Merge branch 'x86/amd-nb' into x86/apic-cleanups

Reason: apic cleanup series depends on x86/apic, x86/amd-nb x86/platform

Conflicts:
	arch/x86/include/asm/io_apic.h

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

25 files changed, 5464 insertions, 0 deletions

diff --git a/arch/x86/platform/Makefile b/arch/x86/platform/Makefile
new file mode 100644
index 00000000000..7bf70b812fa
--- /dev/null
+++ b/arch/x86/platform/Makefile
@@ -0,0 +1,8 @@
+# Platform specific code goes here
+obj-y	+= efi/
+obj-y	+= mrst/
+obj-y	+= olpc/
+obj-y	+= scx200/
+obj-y	+= sfi/
+obj-y	+= visws/
+obj-y	+= uv/
diff --git a/arch/x86/platform/efi/Makefile b/arch/x86/platform/efi/Makefile
new file mode 100644
index 00000000000..73b8be0f367
--- /dev/null
+++ b/arch/x86/platform/efi/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_EFI) 		+= efi.o efi_$(BITS).o efi_stub_$(BITS).o
diff --git a/arch/x86/platform/efi/efi.c b/arch/x86/platform/efi/efi.c
new file mode 100644
index 00000000000..0fe27d7c625
--- /dev/null
+++ b/arch/x86/platform/efi/efi.c
@@ -0,0 +1,613 @@
+/*
+ * Common EFI (Extensible Firmware Interface) support functions
+ * Based on Extensible Firmware Interface Specification version 1.0
+ *
+ * Copyright (C) 1999 VA Linux Systems
+ * Copyright (C) 1999 Walt Drummond <drummond@valinux.com>
+ * Copyright (C) 1999-2002 Hewlett-Packard Co.
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ *	Stephane Eranian <eranian@hpl.hp.com>
+ * Copyright (C) 2005-2008 Intel Co.
+ *	Fenghua Yu <fenghua.yu@intel.com>
+ *	Bibo Mao <bibo.mao@intel.com>
+ *	Chandramouli Narayanan <mouli@linux.intel.com>
+ *	Huang Ying <ying.huang@intel.com>
+ *
+ * Copied from efi_32.c to eliminate the duplicated code between EFI
+ * 32/64 support code. --ying 2007-10-26
+ *
+ * All EFI Runtime Services are not implemented yet as EFI only
+ * supports physical mode addressing on SoftSDV. This is to be fixed
+ * in a future version.  --drummond 1999-07-20
+ *
+ * Implemented EFI runtime services and virtual mode calls.  --davidm
+ *
+ * Goutham Rao: <goutham.rao@intel.com>
+ *	Skip non-WB memory and ignore empty memory ranges.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/efi.h>
+#include <linux/bootmem.h>
+#include <linux/memblock.h>
+#include <linux/spinlock.h>
+#include <linux/uaccess.h>
+#include <linux/time.h>
+#include <linux/io.h>
+#include <linux/reboot.h>
+#include <linux/bcd.h>
+
+#include <asm/setup.h>
+#include <asm/efi.h>
+#include <asm/time.h>
+#include <asm/cacheflush.h>
+#include <asm/tlbflush.h>
+#include <asm/x86_init.h>
+
+#define EFI_DEBUG	1
+#define PFX 		"EFI: "
+
+int efi_enabled;
+EXPORT_SYMBOL(efi_enabled);
+
+struct efi efi;
+EXPORT_SYMBOL(efi);
+
+struct efi_memory_map memmap;
+
+static struct efi efi_phys __initdata;
+static efi_system_table_t efi_systab __initdata;
+
+static int __init setup_noefi(char *arg)
+{
+	efi_enabled = 0;
+	return 0;
+}
+early_param("noefi", setup_noefi);
+
+int add_efi_memmap;
+EXPORT_SYMBOL(add_efi_memmap);
+
+static int __init setup_add_efi_memmap(char *arg)
+{
+	add_efi_memmap = 1;
+	return 0;
+}
+early_param("add_efi_memmap", setup_add_efi_memmap);
+
+
+static efi_status_t virt_efi_get_time(efi_time_t *tm, efi_time_cap_t *tc)
+{
+	return efi_call_virt2(get_time, tm, tc);
+}
+
+static efi_status_t virt_efi_set_time(efi_time_t *tm)
+{
+	return efi_call_virt1(set_time, tm);
+}
+
+static efi_status_t virt_efi_get_wakeup_time(efi_bool_t *enabled,
+					     efi_bool_t *pending,
+					     efi_time_t *tm)
+{
+	return efi_call_virt3(get_wakeup_time,
+			      enabled, pending, tm);
+}
+
+static efi_status_t virt_efi_set_wakeup_time(efi_bool_t enabled, efi_time_t *tm)
+{
+	return efi_call_virt2(set_wakeup_time,
+			      enabled, tm);
+}
+
+static efi_status_t virt_efi_get_variable(efi_char16_t *name,
+					  efi_guid_t *vendor,
+					  u32 *attr,
+					  unsigned long *data_size,
+					  void *data)
+{
+	return efi_call_virt5(get_variable,
+			      name, vendor, attr,
+			      data_size, data);
+}
+
+static efi_status_t virt_efi_get_next_variable(unsigned long *name_size,
+					       efi_char16_t *name,
+					       efi_guid_t *vendor)
+{
+	return efi_call_virt3(get_next_variable,
+			      name_size, name, vendor);
+}
+
+static efi_status_t virt_efi_set_variable(efi_char16_t *name,
+					  efi_guid_t *vendor,
+					  unsigned long attr,
+					  unsigned long data_size,
+					  void *data)
+{
+	return efi_call_virt5(set_variable,
+			      name, vendor, attr,
+			      data_size, data);
+}
+
+static efi_status_t virt_efi_get_next_high_mono_count(u32 *count)
+{
+	return efi_call_virt1(get_next_high_mono_count, count);
+}
+
+static void virt_efi_reset_system(int reset_type,
+				  efi_status_t status,
+				  unsigned long data_size,
+				  efi_char16_t *data)
+{
+	efi_call_virt4(reset_system, reset_type, status,
+		       data_size, data);
+}
+
+static efi_status_t virt_efi_set_virtual_address_map(
+	unsigned long memory_map_size,
+	unsigned long descriptor_size,
+	u32 descriptor_version,
+	efi_memory_desc_t *virtual_map)
+{
+	return efi_call_virt4(set_virtual_address_map,
+			      memory_map_size, descriptor_size,
+			      descriptor_version, virtual_map);
+}
+
+static efi_status_t __init phys_efi_set_virtual_address_map(
+	unsigned long memory_map_size,
+	unsigned long descriptor_size,
+	u32 descriptor_version,
+	efi_memory_desc_t *virtual_map)
+{
+	efi_status_t status;
+
+	efi_call_phys_prelog();
+	status = efi_call_phys4(efi_phys.set_virtual_address_map,
+				memory_map_size, descriptor_size,
+				descriptor_version, virtual_map);
+	efi_call_phys_epilog();
+	return status;
+}
+
+static efi_status_t __init phys_efi_get_time(efi_time_t *tm,
+					     efi_time_cap_t *tc)
+{
+	efi_status_t status;
+
+	efi_call_phys_prelog();
+	status = efi_call_phys2(efi_phys.get_time, tm, tc);
+	efi_call_phys_epilog();
+	return status;
+}
+
+int efi_set_rtc_mmss(unsigned long nowtime)
+{
+	int real_seconds, real_minutes;
+	efi_status_t 	status;
+	efi_time_t 	eft;
+	efi_time_cap_t 	cap;
+
+	status = efi.get_time(&eft, &cap);
+	if (status != EFI_SUCCESS) {
+		printk(KERN_ERR "Oops: efitime: can't read time!\n");
+		return -1;
+	}
+
+	real_seconds = nowtime % 60;
+	real_minutes = nowtime / 60;
+	if (((abs(real_minutes - eft.minute) + 15)/30) & 1)
+		real_minutes += 30;
+	real_minutes %= 60;
+	eft.minute = real_minutes;
+	eft.second = real_seconds;
+
+	status = efi.set_time(&eft);
+	if (status != EFI_SUCCESS) {
+		printk(KERN_ERR "Oops: efitime: can't write time!\n");
+		return -1;
+	}
+	return 0;
+}
+
+unsigned long efi_get_time(void)
+{
+	efi_status_t status;
+	efi_time_t eft;
+	efi_time_cap_t cap;
+
+	status = efi.get_time(&eft, &cap);
+	if (status != EFI_SUCCESS)
+		printk(KERN_ERR "Oops: efitime: can't read time!\n");
+
+	return mktime(eft.year, eft.month, eft.day, eft.hour,
+		      eft.minute, eft.second);
+}
+
+/*
+ * Tell the kernel about the EFI memory map.  This might include
+ * more than the max 128 entries that can fit in the e820 legacy
+ * (zeropage) memory map.
