Merge branch 'master' into gfs2

27 files changed, 833 insertions, 286 deletions

diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig
index db274da7dba..0b7f701d5cf 100644
--- a/arch/ia64/Kconfig
+++ b/arch/ia64/Kconfig
@@ -66,15 +66,6 @@ config IA64_UNCACHED_ALLOCATOR
 	bool
 	select GENERIC_ALLOCATOR
 
-config DMA_IS_DMA32
-	bool
-	default y
-
-config DMA_IS_NORMAL
-	bool
-	depends on IA64_SGI_SN2
-	default y
-
 config AUDIT_ARCH
 	bool
 	default y
@@ -365,6 +356,9 @@ config NODES_SHIFT
 	  MAX_NUMNODES will be 2^(This value).
 	  If in doubt, use the default.
 
+config ARCH_POPULATES_NODE_MAP
+	def_bool y
+
 # VIRTUAL_MEM_MAP and FLAT_NODE_MEM_MAP are functionally equivalent.
 # VIRTUAL_MEM_MAP has been retained for historical reasons.
 config VIRTUAL_MEM_MAP
@@ -429,6 +423,14 @@ config IA64_PALINFO
 config SGI_SN
 	def_bool y if (IA64_SGI_SN2 || IA64_GENERIC)
 
+config IA64_ESI
+	bool "ESI (Extensible SAL Interface) support"
+	help
+	  If you say Y here, support is built into the kernel to
+	  make ESI calls.  ESI calls are used to support vendor-specific
+	  firmware extensions, such as the ability to inject memory-errors
+	  for test-purposes.  If you're unsure, say N.
+
 source "drivers/sn/Kconfig"
 
 source "drivers/firmware/Kconfig"
diff --git a/arch/ia64/ia32/sys_ia32.c b/arch/ia64/ia32/sys_ia32.c
index 6aa3c51619c..bddbd22706e 100644
--- a/arch/ia64/ia32/sys_ia32.c
+++ b/arch/ia64/ia32/sys_ia32.c
@@ -1942,7 +1942,7 @@ struct sysctl32 {
 	unsigned int	__unused[4];
 };
 
-#ifdef CONFIG_SYSCTL
+#ifdef CONFIG_SYSCTL_SYSCALL
 asmlinkage long
 sys32_sysctl (struct sysctl32 __user *args)
 {
diff --git a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
index ad8215a3c58..31497496eb4 100644
--- a/arch/ia64/kernel/Makefile
+++ b/arch/ia64/kernel/Makefile
@@ -32,6 +32,11 @@ obj-$(CONFIG_IA64_UNCACHED_ALLOCATOR)	+= uncached.o
 obj-$(CONFIG_AUDIT)		+= audit.o
 mca_recovery-y			+= mca_drv.o mca_drv_asm.o
 
+obj-$(CONFIG_IA64_ESI)		+= esi.o
+ifneq ($(CONFIG_IA64_ESI),)
+obj-y				+= esi_stub.o	# must be in kernel proper
+endif
+
 # The gate DSO image is built using a special linker script.
 targets += gate.so gate-syms.o
 
diff --git a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
index 0176556aeec..32c3abededc 100644
--- a/arch/ia64/kernel/acpi.c
+++ b/arch/ia64/kernel/acpi.c
@@ -771,16 +771,19 @@ int acpi_map_cpu2node(acpi_handle handle, int cpu, long physid)
 {
 #ifdef CONFIG_ACPI_NUMA
 	int pxm_id;
+	int nid;
 
 	pxm_id = acpi_get_pxm(handle);
-
 	/*
-	 * Assuming that the container driver would have set the proximity
-	 * domain and would have initialized pxm_to_node(pxm_id) && pxm_flag
+	 * We don't have cpu-only-node hotadd. But if the system equips
+	 * SRAT table, pxm is already found and node is ready.
+  	 * So, just pxm_to_nid(pxm) is OK.
+	 * This code here is for the system which doesn't have full SRAT
+  	 * table for possible cpus.
 	 */
-	node_cpuid[cpu].nid = (pxm_id < 0) ? 0 : pxm_to_node(pxm_id);
-
+	nid = acpi_map_pxm_to_node(pxm_id);
 	node_cpuid[cpu].phys_id = physid;
+	node_cpuid[cpu].nid = nid;
 #endif
 	return (0);
 }
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index fef06571be9..12701cf32d9 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -1605,8 +1605,8 @@ sys_call_table:
 	data8 sys_ni_syscall			// 1295 reserved for ppoll
 	data8 sys_unshare
 	data8 sys_splice
-	data8 sys_ni_syscall			// reserved for set_robust_list
-	data8 sys_ni_syscall			// reserved for get_robust_list
+	data8 sys_set_robust_list
+	data8 sys_get_robust_list
 	data8 sys_sync_file_range		// 1300
 	data8 sys_tee
 	data8 sys_vmsplice
diff --git a/arch/ia64/kernel/esi.c b/arch/ia64/kernel/esi.c
new file mode 100644
index 00000000000..ebf4e988e78
--- /dev/null
+++ b/arch/ia64/kernel/esi.c
@@ -0,0 +1,205 @@
+/*
+ * Extensible SAL Interface (ESI) support routines.
+ *
+ * Copyright (C) 2006 Hewlett-Packard Co
+ * 	Alex Williamson <alex.williamson@hp.com>
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/string.h>
+
+#include <asm/esi.h>
+#include <asm/sal.h>
+
+MODULE_AUTHOR("Alex Williamson <alex.williamson@hp.com>");
+MODULE_DESCRIPTION("Extensible SAL Interface (ESI) support");
+MODULE_LICENSE("GPL");
+
+#define MODULE_NAME	"esi"
+
+#define ESI_TABLE_GUID					\
+    EFI_GUID(0x43EA58DC, 0xCF28, 0x4b06, 0xB3,		\
+	     0x91, 0xB7, 0x50, 0x59, 0x34, 0x2B, 0xD4)
+
+enum esi_systab_entry_type {
+	ESI_DESC_ENTRY_POINT = 0
+};
+
+/*
+ * Entry type:	Size:
+ *	0	48
+ */
+#define ESI_DESC_SIZE(type)	"\060"[(unsigned) (type)]
+
+typedef struct ia64_esi_desc_entry_point {
+	u8 type;
+	u8 reserved1[15];
+	u64 esi_proc;
+	u64 gp;
+	efi_guid_t guid;
+} ia64_esi_desc_entry_point_t;
+
+struct pdesc {
+	void *addr;
+	void *gp;
+};
+
+static struct ia64_sal_systab *esi_systab;
+
+static int __init esi_init (void)
+{
+	efi_config_table_t *config_tables;
+	struct ia64_sal_systab *systab;
+	unsigned long esi = 0;
+	char *p;
+	int i;
+
+	config_tables = __va(efi.systab->tables);
+
+	for (i = 0; i < (int) efi.systab->nr_tables; ++i) {
+		if (efi_guidcmp(config_tables[i].guid, ESI_TABLE_GUID) == 0) {
+			esi = config_tables[i].table;
+			break;
+		}
+	}
+
+	if (!esi)
+		return -ENODEV;;
+
+	systab = __va(esi);
+
+	if (strncmp(systab->signature, "ESIT", 4) != 0) {
+		printk(KERN_ERR "bad signature in ESI system table!");
+		return -ENODEV;
+	}
+
+	p = (char *) (systab + 1);
+	for (i = 0; i < systab->entry_count; i++) {
+		/*
+		 * The first byte of each entry type contains the type
+		 * descriptor.
+		 */
+		switch (*p) {
+		      case ESI_DESC_ENTRY_POINT:
+			break;
+		      default:
+			printk(KERN_WARNING "Unkown table type %d found in "
+			       "ESI table, ignoring rest of table\n", *p);
+			return -ENODEV;
+		}
+
+		p += ESI_DESC_SIZE(*p);
+	}
+
+	esi_systab = systab;
+	return 0;
+}
+
+
+int ia64_esi_call (efi_guid_t guid, struct ia64_sal_retval *isrvp,
+		   enum esi_proc_type proc_type, u64 func,
+		   u64 arg1, u64 arg2, u64 arg3, u64 arg4, u64 arg5, u64 arg6,
+		   u64 arg7)
+{
+	struct ia64_fpreg fr[6];
+	unsigned long flags = 0;
+	int i;
+	char *p;
+
+	if (!esi_systab)
+		return -1;
+
+	p = (char *) (esi_systab + 1);
+	for (i = 0; i < esi_systab->entry_count; i++) {
+		if (*p == ESI_DESC_ENTRY_POINT) {
+			ia64_esi_desc_entry_point_t *esi = (void *)p;
+			if (!efi_guidcmp(guid, esi->guid)) {
+				ia64_sal_handler esi_proc;
+				struct pdesc pdesc;
+
+				pdesc.addr = __va(esi->esi_proc);
+				pdesc.gp = __va(esi->gp);
+
+				esi_proc = (ia64_sal_handler) &pdesc;
+
+				ia64_save_scratch_fpregs(fr);
+				if (proc_type == ESI_PROC_SERIALIZED)
+					spin_lock_irqsave(&sal_lock, flags);
+				else if (proc_type == ESI_PROC_MP_SAFE)
+					local_irq_save(flags);
+				else
+					preempt_disable();
+				*isrvp = (*esi_proc)(func, arg1, arg2, arg3,
+						     arg4, arg5, arg6, arg7);
+				if (proc_type == ESI_PROC_SERIALIZED)
+					spin_unlock_irqrestore(&sal_lock,
+							       flags);
+				else if (proc_type == ESI_PROC_MP_SAFE)
+					local_irq_restore(flags);
+				else
+					preempt_enable();
+				ia64_load_scratch_fpregs(fr);
+				return 0;
+			}
+		}
+		p += ESI_DESC_SIZE(*p);
+	}
+	return -1;
+}
+EXPORT_SYMBOL_GPL(ia64_esi_call);
+
+int ia64_esi_call_phys (efi_guid_t guid, struct ia64_sal_retval *isrvp,
+			u64 func, u64 arg1, u64 arg2, u64 arg3, u64 arg4,
+			u64 arg5, u64 arg6, u64 arg7)
+{
+	struct ia64_fpreg fr[6];
+	unsigned long flags;
+	u64 esi_params[8];
+	char *p;
+	int i;
+
+	if (!esi_systab)
+		return -1;
+
+	p = (char *) (esi_systab + 1);
+	for (i = 0; i < esi_systab->entry_count; i++) {
+		if (*p == ESI_DESC_ENTRY_POINT) {
+			ia64_esi_desc_entry_point_t *esi = (void *)p;
+			if (!efi_guidcmp(guid, esi->guid)) {
+				ia64_sal_handler esi_proc;
+				struct pdesc pdesc;
+
+				pdesc.addr = (void *)esi->esi_proc;
+				pdesc.gp = (void *)esi->gp;
+
+				esi_proc = (ia64_sal_handler) &pdesc;
+
+				esi_params[0] = func;
+				esi_params[1] = arg1;
+				esi_params[2] = arg2;
+				esi_params[3] = arg3;
+				esi_params[4] = arg4;
+				esi_params[5] = arg5;
+				esi_params[6] = arg6;
+				esi_params[7] = arg7;
+				ia64_save_scratch_fpregs(fr);
+				spin_lock_irqsave(&sal_lock, flags);
+				*isrvp = esi_call_phys(esi_proc, esi_params);
+				spin_unlock_irqrestore(&sal_lock, flags);
+				ia64_load_scratch_fpregs(fr);
+				return 0;
+			}
+		}
+		p += ESI_DESC_SIZE(*p);
+	}
+	return -1;
+}
+EXPORT_SYMBOL_GPL(ia64_esi_call_phys);
+
+static void __exit esi_exit (void)
+{
+}
+
+module_init(esi_init);
+module_exit(esi_exit);	/* makes module removable... */
diff --git a/arch/ia64/kernel/esi_stub.S b/arch/ia64/kernel/esi_stub.S
new file mode 100644
index 00000000000..6b3d6c1f99b
--- /dev/null
+++ b/arch/ia64/kernel/esi_stub.S
@@ -0,0 +1,96 @@
+/*
+ * ESI call stub.
+ *
+ * Copyright (C) 2005 Hewlett-Packard Co
+ *	Alex Williamson <alex.williamson@hp.com>
+ *
+ * Based on EFI call stub by David Mosberger.  The stub is virtually
+ * identical to the one for EFI phys-mode calls, except that ESI
+ * calls may have up to 8 arguments, so they get passed to this routine
+ * through memory.
+ *
+ * This stub allows us to make ESI calls in physical mode with interrupts
+ * turned off.  ESI calls may not support calling from virtual mode.
+ *
+ * Google for "Extensible SAL specification" for a document describing the
+ * ESI standard.
+ */
+
+/*
+ * PSR settings as per SAL spec (Chapter 8 in the "IA-64 System
+ * Abstraction Layer Specification", revision 2.6e).  Note that
+ * psr.dfl and psr.dfh MUST be cleared, despite what this manual says.
+ * Otherwise, SAL dies whenever it's trying to do an IA-32 BIOS call
+ * (the br.ia instruction fails unless psr.dfl and psr.dfh are
+ * cleared).  Fortunately, SAL promises not to touch the floating
+ * point regs, so at least we don't have to save f2-f127.
+ */
+#define PSR_BITS_TO_CLEAR						\
+	(IA64_PSR_I | IA64_PSR_IT | IA64_PSR_DT | IA64_PSR_RT |		\
+	 IA64_PSR_DD | IA64_PSR_SS | IA64_PSR_RI | IA64_PSR_ED |	\
+	 IA64_PSR_DFL | IA64_PSR_DFH)
+
+#define PSR_BITS_TO_SET							\
+	(IA64_PSR_BN)
+
+#include <asm/processor.h>
+#include <asm/asmmacro.h>
+
+/*
+ * Inputs:
+ *	in0 = address of function descriptor of ESI routine to call
+ *	in1 = address of array of ESI parameters
+ *
+ * Outputs:
+ *	r8 = result returned by called function
+ */
+GLOBAL_ENTRY(esi_call_phys)
+	.prologue ASM_UNW_PRLG_RP|ASM_UNW_PRLG_PFS, ASM_UNW_PRLG_GRSAVE(2)
+	alloc loc1=ar.pfs,2,7,8,0
+	ld8 r2=[in0],8			// load ESI function's entry point
+	mov loc0=rp
+	.body
+	;;
+	ld8 out0=[in1],8		// ESI params loaded from array
+	;;				// passing all as inputs doesn't work
+	ld8 out1=[in1],8
+	;;
+	ld8 out2=[in1],8
+	;;
+	ld8 out3=[in1],8
+	;;
+	ld8 out4=[in1],8
+	;;
+	ld8 out5=[in1],8
+	;;
+	ld8 out6=[in1],8
+	;;
+	ld8 out7=[in1]
+	mov loc2=gp			// save global pointer
+	mov loc4=ar.rsc			// save RSE configuration
+	mov ar.rsc=0			// put RSE in enforced lazy, LE mode
+	;;
+	ld8 gp=[in0]			// load ESI function's global pointer
+	movl r16=PSR_BITS_TO_CLEAR
+	mov loc3=psr			// save processor status word
+	movl r17=PSR_BITS_TO_SET
+	;;
+	or loc3=loc3,r17
+	mov b6=r2
+	;;
+	andcm r16=loc3,r16	// get psr with IT, DT, and RT bits cleared
+	br.call.sptk.many rp=ia64_switch_mode_phys
+.ret0:	mov loc5=r19			// old ar.bsp
+	mov loc6=r20			// old sp
+	br.call.sptk.many rp=b6		// call the ESI function
+.ret1:	mov ar.rsc=0			// put RSE in enforced lazy, LE mode
+	mov r16=loc3			// save virtual mode psr
+	mov r19=loc5			// save virtual mode bspstore
+	mov r20=loc6			// save virtual mode sp
+	br.call.sptk.many rp=ia64_switch_mode_virt // return to virtual mode
+.ret2:	mov ar.rsc=loc4			// restore RSE configuration
+	mov ar.pfs=loc1
+	mov rp=loc0
+	mov gp=loc2
+	br.ret.sptk.many rp
+END(esi_call_phys)
diff --git a/arch/ia64/kernel/ia64_ksyms.c b/arch/ia64/kernel/ia64_ksyms.c
index 3ead20fb6f4..879c1817bd1 100644
--- a/arch/ia64/kernel/ia64_ksyms.c
+++ b/arch/ia64/kernel/ia64_ksyms.c
@@ -105,5 +105,9 @@ EXPORT_SYMBOL(ia64_spinlock_contention);
 # endif
 #endif
 
