Pull pvops into release branch

20 files changed, 1506 insertions, 174 deletions

diff --git a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
index f2778f2c4fd..5628e9a990a 100644
--- a/arch/ia64/kernel/Makefile
+++ b/arch/ia64/kernel/Makefile
@@ -5,7 +5,7 @@
 extra-y	:= head.o init_task.o vmlinux.lds
 
 obj-y := acpi.o entry.o efi.o efi_stub.o gate-data.o fsys.o ia64_ksyms.o irq.o irq_ia64.o	\
-	 irq_lsapic.o ivt.o machvec.o pal.o patch.o process.o perfmon.o ptrace.o sal.o		\
+	 irq_lsapic.o ivt.o machvec.o pal.o paravirt_patchlist.o patch.o process.o perfmon.o ptrace.o sal.o		\
 	 salinfo.o setup.o signal.o sys_ia64.o time.o traps.o unaligned.o \
 	 unwind.o mca.o mca_asm.o topology.o dma-mapping.o
 
@@ -36,7 +36,8 @@ obj-$(CONFIG_PCI_MSI)		+= msi_ia64.o
 mca_recovery-y			+= mca_drv.o mca_drv_asm.o
 obj-$(CONFIG_IA64_MC_ERR_INJECT)+= err_inject.o
 
-obj-$(CONFIG_PARAVIRT)		+= paravirt.o paravirtentry.o
+obj-$(CONFIG_PARAVIRT)		+= paravirt.o paravirtentry.o \
+				   paravirt_patch.o
 
 obj-$(CONFIG_IA64_ESI)		+= esi.o
 ifneq ($(CONFIG_IA64_ESI),)
@@ -45,35 +46,13 @@ endif
 obj-$(CONFIG_DMAR)		+= pci-dma.o
 obj-$(CONFIG_SWIOTLB)		+= pci-swiotlb.o
 
-# The gate DSO image is built using a special linker script.
-targets += gate.so gate-syms.o
-
-extra-y += gate.so gate-syms.o gate.lds gate.o
-
 # fp_emulate() expects f2-f5,f16-f31 to contain the user-level state.
 CFLAGS_traps.o  += -mfixed-range=f2-f5,f16-f31
 
-CPPFLAGS_gate.lds := -P -C -U$(ARCH)
-
-quiet_cmd_gate = GATE $@
-      cmd_gate = $(CC) -nostdlib $(GATECFLAGS_$(@F)) -Wl,-T,$(filter-out FORCE,$^) -o $@
-
-GATECFLAGS_gate.so = -shared -s -Wl,-soname=linux-gate.so.1 \
-		     $(call ld-option, -Wl$(comma)--hash-style=sysv)
-$(obj)/gate.so: $(obj)/gate.lds $(obj)/gate.o FORCE
-	$(call if_changed,gate)
-
-$(obj)/built-in.o: $(obj)/gate-syms.o
-$(obj)/built-in.o: ld_flags += -R $(obj)/gate-syms.o
-
-GATECFLAGS_gate-syms.o = -r
-$(obj)/gate-syms.o: $(obj)/gate.lds $(obj)/gate.o FORCE
-	$(call if_changed,gate)
-
-# gate-data.o contains the gate DSO image as data in section .data.gate.
-# We must build gate.so before we can assemble it.
-# Note: kbuild does not track this dependency due to usage of .incbin
-$(obj)/gate-data.o: $(obj)/gate.so
+# The gate DSO image is built using a special linker script.
+include $(srctree)/arch/ia64/kernel/Makefile.gate
+# tell compiled for native
+CPPFLAGS_gate.lds += -D__IA64_GATE_PARAVIRTUALIZED_NATIVE
 
 # Calculate NR_IRQ = max(IA64_NATIVE_NR_IRQS, XEN_NR_IRQS, ...) based on config
 define sed-y
@@ -109,9 +88,9 @@ include/asm-ia64/nr-irqs.h: arch/$(SRCARCH)/kernel/nr-irqs.s
 clean-files += $(objtree)/include/asm-ia64/nr-irqs.h
 