+ */
+
+static void __init do_add_efi_memmap(void)
+{
+	void *p;
+
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		efi_memory_desc_t *md = p;
+		unsigned long long start = md->phys_addr;
+		unsigned long long size = md->num_pages << EFI_PAGE_SHIFT;
+		int e820_type;
+
+		switch (md->type) {
+		case EFI_LOADER_CODE:
+		case EFI_LOADER_DATA:
+		case EFI_BOOT_SERVICES_CODE:
+		case EFI_BOOT_SERVICES_DATA:
+		case EFI_CONVENTIONAL_MEMORY:
+			if (md->attribute & EFI_MEMORY_WB)
+				e820_type = E820_RAM;
+			else
+				e820_type = E820_RESERVED;
+			break;
+		case EFI_ACPI_RECLAIM_MEMORY:
+			e820_type = E820_ACPI;
+			break;
+		case EFI_ACPI_MEMORY_NVS:
+			e820_type = E820_NVS;
+			break;
+		case EFI_UNUSABLE_MEMORY:
+			e820_type = E820_UNUSABLE;
+			break;
+		default:
+			/*
+			 * EFI_RESERVED_TYPE EFI_RUNTIME_SERVICES_CODE
+			 * EFI_RUNTIME_SERVICES_DATA EFI_MEMORY_MAPPED_IO
+			 * EFI_MEMORY_MAPPED_IO_PORT_SPACE EFI_PAL_CODE
+			 */
+			e820_type = E820_RESERVED;
+			break;
+		}
+		e820_add_region(start, size, e820_type);
+	}
+	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
+}
+
+void __init efi_memblock_x86_reserve_range(void)
+{
+	unsigned long pmap;
+
+#ifdef CONFIG_X86_32
+	pmap = boot_params.efi_info.efi_memmap;
+#else
+	pmap = (boot_params.efi_info.efi_memmap |
+		((__u64)boot_params.efi_info.efi_memmap_hi<<32));
+#endif
+	memmap.phys_map = (void *)pmap;
+	memmap.nr_map = boot_params.efi_info.efi_memmap_size /
+		boot_params.efi_info.efi_memdesc_size;
+	memmap.desc_version = boot_params.efi_info.efi_memdesc_version;
+	memmap.desc_size = boot_params.efi_info.efi_memdesc_size;
+	memblock_x86_reserve_range(pmap, pmap + memmap.nr_map * memmap.desc_size,
+		      "EFI memmap");
+}
+
+#if EFI_DEBUG
+static void __init print_efi_memmap(void)
+{
+	efi_memory_desc_t *md;
+	void *p;
+	int i;
+
+	for (p = memmap.map, i = 0;
+	     p < memmap.map_end;
+	     p += memmap.desc_size, i++) {
+		md = p;
+		printk(KERN_INFO PFX "mem%02u: type=%u, attr=0x%llx, "
+			"range=[0x%016llx-0x%016llx) (%lluMB)\n",
+			i, md->type, md->attribute, md->phys_addr,
+			md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT),
+			(md->num_pages >> (20 - EFI_PAGE_SHIFT)));
+	}
+}
+#endif  /*  EFI_DEBUG  */
+
+void __init efi_init(void)
+{
+	efi_config_table_t *config_tables;
+	efi_runtime_services_t *runtime;
+	efi_char16_t *c16;
+	char vendor[100] = "unknown";
+	int i = 0;
+	void *tmp;
+
+#ifdef CONFIG_X86_32
+	efi_phys.systab = (efi_system_table_t *)boot_params.efi_info.efi_systab;
+#else
+	efi_phys.systab = (efi_system_table_t *)
+		(boot_params.efi_info.efi_systab |
+		 ((__u64)boot_params.efi_info.efi_systab_hi<<32));
+#endif
+
+	efi.systab = early_ioremap((unsigned long)efi_phys.systab,
+				   sizeof(efi_system_table_t));
+	if (efi.systab == NULL)
+		printk(KERN_ERR "Couldn't map the EFI system table!\n");
+	memcpy(&efi_systab, efi.systab, sizeof(efi_system_table_t));
+	early_iounmap(efi.systab, sizeof(efi_system_table_t));
+	efi.systab = &efi_systab;
+
+	/*
+	 * Verify the EFI Table
+	 */
+	if (efi.systab->hdr.signature != EFI_SYSTEM_TABLE_SIGNATURE)
+		printk(KERN_ERR "EFI system table signature incorrect!\n");
+	if ((efi.systab->hdr.revision >> 16) == 0)
+		printk(KERN_ERR "Warning: EFI system table version "
+		       "%d.%02d, expected 1.00 or greater!\n",
+		       efi.systab->hdr.revision >> 16,
+		       efi.systab->hdr.revision & 0xffff);
+
+	/*
+	 * Show what we know for posterity
+	 */
+	c16 = tmp = early_ioremap(efi.systab->fw_vendor, 2);
+	if (c16) {
+		for (i = 0; i < sizeof(vendor) - 1 && *c16; ++i)
+			vendor[i] = *c16++;
+		vendor[i] = '\0';
+	} else
+		printk(KERN_ERR PFX "Could not map the firmware vendor!\n");
+	early_iounmap(tmp, 2);
+
+	printk(KERN_INFO "EFI v%u.%.02u by %s\n",
+	       efi.systab->hdr.revision >> 16,
+	       efi.systab->hdr.revision & 0xffff, vendor);
+
+	/*
+	 * Let's see what config tables the firmware passed to us.
+	 */
+	config_tables = early_ioremap(
+		efi.systab->tables,
+		efi.systab->nr_tables * sizeof(efi_config_table_t));
+	if (config_tables == NULL)
+		printk(KERN_ERR "Could not map EFI Configuration Table!\n");
+
+	printk(KERN_INFO);
+	for (i = 0; i < efi.systab->nr_tables; i++) {
+		if (!efi_guidcmp(config_tables[i].guid, MPS_TABLE_GUID)) {
+			efi.mps = config_tables[i].table;
+			printk(" MPS=0x%lx ", config_tables[i].table);
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					ACPI_20_TABLE_GUID)) {
+			efi.acpi20 = config_tables[i].table;
+			printk(" ACPI 2.0=0x%lx ", config_tables[i].table);
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					ACPI_TABLE_GUID)) {
+			efi.acpi = config_tables[i].table;
+			printk(" ACPI=0x%lx ", config_tables[i].table);
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					SMBIOS_TABLE_GUID)) {
+			efi.smbios = config_tables[i].table;
+			printk(" SMBIOS=0x%lx ", config_tables[i].table);
+#ifdef CONFIG_X86_UV
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					UV_SYSTEM_TABLE_GUID)) {
+			efi.uv_systab = config_tables[i].table;
+			printk(" UVsystab=0x%lx ", config_tables[i].table);
+#endif
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					HCDP_TABLE_GUID)) {
+			efi.hcdp = config_tables[i].table;
+			printk(" HCDP=0x%lx ", config_tables[i].table);
+		} else if (!efi_guidcmp(config_tables[i].guid,
+					UGA_IO_PROTOCOL_GUID)) {
+			efi.uga = config_tables[i].table;
+			printk(" UGA=0x%lx ", config_tables[i].table);
+		}
+	}
+	printk("\n");
+	early_iounmap(config_tables,
+			  efi.systab->nr_tables * sizeof(efi_config_table_t));
+
+	/*
+	 * Check out the runtime services table. We need to map
+	 * the runtime services table so that we can grab the physical
+	 * address of several of the EFI runtime functions, needed to
+	 * set the firmware into virtual mode.
+	 */
+	runtime = early_ioremap((unsigned long)efi.systab->runtime,
+				sizeof(efi_runtime_services_t));
+	if (runtime != NULL) {
+		/*
+		 * We will only need *early* access to the following
+		 * two EFI runtime services before set_virtual_address_map
+		 * is invoked.
+		 */
+		efi_phys.get_time = (efi_get_time_t *)runtime->get_time;
+		efi_phys.set_virtual_address_map =
+			(efi_set_virtual_address_map_t *)
+			runtime->set_virtual_address_map;
+		/*
+		 * Make efi_get_time can be called before entering
+		 * virtual mode.
+		 */
+		efi.get_time = phys_efi_get_time;
+	} else
+		printk(KERN_ERR "Could not map the EFI runtime service "
+		       "table!\n");
+	early_iounmap(runtime, sizeof(efi_runtime_services_t));
+
+	/* Map the EFI memory map */
+	memmap.map = early_ioremap((unsigned long)memmap.phys_map,
+				   memmap.nr_map * memmap.desc_size);
+	if (memmap.map == NULL)
+		printk(KERN_ERR "Could not map the EFI memory map!\n");
+	memmap.map_end = memmap.map + (memmap.nr_map * memmap.desc_size);
+
+	if (memmap.desc_size != sizeof(efi_memory_desc_t))
+		printk(KERN_WARNING
+		  "Kernel-defined memdesc doesn't match the one from EFI!\n");
+
+	if (add_efi_memmap)
+		do_add_efi_memmap();
+
+#ifdef CONFIG_X86_32
+	x86_platform.get_wallclock = efi_get_time;
+	x86_platform.set_wallclock = efi_set_rtc_mmss;
+#endif
+
+	/* Setup for EFI runtime service */
+	reboot_type = BOOT_EFI;
+
+#if EFI_DEBUG
+	print_efi_memmap();
+#endif
+}
+
+static void __init runtime_code_page_mkexec(void)
+{
+	efi_memory_desc_t *md;
+	void *p;
+	u64 addr, npages;
+
+	/* Make EFI runtime service code area executable */
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
+
+		if (md->type != EFI_RUNTIME_SERVICES_CODE)
+			continue;
+
+		addr = md->virt_addr;
+		npages = md->num_pages;
+		memrange_efi_to_native(&addr, &npages);
+		set_memory_x(addr, npages);
+	}
+}
+
+/*
+ * This function will switch the EFI runtime services to virtual mode.
+ * Essentially, look through the EFI memmap and map every region that
+ * has the runtime attribute bit set in its memory descriptor and update
+ * that memory descriptor with the virtual address obtained from ioremap().
+ * This enables the runtime services to be called without having to
+ * thunk back into physical mode for every invocation.
+ */
+void __init efi_enter_virtual_mode(void)
+{
+	efi_memory_desc_t *md;
+	efi_status_t status;
+	unsigned long size;
+	u64 end, systab, addr, npages, end_pfn;
+	void *p, *va;
+
+	efi.systab = NULL;
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
+		if (!(md->attribute & EFI_MEMORY_RUNTIME))
+			continue;
+
+		size = md->num_pages << EFI_PAGE_SHIFT;
+		end = md->phys_addr + size;
+
+		end_pfn = PFN_UP(end);
+		if (end_pfn <= max_low_pfn_mapped
+		    || (end_pfn > (1UL << (32 - PAGE_SHIFT))
+			&& end_pfn <= max_pfn_mapped))
+			va = __va(md->phys_addr);
+		else
+			va = efi_ioremap(md->phys_addr, size, md->type);
+
+		md->virt_addr = (u64) (unsigned long) va;
+
+		if (!va) {
+			printk(KERN_ERR PFX "ioremap of 0x%llX failed!\n",
+			       (unsigned long long)md->phys_addr);
+			continue;
+		}
+
+		if (!(md->attribute & EFI_MEMORY_WB)) {
+			addr = md->virt_addr;
+			npages = md->num_pages;
+			memrange_efi_to_native(&addr, &npages);
+			set_memory_uc(addr, npages);
+		}
+
+		systab = (u64) (unsigned long) efi_phys.systab;
+		if (md->phys_addr <= systab && systab < end) {
+			systab += md->virt_addr - md->phys_addr;
+			efi.systab = (efi_system_table_t *) (unsigned long) systab;
+		}
+	}
+
+	BUG_ON(!efi.systab);
+
+	status = phys_efi_set_virtual_address_map(
+		memmap.desc_size * memmap.nr_map,
+		memmap.desc_size,
+		memmap.desc_version,
+		memmap.phys_map);
+
+	if (status != EFI_SUCCESS) {
+		printk(KERN_ALERT "Unable to switch EFI into virtual mode "
+		       "(status=%lx)!\n", status);
+		panic("EFI call to SetVirtualAddressMap() failed!");
+	}
+
+	/*
+	 * Now that EFI is in virtual mode, update the function
+	 * pointers in the runtime service table to the new virtual addresses.
+	 *
+	 * Call EFI services through wrapper functions.
+	 */
+	efi.get_time = virt_efi_get_time;
+	efi.set_time = virt_efi_set_time;
+	efi.get_wakeup_time = virt_efi_get_wakeup_time;
+	efi.set_wakeup_time = virt_efi_set_wakeup_time;
+	efi.get_variable = virt_efi_get_variable;
+	efi.get_next_variable = virt_efi_get_next_variable;
+	efi.set_variable = virt_efi_set_variable;
+	efi.get_next_high_mono_count = virt_efi_get_next_high_mono_count;
+	efi.reset_system = virt_efi_reset_system;
+	efi.set_virtual_address_map = virt_efi_set_virtual_address_map;
+	if (__supported_pte_mask & _PAGE_NX)
+		runtime_code_page_mkexec();
+	early_iounmap(memmap.map, memmap.nr_map * memmap.desc_size);
+	memmap.map = NULL;
+}
+
+/*
+ * Convenience functions to obtain memory types and attributes
+ */
+u32 efi_mem_type(unsigned long phys_addr)
+{
+	efi_memory_desc_t *md;
+	void *p;
+
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
+		if ((md->phys_addr <= phys_addr) &&
+		    (phys_addr < (md->phys_addr +
+				  (md->num_pages << EFI_PAGE_SHIFT))))
+			return md->type;
+	}
+	return 0;
+}
+
+u64 efi_mem_attributes(unsigned long phys_addr)
+{
+	efi_memory_desc_t *md;
+	void *p;
+
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
+		if ((md->phys_addr <= phys_addr) &&
+		    (phys_addr < (md->phys_addr +
+				  (md->num_pages << EFI_PAGE_SHIFT))))
+			return md->attribute;
+	}
+	return 0;
+}
diff --git a/arch/x86/platform/efi/efi_32.c b/arch/x86/platform/efi/efi_32.c
new file mode 100644
index 00000000000..5cab48ee61a
--- /dev/null
+++ b/arch/x86/platform/efi/efi_32.c
@@ -0,0 +1,112 @@
+/*
+ * Extensible Firmware Interface
+ *
+ * Based on Extensible Firmware Interface Specification version 1.0
+ *
+ * Copyright (C) 1999 VA Linux Systems
+ * Copyright (C) 1999 Walt Drummond <drummond@valinux.com>
+ * Copyright (C) 1999-2002 Hewlett-Packard Co.