+#if defined(CONFIG_IA64_ESI) || defined(CONFIG_IA64_ESI_MODULE)
+extern void esi_call_phys (void);
+EXPORT_SYMBOL_GPL(esi_call_phys);
+#endif
 extern char ia64_ivt[];
 EXPORT_SYMBOL(ia64_ivt);
diff --git a/arch/ia64/kernel/kprobes.c b/arch/ia64/kernel/kprobes.c
index 781960f80b6..169ec3a7156 100644
--- a/arch/ia64/kernel/kprobes.c
+++ b/arch/ia64/kernel/kprobes.c
@@ -136,10 +136,8 @@ static void __kprobes update_kprobe_inst_flag(uint template, uint  slot,
 static int __kprobes unsupported_inst(uint template, uint  slot,
 				      uint major_opcode,
 				      unsigned long kprobe_inst,
-				      struct kprobe *p)
+				      unsigned long addr)
 {
-	unsigned long addr = (unsigned long)p->addr;
-
 	if (bundle_encoding[template][slot] == I) {
 		switch (major_opcode) {
 			case 0x0: //I_UNIT_MISC_OPCODE:
@@ -217,7 +215,7 @@ static void __kprobes prepare_break_inst(uint template, uint  slot,
 					 struct kprobe *p)
 {
 	unsigned long break_inst = BREAK_INST;
-	bundle_t *bundle = &p->ainsn.insn.bundle;
+	bundle_t *bundle = &p->opcode.bundle;
 
 	/*
 	 * Copy the original kprobe_inst qualifying predicate(qp)
@@ -423,11 +421,9 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
 	unsigned long *kprobe_addr = (unsigned long *)(addr & ~0xFULL);
 	unsigned long kprobe_inst=0;
 	unsigned int slot = addr & 0xf, template, major_opcode = 0;
-	bundle_t *bundle = &p->ainsn.insn.bundle;
-
-	memcpy(&p->opcode.bundle, kprobe_addr, sizeof(bundle_t));
-	memcpy(&p->ainsn.insn.bundle, kprobe_addr, sizeof(bundle_t));
+	bundle_t *bundle;
 
+	bundle = &((kprobe_opcode_t *)kprobe_addr)->bundle;
  	template = bundle->quad0.template;
 
 	if(valid_kprobe_addr(template, slot, addr))
@@ -440,20 +436,19 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
 	/* Get kprobe_inst and major_opcode from the bundle */
 	get_kprobe_inst(bundle, slot, &kprobe_inst, &major_opcode);
 
-	if (unsupported_inst(template, slot, major_opcode, kprobe_inst, p))
+	if (unsupported_inst(template, slot, major_opcode, kprobe_inst, addr))
 			return -EINVAL;
 
-	prepare_break_inst(template, slot, major_opcode, kprobe_inst, p);
 
-	return 0;
-}
+	p->ainsn.insn = get_insn_slot();
+	if (!p->ainsn.insn)
+		return -ENOMEM;
+	memcpy(&p->opcode, kprobe_addr, sizeof(kprobe_opcode_t));
+	memcpy(p->ainsn.insn, kprobe_addr, sizeof(kprobe_opcode_t));
 
-void __kprobes flush_insn_slot(struct kprobe *p)
-{
-	unsigned long arm_addr;
+	prepare_break_inst(template, slot, major_opcode, kprobe_inst, p);
 
-	arm_addr = ((unsigned long)&p->opcode.bundle) & ~0xFULL;
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
+	return 0;
 }
 
 void __kprobes arch_arm_kprobe(struct kprobe *p)
@@ -461,9 +456,10 @@ void __kprobes arch_arm_kprobe(struct kprobe *p)
 	unsigned long addr = (unsigned long)p->addr;
 	unsigned long arm_addr = addr & ~0xFULL;
 
-	flush_insn_slot(p);
-	memcpy((char *)arm_addr, &p->ainsn.insn.bundle, sizeof(bundle_t));
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
+	flush_icache_range((unsigned long)p->ainsn.insn,
+			(unsigned long)p->ainsn.insn + sizeof(kprobe_opcode_t));
+	memcpy((char *)arm_addr, &p->opcode, sizeof(kprobe_opcode_t));
+	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
 }
 
 void __kprobes arch_disarm_kprobe(struct kprobe *p)
@@ -471,11 +467,18 @@ void __kprobes arch_disarm_kprobe(struct kprobe *p)
 	unsigned long addr = (unsigned long)p->addr;
 	unsigned long arm_addr = addr & ~0xFULL;
 
-	/* p->opcode contains the original unaltered bundle */
-	memcpy((char *) arm_addr, (char *) &p->opcode.bundle, sizeof(bundle_t));
-	flush_icache_range(arm_addr, arm_addr + sizeof(bundle_t));
+	/* p->ainsn.insn contains the original unaltered kprobe_opcode_t */
+	memcpy((char *) arm_addr, (char *) p->ainsn.insn,
+					 sizeof(kprobe_opcode_t));
+	flush_icache_range(arm_addr, arm_addr + sizeof(kprobe_opcode_t));
 }
 
+void __kprobes arch_remove_kprobe(struct kprobe *p)
+{
+	mutex_lock(&kprobe_mutex);
+	free_insn_slot(p->ainsn.insn);
+	mutex_unlock(&kprobe_mutex);
+}
 /*
  * We are resuming execution after a single step fault, so the pt_regs
  * structure reflects the register state after we executed the instruction
@@ -486,12 +489,12 @@ void __kprobes arch_disarm_kprobe(struct kprobe *p)
  */
 static void __kprobes resume_execution(struct kprobe *p, struct pt_regs *regs)
 {
-  	unsigned long bundle_addr = ((unsigned long) (&p->opcode.bundle)) & ~0xFULL;
+  	unsigned long bundle_addr = (unsigned long) (&p->ainsn.insn->bundle);
   	unsigned long resume_addr = (unsigned long)p->addr & ~0xFULL;
  	unsigned long template;
  	int slot = ((unsigned long)p->addr & 0xf);
 
-	template = p->opcode.bundle.quad0.template;
+	template = p->ainsn.insn->bundle.quad0.template;
 
  	if (slot == 1 && bundle_encoding[template][1] == L)
  		slot = 2;
@@ -553,7 +556,7 @@ turn_ss_off:
 
 static void __kprobes prepare_ss(struct kprobe *p, struct pt_regs *regs)
 {
-	unsigned long bundle_addr = (unsigned long) &p->opcode.bundle;
+	unsigned long bundle_addr = (unsigned long) &p->ainsn.insn->bundle;
 	unsigned long slot = (unsigned long)p->addr & 0xf;
 
 	/* single step inline if break instruction */
@@ -768,6 +771,12 @@ static int __kprobes kprobes_fault_handler(struct pt_regs *regs, int trapnr)
 		 */
 		if (cur->fault_handler && cur->fault_handler(cur, regs, trapnr))
 			return 1;
+		/*
+		 * In case the user-specified fault handler returned
+		 * zero, try to fix up.
+		 */
+		if (ia64_done_with_exception(regs))
+			return 1;
 