 #
-# native ivt.S and entry.S
+# native ivt.S, entry.S and fsys.S
 #
-ASM_PARAVIRT_OBJS = ivt.o entry.o
+ASM_PARAVIRT_OBJS = ivt.o entry.o fsys.o
 define paravirtualized_native
 AFLAGS_$(1) += -D__IA64_ASM_PARAVIRTUALIZED_NATIVE
 AFLAGS_pvchk-sed-$(1) += -D__IA64_ASM_PARAVIRTUALIZED_PVCHECK
diff --git a/arch/ia64/kernel/Makefile.gate b/arch/ia64/kernel/Makefile.gate
new file mode 100644
index 00000000000..1d87f84069b
--- /dev/null
+++ b/arch/ia64/kernel/Makefile.gate
@@ -0,0 +1,27 @@
+# The gate DSO image is built using a special linker script.
+
+targets += gate.so gate-syms.o
+
+extra-y += gate.so gate-syms.o gate.lds gate.o
+
+CPPFLAGS_gate.lds := -P -C -U$(ARCH)
+
+quiet_cmd_gate = GATE $@
+      cmd_gate = $(CC) -nostdlib $(GATECFLAGS_$(@F)) -Wl,-T,$(filter-out FORCE,$^) -o $@
+
+GATECFLAGS_gate.so = -shared -s -Wl,-soname=linux-gate.so.1 \
+		     $(call ld-option, -Wl$(comma)--hash-style=sysv)
+$(obj)/gate.so: $(obj)/gate.lds $(obj)/gate.o FORCE
+	$(call if_changed,gate)
+
+$(obj)/built-in.o: $(obj)/gate-syms.o
+$(obj)/built-in.o: ld_flags += -R $(obj)/gate-syms.o
+
+GATECFLAGS_gate-syms.o = -r
+$(obj)/gate-syms.o: $(obj)/gate.lds $(obj)/gate.o FORCE
+	$(call if_changed,gate)
+
+# gate-data.o contains the gate DSO image as data in section .data.gate.
+# We must build gate.so before we can assemble it.
+# Note: kbuild does not track this dependency due to usage of .incbin
+$(obj)/gate-data.o: $(obj)/gate.so
diff --git a/arch/ia64/kernel/asm-offsets.c b/arch/ia64/kernel/asm-offsets.c
index 742dbb1d5a4..af565016904 100644
--- a/arch/ia64/kernel/asm-offsets.c
+++ b/arch/ia64/kernel/asm-offsets.c
@@ -316,5 +316,7 @@ void foo(void)
 	DEFINE_MAPPED_REG_OFS(XSI_BANK1_R16_OFS, bank1_regs[0]);
 	DEFINE_MAPPED_REG_OFS(XSI_B0NATS_OFS, vbnat);
 	DEFINE_MAPPED_REG_OFS(XSI_B1NATS_OFS, vnat);
+	DEFINE_MAPPED_REG_OFS(XSI_ITC_OFFSET_OFS, itc_offset);
+	DEFINE_MAPPED_REG_OFS(XSI_ITC_LAST_OFS, itc_last);
 #endif /* CONFIG_XEN */
 }
diff --git a/arch/ia64/kernel/efi.c b/arch/ia64/kernel/efi.c
index efaff15d8cf..7ef80e8161c 100644
--- a/arch/ia64/kernel/efi.c
+++ b/arch/ia64/kernel/efi.c
@@ -456,6 +456,7 @@ efi_map_pal_code (void)
 		 GRANULEROUNDDOWN((unsigned long) pal_vaddr),
 		 pte_val(pfn_pte(__pa(pal_vaddr) >> PAGE_SHIFT, PAGE_KERNEL)),
 		 IA64_GRANULE_SHIFT);
+	paravirt_dv_serialize_data();
 	ia64_set_psr(psr);		/* restore psr */
 }
 
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index e5341e2c117..ccfdeee9d89 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -735,7 +735,7 @@ GLOBAL_ENTRY(__paravirt_leave_syscall)
 __paravirt_work_processed_syscall:
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 	adds r2=PT(LOADRS)+16,r12
-(pUStk)	mov.m r22=ar.itc			// fetch time at leave
+	MOV_FROM_ITC(pUStk, p9, r22, r19)	// fetch time at leave
 	adds r18=TI_FLAGS+IA64_TASK_SIZE,r13
 	;;
 (p6)	ld4 r31=[r18]				// load current_thread_info()->flags
@@ -984,7 +984,7 @@ GLOBAL_ENTRY(__paravirt_leave_kernel)
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 	.pred.rel.mutex pUStk,pKStk
 	MOV_FROM_PSR(pKStk, r22, r29)	// M2 read PSR now that interrupts are disabled
-(pUStk)	mov.m r22=ar.itc	// M  fetch time at leave
+	MOV_FROM_ITC(pUStk, p9, r22, r29)	// M  fetch time at leave
 	nop.i 0
 	;;
 #else
diff --git a/arch/ia64/kernel/fsys.S b/arch/ia64/kernel/fsys.S
index c1625c7e177..3567d54f8ce 100644
--- a/arch/ia64/kernel/fsys.S
+++ b/arch/ia64/kernel/fsys.S
@@ -25,6 +25,7 @@
 #include <asm/unistd.h>
 
 #include "entry.h"
+#include "paravirt_inst.h"
 
 /*
  * See Documentation/ia64/fsys.txt for details on fsyscalls.
@@ -279,7 +280,7 @@ ENTRY(fsys_gettimeofday)
 (p9)	cmp.eq p13,p0 = 0,r30	// if mmio_ptr, clear p13 jitter control
 	;;
 	.pred.rel.mutex p8,p9
-(p8)	mov r2 = ar.itc		// CPU_TIMER. 36 clocks latency!!!
+	MOV_FROM_ITC(p8, p6, r2, r10)	// CPU_TIMER. 36 clocks latency!!!
 (p9)	ld8 r2 = [r30]		// MMIO_TIMER. Could also have latency issues..
 (p13)	ld8 r25 = [r19]		// get itc_lastcycle value
 	ld8 r9 = [r22],IA64_TIMESPEC_TV_NSEC_OFFSET	// tv_sec
@@ -418,7 +419,7 @@ EX(.fail_efault, ld8 r14=[r33])			// r14 <- *set
 	mov r17=(1 << (SIGKILL - 1)) | (1 << (SIGSTOP - 1))
 	;;
 
-	rsm psr.i				// mask interrupt delivery
+	RSM_PSR_I(p0, r18, r19)			// mask interrupt delivery
 	mov ar.ccv=0
 	andcm r14=r14,r17			// filter out SIGKILL & SIGSTOP
 