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ *	Stephane Eranian <eranian@hpl.hp.com>
+ *
+ * All EFI Runtime Services are not implemented yet as EFI only
+ * supports physical mode addressing on SoftSDV. This is to be fixed
+ * in a future version.  --drummond 1999-07-20
+ *
+ * Implemented EFI runtime services and virtual mode calls.  --davidm
+ *
+ * Goutham Rao: <goutham.rao@intel.com>
+ *	Skip non-WB memory and ignore empty memory ranges.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/ioport.h>
+#include <linux/efi.h>
+
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+#include <asm/efi.h>
+
+/*
+ * To make EFI call EFI runtime service in physical addressing mode we need
+ * prelog/epilog before/after the invocation to disable interrupt, to
+ * claim EFI runtime service handler exclusively and to duplicate a memory in
+ * low memory space say 0 - 3G.
+ */
+
+static unsigned long efi_rt_eflags;
+static pgd_t efi_bak_pg_dir_pointer[2];
+
+void efi_call_phys_prelog(void)
+{
+	unsigned long cr4;
+	unsigned long temp;
+	struct desc_ptr gdt_descr;
+
+	local_irq_save(efi_rt_eflags);
+
+	/*
+	 * If I don't have PAE, I should just duplicate two entries in page
+	 * directory. If I have PAE, I just need to duplicate one entry in
+	 * page directory.
+	 */
+	cr4 = read_cr4_safe();
+
+	if (cr4 & X86_CR4_PAE) {
+		efi_bak_pg_dir_pointer[0].pgd =
+		    swapper_pg_dir[pgd_index(0)].pgd;
+		swapper_pg_dir[0].pgd =
+		    swapper_pg_dir[pgd_index(PAGE_OFFSET)].pgd;
+	} else {
+		efi_bak_pg_dir_pointer[0].pgd =
+		    swapper_pg_dir[pgd_index(0)].pgd;
+		efi_bak_pg_dir_pointer[1].pgd =
+		    swapper_pg_dir[pgd_index(0x400000)].pgd;
+		swapper_pg_dir[pgd_index(0)].pgd =
+		    swapper_pg_dir[pgd_index(PAGE_OFFSET)].pgd;
+		temp = PAGE_OFFSET + 0x400000;
+		swapper_pg_dir[pgd_index(0x400000)].pgd =
+		    swapper_pg_dir[pgd_index(temp)].pgd;
+	}
+
+	/*
+	 * After the lock is released, the original page table is restored.
+	 */
+	__flush_tlb_all();
+
+	gdt_descr.address = __pa(get_cpu_gdt_table(0));
+	gdt_descr.size = GDT_SIZE - 1;
+	load_gdt(&gdt_descr);
+}
+
+void efi_call_phys_epilog(void)
+{
+	unsigned long cr4;
+	struct desc_ptr gdt_descr;
+
+	gdt_descr.address = (unsigned long)get_cpu_gdt_table(0);
+	gdt_descr.size = GDT_SIZE - 1;
+	load_gdt(&gdt_descr);
+
+	cr4 = read_cr4_safe();
+
+	if (cr4 & X86_CR4_PAE) {
+		swapper_pg_dir[pgd_index(0)].pgd =
+		    efi_bak_pg_dir_pointer[0].pgd;
+	} else {
+		swapper_pg_dir[pgd_index(0)].pgd =
+		    efi_bak_pg_dir_pointer[0].pgd;
+		swapper_pg_dir[pgd_index(0x400000)].pgd =
+		    efi_bak_pg_dir_pointer[1].pgd;
+	}
+
+	/*
+	 * After the lock is released, the original page table is restored.
+	 */
+	__flush_tlb_all();
+
+	local_irq_restore(efi_rt_eflags);
+}
diff --git a/arch/x86/platform/efi/efi_64.c b/arch/x86/platform/efi/efi_64.c
new file mode 100644
index 00000000000..ac0621a7ac3
--- /dev/null
+++ b/arch/x86/platform/efi/efi_64.c
@@ -0,0 +1,114 @@
+/*
+ * x86_64 specific EFI support functions
+ * Based on Extensible Firmware Interface Specification version 1.0
+ *
+ * Copyright (C) 2005-2008 Intel Co.
+ *	Fenghua Yu <fenghua.yu@intel.com>
+ *	Bibo Mao <bibo.mao@intel.com>
+ *	Chandramouli Narayanan <mouli@linux.intel.com>
+ *	Huang Ying <ying.huang@intel.com>
+ *
+ * Code to convert EFI to E820 map has been implemented in elilo bootloader
+ * based on a EFI patch by Edgar Hucek. Based on the E820 map, the page table
+ * is setup appropriately for EFI runtime code.
+ * - mouli 06/14/2007.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/mm.h>
+#include <linux/types.h>
+#include <linux/spinlock.h>
+#include <linux/bootmem.h>
+#include <linux/ioport.h>
+#include <linux/module.h>
+#include <linux/efi.h>
+#include <linux/uaccess.h>
+#include <linux/io.h>
+#include <linux/reboot.h>
+
+#include <asm/setup.h>
+#include <asm/page.h>
+#include <asm/e820.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+#include <asm/proto.h>
+#include <asm/efi.h>
+#include <asm/cacheflush.h>
+#include <asm/fixmap.h>
+
+static pgd_t save_pgd __initdata;
+static unsigned long efi_flags __initdata;
+
+static void __init early_mapping_set_exec(unsigned long start,
+					  unsigned long end,
+					  int executable)
+{
+	unsigned long num_pages;
+
+	start &= PMD_MASK;
+	end = (end + PMD_SIZE - 1) & PMD_MASK;
+	num_pages = (end - start) >> PAGE_SHIFT;
+	if (executable)
+		set_memory_x((unsigned long)__va(start), num_pages);
+	else
+		set_memory_nx((unsigned long)__va(start), num_pages);
+}
+
+static void __init early_runtime_code_mapping_set_exec(int executable)
+{
+	efi_memory_desc_t *md;
+	void *p;
+
+	if (!(__supported_pte_mask & _PAGE_NX))
+		return;
+
+	/* Make EFI runtime service code area executable */
+	for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
+		md = p;
+		if (md->type == EFI_RUNTIME_SERVICES_CODE) {
+			unsigned long end;
+			end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT);
+			early_mapping_set_exec(md->phys_addr, end, executable);
+		}
+	}
+}
+
+void __init efi_call_phys_prelog(void)
+{
+	unsigned long vaddress;
+
+	early_runtime_code_mapping_set_exec(1);
+	local_irq_save(efi_flags);
+	vaddress = (unsigned long)__va(0x0UL);
+	save_pgd = *pgd_offset_k(0x0UL);
+	set_pgd(pgd_offset_k(0x0UL), *pgd_offset_k(vaddress));
+	__flush_tlb_all();
+}
+
+void __init efi_call_phys_epilog(void)
+{
+	/*
+	 * After the lock is released, the original page table is restored.
+	 */
+	set_pgd(pgd_offset_k(0x0UL), save_pgd);
+	__flush_tlb_all();
+	local_irq_restore(efi_flags);
+	early_runtime_code_mapping_set_exec(0);
+}
+
+void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size,
+				 u32 type)
+{
+	unsigned long last_map_pfn;
+
+	if (type == EFI_MEMORY_MAPPED_IO)
+		return ioremap(phys_addr, size);
+
+	last_map_pfn = init_memory_mapping(phys_addr, phys_addr + size);
+	if ((last_map_pfn << PAGE_SHIFT) < phys_addr + size)
+		return NULL;
+
+	return (void __iomem *)__va(phys_addr);
+}
diff --git a/arch/x86/platform/efi/efi_stub_32.S b/arch/x86/platform/efi/efi_stub_32.S
new file mode 100644
index 00000000000..fbe66e626c0
--- /dev/null
+++ b/arch/x86/platform/efi/efi_stub_32.S
@@ -0,0 +1,123 @@
+/*
+ * EFI call stub for IA32.
+ *
+ * This stub allows us to make EFI calls in physical mode with interrupts
+ * turned off.
+ */
+
+#include <linux/linkage.h>
+#include <asm/page_types.h>
+
+/*
+ * efi_call_phys(void *, ...) is a function with variable parameters.
+ * All the callers of this function assure that all the parameters are 4-bytes.
+ */
+
+/*
+ * In gcc calling convention, EBX, ESP, EBP, ESI and EDI are all callee save.
+ * So we'd better save all of them at the beginning of this function and restore
+ * at the end no matter how many we use, because we can not assure EFI runtime
+ * service functions will comply with gcc calling convention, too.
+ */
+
+.text
+ENTRY(efi_call_phys)
+	/*
+	 * 0. The function can only be called in Linux kernel. So CS has been
+	 * set to 0x0010, DS and SS have been set to 0x0018. In EFI, I found
+	 * the values of these registers are the same. And, the corresponding
+	 * GDT entries are identical. So I will do nothing about segment reg
+	 * and GDT, but change GDT base register in prelog and epilog.
+	 */
+
+	/*
+	 * 1. Now I am running with EIP = <physical address> + PAGE_OFFSET.
+	 * But to make it smoothly switch from virtual mode to flat mode.
+	 * The mapping of lower virtual memory has been created in prelog and
+	 * epilog.
+	 */
+	movl	$1f, %edx
+	subl	$__PAGE_OFFSET, %edx
+	jmp	*%edx
+1:
+
+	/*
+	 * 2. Now on the top of stack is the return
+	 * address in the caller of efi_call_phys(), then parameter 1,
+	 * parameter 2, ..., param n. To make things easy, we save the return
+	 * address of efi_call_phys in a global variable.
+	 */
+	popl	%edx
+	movl	%edx, saved_return_addr
+	/* get the function pointer into ECX*/
+	popl	%ecx
+	movl	%ecx, efi_rt_function_ptr
+	movl	$2f, %edx
+	subl	$__PAGE_OFFSET, %edx
+	pushl	%edx
+
+	/*
+	 * 3. Clear PG bit in %CR0.
+	 */
+	movl	%cr0, %edx
+	andl	$0x7fffffff, %edx
+	movl	%edx, %cr0
+	jmp	1f
+1:
+
+	/*
+	 * 4. Adjust stack pointer.
+	 */
+	subl	$__PAGE_OFFSET, %esp
+
+	/*
+	 * 5. Call the physical function.
+	 */
+	jmp	*%ecx
+
+2:
+	/*
+	 * 6. After EFI runtime service returns, control will return to
+	 * following instruction. We'd better readjust stack pointer first.
+	 */
+	addl	$__PAGE_OFFSET, %esp
+
+	/*
+	 * 7. Restore PG bit
+	 */
+	movl	%cr0, %edx
+	orl	$0x80000000, %edx
+	movl	%edx, %cr0
+	jmp	1f
+1:
+	/*
+	 * 8. Now restore the virtual mode from flat mode by
+	 * adding EIP with PAGE_OFFSET.