 		/*
 		 * Let ia64_do_page_fault() fix it.
diff --git a/arch/ia64/kernel/mca.c b/arch/ia64/kernel/mca.c
index 2fbe4536fe1..bfbd8986153 100644
--- a/arch/ia64/kernel/mca.c
+++ b/arch/ia64/kernel/mca.c
@@ -54,6 +54,9 @@
  *
  * 2005-10-07 Keith Owens <kaos@sgi.com>
  *	      Add notify_die() hooks.
+ *
+ * 2006-09-15 Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
+ * 	      Add printing support for MCA/INIT.
  */
 #include <linux/types.h>
 #include <linux/init.h>
@@ -136,11 +139,175 @@ extern void salinfo_log_wakeup(int type, u8 *buffer, u64 size, int irqsafe);
 
 static int mca_init __initdata;
 
+/*
+ * limited & delayed printing support for MCA/INIT handler
+ */
+
+#define mprintk(fmt...) ia64_mca_printk(fmt)
+
+#define MLOGBUF_SIZE (512+256*NR_CPUS)
+#define MLOGBUF_MSGMAX 256
+static char mlogbuf[MLOGBUF_SIZE];
+static DEFINE_SPINLOCK(mlogbuf_wlock);	/* mca context only */
+static DEFINE_SPINLOCK(mlogbuf_rlock);	/* normal context only */
+static unsigned long mlogbuf_start;
+static unsigned long mlogbuf_end;
+static unsigned int mlogbuf_finished = 0;
+static unsigned long mlogbuf_timestamp = 0;
+
+static int loglevel_save = -1;
+#define BREAK_LOGLEVEL(__console_loglevel)		\
+	oops_in_progress = 1;				\
+	if (loglevel_save < 0)				\
+		loglevel_save = __console_loglevel;	\
+	__console_loglevel = 15;
+
+#define RESTORE_LOGLEVEL(__console_loglevel)		\
+	if (loglevel_save >= 0) {			\
+		__console_loglevel = loglevel_save;	\
+		loglevel_save = -1;			\
+	}						\
+	mlogbuf_finished = 0;				\
+	oops_in_progress = 0;
+
+/*
+ * Push messages into buffer, print them later if not urgent.
+ */
+void ia64_mca_printk(const char *fmt, ...)
+{
+	va_list args;
+	int printed_len;
+	char temp_buf[MLOGBUF_MSGMAX];
+	char *p;
+
+	va_start(args, fmt);
+	printed_len = vscnprintf(temp_buf, sizeof(temp_buf), fmt, args);
+	va_end(args);
+
+	/* Copy the output into mlogbuf */
+	if (oops_in_progress) {
+		/* mlogbuf was abandoned, use printk directly instead. */
+		printk(temp_buf);
+	} else {
+		spin_lock(&mlogbuf_wlock);
+		for (p = temp_buf; *p; p++) {
+			unsigned long next = (mlogbuf_end + 1) % MLOGBUF_SIZE;
+			if (next != mlogbuf_start) {
+				mlogbuf[mlogbuf_end] = *p;
+				mlogbuf_end = next;
+			} else {
+				/* buffer full */
+				break;
+			}
+		}
+		mlogbuf[mlogbuf_end] = '\0';
+		spin_unlock(&mlogbuf_wlock);
+	}
+}
+EXPORT_SYMBOL(ia64_mca_printk);
+
+/*
+ * Print buffered messages.
+ *  NOTE: call this after returning normal context. (ex. from salinfod)
+ */
+void ia64_mlogbuf_dump(void)
+{
+	char temp_buf[MLOGBUF_MSGMAX];
+	char *p;
+	unsigned long index;
+	unsigned long flags;
+	unsigned int printed_len;
+
+	/* Get output from mlogbuf */
+	while (mlogbuf_start != mlogbuf_end) {
+		temp_buf[0] = '\0';
+		p = temp_buf;
+		printed_len = 0;
+
+		spin_lock_irqsave(&mlogbuf_rlock, flags);
+
+		index = mlogbuf_start;
+		while (index != mlogbuf_end) {
+			*p = mlogbuf[index];
+			index = (index + 1) % MLOGBUF_SIZE;
+			if (!*p)
+				break;
+			p++;
+			if (++printed_len >= MLOGBUF_MSGMAX - 1)
+				break;
+		}
+		*p = '\0';
+		if (temp_buf[0])
+			printk(temp_buf);
+		mlogbuf_start = index;
+
+		mlogbuf_timestamp = 0;
+		spin_unlock_irqrestore(&mlogbuf_rlock, flags);
+	}
+}
+EXPORT_SYMBOL(ia64_mlogbuf_dump);
+
+/*
+ * Call this if system is going to down or if immediate flushing messages to
+ * console is required. (ex. recovery was failed, crash dump is going to be
+ * invoked, long-wait rendezvous etc.)
+ *  NOTE: this should be called from monarch.
+ */
+static void ia64_mlogbuf_finish(int wait)
+{
+	BREAK_LOGLEVEL(console_loglevel);
+
+	spin_lock_init(&mlogbuf_rlock);
+	ia64_mlogbuf_dump();
+	printk(KERN_EMERG "mlogbuf_finish: printing switched to urgent mode, "
+		"MCA/INIT might be dodgy or fail.\n");
+
+	if (!wait)
+		return;
+
+	/* wait for console */
+	printk("Delaying for 5 seconds...\n");
+	udelay(5*1000000);
+
+	mlogbuf_finished = 1;
+}
+EXPORT_SYMBOL(ia64_mlogbuf_finish);
+
+/*
+ * Print buffered messages from INIT context.
+ */
+static void ia64_mlogbuf_dump_from_init(void)
+{
+	if (mlogbuf_finished)
+		return;
+
+	if (mlogbuf_timestamp && (mlogbuf_timestamp + 30*HZ > jiffies)) {
+		printk(KERN_ERR "INIT: mlogbuf_dump is interrupted by INIT "
+			" and the system seems to be messed up.\n");
+		ia64_mlogbuf_finish(0);
+		return;
+	}
+
+	if (!spin_trylock(&mlogbuf_rlock)) {
+		printk(KERN_ERR "INIT: mlogbuf_dump is interrupted by INIT. "
+			"Generated messages other than stack dump will be "
+			"buffered to mlogbuf and will be printed later.\n");
+		printk(KERN_ERR "INIT: If messages would not printed after "
+			"this INIT, wait 30sec and assert INIT again.\n");
+		if (!mlogbuf_timestamp)
+			mlogbuf_timestamp = jiffies;
+		return;
+	}
+	spin_unlock(&mlogbuf_rlock);
+	ia64_mlogbuf_dump();
+}
 
 static void inline
 ia64_mca_spin(const char *func)
 {
-	printk(KERN_EMERG "%s: spinning here, not returning to SAL\n", func);
+	if (monarch_cpu == smp_processor_id())
+		ia64_mlogbuf_finish(0);
+	mprintk(KERN_EMERG "%s: spinning here, not returning to SAL\n", func);
 	while (1)
 		cpu_relax();
 }
@@ -344,9 +511,6 @@ ia64_mca_cpe_int_handler (int cpe_irq, void *arg, struct pt_regs *ptregs)
 	/* SAL spec states this should run w/ interrupts enabled */
 	local_irq_enable();
 
-	/* Get the CPE error record and log it */
-	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CPE);
-
 	spin_lock(&cpe_history_lock);
 	if (!cpe_poll_enabled && cpe_vector >= 0) {
 
@@ -375,7 +539,7 @@ ia64_mca_cpe_int_handler (int cpe_irq, void *arg, struct pt_regs *ptregs)
 			mod_timer(&cpe_poll_timer, jiffies + MIN_CPE_POLL_INTERVAL);
 
 			/* lock already released, get out now */
-			return IRQ_HANDLED;
+			goto out;
 		} else {
 			cpe_history[index++] = now;
 			if (index == CPE_HISTORY_LENGTH)
@@ -383,6 +547,10 @@ ia64_mca_cpe_int_handler (int cpe_irq, void *arg, struct pt_regs *ptregs)
 		}
 	}
 	spin_unlock(&cpe_history_lock);
+out:
+	/* Get the CPE error record and log it */
+	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CPE);
+
 	return IRQ_HANDLED;
 }
 
@@ -988,18 +1156,22 @@ ia64_wait_for_slaves(int monarch, const char *type)
 	}
 	if (!missing)
 		goto all_in;
-	printk(KERN_INFO "OS %s slave did not rendezvous on cpu", type);
+	/*
+	 * Maybe slave(s) dead. Print buffered messages immediately.
+	 */
+	ia64_mlogbuf_finish(0);
+	mprintk(KERN_INFO "OS %s slave did not rendezvous on cpu", type);
 	for_each_online_cpu(c) {
 		if (c == monarch)
 			continue;
 		if (ia64_mc_info.imi_rendez_checkin[c] == IA64_MCA_RENDEZ_CHECKIN_NOTDONE)
-			printk(" %d", c);
+			mprintk(" %d", c);
 	}
-	printk("\n");
+	mprintk("\n");
 	return;
 
 all_in:
-	printk(KERN_INFO "All OS %s slaves have reached rendezvous\n", type);
+	mprintk(KERN_INFO "All OS %s slaves have reached rendezvous\n", type);
 	return;
 }
 
@@ -1027,10 +1199,8 @@ ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
 	struct ia64_mca_notify_die nd =
 		{ .sos = sos, .monarch_cpu = &monarch_cpu };
 
-	oops_in_progress = 1;	/* FIXME: make printk NMI/MCA/INIT safe */
-	console_loglevel = 15;	/* make sure printks make it to console */
-	printk(KERN_INFO "Entered OS MCA handler. PSP=%lx cpu=%d monarch=%ld\n",
-		sos->proc_state_param, cpu, sos->monarch);
+	mprintk(KERN_INFO "Entered OS MCA handler. PSP=%lx cpu=%d "
+		"monarch=%ld\n", sos->proc_state_param, cpu, sos->monarch);
 
 	previous_current = ia64_mca_modify_original_stack(regs, sw, sos, "MCA");
 	monarch_cpu = cpu;
@@ -1066,6 +1236,9 @@ ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
 		rh->severity = sal_log_severity_corrected;
 		ia64_sal_clear_state_info(SAL_INFO_TYPE_MCA);
 		sos->os_status = IA64_MCA_CORRECTED;
+	} else {
+		/* Dump buffered message to console */
+		ia64_mlogbuf_finish(1);
 	}
 	if (notify_die(DIE_MCA_MONARCH_LEAVE, "MCA", regs, (long)&nd, 0, recover)
 			== NOTIFY_STOP)
@@ -1106,9 +1279,6 @@ ia64_mca_cmc_int_handler(int cmc_irq, void *arg, struct pt_regs *ptregs)
 	/* SAL spec states this should run w/ interrupts enabled */
 	local_irq_enable();
 
-	/* Get the CMC error record and log it */
-	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CMC);
-
 	spin_lock(&cmc_history_lock);
 	if (!cmc_polling_enabled) {
 		int i, count = 1; /* we know 1 happened now */
@@ -1141,7 +1311,7 @@ ia64_mca_cmc_int_handler(int cmc_irq, void *arg, struct pt_regs *ptregs)
 			mod_timer(&cmc_poll_timer, jiffies + CMC_POLL_INTERVAL);
 
 			/* lock already released, get out now */
-			return IRQ_HANDLED;
+			goto out;
 		} else {
 			cmc_history[index++] = now;
 			if (index == CMC_HISTORY_LENGTH)
@@ -1149,6 +1319,10 @@ ia64_mca_cmc_int_handler(int cmc_irq, void *arg, struct pt_regs *ptregs)
 		}
 	}
 	spin_unlock(&cmc_history_lock);
+out:
+	/* Get the CMC error record and log it */
+	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CMC);
+
 	return IRQ_HANDLED;
 }
 
