struct sched_param {
 #include <linux/errno.h>
 #include <linux/nodemask.h>
 #include <linux/mm_types.h>
+#include <linux/ipipe.h>
 
 #include <asm/system.h>
 #include <asm/page.h>
@@ -182,6 +183,13 @@ extern unsigned long long time_sync_thresh;
 /* in tsk->state again */
 #define TASK_DEAD		64
 #define TASK_WAKEKILL		128
+#ifdef CONFIG_IPIPE
+#define TASK_ATOMICSWITCH	512
+#define TASK_NOWAKEUP		1024
+#else  /* !CONFIG_IPIPE */
+#define TASK_ATOMICSWITCH	0
+#define TASK_NOWAKEUP		0
+#endif /* CONFIG_IPIPE */
 
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
@@ -288,6 +296,15 @@ extern void trap_init(void);
 extern void update_process_times(int user);
 extern void scheduler_tick(void);
 
+#ifdef CONFIG_IPIPE
+extern void update_root_process_times(struct pt_regs *regs);
+#else  /* !CONFIG_IPIPE */
+static inline void update_root_process_times(struct pt_regs *regs)
+{
+	update_process_times(user_mode(regs));
+}
+#endif /* CONFIG_IPIPE */
+
 extern void sched_show_task(struct task_struct *p);
 
 #ifdef CONFIG_DETECT_SOFTLOCKUP
@@ -1381,6 +1398,9 @@ struct task_struct {
 #endif
 	atomic_t fs_excl;	/* holding fs exclusive resources */
 	struct rcu_head rcu;
+#ifdef CONFIG_IPIPE
+	void *ptd[IPIPE_ROOT_NPTDKEYS];
+#endif
 
 	/*
 	 * cache last used pipe for splice
@@ -1612,6 +1632,11 @@ extern cputime_t task_gtime(struct task_struct *p);
 #define PF_EXITING	0x00000004	/* getting shut down */
 #define PF_EXITPIDONE	0x00000008	/* pi exit done on shut down */
 #define PF_VCPU		0x00000010	/* I'm a virtual CPU */
+#ifdef CONFIG_IPIPE
+#define PF_EVNOTIFY	0x00000020	/* Notify other domains about internal events */
+#else
+#define PF_EVNOTIFY	0
+#endif /* CONFIG_IPIPE */
 #define PF_FORKNOEXEC	0x00000040	/* forked but didn't exec */
 #define PF_SUPERPRIV	0x00000100	/* used super-user privileges */
 #define PF_DUMPCORE	0x00000200	/* dumped core */
diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h
index a0c66a2..9b44975 100644
--- a/include/linux/spinlock.h
+++ b/include/linux/spinlock.h
@@ -90,10 +90,14 @@ extern int __lockfunc generic__raw_read_trylock(raw_rwlock_t *lock);
 # include <linux/spinlock_up.h>
 #endif
 
+#undef TYPE_EQUAL
+#define TYPE_EQUAL(lock, type) \
+	__builtin_types_compatible_p(typeof(lock), type *)
+
 #ifdef CONFIG_DEBUG_SPINLOCK
   extern void __spin_lock_init(spinlock_t *lock, const char *name,
 			       struct lock_class_key *key);
-# define spin_lock_init(lock)					\
+# define _spin_lock_init(lock)				\
 do {								\
 	static struct lock_class_key __key;			\
 								\
@@ -101,10 +105,21 @@ do {								\
 } while (0)
 
 #else
-# define spin_lock_init(lock)					\
+# define _spin_lock_init(lock)				\
 	do { *(lock) = SPIN_LOCK_UNLOCKED; } while (0)
 #endif
 
+# define spin_lock_init(lock)						\
+	do {								\
+		if (TYPE_EQUAL((lock), __ipipe_spinlock_t))		\
+			do {						\
+				IPIPE_DEFINE_SPINLOCK(__lock__);	\
+				*((ipipe_spinlock_t *)lock) = __lock__; \
+			} while(0);					\
+		else							\
+			_spin_lock_init((spinlock_t *)lock);		\
+	} while(0)
+
 #ifdef CONFIG_DEBUG_SPINLOCK
   extern void __rwlock_init(rwlock_t *lock, const char *name,
 			    struct lock_class_key *key);
@@ -184,7 +199,94 @@ do {								\
 #define read_trylock(lock)		__cond_lock(lock, _read_trylock(lock))
 #define write_trylock(lock)		__cond_lock(lock, _write_trylock(lock))
 
-#define spin_lock(lock)			_spin_lock(lock)
+#define PICK_SPINOP(op, lock)						\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t))			\
+		__raw_spin##op(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	else if (TYPE_EQUAL(lock, spinlock_t))				\
+		_spin##op((spinlock_t *)(lock));			\
+} while (0)
+
+#define PICK_SPINOP_RAW(op, lock)					\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t))			\
+		__raw_spin##op(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	else if (TYPE_EQUAL(lock, spinlock_t))				\
+		__raw_spin##op(&((spinlock_t *)(lock))->raw_lock);	\
+} while (0)
+
+#define PICK_SPINLOCK_IRQ(lock)						\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		__ipipe_spin_lock_irq(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		_spin_lock_irq((spinlock_t *)(lock));			\
+} while (0)
+
+#define PICK_SPINUNLOCK_IRQ(lock)					\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		__ipipe_spin_unlock_irq(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		_spin_unlock_irq((spinlock_t *)(lock));			\
+} while (0)
+
+#define PICK_SPINLOCK_IRQ_RAW(lock)					\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		__ipipe_spin_lock_irq(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		local_irq_disable();					\
+		__raw_spin_lock(&((spinlock_t *)(lock))->raw_lock);	\
+} while (0)
+
+#define PICK_SPINUNLOCK_IRQ_RAW(lock)					\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		__ipipe_spin_unlock_irq(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		__raw_spin_unlock(&((spinlock_t *)(lock))->raw_lock);	\
+		local_irq_enable();					\
+} while (0)
+
+#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
+extern int __bad_spinlock_type(void);
+
+#define PICK_SPINLOCK_IRQSAVE(lock, flags)				\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		(flags) = __ipipe_spin_lock_irqsave(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		flags = _spin_lock_irqsave((spinlock_t *)(lock));	\
+	else __bad_spinlock_type();					\
+} while (0)
+#define PICK_SPINLOCK_IRQSAVE_NESTED(lock, flags, subclass)		\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		(flags) = __ipipe_spin_lock_irqsave(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		flags = _spin_lock_irqsave_nested((spinlock_t *)(lock), subclass); \
+	else __bad_spinlock_type();					\
+} while (0)
+#else
+#define PICK_SPINLOCK_IRQSAVE(lock, flags)				\
+do {									\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		(flags) = __ipipe_spin_lock_irqsave(&((__ipipe_spinlock_t *)(lock))->__raw_lock); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		_spin_lock_irqsave((spinlock_t *)(lock), flags);	\
+} while (0)
+#endif
+
+#define PICK_SPINUNLOCK_IRQRESTORE(lock, flags)				\
+	do {								\
+	if (TYPE_EQUAL((lock), __ipipe_spinlock_t)) {			\
+		__ipipe_spin_unlock_irqrestore(&((__ipipe_spinlock_t *)(lock))->__raw_lock, flags); \
+	} else if (TYPE_EQUAL(lock, spinlock_t))			\
+		_spin_unlock_irqrestore((spinlock_t *)(lock), flags);	\
+} while (0)
+
+#define spin_lock(lock)	PICK_SPINOP(_lock, lock)
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # define spin_lock_nested(lock, subclass) _spin_lock_nested(lock, subclass)
@@ -206,7 +308,7 @@ do {								\
 #define spin_lock_irqsave(lock, flags)			\
 	do {						\
 		typecheck(unsigned long, flags);	\
-		flags = _spin_lock_irqsave(lock);	\
+		PICK_SPINLOCK_IRQSAVE(lock, flags);	\
 	} while (0)
 #define read_lock_irqsave(lock, flags)			\
 	do {						\
@@ -223,13 +325,13 @@ do {								\
 #define spin_lock_irqsave_nested(lock, flags, subclass)			\
 	do {								\
 		typecheck(unsigned long, flags);			\
-		flags = _spin_lock_irqsave_nested(lock, subclass);	\
+		PICK_SPINLOCK_IRQSAVE_NESTED(lock, flags, subclass);	\
 	} while (0)
 #else
 #define spin_lock_irqsave_nested(lock, flags, subclass)			\
 	do {								\
 		typecheck(unsigned long, flags);			\
-		flags = _spin_lock_irqsave(lock);			\
+		PICK_SPINLOCK_IRQSAVE(lock, flags);			\
 	} while (0)
 #endif
 
@@ -238,7 +340,7 @@ do {								\
 #define spin_lock_irqsave(lock, flags)			\
 	do {						\
 		typecheck(unsigned long, flags);	\
-		_spin_lock_irqsave(lock, flags);	\
+		PICK_SPINLOCK_IRQSAVE(lock, flags);	\
 	} while (0)
 #define read_lock_irqsave(lock, flags)			\
 	do {						\
@@ -255,7 +357,7 @@ do {								\
 
 #endif
 
-#define spin_lock_irq(lock)		_spin_lock_irq(lock)
+#define spin_lock_irq(lock)		PICK_SPINLOCK_IRQ(lock)
 #define spin_lock_bh(lock)		_spin_lock_bh(lock)
 
 #define read_lock_irq(lock)		_read_lock_irq(lock)
@@ -269,32 +371,40 @@ do {								\
  */
 #if defined(CONFIG_DEBUG_SPINLOCK) || defined(CONFIG_PREEMPT) || \
 	!defined(CONFIG_SMP)
-# define spin_unlock(lock)		_spin_unlock(lock)
+#define spin_unlock(lock)		PICK_SPINOP(_unlock, lock)
 # define read_unlock(lock)		_read_unlock(lock)
 # define write_unlock(lock)		_write_unlock(lock)
-# define spin_unlock_irq(lock)		_spin_unlock_irq(lock)
-# define read_unlock_irq(lock)		_read_unlock_irq(lock)
-# define write_unlock_irq(lock)		_write_unlock_irq(lock)
+# define spin_unlock_irq(lock)	PICK_SPINUNLOCK_IRQ(lock)
+# define read_unlock_irq(lock)	_read_unlock_irq(lock)
+# define write_unlock_irq(lock)	_write_unlock_irq(lock)
 #else
-# define spin_unlock(lock) \
-    do {__raw_spin_unlock(&(lock)->raw_lock); __release(lock); } while (0)
-# define read_unlock(lock) \
-    do {__raw_read_unlock(&(lock)->raw_lock); __release(lock); } while (0)
-# define write_unlock(lock) \
-    do {__raw_write_unlock(&(lock)->raw_lock); __release(lock); } while (0)
-# define spin_unlock_irq(lock)			\
+# define spin_unlock(lock)			\
 do {						\
-	__raw_spin_unlock(&(lock)->raw_lock);	\
+	PICK_SPINOP_RAW(_unlock, lock); 	\
+	__release(lock);			\
+} while(0)
+# define read_unlock(lock)			\
+do {						\
+	__raw_read_unlock(&(lock)->raw_lock);	\
 	__release(lock);			\
-	local_irq_enable();			\
 } while (0)
-# define read_unlock_irq(lock)			\
+# define write_unlock(lock)			\
+do {						\
+	__raw_write_unlock(&(lock)->raw_lock);	\
+	__release(lock);			\
+} while (0)
+# define spin_unlock_irq(lock)		\
+do {						\
+	PICK_SPINUNLOCK_IRQ_RAW(lock);		\
+	__release(lock);			\
+} while(0)
+# define read_unlock_irq(lock)		\
 do {						\
 	__raw_read_unlock(&(lock)->raw_lock);	\
 	__release(lock);			\
 	local_irq_enable();			\
 } while (0)
-# define write_unlock_irq(lock)			\
+# define write_unlock_irq(lock)		\
 do {						\
 	__raw_write_unlock(&(lock)->raw_lock);	\
 	__release(lock);			\
@@ -305,7 +415,7 @@ do {						\
 #define spin_unlock_irqrestore(lock, flags)		\
 	do {						\
 		typecheck(unsigned long, flags);	\
-		_spin_unlock_irqrestore(lock, flags);	\
+		PICK_SPINUNLOCK_IRQRESTORE(lock, flags);	\
 	} while (0)
 #define spin_unlock_bh(lock)		_spin_unlock_bh(lock)
 
@@ -369,4 +479,29 @@ extern int _atomic_dec_and_lock(atomic_t *atomic, spinlock_t *lock);
  */
 #define spin_can_lock(lock)	(!spin_is_locked(lock))
 
+#ifdef CONFIG_IPIPE
+void __ipipe_spin_lock_irq(raw_spinlock_t *lock);
+void __ipipe_spin_unlock_irq(raw_spinlock_t *lock);
+unsigned long __ipipe_spin_lock_irqsave(raw_spinlock_t *lock);
+void __ipipe_spin_unlock_irqrestore(raw_spinlock_t *lock,
+					     unsigned long x);
+void __ipipe_spin_unlock_irqbegin(ipipe_spinlock_t *lock);
+void __ipipe_spin_unlock_irqcomplete(unsigned long x);
+#define spin_lock_irqsave_cond(lock, flags) \
+	spin_lock_irqsave(lock, flags)
+#define spin_unlock_irqrestore_cond(lock, flags) \
+	spin_unlock_irqrestore(lock, flags)
+#else
+#define spin_lock_irqsave_cond(lock, flags) \
+	do { (void)(flags); spin_lock(lock); } while(0)
+#define spin_unlock_irqrestore_cond(lock, flags) \
+	spin_unlock(lock)
+#define __ipipe_spin_lock_irq(lock)		do { } while(0)
+#define __ipipe_spin_unlock_irq(lock)		do { } while(0)
+#define __ipipe_spin_lock_irqsave(lock)		0
+#define __ipipe_spin_unlock_irqrestore(lock, x)	do { (void)(x); } while(0)
+#define __ipipe_spin_unlock_irqbegin(lock)	do { } while(0)
+#define __ipipe_spin_unlock_irqcomplete(x)	do { (void)(x); } while(0)
+#endif
+
 #endif /* __LINUX_SPINLOCK_H */
diff --git a/include/linux/spinlock_types.h b/include/linux/spinlock_types.h
index 68d88f7..f5ce3c4 100644
--- a/include/linux/spinlock_types.h
+++ b/include/linux/spinlock_types.h
@@ -31,6 +31,10 @@ typedef struct {
 #endif
 } spinlock_t;
 
+typedef struct {
+	raw_spinlock_t __raw_lock;
+} __ipipe_spinlock_t;
+
 #define SPINLOCK_MAGIC		0xdead4ead
 
 typedef struct {
@@ -92,9 +96,21 @@ typedef struct {
  * __SPIN_LOCK_UNLOCKED()/__RW_LOCK_UNLOCKED() as appropriate.
  */
 #define SPIN_LOCK_UNLOCKED	__SPIN_LOCK_UNLOCKED(old_style_spin_init)
+#define IPIPE_SPIN_LOCK_UNLOCKED					\
+	(__ipipe_spinlock_t) {	.__raw_lock = __RAW_SPIN_LOCK_UNLOCKED }
 #define RW_LOCK_UNLOCKED	__RW_LOCK_UNLOCKED(old_style_rw_init)
 
 #define DEFINE_SPINLOCK(x)	spinlock_t x = __SPIN_LOCK_UNLOCKED(x)
 #define DEFINE_RWLOCK(x)	rwlock_t x = __RW_LOCK_UNLOCKED(x)
 
+#ifdef CONFIG_IPIPE
+# define ipipe_spinlock_t		__ipipe_spinlock_t
+# define IPIPE_DEFINE_SPINLOCK(x)	ipipe_spinlock_t x = IPIPE_SPIN_LOCK_UNLOCKED
+# define IPIPE_DECLARE_SPINLOCK(x)	extern ipipe_spinlock_t x
+#else
+# define ipipe_spinlock_t		spinlock_t
+# define IPIPE_DEFINE_SPINLOCK(x)	DEFINE_SPINLOCK(x)
+# define IPIPE_DECLARE_SPINLOCK(x)	extern spinlock_t x
+#endif
+
 #endif /* __LINUX_SPINLOCK_TYPES_H */
diff --git a/init/Kconfig b/init/Kconfig
index 5f3be19..0e4ec13 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -73,6 +73,7 @@ config INIT_ENV_ARG_LIMIT
 
 config LOCALVERSION
 	string "Local version - append to kernel release"
+	default "-ipipe"
 	help
 	  Append an extra string to the end of your kernel version.
 	  This will show up when you type uname, for example.
@@ -1076,4 +1077,3 @@ source "block/Kconfig"
 
 config PREEMPT_NOTIFIERS
 	bool
-
diff --git a/init/main.c b/init/main.c
index 4617fe1..81457bc 100644
--- a/init/main.c
+++ b/init/main.c
@@ -543,7 +543,7 @@ asmlinkage void __init start_kernel(void)
 	debug_objects_early_init();
 	cgroup_init_early();
 
-	local_irq_disable();
+	local_irq_disable_hw();
 	early_boot_irqs_off();
 	early_init_irq_lock_class();
 
@@ -603,6 +603,11 @@ asmlinkage void __init start_kernel(void)
 	timekeeping_init();
 	time_init();
 	sched_clock_init();
+ 	/*
+ 	 * We need to wait for the interrupt and time subsystems to be
+ 	 * initialized before enabling the pipeline.
+ 	 */
+  	ipipe_init();
 	profile_init();
 	if (!irqs_disabled())
 		printk(KERN_CRIT "start_kernel(): bug: interrupts were "
@@ -770,6 +775,7 @@ static void __init do_basic_setup(void)
 	usermodehelper_init();
 	driver_init();
 	init_irq_proc();
+ 	ipipe_init_proc();
 	do_initcalls();
 }
 
diff --git a/kernel/Makefile b/kernel/Makefile
index e4791b3..9517339 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -83,6 +83,7 @@ obj-$(CONFIG_PREEMPT_RCU) += rcupreempt.o
 obj-$(CONFIG_TREE_RCU_TRACE) += rcutree_trace.o
 obj-$(CONFIG_PREEMPT_RCU_TRACE) += rcupreempt_trace.o
 obj-$(CONFIG_RELAY) += relay.o
+obj-$(CONFIG_IPIPE) += ipipe/
 obj-$(CONFIG_SYSCTL) += utsname_sysctl.o
 obj-$(CONFIG_TASK_DELAY_ACCT) += delayacct.o
 obj-$(CONFIG_TASKSTATS) += taskstats.o tsacct.o
diff --git a/kernel/exit.c b/kernel/exit.c
index 467ffcd..61c67e7 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1045,6 +1045,7 @@ NORET_TYPE void do_exit(long code)
 		acct_process();
 	trace_sched_process_exit(tsk);
 
+  	ipipe_exit_notify(tsk);
 	exit_sem(tsk);
 	exit_files(tsk);
 	exit_fs(tsk);
@@ -1811,3 +1812,37 @@ SYSCALL_DEFINE3(waitpid, pid_t, pid, int
 }
 
 #endif
+
+void rt_daemonize(void)
+{
+	sigset_t blocked;
+
+	/*
+	 * We don't want to have TIF_FREEZE set if the system-wide hibernation
+	 * or suspend transition begins right now.
+	 */
+	current->flags |= (PF_NOFREEZE | PF_KTHREAD);
+
+	if (current->nsproxy != &init_nsproxy) {
+		get_nsproxy(&init_nsproxy);
+		switch_task_namespaces(current, &init_nsproxy);
+	}
+	set_special_pids(&init_struct_pid);
+	proc_clear_tty(current);
+
+	/* Block and flush all signals */
+	sigfillset(&blocked);
+	sigprocmask(SIG_BLOCK, &blocked, NULL);
+	flush_signals(current);
+
+	/* Become as one with the init task */
+
+	daemonize_fs_struct();
+	exit_files(current);
+	current->files = init_task.files;
+	atomic_inc(&current->files->count);
+
+	reparent_to_kthreadd();
+}
+
+EXPORT_SYMBOL(rt_daemonize);
diff --git a/kernel/fork.c b/kernel/fork.c
index 4cf4e76..c3f87ca 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -507,6 +507,7 @@ void mmput(struct mm_struct *mm)
 	if (atomic_dec_and_test(&mm->mm_users)) {
 		exit_aio(mm);
 		exit_mmap(mm);
+ 		ipipe_cleanup_notify(mm);
 		set_mm_exe_file(mm, NULL);
 		if (!list_empty(&mm->mmlist)) {
 			spin_lock(&mmlist_lock);
@@ -906,7 +907,7 @@ static void copy_flags(unsigned long clone_flags, struct task_struct *p)
 {
 	unsigned long new_flags = p->flags;
 