@@ -491,7 +492,7 @@ EX(.fail_efault, ld8 r14=[r33])			// r14 <- *set
 #ifdef CONFIG_SMP
 	st4.rel [r31]=r0			// release the lock
 #endif
-	ssm psr.i
+	SSM_PSR_I(p0, p9, r31)
 	;;
 
 	srlz.d					// ensure psr.i is set again
@@ -513,7 +514,7 @@ EX(.fail_efault, (p15) st8 [r34]=r3)
 #ifdef CONFIG_SMP
 	st4.rel [r31]=r0			// release the lock
 #endif
-	ssm psr.i
+	SSM_PSR_I(p0, p9, r17)
 	;;
 	srlz.d
 	br.sptk.many fsys_fallback_syscall	// with signal pending, do the heavy-weight syscall
@@ -521,7 +522,7 @@ EX(.fail_efault, (p15) st8 [r34]=r3)
 #ifdef CONFIG_SMP
 .lock_contention:
 	/* Rather than spinning here, fall back on doing a heavy-weight syscall.  */
-	ssm psr.i
+	SSM_PSR_I(p0, p9, r17)
 	;;
 	srlz.d
 	br.sptk.many fsys_fallback_syscall
@@ -592,17 +593,17 @@ ENTRY(fsys_fallback_syscall)
 	adds r17=-1024,r15
 	movl r14=sys_call_table
 	;;
-	rsm psr.i
+	RSM_PSR_I(p0, r26, r27)
 	shladd r18=r17,3,r14
 	;;
 	ld8 r18=[r18]				// load normal (heavy-weight) syscall entry-point
-	mov r29=psr				// read psr (12 cyc load latency)
+	MOV_FROM_PSR(p0, r29, r26)		// read psr (12 cyc load latency)
 	mov r27=ar.rsc
 	mov r21=ar.fpsr
 	mov r26=ar.pfs
 END(fsys_fallback_syscall)
 	/* FALL THROUGH */
-GLOBAL_ENTRY(fsys_bubble_down)
+GLOBAL_ENTRY(paravirt_fsys_bubble_down)
 	.prologue
 	.altrp b6
 	.body
@@ -640,7 +641,7 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	 *
 	 * PSR.BE : already is turned off in __kernel_syscall_via_epc()
 	 * PSR.AC : don't care (kernel normally turns PSR.AC on)
-	 * PSR.I  : already turned off by the time fsys_bubble_down gets
+	 * PSR.I  : already turned off by the time paravirt_fsys_bubble_down gets
 	 *	    invoked
 	 * PSR.DFL: always 0 (kernel never turns it on)
 	 * PSR.DFH: don't care --- kernel never touches f32-f127 on its own
@@ -650,7 +651,7 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	 * PSR.DB : don't care --- kernel never enables kernel-level
 	 *	    breakpoints
 	 * PSR.TB : must be 0 already; if it wasn't zero on entry to
-	 *          __kernel_syscall_via_epc, the branch to fsys_bubble_down
+	 *          __kernel_syscall_via_epc, the branch to paravirt_fsys_bubble_down
 	 *          will trigger a taken branch; the taken-trap-handler then
 	 *          converts the syscall into a break-based system-call.
 	 */
@@ -683,7 +684,7 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	;;
 	mov ar.rsc=0				// M2   set enforced lazy mode, pl 0, LE, loadrs=0
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
-	mov.m r30=ar.itc			// M    get cycle for accounting
+	MOV_FROM_ITC(p0, p6, r30, r23)		// M    get cycle for accounting
 #else
 	nop.m 0
 #endif
@@ -734,21 +735,21 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	mov rp=r14				// I0   set the real return addr
 	and r3=_TIF_SYSCALL_TRACEAUDIT,r3	// A
 	;;
-	ssm psr.i				// M2   we're on kernel stacks now, reenable irqs
+	SSM_PSR_I(p0, p6, r22)			// M2   we're on kernel stacks now, reenable irqs
 	cmp.eq p8,p0=r3,r0			// A
 (p10)	br.cond.spnt.many ia64_ret_from_syscall	// B    return if bad call-frame or r15 is a NaT
 
 	nop.m 0
 (p8)	br.call.sptk.many b6=b6			// B    (ignore return address)
 	br.cond.spnt ia64_trace_syscall		// B
-END(fsys_bubble_down)
+END(paravirt_fsys_bubble_down)
 
 	.rodata
 	.align 8
-	.globl fsyscall_table
+	.globl paravirt_fsyscall_table
 
-	data8 fsys_bubble_down
-fsyscall_table:
+	data8 paravirt_fsys_bubble_down
+paravirt_fsyscall_table:
 	data8 fsys_ni_syscall
 	data8 0				// exit			// 1025
 	data8 0				// read
@@ -1033,4 +1034,4 @@ fsyscall_table:
 
 	// fill in zeros for the remaining entries
 	.zero:
-	.space fsyscall_table + 8*NR_syscalls - .zero, 0
+	.space paravirt_fsyscall_table + 8*NR_syscalls - .zero, 0
diff --git a/arch/ia64/kernel/gate.S b/arch/ia64/kernel/gate.S
index 74b1ccce4e8..cf5e0a105e1 100644
--- a/arch/ia64/kernel/gate.S
+++ b/arch/ia64/kernel/gate.S
@@ -13,6 +13,7 @@
 #include <asm/sigcontext.h>
 #include <asm/system.h>
 #include <asm/unistd.h>
+#include "paravirt_inst.h"
 