@@ -1305,6 +1479,15 @@ default_monarch_init_process(struct notifier_block *self, unsigned long val, voi
 	struct task_struct *g, *t;
 	if (val != DIE_INIT_MONARCH_PROCESS)
 		return NOTIFY_DONE;
+
+	/*
+	 * FIXME: mlogbuf will brim over with INIT stack dumps.
+	 * To enable show_stack from INIT, we use oops_in_progress which should
+	 * be used in real oops. This would cause something wrong after INIT.
+	 */
+	BREAK_LOGLEVEL(console_loglevel);
+	ia64_mlogbuf_dump_from_init();
+
 	printk(KERN_ERR "Processes interrupted by INIT -");
 	for_each_online_cpu(c) {
 		struct ia64_sal_os_state *s;
@@ -1326,6 +1509,8 @@ default_monarch_init_process(struct notifier_block *self, unsigned long val, voi
 		} while_each_thread (g, t);
 		read_unlock(&tasklist_lock);
 	}
+	/* FIXME: This will not restore zapped printk locks. */
+	RESTORE_LOGLEVEL(console_loglevel);
 	return NOTIFY_DONE;
 }
 
@@ -1357,12 +1542,9 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 	struct ia64_mca_notify_die nd =
 		{ .sos = sos, .monarch_cpu = &monarch_cpu };
 
-	oops_in_progress = 1;	/* FIXME: make printk NMI/MCA/INIT safe */
-	console_loglevel = 15;	/* make sure printks make it to console */
-
 	(void) notify_die(DIE_INIT_ENTER, "INIT", regs, (long)&nd, 0, 0);
 
-	printk(KERN_INFO "Entered OS INIT handler. PSP=%lx cpu=%d monarch=%ld\n",
+	mprintk(KERN_INFO "Entered OS INIT handler. PSP=%lx cpu=%d monarch=%ld\n",
 		sos->proc_state_param, cpu, sos->monarch);
 	salinfo_log_wakeup(SAL_INFO_TYPE_INIT, NULL, 0, 0);
 
@@ -1375,7 +1557,7 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 	 * fix their proms and get their customers updated.
 	 */
 	if (!sos->monarch && atomic_add_return(1, &slaves) == num_online_cpus()) {
-		printk(KERN_WARNING "%s: Promoting cpu %d to monarch.\n",
+		mprintk(KERN_WARNING "%s: Promoting cpu %d to monarch.\n",
 		       __FUNCTION__, cpu);
 		atomic_dec(&slaves);
 		sos->monarch = 1;
@@ -1387,7 +1569,7 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 	 * fix their proms and get their customers updated.
 	 */
 	if (sos->monarch && atomic_add_return(1, &monarchs) > 1) {
-		printk(KERN_WARNING "%s: Demoting cpu %d to slave.\n",
+		mprintk(KERN_WARNING "%s: Demoting cpu %d to slave.\n",
 			       __FUNCTION__, cpu);
 		atomic_dec(&monarchs);
 		sos->monarch = 0;
@@ -1408,7 +1590,7 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 		if (notify_die(DIE_INIT_SLAVE_LEAVE, "INIT", regs, (long)&nd, 0, 0)
 				== NOTIFY_STOP)
 			ia64_mca_spin(__FUNCTION__);
-		printk("Slave on cpu %d returning to normal service.\n", cpu);
+		mprintk("Slave on cpu %d returning to normal service.\n", cpu);
 		set_curr_task(cpu, previous_current);
 		ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;
 		atomic_dec(&slaves);
@@ -1426,7 +1608,7 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 	 * same serial line, the user will need some time to switch out of the BMC before
 	 * the dump begins.
 	 */
-	printk("Delaying for 5 seconds...\n");
+	mprintk("Delaying for 5 seconds...\n");
 	udelay(5*1000000);
 	ia64_wait_for_slaves(cpu, "INIT");
 	/* If nobody intercepts DIE_INIT_MONARCH_PROCESS then we drop through
@@ -1439,7 +1621,7 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
 	if (notify_die(DIE_INIT_MONARCH_LEAVE, "INIT", regs, (long)&nd, 0, 0)
 			== NOTIFY_STOP)
 		ia64_mca_spin(__FUNCTION__);
-	printk("\nINIT dump complete.  Monarch on cpu %d returning to normal service.\n", cpu);
+	mprintk("\nINIT dump complete.  Monarch on cpu %d returning to normal service.\n", cpu);
 	atomic_dec(&monarchs);
 	set_curr_task(cpu, previous_current);
 	monarch_cpu = -1;
diff --git a/arch/ia64/kernel/mca_asm.S b/arch/ia64/kernel/mca_asm.S
index 96047491d1b..c6b607c00de 100644
--- a/arch/ia64/kernel/mca_asm.S
+++ b/arch/ia64/kernel/mca_asm.S
@@ -1025,18 +1025,13 @@ ia64_old_stack:
 
 ia64_set_kernel_registers:
 	add temp3=MCA_SP_OFFSET, r3
-	add temp4=MCA_SOS_OFFSET+SOS(OS_GP), r3
 	mov b0=r2		// save return address
 	GET_IA64_MCA_DATA(temp1)
 	;;
-	add temp4=temp4, temp1	// &struct ia64_sal_os_state.os_gp
 	add r12=temp1, temp3	// kernel stack pointer on MCA/INIT stack
 	add r13=temp1, r3	// set current to start of MCA/INIT stack
 	add r20=temp1, r3	// physical start of MCA/INIT stack
 	;;
-	ld8 r1=[temp4]		// OS GP from SAL OS state
-	;;
-	DATA_PA_TO_VA(r1,temp1)
 	DATA_PA_TO_VA(r12,temp2)
 	DATA_PA_TO_VA(r13,temp3)
 	;;
@@ -1067,6 +1062,10 @@ ia64_set_kernel_registers:
 	mov cr.itir=r18
 	mov cr.ifa=r13
 	mov r20=IA64_TR_CURRENT_STACK
+
+	movl r17=FPSR_DEFAULT
+	;;
+	mov.m ar.fpsr=r17			// set ar.fpsr to kernel default value
 	;;
 	itr.d dtr[r20]=r21
 	;;
diff --git a/arch/ia64/kernel/mca_drv.c b/arch/ia64/kernel/mca_drv.c
index 8db6e0cedad..a45009d2bc9 100644
--- a/arch/ia64/kernel/mca_drv.c
+++ b/arch/ia64/kernel/mca_drv.c
@@ -79,14 +79,30 @@ static int
 fatal_mca(const char *fmt, ...)
 {
 	va_list args;
+	char buf[256];
 
 	va_start(args, fmt);
-	vprintk(fmt, args);
+	vsnprintf(buf, sizeof(buf), fmt, args);
 	va_end(args);
+	ia64_mca_printk(KERN_ALERT "MCA: %s\n", buf);
 
 	return MCA_NOT_RECOVERED;
 }
 
+static int
+mca_recovered(const char *fmt, ...)
+{
+	va_list args;
+	char buf[256];
+
+	va_start(args, fmt);
+	vsnprintf(buf, sizeof(buf), fmt, args);
+	va_end(args);
+	ia64_mca_printk(KERN_INFO "MCA: %s\n", buf);
+
+	return MCA_RECOVERED;
+}
+
 /**
  * mca_page_isolate - isolate a poisoned page in order not to use it later
  * @paddr:	poisoned memory location
@@ -140,6 +156,7 @@ mca_page_isolate(unsigned long paddr)
 void
 mca_handler_bh(unsigned long paddr, void *iip, unsigned long ipsr)
 {
+	ia64_mlogbuf_dump();
 	printk(KERN_ERR "OS_MCA: process [cpu %d, pid: %d, uid: %d, "
 		"iip: %p, psr: 0x%lx,paddr: 0x%lx](%s) encounters MCA.\n",
 		raw_smp_processor_id(), current->pid, current->uid,
@@ -440,7 +457,7 @@ recover_from_read_error(slidx_table_t *slidx,
 
 	/* Is target address valid? */
 	if (!pbci->tv)
-		return fatal_mca(KERN_ALERT "MCA: target address not valid\n");
+		return fatal_mca("target address not valid");
 
 	/*
 	 * cpu read or memory-mapped io read
@@ -458,7 +475,7 @@ recover_from_read_error(slidx_table_t *slidx,
 
 	/* Is minstate valid? */
 	if (!peidx_bottom(peidx) || !(peidx_bottom(peidx)->valid.minstate))
-		return fatal_mca(KERN_ALERT "MCA: minstate not valid\n");
+		return fatal_mca("minstate not valid");
 	psr1 =(struct ia64_psr *)&(peidx_minstate_area(peidx)->pmsa_ipsr);
 	psr2 =(struct ia64_psr *)&(peidx_minstate_area(peidx)->pmsa_xpsr);
 
@@ -492,13 +509,14 @@ recover_from_read_error(slidx_table_t *slidx,
 			psr2->bn  = 1;
 			psr2->i  = 0;
 
-			return MCA_RECOVERED;
+			return mca_recovered("user memory corruption. "
+				"kill affected process - recovered.");
 		}
 
 	}
 
-	return fatal_mca(KERN_ALERT "MCA: kernel context not recovered,"
-			  " iip 0x%lx\n", pmsa->pmsa_iip);
+	return fatal_mca("kernel context not recovered, iip 0x%lx\n",
+			 pmsa->pmsa_iip);
 }
 
 /**
@@ -584,13 +602,13 @@ recover_from_processor_error(int platform, slidx_table_t *slidx,
 	 * The machine check is corrected.
 	 */
 	if (psp->cm == 1)
-		return MCA_RECOVERED;
+		return mca_recovered("machine check is already corrected.");
 
 	/*
 	 * The error was not contained.  Software must be reset.
 	 */
 	if (psp->us || psp->ci == 0)
-		return fatal_mca(KERN_ALERT "MCA: error not contained\n");
+		return fatal_mca("error not contained");
 
 	/*
 	 * The cache check and bus check bits have four possible states
@@ -601,22 +619,22 @@ recover_from_processor_error(int platform, slidx_table_t *slidx,
 	 *    1  1	Memory error, attempt recovery
 	 */
 	if (psp->bc == 0 || pbci == NULL)
-		return fatal_mca(KERN_ALERT "MCA: No bus check\n");
+		return fatal_mca("No bus check");
 
 	/*
 	 * Sorry, we cannot handle so many.
 	 */
 	if (peidx_bus_check_num(peidx) > 1)
-		return fatal_mca(KERN_ALERT "MCA: Too many bus checks\n");
+		return fatal_mca("Too many bus checks");
 	/*
 	 * Well, here is only one bus error.
 	 */
 	if (pbci->ib)
-		return fatal_mca(KERN_ALERT "MCA: Internal Bus error\n");
+		return fatal_mca("Internal Bus error");
 	if (pbci->cc)
-		return fatal_mca(KERN_ALERT "MCA: Cache-cache error\n");
+		return fatal_mca("Cache-cache error");
 	if (pbci->eb && pbci->bsi > 0)
-		return fatal_mca(KERN_ALERT "MCA: External bus check fatal status\n");
+		return fatal_mca("External bus check fatal status");
 
 	/*
 	 * This is a local MCA and estimated as recoverble external bus error.
@@ -628,7 +646,7 @@ recover_from_processor_error(int platform, slidx_table_t *slidx,
 	/*
 	 * On account of strange SAL error record, we cannot recover.
 	 */
-	return fatal_mca(KERN_ALERT "MCA: Strange SAL record\n");
+	return fatal_mca("Strange SAL record");
 }
 
 /**
@@ -657,10 +675,10 @@ mca_try_to_recover(void *rec, struct ia64_sal_os_state *sos)
 
 	 /* Now, OS can recover when there is one processor error section */
 	if (n_proc_err > 1)
-		return fatal_mca(KERN_ALERT "MCA: Too Many Errors\n");
+		return fatal_mca("Too Many Errors");
 	else if (n_proc_err == 0)
-		/* Weird SAL record ... We need not to recover */
-		return fatal_mca(KERN_ALERT "MCA: Weird SAL record\n");
+		/* Weird SAL record ... We can't do anything */
+		return fatal_mca("Weird SAL record");
 
 	/* Make index of processor error section */
 	mca_make_peidx((sal_log_processor_info_t*)
@@ -671,7 +689,7 @@ mca_try_to_recover(void *rec, struct ia64_sal_os_state *sos)
 