-	new_flags &= ~PF_SUPERPRIV;
+ 	new_flags &= ~(PF_SUPERPRIV | PF_EVNOTIFY);
 	new_flags |= PF_FORKNOEXEC;
 	new_flags |= PF_STARTING;
 	p->flags = new_flags;
@@ -1287,6 +1288,9 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	write_unlock_irq(&tasklist_lock);
 	proc_fork_connector(p);
 	cgroup_post_fork(p);
+#ifdef CONFIG_IPIPE
+	memset(p->ptd, 0, sizeof(p->ptd));
+#endif /* CONFIG_IPIPE */
 	return p;
 
 bad_fork_free_graph:
diff --git a/kernel/ipipe/Kconfig b/kernel/ipipe/Kconfig
new file mode 100644
index 0000000..de5e6a3
--- /dev/null
+++ b/kernel/ipipe/Kconfig
@@ -0,0 +1,35 @@
+config IPIPE
+	bool "Interrupt pipeline"
+	default y
+	---help---
+	  Activate this option if you want the interrupt pipeline to be
+	  compiled in.
+
+config IPIPE_DOMAINS
+	int "Max domains"
+	depends on IPIPE
+	default 4
+	---help---
+	The maximum number of I-pipe domains to run concurrently.
+
+config IPIPE_COMPAT
+	bool "Maintain code compatibility with older releases"
+	depends on IPIPE
+	default y
+	---help---
+	Activate this option if you want the compatibility code to be
+	defined, so that older I-pipe clients may use obsolete
+	constructs. WARNING: obsolete code will be eventually
+	deprecated in future I-pipe releases, and removed from the
+	compatibility support as time passes. Please fix I-pipe
+	clients to get rid of such uses as soon as possible.
+
+config IPIPE_DELAYED_ATOMICSW
+       bool
+       depends on IPIPE
+       default n
+
+config IPIPE_UNMASKED_CONTEXT_SWITCH
+       bool
+       depends on IPIPE
+       default n
diff --git a/kernel/ipipe/Kconfig.debug b/kernel/ipipe/Kconfig.debug
new file mode 100644
index 0000000..fc3e1dd
--- /dev/null
+++ b/kernel/ipipe/Kconfig.debug
@@ -0,0 +1,86 @@
+config IPIPE_DEBUG
+	bool "I-pipe debugging"
+	depends on IPIPE
+
+config IPIPE_DEBUG_CONTEXT
+	bool "Check for illicit cross-domain calls"
+	depends on IPIPE_DEBUG
+	default y
+	---help---
+	  Enable this feature to arm checkpoints in the kernel that
+	  verify the correct invocation context. On entry of critical
+	  Linux services a warning is issued if the caller is not
+	  running over the root domain.
+
+config IPIPE_TRACE
+	bool "Latency tracing"
+	depends on IPIPE_DEBUG
+	select FRAME_POINTER
+	select KALLSYMS
+	select PROC_FS
+	---help---
+	  Activate this option if you want to use per-function tracing of
+	  the kernel. The tracer will collect data via instrumentation
+	  features like the one below or with the help of explicite calls
+	  of ipipe_trace_xxx(). See include/linux/ipipe_trace.h for the
+	  in-kernel tracing API. The collected data and runtime control
+	  is available via /proc/ipipe/trace/*.
+
+if IPIPE_TRACE
+
+config IPIPE_TRACE_ENABLE
+	bool "Enable tracing on boot"
+	default y
+	---help---
+	  Disable this option if you want to arm the tracer after booting
+	  manually ("echo 1 > /proc/ipipe/tracer/enable"). This can reduce
+	  boot time on slow embedded devices due to the tracer overhead.
+
+config IPIPE_TRACE_MCOUNT
+	bool "Instrument function entries"
+	default y
+	select FUNCTION_TRACER
+	select TRACING
+	select CONTEXT_SWITCH_TRACER
+	---help---
+	  When enabled, records every kernel function entry in the tracer
+	  log. While this slows down the system noticeably, it provides
+	  the highest level of information about the flow of events.
+	  However, it can be switch off in order to record only explicit
+	  I-pipe trace points.
+
+config IPIPE_TRACE_IRQSOFF
+	bool "Trace IRQs-off times"
+	default y
+	---help---
+	  Activate this option if I-pipe shall trace the longest path
+	  with hard-IRQs switched off.
+
+config IPIPE_TRACE_SHIFT
+	int "Depth of trace log (14 => 16Kpoints, 15 => 32Kpoints)"
+	range 10 18
+	default 14
+	---help---
+	  The number of trace points to hold tracing data for each
+	  trace path, as a power of 2.
+
+config IPIPE_TRACE_VMALLOC
+	bool "Use vmalloc'ed trace buffer"
+	default y if EMBEDDED
+	---help---
+	  Instead of reserving static kernel data, the required buffer
+	  is allocated via vmalloc during boot-up when this option is
+	  enabled. This can help to start systems that are low on memory,
+	  but it slightly degrades overall performance. Try this option
+	  when a traced kernel hangs unexpectedly at boot time.
+
+config IPIPE_TRACE_PANIC
+	bool "Enable panic back traces"
+	default y
+	---help---
+	  Provides services to freeze and dump a back trace on panic
+	  situations. This is used on IPIPE_DEBUG_CONTEXT exceptions
+	  as well as ordinary kernel oopses. You can control the number
+	  of printed back trace points via /proc/ipipe/trace.
+
+endif
diff --git a/kernel/ipipe/Makefile b/kernel/ipipe/Makefile
new file mode 100644
index 0000000..6257dfa
--- /dev/null
+++ b/kernel/ipipe/Makefile
@@ -0,0 +1,3 @@
+
+obj-$(CONFIG_IPIPE)	+= core.o
+obj-$(CONFIG_IPIPE_TRACE) += tracer.o
diff --git a/kernel/ipipe/core.c b/kernel/ipipe/core.c
new file mode 100644
index 0000000..dc79d03
--- /dev/null
+++ b/kernel/ipipe/core.c
@@ -0,0 +1,1799 @@
+/* -*- linux-c -*-
+ * linux/kernel/ipipe/core.c
+ *
+ * Copyright (C) 2002-2005 Philippe Gerum.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Architecture-independent I-PIPE core support.
+ */
+
+#include <linux/version.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/sched.h>
+#include <linux/kallsyms.h>
+#include <linux/interrupt.h>
+#include <linux/bitops.h>
+#include <linux/tick.h>
+#include <linux/prefetch.h>
+#ifdef CONFIG_PROC_FS
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#endif	/* CONFIG_PROC_FS */
+#include <linux/ipipe_trace.h>
+#include <linux/ipipe_tickdev.h>
+#include <linux/irq.h>
+
+static int __ipipe_ptd_key_count;
+
+static unsigned long __ipipe_ptd_key_map;
+
+static unsigned long __ipipe_domain_slot_map;
+
+struct ipipe_domain ipipe_root;
+
+#ifndef CONFIG_SMP
+/*
+ * Create an alias to the unique root status, so that arch-dep code
+ * may get simple and easy access to this percpu variable.  We also
+ * create an array of pointers to the percpu domain data; this tends
+ * to produce a better code when reaching non-root domains. We make
+ * sure that the early boot code would be able to dereference the
+ * pointer to the root domain data safely by statically initializing
+ * its value (local_irq*() routines depend on this).
+ */
+#if __GNUC__ >= 4
+extern unsigned long __ipipe_root_status
+__attribute__((alias(__stringify(__raw_get_cpu_var(ipipe_percpu_darray)))));
+EXPORT_SYMBOL(__ipipe_root_status);
+#else /* __GNUC__ < 4 */
+/*
+ * Work around a GCC 3.x issue making alias symbols unusable as
+ * constant initializers.
+ */
+unsigned long *const __ipipe_root_status_addr =
+	&__raw_get_cpu_var(ipipe_percpu_darray)[IPIPE_ROOT_SLOT].status;
+EXPORT_SYMBOL(__ipipe_root_status_addr);
+#endif /* __GNUC__ < 4 */
+
+DEFINE_PER_CPU(struct ipipe_percpu_domain_data *, ipipe_percpu_daddr[CONFIG_IPIPE_DOMAINS]) =
+{ [IPIPE_ROOT_SLOT] = (struct ipipe_percpu_domain_data *)&__raw_get_cpu_var(ipipe_percpu_darray) };
+EXPORT_PER_CPU_SYMBOL(ipipe_percpu_daddr);
+#endif /* !CONFIG_SMP */
+
+DEFINE_PER_CPU(struct ipipe_percpu_domain_data, ipipe_percpu_darray[CONFIG_IPIPE_DOMAINS]) =
+{ [IPIPE_ROOT_SLOT] = { .status = IPIPE_STALL_MASK } }; /* Root domain stalled on each CPU at startup. */
+
+DEFINE_PER_CPU(struct ipipe_domain *, ipipe_percpu_domain) = { &ipipe_root };
+
+DEFINE_PER_CPU(unsigned long, ipipe_nmi_saved_root); /* Copy of root status during NMI */
+
+static IPIPE_DEFINE_SPINLOCK(__ipipe_pipelock);
+
+LIST_HEAD(__ipipe_pipeline);
+
+unsigned long __ipipe_virtual_irq_map;
+
+#ifdef CONFIG_PRINTK
+unsigned __ipipe_printk_virq;
+#endif /* CONFIG_PRINTK */
+
+int __ipipe_event_monitors[IPIPE_NR_EVENTS];
+
+#ifdef CONFIG_GENERIC_CLOCKEVENTS
+
+DECLARE_PER_CPU(struct tick_device, tick_cpu_device);
+
+static DEFINE_PER_CPU(struct ipipe_tick_device, ipipe_tick_cpu_device);
+
+int ipipe_request_tickdev(const char *devname,
+			  void (*emumode)(enum clock_event_mode mode,
+					  struct clock_event_device *cdev),
+			  int (*emutick)(unsigned long delta,
+					 struct clock_event_device *cdev),
+			  int cpu, unsigned long *tmfreq)
+{
+	struct ipipe_tick_device *itd;
+	struct tick_device *slave;
+	struct clock_event_device *evtdev;
+	unsigned long long freq;
+	unsigned long flags;
+	int status;
+
+	flags = ipipe_critical_enter(NULL);
+
+	itd = &per_cpu(ipipe_tick_cpu_device, cpu);
+
+	if (itd->slave != NULL) {
+		status = -EBUSY;
+		goto out;
+	}
+
+	slave = &per_cpu(tick_cpu_device, cpu);
+
+	if (strcmp(slave->evtdev->name, devname)) {
+		/*
+		 * No conflict so far with the current tick device,
+		 * check whether the requested device is sane and has
+		 * been blessed by the kernel.
+		 */
+		status = __ipipe_check_tickdev(devname) ?
+			CLOCK_EVT_MODE_UNUSED : CLOCK_EVT_MODE_SHUTDOWN;
+		goto out;
+	}
+
+	/*
+	 * Our caller asks for using the same clock event device for
+	 * ticking than we do, let's create a tick emulation device to
+	 * interpose on the set_next_event() method, so that we may
+	 * both manage the device in oneshot mode. Only the tick
+	 * emulation code will actually program the clockchip hardware
+	 * for the next shot, though.
+	 *
+	 * CAUTION: we still have to grab the tick device even when it
+	 * current runs in periodic mode, since the kernel may switch
+	 * to oneshot dynamically (highres/no_hz tick mode).
+	 */
+
+	evtdev = slave->evtdev;
+	status = evtdev->mode;
+
+        if (status == CLOCK_EVT_MODE_SHUTDOWN)
+                goto out;
+
+	itd->slave = slave;
+	itd->emul_set_mode = emumode;
+	itd->emul_set_tick = emutick;
+	itd->real_set_mode = evtdev->set_mode;
+	itd->real_set_tick = evtdev->set_next_event;
+	itd->real_max_delta_ns = evtdev->max_delta_ns;
+	itd->real_mult = evtdev->mult;
+	itd->real_shift = evtdev->shift;
+	freq = (1000000000ULL * evtdev->mult) >> evtdev->shift;
+	*tmfreq = (unsigned long)freq;
+	evtdev->set_mode = emumode;
+	evtdev->set_next_event = emutick;
+	evtdev->max_delta_ns = ULONG_MAX;
+	evtdev->mult = 1;
+	evtdev->shift = 0;
+out:
+	ipipe_critical_exit(flags);
+
+	return status;
+}
+
+void ipipe_release_tickdev(int cpu)
+{
+	struct ipipe_tick_device *itd;
+	struct tick_device *slave;
+	struct clock_event_device *evtdev;
+	unsigned long flags;
+
+	flags = ipipe_critical_enter(NULL);
+
+	itd = &per_cpu(ipipe_tick_cpu_device, cpu);
+
+	if (itd->slave != NULL) {
+		slave = &per_cpu(tick_cpu_device, cpu);
+		evtdev = slave->evtdev;
+		evtdev->set_mode = itd->real_set_mode;
+		evtdev->set_next_event = itd->real_set_tick;
+		evtdev->max_delta_ns = itd->real_max_delta_ns;
+		evtdev->mult = itd->real_mult;
+		evtdev->shift = itd->real_shift;
+		itd->slave = NULL;
+	}
+
+	ipipe_critical_exit(flags);
+}
+
+#endif /* CONFIG_GENERIC_CLOCKEVENTS */
+
+/*
+ * ipipe_init() -- Initialization routine of the IPIPE layer. Called
+ * by the host kernel early during the boot procedure.
+ */
+void __init ipipe_init(void)
+{
+	struct ipipe_domain *ipd = &ipipe_root;
+
+	__ipipe_check_platform();	/* Do platform dependent checks first. */
+
+	/*
+	 * A lightweight registration code for the root domain. We are
+	 * running on the boot CPU, hw interrupts are off, and
+	 * secondary CPUs are still lost in space.
+	 */
+
+	/* Reserve percpu data slot #0 for the root domain. */
+	ipd->slot = 0;
+	set_bit(0, &__ipipe_domain_slot_map);
+
+	ipd->name = "Linux";
+	ipd->domid = IPIPE_ROOT_ID;
+	ipd->priority = IPIPE_ROOT_PRIO;
+
+	__ipipe_init_stage(ipd);
+
+	INIT_LIST_HEAD(&ipd->p_link);
+	list_add_tail(&ipd->p_link, &__ipipe_pipeline);
+
+	__ipipe_init_platform();
+
+#ifdef CONFIG_PRINTK
+	__ipipe_printk_virq = ipipe_alloc_virq();	/* Cannot fail here. */
+	ipd->irqs[__ipipe_printk_virq].handler = &__ipipe_flush_printk;
+	ipd->irqs[__ipipe_printk_virq].cookie = NULL;
+	ipd->irqs[__ipipe_printk_virq].acknowledge = NULL;
+	ipd->irqs[__ipipe_printk_virq].control = IPIPE_HANDLE_MASK;
+#endif /* CONFIG_PRINTK */
+
+	__ipipe_enable_pipeline();
+
+	printk(KERN_INFO "I-pipe %s: pipeline enabled.\n",
+	       IPIPE_VERSION_STRING);
+}
+
+void __ipipe_init_stage(struct ipipe_domain *ipd)
+{
+	int cpu, n;
+
+	for_each_online_cpu(cpu) {
+
+		ipipe_percpudom(ipd, irqpend_himask, cpu) = 0;
+
+		for (n = 0; n < IPIPE_IRQ_IWORDS; n++) {
+			ipipe_percpudom(ipd, irqpend_lomask, cpu)[n] = 0;
+			ipipe_percpudom(ipd, irqheld_mask, cpu)[n] = 0;
+		}
+
+		for (n = 0; n < IPIPE_NR_IRQS; n++)
+			ipipe_percpudom(ipd, irqall, cpu)[n] = 0;
+
+		ipipe_percpudom(ipd, evsync, cpu) = 0;
+	}
+
+	for (n = 0; n < IPIPE_NR_IRQS; n++) {
+		ipd->irqs[n].acknowledge = NULL;
+		ipd->irqs[n].handler = NULL;
+		ipd->irqs[n].control = IPIPE_PASS_MASK;	/* Pass but don't handle */
+	}
+
+	for (n = 0; n < IPIPE_NR_EVENTS; n++)
+		ipd->evhand[n] = NULL;
+
+	ipd->evself = 0LL;
+	mutex_init(&ipd->mutex);
+
+	__ipipe_hook_critical_ipi(ipd);
+}
+
+void __ipipe_cleanup_domain(struct ipipe_domain *ipd)
+{
+	ipipe_unstall_pipeline_from(ipd);
+
+#ifdef CONFIG_SMP
+	{
+		int cpu;
+
+		for_each_online_cpu(cpu) {
+			while (ipipe_percpudom(ipd, irqpend_himask, cpu) != 0)
+				cpu_relax();
+		}
+	}
+#else
+	__raw_get_cpu_var(ipipe_percpu_daddr)[ipd->slot] = NULL;
+#endif
+
+	clear_bit(ipd->slot, &__ipipe_domain_slot_map);
+}
+
+void __ipipe_unstall_root(void)
+{
+	struct ipipe_percpu_domain_data *p;
+
+        local_irq_disable_hw();
+
+#ifdef CONFIG_IPIPE_DEBUG
+	/* This helps catching bad usage from assembly call sites. */
+	BUG_ON(!__ipipe_root_domain_p);
+#endif
+
+	p = ipipe_root_cpudom_ptr();
+
+        __clear_bit(IPIPE_STALL_FLAG, &p->status);
+
+        if (unlikely(p->irqpend_himask != 0))
+                __ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+
+        local_irq_enable_hw();
+}
+
+void __ipipe_restore_root(unsigned long x)
+{
+#ifdef CONFIG_IPIPE_DEBUG
+	BUG_ON(!ipipe_root_domain_p);
+#endif
+
+	if (x)
+		__ipipe_stall_root();
+	else
+		__ipipe_unstall_root();
+}
+
+void ipipe_stall_pipeline_from(struct ipipe_domain *ipd)
+{
+	unsigned long flags;
+	/*
+	 * We have to prevent against race on updating the status
+	 * variable _and_ CPU migration at the same time, so disable
+	 * hw IRQs here.
+	 */
+	local_irq_save_hw(flags);
+
+	__set_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(ipd, status));
+
+	if (!__ipipe_pipeline_head_p(ipd))
+		local_irq_restore_hw(flags);
+}
+
+unsigned long ipipe_test_and_stall_pipeline_from(struct ipipe_domain *ipd)
+{
+	unsigned long flags, x;
+
+	/* See ipipe_stall_pipeline_from() */
+	local_irq_save_hw(flags);
+
+	x = __test_and_set_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(ipd, status));
+
+	if (!__ipipe_pipeline_head_p(ipd))
+		local_irq_restore_hw(flags);
+
+	return x;
+}
+
+unsigned long ipipe_test_and_unstall_pipeline_from(struct ipipe_domain *ipd)
+{
+	unsigned long flags, x;
+	struct list_head *pos;
+
+	local_irq_save_hw(flags);
+
+	x = __test_and_clear_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(ipd, status));
+
+	if (ipd == __ipipe_current_domain)
+		pos = &ipd->p_link;
+	else
+		pos = __ipipe_pipeline.next;
+
+	__ipipe_walk_pipeline(pos);
+
+	if (likely(__ipipe_pipeline_head_p(ipd)))
+		local_irq_enable_hw();
+	else
+		local_irq_restore_hw(flags);
+
+	return x;
+}
+
+void ipipe_restore_pipeline_from(struct ipipe_domain *ipd,
+					  unsigned long x)
+{
+	if (x)
+		ipipe_stall_pipeline_from(ipd);
+	else
+		ipipe_unstall_pipeline_from(ipd);
+}
+
+void ipipe_unstall_pipeline_head(void)
+{
+	struct ipipe_percpu_domain_data *p = ipipe_head_cpudom_ptr();
+
+	local_irq_disable_hw();
+
+	__clear_bit(IPIPE_STALL_FLAG, &p->status);
+
+	if (unlikely(p->irqpend_himask != 0)) {
+		struct ipipe_domain *head_domain = __ipipe_pipeline_head();
+		if (likely(head_domain == __ipipe_current_domain))
+			__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+		else
+			__ipipe_walk_pipeline(&head_domain->p_link);
+        }
+
+	local_irq_enable_hw();
+}
+
+void __ipipe_restore_pipeline_head(unsigned long x)
+{
+	struct ipipe_percpu_domain_data *p = ipipe_head_cpudom_ptr();
+
+	local_irq_disable_hw();
+
+	if (x) {
+#ifdef CONFIG_DEBUG_KERNEL
+		static int warned;
+		if (!warned && test_and_set_bit(IPIPE_STALL_FLAG, &p->status)) {
+			/*
+			 * Already stalled albeit ipipe_restore_pipeline_head()
+			 * should have detected it? Send a warning once.
+			 */
+			warned = 1;
+			printk(KERN_WARNING
+				   "I-pipe: ipipe_restore_pipeline_head() optimization failed.\n");
+			dump_stack();
+		}
+#else /* !CONFIG_DEBUG_KERNEL */
+		set_bit(IPIPE_STALL_FLAG, &p->status);
+#endif /* CONFIG_DEBUG_KERNEL */
+	}
+	else {
+		__clear_bit(IPIPE_STALL_FLAG, &p->status);
+		if (unlikely(p->irqpend_himask != 0)) {
+			struct ipipe_domain *head_domain = __ipipe_pipeline_head();
+			if (likely(head_domain == __ipipe_current_domain))
+				__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+			else
+				__ipipe_walk_pipeline(&head_domain->p_link);
+		}
+		local_irq_enable_hw();
+	}
+}
+
+void __ipipe_spin_lock_irq(raw_spinlock_t *lock)
+{
+	local_irq_disable_hw();
+	__raw_spin_lock(lock);
+	__set_bit(IPIPE_STALL_FLAG, &ipipe_this_cpudom_var(status));
+}
+
+void __ipipe_spin_unlock_irq(raw_spinlock_t *lock)
+{
+	__raw_spin_unlock(lock);
+	__clear_bit(IPIPE_STALL_FLAG, &ipipe_this_cpudom_var(status));
+	local_irq_enable_hw();
+}
+
+unsigned long __ipipe_spin_lock_irqsave(raw_spinlock_t *lock)
+{
+	unsigned long flags;
+	int s;
+
+	local_irq_save_hw(flags);
+	__raw_spin_lock(lock);
+	s = __test_and_set_bit(IPIPE_STALL_FLAG, &ipipe_this_cpudom_var(status));
+
+	return raw_mangle_irq_bits(s, flags);
+}
+
+void __ipipe_spin_unlock_irqrestore(raw_spinlock_t *lock, unsigned long x)
+{
+	__raw_spin_unlock(lock);
+	if (!raw_demangle_irq_bits(&x))
+		__clear_bit(IPIPE_STALL_FLAG, &ipipe_this_cpudom_var(status));
+	local_irq_restore_hw(x);
+}
+
+void __ipipe_spin_unlock_irqbegin(ipipe_spinlock_t *lock)
+{
+	__raw_spin_unlock(&lock->__raw_lock);
+}
+
+void __ipipe_spin_unlock_irqcomplete(unsigned long x)
+{
+	if (!raw_demangle_irq_bits(&x))
+		__clear_bit(IPIPE_STALL_FLAG, &ipipe_this_cpudom_var(status));
+	local_irq_restore_hw(x);
+}
+
+/* Must be called hw IRQs off. */
+void __ipipe_set_irq_pending(struct ipipe_domain *ipd, unsigned irq)
+{
+	int level = irq >> IPIPE_IRQ_ISHIFT, rank = irq & IPIPE_IRQ_IMASK;
+	struct ipipe_percpu_domain_data *p = ipipe_cpudom_ptr(ipd);
+
+	prefetchw(p);
+	
+	if (likely(!test_bit(IPIPE_LOCK_FLAG, &ipd->irqs[irq].control))) {
+		__set_bit(rank, &p->irqpend_lomask[level]);
+		__set_bit(level, &p->irqpend_himask);
+	} else
+		__set_bit(rank, &p->irqheld_mask[level]);
+
+	p->irqall[irq]++;
+}
+
+/* Must be called hw IRQs off. */
+void __ipipe_lock_irq(struct ipipe_domain *ipd, int cpu, unsigned irq)
+{
+	struct ipipe_percpu_domain_data *p;
+	int level, rank;
+