 /*
  * We can't easily refer to symbols inside the kernel.  To avoid full runtime relocation,
@@ -48,87 +49,6 @@ GLOBAL_ENTRY(__kernel_syscall_via_break)
 }
 END(__kernel_syscall_via_break)
 
-/*
- * On entry:
- *	r11 = saved ar.pfs
- *	r15 = system call #
- *	b0  = saved return address
- *	b6  = return address
- * On exit:
- *	r11 = saved ar.pfs
- *	r15 = system call #
- *	b0  = saved return address
- *	all other "scratch" registers:	undefined
- *	all "preserved" registers:	same as on entry
- */
-
-GLOBAL_ENTRY(__kernel_syscall_via_epc)
-	.prologue
-	.altrp b6
-	.body
-{
-	/*
-	 * Note: the kernel cannot assume that the first two instructions in this
-	 * bundle get executed.  The remaining code must be safe even if
-	 * they do not get executed.
-	 */
-	adds r17=-1024,r15			// A
-	mov r10=0				// A    default to successful syscall execution
-	epc					// B	causes split-issue
-}
-	;;
-	rsm psr.be | psr.i			// M2 (5 cyc to srlz.d)
-	LOAD_FSYSCALL_TABLE(r14)		// X
-	;;
-	mov r16=IA64_KR(CURRENT)		// M2 (12 cyc)
-	shladd r18=r17,3,r14			// A
-	mov r19=NR_syscalls-1			// A
-	;;
-	lfetch [r18]				// M0|1
-	mov r29=psr				// M2 (12 cyc)
-	// If r17 is a NaT, p6 will be zero
-	cmp.geu p6,p7=r19,r17			// A    (sysnr > 0 && sysnr < 1024+NR_syscalls)?
-	;;
-	mov r21=ar.fpsr				// M2 (12 cyc)
-	tnat.nz p10,p9=r15			// I0
-	mov.i r26=ar.pfs			// I0 (would stall anyhow due to srlz.d...)
-	;;
-	srlz.d					// M0 (forces split-issue) ensure PSR.BE==0
-(p6)	ld8 r18=[r18]				// M0|1
-	nop.i 0
-	;;
-	nop.m 0
-(p6)	tbit.z.unc p8,p0=r18,0			// I0 (dual-issues with "mov b7=r18"!)
-	nop.i 0
-	;;
-(p8)	ssm psr.i
-(p6)	mov b7=r18				// I0
-(p8)	br.dptk.many b7				// B
-
-	mov r27=ar.rsc				// M2 (12 cyc)
-/*
- * brl.cond doesn't work as intended because the linker would convert this branch
- * into a branch to a PLT.  Perhaps there will be a way to avoid this with some
- * future version of the linker.  In the meantime, we just use an indirect branch
- * instead.
- */
-#ifdef CONFIG_ITANIUM
-(p6)	add r14=-8,r14				// r14 <- addr of fsys_bubble_down entry
-	;;
-(p6)	ld8 r14=[r14]				// r14 <- fsys_bubble_down
-	;;
-(p6)	mov b7=r14
-(p6)	br.sptk.many b7
-#else
-	BRL_COND_FSYS_BUBBLE_DOWN(p6)
-#endif
-	ssm psr.i
-	mov r10=-1
-(p10)	mov r8=EINVAL
-(p9)	mov r8=ENOSYS
-	FSYS_RETURN
-END(__kernel_syscall_via_epc)
-
 #	define ARG0_OFF		(16 + IA64_SIGFRAME_ARG0_OFFSET)
 #	define ARG1_OFF		(16 + IA64_SIGFRAME_ARG1_OFFSET)
 #	define ARG2_OFF		(16 + IA64_SIGFRAME_ARG2_OFFSET)
@@ -374,3 +294,92 @@ restore_rbs:
 	// invala not necessary as that will happen when returning to user-mode
 	br.cond.sptk back_from_restore_rbs
 END(__kernel_sigtramp)
+
+/*
+ * On entry:
+ *	r11 = saved ar.pfs
+ *	r15 = system call #
+ *	b0  = saved return address
+ *	b6  = return address
+ * On exit:
+ *	r11 = saved ar.pfs
+ *	r15 = system call #
+ *	b0  = saved return address
+ *	all other "scratch" registers:	undefined
+ *	all "preserved" registers:	same as on entry
+ */
+
+GLOBAL_ENTRY(__kernel_syscall_via_epc)
+	.prologue
+	.altrp b6
+	.body
+{
+	/*
+	 * Note: the kernel cannot assume that the first two instructions in this
+	 * bundle get executed.  The remaining code must be safe even if
+	 * they do not get executed.
+	 */
+	adds r17=-1024,r15			// A
+	mov r10=0				// A    default to successful syscall execution
+	epc					// B	causes split-issue
+}
+	;;
+	RSM_PSR_BE_I(r20, r22)			// M2 (5 cyc to srlz.d)
+	LOAD_FSYSCALL_TABLE(r14)		// X
+	;;
+	mov r16=IA64_KR(CURRENT)		// M2 (12 cyc)
+	shladd r18=r17,3,r14			// A
+	mov r19=NR_syscalls-1			// A
+	;;
+	lfetch [r18]				// M0|1
+	MOV_FROM_PSR(p0, r29, r8)		// M2 (12 cyc)
+	// If r17 is a NaT, p6 will be zero
+	cmp.geu p6,p7=r19,r17			// A    (sysnr > 0 && sysnr < 1024+NR_syscalls)?
+	;;
+	mov r21=ar.fpsr				// M2 (12 cyc)
+	tnat.nz p10,p9=r15			// I0
+	mov.i r26=ar.pfs			// I0 (would stall anyhow due to srlz.d...)
+	;;
+	srlz.d					// M0 (forces split-issue) ensure PSR.BE==0
+(p6)	ld8 r18=[r18]				// M0|1
+	nop.i 0
+	;;
+	nop.m 0
+(p6)	tbit.z.unc p8,p0=r18,0			// I0 (dual-issues with "mov b7=r18"!)
+	nop.i 0
+	;;
+	SSM_PSR_I(p8, p14, r25)
+(p6)	mov b7=r18				// I0
+(p8)	br.dptk.many b7				// B
+
+	mov r27=ar.rsc				// M2 (12 cyc)
+/*
+ * brl.cond doesn't work as intended because the linker would convert this branch
+ * into a branch to a PLT.  Perhaps there will be a way to avoid this with some
+ * future version of the linker.  In the meantime, we just use an indirect branch
+ * instead.
+ */
+#ifdef CONFIG_ITANIUM
+(p6)	add r14=-8,r14				// r14 <- addr of fsys_bubble_down entry
+	;;
+(p6)	ld8 r14=[r14]				// r14 <- fsys_bubble_down
+	;;
+(p6)	mov b7=r14
+(p6)	br.sptk.many b7
+#else
+	BRL_COND_FSYS_BUBBLE_DOWN(p6)
+#endif
+	SSM_PSR_I(p0, p14, r10)
+	mov r10=-1
+(p10)	mov r8=EINVAL
+(p9)	mov r8=ENOSYS
+	FSYS_RETURN
+
+#ifdef CONFIG_PARAVIRT
+	/*
+	 * padd to make the size of this symbol constant
+	 * independent of paravirtualization.
+	 */
+	.align PAGE_SIZE / 8
+#endif
+END(__kernel_syscall_via_epc)
diff --git a/arch/ia64/kernel/gate.lds.S b/arch/ia64/kernel/gate.lds.S
index 3cb1abc00e2..88c64ed47c3 100644
--- a/arch/ia64/kernel/gate.lds.S
+++ b/arch/ia64/kernel/gate.lds.S
@@ -7,6 +7,7 @@
 