+	 */
+	movl	$1f, %edx
+	jmp	*%edx
+1:
+
+	/*
+	 * 9. Balance the stack. And because EAX contain the return value,
+	 * we'd better not clobber it.
+	 */
+	leal	efi_rt_function_ptr, %edx
+	movl	(%edx), %ecx
+	pushl	%ecx
+
+	/*
+	 * 10. Push the saved return address onto the stack and return.
+	 */
+	leal	saved_return_addr, %edx
+	movl	(%edx), %ecx
+	pushl	%ecx
+	ret
+ENDPROC(efi_call_phys)
+.previous
+
+.data
+saved_return_addr:
+	.long 0
+efi_rt_function_ptr:
+	.long 0
diff --git a/arch/x86/platform/efi/efi_stub_64.S b/arch/x86/platform/efi/efi_stub_64.S
new file mode 100644
index 00000000000..4c07ccab814
--- /dev/null
+++ b/arch/x86/platform/efi/efi_stub_64.S
@@ -0,0 +1,116 @@
+/*
+ * Function calling ABI conversion from Linux to EFI for x86_64
+ *
+ * Copyright (C) 2007 Intel Corp
+ *	Bibo Mao <bibo.mao@intel.com>
+ *	Huang Ying <ying.huang@intel.com>
+ */
+
+#include <linux/linkage.h>
+
+#define SAVE_XMM			\
+	mov %rsp, %rax;			\
+	subq $0x70, %rsp;		\
+	and $~0xf, %rsp;		\
+	mov %rax, (%rsp);		\
+	mov %cr0, %rax;			\
+	clts;				\
+	mov %rax, 0x8(%rsp);		\
+	movaps %xmm0, 0x60(%rsp);	\
+	movaps %xmm1, 0x50(%rsp);	\
+	movaps %xmm2, 0x40(%rsp);	\
+	movaps %xmm3, 0x30(%rsp);	\
+	movaps %xmm4, 0x20(%rsp);	\
+	movaps %xmm5, 0x10(%rsp)
+
+#define RESTORE_XMM			\
+	movaps 0x60(%rsp), %xmm0;	\
+	movaps 0x50(%rsp), %xmm1;	\
+	movaps 0x40(%rsp), %xmm2;	\
+	movaps 0x30(%rsp), %xmm3;	\
+	movaps 0x20(%rsp), %xmm4;	\
+	movaps 0x10(%rsp), %xmm5;	\
+	mov 0x8(%rsp), %rsi;		\
+	mov %rsi, %cr0;			\
+	mov (%rsp), %rsp
+
+ENTRY(efi_call0)
+	SAVE_XMM
+	subq $32, %rsp
+	call *%rdi
+	addq $32, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call0)
+
+ENTRY(efi_call1)
+	SAVE_XMM
+	subq $32, %rsp
+	mov  %rsi, %rcx
+	call *%rdi
+	addq $32, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call1)
+
+ENTRY(efi_call2)
+	SAVE_XMM
+	subq $32, %rsp
+	mov  %rsi, %rcx
+	call *%rdi
+	addq $32, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call2)
+
+ENTRY(efi_call3)
+	SAVE_XMM
+	subq $32, %rsp
+	mov  %rcx, %r8
+	mov  %rsi, %rcx
+	call *%rdi
+	addq $32, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call3)
+
+ENTRY(efi_call4)
+	SAVE_XMM
+	subq $32, %rsp
+	mov %r8, %r9
+	mov %rcx, %r8
+	mov %rsi, %rcx
+	call *%rdi
+	addq $32, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call4)
+
+ENTRY(efi_call5)
+	SAVE_XMM
+	subq $48, %rsp
+	mov %r9, 32(%rsp)
+	mov %r8, %r9
+	mov %rcx, %r8
+	mov %rsi, %rcx
+	call *%rdi
+	addq $48, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call5)
+
+ENTRY(efi_call6)
+	SAVE_XMM
+	mov (%rsp), %rax
+	mov 8(%rax), %rax
+	subq $48, %rsp
+	mov %r9, 32(%rsp)
+	mov %rax, 40(%rsp)
+	mov %r8, %r9
+	mov %rcx, %r8
+	mov %rsi, %rcx
+	call *%rdi
+	addq $48, %rsp
+	RESTORE_XMM
+	ret
+ENDPROC(efi_call6)
diff --git a/arch/x86/platform/mrst/Makefile b/arch/x86/platform/mrst/Makefile
new file mode 100644
index 00000000000..efbbc552fa9
--- /dev/null
+++ b/arch/x86/platform/mrst/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_X86_MRST)		+= mrst.o
diff --git a/arch/x86/platform/mrst/mrst.c b/arch/x86/platform/mrst/mrst.c
new file mode 100644
index 00000000000..79ae68154e8
--- /dev/null
+++ b/arch/x86/platform/mrst/mrst.c
@@ -0,0 +1,311 @@
+/*
+ * mrst.c: Intel Moorestown platform specific setup code
+ *
+ * (C) Copyright 2008 Intel Corporation
+ * Author: Jacob Pan (jacob.jun.pan@intel.com)
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sfi.h>
+#include <linux/irq.h>
+#include <linux/module.h>
+
+#include <asm/setup.h>
+#include <asm/mpspec_def.h>
+#include <asm/hw_irq.h>
+#include <asm/apic.h>
+#include <asm/io_apic.h>
+#include <asm/mrst.h>
+#include <asm/io.h>
+#include <asm/i8259.h>
+#include <asm/apb_timer.h>
+
+/*
+ * the clockevent devices on Moorestown/Medfield can be APBT or LAPIC clock,
+ * cmdline option x86_mrst_timer can be used to override the configuration
+ * to prefer one or the other.
+ * at runtime, there are basically three timer configurations:
+ * 1. per cpu apbt clock only
+ * 2. per cpu always-on lapic clocks only, this is Penwell/Medfield only
+ * 3. per cpu lapic clock (C3STOP) and one apbt clock, with broadcast.
+ *
+ * by default (without cmdline option), platform code first detects cpu type
+ * to see if we are on lincroft or penwell, then set up both lapic or apbt
+ * clocks accordingly.
+ * i.e. by default, medfield uses configuration #2, moorestown uses #1.
+ * config #3 is supported but not recommended on medfield.
+ *
+ * rating and feature summary:
+ * lapic (with C3STOP) --------- 100
+ * apbt (always-on) ------------ 110
+ * lapic (always-on,ARAT) ------ 150
+ */
+
+__cpuinitdata enum mrst_timer_options mrst_timer_options;
+
+static u32 sfi_mtimer_usage[SFI_MTMR_MAX_NUM];
+static struct sfi_timer_table_entry sfi_mtimer_array[SFI_MTMR_MAX_NUM];
+enum mrst_cpu_type __mrst_cpu_chip;
+EXPORT_SYMBOL_GPL(__mrst_cpu_chip);
+
+int sfi_mtimer_num;
+
+struct sfi_rtc_table_entry sfi_mrtc_array[SFI_MRTC_MAX];
+EXPORT_SYMBOL_GPL(sfi_mrtc_array);
+int sfi_mrtc_num;
+
+static inline void assign_to_mp_irq(struct mpc_intsrc *m,
+				    struct mpc_intsrc *mp_irq)
+{
+	memcpy(mp_irq, m, sizeof(struct mpc_intsrc));
+}
+
+static inline int mp_irq_cmp(struct mpc_intsrc *mp_irq,
+				struct mpc_intsrc *m)
+{
+	return memcmp(mp_irq, m, sizeof(struct mpc_intsrc));
+}
+
+static void save_mp_irq(struct mpc_intsrc *m)
+{
+	int i;
+
+	for (i = 0; i < mp_irq_entries; i++) {
+		if (!mp_irq_cmp(&mp_irqs[i], m))
+			return;
+	}
+
+	assign_to_mp_irq(m, &mp_irqs[mp_irq_entries]);
+	if (++mp_irq_entries == MAX_IRQ_SOURCES)
+		panic("Max # of irq sources exceeded!!\n");
+}
+
+/* parse all the mtimer info to a static mtimer array */
+static int __init sfi_parse_mtmr(struct sfi_table_header *table)
+{
+	struct sfi_table_simple *sb;
+	struct sfi_timer_table_entry *pentry;
+	struct mpc_intsrc mp_irq;
+	int totallen;
+
+	sb = (struct sfi_table_simple *)table;
+	if (!sfi_mtimer_num) {
+		sfi_mtimer_num = SFI_GET_NUM_ENTRIES(sb,
+					struct sfi_timer_table_entry);
+		pentry = (struct sfi_timer_table_entry *) sb->pentry;
+		totallen = sfi_mtimer_num * sizeof(*pentry);
+		memcpy(sfi_mtimer_array, pentry, totallen);
+	}
+
+	printk(KERN_INFO "SFI: MTIMER info (num = %d):\n", sfi_mtimer_num);
+	pentry = sfi_mtimer_array;
+	for (totallen = 0; totallen < sfi_mtimer_num; totallen++, pentry++) {
+		printk(KERN_INFO "timer[%d]: paddr = 0x%08x, freq = %dHz,"
+			" irq = %d\n", totallen, (u32)pentry->phys_addr,
+			pentry->freq_hz, pentry->irq);
+			if (!pentry->irq)
+				continue;
+			mp_irq.type = MP_IOAPIC;
+			mp_irq.irqtype = mp_INT;
+/* triggering mode edge bit 2-3, active high polarity bit 0-1 */
+			mp_irq.irqflag = 5;
+			mp_irq.srcbus = 0;
+			mp_irq.srcbusirq = pentry->irq;	/* IRQ */
+			mp_irq.dstapic = MP_APIC_ALL;
+			mp_irq.dstirq = pentry->irq;
+			save_mp_irq(&mp_irq);
+	}
+
+	return 0;
+}
+
+struct sfi_timer_table_entry *sfi_get_mtmr(int hint)
+{
+	int i;
+	if (hint < sfi_mtimer_num) {
+		if (!sfi_mtimer_usage[hint]) {
+			pr_debug("hint taken for timer %d irq %d\n",\
+				hint, sfi_mtimer_array[hint].irq);
+			sfi_mtimer_usage[hint] = 1;
+			return &sfi_mtimer_array[hint];
+		}
+	}
+	/* take the first timer available */
+	for (i = 0; i < sfi_mtimer_num;) {
+		if (!sfi_mtimer_usage[i]) {
+			sfi_mtimer_usage[i] = 1;
+			return &sfi_mtimer_array[i];
+		}
+		i++;
+	}
+	return NULL;
+}
+
+void sfi_free_mtmr(struct sfi_timer_table_entry *mtmr)
+{
+	int i;
+	for (i = 0; i < sfi_mtimer_num;) {
+		if (mtmr->irq == sfi_mtimer_array[i].irq) {
+			sfi_mtimer_usage[i] = 0;
+			return;
+		}
+		i++;
+	}
+}
+
+/* parse all the mrtc info to a global mrtc array */
+int __init sfi_parse_mrtc(struct sfi_table_header *table)
+{
+	struct sfi_table_simple *sb;
+	struct sfi_rtc_table_entry *pentry;
+	struct mpc_intsrc mp_irq;
+
+	int totallen;
+
+	sb = (struct sfi_table_simple *)table;
+	if (!sfi_mrtc_num) {
+		sfi_mrtc_num = SFI_GET_NUM_ENTRIES(sb,
+						struct sfi_rtc_table_entry);
+		pentry = (struct sfi_rtc_table_entry *)sb->pentry;
+		totallen = sfi_mrtc_num * sizeof(*pentry);
+		memcpy(sfi_mrtc_array, pentry, totallen);
+	}
+
+	printk(KERN_INFO "SFI: RTC info (num = %d):\n", sfi_mrtc_num);
+	pentry = sfi_mrtc_array;
+	for (totallen = 0; totallen < sfi_mrtc_num; totallen++, pentry++) {
+		printk(KERN_INFO "RTC[%d]: paddr = 0x%08x, irq = %d\n",
+			totallen, (u32)pentry->phys_addr, pentry->irq);
+		mp_irq.type = MP_IOAPIC;
+		mp_irq.irqtype = mp_INT;
+		mp_irq.irqflag = 0;
+		mp_irq.srcbus = 0;
+		mp_irq.srcbusirq = pentry->irq;	/* IRQ */
+		mp_irq.dstapic = MP_APIC_ALL;
+		mp_irq.dstirq = pentry->irq;
+		save_mp_irq(&mp_irq);
+	}
+	return 0;
+}
+
+static unsigned long __init mrst_calibrate_tsc(void)
+{
+	unsigned long flags, fast_calibrate;
+
+	local_irq_save(flags);
+	fast_calibrate = apbt_quick_calibrate();
+	local_irq_restore(flags);
+
+	if (fast_calibrate)
+		return fast_calibrate;
+
+	return 0;
+}
+
+void __init mrst_time_init(void)
+{
+	switch (mrst_timer_options) {
+	case MRST_TIMER_APBT_ONLY:
+		break;
+	case MRST_TIMER_LAPIC_APBT:
+		x86_init.timers.setup_percpu_clockev = setup_boot_APIC_clock;
+		x86_cpuinit.setup_percpu_clockev = setup_secondary_APIC_clock;
+		break;
+	default:
+		if (!boot_cpu_has(X86_FEATURE_ARAT))
+			break;
+		x86_init.timers.setup_percpu_clockev = setup_boot_APIC_clock;
+		x86_cpuinit.setup_percpu_clockev = setup_secondary_APIC_clock;
+		return;
+	}
+	/* we need at least one APB timer */
+	sfi_table_parse(SFI_SIG_MTMR, NULL, NULL, sfi_parse_mtmr);
+	pre_init_apic_IRQ0();
+	apbt_time_init();
+}
+
+void __init mrst_rtc_init(void)
+{
+	sfi_table_parse(SFI_SIG_MRTC, NULL, NULL, sfi_parse_mrtc);
+}
+
+void __cpuinit mrst_arch_setup(void)
+{
+	if (boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 0x27)
+		__mrst_cpu_chip = MRST_CPU_CHIP_PENWELL;
+	else if (boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 0x26)
+		__mrst_cpu_chip = MRST_CPU_CHIP_LINCROFT;
+	else {
+		pr_err("Unknown Moorestown CPU (%d:%d), default to Lincroft\n",
+			boot_cpu_data.x86, boot_cpu_data.x86_model);
+		__mrst_cpu_chip = MRST_CPU_CHIP_LINCROFT;
+	}
+	pr_debug("Moorestown CPU %s identified\n",
+		(__mrst_cpu_chip == MRST_CPU_CHIP_LINCROFT) ?