 	/* Check whether MCA is global or not */
 	if (is_mca_global(&peidx, &pbci, sos))
-		return fatal_mca(KERN_ALERT "MCA: global MCA\n");
+		return fatal_mca("global MCA");
 	
 	/* Try to recover a processor error */
 	return recover_from_processor_error(platform_err, &slidx, &peidx,
diff --git a/arch/ia64/kernel/mca_drv.h b/arch/ia64/kernel/mca_drv.h
index 31a2e52bb16..c85e943ba5f 100644
--- a/arch/ia64/kernel/mca_drv.h
+++ b/arch/ia64/kernel/mca_drv.h
@@ -118,3 +118,7 @@ struct mca_table_entry {
 
 extern const struct mca_table_entry *search_mca_tables (unsigned long addr);
 extern int mca_recover_range(unsigned long);
+extern void ia64_mca_printk(const char * fmt, ...)
+	 __attribute__ ((format (printf, 1, 2)));
+extern void ia64_mlogbuf_dump(void);
+
diff --git a/arch/ia64/kernel/numa.c b/arch/ia64/kernel/numa.c
index 1cc360c83e7..20340631179 100644
--- a/arch/ia64/kernel/numa.c
+++ b/arch/ia64/kernel/numa.c
@@ -29,6 +29,36 @@ EXPORT_SYMBOL(cpu_to_node_map);
 
 cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
 
+void __cpuinit map_cpu_to_node(int cpu, int nid)
+{
+	int oldnid;
+	if (nid < 0) { /* just initialize by zero */
+		cpu_to_node_map[cpu] = 0;
+		return;
+	}
+	/* sanity check first */
+	oldnid = cpu_to_node_map[cpu];
+	if (cpu_isset(cpu, node_to_cpu_mask[oldnid])) {
+		return; /* nothing to do */
+	}
+	/* we don't have cpu-driven node hot add yet...
+	   In usual case, node is created from SRAT at boot time. */
+	if (!node_online(nid))
+		nid = first_online_node;
+	cpu_to_node_map[cpu] = nid;
+	cpu_set(cpu, node_to_cpu_mask[nid]);
+	return;
+}
+
+void __cpuinit unmap_cpu_from_node(int cpu, int nid)
+{
+	WARN_ON(!cpu_isset(cpu, node_to_cpu_mask[nid]));
+	WARN_ON(cpu_to_node_map[cpu] != nid);
+	cpu_to_node_map[cpu] = 0;
+	cpu_clear(cpu, node_to_cpu_mask[nid]);
+}
+
+
 /**
  * build_cpu_to_node_map - setup cpu to node and node to cpumask arrays
  *
@@ -49,8 +79,6 @@ void __init build_cpu_to_node_map(void)
 				node = node_cpuid[i].nid;
 				break;
 			}
-		cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
-		if (node >= 0)
-			cpu_set(cpu, node_to_cpu_mask[node]);
+		map_cpu_to_node(cpu, node);
 	}
 }
diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index 84a7e52f56f..281004ff7b0 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -34,6 +34,7 @@
 #include <linux/file.h>
 #include <linux/poll.h>
 #include <linux/vfs.h>
+#include <linux/smp.h>
 #include <linux/pagemap.h>
 #include <linux/mount.h>
 #include <linux/bitops.h>
@@ -62,6 +63,9 @@
 
 #define PFM_INVALID_ACTIVATION	(~0UL)
 
+#define PFM_NUM_PMC_REGS	64	/* PMC save area for ctxsw */
+#define PFM_NUM_PMD_REGS	64	/* PMD save area for ctxsw */
+
 /*
  * depth of message queue
  */
@@ -296,14 +300,17 @@ typedef struct pfm_context {
 	unsigned long		ctx_reload_pmcs[4];	/* bitmask of force reload PMC on ctxsw in */
 	unsigned long		ctx_used_monitors[4];	/* bitmask of monitor PMC being used */
 
-	unsigned long		ctx_pmcs[IA64_NUM_PMC_REGS];	/*  saved copies of PMC values */
+	unsigned long		ctx_pmcs[PFM_NUM_PMC_REGS];	/*  saved copies of PMC values */
 
 	unsigned int		ctx_used_ibrs[1];		/* bitmask of used IBR (speedup ctxsw in) */
 	unsigned int		ctx_used_dbrs[1];		/* bitmask of used DBR (speedup ctxsw in) */
 	unsigned long		ctx_dbrs[IA64_NUM_DBG_REGS];	/* DBR values (cache) when not loaded */
 	unsigned long		ctx_ibrs[IA64_NUM_DBG_REGS];	/* IBR values (cache) when not loaded */
 
-	pfm_counter_t		ctx_pmds[IA64_NUM_PMD_REGS]; /* software state for PMDS */
+	pfm_counter_t		ctx_pmds[PFM_NUM_PMD_REGS]; /* software state for PMDS */
+
+	unsigned long		th_pmcs[PFM_NUM_PMC_REGS];	/* PMC thread save state */
+	unsigned long		th_pmds[PFM_NUM_PMD_REGS];	/* PMD thread save state */
 
 	u64			ctx_saved_psr_up;	/* only contains psr.up value */
 
@@ -867,7 +874,6 @@ static void
 pfm_mask_monitoring(struct task_struct *task)
 {
 	pfm_context_t *ctx = PFM_GET_CTX(task);
-	struct thread_struct *th = &task->thread;
 	unsigned long mask, val, ovfl_mask;
 	int i;
 
@@ -888,7 +894,7 @@ pfm_mask_monitoring(struct task_struct *task)
 	 * So in both cases, the live register contains the owner's
 	 * state. We can ONLY touch the PMU registers and NOT the PSR.
 	 *
-	 * As a consequence to this call, the thread->pmds[] array
+	 * As a consequence to this call, the ctx->th_pmds[] array
 	 * contains stale information which must be ignored
 	 * when context is reloaded AND monitoring is active (see
 	 * pfm_restart).
@@ -923,9 +929,9 @@ pfm_mask_monitoring(struct task_struct *task)
 	mask = ctx->ctx_used_monitors[0] >> PMU_FIRST_COUNTER;
 	for(i= PMU_FIRST_COUNTER; mask; i++, mask>>=1) {
 		if ((mask & 0x1) == 0UL) continue;
-		ia64_set_pmc(i, th->pmcs[i] & ~0xfUL);
-		th->pmcs[i] &= ~0xfUL;
-		DPRINT_ovfl(("pmc[%d]=0x%lx\n", i, th->pmcs[i]));
+		ia64_set_pmc(i, ctx->th_pmcs[i] & ~0xfUL);
+		ctx->th_pmcs[i] &= ~0xfUL;
+		DPRINT_ovfl(("pmc[%d]=0x%lx\n", i, ctx->th_pmcs[i]));
 	}
 	/*
 	 * make all of this visible
@@ -942,7 +948,6 @@ static void
 pfm_restore_monitoring(struct task_struct *task)
 {
 	pfm_context_t *ctx = PFM_GET_CTX(task);
-	struct thread_struct *th = &task->thread;
 	unsigned long mask, ovfl_mask;
 	unsigned long psr, val;
 	int i, is_system;
@@ -1008,9 +1013,9 @@ pfm_restore_monitoring(struct task_struct *task)
 	mask = ctx->ctx_used_monitors[0] >> PMU_FIRST_COUNTER;
 	for(i= PMU_FIRST_COUNTER; mask; i++, mask>>=1) {
 		if ((mask & 0x1) == 0UL) continue;
-		th->pmcs[i] = ctx->ctx_pmcs[i];
-		ia64_set_pmc(i, th->pmcs[i]);
-		DPRINT(("[%d] pmc[%d]=0x%lx\n", task->pid, i, th->pmcs[i]));
+		ctx->th_pmcs[i] = ctx->ctx_pmcs[i];
+		ia64_set_pmc(i, ctx->th_pmcs[i]);
+		DPRINT(("[%d] pmc[%d]=0x%lx\n", task->pid, i, ctx->th_pmcs[i]));
 	}
 	ia64_srlz_d();
 
@@ -1069,7 +1074,6 @@ pfm_restore_pmds(unsigned long *pmds, unsigned long mask)
 static inline void
 pfm_copy_pmds(struct task_struct *task, pfm_context_t *ctx)
 {
-	struct thread_struct *thread = &task->thread;
 	unsigned long ovfl_val = pmu_conf->ovfl_val;
 	unsigned long mask = ctx->ctx_all_pmds[0];
 	unsigned long val;
@@ -1091,11 +1095,11 @@ pfm_copy_pmds(struct task_struct *task, pfm_context_t *ctx)
 			ctx->ctx_pmds[i].val = val & ~ovfl_val;
 			 val &= ovfl_val;
 		}
-		thread->pmds[i] = val;
+		ctx->th_pmds[i] = val;
 
 		DPRINT(("pmd[%d]=0x%lx soft_val=0x%lx\n",
 			i,
-			thread->pmds[i],
+			ctx->th_pmds[i],
 			ctx->ctx_pmds[i].val));
 	}
 }
@@ -1106,7 +1110,6 @@ pfm_copy_pmds(struct task_struct *task, pfm_context_t *ctx)
 static inline void
 pfm_copy_pmcs(struct task_struct *task, pfm_context_t *ctx)
 {
-	struct thread_struct *thread = &task->thread;
 	unsigned long mask = ctx->ctx_all_pmcs[0];
 	int i;
 
@@ -1114,8 +1117,8 @@ pfm_copy_pmcs(struct task_struct *task, pfm_context_t *ctx)
 
 	for (i=0; mask; i++, mask>>=1) {
 		/* masking 0 with ovfl_val yields 0 */
-		thread->pmcs[i] = ctx->ctx_pmcs[i];
-		DPRINT(("pmc[%d]=0x%lx\n", i, thread->pmcs[i]));
+		ctx->th_pmcs[i] = ctx->ctx_pmcs[i];
+		DPRINT(("pmc[%d]=0x%lx\n", i, ctx->th_pmcs[i]));
 	}
 }
 
@@ -2859,7 +2862,6 @@ pfm_reset_regs(pfm_context_t *ctx, unsigned long *ovfl_regs, int is_long_reset)
 static int
 pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
-	struct thread_struct *thread = NULL;
 	struct task_struct *task;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
 	unsigned long value, pmc_pm;
@@ -2880,7 +2882,6 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	if (state == PFM_CTX_ZOMBIE) return -EINVAL;
 
 	if (is_loaded) {
-		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
@@ -3035,7 +3036,7 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 		 *
 		 * The value in ctx_pmcs[] can only be changed in pfm_write_pmcs().
 		 *
-		 * The value in thread->pmcs[] may be modified on overflow, i.e.,  when
+		 * The value in th_pmcs[] may be modified on overflow, i.e.,  when
 		 * monitoring needs to be stopped.
 		 */
 		if (is_monitor) CTX_USED_MONITOR(ctx, 1UL << cnum);
@@ -3049,7 +3050,7 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			/*
 			 * write thread state
 			 */
-			if (is_system == 0) thread->pmcs[cnum] = value;
+			if (is_system == 0) ctx->th_pmcs[cnum] = value;
 
 			/*
 			 * write hardware register if we can
@@ -3101,7 +3102,6 @@ error:
 static int
 pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
-	struct thread_struct *thread = NULL;
 	struct task_struct *task;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
 	unsigned long value, hw_value, ovfl_mask;
@@ -3125,7 +3125,6 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	 * the owner of the local PMU.
 	 */
 	if (likely(is_loaded)) {
-		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
@@ -3233,7 +3232,7 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			/*
 		 	 * write thread state
 		 	 */
-			if (is_system == 0) thread->pmds[cnum] = hw_value;
+			if (is_system == 0) ctx->th_pmds[cnum] = hw_value;
 