 
 #include <asm/system.h>
+#include "paravirt_patchlist.h"
 
 SECTIONS
 {
@@ -33,21 +34,21 @@ SECTIONS
 	. = GATE_ADDR + 0x600;
 
 	.data.patch		: {
-		__start_gate_mckinley_e9_patchlist = .;
+		__paravirt_start_gate_mckinley_e9_patchlist = .;
 		*(.data.patch.mckinley_e9)
-		__end_gate_mckinley_e9_patchlist = .;
+		__paravirt_end_gate_mckinley_e9_patchlist = .;
 
-		__start_gate_vtop_patchlist = .;
+		__paravirt_start_gate_vtop_patchlist = .;
 		*(.data.patch.vtop)
-		__end_gate_vtop_patchlist = .;
+		__paravirt_end_gate_vtop_patchlist = .;
 
-		__start_gate_fsyscall_patchlist = .;
+		__paravirt_start_gate_fsyscall_patchlist = .;
 		*(.data.patch.fsyscall_table)
-		__end_gate_fsyscall_patchlist = .;
+		__paravirt_end_gate_fsyscall_patchlist = .;
 
-		__start_gate_brl_fsys_bubble_down_patchlist = .;
+		__paravirt_start_gate_brl_fsys_bubble_down_patchlist = .;
 		*(.data.patch.brl_fsys_bubble_down)
-		__end_gate_brl_fsys_bubble_down_patchlist = .;
+		__paravirt_end_gate_brl_fsys_bubble_down_patchlist = .;
 	}						:readable
 
 	.IA_64.unwind_info	: { *(.IA_64.unwind_info*) }
diff --git a/arch/ia64/kernel/head.S b/arch/ia64/kernel/head.S
index 59301c47280..23f846de62d 100644
--- a/arch/ia64/kernel/head.S
+++ b/arch/ia64/kernel/head.S
@@ -1050,7 +1050,7 @@ END(ia64_delay_loop)
  * except that the multiplication and the shift are done with 128-bit
  * intermediate precision so that we can produce a full 64-bit result.
  */
-GLOBAL_ENTRY(sched_clock)
+GLOBAL_ENTRY(ia64_native_sched_clock)
 	addl r8=THIS_CPU(cpu_info) + IA64_CPUINFO_NSEC_PER_CYC_OFFSET,r0
 	mov.m r9=ar.itc		// fetch cycle-counter				(35 cyc)
 	;;
@@ -1066,7 +1066,13 @@ GLOBAL_ENTRY(sched_clock)
 	;;
 	shrp r8=r9,r8,IA64_NSEC_PER_CYC_SHIFT
 	br.ret.sptk.many rp
-END(sched_clock)
+END(ia64_native_sched_clock)
+#ifndef CONFIG_PARAVIRT
+	//unsigned long long
+	//sched_clock(void) __attribute__((alias("ia64_native_sched_clock")));
+	.global sched_clock
+sched_clock = ia64_native_sched_clock
+#endif
 