+		"Lincroft" : "Penwell");
+}
+
+/* MID systems don't have i8042 controller */
+static int mrst_i8042_detect(void)
+{
+	return 0;
+}
+
+/*
+ * Moorestown specific x86_init function overrides and early setup
+ * calls.
+ */
+void __init x86_mrst_early_setup(void)
+{
+	x86_init.resources.probe_roms = x86_init_noop;
+	x86_init.resources.reserve_resources = x86_init_noop;
+
+	x86_init.timers.timer_init = mrst_time_init;
+	x86_init.timers.setup_percpu_clockev = x86_init_noop;
+
+	x86_init.irqs.pre_vector_init = x86_init_noop;
+
+	x86_init.oem.arch_setup = mrst_arch_setup;
+
+	x86_cpuinit.setup_percpu_clockev = apbt_setup_secondary_clock;
+
+	x86_platform.calibrate_tsc = mrst_calibrate_tsc;
+	x86_platform.i8042_detect = mrst_i8042_detect;
+	x86_init.pci.init = pci_mrst_init;
+	x86_init.pci.fixup_irqs = x86_init_noop;
+
+	legacy_pic = &null_legacy_pic;
+
+	/* Avoid searching for BIOS MP tables */
+	x86_init.mpparse.find_smp_config = x86_init_noop;
+	x86_init.mpparse.get_smp_config = x86_init_uint_noop;
+
+}
+
+/*
+ * if user does not want to use per CPU apb timer, just give it a lower rating
+ * than local apic timer and skip the late per cpu timer init.
+ */
+static inline int __init setup_x86_mrst_timer(char *arg)
+{
+	if (!arg)
+		return -EINVAL;
+
+	if (strcmp("apbt_only", arg) == 0)
+		mrst_timer_options = MRST_TIMER_APBT_ONLY;
+	else if (strcmp("lapic_and_apbt", arg) == 0)
+		mrst_timer_options = MRST_TIMER_LAPIC_APBT;
+	else {
+		pr_warning("X86 MRST timer option %s not recognised"
+			   " use x86_mrst_timer=apbt_only or lapic_and_apbt\n",
+			   arg);
+		return -EINVAL;
+	}
+	return 0;
+}
+__setup("x86_mrst_timer=", setup_x86_mrst_timer);
diff --git a/arch/x86/platform/olpc/Makefile b/arch/x86/platform/olpc/Makefile
new file mode 100644
index 00000000000..c31b8fcb5a8
--- /dev/null
+++ b/arch/x86/platform/olpc/Makefile
@@ -0,0 +1,3 @@
+obj-$(CONFIG_OLPC)		+= olpc.o
+obj-$(CONFIG_OLPC_XO1)		+= olpc-xo1.o
+obj-$(CONFIG_OLPC_OPENFIRMWARE)	+= olpc_ofw.o
diff --git a/arch/x86/platform/olpc/olpc-xo1.c b/arch/x86/platform/olpc/olpc-xo1.c
new file mode 100644
index 00000000000..f5442c03abc
--- /dev/null
+++ b/arch/x86/platform/olpc/olpc-xo1.c
@@ -0,0 +1,140 @@
+/*
+ * Support for features of the OLPC XO-1 laptop
+ *
+ * Copyright (C) 2010 One Laptop per Child
+ * Copyright (C) 2006 Red Hat, Inc.
+ * Copyright (C) 2006 Advanced Micro Devices, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/platform_device.h>
+#include <linux/pm.h>
+
+#include <asm/io.h>
+#include <asm/olpc.h>
+
+#define DRV_NAME "olpc-xo1"
+
+#define PMS_BAR		4
+#define ACPI_BAR	5
+
+/* PMC registers (PMS block) */
+#define PM_SCLK		0x10
+#define PM_IN_SLPCTL	0x20
+#define PM_WKXD		0x34
+#define PM_WKD		0x30
+#define PM_SSC		0x54
+
+/* PM registers (ACPI block) */
+#define PM1_CNT		0x08
+#define PM_GPE0_STS	0x18
+
+static unsigned long acpi_base;
+static unsigned long pms_base;
+
+static void xo1_power_off(void)
+{
+	printk(KERN_INFO "OLPC XO-1 power off sequence...\n");
+
+	/* Enable all of these controls with 0 delay */
+	outl(0x40000000, pms_base + PM_SCLK);
+	outl(0x40000000, pms_base + PM_IN_SLPCTL);
+	outl(0x40000000, pms_base + PM_WKXD);
+	outl(0x40000000, pms_base + PM_WKD);
+
+	/* Clear status bits (possibly unnecessary) */
+	outl(0x0002ffff, pms_base  + PM_SSC);
+	outl(0xffffffff, acpi_base + PM_GPE0_STS);
+
+	/* Write SLP_EN bit to start the machinery */
+	outl(0x00002000, acpi_base + PM1_CNT);
+}
+
+/* Read the base addresses from the PCI BAR info */
+static int __devinit setup_bases(struct pci_dev *pdev)
+{
+	int r;
+
+	r = pci_enable_device_io(pdev);
+	if (r) {
+		dev_err(&pdev->dev, "can't enable device IO\n");
+		return r;
+	}
+
+	r = pci_request_region(pdev, ACPI_BAR, DRV_NAME);
+	if (r) {
+		dev_err(&pdev->dev, "can't alloc PCI BAR #%d\n", ACPI_BAR);
+		return r;
+	}
+
+	r = pci_request_region(pdev, PMS_BAR, DRV_NAME);
+	if (r) {
+		dev_err(&pdev->dev, "can't alloc PCI BAR #%d\n", PMS_BAR);
+		pci_release_region(pdev, ACPI_BAR);
+		return r;
+	}
+
+	acpi_base = pci_resource_start(pdev, ACPI_BAR);
+	pms_base = pci_resource_start(pdev, PMS_BAR);
+
+	return 0;
+}
+
+static int __devinit olpc_xo1_probe(struct platform_device *pdev)
+{
+	struct pci_dev *pcidev;
+	int r;
+
+	pcidev = pci_get_device(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_CS5536_ISA,
+				NULL);
+	if (!pdev)
+		return -ENODEV;
+
+	r = setup_bases(pcidev);
+	if (r)
+		return r;
+
+	pm_power_off = xo1_power_off;
+
+	printk(KERN_INFO "OLPC XO-1 support registered\n");
+	return 0;
+}
+
+static int __devexit olpc_xo1_remove(struct platform_device *pdev)
+{
+	pm_power_off = NULL;
+	return 0;
+}
+
+static struct platform_driver olpc_xo1_driver = {
+	.driver = {
+		.name = DRV_NAME,
+		.owner = THIS_MODULE,
+	},
+	.probe = olpc_xo1_probe,
+	.remove = __devexit_p(olpc_xo1_remove),
+};
+
+static int __init olpc_xo1_init(void)
+{
+	return platform_driver_register(&olpc_xo1_driver);
+}
+
+static void __exit olpc_xo1_exit(void)
+{
+	platform_driver_unregister(&olpc_xo1_driver);
+}
+
+MODULE_AUTHOR("Daniel Drake <dsd@laptop.org>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:olpc-xo1");
+
+module_init(olpc_xo1_init);
+module_exit(olpc_xo1_exit);
diff --git a/arch/x86/platform/olpc/olpc.c b/arch/x86/platform/olpc/olpc.c
new file mode 100644
index 00000000000..edaf3fe8dc5
--- /dev/null
+++ b/arch/x86/platform/olpc/olpc.c
@@ -0,0 +1,281 @@
+/*
+ * Support for the OLPC DCON and OLPC EC access
+ *
+ * Copyright © 2006  Advanced Micro Devices, Inc.
+ * Copyright © 2007-2008  Andres Salomon <dilinger@debian.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/delay.h>
+#include <linux/spinlock.h>
+#include <linux/io.h>
+#include <linux/string.h>
+#include <linux/platform_device.h>
+
+#include <asm/geode.h>
+#include <asm/setup.h>
+#include <asm/olpc.h>
+#include <asm/olpc_ofw.h>
+
+struct olpc_platform_t olpc_platform_info;
+EXPORT_SYMBOL_GPL(olpc_platform_info);
+
+static DEFINE_SPINLOCK(ec_lock);
+
+/* what the timeout *should* be (in ms) */
+#define EC_BASE_TIMEOUT 20
+
+/* the timeout that bugs in the EC might force us to actually use */
+static int ec_timeout = EC_BASE_TIMEOUT;
+
+static int __init olpc_ec_timeout_set(char *str)
+{
+	if (get_option(&str, &ec_timeout) != 1) {
+		ec_timeout = EC_BASE_TIMEOUT;
+		printk(KERN_ERR "olpc-ec:  invalid argument to "
+				"'olpc_ec_timeout=', ignoring!\n");
+	}
+	printk(KERN_DEBUG "olpc-ec:  using %d ms delay for EC commands.\n",
+			ec_timeout);
+	return 1;
+}
+__setup("olpc_ec_timeout=", olpc_ec_timeout_set);
+
+/*
+ * These {i,o}bf_status functions return whether the buffers are full or not.