 			/*
 			 * write hardware register if we can
@@ -3299,7 +3298,6 @@ abort_mission:
 static int
 pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 {
-	struct thread_struct *thread = NULL;
 	struct task_struct *task;
 	unsigned long val = 0UL, lval, ovfl_mask, sval;
 	pfarg_reg_t *req = (pfarg_reg_t *)arg;
@@ -3323,7 +3321,6 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	if (state == PFM_CTX_ZOMBIE) return -EINVAL;
 
 	if (likely(is_loaded)) {
-		thread = &task->thread;
 		/*
 		 * In system wide and when the context is loaded, access can only happen
 		 * when the caller is running on the CPU being monitored by the session.
@@ -3385,7 +3382,7 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			 * if context is zombie, then task does not exist anymore.
 			 * In this case, we use the full value saved in the context (pfm_flush_regs()).
 			 */
-			val = is_loaded ? thread->pmds[cnum] : 0UL;
+			val = is_loaded ? ctx->th_pmds[cnum] : 0UL;
 		}
 		rd_func = pmu_conf->pmd_desc[cnum].read_check;
 
@@ -4354,8 +4351,8 @@ pfm_context_load(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	pfm_copy_pmds(task, ctx);
 	pfm_copy_pmcs(task, ctx);
 
-	pmcs_source = thread->pmcs;
-	pmds_source = thread->pmds;
+	pmcs_source = ctx->th_pmcs;
+	pmds_source = ctx->th_pmds;
 
 	/*
 	 * always the case for system-wide
@@ -5864,14 +5861,12 @@ void
 pfm_save_regs(struct task_struct *task)
 {
 	pfm_context_t *ctx;
-	struct thread_struct *t;
 	unsigned long flags;
 	u64 psr;
 
 
 	ctx = PFM_GET_CTX(task);
 	if (ctx == NULL) return;
-	t = &task->thread;
 
 	/*
  	 * we always come here with interrupts ALREADY disabled by
@@ -5929,19 +5924,19 @@ pfm_save_regs(struct task_struct *task)
 	 * guarantee we will be schedule at that same
 	 * CPU again.
 	 */
-	pfm_save_pmds(t->pmds, ctx->ctx_used_pmds[0]);
+	pfm_save_pmds(ctx->th_pmds, ctx->ctx_used_pmds[0]);
 
 	/*
 	 * save pmc0 ia64_srlz_d() done in pfm_save_pmds()
 	 * we will need it on the restore path to check
 	 * for pending overflow.
 	 */
-	t->pmcs[0] = ia64_get_pmc(0);
+	ctx->th_pmcs[0] = ia64_get_pmc(0);
 
 	/*
 	 * unfreeze PMU if had pending overflows
 	 */
-	if (t->pmcs[0] & ~0x1UL) pfm_unfreeze_pmu();
+	if (ctx->th_pmcs[0] & ~0x1UL) pfm_unfreeze_pmu();
 
 	/*
 	 * finally, allow context access.
@@ -5986,7 +5981,6 @@ static void
 pfm_lazy_save_regs (struct task_struct *task)
 {
 	pfm_context_t *ctx;
-	struct thread_struct *t;
 	unsigned long flags;
 
 	{ u64 psr  = pfm_get_psr();
@@ -5994,7 +5988,6 @@ pfm_lazy_save_regs (struct task_struct *task)
 	}
 
 	ctx = PFM_GET_CTX(task);
-	t   = &task->thread;
 
 	/*
 	 * we need to mask PMU overflow here to
@@ -6019,19 +6012,19 @@ pfm_lazy_save_regs (struct task_struct *task)
 	/*
 	 * save all the pmds we use
 	 */
-	pfm_save_pmds(t->pmds, ctx->ctx_used_pmds[0]);
+	pfm_save_pmds(ctx->th_pmds, ctx->ctx_used_pmds[0]);
 
 	/*
 	 * save pmc0 ia64_srlz_d() done in pfm_save_pmds()
 	 * it is needed to check for pended overflow
 	 * on the restore path
 	 */
-	t->pmcs[0] = ia64_get_pmc(0);
+	ctx->th_pmcs[0] = ia64_get_pmc(0);
 
 	/*
 	 * unfreeze PMU if had pending overflows
 	 */
-	if (t->pmcs[0] & ~0x1UL) pfm_unfreeze_pmu();
+	if (ctx->th_pmcs[0] & ~0x1UL) pfm_unfreeze_pmu();
 
 	/*
 	 * now get can unmask PMU interrupts, they will
@@ -6050,7 +6043,6 @@ void
 pfm_load_regs (struct task_struct *task)
 {
 	pfm_context_t *ctx;
-	struct thread_struct *t;
 	unsigned long pmc_mask = 0UL, pmd_mask = 0UL;
 	unsigned long flags;
 	u64 psr, psr_up;
@@ -6061,11 +6053,10 @@ pfm_load_regs (struct task_struct *task)
 
 	BUG_ON(GET_PMU_OWNER());
 
-	t     = &task->thread;
 	/*
 	 * possible on unload
 	 */
-	if (unlikely((t->flags & IA64_THREAD_PM_VALID) == 0)) return;
+	if (unlikely((task->thread.flags & IA64_THREAD_PM_VALID) == 0)) return;
 
 	/*
  	 * we always come here with interrupts ALREADY disabled by
@@ -6147,21 +6138,21 @@ pfm_load_regs (struct task_struct *task)
 	 *
 	 * XXX: optimize here
 	 */
-	if (pmd_mask) pfm_restore_pmds(t->pmds, pmd_mask);
-	if (pmc_mask) pfm_restore_pmcs(t->pmcs, pmc_mask);
+	if (pmd_mask) pfm_restore_pmds(ctx->th_pmds, pmd_mask);
+	if (pmc_mask) pfm_restore_pmcs(ctx->th_pmcs, pmc_mask);
 
 	/*
 	 * check for pending overflow at the time the state
 	 * was saved.
 	 */
-	if (unlikely(PMC0_HAS_OVFL(t->pmcs[0]))) {
+	if (unlikely(PMC0_HAS_OVFL(ctx->th_pmcs[0]))) {
 		/*
 		 * reload pmc0 with the overflow information
 		 * On McKinley PMU, this will trigger a PMU interrupt
 		 */
-		ia64_set_pmc(0, t->pmcs[0]);
+		ia64_set_pmc(0, ctx->th_pmcs[0]);
 		ia64_srlz_d();
-		t->pmcs[0] = 0UL;
+		ctx->th_pmcs[0] = 0UL;
 
 		/*
 		 * will replay the PMU interrupt
@@ -6214,7 +6205,6 @@ pfm_load_regs (struct task_struct *task)
 void
 pfm_load_regs (struct task_struct *task)
 {
-	struct thread_struct *t;
 	pfm_context_t *ctx;
 	struct task_struct *owner;
 	unsigned long pmd_mask, pmc_mask;
@@ -6223,7 +6213,6 @@ pfm_load_regs (struct task_struct *task)
 
 	owner = GET_PMU_OWNER();
 	ctx   = PFM_GET_CTX(task);
-	t     = &task->thread;
 	psr   = pfm_get_psr();
 
 	BUG_ON(psr & (IA64_PSR_UP|IA64_PSR_PP));
@@ -6286,22 +6275,22 @@ pfm_load_regs (struct task_struct *task)
 	 */
 	pmc_mask = ctx->ctx_all_pmcs[0];
 
-	pfm_restore_pmds(t->pmds, pmd_mask);
-	pfm_restore_pmcs(t->pmcs, pmc_mask);
+	pfm_restore_pmds(ctx->th_pmds, pmd_mask);
+	pfm_restore_pmcs(ctx->th_pmcs, pmc_mask);
 
 	/*
 	 * check for pending overflow at the time the state
 	 * was saved.
 	 */
-	if (unlikely(PMC0_HAS_OVFL(t->pmcs[0]))) {
+	if (unlikely(PMC0_HAS_OVFL(ctx->th_pmcs[0]))) {
 		/*
 		 * reload pmc0 with the overflow information
 		 * On McKinley PMU, this will trigger a PMU interrupt
 		 */
-		ia64_set_pmc(0, t->pmcs[0]);
+		ia64_set_pmc(0, ctx->th_pmcs[0]);
 		ia64_srlz_d();
 
-		t->pmcs[0] = 0UL;
+		ctx->th_pmcs[0] = 0UL;
 
 		/*
 		 * will replay the PMU interrupt
@@ -6376,11 +6365,11 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx)
 		 */
 		pfm_unfreeze_pmu();
 	} else {
-		pmc0 = task->thread.pmcs[0];
+		pmc0 = ctx->th_pmcs[0];
 		/*
 		 * clear whatever overflow status bits there were
 		 */
-		task->thread.pmcs[0] = 0;
+		ctx->th_pmcs[0] = 0;
 	}
 	ovfl_val = pmu_conf->ovfl_val;
 	/*
@@ -6401,7 +6390,7 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx)
 		/*
 		 * can access PMU always true in system wide mode
 		 */
-		val = pmd_val = can_access_pmu ? ia64_get_pmd(i) : task->thread.pmds[i];
+		val = pmd_val = can_access_pmu ? ia64_get_pmd(i) : ctx->th_pmds[i];
 
 		if (PMD_IS_COUNTING(i)) {
 			DPRINT(("[%d] pmd[%d] ctx_pmd=0x%lx hw_pmd=0x%lx\n",
@@ -6433,7 +6422,7 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx)
 
 		DPRINT(("[%d] ctx_pmd[%d]=0x%lx  pmd_val=0x%lx\n", task->pid, i, val, pmd_val));
 
-		if (is_self) task->thread.pmds[i] = pmd_val;
+		if (is_self) ctx->th_pmds[i] = pmd_val;
 
 		ctx->ctx_pmds[i].val = val;
 	}
@@ -6677,7 +6666,7 @@ pfm_init(void)
 	       ffz(pmu_conf->ovfl_val));
 
 	/* sanity check */
-	if (pmu_conf->num_pmds >= IA64_NUM_PMD_REGS || pmu_conf->num_pmcs >= IA64_NUM_PMC_REGS) {
+	if (pmu_conf->num_pmds >= PFM_NUM_PMD_REGS || pmu_conf->num_pmcs >= PFM_NUM_PMC_REGS) {
 		printk(KERN_ERR "perfmon: not enough pmc/pmd, perfmon disabled\n");
 		pmu_conf = NULL;
 		return -1;
@@ -6752,7 +6741,6 @@ void
 dump_pmu_state(const char *from)
 {
 	struct task_struct *task;
-	struct thread_struct *t;
 	struct pt_regs *regs;
 	pfm_context_t *ctx;
 	unsigned long psr, dcr, info, flags;
@@ -6797,16 +6785,14 @@ dump_pmu_state(const char *from)
 	ia64_psr(regs)->up = 0;
 	ia64_psr(regs)->pp = 0;
 
-	t = &current->thread;
-
 	for (i=1; PMC_IS_LAST(i) == 0; i++) {
 		if (PMC_IS_IMPL(i) == 0) continue;
-		printk("->CPU%d pmc[%d]=0x%lx thread_pmc[%d]=0x%lx\n", this_cpu, i, ia64_get_pmc(i), i, t->pmcs[i]);
+		printk("->CPU%d pmc[%d]=0x%lx thread_pmc[%d]=0x%lx\n", this_cpu, i, ia64_get_pmc(i), i, ctx->th_pmcs[i]);
 	}
 
 	for (i=1; PMD_IS_LAST(i) == 0; i++) {
 		if (PMD_IS_IMPL(i) == 0) continue;
-		printk("->CPU%d pmd[%d]=0x%lx thread_pmd[%d]=0x%lx\n", this_cpu, i, ia64_get_pmd(i), i, t->pmds[i]);
+		printk("->CPU%d pmd[%d]=0x%lx thread_pmd[%d]=0x%lx\n", this_cpu, i, ia64_get_pmd(i), i, ctx->th_pmds[i]);
 	}
 
 	if (ctx) {
diff --git a/arch/ia64/kernel/salinfo.c b/arch/ia64/kernel/salinfo.c
index 9065f0f01ba..e63b8ca5344 100644
--- a/arch/ia64/kernel/salinfo.c
+++ b/arch/ia64/kernel/salinfo.c
@@ -266,6 +266,7 @@ salinfo_log_wakeup(int type, u8 *buffer, u64 size, int irqsafe)
 /* Check for outstanding MCA/INIT records every minute (arbitrary) */
 #define SALINFO_TIMER_DELAY (60*HZ)
 static struct timer_list salinfo_timer;
+extern void ia64_mlogbuf_dump(void);
 
 static void
 salinfo_timeout_check(struct salinfo_data *data)
@@ -283,6 +284,7 @@ salinfo_timeout_check(struct salinfo_data *data)
 static void
 salinfo_timeout (unsigned long arg)
 {
+	ia64_mlogbuf_dump();
 	salinfo_timeout_check(salinfo_data + SAL_INFO_TYPE_MCA);
 	salinfo_timeout_check(salinfo_data + SAL_INFO_TYPE_INIT);
 	salinfo_timer.expires = jiffies + SALINFO_TIMER_DELAY;
@@ -332,6 +334,8 @@ retry:
 	if (cpu == -1)
 		goto retry;
 