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 GLOBAL_ENTRY(cycle_to_cputime)
diff --git a/arch/ia64/kernel/ivt.S b/arch/ia64/kernel/ivt.S
index f675d8e3385..ec9a5fdfa1b 100644
--- a/arch/ia64/kernel/ivt.S
+++ b/arch/ia64/kernel/ivt.S
@@ -804,7 +804,7 @@ ENTRY(break_fault)
 ///////////////////////////////////////////////////////////////////////
 	st1 [r16]=r0				// M2|3 clear current->thread.on_ustack flag
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
-	mov.m r30=ar.itc			// M    get cycle for accounting
+	MOV_FROM_ITC(p0, p14, r30, r18)		// M    get cycle for accounting
 #else
 	mov b6=r30				// I0   setup syscall handler branch reg early
 #endif
diff --git a/arch/ia64/kernel/module.c b/arch/ia64/kernel/module.c
index aaa7d901521..34fe4259a14 100644
--- a/arch/ia64/kernel/module.c
+++ b/arch/ia64/kernel/module.c
@@ -446,6 +446,14 @@ module_frob_arch_sections (Elf_Ehdr *ehdr, Elf_Shdr *sechdrs, char *secstrings,
 			mod->arch.opd = s;
 		else if (strcmp(".IA_64.unwind", secstrings + s->sh_name) == 0)
 			mod->arch.unwind = s;
+#ifdef CONFIG_PARAVIRT
+		else if (strcmp(".paravirt_bundles",
+				secstrings + s->sh_name) == 0)
+			mod->arch.paravirt_bundles = s;
+		else if (strcmp(".paravirt_insts",
+				secstrings + s->sh_name) == 0)
+			mod->arch.paravirt_insts = s;
+#endif
 
 	if (!mod->arch.core_plt || !mod->arch.init_plt || !mod->arch.got || !mod->arch.opd) {
 		printk(KERN_ERR "%s: sections missing\n", mod->name);
@@ -921,6 +929,30 @@ module_finalize (const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs, struct module *mo
 	DEBUGP("%s: init: entry=%p\n", __func__, mod->init);
 	if (mod->arch.unwind)
 		register_unwind_table(mod);
+#ifdef CONFIG_PARAVIRT
+        if (mod->arch.paravirt_bundles) {
+                struct paravirt_patch_site_bundle *start =
+                        (struct paravirt_patch_site_bundle *)
+                        mod->arch.paravirt_bundles->sh_addr;
+                struct paravirt_patch_site_bundle *end =
+                        (struct paravirt_patch_site_bundle *)
+                        (mod->arch.paravirt_bundles->sh_addr +
+                         mod->arch.paravirt_bundles->sh_size);
+
+                paravirt_patch_apply_bundle(start, end);
+        }
+        if (mod->arch.paravirt_insts) {
+                struct paravirt_patch_site_inst *start =
+                        (struct paravirt_patch_site_inst *)
+                        mod->arch.paravirt_insts->sh_addr;
+                struct paravirt_patch_site_inst *end =
+                        (struct paravirt_patch_site_inst *)
+                        (mod->arch.paravirt_insts->sh_addr +
+                         mod->arch.paravirt_insts->sh_size);
+
+                paravirt_patch_apply_inst(start, end);
+        }
+#endif
 	return 0;
 }
 
diff --git a/arch/ia64/kernel/paravirt.c b/arch/ia64/kernel/paravirt.c
index 9f14c16f636..a21d7bb9c69 100644
--- a/arch/ia64/kernel/paravirt.c
+++ b/arch/ia64/kernel/paravirt.c
@@ -46,13 +46,23 @@ struct pv_info pv_info = {
  * initialization hooks.
  */
 
-struct pv_init_ops pv_init_ops;
+static void __init
+ia64_native_patch_branch(unsigned long tag, unsigned long type);
+
+struct pv_init_ops pv_init_ops =
+{
+#ifdef ASM_SUPPORTED
+	.patch_bundle = ia64_native_patch_bundle,
+#endif
+	.patch_branch = ia64_native_patch_branch,
+};
 
 /***************************************************************************
  * pv_cpu_ops
  * intrinsics hooks.
  */
 
+#ifndef ASM_SUPPORTED
 /* ia64_native_xxx are macros so that we have to make them real functions */
 
 #define DEFINE_VOID_FUNC1(name)					\
@@ -60,7 +70,14 @@ struct pv_init_ops pv_init_ops;
 	ia64_native_ ## name ## _func(unsigned long arg)	\
 	{							\
 		ia64_native_ ## name(arg);			\
-	}							\
+	}
+
+#define DEFINE_VOID_FUNC1_VOID(name)				\
+	static void						\
+	ia64_native_ ## name ## _func(void *arg)		\
+	{							\
+		ia64_native_ ## name(arg);			\
+	}
 
 #define DEFINE_VOID_FUNC2(name)					\
 	static void						\
@@ -68,7 +85,7 @@ struct pv_init_ops pv_init_ops;
 				      unsigned long arg1)	\
 	{							\
 		ia64_native_ ## name(arg0, arg1);		\
-	}							\
+	}
 
 #define DEFINE_FUNC0(name)			\
 	static unsigned long			\
@@ -84,7 +101,7 @@ struct pv_init_ops pv_init_ops;
 		return ia64_native_ ## name(arg);	\
 	}						\
 