+	if (unlikely(test_and_set_bit(IPIPE_LOCK_FLAG, &ipd->irqs[irq].control)))
+		return;
+
+	level = irq >> IPIPE_IRQ_ISHIFT;
+	rank = irq & IPIPE_IRQ_IMASK;
+	p = ipipe_percpudom_ptr(ipd, cpu);
+
+	if (__test_and_clear_bit(rank, &p->irqpend_lomask[level]))
+		__set_bit(rank, &p->irqheld_mask[level]);
+	if (p->irqpend_lomask[level] == 0)
+		__clear_bit(level, &p->irqpend_himask);
+}
+
+/* Must be called hw IRQs off. */
+void __ipipe_unlock_irq(struct ipipe_domain *ipd, unsigned irq)
+{
+	struct ipipe_percpu_domain_data *p;
+	int cpu, level, rank;
+
+	if (unlikely(!test_and_clear_bit(IPIPE_LOCK_FLAG, &ipd->irqs[irq].control)))
+		return;
+
+	level = irq >> IPIPE_IRQ_ISHIFT, rank = irq & IPIPE_IRQ_IMASK;
+	for_each_online_cpu(cpu) {
+		p = ipipe_percpudom_ptr(ipd, cpu);
+		if (test_and_clear_bit(rank, &p->irqheld_mask[level])) {
+			/* We need atomic ops here: */
+			set_bit(rank, &p->irqpend_lomask[level]);
+			set_bit(level, &p->irqpend_himask);
+		}
+	}
+}
+
+/*
+ * __ipipe_walk_pipeline(): Plays interrupts pending in the log. Must
+ * be called with local hw interrupts disabled.
+ */
+void __ipipe_walk_pipeline(struct list_head *pos)
+{
+	struct ipipe_domain *this_domain = __ipipe_current_domain, *next_domain;
+	struct ipipe_percpu_domain_data *p, *np;
+
+	p = ipipe_cpudom_ptr(this_domain);
+
+	while (pos != &__ipipe_pipeline) {
+
+		next_domain = list_entry(pos, struct ipipe_domain, p_link);
+		np = ipipe_cpudom_ptr(next_domain);
+
+		if (test_bit(IPIPE_STALL_FLAG, &np->status))
+			break;	/* Stalled stage -- do not go further. */
+
+		if (np->irqpend_himask) {
+			if (next_domain == this_domain)
+				__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+			else {
+
+				p->evsync = 0;
+				__ipipe_current_domain = next_domain;
+				ipipe_suspend_domain();	/* Sync stage and propagate interrupts. */
+
+				if (__ipipe_current_domain == next_domain)
+					__ipipe_current_domain = this_domain;
+				/*
+				 * Otherwise, something changed the current domain under our
+				 * feet recycling the register set; do not override the new
+				 * domain.
+				 */
+
+				if (p->irqpend_himask &&
+				    !test_bit(IPIPE_STALL_FLAG, &p->status))
+					__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+			}
+			break;
+		} else if (next_domain == this_domain)
+			break;
+
+		pos = next_domain->p_link.next;
+	}
+}
+
+/*
+ * ipipe_suspend_domain() -- Suspend the current domain, switching to
+ * the next one which has pending work down the pipeline.
+ */
+void ipipe_suspend_domain(void)
+{
+	struct ipipe_domain *this_domain, *next_domain;
+	struct ipipe_percpu_domain_data *p;
+	struct list_head *ln;
+	unsigned long flags;
+
+	local_irq_save_hw(flags);
+
+	this_domain = next_domain = __ipipe_current_domain;
+	p = ipipe_cpudom_ptr(this_domain);
+	p->status &= ~(IPIPE_STALL_MASK|IPIPE_SYNC_MASK);
+
+	if (p->irqpend_himask != 0)
+		goto sync_stage;
+
+	for (;;) {
+		ln = next_domain->p_link.next;
+
+		if (ln == &__ipipe_pipeline)
+			break;
+
+		next_domain = list_entry(ln, struct ipipe_domain, p_link);
+		p = ipipe_cpudom_ptr(next_domain);
+
+		if (p->status & IPIPE_STALL_MASK)
+			break;
+
+		if (p->irqpend_himask == 0)
+			continue;
+
+		__ipipe_current_domain = next_domain;
+sync_stage:
+		__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+
+		if (__ipipe_current_domain != next_domain)
+			/*
+			 * Something has changed the current domain under our
+			 * feet, recycling the register set; take note.
+			 */
+			this_domain = __ipipe_current_domain;
+	}
+
+	__ipipe_current_domain = this_domain;
+
+	local_irq_restore_hw(flags);
+}
+
+
+/* ipipe_alloc_virq() -- Allocate a pipelined virtual/soft interrupt.
+ * Virtual interrupts are handled in exactly the same way than their
+ * hw-generated counterparts wrt pipelining.
+ */
+unsigned ipipe_alloc_virq(void)
+{
+	unsigned long flags, irq = 0;
+	int ipos;
+
+	spin_lock_irqsave(&__ipipe_pipelock, flags);
+
+	if (__ipipe_virtual_irq_map != ~0) {
+		ipos = ffz(__ipipe_virtual_irq_map);
+		set_bit(ipos, &__ipipe_virtual_irq_map);
+		irq = ipos + IPIPE_VIRQ_BASE;
+	}
+
+	spin_unlock_irqrestore(&__ipipe_pipelock, flags);
+
+	return irq;
+}
+
+/* ipipe_virtualize_irq() -- Attach a handler (and optionally a hw
+   acknowledge routine) to an interrupt for a given domain. */
+
+int ipipe_virtualize_irq(struct ipipe_domain *ipd,
+			 unsigned irq,
+			 ipipe_irq_handler_t handler,
+			 void *cookie,
+			 ipipe_irq_ackfn_t acknowledge,
+			 unsigned modemask)
+{
+	ipipe_irq_handler_t old_handler;
+	unsigned long flags;
+	int err;
+
+	if (irq >= IPIPE_NR_IRQS)
+		return -EINVAL;
+
+	if (ipd->irqs[irq].control & IPIPE_SYSTEM_MASK)
+		return -EPERM;
+
+	if (!test_bit(IPIPE_AHEAD_FLAG, &ipd->flags))
+		/* Silently unwire interrupts for non-heading domains. */
+		modemask &= ~IPIPE_WIRED_MASK;
+
+	spin_lock_irqsave(&__ipipe_pipelock, flags);
+
+	old_handler = ipd->irqs[irq].handler;
+
+	if (handler != NULL) {
+		if (handler == IPIPE_SAME_HANDLER) {
+			handler = old_handler;
+			cookie = ipd->irqs[irq].cookie;
+
+			if (handler == NULL) {
+				err = -EINVAL;
+				goto unlock_and_exit;
+			}
+		} else if ((modemask & IPIPE_EXCLUSIVE_MASK) != 0 &&
+			   old_handler != NULL) {
+			err = -EBUSY;
+			goto unlock_and_exit;
+		}
+
+		/* Wired interrupts can only be delivered to domains
+		 * always heading the pipeline, and using dynamic
+		 * propagation. */
+
+		if ((modemask & IPIPE_WIRED_MASK) != 0) {
+			if ((modemask & (IPIPE_PASS_MASK | IPIPE_STICKY_MASK)) != 0) {
+				err = -EINVAL;
+				goto unlock_and_exit;
+			}
+			modemask |= (IPIPE_HANDLE_MASK);
+		}
+
+		if ((modemask & IPIPE_STICKY_MASK) != 0)
+			modemask |= IPIPE_HANDLE_MASK;
+	} else
+		modemask &=
+		    ~(IPIPE_HANDLE_MASK | IPIPE_STICKY_MASK |
+		      IPIPE_EXCLUSIVE_MASK | IPIPE_WIRED_MASK);
+
+	if (acknowledge == NULL && !ipipe_virtual_irq_p(irq))
+		/*
+		 * Acknowledge handler unspecified for a hw interrupt:
+		 * use the Linux-defined handler instead.
+		 */
+		acknowledge = ipipe_root_domain->irqs[irq].acknowledge;
+
+	ipd->irqs[irq].handler = handler;
+	ipd->irqs[irq].cookie = cookie;
+	ipd->irqs[irq].acknowledge = acknowledge;
+	ipd->irqs[irq].control = modemask;
+
+	if (irq < NR_IRQS && !ipipe_virtual_irq_p(irq)) {
+		if (handler != NULL) {
+			__ipipe_enable_irqdesc(ipd, irq);
+
+			if ((modemask & IPIPE_ENABLE_MASK) != 0) {
+				if (ipd != __ipipe_current_domain) {
+		/*
+		 * IRQ enable/disable state is domain-sensitive, so we
+		 * may not change it for another domain. What is
+		 * allowed however is forcing some domain to handle an
+		 * interrupt source, by passing the proper 'ipd'
+		 * descriptor which thus may be different from
+		 * __ipipe_current_domain.
+		 */
+					err = -EPERM;
+					goto unlock_and_exit;
+				}
+				__ipipe_enable_irq(irq);
+			}
+		} else if (old_handler != NULL)
+				__ipipe_disable_irqdesc(ipd, irq);
+	}
+
+	err = 0;
+
+      unlock_and_exit:
+
+	spin_unlock_irqrestore(&__ipipe_pipelock, flags);
+
+	return err;
+}
+
+/* ipipe_control_irq() -- Change modes of a pipelined interrupt for
+ * the current domain. */
+
+int ipipe_control_irq(unsigned irq, unsigned clrmask, unsigned setmask)
+{
+	struct ipipe_domain *ipd;
+	unsigned long flags;
+
+	if (irq >= IPIPE_NR_IRQS)
+		return -EINVAL;
+
+	spin_lock_irqsave(&__ipipe_pipelock, flags);
+
+	ipd = __ipipe_current_domain;
+
+	if (ipd->irqs[irq].control & IPIPE_SYSTEM_MASK) {
+		spin_unlock_irqrestore(&__ipipe_pipelock, flags);
+		return -EPERM;
+	}
+
+	if (ipd->irqs[irq].handler == NULL)
+		setmask &= ~(IPIPE_HANDLE_MASK | IPIPE_STICKY_MASK);
+
+	if ((setmask & IPIPE_STICKY_MASK) != 0)
+		setmask |= IPIPE_HANDLE_MASK;
+
+	if ((clrmask & (IPIPE_HANDLE_MASK | IPIPE_STICKY_MASK)) != 0)	/* If one goes, both go. */
+		clrmask |= (IPIPE_HANDLE_MASK | IPIPE_STICKY_MASK);
+
+	ipd->irqs[irq].control &= ~clrmask;
+	ipd->irqs[irq].control |= setmask;
+
+	if ((setmask & IPIPE_ENABLE_MASK) != 0)
+		__ipipe_enable_irq(irq);
+	else if ((clrmask & IPIPE_ENABLE_MASK) != 0)
+		__ipipe_disable_irq(irq);
+
+	spin_unlock_irqrestore(&__ipipe_pipelock, flags);
+
+	return 0;
+}
+
+/* __ipipe_dispatch_event() -- Low-level event dispatcher. */
+
+int ipipe_events_diverted;
+EXPORT_SYMBOL(ipipe_events_diverted);
+
+int __ipipe_dispatch_event (unsigned event, void *data)
+{
+void *handler; if (ipipe_events_diverted && (handler = ipipe_root_domain->evhand[event])) { return ((int (*)(unsigned long, void *))handler)(event, data); } else {
+	struct ipipe_domain *start_domain, *this_domain, *next_domain;
+	ipipe_event_handler_t evhand;
+	struct list_head *pos, *npos;
+	unsigned long flags;
+	int propagate = 1;
+
+	local_irq_save_hw(flags);
+
+	start_domain = this_domain = __ipipe_current_domain;
+
+	list_for_each_safe(pos, npos, &__ipipe_pipeline) {
+		/*
+		 * Note: Domain migration may occur while running
+		 * event or interrupt handlers, in which case the
+		 * current register set is going to be recycled for a
+		 * different domain than the initiating one. We do
+		 * care for that, always tracking the current domain
+		 * descriptor upon return from those handlers.
+		 */
+		next_domain = list_entry(pos, struct ipipe_domain, p_link);
+
+		/*
+		 * Keep a cached copy of the handler's address since
+		 * ipipe_catch_event() may clear it under our feet.
+		 */
+		evhand = next_domain->evhand[event];
+
+		if (evhand != NULL) {
+			__ipipe_current_domain = next_domain;
+			ipipe_cpudom_var(next_domain, evsync) |= (1LL << event);
+			local_irq_restore_hw(flags);
+			propagate = !evhand(event, start_domain, data);
+			local_irq_save_hw(flags);
+			/*
+			 * We may have a migration issue here, if the
+			 * current task is migrated to another CPU on
+			 * behalf of the invoked handler, usually when
+			 * a syscall event is processed. However,
+			 * ipipe_catch_event() will make sure that a
+			 * CPU that clears a handler for any given
+			 * event will not attempt to wait for itself
+			 * to clear the evsync bit for that event,
+			 * which practically plugs the hole, without
+			 * resorting to a much more complex strategy.
+			 */
+			ipipe_cpudom_var(next_domain, evsync) &= ~(1LL << event);
+			if (__ipipe_current_domain != next_domain)
+				this_domain = __ipipe_current_domain;
+		}
+
+		if (next_domain != ipipe_root_domain &&	/* NEVER sync the root stage here. */
+		    ipipe_cpudom_var(next_domain, irqpend_himask) != 0 &&
+		    !test_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(next_domain, status))) {
+			__ipipe_current_domain = next_domain;
+			__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+			if (__ipipe_current_domain != next_domain)
+				this_domain = __ipipe_current_domain;
+		}
+
+		__ipipe_current_domain = this_domain;
+
+		if (next_domain == this_domain || !propagate)
+			break;
+	}
+
+	local_irq_restore_hw(flags);
+
+	return !propagate;
+} }
+
+/*
+ * __ipipe_dispatch_wired -- Wired interrupt dispatcher. Wired
+ * interrupts are immediately and unconditionally delivered to the
+ * domain heading the pipeline upon receipt, and such domain must have
+ * been registered as an invariant head for the system (priority ==
+ * IPIPE_HEAD_PRIORITY). The motivation for using wired interrupts is
+ * to get an extra-fast dispatching path for those IRQs, by relying on
+ * a straightforward logic based on assumptions that must always be
+ * true for invariant head domains.  The following assumptions are
+ * made when dealing with such interrupts:
+ *
+ * 1- Wired interrupts are purely dynamic, i.e. the decision to
+ * propagate them down the pipeline must be done from the head domain
+ * ISR.
+ * 2- Wired interrupts cannot be shared or sticky.
+ * 3- The root domain cannot be an invariant pipeline head, in
+ * consequence of what the root domain cannot handle wired
+ * interrupts.
+ * 4- Wired interrupts must have a valid acknowledge handler for the
+ * head domain (if needed, see __ipipe_handle_irq).
+ *
+ * Called with hw interrupts off.
+ */
+
+void __ipipe_dispatch_wired(struct ipipe_domain *head, unsigned irq)
+{
+	struct ipipe_percpu_domain_data *p = ipipe_cpudom_ptr(head);
+
+	prefetchw(p);
+
+	if (unlikely(test_bit(IPIPE_LOCK_FLAG, &head->irqs[irq].control))) {
+		/*
+		 * If we can't process this IRQ right now, we must
+		 * mark it as held, so that it will get played during
+		 * normal log sync when the corresponding interrupt
+		 * source is eventually unlocked.
+		 */
+		p->irqall[irq]++;
+		__set_bit(irq & IPIPE_IRQ_IMASK, &p->irqheld_mask[irq >> IPIPE_IRQ_ISHIFT]);
+		return;
+	}
+
+	if (test_bit(IPIPE_STALL_FLAG, &p->status)) {
+		__ipipe_set_irq_pending(head, irq);
+		return;
+	}
+
+	__ipipe_dispatch_wired_nocheck(head, irq);
+}
+
+void __ipipe_dispatch_wired_nocheck(struct ipipe_domain *head, unsigned irq) /* hw interrupts off */
+{
+	struct ipipe_percpu_domain_data *p = ipipe_cpudom_ptr(head);
+	struct ipipe_domain *old;
+
+	prefetchw(p);
+
+	old = __ipipe_current_domain;
+	__ipipe_current_domain = head; /* Switch to the head domain. */
+
+	p->irqall[irq]++;
+	__set_bit(IPIPE_STALL_FLAG, &p->status);
+	head->irqs[irq].handler(irq, head->irqs[irq].cookie); /* Call the ISR. */
+	__ipipe_run_irqtail();
+	__clear_bit(IPIPE_STALL_FLAG, &p->status);
+
+	if (__ipipe_current_domain == head) {
+		__ipipe_current_domain = old;
+		if (old == head) {
+			if (p->irqpend_himask)
+				__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+			return;
+		}
+	}
+
+	__ipipe_walk_pipeline(&head->p_link);
+}
+
+/*
+ * __ipipe_sync_stage() -- Flush the pending IRQs for the current
+ * domain (and processor). This routine flushes the interrupt log
+ * (see "Optimistic interrupt protection" from D. Stodolsky et al. for
+ * more on the deferred interrupt scheme). Every interrupt that
+ * occurred while the pipeline was stalled gets played. WARNING:
+ * callers on SMP boxen should always check for CPU migration on
+ * return of this routine. One can control the kind of interrupts
+ * which are going to be sync'ed using the syncmask
+ * parameter. IPIPE_IRQMASK_ANY plays them all, IPIPE_IRQMASK_VIRT
+ * plays virtual interrupts only.
+ *
+ * This routine must be called with hw interrupts off.
+ */
+void __ipipe_sync_stage(unsigned long syncmask)
+{
+	struct ipipe_percpu_domain_data *p;
+	unsigned long mask, submask;
+	struct ipipe_domain *ipd;
+	int level, rank, cpu;
+	unsigned irq;
+
+	ipd = __ipipe_current_domain;
+	p = ipipe_cpudom_ptr(ipd);
+
+	if (__test_and_set_bit(IPIPE_SYNC_FLAG, &p->status))
+		return;
+
+	cpu = ipipe_processor_id();
+
+	/*
+	 * The policy here is to keep the dispatching code interrupt-free
+	 * by stalling the current stage. If the upper domain handler
+	 * (which we call) wants to re-enable interrupts while in a safe
+	 * portion of the code (e.g. SA_INTERRUPT flag unset for Linux's
+	 * sigaction()), it will have to unstall (then stall again before
+	 * returning to us!) the stage when it sees fit.
+	 */
+	while ((mask = (p->irqpend_himask & syncmask)) != 0) {
+		level = __ipipe_ffnz(mask);
+
+		while ((submask = p->irqpend_lomask[level]) != 0) {
+			rank = __ipipe_ffnz(submask);
+			irq = (level << IPIPE_IRQ_ISHIFT) + rank;
+
+			__clear_bit(rank, &p->irqpend_lomask[level]);
+
+			if (p->irqpend_lomask[level] == 0)
+				__clear_bit(level, &p->irqpend_himask);
+			/*
+			 * Make sure the compiler will not postpone
+			 * the pending bitmask updates before calling
+			 * the interrupt handling routine. Otherwise,
+			 * those late updates could overwrite any
+			 * change to irqpend_hi/lomask due to a nested
+			 * interrupt, leaving the latter unprocessed
+			 * (seen on mpc836x).
+			 */
+			barrier();
+
+			if (test_bit(IPIPE_LOCK_FLAG, &ipd->irqs[irq].control))
+				continue;
+
+			__set_bit(IPIPE_STALL_FLAG, &p->status);
+			barrier();
+
+			if (ipd == ipipe_root_domain)
+				trace_hardirqs_off();
+
+			__ipipe_run_isr(ipd, irq);
+			barrier();
+			p = ipipe_cpudom_ptr(__ipipe_current_domain);
+#ifdef CONFIG_SMP
+			{
+				int newcpu = ipipe_processor_id();
+
+				if (newcpu != cpu) {	/* Handle CPU migration. */
+					/*
+					 * We expect any domain to clear the SYNC bit each
+					 * time it switches in a new task, so that preemptions
+					 * and/or CPU migrations (in the SMP case) over the
+					 * ISR do not lock out the log syncer for some
+					 * indefinite amount of time. In the Linux case,
+					 * schedule() handles this (see kernel/sched.c). For
+					 * this reason, we don't bother clearing it here for
+					 * the source CPU in the migration handling case,
+					 * since it must have scheduled another task in by
+					 * now.
+					 */
+					__set_bit(IPIPE_SYNC_FLAG, &p->status);
+					cpu = newcpu;
+				}
+			}
+#endif	/* CONFIG_SMP */
+#ifdef CONFIG_TRACE_IRQFLAGS
+			if (__ipipe_root_domain_p &&
+			    test_bit(IPIPE_STALL_FLAG, &p->status))
+				trace_hardirqs_on();
+#endif
+			__clear_bit(IPIPE_STALL_FLAG, &p->status);
+		}
+	}
+
+	__clear_bit(IPIPE_SYNC_FLAG, &p->status);
+}
+
+/* ipipe_register_domain() -- Link a new domain to the pipeline. */
+
+int ipipe_register_domain(struct ipipe_domain *ipd,
+			  struct ipipe_domain_attr *attr)
+{
+	struct ipipe_domain *_ipd;
+	struct list_head *pos = NULL;
+	unsigned long flags;
+
+	if (!ipipe_root_domain_p) {
+		printk(KERN_WARNING
+		       "I-pipe: Only the root domain may register a new domain.\n");
+		return -EPERM;
+	}
+
+	flags = ipipe_critical_enter(NULL);
+
+	if (attr->priority == IPIPE_HEAD_PRIORITY) {
+		if (test_bit(IPIPE_HEAD_SLOT, &__ipipe_domain_slot_map)) {
+			ipipe_critical_exit(flags);
+			return -EAGAIN;	/* Cannot override current head. */
+		}
+		ipd->slot = IPIPE_HEAD_SLOT;
+	} else
+		ipd->slot = ffz(__ipipe_domain_slot_map);
+
+	if (ipd->slot < CONFIG_IPIPE_DOMAINS) {
+		set_bit(ipd->slot, &__ipipe_domain_slot_map);
+		list_for_each(pos, &__ipipe_pipeline) {
+			_ipd = list_entry(pos, struct ipipe_domain, p_link);
+			if (_ipd->domid == attr->domid)
+				break;
+		}
+	}
+
+	ipipe_critical_exit(flags);
+
+	if (pos != &__ipipe_pipeline) {
+		if (ipd->slot < CONFIG_IPIPE_DOMAINS)
+			clear_bit(ipd->slot, &__ipipe_domain_slot_map);
+		return -EBUSY;
+	}
+
+#ifndef CONFIG_SMP
+	/*
+	 * Set up the perdomain pointers for direct access to the
+	 * percpu domain data. This saves a costly multiply each time
+	 * we need to refer to the contents of the percpu domain data
+	 * array.
+	 */
+	__raw_get_cpu_var(ipipe_percpu_daddr)[ipd->slot] = &__raw_get_cpu_var(ipipe_percpu_darray)[ipd->slot];
+#endif
+
+	ipd->name = attr->name;
+	ipd->domid = attr->domid;
+	ipd->pdd = attr->pdd;
+	ipd->flags = 0;
+
+	if (attr->priority == IPIPE_HEAD_PRIORITY) {
+		ipd->priority = INT_MAX;
+		__set_bit(IPIPE_AHEAD_FLAG,&ipd->flags);
+	}
+	else
+		ipd->priority = attr->priority;
+
+	__ipipe_init_stage(ipd);