+	ia64_mlogbuf_dump();
+
 	/* for next read, start checking at next CPU */
 	data->cpu_check = cpu;
 	if (++data->cpu_check == NR_CPUS)
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index 7ad0d9cc6db..84f93c0f2c6 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -509,7 +509,7 @@ show_cpuinfo (struct seq_file *m, void *v)
 		{ 1UL << 1, "spontaneous deferral"},
 		{ 1UL << 2, "16-byte atomic ops" }
 	};
-	char family[32], features[128], *cp, sep;
+	char features[128], *cp, sep;
 	struct cpuinfo_ia64 *c = v;
 	unsigned long mask;
 	unsigned long proc_freq;
@@ -517,12 +517,6 @@ show_cpuinfo (struct seq_file *m, void *v)
 
 	mask = c->features;
 
-	switch (c->family) {
-	      case 0x07:	memcpy(family, "Itanium", 8); break;
-	      case 0x1f:	memcpy(family, "Itanium 2", 10); break;
-	      default:		sprintf(family, "%u", c->family); break;
-	}
-
 	/* build the feature string: */
 	memcpy(features, " standard", 10);
 	cp = features;
@@ -553,8 +547,9 @@ show_cpuinfo (struct seq_file *m, void *v)
 		   "processor  : %d\n"
 		   "vendor     : %s\n"
 		   "arch       : IA-64\n"
-		   "family     : %s\n"
+		   "family     : %u\n"
 		   "model      : %u\n"
+		   "model name : %s\n"
 		   "revision   : %u\n"
 		   "archrev    : %u\n"
 		   "features   :%s\n"	/* don't change this---it _is_ right! */
@@ -563,7 +558,8 @@ show_cpuinfo (struct seq_file *m, void *v)
 		   "cpu MHz    : %lu.%06lu\n"
 		   "itc MHz    : %lu.%06lu\n"
 		   "BogoMIPS   : %lu.%02lu\n",
-		   cpunum, c->vendor, family, c->model, c->revision, c->archrev,
+		   cpunum, c->vendor, c->family, c->model,
+		   c->model_name, c->revision, c->archrev,
 		   features, c->ppn, c->number,
 		   proc_freq / 1000, proc_freq % 1000,
 		   c->itc_freq / 1000000, c->itc_freq % 1000000,
@@ -611,6 +607,31 @@ struct seq_operations cpuinfo_op = {
 	.show =		show_cpuinfo
 };
 
+static char brandname[128];
+
+static char * __cpuinit
+get_model_name(__u8 family, __u8 model)
+{
+	char brand[128];
+
+	if (ia64_pal_get_brand_info(brand)) {
+		if (family == 0x7)
+			memcpy(brand, "Merced", 7);
+		else if (family == 0x1f) switch (model) {
+			case 0: memcpy(brand, "McKinley", 9); break;
+			case 1: memcpy(brand, "Madison", 8); break;
+			case 2: memcpy(brand, "Madison up to 9M cache", 23); break;
+		} else
+			memcpy(brand, "Unknown", 8);
+	}
+	if (brandname[0] == '\0')
+		return strcpy(brandname, brand);
+	else if (strcmp(brandname, brand) == 0)
+		return brandname;
+	else
+		return kstrdup(brand, GFP_KERNEL);
+}
+
 static void __cpuinit
 identify_cpu (struct cpuinfo_ia64 *c)
 {
@@ -640,7 +661,6 @@ identify_cpu (struct cpuinfo_ia64 *c)
 	pal_status_t status;
 	unsigned long impl_va_msb = 50, phys_addr_size = 44;	/* Itanium defaults */
 	int i;
-
 	for (i = 0; i < 5; ++i)
 		cpuid.bits[i] = ia64_get_cpuid(i);
 
@@ -663,6 +683,7 @@ identify_cpu (struct cpuinfo_ia64 *c)
 	c->family = cpuid.field.family;
 	c->archrev = cpuid.field.archrev;
 	c->features = cpuid.field.features;
+	c->model_name = get_model_name(c->family, c->model);
 
 	status = ia64_pal_vm_summary(&vm1, &vm2);
 	if (status == PAL_STATUS_SUCCESS) {
diff --git a/arch/ia64/kernel/smpboot.c b/arch/ia64/kernel/smpboot.c
index 6203ed4ec8c..f7d7f566814 100644
--- a/arch/ia64/kernel/smpboot.c
+++ b/arch/ia64/kernel/smpboot.c
@@ -879,3 +879,27 @@ identify_siblings(struct cpuinfo_ia64 *c)
 	c->core_id = info.log1_cid;
 	c->thread_id = info.log1_tid;
 }
+
+/*
+ * returns non zero, if multi-threading is enabled
+ * on at least one physical package. Due to hotplug cpu
+ * and (maxcpus=), all threads may not necessarily be enabled
+ * even though the processor supports multi-threading.
+ */
+int is_multithreading_enabled(void)
+{
+	int i, j;
+
+	for_each_present_cpu(i) {
+		for_each_present_cpu(j) {
+			if (j == i)
+				continue;
+			if ((cpu_data(j)->socket_id == cpu_data(i)->socket_id)) {
+				if (cpu_data(j)->core_id == cpu_data(i)->core_id)
+					return 1;
+			}
+		}
+	}
+	return 0;
+}
+EXPORT_SYMBOL_GPL(is_multithreading_enabled);
diff --git a/arch/ia64/kernel/topology.c b/arch/ia64/kernel/topology.c
index f648c610b10..5629b45e89c 100644
--- a/arch/ia64/kernel/topology.c
+++ b/arch/ia64/kernel/topology.c
@@ -36,6 +36,7 @@ int arch_register_cpu(int num)
 	 */
 	if (!can_cpei_retarget() && is_cpu_cpei_target(num))
 		sysfs_cpus[num].cpu.no_control = 1;
+	map_cpu_to_node(num, node_cpuid[num].nid);
 #endif
 
 	return register_cpu(&sysfs_cpus[num].cpu, num);
@@ -45,7 +46,8 @@ int arch_register_cpu(int num)
 
 void arch_unregister_cpu(int num)
 {
-	return unregister_cpu(&sysfs_cpus[num].cpu);
+	unregister_cpu(&sysfs_cpus[num].cpu);
+	unmap_cpu_from_node(num, cpu_to_node(num));
 }
 EXPORT_SYMBOL(arch_register_cpu);
 EXPORT_SYMBOL(arch_unregister_cpu);
diff --git a/arch/ia64/kernel/uncached.c b/arch/ia64/kernel/uncached.c
index 4c73a676366..c58e933694d 100644
--- a/arch/ia64/kernel/uncached.c
+++ b/arch/ia64/kernel/uncached.c
@@ -98,7 +98,7 @@ static int uncached_add_chunk(struct uncached_pool *uc_pool, int nid)
 
 	/* attempt to allocate a granule's worth of cached memory pages */
 
-	page = alloc_pages_node(nid, GFP_KERNEL | __GFP_ZERO,
+	page = alloc_pages_node(nid, GFP_KERNEL | __GFP_ZERO | GFP_THISNODE,
 				IA64_GRANULE_SHIFT-PAGE_SHIFT);
 	if (!page) {
 		mutex_unlock(&uc_pool->add_chunk_mutex);
diff --git a/arch/ia64/kernel/vmlinux.lds.S b/arch/ia64/kernel/vmlinux.lds.S
index 5b0d5f64a9b..b3b2e389d6b 100644
--- a/arch/ia64/kernel/vmlinux.lds.S
+++ b/arch/ia64/kernel/vmlinux.lds.S
@@ -184,7 +184,9 @@ SECTIONS
 	  *(.data.gate)
 	  __stop_gate_section = .;
 	}
-  . = ALIGN(PAGE_SIZE);		/* make sure the gate page doesn't expose kernel data */
+  . = ALIGN(PAGE_SIZE);		/* make sure the gate page doesn't expose
+  				 * kernel data
+				 */
 
   .data.read_mostly : AT(ADDR(.data.read_mostly) - LOAD_OFFSET)
         { *(.data.read_mostly) }
@@ -202,7 +204,9 @@ SECTIONS
 		*(.data.percpu)
 		__per_cpu_end = .;
 	}
-  . = __phys_per_cpu_start + PERCPU_PAGE_SIZE;	/* ensure percpu data fits into percpu page size */
+  . = __phys_per_cpu_start + PERCPU_PAGE_SIZE;	/* ensure percpu data fits
+  						 * into percpu page size
+						 */
 
   data : { } :data
   .data : AT(ADDR(.data) - LOAD_OFFSET)
diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c
index e004143ba86..daf977ff292 100644
--- a/arch/ia64/mm/contig.c
+++ b/arch/ia64/mm/contig.c
@@ -26,7 +26,6 @@
 #include <asm/mca.h>
 
 #ifdef CONFIG_VIRTUAL_MEM_MAP
-static unsigned long num_dma_physpages;
 static unsigned long max_gap;
 #endif
 
@@ -41,10 +40,11 @@ show_mem (void)
 	int i, total = 0, reserved = 0;
 	int shared = 0, cached = 0;
 
-	printk("Mem-info:\n");
+	printk(KERN_INFO "Mem-info:\n");
 	show_free_areas();
 
-	printk("Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
+	printk(KERN_INFO "Free swap:       %6ldkB\n",
+	       nr_swap_pages<<(PAGE_SHIFT-10));
 	i = max_mapnr;
 	for (i = 0; i < max_mapnr; i++) {
 		if (!pfn_valid(i)) {
@@ -63,12 +63,12 @@ show_mem (void)
 		else if (page_count(mem_map + i))
 			shared += page_count(mem_map + i) - 1;
 	}
-	printk("%d pages of RAM\n", total);
-	printk("%d reserved pages\n", reserved);
-	printk("%d pages shared\n", shared);
-	printk("%d pages swap cached\n", cached);
-	printk("%ld pages in page table cache\n",
-		pgtable_quicklist_total_size());
+	printk(KERN_INFO "%d pages of RAM\n", total);
+	printk(KERN_INFO "%d reserved pages\n", reserved);
+	printk(KERN_INFO "%d pages shared\n", shared);
+	printk(KERN_INFO "%d pages swap cached\n", cached);
+	printk(KERN_INFO "%ld pages in page table cache\n",
+	       pgtable_quicklist_total_size());
 }
 
 /* physical address where the bootmem map is located */
@@ -218,18 +218,6 @@ count_pages (u64 start, u64 end, void *arg)
 	return 0;
 }
 