-DEFINE_VOID_FUNC1(fc);
+DEFINE_VOID_FUNC1_VOID(fc);
 DEFINE_VOID_FUNC1(intrin_local_irq_restore);
 
 DEFINE_VOID_FUNC2(ptcga);
@@ -274,6 +291,266 @@ ia64_native_setreg_func(int regnum, unsigned long val)
 		break;
 	}
 }
+#else
+
+#define __DEFINE_FUNC(name, code)					\
+	extern const char ia64_native_ ## name ## _direct_start[];	\
+	extern const char ia64_native_ ## name ## _direct_end[];	\
+	asm (".align 32\n"						\
+	     ".proc ia64_native_" #name "_func\n"			\
+	     "ia64_native_" #name "_func:\n"				\
+	     "ia64_native_" #name "_direct_start:\n"			\
+	     code							\
+	     "ia64_native_" #name "_direct_end:\n"			\
+	     "br.cond.sptk.many b6\n"					\
+	     ".endp ia64_native_" #name "_func\n")
+
+#define DEFINE_VOID_FUNC0(name, code)				\
+	extern void						\
+	ia64_native_ ## name ## _func(void);			\
+	__DEFINE_FUNC(name, code)
+
+#define DEFINE_VOID_FUNC1(name, code)				\
+	extern void						\
+	ia64_native_ ## name ## _func(unsigned long arg);	\
+	__DEFINE_FUNC(name, code)
+
+#define DEFINE_VOID_FUNC1_VOID(name, code)			\
+	extern void						\
+	ia64_native_ ## name ## _func(void *arg);		\
+	__DEFINE_FUNC(name, code)
+
+#define DEFINE_VOID_FUNC2(name, code)				\
+	extern void						\
+	ia64_native_ ## name ## _func(unsigned long arg0,	\
+				      unsigned long arg1);	\
+	__DEFINE_FUNC(name, code)
+
+#define DEFINE_FUNC0(name, code)		\
+	extern unsigned long			\
+	ia64_native_ ## name ## _func(void);	\
+	__DEFINE_FUNC(name, code)
+
+#define DEFINE_FUNC1(name, type, code)			\
+	extern unsigned long				\
+	ia64_native_ ## name ## _func(type arg);	\
+	__DEFINE_FUNC(name, code)
+
+DEFINE_VOID_FUNC1_VOID(fc,
+		       "fc r8\n");
+DEFINE_VOID_FUNC1(intrin_local_irq_restore,
+		  ";;\n"
+		  "     cmp.ne p6, p7 = r8, r0\n"
+		  ";;\n"
+		  "(p6) ssm psr.i\n"
+		  "(p7) rsm psr.i\n"
+		  ";;\n"
+		  "(p6) srlz.d\n");
+
+DEFINE_VOID_FUNC2(ptcga,
+		  "ptc.ga r8, r9\n");
+DEFINE_VOID_FUNC2(set_rr,
+		  "mov rr[r8] = r9\n");
+
+/* ia64_native_getreg(_IA64_REG_PSR) & IA64_PSR_I */
+DEFINE_FUNC0(get_psr_i,
+	     "mov r2 = " __stringify(1 << IA64_PSR_I_BIT) "\n"
+	     "mov r8 = psr\n"
+	     ";;\n"
+	     "and r8 = r2, r8\n");
+
+DEFINE_FUNC1(thash, unsigned long,
+	     "thash r8 = r8\n");
+DEFINE_FUNC1(get_cpuid, int,
+	     "mov r8 = cpuid[r8]\n");
+DEFINE_FUNC1(get_pmd, int,
+	     "mov r8 = pmd[r8]\n");
+DEFINE_FUNC1(get_rr, unsigned long,
+	     "mov r8 = rr[r8]\n");
+
+DEFINE_VOID_FUNC0(ssm_i,
+		  "ssm psr.i\n");
+DEFINE_VOID_FUNC0(rsm_i,
+		  "rsm psr.i\n");
+
+extern void
+ia64_native_set_rr0_to_rr4_func(unsigned long val0, unsigned long val1,
+				unsigned long val2, unsigned long val3,
+				unsigned long val4);
+__DEFINE_FUNC(set_rr0_to_rr4,
+	      "mov rr[r0] = r8\n"
+	      "movl r2 = 0x2000000000000000\n"
+	      ";;\n"
+	      "mov rr[r2] = r9\n"
+	      "shl r3 = r2, 1\n"	/* movl r3 = 0x4000000000000000 */
+	      ";;\n"
+	      "add r2 = r2, r3\n"	/* movl r2 = 0x6000000000000000 */
+	      "mov rr[r3] = r10\n"
+	      ";;\n"
+	      "mov rr[r2] = r11\n"
+	      "shl r3 = r3, 1\n"	/* movl r3 = 0x8000000000000000 */
+	      ";;\n"
+	      "mov rr[r3] = r14\n");
+
+extern unsigned long ia64_native_getreg_func(int regnum);
+asm(".global ia64_native_getreg_func\n");
+#define __DEFINE_GET_REG(id, reg)			\