+
+	INIT_LIST_HEAD(&ipd->p_link);
+
+#ifdef CONFIG_PROC_FS
+	__ipipe_add_domain_proc(ipd);
+#endif /* CONFIG_PROC_FS */
+
+	flags = ipipe_critical_enter(NULL);
+
+	list_for_each(pos, &__ipipe_pipeline) {
+		_ipd = list_entry(pos, struct ipipe_domain, p_link);
+		if (ipd->priority > _ipd->priority)
+			break;
+	}
+
+	list_add_tail(&ipd->p_link, pos);
+
+	ipipe_critical_exit(flags);
+
+	printk(KERN_INFO "I-pipe: Domain %s registered.\n", ipd->name);
+
+	/*
+	 * Finally, allow the new domain to perform its initialization
+	 * chores.
+	 */
+
+	if (attr->entry != NULL) {
+		local_irq_save_hw_smp(flags);
+		__ipipe_current_domain = ipd;
+		local_irq_restore_hw_smp(flags);
+		attr->entry();
+		local_irq_save_hw(flags);
+		__ipipe_current_domain = ipipe_root_domain;
+
+		if (ipipe_root_cpudom_var(irqpend_himask) != 0 &&
+		    !test_bit(IPIPE_STALL_FLAG, &ipipe_root_cpudom_var(status)))
+			__ipipe_sync_pipeline(IPIPE_IRQMASK_ANY);
+
+		local_irq_restore_hw(flags);
+	}
+
+	return 0;
+}
+
+/* ipipe_unregister_domain() -- Remove a domain from the pipeline. */
+
+int ipipe_unregister_domain(struct ipipe_domain *ipd)
+{
+	unsigned long flags;
+
+	if (!ipipe_root_domain_p) {
+		printk(KERN_WARNING
+		       "I-pipe: Only the root domain may unregister a domain.\n");
+		return -EPERM;
+	}
+
+	if (ipd == ipipe_root_domain) {
+		printk(KERN_WARNING
+		       "I-pipe: Cannot unregister the root domain.\n");
+		return -EPERM;
+	}
+#ifdef CONFIG_SMP
+	{
+		unsigned irq;
+		int cpu;
+
+		/*
+		 * In the SMP case, wait for the logged events to drain on
+		 * other processors before eventually removing the domain
+		 * from the pipeline.
+		 */
+
+		ipipe_unstall_pipeline_from(ipd);
+
+		flags = ipipe_critical_enter(NULL);
+
+		for (irq = 0; irq < IPIPE_NR_IRQS; irq++) {
+			clear_bit(IPIPE_HANDLE_FLAG, &ipd->irqs[irq].control);
+			clear_bit(IPIPE_STICKY_FLAG, &ipd->irqs[irq].control);
+			set_bit(IPIPE_PASS_FLAG, &ipd->irqs[irq].control);
+		}
+
+		ipipe_critical_exit(flags);
+
+		for_each_online_cpu(cpu) {
+			while (ipipe_percpudom(ipd, irqpend_himask, cpu) > 0)
+				cpu_relax();
+		}
+	}
+#endif	/* CONFIG_SMP */
+
+	mutex_lock(&ipd->mutex);
+
+#ifdef CONFIG_PROC_FS
+	__ipipe_remove_domain_proc(ipd);
+#endif /* CONFIG_PROC_FS */
+
+	/*
+	 * Simply remove the domain from the pipeline and we are almost done.
+	 */
+
+	flags = ipipe_critical_enter(NULL);
+	list_del_init(&ipd->p_link);
+	ipipe_critical_exit(flags);
+
+	__ipipe_cleanup_domain(ipd);
+
+	mutex_unlock(&ipd->mutex);
+
+	printk(KERN_INFO "I-pipe: Domain %s unregistered.\n", ipd->name);
+
+	return 0;
+}
+
+/*
+ * ipipe_propagate_irq() -- Force a given IRQ propagation on behalf of
+ * a running interrupt handler to the next domain down the pipeline.
+ * ipipe_schedule_irq() -- Does almost the same as above, but attempts
+ * to pend the interrupt for the current domain first.
+ * Must be called hw IRQs off.
+ */
+void __ipipe_pend_irq(unsigned irq, struct list_head *head)
+{
+	struct ipipe_domain *ipd;
+	struct list_head *ln;
+
+#ifdef CONFIG_IPIPE_DEBUG
+	BUG_ON(irq >= IPIPE_NR_IRQS ||
+	       (ipipe_virtual_irq_p(irq)
+		&& !test_bit(irq - IPIPE_VIRQ_BASE, &__ipipe_virtual_irq_map)));
+#endif
+	for (ln = head; ln != &__ipipe_pipeline; ln = ipd->p_link.next) {
+		ipd = list_entry(ln, struct ipipe_domain, p_link);
+		if (test_bit(IPIPE_HANDLE_FLAG, &ipd->irqs[irq].control)) {
+			__ipipe_set_irq_pending(ipd, irq);
+			return;
+		}
+	}
+}
+
+/* ipipe_free_virq() -- Release a virtual/soft interrupt. */
+
+int ipipe_free_virq(unsigned virq)
+{
+	if (!ipipe_virtual_irq_p(virq))
+		return -EINVAL;
+
+	clear_bit(virq - IPIPE_VIRQ_BASE, &__ipipe_virtual_irq_map);
+
+	return 0;
+}
+
+void ipipe_init_attr(struct ipipe_domain_attr *attr)
+{
+	attr->name = "anon";
+	attr->domid = 1;
+	attr->entry = NULL;
+	attr->priority = IPIPE_ROOT_PRIO;
+	attr->pdd = NULL;
+}
+
+/*
+ * ipipe_catch_event() -- Interpose or remove an event handler for a
+ * given domain.
+ */
+ipipe_event_handler_t ipipe_catch_event(struct ipipe_domain *ipd,
+					unsigned event,
+					ipipe_event_handler_t handler)
+{
+	ipipe_event_handler_t old_handler;
+	unsigned long flags;
+	int self = 0, cpu;
+
+	if (event & IPIPE_EVENT_SELF) {
+		event &= ~IPIPE_EVENT_SELF;
+		self = 1;
+	}
+
+	if (event >= IPIPE_NR_EVENTS)
+		return NULL;
+
+	flags = ipipe_critical_enter(NULL);
+
+	if (!(old_handler = xchg(&ipd->evhand[event],handler)))	{
+		if (handler) {
+			if (self)
+				ipd->evself |= (1LL << event);
+			else
+				__ipipe_event_monitors[event]++;
+		}
+	}
+	else if (!handler) {
+		if (ipd->evself & (1LL << event))
+			ipd->evself &= ~(1LL << event);
+		else
+			__ipipe_event_monitors[event]--;
+	} else if ((ipd->evself & (1LL << event)) && !self) {
+			__ipipe_event_monitors[event]++;
+			ipd->evself &= ~(1LL << event);
+	} else if (!(ipd->evself & (1LL << event)) && self) {
+			__ipipe_event_monitors[event]--;
+			ipd->evself |= (1LL << event);
+	}
+
+	ipipe_critical_exit(flags);
+
+	if (!handler && ipipe_root_domain_p) {
+		/*
+		 * If we cleared a handler on behalf of the root
+		 * domain, we have to wait for any current invocation
+		 * to drain, since our caller might subsequently unmap
+		 * the target domain. To this aim, this code
+		 * synchronizes with __ipipe_dispatch_event(),
+		 * guaranteeing that either the dispatcher sees a null
+		 * handler in which case it discards the invocation
+		 * (which also prevents from entering a livelock), or
+		 * finds a valid handler and calls it. Symmetrically,
+		 * ipipe_catch_event() ensures that the called code
+		 * won't be unmapped under our feet until the event
+		 * synchronization flag is cleared for the given event
+		 * on all CPUs.
+		 */
+		preempt_disable();
+		cpu = smp_processor_id();
+		/*
+		 * Hack: this solves the potential migration issue
+		 * raised in __ipipe_dispatch_event(). This is a
+		 * work-around which makes the assumption that other
+		 * CPUs will subsequently, either process at least one
+		 * interrupt for the target domain, or call
+		 * __ipipe_dispatch_event() without going through a
+		 * migration while running the handler at least once;
+		 * practically, this is safe on any normally running
+		 * system.
+		 */
+		ipipe_percpudom(ipd, evsync, cpu) &= ~(1LL << event);
+		preempt_enable();
+
+		for_each_online_cpu(cpu) {
+			while (ipipe_percpudom(ipd, evsync, cpu) & (1LL << event))
+				schedule_timeout_interruptible(HZ / 50);
+		}
+	}
+
+	return old_handler;
+}
+
+cpumask_t ipipe_set_irq_affinity (unsigned irq, cpumask_t cpumask)
+{
+#ifdef CONFIG_SMP
+	if (irq >= IPIPE_NR_XIRQS)
+		/* Allow changing affinity of external IRQs only. */
+		return CPU_MASK_NONE;
+
+	if (num_online_cpus() > 1)
+		return __ipipe_set_irq_affinity(irq,cpumask);
+#endif /* CONFIG_SMP */
+
+	return CPU_MASK_NONE;
+}
+
+int ipipe_send_ipi (unsigned ipi, cpumask_t cpumask)
+
+{
+#ifdef CONFIG_SMP
+	return __ipipe_send_ipi(ipi,cpumask);
+#else /* !CONFIG_SMP */
+	return -EINVAL;
+#endif /* CONFIG_SMP */
+}
+
+int ipipe_alloc_ptdkey (void)
+{
+	unsigned long flags;
+	int key = -1;
+
+	spin_lock_irqsave(&__ipipe_pipelock,flags);
+
+	if (__ipipe_ptd_key_count < IPIPE_ROOT_NPTDKEYS) {
+		key = ffz(__ipipe_ptd_key_map);
+		set_bit(key,&__ipipe_ptd_key_map);
+		__ipipe_ptd_key_count++;
+	}
+
+	spin_unlock_irqrestore(&__ipipe_pipelock,flags);
+
+	return key;
+}
+
+int ipipe_free_ptdkey (int key)
+{
+	unsigned long flags;
+
+	if (key < 0 || key >= IPIPE_ROOT_NPTDKEYS)
+		return -EINVAL;
+
+	spin_lock_irqsave(&__ipipe_pipelock,flags);
+
+	if (test_and_clear_bit(key,&__ipipe_ptd_key_map))
+		__ipipe_ptd_key_count--;
+
+	spin_unlock_irqrestore(&__ipipe_pipelock,flags);
+
+	return 0;
+}
+
+int ipipe_set_ptd (int key, void *value)
+
+{
+	if (key < 0 || key >= IPIPE_ROOT_NPTDKEYS)
+		return -EINVAL;
+
+	current->ptd[key] = value;
+
+	return 0;
+}
+
+void *ipipe_get_ptd (int key)
+
+{
+	if (key < 0 || key >= IPIPE_ROOT_NPTDKEYS)
+		return NULL;
+
+	return current->ptd[key];
+}
+
+#ifdef CONFIG_PROC_FS
+
+struct proc_dir_entry *ipipe_proc_root;
+
+static int __ipipe_version_info_proc(char *page,
+				     char **start,
+				     off_t off, int count, int *eof, void *data)
+{
+	int len = sprintf(page, "%s\n", IPIPE_VERSION_STRING);
+
+	len -= off;
+
+	if (len <= off + count)
+		*eof = 1;
+
+	*start = page + off;
+
+	if(len > count)
+		len = count;
+
+	if(len < 0)
+		len = 0;
+
+	return len;
+}
+
+static int __ipipe_common_info_show(struct seq_file *p, void *data)
+{
+	struct ipipe_domain *ipd = (struct ipipe_domain *)p->private;
+	char handling, stickiness, lockbit, exclusive, virtuality;
+
+	unsigned long ctlbits;
+	unsigned irq;
+
+	seq_printf(p, "       +----- Handling ([A]ccepted, [G]rabbed, [W]ired, [D]iscarded)\n");
+	seq_printf(p, "       |+---- Sticky\n");
+	seq_printf(p, "       ||+--- Locked\n");
+	seq_printf(p, "       |||+-- Exclusive\n");
+	seq_printf(p, "       ||||+- Virtual\n");
+	seq_printf(p, "[IRQ]  |||||\n");
+
+	mutex_lock(&ipd->mutex);
+
+	for (irq = 0; irq < IPIPE_NR_IRQS; irq++) {
+		/* Remember to protect against
+		 * ipipe_virtual_irq/ipipe_control_irq if more fields
+		 * get involved. */
+		ctlbits = ipd->irqs[irq].control;
+
+		if (irq >= IPIPE_NR_XIRQS && !ipipe_virtual_irq_p(irq))
+			/*
+			 * There might be a hole between the last external
+			 * IRQ and the first virtual one; skip it.
+			 */
+			continue;
+
+		if (ipipe_virtual_irq_p(irq)
+		    && !test_bit(irq - IPIPE_VIRQ_BASE, &__ipipe_virtual_irq_map))
+			/* Non-allocated virtual IRQ; skip it. */
+			continue;
+
+		/*
+		 * Statuses are as follows:
+		 * o "accepted" means handled _and_ passed down the pipeline.
+		 * o "grabbed" means handled, but the interrupt might be
+		 * terminated _or_ passed down the pipeline depending on
+		 * what the domain handler asks for to the I-pipe.
+		 * o "wired" is basically the same as "grabbed", except that
+		 * the interrupt is unconditionally delivered to an invariant
+		 * pipeline head domain.
+		 * o "passed" means unhandled by the domain but passed
+		 * down the pipeline.
+		 * o "discarded" means unhandled and _not_ passed down the
+		 * pipeline. The interrupt merely disappears from the
+		 * current domain down to the end of the pipeline.
+		 */
+		if (ctlbits & IPIPE_HANDLE_MASK) {
+			if (ctlbits & IPIPE_PASS_MASK)
+				handling = 'A';
+			else if (ctlbits & IPIPE_WIRED_MASK)
+				handling = 'W';
+			else
+				handling = 'G';
+		} else if (ctlbits & IPIPE_PASS_MASK)
+			/* Do not output if no major action is taken. */
+			continue;
+		else
+			handling = 'D';
+
+		if (ctlbits & IPIPE_STICKY_MASK)
+			stickiness = 'S';
+		else
+			stickiness = '.';
+
+		if (ctlbits & IPIPE_LOCK_MASK)
+			lockbit = 'L';
+		else
+			lockbit = '.';
+
+		if (ctlbits & IPIPE_EXCLUSIVE_MASK)
+			exclusive = 'X';
+		else
+			exclusive = '.';
+
+		if (ipipe_virtual_irq_p(irq))
+			virtuality = 'V';
+		else
+			virtuality = '.';
+
+		seq_printf(p, " %3u:  %c%c%c%c%c\n",
+			     irq, handling, stickiness, lockbit, exclusive, virtuality);
+	}
+
+	seq_printf(p, "[Domain info]\n");
+
+	seq_printf(p, "id=0x%.8x\n", ipd->domid);
+
+	if (test_bit(IPIPE_AHEAD_FLAG,&ipd->flags))
+		seq_printf(p, "priority=topmost\n");
+	else
+		seq_printf(p, "priority=%d\n", ipd->priority);
+
+	mutex_unlock(&ipd->mutex);
+
+	return 0;
+}
+
+static int __ipipe_common_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, __ipipe_common_info_show, PROC_I(inode)->pde->data);
+}
+
+static struct file_operations __ipipe_info_proc_ops = {
+	.owner		= THIS_MODULE,
+	.open		= __ipipe_common_info_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+void __ipipe_add_domain_proc(struct ipipe_domain *ipd)
+{
+	struct proc_dir_entry *e = create_proc_entry(ipd->name, 0444, ipipe_proc_root);
+	if (e) {
+		e->proc_fops = &__ipipe_info_proc_ops;
+		e->data = (void*) ipd;
+	}
+}
+
+void __ipipe_remove_domain_proc(struct ipipe_domain *ipd)
+{
+	remove_proc_entry(ipd->name,ipipe_proc_root);
+}
+
+void __init ipipe_init_proc(void)
+{
+	ipipe_proc_root = create_proc_entry("ipipe",S_IFDIR, 0);
+	create_proc_read_entry("version",0444,ipipe_proc_root,&__ipipe_version_info_proc,NULL);
+	__ipipe_add_domain_proc(ipipe_root_domain);
+
+	__ipipe_init_tracer();
+}
+
+#endif	/* CONFIG_PROC_FS */
+
+#ifdef CONFIG_IPIPE_DEBUG_CONTEXT
+
+DEFINE_PER_CPU(int, ipipe_percpu_context_check) = { 1 };
+DEFINE_PER_CPU(int, ipipe_saved_context_check_state);
+
+void ipipe_check_context(struct ipipe_domain *border_domain)
+{
+        struct ipipe_percpu_domain_data *p; 
+        struct ipipe_domain *this_domain; 
+        unsigned long flags;
+	int cpu;
+ 
+        local_irq_save_hw_smp(flags); 
+
+        this_domain = __ipipe_current_domain; 
+        p = ipipe_head_cpudom_ptr(); 
+        if (likely(this_domain->priority <= border_domain->priority && 
+		   !test_bit(IPIPE_STALL_FLAG, &p->status))) { 
+                local_irq_restore_hw_smp(flags); 
+                return; 
+        } 
+ 
+	cpu = ipipe_processor_id();
+        if (!per_cpu(ipipe_percpu_context_check, cpu)) { 
+                local_irq_restore_hw_smp(flags); 
+                return; 
+        } 
+ 
+        local_irq_restore_hw_smp(flags); 
+
+	ipipe_context_check_off();
+	ipipe_trace_panic_freeze();
+	ipipe_set_printk_sync(__ipipe_current_domain);
+
+	if (this_domain->priority > border_domain->priority)
+		printk(KERN_ERR "I-pipe: Detected illicit call from domain "
+				"'%s'\n"
+		       KERN_ERR "        into a service reserved for domain "
+				"'%s' and below.\n",
+		       this_domain->name, border_domain->name);
+	else
+		printk(KERN_ERR "I-pipe: Detected stalled topmost domain, "
+				"probably caused by a bug.\n"
+				"        A critical section may have been "
+				"left unterminated.\n");
+	dump_stack();
+	ipipe_trace_panic_dump();
+}
+
+EXPORT_SYMBOL(ipipe_check_context);
+
+#endif /* CONFIG_IPIPE_DEBUG_CONTEXT */
+
+#if defined(CONFIG_IPIPE_DEBUG) && defined(CONFIG_SMP)
+
+int __ipipe_check_percpu_access(void)
+{
+	struct ipipe_percpu_domain_data *p;
+	struct ipipe_domain *this_domain;
+	unsigned long flags;
+	int ret = 0;
+
+	local_irq_save_hw(flags);
+
+	this_domain = __raw_get_cpu_var(ipipe_percpu_domain);
+
+	/*
+	 * Only the root domain may implement preemptive CPU migration
+	 * of tasks, so anything above in the pipeline should be fine.
+	 */
+	if (this_domain->priority > IPIPE_ROOT_PRIO)
+		goto out;
+
+	if (raw_irqs_disabled_flags(flags))
+		goto out;
+
+	/*
+	 * Last chance: hw interrupts were enabled on entry while
+	 * running over the root domain, but the root stage might be
+	 * currently stalled, in which case preemption would be
+	 * disabled, and no migration could occur.
+	 */
+	if (this_domain == ipipe_root_domain) {
+		p = ipipe_root_cpudom_ptr(); 
+		if (test_bit(IPIPE_STALL_FLAG, &p->status))
+			goto out;
+	}
+	/*
+	 * Our caller may end up accessing the wrong per-cpu variable
+	 * instance due to CPU migration; tell it to complain about
+	 * this.
+	 */
+	ret = 1;
+out:
+	local_irq_restore_hw(flags);
+
+	return ret;
+}
+
+#endif /* CONFIG_IPIPE_DEBUG && CONFIG_SMP */
+
+EXPORT_SYMBOL(ipipe_virtualize_irq);
+EXPORT_SYMBOL(ipipe_control_irq);
+EXPORT_SYMBOL(ipipe_suspend_domain);
+EXPORT_SYMBOL(ipipe_alloc_virq);
+EXPORT_PER_CPU_SYMBOL(ipipe_percpu_domain);
+EXPORT_PER_CPU_SYMBOL(ipipe_percpu_darray);
+EXPORT_SYMBOL(ipipe_root);
+EXPORT_SYMBOL(ipipe_stall_pipeline_from);
+EXPORT_SYMBOL(ipipe_test_and_stall_pipeline_from);
+EXPORT_SYMBOL(ipipe_test_and_unstall_pipeline_from);
+EXPORT_SYMBOL(ipipe_restore_pipeline_from);
+EXPORT_SYMBOL(ipipe_unstall_pipeline_head);
+EXPORT_SYMBOL(__ipipe_restore_pipeline_head);
+EXPORT_SYMBOL(__ipipe_unstall_root);
+EXPORT_SYMBOL(__ipipe_restore_root);
+EXPORT_SYMBOL(__ipipe_spin_lock_irq);
+EXPORT_SYMBOL(__ipipe_spin_unlock_irq);
+EXPORT_SYMBOL(__ipipe_spin_lock_irqsave);
+EXPORT_SYMBOL(__ipipe_spin_unlock_irqrestore);
+EXPORT_SYMBOL(__ipipe_pipeline);
+EXPORT_SYMBOL(__ipipe_lock_irq);
+EXPORT_SYMBOL(__ipipe_unlock_irq);
+EXPORT_SYMBOL(ipipe_register_domain);
+EXPORT_SYMBOL(ipipe_unregister_domain);
+EXPORT_SYMBOL(ipipe_free_virq);
+EXPORT_SYMBOL(ipipe_init_attr);
+EXPORT_SYMBOL(ipipe_catch_event);
+EXPORT_SYMBOL(ipipe_alloc_ptdkey);
+EXPORT_SYMBOL(ipipe_free_ptdkey);
+EXPORT_SYMBOL(ipipe_set_ptd);
+EXPORT_SYMBOL(ipipe_get_ptd);
+EXPORT_SYMBOL(ipipe_set_irq_affinity);
+EXPORT_SYMBOL(ipipe_send_ipi);
+EXPORT_SYMBOL(__ipipe_pend_irq);
+EXPORT_SYMBOL(__ipipe_set_irq_pending);
+#if defined(CONFIG_IPIPE_DEBUG) && defined(CONFIG_SMP)
+EXPORT_SYMBOL(__ipipe_check_percpu_access);
+#endif
+#ifdef CONFIG_GENERIC_CLOCKEVENTS
+EXPORT_SYMBOL(ipipe_request_tickdev);
+EXPORT_SYMBOL(ipipe_release_tickdev);
+#endif
+
+EXPORT_SYMBOL(ipipe_critical_enter);
+EXPORT_SYMBOL(ipipe_critical_exit);
+EXPORT_SYMBOL(ipipe_trigger_irq);
+EXPORT_SYMBOL(ipipe_get_sysinfo);
+EXPORT_SYMBOL(__ipipe_sync_stage);
diff --git a/kernel/ipipe/tracer.c b/kernel/ipipe/tracer.c
new file mode 100644
index 0000000..3f6a97c
--- /dev/null
+++ b/kernel/ipipe/tracer.c
@@ -0,0 +1,1443 @@
+/* -*- linux-c -*-
+ * kernel/ipipe/tracer.c
+ *
+ * Copyright (C) 2005 Luotao Fu.
+ *               2005-2008 Jan Kiszka.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, Inc., 675 Mass Ave, Cambridge MA 02139,
+ * USA; either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/version.h>
+#include <linux/kallsyms.h>
+#include <linux/seq_file.h>
+#include <linux/proc_fs.h>
+#include <linux/ctype.h>
+#include <linux/vmalloc.h>
+#include <linux/pid.h>
+#include <linux/utsrelease.h>
+#include <linux/sched.h>
+#include <linux/ipipe.h>
+#include <linux/ftrace.h>
+#include <asm/uaccess.h>
+
+#define IPIPE_TRACE_PATHS           4 /* <!> Do not lower below 3 */
+#define IPIPE_DEFAULT_ACTIVE        0
+#define IPIPE_DEFAULT_MAX           1
+#define IPIPE_DEFAULT_FROZEN        2
+
+#define IPIPE_TRACE_POINTS          (1 << CONFIG_IPIPE_TRACE_SHIFT)
+#define WRAP_POINT_NO(point)        ((point) & (IPIPE_TRACE_POINTS-1))
+
+#define IPIPE_DEFAULT_PRE_TRACE     10
+#define IPIPE_DEFAULT_POST_TRACE    10
+#define IPIPE_DEFAULT_BACK_TRACE    100
+
+#define IPIPE_DELAY_NOTE            1000  /* in nanoseconds */
+#define IPIPE_DELAY_WARN            10000 /* in nanoseconds */
+
+#define IPIPE_TFLG_NMI_LOCK         0x0001
+#define IPIPE_TFLG_NMI_HIT          0x0002
+#define IPIPE_TFLG_NMI_FREEZE_REQ   0x0004
+
+#define IPIPE_TFLG_HWIRQ_OFF        0x0100
+#define IPIPE_TFLG_FREEZING         0x0200
+#define IPIPE_TFLG_CURRDOM_SHIFT    10   /* bits 10..11: current domain */
+#define IPIPE_TFLG_CURRDOM_MASK     0x0C00
+#define IPIPE_TFLG_DOMSTATE_SHIFT   12   /* bits 12..15: domain stalled? */
+#define IPIPE_TFLG_DOMSTATE_BITS    3
+
+#define IPIPE_TFLG_DOMAIN_STALLED(point, n) \