+ */
+
+static inline unsigned int ibf_status(unsigned int port)
+{
+	return !!(inb(port) & 0x02);
+}
+
+static inline unsigned int obf_status(unsigned int port)
+{
+	return inb(port) & 0x01;
+}
+
+#define wait_on_ibf(p, d) __wait_on_ibf(__LINE__, (p), (d))
+static int __wait_on_ibf(unsigned int line, unsigned int port, int desired)
+{
+	unsigned int timeo;
+	int state = ibf_status(port);
+
+	for (timeo = ec_timeout; state != desired && timeo; timeo--) {
+		mdelay(1);
+		state = ibf_status(port);
+	}
+
+	if ((state == desired) && (ec_timeout > EC_BASE_TIMEOUT) &&
+			timeo < (ec_timeout - EC_BASE_TIMEOUT)) {
+		printk(KERN_WARNING "olpc-ec:  %d: waited %u ms for IBF!\n",
+				line, ec_timeout - timeo);
+	}
+
+	return !(state == desired);
+}
+
+#define wait_on_obf(p, d) __wait_on_obf(__LINE__, (p), (d))
+static int __wait_on_obf(unsigned int line, unsigned int port, int desired)
+{
+	unsigned int timeo;
+	int state = obf_status(port);
+
+	for (timeo = ec_timeout; state != desired && timeo; timeo--) {
+		mdelay(1);
+		state = obf_status(port);
+	}
+
+	if ((state == desired) && (ec_timeout > EC_BASE_TIMEOUT) &&
+			timeo < (ec_timeout - EC_BASE_TIMEOUT)) {
+		printk(KERN_WARNING "olpc-ec:  %d: waited %u ms for OBF!\n",
+				line, ec_timeout - timeo);
+	}
+
+	return !(state == desired);
+}
+
+/*
+ * This allows the kernel to run Embedded Controller commands.  The EC is
+ * documented at <http://wiki.laptop.org/go/Embedded_controller>, and the
+ * available EC commands are here:
+ * <http://wiki.laptop.org/go/Ec_specification>.  Unfortunately, while
+ * OpenFirmware's source is available, the EC's is not.
+ */
+int olpc_ec_cmd(unsigned char cmd, unsigned char *inbuf, size_t inlen,
+		unsigned char *outbuf,  size_t outlen)
+{
+	unsigned long flags;
+	int ret = -EIO;
+	int i;
+	int restarts = 0;
+
+	spin_lock_irqsave(&ec_lock, flags);
+
+	/* Clear OBF */
+	for (i = 0; i < 10 && (obf_status(0x6c) == 1); i++)
+		inb(0x68);
+	if (i == 10) {
+		printk(KERN_ERR "olpc-ec:  timeout while attempting to "
+				"clear OBF flag!\n");
+		goto err;
+	}
+
+	if (wait_on_ibf(0x6c, 0)) {
+		printk(KERN_ERR "olpc-ec:  timeout waiting for EC to "
+				"quiesce!\n");
+		goto err;
+	}
+
+restart:
+	/*
+	 * Note that if we time out during any IBF checks, that's a failure;
+	 * we have to return.  There's no way for the kernel to clear that.
+	 *
+	 * If we time out during an OBF check, we can restart the command;
+	 * reissuing it will clear the OBF flag, and we should be alright.
+	 * The OBF flag will sometimes misbehave due to what we believe
+	 * is a hardware quirk..
+	 */
+	pr_devel("olpc-ec:  running cmd 0x%x\n", cmd);
+	outb(cmd, 0x6c);
+
+	if (wait_on_ibf(0x6c, 0)) {
+		printk(KERN_ERR "olpc-ec:  timeout waiting for EC to read "
+				"command!\n");
+		goto err;
+	}
+
+	if (inbuf && inlen) {
+		/* write data to EC */
+		for (i = 0; i < inlen; i++) {
+			if (wait_on_ibf(0x6c, 0)) {
+				printk(KERN_ERR "olpc-ec:  timeout waiting for"
+						" EC accept data!\n");
+				goto err;
+			}
+			pr_devel("olpc-ec:  sending cmd arg 0x%x\n", inbuf[i]);
+			outb(inbuf[i], 0x68);
+		}
+	}
+	if (outbuf && outlen) {
+		/* read data from EC */
+		for (i = 0; i < outlen; i++) {
+			if (wait_on_obf(0x6c, 1)) {
+				printk(KERN_ERR "olpc-ec:  timeout waiting for"
+						" EC to provide data!\n");
+				if (restarts++ < 10)
+					goto restart;
+				goto err;
+			}
+			outbuf[i] = inb(0x68);
+			pr_devel("olpc-ec:  received 0x%x\n", outbuf[i]);
+		}
+	}
+
+	ret = 0;
+err:
+	spin_unlock_irqrestore(&ec_lock, flags);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(olpc_ec_cmd);
+
+static bool __init check_ofw_architecture(void)
+{
+	size_t propsize;
+	char olpc_arch[5];
+	const void *args[] = { NULL, "architecture", olpc_arch, (void *)5 };
+	void *res[] = { &propsize };
+
+	if (olpc_ofw("getprop", args, res)) {
+		printk(KERN_ERR "ofw: getprop call failed!\n");
+		return false;
+	}
+	return propsize == 5 && strncmp("OLPC", olpc_arch, 5) == 0;
+}
+
+static u32 __init get_board_revision(void)
+{
+	size_t propsize;
+	__be32 rev;
+	const void *args[] = { NULL, "board-revision-int", &rev, (void *)4 };
+	void *res[] = { &propsize };
+
+	if (olpc_ofw("getprop", args, res) || propsize != 4) {
+		printk(KERN_ERR "ofw: getprop call failed!\n");
+		return cpu_to_be32(0);
+	}
+	return be32_to_cpu(rev);
+}
+
+static bool __init platform_detect(void)
+{
+	if (!check_ofw_architecture())
+		return false;
+	olpc_platform_info.flags |= OLPC_F_PRESENT;
+	olpc_platform_info.boardrev = get_board_revision();
+	return true;
+}
+
+static int __init add_xo1_platform_devices(void)
+{
+	struct platform_device *pdev;
+
+	pdev = platform_device_register_simple("xo1-rfkill", -1, NULL, 0);
+	if (IS_ERR(pdev))
+		return PTR_ERR(pdev);
+
+	pdev = platform_device_register_simple("olpc-xo1", -1, NULL, 0);
+	if (IS_ERR(pdev))
+		return PTR_ERR(pdev);
+
+	return 0;
+}
+
+static int __init olpc_init(void)
+{
+	int r = 0;
+
+	if (!olpc_ofw_present() || !platform_detect())
+		return 0;
+
+	spin_lock_init(&ec_lock);
+
+	/* assume B1 and above models always have a DCON */
+	if (olpc_board_at_least(olpc_board(0xb1)))
+		olpc_platform_info.flags |= OLPC_F_DCON;
+
+	/* get the EC revision */
+	olpc_ec_cmd(EC_FIRMWARE_REV, NULL, 0,
+			(unsigned char *) &olpc_platform_info.ecver, 1);
+
+#ifdef CONFIG_PCI_OLPC
+	/* If the VSA exists let it emulate PCI, if not emulate in kernel.
+	 * XO-1 only. */
+	if (olpc_platform_info.boardrev < olpc_board_pre(0xd0) &&
+			!cs5535_has_vsa2())
+		x86_init.pci.arch_init = pci_olpc_init;
+#endif
+
+	printk(KERN_INFO "OLPC board revision %s%X (EC=%x)\n",
+			((olpc_platform_info.boardrev & 0xf) < 8) ? "pre" : "",
+			olpc_platform_info.boardrev >> 4,
+			olpc_platform_info.ecver);
+
+	if (olpc_platform_info.boardrev < olpc_board_pre(0xd0)) { /* XO-1 */
+		r = add_xo1_platform_devices();
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+postcore_initcall(olpc_init);
diff --git a/arch/x86/platform/olpc/olpc_ofw.c b/arch/x86/platform/olpc/olpc_ofw.c
new file mode 100644
index 00000000000..78732046437
--- /dev/null
+++ b/arch/x86/platform/olpc/olpc_ofw.c
@@ -0,0 +1,112 @@
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <asm/page.h>
+#include <asm/setup.h>
+#include <asm/io.h>
+#include <asm/pgtable.h>
+#include <asm/olpc_ofw.h>
+
+/* address of OFW callback interface; will be NULL if OFW isn't found */
+static int (*olpc_ofw_cif)(int *);
+
+/* page dir entry containing OFW's pgdir table; filled in by head_32.S */
+u32 olpc_ofw_pgd __initdata;
+
+static DEFINE_SPINLOCK(ofw_lock);
+
+#define MAXARGS 10
+
+void __init setup_olpc_ofw_pgd(void)
+{
+	pgd_t *base, *ofw_pde;
+
+	if (!olpc_ofw_cif)
+		return;
+
+	/* fetch OFW's PDE */
+	base = early_ioremap(olpc_ofw_pgd, sizeof(olpc_ofw_pgd) * PTRS_PER_PGD);
+	if (!base) {
+		printk(KERN_ERR "failed to remap OFW's pgd - disabling OFW!\n");
+		olpc_ofw_cif = NULL;
+		return;
+	}
+	ofw_pde = &base[OLPC_OFW_PDE_NR];
+
+	/* install OFW's PDE permanently into the kernel's pgtable */
+	set_pgd(&swapper_pg_dir[OLPC_OFW_PDE_NR], *ofw_pde);
+	/* implicit optimization barrier here due to uninline function return */
+
+	early_iounmap(base, sizeof(olpc_ofw_pgd) * PTRS_PER_PGD);
+}
+
+int __olpc_ofw(const char *name, int nr_args, const void **args, int nr_res,
+		void **res)
+{
+	int ofw_args[MAXARGS + 3];
+	unsigned long flags;
+	int ret, i, *p;
+
+	BUG_ON(nr_args + nr_res > MAXARGS);
+
+	if (!olpc_ofw_cif)
+		return -EIO;
+
+	ofw_args[0] = (int)name;
+	ofw_args[1] = nr_args;
+	ofw_args[2] = nr_res;
+
+	p = &ofw_args[3];
+	for (i = 0; i < nr_args; i++, p++)
+		*p = (int)args[i];
+
+	/* call into ofw */
+	spin_lock_irqsave(&ofw_lock, flags);
+	ret = olpc_ofw_cif(ofw_args);
+	spin_unlock_irqrestore(&ofw_lock, flags);
+
+	if (!ret) {
+		for (i = 0; i < nr_res; i++, p++)
+			*((int *)res[i]) = *p;
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(__olpc_ofw);
+
+bool olpc_ofw_present(void)
+{
+	return olpc_ofw_cif != NULL;
+}
+EXPORT_SYMBOL_GPL(olpc_ofw_present);
+
+/* OFW cif _should_ be above this address */
+#define OFW_MIN 0xff000000
+
+/* OFW starts on a 1MB boundary */
+#define OFW_BOUND (1<<20)
+
+void __init olpc_ofw_detect(void)
+{
+	struct olpc_ofw_header *hdr = &boot_params.olpc_ofw_header;
+	unsigned long start;
+
+	/* ensure OFW booted us by checking for "OFW " string */
+	if (hdr->ofw_magic != OLPC_OFW_SIG)
+		return;
+
+	olpc_ofw_cif = (int (*)(int *))hdr->cif_handler;
+
+	if ((unsigned long)olpc_ofw_cif < OFW_MIN) {
+		printk(KERN_ERR "OFW detected, but cif has invalid address 0x%lx - disabling.\n",
+				(unsigned long)olpc_ofw_cif);
+		olpc_ofw_cif = NULL;
+		return;
+	}
+
+	/* determine where OFW starts in memory */
+	start = round_down((unsigned long)olpc_ofw_cif, OFW_BOUND);
+	printk(KERN_INFO "OFW detected in memory, cif @ 0x%lx (reserving top %ldMB)\n",
+			(unsigned long)olpc_ofw_cif, (-start) >> 20);
+	reserve_top_address(-start);
+}
diff --git a/arch/x86/platform/scx200/Makefile b/arch/x86/platform/scx200/Makefile
new file mode 100644
index 00000000000..762b4c7f431
--- /dev/null
+++ b/arch/x86/platform/scx200/Makefile
@@ -0,0 +1,2 @@
+obj-$(CONFIG_SCx200)		+= scx200.o
+scx200-y			+= scx200_32.o
diff --git a/arch/x86/platform/scx200/scx200_32.c b/arch/x86/platform/scx200/scx200_32.c
new file mode 100644
index 00000000000..7e004acbe52
--- /dev/null
+++ b/arch/x86/platform/scx200/scx200_32.c
@@ -0,0 +1,131 @@
+/*
+ *  Copyright (c) 2001,2002 Christer Weinigel <wingel@nano-system.com>
+ *
+ *  National Semiconductor SCx200 support.