+	"mov r2 = " __stringify(_IA64_REG_ ## id) "\n"	\
+	";;\n"						\
+	"cmp.eq p6, p0 = r2, r8\n"			\
+	";;\n"						\
+	"(p6) mov r8 = " #reg "\n"			\
+	"(p6) br.cond.sptk.many b6\n"			\
+	";;\n"
+#define __DEFINE_GET_AR(id, reg)	__DEFINE_GET_REG(AR_ ## id, ar.reg)
+#define __DEFINE_GET_CR(id, reg)	__DEFINE_GET_REG(CR_ ## id, cr.reg)
+
+__DEFINE_FUNC(getreg,
+	      __DEFINE_GET_REG(GP, gp)
+	      /*__DEFINE_GET_REG(IP, ip)*/ /* returned ip value shouldn't be constant */
+	      __DEFINE_GET_REG(PSR, psr)
+	      __DEFINE_GET_REG(TP, tp)
+	      __DEFINE_GET_REG(SP, sp)
+
+	      __DEFINE_GET_REG(AR_KR0, ar0)
+	      __DEFINE_GET_REG(AR_KR1, ar1)
+	      __DEFINE_GET_REG(AR_KR2, ar2)
+	      __DEFINE_GET_REG(AR_KR3, ar3)
+	      __DEFINE_GET_REG(AR_KR4, ar4)
+	      __DEFINE_GET_REG(AR_KR5, ar5)
+	      __DEFINE_GET_REG(AR_KR6, ar6)
+	      __DEFINE_GET_REG(AR_KR7, ar7)
+	      __DEFINE_GET_AR(RSC, rsc)
+	      __DEFINE_GET_AR(BSP, bsp)
+	      __DEFINE_GET_AR(BSPSTORE, bspstore)
+	      __DEFINE_GET_AR(RNAT, rnat)
+	      __DEFINE_GET_AR(FCR, fcr)
+	      __DEFINE_GET_AR(EFLAG, eflag)
+	      __DEFINE_GET_AR(CSD, csd)
+	      __DEFINE_GET_AR(SSD, ssd)
+	      __DEFINE_GET_REG(AR_CFLAG, ar27)
+	      __DEFINE_GET_AR(FSR, fsr)
+	      __DEFINE_GET_AR(FIR, fir)
+	      __DEFINE_GET_AR(FDR, fdr)
+	      __DEFINE_GET_AR(CCV, ccv)
+	      __DEFINE_GET_AR(UNAT, unat)
+	      __DEFINE_GET_AR(FPSR, fpsr)
+	      __DEFINE_GET_AR(ITC, itc)
+	      __DEFINE_GET_AR(PFS, pfs)
+	      __DEFINE_GET_AR(LC, lc)
+	      __DEFINE_GET_AR(EC, ec)
+
+	      __DEFINE_GET_CR(DCR, dcr)
+	      __DEFINE_GET_CR(ITM, itm)
+	      __DEFINE_GET_CR(IVA, iva)
+	      __DEFINE_GET_CR(PTA, pta)
+	      __DEFINE_GET_CR(IPSR, ipsr)
+	      __DEFINE_GET_CR(ISR, isr)
+	      __DEFINE_GET_CR(IIP, iip)
+	      __DEFINE_GET_CR(IFA, ifa)
+	      __DEFINE_GET_CR(ITIR, itir)
+	      __DEFINE_GET_CR(IIPA, iipa)
+	      __DEFINE_GET_CR(IFS, ifs)
+	      __DEFINE_GET_CR(IIM, iim)
+	      __DEFINE_GET_CR(IHA, iha)
+	      __DEFINE_GET_CR(LID, lid)
+	      __DEFINE_GET_CR(IVR, ivr)
+	      __DEFINE_GET_CR(TPR, tpr)
+	      __DEFINE_GET_CR(EOI, eoi)
+	      __DEFINE_GET_CR(IRR0, irr0)
+	      __DEFINE_GET_CR(IRR1, irr1)
+	      __DEFINE_GET_CR(IRR2, irr2)
+	      __DEFINE_GET_CR(IRR3, irr3)
+	      __DEFINE_GET_CR(ITV, itv)
+	      __DEFINE_GET_CR(PMV, pmv)
+	      __DEFINE_GET_CR(CMCV, cmcv)
+	      __DEFINE_GET_CR(LRR0, lrr0)
+	      __DEFINE_GET_CR(LRR1, lrr1)
+
+	      "mov r8 = -1\n"	/* unsupported case */
+	);
+
+extern void ia64_native_setreg_func(int regnum, unsigned long val);
+asm(".global ia64_native_setreg_func\n");
+#define __DEFINE_SET_REG(id, reg)			\
+	"mov r2 = " __stringify(_IA64_REG_ ## id) "\n"	\
+	";;\n"						\
+	"cmp.eq p6, p0 = r2, r9\n"			\
+	";;\n"						\
+	"(p6) mov " #reg " = r8\n"			\
+	"(p6) br.cond.sptk.many b6\n"			\
+	";;\n"
+#define __DEFINE_SET_AR(id, reg)	__DEFINE_SET_REG(AR_ ## id, ar.reg)
+#define __DEFINE_SET_CR(id, reg)	__DEFINE_SET_REG(CR_ ## id, cr.reg)
+__DEFINE_FUNC(setreg,
+	      "mov r2 = " __stringify(_IA64_REG_PSR_L) "\n"
+	      ";;\n"
+	      "cmp.eq p6, p0 = r2, r9\n"
+	      ";;\n"
+	      "(p6) mov psr.l = r8\n"
+#ifdef HAVE_SERIALIZE_DIRECTIVE
+	      ".serialize.data\n"
+#endif
+	      "(p6) b