+	(point->flags & (1 << (n + IPIPE_TFLG_DOMSTATE_SHIFT)))
+#define IPIPE_TFLG_CURRENT_DOMAIN(point) \
+	((point->flags & IPIPE_TFLG_CURRDOM_MASK) >> IPIPE_TFLG_CURRDOM_SHIFT)
+
+struct ipipe_trace_point {
+	short type;
+	short flags;
+	unsigned long eip;
+	unsigned long parent_eip;
+	unsigned long v;
+	unsigned long long timestamp;
+};
+
+struct ipipe_trace_path {
+	volatile int flags;
+	int dump_lock; /* separated from flags due to cross-cpu access */
+	int trace_pos; /* next point to fill */
+	int begin, end; /* finalised path begin and end */
+	int post_trace; /* non-zero when in post-trace phase */
+	unsigned long long length; /* max path length in cycles */
+	unsigned long nmi_saved_eip; /* for deferred requests from NMIs */
+	unsigned long nmi_saved_parent_eip;
+	unsigned long nmi_saved_v;
+	struct ipipe_trace_point point[IPIPE_TRACE_POINTS];
+} ____cacheline_aligned_in_smp;
+
+enum ipipe_trace_type
+{
+	IPIPE_TRACE_FUNC = 0,
+	IPIPE_TRACE_BEGIN,
+	IPIPE_TRACE_END,
+	IPIPE_TRACE_FREEZE,
+	IPIPE_TRACE_SPECIAL,
+	IPIPE_TRACE_PID,
+	IPIPE_TRACE_EVENT,
+};
+
+#define IPIPE_TYPE_MASK             0x0007
+#define IPIPE_TYPE_BITS             3
+
+#ifdef CONFIG_IPIPE_TRACE_VMALLOC
+static DEFINE_PER_CPU(struct ipipe_trace_path *, trace_path);
+#else /* !CONFIG_IPIPE_TRACE_VMALLOC */
+static DEFINE_PER_CPU(struct ipipe_trace_path, trace_path[IPIPE_TRACE_PATHS]) =
+	{ [0 ... IPIPE_TRACE_PATHS-1] = { .begin = -1, .end = -1 } };
+#endif /* CONFIG_IPIPE_TRACE_VMALLOC */
+
+int ipipe_trace_enable = 0;
+
+static DEFINE_PER_CPU(int, active_path) = { IPIPE_DEFAULT_ACTIVE };
+static DEFINE_PER_CPU(int, max_path) = { IPIPE_DEFAULT_MAX };
+static DEFINE_PER_CPU(int, frozen_path) = { IPIPE_DEFAULT_FROZEN };
+static IPIPE_DEFINE_SPINLOCK(global_path_lock);
+static int pre_trace = IPIPE_DEFAULT_PRE_TRACE;
+static int post_trace = IPIPE_DEFAULT_POST_TRACE;
+static int back_trace = IPIPE_DEFAULT_BACK_TRACE;
+static int verbose_trace = 1;
+static unsigned long trace_overhead;
+
+static unsigned long trigger_begin;
+static unsigned long trigger_end;
+
+static DEFINE_MUTEX(out_mutex);
+static struct ipipe_trace_path *print_path;
+#ifdef CONFIG_IPIPE_TRACE_PANIC
+static struct ipipe_trace_path *panic_path;
+#endif /* CONFIG_IPIPE_TRACE_PANIC */
+static int print_pre_trace;
+static int print_post_trace;
+
+
+static long __ipipe_signed_tsc2us(long long tsc);
+static void
+__ipipe_trace_point_type(char *buf, struct ipipe_trace_point *point);
+static void __ipipe_print_symname(struct seq_file *m, unsigned long eip);
+
+
+static notrace void
+__ipipe_store_domain_states(struct ipipe_trace_point *point)
+{
+	struct ipipe_domain *ipd;
+	struct list_head *pos;
+	int i = 0;
+
+	list_for_each_prev(pos, &__ipipe_pipeline) {
+		ipd = list_entry(pos, struct ipipe_domain, p_link);
+
+		if (test_bit(IPIPE_STALL_FLAG, &ipipe_cpudom_var(ipd, status)))
+			point->flags |= 1 << (i + IPIPE_TFLG_DOMSTATE_SHIFT);
+
+		if (ipd == __ipipe_current_domain)
+			point->flags |= i << IPIPE_TFLG_CURRDOM_SHIFT;
+
+		if (++i > IPIPE_TFLG_DOMSTATE_BITS)
+			break;
+	}
+}
+
+static notrace int __ipipe_get_free_trace_path(int old, int cpu)
+{
+	int new_active = old;
+	struct ipipe_trace_path *tp;
+
+	do {
+		if (++new_active == IPIPE_TRACE_PATHS)
+			new_active = 0;
+		tp = &per_cpu(trace_path, cpu)[new_active];
+	} while (new_active == per_cpu(max_path, cpu) ||
+	         new_active == per_cpu(frozen_path, cpu) ||
+	         tp->dump_lock);
+
+	return new_active;
+}
+
+static notrace void
+__ipipe_migrate_pre_trace(struct ipipe_trace_path *new_tp,
+                          struct ipipe_trace_path *old_tp, int old_pos)
+{
+	int i;
+
+	new_tp->trace_pos = pre_trace+1;
+
+	for (i = new_tp->trace_pos; i > 0; i--)
+		memcpy(&new_tp->point[WRAP_POINT_NO(new_tp->trace_pos-i)],
+		       &old_tp->point[WRAP_POINT_NO(old_pos-i)],
+		       sizeof(struct ipipe_trace_point));
+
+	/* mark the end (i.e. the point before point[0]) invalid */
+	new_tp->point[IPIPE_TRACE_POINTS-1].eip = 0;
+}
+
+static notrace struct ipipe_trace_path *
+__ipipe_trace_end(int cpu, struct ipipe_trace_path *tp, int pos)
+{
+	struct ipipe_trace_path *old_tp = tp;
+	long active = per_cpu(active_path, cpu);
+	unsigned long long length;
+
+	/* do we have a new worst case? */
+	length = tp->point[tp->end].timestamp -
+	         tp->point[tp->begin].timestamp;
+	if (length > per_cpu(trace_path, cpu)[per_cpu(max_path, cpu)].length) {
+		/* we need protection here against other cpus trying
+		   to start a proc dump */
+		spin_lock(&global_path_lock);
+
+		/* active path holds new worst case */
+		tp->length = length;
+		per_cpu(max_path, cpu) = active;
+
+		/* find next unused trace path */
+		active = __ipipe_get_free_trace_path(active, cpu);
+
+		spin_unlock(&global_path_lock);
+
+		tp = &per_cpu(trace_path, cpu)[active];
+
+		/* migrate last entries for pre-tracing */
+		__ipipe_migrate_pre_trace(tp, old_tp, pos);
+	}
+
+	return tp;
+}
+
+static notrace struct ipipe_trace_path *
+__ipipe_trace_freeze(int cpu, struct ipipe_trace_path *tp, int pos)
+{
+	struct ipipe_trace_path *old_tp = tp;
+	long active = per_cpu(active_path, cpu);
+	int n;
+
+	/* frozen paths have no core (begin=end) */
+	tp->begin = tp->end;
+
+	/* we need protection here against other cpus trying
+	 * to set their frozen path or to start a proc dump */
+	spin_lock(&global_path_lock);
+
+	per_cpu(frozen_path, cpu) = active;
+
+	/* find next unused trace path */
+	active = __ipipe_get_free_trace_path(active, cpu);
+
+	/* check if this is the first frozen path */
+	for_each_possible_cpu(n) {
+		if (n != cpu &&
+		    per_cpu(trace_path, n)[per_cpu(frozen_path, n)].end >= 0)
+			tp->end = -1;
+	}
+
+	spin_unlock(&global_path_lock);
+
+	tp = &per_cpu(trace_path, cpu)[active];
+
+	/* migrate last entries for pre-tracing */
+	__ipipe_migrate_pre_trace(tp, old_tp, pos);
+
+	return tp;
+}
+
+void notrace
+__ipipe_trace(enum ipipe_trace_type type, unsigned long eip,
+              unsigned long parent_eip, unsigned long v)
+{
+	struct ipipe_trace_path *tp, *old_tp;
+	int pos, next_pos, begin;
+	struct ipipe_trace_point *point;
+	unsigned long flags;
+	int cpu;
+
+	local_irq_save_hw_notrace(flags);
+
+	cpu = ipipe_processor_id();
+ restart:
+	tp = old_tp = &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)];
+
+	/* here starts a race window with NMIs - catched below */
+
+	/* check for NMI recursion */
+	if (unlikely(tp->flags & IPIPE_TFLG_NMI_LOCK)) {
+		tp->flags |= IPIPE_TFLG_NMI_HIT;
+
+		/* first freeze request from NMI context? */
+		if ((type == IPIPE_TRACE_FREEZE) &&
+		    !(tp->flags & IPIPE_TFLG_NMI_FREEZE_REQ)) {
+			/* save arguments and mark deferred freezing */
+			tp->flags |= IPIPE_TFLG_NMI_FREEZE_REQ;
+			tp->nmi_saved_eip = eip;
+			tp->nmi_saved_parent_eip = parent_eip;
+			tp->nmi_saved_v = v;
+		}
+		return; /* no need for restoring flags inside IRQ */
+	}
+
+	/* clear NMI events and set lock (atomically per cpu) */
+	tp->flags = (tp->flags & ~(IPIPE_TFLG_NMI_HIT |
+	                           IPIPE_TFLG_NMI_FREEZE_REQ))
+	                       | IPIPE_TFLG_NMI_LOCK;
+
+	/* check active_path again - some nasty NMI may have switched
+	 * it meanwhile */
+	if (unlikely(tp !=
+		     &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)])) {
+		/* release lock on wrong path and restart */
+		tp->flags &= ~IPIPE_TFLG_NMI_LOCK;
+
+		/* there is no chance that the NMI got deferred
+		 * => no need to check for pending freeze requests */
+		goto restart;
+	}
+
+	/* get the point buffer */
+	pos = tp->trace_pos;
+	point = &tp->point[pos];
+
+	/* store all trace point data */
+	point->type = type;
+	point->flags = raw_irqs_disabled_flags(flags) ? IPIPE_TFLG_HWIRQ_OFF : 0;
+	point->eip = eip;
+	point->parent_eip = parent_eip;
+	point->v = v;
+	ipipe_read_tsc(point->timestamp);
+
+	__ipipe_store_domain_states(point);
+
+	/* forward to next point buffer */
+	next_pos = WRAP_POINT_NO(pos+1);
+	tp->trace_pos = next_pos;
+
+	/* only mark beginning if we haven't started yet */
+	begin = tp->begin;
+	if (unlikely(type == IPIPE_TRACE_BEGIN) && (begin < 0))
+		tp->begin = pos;
+
+	/* end of critical path, start post-trace if not already started */
+	if (unlikely(type == IPIPE_TRACE_END) &&
+	    (begin >= 0) && !tp->post_trace)
+		tp->post_trace = post_trace + 1;
+
+	/* freeze only if the slot is free and we are not already freezing */
+	if ((unlikely(type == IPIPE_TRACE_FREEZE) ||
+	     (unlikely(eip >= trigger_begin && eip <= trigger_end) &&
+	     type == IPIPE_TRACE_FUNC)) &&
+	    per_cpu(trace_path, cpu)[per_cpu(frozen_path, cpu)].begin < 0 &&
+	    !(tp->flags & IPIPE_TFLG_FREEZING)) {
+		tp->post_trace = post_trace + 1;
+		tp->flags |= IPIPE_TFLG_FREEZING;
+	}
+
+	/* enforce end of trace in case of overflow */
+	if (unlikely(WRAP_POINT_NO(next_pos + 1) == begin)) {
+		tp->end = pos;
+		goto enforce_end;
+	}
+
+	/* stop tracing this path if we are in post-trace and
+	 *  a) that phase is over now or
+	 *  b) a new TRACE_BEGIN came in but we are not freezing this path */
+	if (unlikely((tp->post_trace > 0) && ((--tp->post_trace == 0) ||
+	             ((type == IPIPE_TRACE_BEGIN) &&
+	              !(tp->flags & IPIPE_TFLG_FREEZING))))) {
+		/* store the path's end (i.e. excluding post-trace) */
+		tp->end = WRAP_POINT_NO(pos - post_trace + tp->post_trace);
+
+ enforce_end:
+		if (tp->flags & IPIPE_TFLG_FREEZING)
+			tp = __ipipe_trace_freeze(cpu, tp, pos);
+		else
+			tp = __ipipe_trace_end(cpu, tp, pos);
+
+		/* reset the active path, maybe already start a new one */
+		tp->begin = (type == IPIPE_TRACE_BEGIN) ?
+			WRAP_POINT_NO(tp->trace_pos - 1) : -1;
+		tp->end = -1;
+		tp->post_trace = 0;
+		tp->flags = 0;
+
+		/* update active_path not earlier to avoid races with NMIs */
+		per_cpu(active_path, cpu) = tp - per_cpu(trace_path, cpu);
+	}
+
+	/* we still have old_tp and point,
+	 * let's reset NMI lock and check for catches */
+	old_tp->flags &= ~IPIPE_TFLG_NMI_LOCK;
+	if (unlikely(old_tp->flags & IPIPE_TFLG_NMI_HIT)) {
+		/* well, this late tagging may not immediately be visible for
+		 * other cpus already dumping this path - a minor issue */
+		point->flags |= IPIPE_TFLG_NMI_HIT;
+
+		/* handle deferred freezing from NMI context */
+		if (old_tp->flags & IPIPE_TFLG_NMI_FREEZE_REQ)
+			__ipipe_trace(IPIPE_TRACE_FREEZE, old_tp->nmi_saved_eip,
+			              old_tp->nmi_saved_parent_eip,
+			              old_tp->nmi_saved_v);
+	}
+
+	local_irq_restore_hw_notrace(flags);
+}
+
+static unsigned long __ipipe_global_path_lock(void)
+{
+	unsigned long flags;
+	int cpu;
+	struct ipipe_trace_path *tp;
+
+	spin_lock_irqsave(&global_path_lock, flags);
+
+	cpu = ipipe_processor_id();
+ restart:
+	tp = &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)];
+
+	/* here is small race window with NMIs - catched below */
+
+	/* clear NMI events and set lock (atomically per cpu) */
+	tp->flags = (tp->flags & ~(IPIPE_TFLG_NMI_HIT |
+	                           IPIPE_TFLG_NMI_FREEZE_REQ))
+	                       | IPIPE_TFLG_NMI_LOCK;
+
+	/* check active_path again - some nasty NMI may have switched
+	 * it meanwhile */
+	if (tp != &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)]) {
+		/* release lock on wrong path and restart */
+		tp->flags &= ~IPIPE_TFLG_NMI_LOCK;
+
+		/* there is no chance that the NMI got deferred
+		 * => no need to check for pending freeze requests */
+		goto restart;
+	}
+
+	return flags;
+}
+
+static void __ipipe_global_path_unlock(unsigned long flags)
+{
+	int cpu;
+	struct ipipe_trace_path *tp;
+
+	/* release spinlock first - it's not involved in the NMI issue */
+	__ipipe_spin_unlock_irqbegin(&global_path_lock);
+
+	cpu = ipipe_processor_id();
+	tp = &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)];
+
+	tp->flags &= ~IPIPE_TFLG_NMI_LOCK;
+
+	/* handle deferred freezing from NMI context */
+	if (tp->flags & IPIPE_TFLG_NMI_FREEZE_REQ)
+		__ipipe_trace(IPIPE_TRACE_FREEZE, tp->nmi_saved_eip,
+		              tp->nmi_saved_parent_eip, tp->nmi_saved_v);
+
+	/* See __ipipe_spin_lock_irqsave() and friends. */
+	__ipipe_spin_unlock_irqcomplete(flags);
+}
+
+void notrace ipipe_trace_begin(unsigned long v)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_BEGIN, __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, v);
+}
+EXPORT_SYMBOL(ipipe_trace_begin);
+
+void notrace ipipe_trace_end(unsigned long v)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_END, __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, v);
+}
+EXPORT_SYMBOL(ipipe_trace_end);
+
+void notrace ipipe_trace_freeze(unsigned long v)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_FREEZE, __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, v);
+}
+EXPORT_SYMBOL(ipipe_trace_freeze);
+
+void notrace ipipe_trace_special(unsigned char id, unsigned long v)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_SPECIAL | (id << IPIPE_TYPE_BITS),
+	              __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, v);
+}
+EXPORT_SYMBOL(ipipe_trace_special);
+
+void notrace ipipe_trace_pid(pid_t pid, short prio)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_PID | (prio << IPIPE_TYPE_BITS),
+	              __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, pid);
+}
+EXPORT_SYMBOL(ipipe_trace_pid);
+
+void notrace ipipe_trace_event(unsigned char id, unsigned long delay_tsc)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_EVENT | (id << IPIPE_TYPE_BITS),
+	              __BUILTIN_RETURN_ADDRESS0,
+	              __BUILTIN_RETURN_ADDRESS1, delay_tsc);
+}
+EXPORT_SYMBOL(ipipe_trace_event);
+
+int ipipe_trace_max_reset(void)
+{
+	int cpu;
+	unsigned long flags;
+	struct ipipe_trace_path *path;
+	int ret = 0;
+
+	flags = __ipipe_global_path_lock();
+
+	for_each_possible_cpu(cpu) {
+		path = &per_cpu(trace_path, cpu)[per_cpu(max_path, cpu)];
+
+		if (path->dump_lock) {
+			ret = -EBUSY;
+			break;
+		}
+
+		path->begin     = -1;
+		path->end       = -1;
+		path->trace_pos = 0;
+		path->length    = 0;
+	}
+
+	__ipipe_global_path_unlock(flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(ipipe_trace_max_reset);
+
+int ipipe_trace_frozen_reset(void)
+{
+	int cpu;
+	unsigned long flags;
+	struct ipipe_trace_path *path;
+	int ret = 0;
+
+	flags = __ipipe_global_path_lock();
+
+	for_each_online_cpu(cpu) {
+		path = &per_cpu(trace_path, cpu)[per_cpu(frozen_path, cpu)];
+
+		if (path->dump_lock) {
+			ret = -EBUSY;
+			break;
+		}
+
+		path->begin = -1;
+		path->end = -1;
+		path->trace_pos = 0;
+		path->length    = 0;
+	}
+
+	__ipipe_global_path_unlock(flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(ipipe_trace_frozen_reset);
+
+static void
+__ipipe_get_task_info(char *task_info, struct ipipe_trace_point *point,
+                      int trylock)
+{
+	struct task_struct *task = NULL;
+	char buf[8];
+	int i;
+	int locked = 1;
+
+	if (trylock) {
+		if (!read_trylock(&tasklist_lock))
+			locked = 0;
+	} else
+		read_lock(&tasklist_lock);
+
+	if (locked)
+		task = find_task_by_pid_type_ns(PIDTYPE_PID, (pid_t)point->v, &init_pid_ns);
+
+	if (task)
+		strncpy(task_info, task->comm, 11);
+	else
+		strcpy(task_info, "-<?>-");
+
+	if (locked)
+		read_unlock(&tasklist_lock);
+
+	for (i = strlen(task_info); i < 11; i++)
+		task_info[i] = ' ';
+
+	sprintf(buf, " %d ", point->type >> IPIPE_TYPE_BITS);
+	strcpy(task_info + (11 - strlen(buf)), buf);
+}
+
+static void
+__ipipe_get_event_date(char *buf,struct ipipe_trace_path *path,
+		       struct ipipe_trace_point *point)
+{
+	long time;
+	int type;
+
+	time = __ipipe_signed_tsc2us(point->timestamp -
+				     path->point[path->begin].timestamp + point->v);
+	type = point->type >> IPIPE_TYPE_BITS;
+
+	if (type == 0)
+		/*
+		 * Event type #0 is predefined, stands for the next
+		 * timer tick.
+		 */
+		sprintf(buf, "tick@%-6ld", time);
+	else
+		sprintf(buf, "%3d@%-7ld", type, time);
+}
+
+#ifdef CONFIG_IPIPE_TRACE_PANIC
+void ipipe_trace_panic_freeze(void)
+{
+	unsigned long flags;
+	int cpu;
+
+	if (!ipipe_trace_enable)
+		return;
+
+	ipipe_trace_enable = 0;
+	local_irq_save_hw_notrace(flags);
+
+	cpu = ipipe_processor_id();
+
+	panic_path = &per_cpu(trace_path, cpu)[per_cpu(active_path, cpu)];
+
+	local_irq_restore_hw(flags);
+}
+EXPORT_SYMBOL(ipipe_trace_panic_freeze);
+
+void ipipe_trace_panic_dump(void)
+{
+	int cnt = back_trace;
+	int start, pos;
+	char buf[16];
+
+	if (!panic_path)
+		return;
+
+	ipipe_context_check_off();
+
+	printk("I-pipe tracer log (%d points):\n", cnt);
+
+	start = pos = WRAP_POINT_NO(panic_path->trace_pos-1);
+
+	while (cnt-- > 0) {
+		struct ipipe_trace_point *point = &panic_path->point[pos];
+		long time;
+		char info[16];
+		int i;
+
+		printk(" %c",
+		       (point->flags & IPIPE_TFLG_HWIRQ_OFF) ? '|' : ' ');
+
+		for (i = IPIPE_TFLG_DOMSTATE_BITS; i >= 0; i--)
+			printk("%c",
+			       (IPIPE_TFLG_CURRENT_DOMAIN(point) == i) ?
+				(IPIPE_TFLG_DOMAIN_STALLED(point, i) ?
+					'#' : '+') :
+				(IPIPE_TFLG_DOMAIN_STALLED(point, i) ?
+					'*' : ' '));
+
+		if (!point->eip)
+			printk("-<invalid>-\n");
+		else {
+			__ipipe_trace_point_type(buf, point);
+			printk("%s", buf);
+
+			switch (point->type & IPIPE_TYPE_MASK) {
+				case IPIPE_TRACE_FUNC:
+					printk("           ");
+					break;
+
+				case IPIPE_TRACE_PID:
+					__ipipe_get_task_info(info,
+							      point, 1);
+					printk("%s", info);
+					break;
+
+				case IPIPE_TRACE_EVENT:
+					__ipipe_get_event_date(info,
+							       panic_path, point);
+					printk("%s", info);
+					break;
+
+				default:
+					printk("0x%08lx ", point->v);
+			}
+
+			time = __ipipe_signed_tsc2us(point->timestamp -
+				panic_path->point[start].timestamp);
+			printk(" %5ld ", time);
+
+			__ipipe_print_symname(NULL, point->eip);
+			printk(" (");
+			__ipipe_print_symname(NULL, point->parent_eip);
+			printk(")\n");
+		}
+		pos = WRAP_POINT_NO(pos - 1);
+	}
+
+	panic_path = NULL;
+}
+EXPORT_SYMBOL(ipipe_trace_panic_dump);
+#endif /* CONFIG_IPIPE_TRACE_PANIC */
+
+
+/* --- /proc output --- */
+
+static notrace int __ipipe_in_critical_trpath(long point_no)
+{
+	return ((WRAP_POINT_NO(point_no-print_path->begin) <
+	         WRAP_POINT_NO(print_path->end-print_path->begin)) ||
+	        ((print_path->end == print_path->begin) &&
+	         (WRAP_POINT_NO(point_no-print_path->end) >
+	          print_post_trace)));
+}
+
+static long __ipipe_signed_tsc2us(long long tsc)
+{
+        unsigned long long abs_tsc;
+        long us;
+
+	/* ipipe_tsc2us works on unsigned => handle sign separately */
+        abs_tsc = (tsc >= 0) ? tsc : -tsc;
+        us = ipipe_tsc2us(abs_tsc);
+        if (tsc < 0)
+                return -us;
+        else
+                return us;
+}
+
+static void