+ */
+
+#include <linux/module.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <linux/pci.h>
+
+#include <linux/scx200.h>
+#include <linux/scx200_gpio.h>
+
+/* Verify that the configuration block really is there */
+#define scx200_cb_probe(base) (inw((base) + SCx200_CBA) == (base))
+
+#define NAME "scx200"
+
+MODULE_AUTHOR("Christer Weinigel <wingel@nano-system.com>");
+MODULE_DESCRIPTION("NatSemi SCx200 Driver");
+MODULE_LICENSE("GPL");
+
+unsigned scx200_gpio_base = 0;
+unsigned long scx200_gpio_shadow[2];
+
+unsigned scx200_cb_base = 0;
+
+static struct pci_device_id scx200_tbl[] = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_NS, PCI_DEVICE_ID_NS_SCx200_BRIDGE) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_NS, PCI_DEVICE_ID_NS_SC1100_BRIDGE) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_NS, PCI_DEVICE_ID_NS_SCx200_XBUS)   },
+	{ PCI_DEVICE(PCI_VENDOR_ID_NS, PCI_DEVICE_ID_NS_SC1100_XBUS)   },
+	{ },
+};
+MODULE_DEVICE_TABLE(pci,scx200_tbl);
+
+static int __devinit scx200_probe(struct pci_dev *, const struct pci_device_id *);
+
+static struct pci_driver scx200_pci_driver = {
+	.name = "scx200",
+	.id_table = scx200_tbl,
+	.probe = scx200_probe,
+};
+
+static DEFINE_MUTEX(scx200_gpio_config_lock);
+
+static void __devinit scx200_init_shadow(void)
+{
+	int bank;
+
+	/* read the current values driven on the GPIO signals */
+	for (bank = 0; bank < 2; ++bank)
+		scx200_gpio_shadow[bank] = inl(scx200_gpio_base + 0x10 * bank);
+}
+
+static int __devinit scx200_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	unsigned base;
+
+	if (pdev->device == PCI_DEVICE_ID_NS_SCx200_BRIDGE ||
+	    pdev->device == PCI_DEVICE_ID_NS_SC1100_BRIDGE) {
+		base = pci_resource_start(pdev, 0);
+		printk(KERN_INFO NAME ": GPIO base 0x%x\n", base);
+
+		if (!request_region(base, SCx200_GPIO_SIZE, "NatSemi SCx200 GPIO")) {
+			printk(KERN_ERR NAME ": can't allocate I/O for GPIOs\n");
+			return -EBUSY;
+		}
+
+		scx200_gpio_base = base;
+		scx200_init_shadow();
+
+	} else {
+		/* find the base of the Configuration Block */
+		if (scx200_cb_probe(SCx200_CB_BASE_FIXED)) {
+			scx200_cb_base = SCx200_CB_BASE_FIXED;
+		} else {
+			pci_read_config_dword(pdev, SCx200_CBA_SCRATCH, &base);
+			if (scx200_cb_probe(base)) {
+				scx200_cb_base = base;
+			} else {
+				printk(KERN_WARNING NAME ": Configuration Block not found\n");
+				return -ENODEV;
+			}
+		}
+		printk(KERN_INFO NAME ": Configuration Block base 0x%x\n", scx200_cb_base);
+	}
+
+	return 0;
+}
+
+u32 scx200_gpio_configure(unsigned index, u32 mask, u32 bits)
+{
+	u32 config, new_config;
+
+	mutex_lock(&scx200_gpio_config_lock);
+
+	outl(index, scx200_gpio_base + 0x20);
+	config = inl(scx200_gpio_base + 0x24);
+
+	new_config = (config & mask) | bits;
+	outl(new_config, scx200_gpio_base + 0x24);
+
+	mutex_unlock(&scx200_gpio_config_lock);
+
+	return config;
+}
+
+static int __init scx200_init(void)
+{
+	printk(KERN_INFO NAME ": NatSemi SCx200 Driver\n");
+
+	return pci_register_driver(&scx200_pci_driver);
+}
+
+static void __exit scx200_cleanup(void)
+{
+	pci_unregister_driver(&scx200_pci_driver);
+	release_region(scx200_gpio_base, SCx200_GPIO_SIZE);
+}
+
+module_init(scx200_init);
+module_exit(scx200_cleanup);
+
+EXPORT_SYMBOL(scx200_gpio_base);
+EXPORT_SYMBOL(scx200_gpio_shadow);
+EXPORT_SYMBOL(scx200_gpio_configure);
+EXPORT_SYMBOL(scx200_cb_base);
diff --git a/arch/x86/platform/sfi/Makefile b/arch/x86/platform/sfi/Makefile
new file mode 100644
index 00000000000..cc5db1168a5
--- /dev/null
+++ b/arch/x86/platform/sfi/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_SFI)		+= sfi.o
diff --git a/arch/x86/platform/sfi/sfi.c b/arch/x86/platform/sfi/sfi.c
new file mode 100644
index 00000000000..dd4c281ffe5
--- /dev/null
+++ b/arch/x86/platform/sfi/sfi.c
@@ -0,0 +1,120 @@
+/*
+ * sfi.c - x86 architecture SFI support.
+ *
+ * Copyright (c) 2009, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ */
+
+#define KMSG_COMPONENT "SFI"
+#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
+
+#include <linux/acpi.h>
+#include <linux/init.h>
+#include <linux/sfi.h>
+#include <linux/io.h>
+
+#include <asm/io_apic.h>
+#include <asm/mpspec.h>
+#include <asm/setup.h>
+#include <asm/apic.h>
+
+#ifdef CONFIG_X86_LOCAL_APIC
+static unsigned long sfi_lapic_addr __initdata = APIC_DEFAULT_PHYS_BASE;
+
+static void __init mp_sfi_register_lapic_address(unsigned long address)
+{
+	mp_lapic_addr = address;
+
+	set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr);
+	if (boot_cpu_physical_apicid == -1U)
+		boot_cpu_physical_apicid = read_apic_id();
+
+	pr_info("Boot CPU = %d\n", boot_cpu_physical_apicid);
+}
+
+/* All CPUs enumerated by SFI must be present and enabled */
+static void __cpuinit mp_sfi_register_lapic(u8 id)
+{
+	if (MAX_APICS - id <= 0) {
+		pr_warning("Processor #%d invalid (max %d)\n",
+			id, MAX_APICS);
+		return;
+	}
+
+	pr_info("registering lapic[%d]\n", id);
+
+	generic_processor_info(id, GET_APIC_VERSION(apic_read(APIC_LVR)));
+}
+
+static int __init sfi_parse_cpus(struct sfi_table_header *table)
+{
+	struct sfi_table_simple *sb;
+	struct sfi_cpu_table_entry *pentry;
+	int i;
+	int cpu_num;
+
+	sb = (struct sfi_table_simple *)table;
+	cpu_num = SFI_GET_NUM_ENTRIES(sb, struct sfi_cpu_table_entry);
+	pentry = (struct sfi_cpu_table_entry *)sb->pentry;
+
+	for (i = 0; i < cpu_num; i++) {
+		mp_sfi_register_lapic(pentry->apic_id);
+		pentry++;
+	}
+
+	smp_found_config = 1;
+	return 0;
+}
+#endif /* CONFIG_X86_LOCAL_APIC */
+
+#ifdef CONFIG_X86_IO_APIC
+
+static int __init sfi_parse_ioapic(struct sfi_table_header *table)
+{
+	struct sfi_table_simple *sb;
+	struct sfi_apic_table_entry *pentry;
+	int i, num;
+
+	sb = (struct sfi_table_simple *)table;
+	num = SFI_GET_NUM_ENTRIES(sb, struct sfi_apic_table_entry);
+	pentry = (struct sfi_apic_table_entry *)sb->pentry;
+
+	for (i = 0; i < num; i++) {
+		mp_register_ioapic(i, pentry->phys_addr, gsi_top);
+		pentry++;
+	}
+
+	WARN(pic_mode, KERN_WARNING
+		"SFI: pic_mod shouldn't be 1 when IOAPIC table is present\n");
+	pic_mode = 0;
+	return 0;
+}
+#endif /* CONFIG_X86_IO_APIC */
+
+/*
+ * sfi_platform_init(): register lapics & io-apics
+ */
+int __init sfi_platform_init(void)
+{
+#ifdef CONFIG_X86_LOCAL_APIC
+	mp_sfi_register_lapic_address(sfi_lapic_addr);
+	sfi_table_parse(SFI_SIG_CPUS, NULL, NULL, sfi_parse_cpus);
+#endif
+#ifdef CONFIG_X86_IO_APIC
+	sfi_table_parse(SFI_SIG_APIC, NULL, NULL, sfi_parse_ioapic);
+#endif
+	return 0;
+}
diff --git a/arch/x86/platform/uv/Makefile b/arch/x86/platform/uv/Makefile
new file mode 100644
index 00000000000..6c40995fefb
--- /dev/null
+++ b/arch/x86/platform/uv/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_X86_UV)		+= tlb_uv.o bios_uv.o uv_irq.o uv_sysfs.o uv_time.o
diff --git a/arch/x86/platform/uv/bios_uv.c b/arch/x86/platform/uv/bios_uv.c
new file mode 100644
index 00000000000..8bc57baaa9a
--- /dev/null
+++ b/arch/x86/platform/uv/bios_uv.c
@@ -0,0 +1,215 @@
+/*
+ * BIOS run time interface routines.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ *
+ *  Copyright (c) 2008-2009 Silicon Graphics, Inc.  All Rights Reserved.