-#ifdef CONFIG_VIRTUAL_MEM_MAP
-static int
-count_dma_pages (u64 start, u64 end, void *arg)
-{
-	unsigned long *count = arg;
-
-	if (start < MAX_DMA_ADDRESS)
-		*count += (min(end, MAX_DMA_ADDRESS) - start) >> PAGE_SHIFT;
-	return 0;
-}
-#endif
-
 /*
  * Set up the page tables.
  */
@@ -238,45 +226,22 @@ void __init
 paging_init (void)
 {
 	unsigned long max_dma;
-	unsigned long zones_size[MAX_NR_ZONES];
-#ifdef CONFIG_VIRTUAL_MEM_MAP
-	unsigned long zholes_size[MAX_NR_ZONES];
-#endif
-
-	/* initialize mem_map[] */
-
-	memset(zones_size, 0, sizeof(zones_size));
+	unsigned long nid = 0;
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
 
 	num_physpages = 0;
 	efi_memmap_walk(count_pages, &num_physpages);
 
 	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
+	max_zone_pfns[ZONE_DMA] = max_dma;
+	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
 
 #ifdef CONFIG_VIRTUAL_MEM_MAP
-	memset(zholes_size, 0, sizeof(zholes_size));
-
-	num_dma_physpages = 0;
-	efi_memmap_walk(count_dma_pages, &num_dma_physpages);
-
-	if (max_low_pfn < max_dma) {
-		zones_size[ZONE_DMA] = max_low_pfn;
-		zholes_size[ZONE_DMA] = max_low_pfn - num_dma_physpages;
-	} else {
-		zones_size[ZONE_DMA] = max_dma;
-		zholes_size[ZONE_DMA] = max_dma - num_dma_physpages;
-		if (num_physpages > num_dma_physpages) {
-			zones_size[ZONE_NORMAL] = max_low_pfn - max_dma;
-			zholes_size[ZONE_NORMAL] =
-				((max_low_pfn - max_dma) -
-				 (num_physpages - num_dma_physpages));
-		}
-	}
-
+	efi_memmap_walk(register_active_ranges, &nid);
 	efi_memmap_walk(find_largest_hole, (u64 *)&max_gap);
 	if (max_gap < LARGE_GAP) {
 		vmem_map = (struct page *) 0;
-		free_area_init_node(0, NODE_DATA(0), zones_size, 0,
-				    zholes_size);
+		free_area_init_nodes(max_zone_pfns);
 	} else {
 		unsigned long map_size;
 
@@ -288,20 +253,19 @@ paging_init (void)
 		vmem_map = (struct page *) vmalloc_end;
 		efi_memmap_walk(create_mem_map_page_table, NULL);
 
-		NODE_DATA(0)->node_mem_map = vmem_map;
-		free_area_init_node(0, NODE_DATA(0), zones_size,
-				    0, zholes_size);
+		/*
+		 * alloc_node_mem_map makes an adjustment for mem_map
+		 * which isn't compatible with vmem_map.
+		 */
+		NODE_DATA(0)->node_mem_map = vmem_map +
+			find_min_pfn_with_active_regions();
+		free_area_init_nodes(max_zone_pfns);
 
 		printk("Virtual mem_map starts at 0x%p\n", mem_map);
 	}
 #else /* !CONFIG_VIRTUAL_MEM_MAP */
-	if (max_low_pfn < max_dma)
-		zones_size[ZONE_DMA] = max_low_pfn;
-	else {
-		zones_size[ZONE_DMA] = max_dma;
-		zones_size[ZONE_NORMAL] = max_low_pfn - max_dma;
-	}
-	free_area_init(zones_size);
+	add_active_range(0, 0, max_low_pfn);
+	free_area_init_nodes(max_zone_pfns);
 #endif /* !CONFIG_VIRTUAL_MEM_MAP */
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index d260bffa01a..d497b6b0f5b 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -547,15 +547,16 @@ void show_mem(void)
 	unsigned long total_present = 0;
 	pg_data_t *pgdat;
 
-	printk("Mem-info:\n");
+	printk(KERN_INFO "Mem-info:\n");
 	show_free_areas();
-	printk("Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
+	printk(KERN_INFO "Free swap:       %6ldkB\n",
+	       nr_swap_pages<<(PAGE_SHIFT-10));
+	printk(KERN_INFO "Node memory in pages:\n");
 	for_each_online_pgdat(pgdat) {
 		unsigned long present;
 		unsigned long flags;
 		int shared = 0, cached = 0, reserved = 0;
 
-		printk("Node ID: %d\n", pgdat->node_id);
 		pgdat_resize_lock(pgdat, &flags);
 		present = pgdat->node_present_pages;
 		for(i = 0; i < pgdat->node_spanned_pages; i++) {
@@ -579,18 +580,17 @@ void show_mem(void)
 		total_reserved += reserved;
 		total_cached += cached;
 		total_shared += shared;
-		printk("\t%ld pages of RAM\n", present);
-		printk("\t%d reserved pages\n", reserved);
-		printk("\t%d pages shared\n", shared);
-		printk("\t%d pages swap cached\n", cached);
+		printk(KERN_INFO "Node %4d:  RAM: %11ld, rsvd: %8d, "
+		       "shrd: %10d, swpd: %10d\n", pgdat->node_id,
+		       present, reserved, shared, cached);
 	}
-	printk("%ld pages of RAM\n", total_present);
-	printk("%d reserved pages\n", total_reserved);
-	printk("%d pages shared\n", total_shared);
-	printk("%d pages swap cached\n", total_cached);
-	printk("Total of %ld pages in page table cache\n",
-		pgtable_quicklist_total_size());
-	printk("%d free buffer pages\n", nr_free_buffer_pages());
+	printk(KERN_INFO "%ld pages of RAM\n", total_present);
+	printk(KERN_INFO "%d reserved pages\n", total_reserved);
+	printk(KERN_INFO "%d pages shared\n", total_shared);
+	printk(KERN_INFO "%d pages swap cached\n", total_cached);
+	printk(KERN_INFO "Total of %ld pages in page table cache\n",
+	       pgtable_quicklist_total_size());
+	printk(KERN_INFO "%d free buffer pages\n", nr_free_buffer_pages());
 }
 
 /**
@@ -654,6 +654,7 @@ static __init int count_node_pages(unsigned long start, unsigned long len, int n
 {
 	unsigned long end = start + len;
 
+	add_active_range(node, start >> PAGE_SHIFT, end >> PAGE_SHIFT);
 	mem_data[node].num_physpages += len >> PAGE_SHIFT;
 	if (start <= __pa(MAX_DMA_ADDRESS))
 		mem_data[node].num_dma_physpages +=
@@ -678,10 +679,10 @@ static __init int count_node_pages(unsigned long start, unsigned long len, int n
 void __init paging_init(void)
 {
 	unsigned long max_dma;
-	unsigned long zones_size[MAX_NR_ZONES];
-	unsigned long zholes_size[MAX_NR_ZONES];
 	unsigned long pfn_offset = 0;
+	unsigned long max_pfn = 0;
 	int node;
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
 
 	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 
@@ -698,47 +699,20 @@ void __init paging_init(void)
 #endif
 
 	for_each_online_node(node) {
-		memset(zones_size, 0, sizeof(zones_size));
-		memset(zholes_size, 0, sizeof(zholes_size));
-
 		num_physpages += mem_data[node].num_physpages;
-
-		if (mem_data[node].min_pfn >= max_dma) {
-			/* All of this node's memory is above ZONE_DMA */
-			zones_size[ZONE_NORMAL] = mem_data[node].max_pfn -
-				mem_data[node].min_pfn;
-			zholes_size[ZONE_NORMAL] = mem_data[node].max_pfn -
-				mem_data[node].min_pfn -
-				mem_data[node].num_physpages;
-		} else if (mem_data[node].max_pfn < max_dma) {
-			/* All of this node's memory is in ZONE_DMA */
-			zones_size[ZONE_DMA] = mem_data[node].max_pfn -
-				mem_data[node].min_pfn;
-			zholes_size[ZONE_DMA] = mem_data[node].max_pfn -
-				mem_data[node].min_pfn -
-				mem_data[node].num_dma_physpages;
-		} else {
-			/* This node has memory in both zones */
-			zones_size[ZONE_DMA] = max_dma -
-				mem_data[node].min_pfn;
-			zholes_size[ZONE_DMA] = zones_size[ZONE_DMA] -
-				mem_data[node].num_dma_physpages;
-			zones_size[ZONE_NORMAL] = mem_data[node].max_pfn -
-				max_dma;
-			zholes_size[ZONE_NORMAL] = zones_size[ZONE_NORMAL] -
-				(mem_data[node].num_physpages -
-				 mem_data[node].num_dma_physpages);
-		}
-
 		pfn_offset = mem_data[node].min_pfn;
 
 #ifdef CONFIG_VIRTUAL_MEM_MAP
 		NODE_DATA(node)->node_mem_map = vmem_map + pfn_offset;
 #endif
-		free_area_init_node(node, NODE_DATA(node), zones_size,
-				    pfn_offset, zholes_size);
+		if (mem_data[node].max_pfn > max_pfn)
+			max_pfn = mem_data[node].max_pfn;
 	}
 
+	max_zone_pfns[ZONE_DMA] = max_dma;
+	max_zone_pfns[ZONE_NORMAL] = max_pfn;
+	free_area_init_nodes(max_zone_pfns);
+
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
 
diff --git a/arch/ia64/mm/init.c b/arch/ia64/mm/init.c
index 30617ccb4f7..ff87a5cba39 100644
--- a/arch/ia64/mm/init.c
+++ b/arch/ia64/mm/init.c
@@ -593,6 +593,18 @@ find_largest_hole (u64 start, u64 end, void *arg)
 	last_end = end;
 	return 0;
 }
+
+int __init
+register_active_ranges(u64 start, u64 end, void *nid)
+{
+	BUG_ON(nid == NULL);
+	BUG_ON(*(unsigned long *)nid >= MAX_NUMNODES);
+
+	add_active_range(*(unsigned long *)nid,
+				__pa(start) >> PAGE_SHIFT,
+				__pa(end) >> PAGE_SHIFT);
+	return 0;
+}
 #endif /* CONFIG_VIRTUAL_MEM_MAP */
 
 static int __init
diff --git a/arch/ia64/pci/pci.c b/arch/ia64/pci/pci.c
index 60b45e79f08..15c7c670da3 100644
--- a/arch/ia64/pci/pci.c
+++ b/arch/ia64/pci/pci.c
@@ -562,7 +562,8 @@ pcibios_enable_device (struct pci_dev *dev, int mask)
 void
 pcibios_disable_device (struct pci_dev *dev)
 {
-	acpi_pci_irq_disable(dev);
+	if (dev->is_enabled)
+		acpi_pci_irq_disable(dev);
 }
 
 void
diff --git a/arch/ia64/sn/kernel/bte.c b/arch/ia64/sn/kernel/bte.c
index 27dee458406..7f73ad4408a 100644
--- a/arch/ia64/sn/kernel/bte.c
+++ b/arch/ia64/sn/kernel/bte.c
@@ -277,8 +277,7 @@ bte_result_t bte_unaligned_copy(u64 src, u64 dest, u64 len, u64 mode)
 	}
 
 	/* temporary buffer used during unaligned transfers */
-	bteBlock_unaligned = kmalloc(len + 3 * L1_CACHE_BYTES,
-				     GFP_KERNEL | GFP_DMA);
+	bteBlock_unaligned = kmalloc(len + 3 * L1_CACHE_BYTES, GFP_KERNEL);
 	if (bteBlock_unaligned == NULL) {
 		return BTEFAIL_NOTAVAIL;
 	}
diff --git a/arch/ia64/sn/kernel/sn2/sn_hwperf.c b/arch/ia64/sn/kernel/sn2/sn_hwperf.c
index 9a8a29339d2..b632b9c1e3b 100644
--- a/arch/ia64/sn/kernel/sn2/sn_hwperf.c
+++ b/arch/ia64/sn/kernel/sn2/sn_hwperf.c
@@ -32,9 +32,10 @@
 #include <linux/cpumask.h>
 #include <linux/smp_lock.h>
 #include <linux/nodemask.h>
+#include <linux/smp.h>
+
 #include <asm/processor.h>
 #include <asm/topology.h>
-#include <asm/smp.h>
 #include <asm/semaphore.h>
 #include <asm/uaccess.h>
 #include <asm/sal.h>