+__ipipe_trace_point_type(char *buf, struct ipipe_trace_point *point)
+{
+	switch (point->type & IPIPE_TYPE_MASK) {
+		case IPIPE_TRACE_FUNC:
+			strcpy(buf, "func    ");
+			break;
+
+		case IPIPE_TRACE_BEGIN:
+			strcpy(buf, "begin   ");
+			break;
+
+		case IPIPE_TRACE_END:
+			strcpy(buf, "end     ");
+			break;
+
+		case IPIPE_TRACE_FREEZE:
+			strcpy(buf, "freeze  ");
+			break;
+
+		case IPIPE_TRACE_SPECIAL:
+			sprintf(buf, "(0x%02x)  ",
+				point->type >> IPIPE_TYPE_BITS);
+			break;
+
+		case IPIPE_TRACE_PID:
+			sprintf(buf, "[%5d] ", (pid_t)point->v);
+			break;
+
+		case IPIPE_TRACE_EVENT:
+			sprintf(buf, "event   ");
+			break;
+	}
+}
+
+static void
+__ipipe_print_pathmark(struct seq_file *m, struct ipipe_trace_point *point)
+{
+	char mark = ' ';
+	int point_no = point - print_path->point;
+	int i;
+
+	if (print_path->end == point_no)
+		mark = '<';
+	else if (print_path->begin == point_no)
+		mark = '>';
+	else if (__ipipe_in_critical_trpath(point_no))
+		mark = ':';
+	seq_printf(m, "%c%c", mark,
+	           (point->flags & IPIPE_TFLG_HWIRQ_OFF) ? '|' : ' ');
+
+	if (!verbose_trace)
+		return;
+
+	for (i = IPIPE_TFLG_DOMSTATE_BITS; i >= 0; i--)
+		seq_printf(m, "%c",
+			(IPIPE_TFLG_CURRENT_DOMAIN(point) == i) ?
+			    (IPIPE_TFLG_DOMAIN_STALLED(point, i) ?
+				'#' : '+') :
+			(IPIPE_TFLG_DOMAIN_STALLED(point, i) ? '*' : ' '));
+}
+
+static void
+__ipipe_print_delay(struct seq_file *m, struct ipipe_trace_point *point)
+{
+	unsigned long delay = 0;
+	int next;
+	char *mark = "  ";
+
+	next = WRAP_POINT_NO(point+1 - print_path->point);
+
+	if (next != print_path->trace_pos)
+		delay = ipipe_tsc2ns(print_path->point[next].timestamp -
+		                     point->timestamp);
+
+	if (__ipipe_in_critical_trpath(point - print_path->point)) {
+		if (delay > IPIPE_DELAY_WARN)
+			mark = "! ";
+		else if (delay > IPIPE_DELAY_NOTE)
+			mark = "+ ";
+	}
+	seq_puts(m, mark);
+
+	if (verbose_trace)
+		seq_printf(m, "%3lu.%03lu%c ", delay/1000, delay%1000,
+		           (point->flags & IPIPE_TFLG_NMI_HIT) ? 'N' : ' ');
+	else
+		seq_puts(m, " ");
+}
+
+static void __ipipe_print_symname(struct seq_file *m, unsigned long eip)
+{
+	char namebuf[KSYM_NAME_LEN+1];
+	unsigned long size, offset;
+	const char *sym_name;
+	char *modname;
+
+	sym_name = kallsyms_lookup(eip, &size, &offset, &modname, namebuf);
+
+#ifdef CONFIG_IPIPE_TRACE_PANIC
+	if (!m) {
+		/* panic dump */
+		if (sym_name) {
+			printk("%s+0x%lx", sym_name, offset);
+			if (modname)
+				printk(" [%s]", modname);
+		}
+	} else
+#endif /* CONFIG_IPIPE_TRACE_PANIC */
+	{
+		if (sym_name) {
+			if (verbose_trace) {
+				seq_printf(m, "%s+0x%lx", sym_name, offset);
+				if (modname)
+					seq_printf(m, " [%s]", modname);
+			} else
+				seq_puts(m, sym_name);
+		} else
+			seq_printf(m, "<%08lx>", eip);
+	}
+}
+
+static void __ipipe_print_headline(struct seq_file *m)
+{
+	seq_printf(m, "Calibrated minimum trace-point overhead: %lu.%03lu "
+		   "us\n\n", trace_overhead/1000, trace_overhead%1000);
+
+	if (verbose_trace) {
+		const char *name[4] = { [0 ... 3] = "<unused>" };
+		struct list_head *pos;
+		int i = 0;
+
+		list_for_each_prev(pos, &__ipipe_pipeline) {
+			struct ipipe_domain *ipd =
+				list_entry(pos, struct ipipe_domain, p_link);
+
+			name[i] = ipd->name;
+			if (++i > 3)
+				break;
+		}
+
+		seq_printf(m,
+		           " +----- Hard IRQs ('|': locked)\n"
+		           " |+---- %s\n"
+		           " ||+--- %s\n"
+		           " |||+-- %s\n"
+		           " ||||+- %s%s\n"
+		           " |||||                        +---------- "
+		               "Delay flag ('+': > %d us, '!': > %d us)\n"
+		           " |||||                        |        +- "
+		               "NMI noise ('N')\n"
+		           " |||||                        |        |\n"
+		           "      Type    User Val.   Time    Delay  Function "
+		               "(Parent)\n",
+		           name[3], name[2], name[1], name[0],
+		           name[0] ? " ('*': domain stalled, '+': current, "
+		               "'#': current+stalled)" : "",
+		           IPIPE_DELAY_NOTE/1000, IPIPE_DELAY_WARN/1000);
+	} else
+		seq_printf(m,
+		           " +--------------- Hard IRQs ('|': locked)\n"
+		           " |             +- Delay flag "
+		               "('+': > %d us, '!': > %d us)\n"
+		           " |             |\n"
+		           "  Type     Time   Function (Parent)\n",
+		           IPIPE_DELAY_NOTE/1000, IPIPE_DELAY_WARN/1000);
+}
+
+static void *__ipipe_max_prtrace_start(struct seq_file *m, loff_t *pos)
+{
+	loff_t n = *pos;
+
+	mutex_lock(&out_mutex);
+
+	if (!n) {
+		struct ipipe_trace_path *tp;
+		unsigned long length_usecs;
+		int points, cpu;
+		unsigned long flags;
+
+		/* protect against max_path/frozen_path updates while we
+		 * haven't locked our target path, also avoid recursively
+		 * taking global_path_lock from NMI context */
+		flags = __ipipe_global_path_lock();
+
+		/* find the longest of all per-cpu paths */
+		print_path = NULL;
+		for_each_online_cpu(cpu) {
+			tp = &per_cpu(trace_path, cpu)[per_cpu(max_path, cpu)];
+			if ((print_path == NULL) ||
+			    (tp->length > print_path->length)) {
+				print_path = tp;
+				break;
+			}
+		}
+		print_path->dump_lock = 1;
+
+		__ipipe_global_path_unlock(flags);
+
+		/* does this path actually contain data? */
+		if (print_path->end == print_path->begin)
+			return NULL;
+
+		/* number of points inside the critical path */
+		points = WRAP_POINT_NO(print_path->end-print_path->begin+1);
+
+		/* pre- and post-tracing length, post-trace length was frozen
+		   in __ipipe_trace, pre-trace may have to be reduced due to
+		   buffer overrun */
+		print_pre_trace  = pre_trace;
+		print_post_trace = WRAP_POINT_NO(print_path->trace_pos -
+		                                 print_path->end - 1);
+		if (points+pre_trace+print_post_trace > IPIPE_TRACE_POINTS - 1)
+			print_pre_trace = IPIPE_TRACE_POINTS - 1 - points -
+				print_post_trace;
+
+		length_usecs = ipipe_tsc2us(print_path->length);
+		seq_printf(m, "I-pipe worst-case tracing service on %s/ipipe-%s\n"
+			"------------------------------------------------------------\n",
+			UTS_RELEASE, IPIPE_ARCH_STRING);
+		seq_printf(m, "CPU: %d, Begin: %lld cycles, Trace Points: "
+			"%d (-%d/+%d), Length: %lu us\n",
+			cpu, print_path->point[print_path->begin].timestamp,
+			points, print_pre_trace, print_post_trace, length_usecs);
+		__ipipe_print_headline(m);
+	}
+
+	/* check if we are inside the trace range */
+	if (n >= WRAP_POINT_NO(print_path->end - print_path->begin + 1 +
+	                       print_pre_trace + print_post_trace))
+		return NULL;
+
+	/* return the next point to be shown */
+	return &print_path->point[WRAP_POINT_NO(print_path->begin -
+	                                        print_pre_trace + n)];
+}
+
+static void *__ipipe_prtrace_next(struct seq_file *m, void *p, loff_t *pos)
+{
+	loff_t n = ++*pos;
+
+	/* check if we are inside the trace range with the next entry */
+	if (n >= WRAP_POINT_NO(print_path->end - print_path->begin + 1 +
+	                       print_pre_trace + print_post_trace))
+		return NULL;
+
+	/* return the next point to be shown */
+	return &print_path->point[WRAP_POINT_NO(print_path->begin -
+	                                        print_pre_trace + *pos)];
+}
+
+static void __ipipe_prtrace_stop(struct seq_file *m, void *p)
+{
+	if (print_path)
+		print_path->dump_lock = 0;
+	mutex_unlock(&out_mutex);
+}
+
+static int __ipipe_prtrace_show(struct seq_file *m, void *p)
+{
+	long time;
+	struct ipipe_trace_point *point = p;
+	char buf[16];
+
+	if (!point->eip) {
+		seq_puts(m, "-<invalid>-\n");
+		return 0;
+	}
+
+	__ipipe_print_pathmark(m, point);
+	__ipipe_trace_point_type(buf, point);
+	seq_puts(m, buf);
+	if (verbose_trace)
+		switch (point->type & IPIPE_TYPE_MASK) {
+			case IPIPE_TRACE_FUNC:
+				seq_puts(m, "           ");
+				break;
+
+			case IPIPE_TRACE_PID:
+				__ipipe_get_task_info(buf, point, 0);
+				seq_puts(m, buf);
+				break;
+
+			case IPIPE_TRACE_EVENT:
+				__ipipe_get_event_date(buf, print_path, point);
+				seq_puts(m, buf);
+				break;
+
+			default:
+				seq_printf(m, "0x%08lx ", point->v);
+		}
+
+	time = __ipipe_signed_tsc2us(point->timestamp -
+		print_path->point[print_path->begin].timestamp);
+	seq_printf(m, "%5ld", time);
+
+	__ipipe_print_delay(m, point);
+	__ipipe_print_symname(m, point->eip);
+	seq_puts(m, " (");
+	__ipipe_print_symname(m, point->parent_eip);
+	seq_puts(m, ")\n");
+
+	return 0;
+}
+
+static struct seq_operations __ipipe_max_ptrace_ops = {
+	.start = __ipipe_max_prtrace_start,
+	.next  = __ipipe_prtrace_next,
+	.stop  = __ipipe_prtrace_stop,
+	.show  = __ipipe_prtrace_show
+};
+
+static int __ipipe_max_prtrace_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &__ipipe_max_ptrace_ops);
+}
+
+static ssize_t
+__ipipe_max_reset(struct file *file, const char __user *pbuffer,
+                  size_t count, loff_t *data)
+{
+	mutex_lock(&out_mutex);
+	ipipe_trace_max_reset();
+	mutex_unlock(&out_mutex);
+
+	return count;
+}
+
+struct file_operations __ipipe_max_prtrace_fops = {
+	.open       = __ipipe_max_prtrace_open,
+	.read       = seq_read,
+	.write      = __ipipe_max_reset,
+	.llseek     = seq_lseek,
+	.release    = seq_release,
+};
+
+static void *__ipipe_frozen_prtrace_start(struct seq_file *m, loff_t *pos)
+{
+	loff_t n = *pos;
+
+	mutex_lock(&out_mutex);
+
+	if (!n) {
+		struct ipipe_trace_path *tp;
+		int cpu;
+		unsigned long flags;
+
+		/* protect against max_path/frozen_path updates while we
+		 * haven't locked our target path, also avoid recursively
+		 * taking global_path_lock from NMI context */
+		flags = __ipipe_global_path_lock();
+
+		/* find the first of all per-cpu frozen paths */
+		print_path = NULL;
+		for_each_online_cpu(cpu) {
+			tp = &per_cpu(trace_path, cpu)[per_cpu(frozen_path, cpu)];
+			if (tp->end >= 0) {
+				print_path = tp;
+				break;
+			}
+		}
+		if (print_path)
+			print_path->dump_lock = 1;
+
+		__ipipe_global_path_unlock(flags);
+
+		if (!print_path)
+			return NULL;
+
+		/* back- and post-tracing length, post-trace length was frozen
+		   in __ipipe_trace, back-trace may have to be reduced due to
+		   buffer overrun */
+		print_pre_trace  = back_trace-1; /* substract freeze point */
+		print_post_trace = WRAP_POINT_NO(print_path->trace_pos -
+		                                 print_path->end - 1);
+		if (1+pre_trace+print_post_trace > IPIPE_TRACE_POINTS - 1)
+			print_pre_trace = IPIPE_TRACE_POINTS - 2 -
+				print_post_trace;
+
+		seq_printf(m, "I-pipe frozen back-tracing service on %s/ipipe-%s\n"
+			"------------------------------------------------------"
+			"------\n",
+			UTS_RELEASE, IPIPE_ARCH_STRING);
+		seq_printf(m, "CPU: %d, Freeze: %lld cycles, Trace Points: %d (+%d)\n",
+			cpu, print_path->point[print_path->begin].timestamp,
+			print_pre_trace+1, print_post_trace);
+		__ipipe_print_headline(m);
+	}
+
+	/* check if we are inside the trace range */
+	if (n >= print_pre_trace + 1 + print_post_trace)
+		return NULL;
+
+	/* return the next point to be shown */
+	return &print_path->point[WRAP_POINT_NO(print_path->begin-
+	                                        print_pre_trace+n)];
+}
+
+static struct seq_operations __ipipe_frozen_ptrace_ops = {
+	.start = __ipipe_frozen_prtrace_start,
+	.next  = __ipipe_prtrace_next,
+	.stop  = __ipipe_prtrace_stop,
+	.show  = __ipipe_prtrace_show
+};
+
+static int __ipipe_frozen_prtrace_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &__ipipe_frozen_ptrace_ops);
+}
+
+static ssize_t
+__ipipe_frozen_ctrl(struct file *file, const char __user *pbuffer,
+                    size_t count, loff_t *data)
+{
+	char *end, buf[16];
+	int val;
+	int n;
+
+	n = (count > sizeof(buf) - 1) ? sizeof(buf) - 1 : count;
+
+	if (copy_from_user(buf, pbuffer, n))
+		return -EFAULT;
+
+	buf[n] = '\0';
+	val = simple_strtol(buf, &end, 0);
+
+	if (((*end != '\0') && !isspace(*end)) || (val < 0))
+		return -EINVAL;
+
+	mutex_lock(&out_mutex);
+	ipipe_trace_frozen_reset();
+	if (val > 0)
+		ipipe_trace_freeze(-1);
+	mutex_unlock(&out_mutex);
+
+	return count;
+}
+
+struct file_operations __ipipe_frozen_prtrace_fops = {
+	.open       = __ipipe_frozen_prtrace_open,
+	.read       = seq_read,
+	.write      = __ipipe_frozen_ctrl,
+	.llseek     = seq_lseek,
+	.release    = seq_release,
+};
+
+static int __ipipe_rd_proc_val(char *page, char **start, off_t off,
+                               int count, int *eof, void *data)
+{
+	int len;
+
+	len = sprintf(page, "%u\n", *(int *)data);
+	len -= off;
+	if (len <= off + count)
+		*eof = 1;
+	*start = page + off;
+	if (len > count)
+		len = count;
+	if (len < 0)
+		len = 0;
+
+	return len;
+}
+
+static int __ipipe_wr_proc_val(struct file *file, const char __user *buffer,
+                               unsigned long count, void *data)
+{
+	char *end, buf[16];
+	int val;
+	int n;
+
+	n = (count > sizeof(buf) - 1) ? sizeof(buf) - 1 : count;
+
+	if (copy_from_user(buf, buffer, n))
+		return -EFAULT;
+
+	buf[n] = '\0';
+	val = simple_strtol(buf, &end, 0);
+
+	if (((*end != '\0') && !isspace(*end)) || (val < 0))
+		return -EINVAL;
+
+	mutex_lock(&out_mutex);
+	*(int *)data = val;
+	mutex_unlock(&out_mutex);
+
+	return count;
+}
+
+static int __ipipe_rd_trigger(char *page, char **start, off_t off, int count,
+			      int *eof, void *data)
+{
+	int len;
+
+	if (!trigger_begin)
+		return 0;
+
+	len = sprint_symbol(page, trigger_begin);
+	page[len++] = '\n';
+
+	len -= off;
+	if (len <= off + count)
+		*eof = 1;
+	*start = page + off;
+	if (len > count)
+		len = count;
+	if (len < 0)
+		len = 0;
+
+	return len;
+}
+
+static int __ipipe_wr_trigger(struct file *file, const char __user *buffer,
+			      unsigned long count, void *data)
+{
+	char buf[KSYM_SYMBOL_LEN];
+	unsigned long begin, end;
+
+	if (count > sizeof(buf) - 1)
+		count = sizeof(buf) - 1;
+	if (copy_from_user(buf, buffer, count))
+		return -EFAULT;
+	buf[count] = 0;
+	if (buf[count-1] == '\n')
+		buf[count-1] = 0;
+
+	begin = kallsyms_lookup_name(buf);
+	if (!begin || !kallsyms_lookup_size_offset(begin, &end, NULL))
+		return -ENOENT;
+	end += begin - 1;
+
+	mutex_lock(&out_mutex);
+	/* invalidate the current range before setting a new one */
+	trigger_end = 0;
+	wmb();
+	ipipe_trace_frozen_reset();
+
+	/* set new range */
+	trigger_begin = begin;
+	wmb();
+	trigger_end = end;
+	mutex_unlock(&out_mutex);
+
+	return count;
+}
+
+#ifdef CONFIG_IPIPE_TRACE_MCOUNT
+static void notrace
+ipipe_trace_function(unsigned long ip, unsigned long parent_ip)
+{
+	if (!ipipe_trace_enable)
+		return;
+	__ipipe_trace(IPIPE_TRACE_FUNC, ip, parent_ip, 0);
+}
+
+static struct ftrace_ops ipipe_trace_ops = {
+	.func = ipipe_trace_function
+};
+
+static int __ipipe_wr_enable(struct file *file, const char __user *buffer,
+			     unsigned long count, void *data)
+{
+	char *end, buf[16];
+	int val;
+	int n;
+
+	n = (count > sizeof(buf) - 1) ? sizeof(buf) - 1 : count;
+
+	if (copy_from_user(buf, buffer, n))
+		return -EFAULT;
+
+	buf[n] = '\0';
+	val = simple_strtol(buf, &end, 0);
+
+	if (((*end != '\0') && !isspace(*end)) || (val < 0))
+		return -EINVAL;
+
+	mutex_lock(&out_mutex);
+
+	if (ipipe_trace_enable) {
+		if (!val)
+			unregister_ftrace_function(&ipipe_trace_ops);
+	} else if (val)
+		register_ftrace_function(&ipipe_trace_ops);
+
+	ipipe_trace_enable = val;
+
+	mutex_unlock(&out_mutex);
+
+	return count;
+}
+#endif /* CONFIG_IPIPE_TRACE_MCOUNT */
+
+extern struct proc_dir_entry *ipipe_proc_root;
+
+static struct proc_dir_entry * __init
+__ipipe_create_trace_proc_val(struct proc_dir_entry *trace_dir,
+                              const char *name, int *value_ptr)
+{
+	struct proc_dir_entry *entry;
+
+	entry = create_proc_entry(name, 0644, trace_dir);
+	if (entry) {
+		entry->data = value_ptr;
+		entry->read_proc = __ipipe_rd_proc_val;
+		entry->write_proc = __ipipe_wr_proc_val;
+		entry->owner = THIS_MODULE;
+	}
+	return entry;
+}
+
+void __init __ipipe_init_tracer(void)
+{
+	struct proc_dir_entry *trace_dir;
+	struct proc_dir_entry *entry;
+	unsigned long long start, end, min = ULLONG_MAX;
+	int i;
+#ifdef CONFIG_IPIPE_TRACE_VMALLOC
+	int cpu, path;
+
+	for_each_possible_cpu(cpu) {
+		struct ipipe_trace_path *tp_buf;
+
+		tp_buf = vmalloc_node(sizeof(struct ipipe_trace_path) *
+				      IPIPE_TRACE_PATHS, cpu_to_node(cpu));
+		if (!tp_buf) {
+			printk(KERN_ERR "I-pipe: "
+			       "insufficient memory for trace buffer.\n");
+			return;
+		}
+		memset(tp_buf, 0,
+		       sizeof(struct ipipe_trace_path) * IPIPE_TRACE_PATHS);
+		for (path = 0; path < IPIPE_TRACE_PATHS; path++) {
+			tp_buf[path].begin = -1;
+			tp_buf[path].end   = -1;
+		}
+		per_cpu(trace_path, cpu) = tp_buf;
+	}
+#endif /* CONFIG_IPIPE_TRACE_VMALLOC */
+
+	/* Calculate minimum overhead of __ipipe_trace() */
+	local_irq_disable_hw();
+	for (i = 0; i < 100; i++) {
+		ipipe_read_tsc(start);
+		__ipipe_trace(IPIPE_TRACE_FUNC, __BUILTIN_RETURN_ADDRESS0,
+			      __BUILTIN_RETURN_ADDRESS1, 0);
+		ipipe_read_tsc(end);
+
+		end -= start;
+		if (end < min)
+			min = end;
+	}
+	local_irq_enable_hw();
+	trace_overhead = ipipe_tsc2ns(min);
+
+#ifdef CONFIG_IPIPE_TRACE_ENABLE
+	ipipe_trace_enable = 1;
+#ifdef CONFIG_IPIPE_TRACE_MCOUNT
+	register_ftrace_function(&ipipe_trace_ops);
+#endif /* CONFIG_IPIPE_TRACE_MCOUNT */
+#endif /* CONFIG_IPIPE_TRACE_ENABLE */
+
+	trace_dir = create_proc_entry("trace", S_IFDIR, ipipe_proc_root);
+
+	entry = create_proc_entry("max", 0644, trace_dir);
+	if (entry)
+		entry->proc_fops = &__ipipe_max_prtrace_fops;
+
+	entry = create_proc_entry("frozen", 0644, trace_dir);
+	if (entry)
+		entry->proc_fops = &__ipipe_frozen_prtrace_fops;
+
+	entry = create_proc_entry("trigger", 0644, trace_dir);
+	if (entry) {
+		entry->read_proc = __ipipe_rd_trigger;
+		entry->write_proc = __ipipe_wr_trigger;
+		entry->owner = THIS_MODULE;
+	}
+
+	__ipipe_create_trace_proc_val(trace_dir, "pre_trace_points",
+	                              &pre_trace);
+	__ipipe_create_trace_proc_val(trace_dir, "post_trace_points",
+	                              &post_trace);
+	__ipipe_create_trace_proc_val(trace_dir, "back_trace_points",
+	                              &back_trace);
+	__ipipe_create_trace_proc_val(trace_dir, "verbose",
+	                              &verbose_trace);
+	entry = __ipipe_create_trace_proc_val(trace_dir, "enable",
+					      &ipipe_trace_enable);
+#ifdef CONFIG_IPIPE_TRACE_MCOUNT
+	if (entry)
+		entry->write_proc = __ipipe_wr_enable;
+#endif /* CONFIG_IPIPE_TRACE_MCOUNT */
+}
diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 7de11bd..b051136 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -353,8 +353,10 @@ handle_level_irq(unsigned int irq, struct irq_desc *desc)
 	irqreturn_t action_ret;
 
 	spin_lock(&desc->lock);
+#ifndef CONFIG_IPIPE
 	mask_ack_irq(desc, irq);
 	desc = irq_remap_to_desc(irq, desc);
+#endif /* CONFIG_IPIPE */
 
 	if (unlikely(desc->status & IRQ_INPROGRESS))
 		goto out_unlock;
@@ -431,10 +433,14 @@ handle_fasteoi_irq(unsigned int irq, struct irq_desc *desc)
 
 	spin_lock(&desc->lock);
 	desc->status &= ~IRQ_INPROGRESS;
+#ifdef CONFIG_IPIPE
+	desc->chip->unmask(irq);
+out:
+#else
 out:
 	desc->chip->eoi(irq);
+#endif
 	desc = irq_remap_to_desc(irq, desc);
-
 	spin_unlock(&desc->lock);
 }
 
@@ -476,8 +482,10 @@ handle_edge_irq(unsigned int irq, struct irq_desc *desc)
 	kstat_incr_irqs_this_cpu(irq, desc);
 
 	/* Start handling the irq */
+#ifndef CONFIG_IPIPE
 	desc->chip->ack(irq);
 	desc = irq_remap_to_desc(irq, desc);
+#endif /* CONFIG_IPIPE */
 
 	/* Mark the IRQ currently in progress.*/
 	desc->status |= IRQ_INPROGRESS;
@@ -517,6 +525,85 @@ out_unlock:
 	spin_unlock(&desc->lock);
 }
 
+#ifdef CONFIG_IPIPE
+
+void __ipipe_ack_simple_irq(unsigned irq, struct irq_desc *desc)
+{
+}
+
+void __ipipe_end_simple_irq(unsigned irq, struct irq_desc *desc)
+{
+}
+
+void __ipipe_ack_level_irq(unsigned irq, struct irq_desc *desc)
+{
+	mask_ack_irq(desc, irq);
+}
+
+void __ipipe_end_level_irq(unsigned irq, struct irq_desc *desc)
+{
+	if (desc->chip->unmask)
+		desc->chip->unmask(irq);
+}
+
+void __ipipe_ack_fasteoi_irq(unsigned irq, struct irq_desc *desc)
+{
+	desc->chip->eoi(irq);
+}
+
+void __ipipe_end_fasteoi_irq(unsigned irq, struct irq_desc *desc)
+{
+	/*
+	 * Non-requestable IRQs should not be masked in EOI handler.
+	 */
+	if (!(desc->status & IRQ_NOREQUEST))
+		desc->chip->unmask(irq);
+}
+
+void __ipipe_ack_edge_irq(unsigned irq, struct irq_desc *desc)
+{
+	desc->chip->ack(irq);
+}
+
+void __ipipe_ack_percpu_irq(unsigned irq, struct irq_desc *desc)
+{
+	if (desc->chip->ack)
+		desc->chip->ack(irq);
+}
+
+void __ipipe_end_percpu_irq(unsigned irq, struct irq_desc *desc)
+{
+	if (desc->chip->eoi)
+		desc->chip->eoi(irq);
+}
+
+void __ipipe_end_edge_irq(unsigned irq, struct irq_desc *desc)
+{
+}
+
+void __ipipe_ack_bad_irq(unsigned irq, struct irq_desc *desc)
+{
+	static int done;
+
+	handle_bad_irq(irq, desc);
+
+	if (!done) {
+		printk(KERN_WARNING "%s: unknown flow handler for IRQ %d\n",
+		       __FUNCTION__, irq);
+		done = 1;
+	}
+}
+
+void __ipipe_noack_irq(unsigned irq, struct irq_desc *desc)
+{
+}
+
+void __ipipe_noend_irq(unsigned irq, struct irq_desc *desc)
+{
+}
+
+#endif /* CONFIG_IPIPE */
+
 /**
  *	handle_percpu_IRQ - Per CPU local irq handler
  *	@irq:	the interrupt number
@@ -531,8 +618,10 @@ handle_percpu_irq(unsigned int irq, struct irq_desc *desc)
 
 	kstat_incr_irqs_this_cpu(irq, desc);
 
+#ifndef CONFIG_IPIPE
 	if (desc->chip->ack)
 		desc->chip->ack(irq);
+#endif /* CONFIG_IPIPE */
 
 	action_ret = handle_IRQ_event(irq, desc->action);
 	if (!noirqdebug)
@@ -559,6 +648,30 @@ __set_irq_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
 
 	if (!handle)
 		handle = handle_bad_irq;
+#ifdef CONFIG_IPIPE
+	else if (handle == &handle_simple_irq) {
+		desc->ipipe_ack = &__ipipe_ack_simple_irq;
+		desc->ipipe_end = &__ipipe_end_simple_irq;
+	}
+	else if (handle == &handle_level_irq) {
+		desc->ipipe_ack = &__ipipe_ack_level_irq;
+		desc->ipipe_end = &__ipipe_end_level_irq;
+	}
+	else if (handle == &handle_edge_irq) {
+		desc->ipipe_ack = &__ipipe_ack_edge_irq;
+		desc->ipipe_end = &__ipipe_end_edge_irq;
+	}
+	else if (handle == &handle_fasteoi_irq) {
+		desc->ipipe_ack = &__ipipe_ack_fasteoi_irq;
+		desc->ipipe_end = &__ipipe_end_fasteoi_irq;
+	}
+#ifdef CONFIG_SMP
+	else if (handle == &handle_percpu_irq) {
+		desc->ipipe_ack = &__ipipe_ack_percpu_irq;
+		desc->ipipe_end = &__ipipe_end_percpu_irq;
+	}
+#endif /* CONFIG_SMP */
+#endif /* CONFIG_IPIPE */
 	else if (desc->chip == &no_irq_chip) {
 		printk(KERN_WARNING "Trying to install %sinterrupt handler "
 		       "for IRQ%d\n", is_chained ? "chained " : "", irq);
@@ -570,7 +683,21 @@ __set_irq_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
 		 * dummy_irq_chip for easy transition.
 		 */
 		desc->chip = &dummy_irq_chip;
+#ifdef CONFIG_IPIPE
+		desc->ipipe_ack = &__ipipe_noack_irq;
+		desc->ipipe_end = &__ipipe_noend_irq;
+#endif /* CONFIG_IPIPE */
+	}
+#ifdef CONFIG_IPIPE
+ 	else if (is_chained) {
+ 		desc->ipipe_ack = handle;
+ 		desc->ipipe_end = &__ipipe_noend_irq;
+		handle = &__ipipe_noack_irq;
+ 	} else {
+ 		desc->ipipe_ack = &__ipipe_ack_bad_irq;
+ 		desc->ipipe_end = &__ipipe_noend_irq;
 	}
+#endif /* CONFIG_IPIPE */
 
 	spin_lock_irqsave(&desc->lock, flags);
 
@@ -582,9 +709,17 @@ __set_irq_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
 		}
 		desc->status |= IRQ_DISABLED;
 		desc->depth = 1;
+#ifdef CONFIG_IPIPE
+		desc->ipipe_ack = &__ipipe_ack_bad_irq;
+		desc->ipipe_end = &__ipipe_noend_irq;
+#endif /* CONFIG_IPIPE */
 	}
 	desc->handle_irq = handle;
 	desc->name = name;
+#ifdef CONFIG_IPIPE
+	/* Suppress intermediate trampoline routine. */
+	ipipe_root_domain->irqs[irq].acknowledge = desc->ipipe_ack;
+#endif /* CONFIG_IPIPE */
 
 	if (handle != handle_bad_irq && is_chained) {
 		desc->status &= ~IRQ_DISABLED;
diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
index 3aba8d1..381df9d 100644
--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -372,11 +372,13 @@ unsigned int __do_IRQ(unsigned int irq)
 		/*
 		 * No locking required for CPU-local interrupts:
 		 */
+#ifndef CONFIG_IPIPE
 		if (desc->chip->ack) {
 			desc->chip->ack(irq);
 			/* get new one */
 			desc = irq_remap_to_desc(irq, desc);
 		}
+#endif
 		if (likely(!(desc->status & IRQ_DISABLED))) {
 			action_ret = handle_IRQ_event(irq, desc->action);
 			if (!noirqdebug)
@@ -387,10 +389,12 @@ unsigned int __do_IRQ(unsigned int irq)
 	}
 
 	spin_lock(&desc->lock);
+#ifndef CONFIG_IPIPE
 	if (desc->chip->ack) {
 		desc->chip->ack(irq);
 		desc = irq_remap_to_desc(irq, desc);
 	}
+#endif
 	/*
 	 * REPLAY is when Linux resends an IRQ that was dropped earlier
 	 * WAITING is used by probe to mark irqs that are being tested
diff --git a/kernel/lockdep.c b/kernel/lockdep.c
index 06b0c35..6c03fd9 100644
--- a/kernel/lockdep.c
+++ b/kernel/lockdep.c
@@ -2192,7 +2192,7 @@ void trace_hardirqs_on_caller(unsigned long ip)
 	/* we'll do an OFF -> ON transition: */
 	curr->hardirqs_enabled = 1;
 
-	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
+	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled() && !irqs_disabled_hw()))
 		return;
 	if (DEBUG_LOCKS_WARN_ON(current->hardirq_context))
 		return;
@@ -2235,7 +2235,7 @@ void trace_hardirqs_off_caller(unsigned long ip)
 	if (unlikely(!debug_locks || current->lockdep_recursion))
 		return;
 
-	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
+	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled() && !irqs_disabled_hw()))
 		return;
 
 	if (curr->hardirqs_enabled) {
diff --git a/kernel/panic.c b/kernel/panic.c
index 2a2ff36..16007a4 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -15,6 +15,7 @@
 #include <linux/notifier.h>
 #include <linux/init.h>
 #include <linux/sysrq.h>
+#include <linux/ipipe_trace.h>
 #include <linux/interrupt.h>
 #include <linux/nmi.h>
 #include <linux/kexec.h>
@@ -286,6 +287,8 @@ int oops_may_print(void)
  */
 void oops_enter(void)
 {
+	ipipe_trace_panic_freeze();
+	ipipe_disable_context_check(ipipe_processor_id());
 	debug_locks_off(); /* can't trust the integrity of the kernel anymore */
 	do_oops_enter_exit();
 }
diff --git a/kernel/power/disk.c b/kernel/power/disk.c
index 4a4a206..d526845 100644
--- a/kernel/power/disk.c
+++ b/kernel/power/disk.c
@@ -215,6 +215,7 @@ static int create_image(int platform_mode)
 
 	device_pm_lock();
 	local_irq_disable();
+ 	local_irq_disable_hw_cond();
 	/* At this point, device_suspend() has been called, but *not*
 	 * device_power_down(). We *must* call device_power_down() now.
 	 * Otherwise, drivers for some devices (e.g. interrupt controllers)
@@ -256,6 +257,7 @@ static int create_image(int platform_mode)
 	device_power_up(in_suspend ?
 		(error ? PMSG_RECOVER : PMSG_THAW) : PMSG_RESTORE);
  Enable_irqs:
+ 	local_irq_enable_hw_cond();
 	local_irq_enable();
 	device_pm_unlock();
 	return error;
@@ -337,6 +339,7 @@ static int resume_target_kernel(void)
 
 	device_pm_lock();
 	local_irq_disable();
+ 	local_irq_disable_hw_cond();
 	error = device_power_down(PMSG_QUIESCE);
 	if (error) {
 		printk(KERN_ERR "PM: Some devices failed to power down, "
@@ -369,6 +372,7 @@ static int resume_target_kernel(void)
 	sysdev_resume();
 	device_power_up(PMSG_RECOVER);
  Enable_irqs:
+	local_irq_enable_hw_cond();
 	local_irq_enable();
 	device_pm_unlock();
 	return error;
diff --git a/kernel/printk.c b/kernel/printk.c
index 63f703c..ed60d77 100644
--- a/kernel/printk.c
+++ b/kernel/printk.c
@@ -637,6 +637,41 @@ static int have_callable_console(void)
 	return 0;
 }
 
+#ifdef CONFIG_IPIPE
+
+static ipipe_spinlock_t __ipipe_printk_lock = IPIPE_SPIN_LOCK_UNLOCKED;
+
+static int __ipipe_printk_fill;
+
+static char __ipipe_printk_buf[__LOG_BUF_LEN];
+
+void __ipipe_flush_printk (unsigned virq, void *cookie)
+{
+	char *p = __ipipe_printk_buf;
+	int len, lmax, out = 0;
+	unsigned long flags;
+
+	goto start;
+
+	do {
+		spin_unlock_irqrestore(&__ipipe_printk_lock, flags);
+ start:
+		lmax = __ipipe_printk_fill;
+		while (out < lmax) {
+			len = strlen(p) + 1;
+			printk("%s",p);
+			p += len;
+			out += len;
+		}
+		spin_lock_irqsave(&__ipipe_printk_lock, flags);
+	}
+	while (__ipipe_printk_fill != lmax);
+
+	__ipipe_printk_fill = 0;
+
+	spin_unlock_irqrestore(&__ipipe_printk_lock, flags);
+}
+
 /**
  * printk - print a kernel message
  * @fmt: format string
@@ -661,6 +696,65 @@ static int have_callable_console(void)
 
 asmlinkage int printk(const char *fmt, ...)
 {
+	int r, fbytes, oldcount;
+	unsigned long flags;
+	int sprintk = 1;
+	int cs = -1;
+	va_list args;
+
+	va_start(args, fmt);
+
+	local_irq_save_hw(flags);
+
+	if (test_bit(IPIPE_SPRINTK_FLAG, &__ipipe_current_domain->flags) ||
+	    oops_in_progress)
+		cs = ipipe_disable_context_check(ipipe_processor_id());
+	else if (__ipipe_current_domain == ipipe_root_domain) {
+		struct ipipe_domain *dom;
+
+		list_for_each_entry(dom, &__ipipe_pipeline, p_link) {
+			if (dom == ipipe_root_domain)
+				break;
+			if (test_bit(IPIPE_STALL_FLAG,
+				     &ipipe_cpudom_var(dom, status)))
+				sprintk = 0;
+		}
+	} else
+		sprintk = 0;
+
+	local_irq_restore_hw(flags);
+
+	if (sprintk) {
+		r = vprintk(fmt, args);
+		if (cs != -1)
+			ipipe_restore_context_check(ipipe_processor_id(), cs);
+		goto out;
+	}
+
+	spin_lock_irqsave(&__ipipe_printk_lock, flags);
+
+	oldcount = __ipipe_printk_fill;
+	fbytes = __LOG_BUF_LEN - oldcount;
+
+	if (fbytes > 1)	{
+		r = vscnprintf(__ipipe_printk_buf + __ipipe_printk_fill,
+			       fbytes, fmt, args) + 1; /* account for the null byte */
+		__ipipe_printk_fill += r;
+	} else
+		r = 0;
+
+	spin_unlock_irqrestore(&__ipipe_printk_lock, flags);
+
+	if (oldcount == 0)
+		ipipe_trigger_irq(__ipipe_printk_virq);
+out:
+	va_end(args);
+
+	return r;
+}
+#else /* !CONFIG_IPIPE */
+asmlinkage int printk(const char *fmt, ...)
+{
 	va_list args;
 	int r;
 
@@ -670,6 +764,7 @@ asmlinkage int printk(const char *fmt, ...)
 
 	return r;
 }
+#endif /* CONFIG_IPIPE */
 
 /* cpu currently holding logbuf_lock */
 static volatile unsigned int printk_cpu = UINT_MAX;
diff --git a/kernel/sched.c b/kernel/sched.c
index 7d13deb..e371aa3 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2294,7 +2294,7 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state, int sync)
 	rq = task_rq_lock(p, &flags);
 	update_rq_clock(rq);
 	old_state = p->state;
-	if (!(old_state & state))
+	if (!(old_state & state) || (old_state & (TASK_NOWAKEUP|TASK_ATOMICSWITCH)))
 		goto out;
 
 	if (p->se.on_rq)
@@ -2647,22 +2647,29 @@ asmlinkage void schedule_tail(struct tas
 #endif
 	if (current->set_child_tid)
 		put_user(task_pid_vnr(current), current->set_child_tid);
+
+ 	ipipe_init_notify(current);
 }
 
 /*
  * context_switch - switch to the new MM and the new
  * thread's register state.
  */
-static inline void
+int
 context_switch(struct rq *rq, struct task_struct *prev,
 	       struct task_struct *next)
 {
 	struct mm_struct *mm, *oldmm;
 
-	prepare_task_switch(rq, prev, next);
-	trace_sched_switch(rq, prev, next);
 	mm = next->mm;
 	oldmm = prev->active_mm;
+
+if (!rq) {
+	switch_mm(oldmm, next->active_mm, next);
+	if (!mm) enter_lazy_tlb(oldmm, next);
+} else {
+	prepare_task_switch(rq, prev, next);
+	trace_sched_switch(rq, prev, next);
 	/*
 	 * For paravirt, this is coupled with an exit in switch_to to
 	 * combine the page table reload and the switch backend into
@@ -2690,11 +2697,24 @@ context_switch(struct rq *rq, struct tas
 #ifndef __ARCH_WANT_UNLOCKED_CTXSW
 	spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
 #endif
-
+}
+#ifdef CONFIG_IPIPE
+	next->ptd[IPIPE_ROOT_NPTDKEYS - 1] = prev;
+#endif /* CONFIG_IPIPE */
 	/* Here we just switch the register state and the stack. */
 	switch_to(prev, next, prev);
 
 	barrier();
+
+if (unlikely(rq)) {
+#if 1 //def CONFIG_IPIPE_DELAYED_ATOMICSW
+	current->state &= ~TASK_ATOMICSWITCH;
+#else
+	prev->state &= ~TASK_ATOMICSWITCH;
+#endif
+	if (task_hijacked(prev))
+		return 1; __ipipe_dispatch_event(IPIPE_FIRST_EVENT - 2, 0);
+
 	/*
 	 * this_rq must be evaluated again because prev may have moved
 	 * CPUs since it called schedule(), thus the 'rq' on its stack
@@ -2702,6 +2722,10 @@ context_switch(struct rq *rq, struct tas
 	 */
 	finish_task_switch(this_rq(), prev);
 }
+	return 0;
+}
+
+EXPORT_SYMBOL(context_switch);
 
 /*
  * nr_running, nr_uninterruptible and nr_context_switches:
@@ -4498,6 +4522,7 @@ EXPORT_SYMBOL(add_preempt_count);
 
 void __kprobes sub_preempt_count(int val)
 {
+ 	ipipe_check_context(ipipe_root_domain);
 #ifdef CONFIG_DEBUG_PREEMPT
 	/*
 	 * Underflow?
@@ -4546,6 +4571,7 @@ static noinline void __schedule_bug(stru
  */
 static inline void schedule_debug(struct task_struct *prev)
 {
+	ipipe_check_context(ipipe_root_domain);
 	/*
 	 * Test if we are atomic. Since do_exit() needs to call into
 	 * schedule() atomically, we ignore that path for now.
@@ -4614,6 +4640,9 @@ need_resched:
 	rcu_qsctr_inc(cpu);
 	prev = rq->curr;
 	switch_count = &prev->nivcsw;
+ 	if (unlikely(prev->state & TASK_ATOMICSWITCH))
+		/* Pop one disable level -- one still remains. */
+		preempt_enable();
 
 	release_kernel_lock(prev);
 need_resched_nonpreemptible:
@@ -4653,15 +4682,18 @@ need_resched_nonpreemptible:
 		rq->curr = next;
 		++*switch_count;
 
-		context_switch(rq, prev, next); /* unlocks the rq */
+		if (context_switch(rq, prev, next)) /* unlocks the rq unless hijacked */
+			return;
 		/*
 		 * the context switch might have flipped the stack from under
 		 * us, hence refresh the local variables.
 		 */
 		cpu = smp_processor_id();
 		rq = cpu_rq(cpu);
-	} else
+	} else {
+		prev->state &= ~TASK_ATOMICSWITCH;
 		spin_unlock_irq(&rq->lock);
+	}
 
 	if (unlikely(reacquire_kernel_lock(current) < 0))
 		goto need_resched_nonpreemptible;
@@ -5419,6 +5451,7 @@ recheck:
 
 	oldprio = p->prio;
 	__setscheduler(rq, p, policy, param->sched_priority);
+  	ipipe_setsched_notify(p);
 
 	if (running)
 		p->sched_class->set_curr_task(rq);
@@ -5766,6 +5799,7 @@ static void __cond_resched(void)
 #ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
 	__might_sleep(__FILE__, __LINE__);
 #endif
+	ipipe_check_context(ipipe_root_domain);
 	/*
 	 * The BKS might be reacquired before we have dropped
 	 * PREEMPT_ACTIVE, which could trigger a second
@@ -9675,3 +9709,60 @@ struct cgroup_subsys cpuacct_subsys = {
 	.subsys_id = cpuacct_subsys_id,
 };
 #endif	/* CONFIG_CGROUP_CPUACCT */
+
+#ifdef CONFIG_IPIPE
+
+int ipipe_setscheduler_root (struct task_struct *p, int policy, int prio)
+{
+	const struct sched_class *prev_class = p->sched_class;
+	int oldprio, on_rq, running;
+	unsigned long flags;
+	struct rq *rq;
+
+	spin_lock_irqsave(&p->pi_lock, flags);
+	rq = __task_rq_lock(p);
+	update_rq_clock(rq);
+	on_rq = p->se.on_rq;
+	running = task_running(rq, p);
+
+	if (on_rq)
+		deactivate_task(rq, p, 0);
+	if (running)
+		p->sched_class->put_prev_task(rq, p);
+
+	oldprio = p->prio;
+	__setscheduler(rq, p, policy, prio);
+	ipipe_setsched_notify(p);
+
+	if (running)
+		p->sched_class->set_curr_task(rq);
+	if (on_rq) {
+		activate_task(rq, p, 0);
+		check_class_changed(rq, p, prev_class, oldprio, running);
+	}
+	__task_rq_unlock(rq);
+	spin_unlock_irqrestore(&p->pi_lock, flags);
+
+	rt_mutex_adjust_pi(p);
+
+	return 0;
+}
+
+EXPORT_SYMBOL(ipipe_setscheduler_root);
+
+int ipipe_reenter_root (struct task_struct *prev, int policy, int prio)
+{
+	finish_task_switch(this_rq(), prev);
+
+	(void)reacquire_kernel_lock(current);
+	preempt_enable_no_resched();
+
+	if (current->policy != policy || current->rt_priority != prio)
+		return ipipe_setscheduler_root(current, policy, prio);
+
+	return 0;
+}
+
+EXPORT_SYMBOL(ipipe_reenter_root);
+
+#endif /* CONFIG_IPIPE */
diff --git a/kernel/signal.c b/kernel/signal.c
index 1c88144..e0dcd3d 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -504,6 +504,7 @@ void signal_wake_up(struct task_struct *t, int resume)
 	unsigned int mask;
 
 	set_tsk_thread_flag(t, TIF_SIGPENDING);
+	ipipe_sigwake_notify(t); /* TIF_SIGPENDING must be set first. */
 
 	/*
 	 * For SIGKILL, we want to wake it up in the stopped/traced/killable
diff --git a/kernel/spinlock.c b/kernel/spinlock.c
index 29ab207..c21e314 100644
--- a/kernel/spinlock.c
+++ b/kernel/spinlock.c
@@ -87,7 +87,7 @@ unsigned long __lockfunc _spin_lock_irqsave(spinlock_t *lock)
 	 * _raw_spin_lock_flags() code, because lockdep assumes
 	 * that interrupts are not re-enabled during lock-acquire:
 	 */
-#ifdef CONFIG_LOCKDEP
+#if defined(CONFIG_LOCKDEP) || defined(CONFIG_IPIPE)
 	LOCK_CONTENDED(lock, _raw_spin_trylock, _raw_spin_lock);
 #else
 	_raw_spin_lock_flags(lock, &flags);
@@ -304,7 +304,7 @@ unsigned long __lockfunc _spin_lock_irqsave_nested(spinlock_t *lock, int subclas
 	 * _raw_spin_lock_flags() code, because lockdep assumes
 	 * that interrupts are not re-enabled during lock-acquire:
 	 */
-#ifdef CONFIG_LOCKDEP
+#if defined(CONFIG_LOCKDEP) || defined(CONFIG_IPIPE)
 	LOCK_CONTENDED(lock, _raw_spin_trylock, _raw_spin_lock);
 #else
 	_raw_spin_lock_flags(lock, &flags);
diff --git a/kernel/time/tick-common.c b/kernel/time/tick-common.c
index 83c4417..782a209 100644
--- a/kernel/time/tick-common.c
+++ b/kernel/time/tick-common.c
@@ -69,7 +69,7 @@ static void tick_periodic(int cpu)
 		write_sequnlock(&xtime_lock);
 	}
 
-	update_process_times(user_mode(get_irq_regs()));
+	update_root_process_times(get_irq_regs());
 	profile_tick(CPU_PROFILING);
 }
 
@@ -177,6 +177,10 @@ static void tick_setup_device(struct tick_device *td,
 
 	td->evtdev = newdev;
 
+	/* I-pipe: derive global tick IRQ from CPU 0 */
+	if (cpu == 0)
+		ipipe_update_tick_evtdev(newdev);
+
 	/*
 	 * When the device is not per cpu, pin the interrupt to the
 	 * current cpu:
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index d3f1ef4..6de336b 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -514,7 +514,7 @@ static void tick_nohz_handler(struct clock_event_device *dev)
 		ts->idle_jiffies++;
 	}
 
-	update_process_times(user_mode(regs));
+	update_root_process_times(regs);
 	profile_tick(CPU_PROFILING);
 
 	while (tick_nohz_reprogram(ts, now)) {
@@ -665,7 +665,7 @@ static enum hrtimer_restart tick_sched_timer(struct hrtimer *timer)
 			touch_softlockup_watchdog();
 			ts->idle_jiffies++;
 		}
-		update_process_times(user_mode(regs));
+		update_root_process_times(regs);
 		profile_tick(CPU_PROFILING);
 	}
 
diff --git a/kernel/timer.c b/kernel/timer.c
index 13dd64f..6769b8e 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -1037,6 +1037,26 @@ void update_process_times(int user_tick)
 	run_posix_cpu_timers(p);
 }
 
+#ifdef CONFIG_IPIPE
+
+void update_root_process_times(struct pt_regs *regs)
+{
+	int cpu, user_tick = user_mode(regs);
+
+	if (__ipipe_root_tick_p(regs)) {
+		update_process_times(user_tick);
+		return;
+	}
+
+	run_local_timers();
+	cpu = smp_processor_id();
+	if (rcu_pending(cpu))
+		rcu_check_callbacks(cpu, user_tick);
+	run_posix_cpu_timers(current);
+}
+
+#endif
+
 /*
  * Nr of active tasks - counted in fixed-point numbers
  */
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index 34e707e..c8d7841 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -271,6 +271,7 @@ config DYNAMIC_FTRACE
 	depends on FUNCTION_TRACER
 	depends on HAVE_DYNAMIC_FTRACE
 	depends on DEBUG_KERNEL
+ 	depends on !IPIPE_TRACE_MCOUNT
 	default y
 	help
          This option will modify all the calls to ftrace dynamically
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 1bcf9cd..eabe65b 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -128,6 +128,8 @@ config DEBUG_SECTION_MISMATCH
 	  - Enable verbose reporting from modpost to help solving
 	    the section mismatches reported.
 
+source "kernel/ipipe/Kconfig.debug"
+
 config DEBUG_KERNEL
 	bool "Kernel debugging"
 	help
diff --git a/lib/bust_spinlocks.c b/lib/bust_spinlocks.c
index 9681d54..0dd7d52 100644
--- a/lib/bust_spinlocks.c
+++ b/lib/bust_spinlocks.c
@@ -8,6 +8,7 @@
  */
 
 #include <linux/kernel.h>
+#include <linux/ipipe_trace.h>
 #include <linux/spinlock.h>
 #include <linux/tty.h>
 #include <linux/wait.h>
@@ -24,6 +25,7 @@ void __attribute__((weak)) bust_spinlocks(int yes)
 		unblank_screen();
 #endif
 		console_unblank();
+  		ipipe_trace_panic_dump();
 		if (--oops_in_progress == 0)
 			wake_up_klogd();
 	}
diff --git a/lib/ioremap.c b/lib/ioremap.c
index 14c6078..a275469 100644
--- a/lib/ioremap.c
+++ b/lib/ioremap.c
@@ -85,8 +85,8 @@ int ioremap_page_range(unsigned long addr,
 		if (err)
 			break;
 	} while (pgd++, addr = next, addr != end);
-
-	flush_cache_vmap(start, end);
+	__ipipe_pin_range_globally(start, end);
+ 	flush_cache_vmap(start, end);
 
 	return err;
 }
diff --git a/lib/smp_processor_id.c b/lib/smp_processor_id.c
index 4689cb0..3d12764 100644
--- a/lib/smp_processor_id.c
+++ b/lib/smp_processor_id.c
@@ -12,10 +12,13 @@ notrace unsigned int debug_smp_processor_id(void)
 	unsigned long preempt_count = preempt_count();
 	int this_cpu = raw_smp_processor_id();
 
+	if (!ipipe_root_domain_p)
+		goto out;
+
 	if (likely(preempt_count))
 		goto out;
 
-	if (irqs_disabled())
+	if (irqs_disabled() || irqs_disabled_hw())
 		goto out;
 
 	/*
diff --git a/lib/spinlock_debug.c b/lib/spinlock_debug.c
index 9c4b025..08f096b 100644
--- a/lib/spinlock_debug.c
+++ b/lib/spinlock_debug.c
@@ -133,6 +133,8 @@ void _raw_spin_lock(spinlock_t *lock)
 	debug_spin_lock_after(lock);
 }
 
+EXPORT_SYMBOL(_raw_spin_lock);
+
 int _raw_spin_trylock(spinlock_t *lock)
 {
 	int ret = __raw_spin_trylock(&lock->raw_lock);
@@ -148,12 +150,16 @@ int _raw_spin_trylock(spinlock_t *lock)
 	return ret;
 }
 
+EXPORT_SYMBOL(_raw_spin_trylock);
+
 void _raw_spin_unlock(spinlock_t *lock)
 {
 	debug_spin_unlock(lock);
 	__raw_spin_unlock(&lock->raw_lock);
 }
 
+EXPORT_SYMBOL(_raw_spin_unlock);
+
 static void rwlock_bug(rwlock_t *lock, const char *msg)
 {
 	if (!debug_locks_off())
@@ -199,6 +205,8 @@ void _raw_read_lock(rwlock_t *lock)
 	__raw_read_lock(&lock->raw_lock);
 }
 
+EXPORT_SYMBOL(_raw_read_lock);
+
 int _raw_read_trylock(rwlock_t *lock)
 {
 	int ret = __raw_read_trylock(&lock->raw_lock);
@@ -212,12 +220,16 @@ int _raw_read_trylock(rwlock_t *lock)
 	return ret;
 }
 
+EXPORT_SYMBOL(_raw_read_trylock);
+
 void _raw_read_unlock(rwlock_t *lock)
 {
 	RWLOCK_BUG_ON(lock->magic != RWLOCK_MAGIC, lock, "bad magic");
 	__raw_read_unlock(&lock->raw_lock);
 }
 
+EXPORT_SYMBOL(_raw_read_unlock);
+
 static inline void debug_write_lock_before(rwlock_t *lock)
 {
 	RWLOCK_BUG_ON(lock->magic != RWLOCK_MAGIC, lock, "bad magic");
@@ -275,6 +287,8 @@ void _raw_write_lock(rwlock_t *lock)
 	debug_write_lock_after(lock);
 }
 
+EXPORT_SYMBOL(_raw_write_lock);
+
 int _raw_write_trylock(rwlock_t *lock)
 {
 	int ret = __raw_write_trylock(&lock->raw_lock);
@@ -290,8 +304,12 @@ int _raw_write_trylock(rwlock_t *lock)
 	return ret;
 }
 
+EXPORT_SYMBOL(_raw_write_trylock);
+
 void _raw_write_unlock(rwlock_t *lock)
 {
 	debug_write_unlock(lock);
 	__raw_write_unlock(&lock->raw_lock);
 }
+
+EXPORT_SYMBOL(_raw_write_unlock);
diff --git a/mm/memory.c b/mm/memory.c
index c304626..f433db1 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -52,6 +52,7 @@
 #include <linux/writeback.h>
 #include <linux/memcontrol.h>
 #include <linux/mmu_notifier.h>
+#include <linux/vmalloc.h>
 #include <linux/kallsyms.h>
 #include <linux/swapops.h>
 #include <linux/elf.h>
@@ -532,6 +533,32 @@ out:
 	return pfn_to_page(pfn);
 }
 
+static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
+{
+	/*
+	 * If the source page was a PFN mapping, we don't have
+	 * a "struct page" for it. We do a best-effort copy by
+	 * just copying from the original user address. If that
+	 * fails, we just zero-fill it. Live with it.
+	 */
+	if (unlikely(!src)) {
+		void *kaddr = kmap_atomic(dst, KM_USER0);
+		void __user *uaddr = (void __user *)(va & PAGE_MASK);
+
+		/*
+		 * This really shouldn't fail, because the page is there
+		 * in the page tables. But it might just be unreadable,
+		 * in which case we just give up and fill the result with
+		 * zeroes.
+		 */
+		if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE))
+			memset(kaddr, 0, PAGE_SIZE);
+		kunmap_atomic(kaddr, KM_USER0);
+		flush_dcache_page(dst);
+	} else
+		copy_user_highpage(dst, src, va, vma);
+}
+
 /*
  * copy one vm_area from one task to the other. Assumes the page tables
  * already present in the new task to be cleared in the whole range
@@ -540,8 +567,8 @@ out:
 
 static inline void
 copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
-		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
-		unsigned long addr, int *rss)
+	     pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
+	     unsigned long addr, int *rss, struct page *uncow_page)
 {
 	unsigned long vm_flags = vma->vm_flags;
 	pte_t pte = *src_pte;
@@ -580,6 +607,21 @@ copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
 	 * in the parent and the child
 	 */
 	if (is_cow_mapping(vm_flags)) {
+#ifdef CONFIG_IPIPE
+		if (uncow_page) {
+			struct page *old_page = vm_normal_page(vma, addr, pte);
+			cow_user_page(uncow_page, old_page, addr, vma);
+			pte = mk_pte(uncow_page, vma->vm_page_prot);
+
+			if (vm_flags & VM_SHARED)
+				pte = pte_mkclean(pte);
+			pte = pte_mkold(pte);
+
+			page_add_new_anon_rmap(uncow_page, vma, addr);
+			rss[!!PageAnon(uncow_page)]++;
+			goto out_set_pte;
+		}
+#endif /* CONFIG_IPIPE */
 		ptep_set_wrprotect(src_mm, addr, src_pte);
 		pte = pte_wrprotect(pte);
 	}
@@ -610,13 +652,27 @@ static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
 	pte_t *src_pte, *dst_pte;
 	spinlock_t *src_ptl, *dst_ptl;
 	int progress = 0;
+	struct page *uncow_page = NULL;
 	int rss[2];
-
+#ifdef CONFIG_IPIPE
+	int do_cow_break = 0;
+again:
+	if (do_cow_break) {
+		uncow_page = alloc_page_vma(GFP_HIGHUSER, vma, addr);
+		if (!uncow_page)
+			return -ENOMEM;
+		do_cow_break = 0;
+	}
+#else
 again:
+#endif
 	rss[1] = rss[0] = 0;
 	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
-	if (!dst_pte)
+	if (!dst_pte) {
+		if (uncow_page)
+			page_cache_release(uncow_page);
 		return -ENOMEM;
+	}
 	src_pte = pte_offset_map_nested(src_pmd, addr);
 	src_ptl = pte_lockptr(src_mm, src_pmd);
 	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
@@ -637,7 +693,26 @@ again:
 			progress++;
 			continue;
 		}
-		copy_one_pte(dst_mm, src_mm, dst_pte, src_pte, vma, addr, rss);
+#ifdef CONFIG_IPIPE
+		if (likely(uncow_page == NULL) && likely(pte_present(*src_pte))) {
+			if (is_cow_mapping(vma->vm_flags)) {
+				if (((vma->vm_flags|src_mm->def_flags) & (VM_LOCKED|VM_PINNED))
+				    == (VM_LOCKED|VM_PINNED)) {
+				    	arch_leave_lazy_mmu_mode();
+				    	spin_unlock(src_ptl);
+				    	pte_unmap_nested(src_pte);
+				    	add_mm_rss(dst_mm, rss[0], rss[1]);
+				    	pte_unmap_unlock(dst_pte, dst_ptl);
+				    	cond_resched();
+					do_cow_break = 1;
+					goto again;
+				}
+			}
+		}
+#endif
+		copy_one_pte(dst_mm, src_mm, dst_pte,
+			     src_pte, vma, addr, rss, uncow_page);
+		uncow_page = NULL;
 		progress += 8;
 	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);
 
@@ -1843,32 +1918,6 @@ static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 	return pte;
 }
 
-static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
-{
-	/*
-	 * If the source page was a PFN mapping, we don't have
-	 * a "struct page" for it. We do a best-effort copy by
-	 * just copying from the original user address. If that
-	 * fails, we just zero-fill it. Live with it.
-	 */
-	if (unlikely(!src)) {
-		void *kaddr = kmap_atomic(dst, KM_USER0);
-		void __user *uaddr = (void __user *)(va & PAGE_MASK);
-
-		/*
-		 * This really shouldn't fail, because the page is there
-		 * in the page tables. But it might just be unreadable,
-		 * in which case we just give up and fill the result with
-		 * zeroes.
-		 */
-		if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE))
-			memset(kaddr, 0, PAGE_SIZE);
-		kunmap_atomic(kaddr, KM_USER0);
-		flush_dcache_page(dst);
-	} else
-		copy_user_highpage(dst, src, va, vma);
-}
-
 /*
  * This routine handles present pages, when users try to write
  * to a shared page. It is done by copying the page to a new address
@@ -3233,6 +3282,113 @@ void print_vma_addr(char *prefix, unsigned long ip)
 	up_read(&current->mm->mmap_sem);
 }
 
+#ifdef CONFIG_IPIPE
+
+static inline int ipipe_pin_pte_range(struct mm_struct *mm, pmd_t *pmd,
+				      struct vm_area_struct *vma,
+				      unsigned long addr, unsigned long end)
+{
+	spinlock_t *ptl;
+	pte_t *pte;
+
+	do {
+		pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
+		if (!pte)
+			continue;
+
+		if (!pte_present(*pte) || pte_write(*pte)) {
+			pte_unmap_unlock(pte, ptl);
+			continue;
+		}
+
+		if (do_wp_page(mm, vma, addr, pte, pmd, ptl, *pte) == VM_FAULT_OOM)
+			return -ENOMEM;
+	} while (addr += PAGE_SIZE, addr != end);
+	return 0;
+}
+
+static inline int ipipe_pin_pmd_range(struct mm_struct *mm, pud_t *pud,
+				      struct vm_area_struct *vma,
+				      unsigned long addr, unsigned long end)
+{
+	unsigned long next;
+	pmd_t *pmd;
+
+	pmd = pmd_offset(pud, addr);
+	do {
+		next = pmd_addr_end(addr, end);
+		if (pmd_none_or_clear_bad(pmd))
+			continue;
+		if (ipipe_pin_pte_range(mm, pmd, vma, addr, next))
+			return -ENOMEM;
+	} while (pmd++, addr = next, addr != end);
+	return 0;
+}
+
+static inline int ipipe_pin_pud_range(struct mm_struct *mm, pgd_t *pgd,
+				      struct vm_area_struct *vma,
+				      unsigned long addr, unsigned long end)
+{
+	unsigned long next;
+	pud_t *pud;
+
+	pud = pud_offset(pgd, addr);
+	do {
+		next = pud_addr_end(addr, end);
+		if (pud_none_or_clear_bad(pud))
+			continue;
+		if (ipipe_pin_pmd_range(mm, pud, vma, addr, next))
+			return -ENOMEM;
+	} while (pud++, addr = next, addr != end);
+	return 0;
+}
+
+int ipipe_disable_ondemand_mappings(struct task_struct *tsk)
+{
+	unsigned long addr, next, end;
+	struct vm_area_struct *vma;
+	struct mm_struct *mm;
+	int result = 0;
+	pgd_t *pgd;
+
+	mm = get_task_mm(tsk);
+	if (!mm)
+		return -EPERM;
+
+	down_write(&mm->mmap_sem);
+	if (mm->def_flags & VM_PINNED)
+		goto done_mm;
+
+	for (vma = mm->mmap; vma; vma = vma->vm_next) {
+		if (!is_cow_mapping(vma->vm_flags)
+		    || !(vma->vm_flags & VM_WRITE))
+			continue;
+
+	addr = vma->vm_start;
+ 		end = vma->vm_end;
+
+		pgd = pgd_offset(mm, addr);
+		do {
+ 			next = pgd_addr_end(addr, end);
+			if (pgd_none_or_clear_bad(pgd))
+				continue;
+			if (ipipe_pin_pud_range(mm, pgd, vma, addr, next)) {
+				result = -ENOMEM;
+				goto done_mm;
+			}
+		} while (pgd++, addr = next, addr != end);
+	}
+	mm->def_flags |= VM_PINNED;
+
+   done_mm:
+	up_write(&mm->mmap_sem);
+	mmput(mm);
+	return result;
+}
+EXPORT_SYMBOL(ipipe_disable_ondemand_mappings);
+
+#endif
+
 #ifdef CONFIG_PROVE_LOCKING
 void might_fault(void)
 {
diff --git a/mm/mlock.c b/mm/mlock.c
index cbe9e05..b9d0ed0 100644
--- a/mm/mlock.c
+++ b/mm/mlock.c
@@ -533,10 +533,10 @@ SYSCALL_DEFINE2(munlock, unsigned long, start, size_t, len)
 static int do_mlockall(int flags)
 {
 	struct vm_area_struct * vma, * prev = NULL;
-	unsigned int def_flags = 0;
+	unsigned int def_flags = current->mm->def_flags & VM_PINNED;
 
 	if (flags & MCL_FUTURE)
-		def_flags = VM_LOCKED;
+		def_flags |= VM_LOCKED;
 	current->mm->def_flags = def_flags;
 	if (flags == MCL_FUTURE)
 		goto out;
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 520a759..f123558 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -173,6 +173,9 @@ static int vmap_page_range(unsigned long start, unsigned long end,
 
 	if (unlikely(err))
 		return err;
+
+ 	__ipipe_pin_range_globally(start, end);
+
 	return nr;
 }