+ *  Copyright (c) Russ Anderson <rja@sgi.com>
+ */
+
+#include <linux/efi.h>
+#include <asm/efi.h>
+#include <linux/io.h>
+#include <asm/uv/bios.h>
+#include <asm/uv/uv_hub.h>
+
+static struct uv_systab uv_systab;
+
+s64 uv_bios_call(enum uv_bios_cmd which, u64 a1, u64 a2, u64 a3, u64 a4, u64 a5)
+{
+	struct uv_systab *tab = &uv_systab;
+	s64 ret;
+
+	if (!tab->function)
+		/*
+		 * BIOS does not support UV systab
+		 */
+		return BIOS_STATUS_UNIMPLEMENTED;
+
+	ret = efi_call6((void *)__va(tab->function), (u64)which,
+			a1, a2, a3, a4, a5);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(uv_bios_call);
+
+s64 uv_bios_call_irqsave(enum uv_bios_cmd which, u64 a1, u64 a2, u64 a3,
+					u64 a4, u64 a5)
+{
+	unsigned long bios_flags;
+	s64 ret;
+
+	local_irq_save(bios_flags);
+	ret = uv_bios_call(which, a1, a2, a3, a4, a5);
+	local_irq_restore(bios_flags);
+
+	return ret;
+}
+
+s64 uv_bios_call_reentrant(enum uv_bios_cmd which, u64 a1, u64 a2, u64 a3,
+					u64 a4, u64 a5)
+{
+	s64 ret;
+
+	preempt_disable();
+	ret = uv_bios_call(which, a1, a2, a3, a4, a5);
+	preempt_enable();
+
+	return ret;
+}
+
+
+long sn_partition_id;
+EXPORT_SYMBOL_GPL(sn_partition_id);
+long sn_coherency_id;
+EXPORT_SYMBOL_GPL(sn_coherency_id);
+long sn_region_size;
+EXPORT_SYMBOL_GPL(sn_region_size);
+long system_serial_number;
+EXPORT_SYMBOL_GPL(system_serial_number);
+int uv_type;
+EXPORT_SYMBOL_GPL(uv_type);
+
+
+s64 uv_bios_get_sn_info(int fc, int *uvtype, long *partid, long *coher,
+		long *region, long *ssn)
+{
+	s64 ret;
+	u64 v0, v1;
+	union partition_info_u part;
+
+	ret = uv_bios_call_irqsave(UV_BIOS_GET_SN_INFO, fc,
+				(u64)(&v0), (u64)(&v1), 0, 0);
+	if (ret != BIOS_STATUS_SUCCESS)
+		return ret;
+
+	part.val = v0;
+	if (uvtype)
+		*uvtype = part.hub_version;
+	if (partid)
+		*partid = part.partition_id;
+	if (coher)
+		*coher = part.coherence_id;
+	if (region)
+		*region = part.region_size;
+	if (ssn)
+		*ssn = v1;
+	return ret;
+}
+EXPORT_SYMBOL_GPL(uv_bios_get_sn_info);
+
+int
+uv_bios_mq_watchlist_alloc(unsigned long addr, unsigned int mq_size,
+			   unsigned long *intr_mmr_offset)
+{
+	u64 watchlist;
+	s64 ret;
+
+	/*
+	 * bios returns watchlist number or negative error number.
+	 */
+	ret = (int)uv_bios_call_irqsave(UV_BIOS_WATCHLIST_ALLOC, addr,
+			mq_size, (u64)intr_mmr_offset,
+			(u64)&watchlist, 0);
+	if (ret < BIOS_STATUS_SUCCESS)
+		return ret;
+
+	return watchlist;
+}
+EXPORT_SYMBOL_GPL(uv_bios_mq_watchlist_alloc);
+
+int
+uv_bios_mq_watchlist_free(int blade, int watchlist_num)
+{
+	return (int)uv_bios_call_irqsave(UV_BIOS_WATCHLIST_FREE,
+				blade, watchlist_num, 0, 0, 0);
+}
+EXPORT_SYMBOL_GPL(uv_bios_mq_watchlist_free);
+
+s64
+uv_bios_change_memprotect(u64 paddr, u64 len, enum uv_memprotect perms)
+{
+	return uv_bios_call_irqsave(UV_BIOS_MEMPROTECT, paddr, len,
+					perms, 0, 0);
+}
+EXPORT_SYMBOL_GPL(uv_bios_change_memprotect);
+
+s64
+uv_bios_reserved_page_pa(u64 buf, u64 *cookie, u64 *addr, u64 *len)
+{
+	s64 ret;
+
+	ret = uv_bios_call_irqsave(UV_BIOS_GET_PARTITION_ADDR, (u64)cookie,
+					(u64)addr, buf, (u64)len, 0);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(uv_bios_reserved_page_pa);
+
+s64 uv_bios_freq_base(u64 clock_type, u64 *ticks_per_second)
+{
+	return uv_bios_call(UV_BIOS_FREQ_BASE, clock_type,
+			   (u64)ticks_per_second, 0, 0, 0);
+}
+EXPORT_SYMBOL_GPL(uv_bios_freq_base);
+
+/*
+ * uv_bios_set_legacy_vga_target - Set Legacy VGA I/O Target
+ * @decode: true to enable target, false to disable target
+ * @domain: PCI domain number
+ * @bus: PCI bus number
+ *
+ * Returns:
+ *    0: Success
+ *    -EINVAL: Invalid domain or bus number
+ *    -ENOSYS: Capability not available
+ *    -EBUSY: Legacy VGA I/O cannot be retargeted at this time
+ */
+int uv_bios_set_legacy_vga_target(bool decode, int domain, int bus)
+{
+	return uv_bios_call(UV_BIOS_SET_LEGACY_VGA_TARGET,
+				(u64)decode, (u64)domain, (u64)bus, 0, 0);
+}
+EXPORT_SYMBOL_GPL(uv_bios_set_legacy_vga_target);
+
+
+#ifdef CONFIG_EFI
+void uv_bios_init(void)
+{
+	struct uv_systab *tab;
+
+	if ((efi.uv_systab == EFI_INVALID_TABLE_ADDR) ||
+	    (efi.uv_systab == (unsigned long)NULL)) {
+		printk(KERN_CRIT "No EFI UV System Table.\n");
+		uv_systab.function = (unsigned long)NULL;
+		return;
+	}
+
+	tab = (struct uv_systab *)ioremap(efi.uv_systab,
+					sizeof(struct uv_systab));
+	if (strncmp(tab->signature, "UVST", 4) != 0)
+		printk(KERN_ERR "bad signature in UV system table!");
+
+	/*
+	 * Copy table to permanent spot for later use.
+	 */
+	memcpy(&uv_systab, tab, sizeof(struct uv_systab));
+	iounmap(tab);
+
+	printk(KERN_INFO "EFI UV System Table Revision %d\n",
+					uv_systab.revision);
+}
+#else	/* !CONFIG_EFI */
+
+void uv_bios_init(void) { }
+#endif
diff --git a/arch/x86/platform/uv/tlb_uv.c b/arch/x86/platform/uv/tlb_uv.c
new file mode 100644
index 00000000000..a318194002b
--- /dev/null
+++ b/arch/x86/platform/uv/tlb_uv.c
@@ -0,0 +1,1660 @@
+/*
+ *	SGI UltraViolet TLB flush routines.
+ *
+ *	(c) 2008-2010 Cliff Wickman <cpw@sgi.com>, SGI.
+ *
+ *	This code is released under the GNU General Public License version 2 or
+ *	later.
+ */
+#include <linux/seq_file.h>
+#include <linux/proc_fs.h>
+#include <linux/debugfs.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+
+#include <asm/mmu_context.h>
+#include <asm/uv/uv.h>
+#include <asm/uv/uv_mmrs.h>
+#include <asm/uv/uv_hub.h>
+#include <asm/uv/uv_bau.h>
+#include <asm/apic.h>
+#include <asm/idle.h>
+#include <asm/tsc.h>
+#include <asm/irq_vectors.h>
+#include <asm/timer.h>
+
+/* timeouts in nanoseconds (indexed by UVH_AGING_PRESCALE_SEL urgency7 30:28) */
+static int timeout_base_ns[] = {
+		20,
+		160,
+		1280,
+		10240,
+		81920,
+		655360,
+		5242880,
+		167772160
+};
+static int timeout_us;
+static int nobau;
+static int baudisabled;
+static spinlock_t disable_lock;
+static cycles_t congested_cycles;
+
+/* tunables: */
+static int max_bau_concurrent = MAX_BAU_CONCURRENT;
+static int max_bau_concurrent_constant = MAX_BAU_CONCURRENT;
+static int plugged_delay = PLUGGED_DELAY;
+static int plugsb4reset = PLUGSB4RESET;
+static int timeoutsb4reset = TIMEOUTSB4RESET;
+static int ipi_reset_limit = IPI_RESET_LIMIT;
+static int complete_threshold = COMPLETE_THRESHOLD;
+static int congested_response_us = CONGESTED_RESPONSE_US;
+static int congested_reps = CONGESTED_REPS;
+static int congested_period = CONGESTED_PERIOD;
+static struct dentry *tunables_dir;
+static struct dentry *tunables_file;
+
+static int __init setup_nobau(char *arg)
+{
+	nobau = 1;
+	return 0;
+}
+early_param("nobau", setup_nobau);
+
+/* base pnode in this partition */
+static int uv_partition_base_pnode __read_mostly;
+/* position of pnode (which is nasid>>1): */
+static int uv_nshift __read_mostly;
+static unsigned long uv_mmask __read_mostly;
+
+static DEFINE_PER_CPU(struct ptc_stats, ptcstats);
+static DEFINE_PER_CPU(struct bau_control, bau_control);
+static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask);
+
+/*
+ * Determine the first node on a uvhub. 'Nodes' are used for kernel
+ * memory allocation.
+ */
+static int __init uvhub_to_first_node(int uvhub)
+{
+	int node, b;
+
+	for_each_online_node(node) {
+		b = uv_node_to_blade_id(node);
+		if (uvhub == b)
+			return node;
+	}
+	return -1;
+}
+
+/*
+ * Determine the apicid of the first cpu on a uvhub.
+ */
+static int __init uvhub_to_first_apicid(int uvhub)
+{
+	int cpu;
+
+	for_each_present_cpu(cpu)
+		if (uvhub == uv_cpu_to_blade_id(cpu))
+			return per_cpu(x86_cpu_to_apicid, cpu);
+	return -1;
+}
+
+/*
+ * Free a software acknowledge hardware resource by clearing its Pending
+ * bit. This will return a reply to the sender.
+ * If the message has timed out, a reply has already been sent by the
+ * hardware but the resource has not been released. In that case our
+ * clear of the Timeout bit (as well) will free the resource. No reply will
+ * be sent (the hardware will only do one reply per message).
+ */
+static inline void uv_reply_to_message(struct msg_desc *mdp,
+				       struct bau_control *bcp)
+{
+	unsigned long dw;
+	struct bau_payload_queue_entry *msg;
+
+	msg = mdp->msg;
+	if (!msg->canceled) {
+		dw = (msg->sw_ack_vector << UV_SW_ACK_NPENDING) |
+						msg->sw_ack_vector;
+		uv_write_local_mmr(
+				UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS, dw);
+	}
+	msg->replied_to = 1;
+	msg->sw_ack_vector = 0;
+}
+
+/*
+ * Process the receipt of a RETRY message
+ */
+static inline void uv_bau_process_retry_msg(struct msg_desc *mdp,
+					    struct bau_control *bcp)
+{
+	int i;
+	int cancel_count = 0;
+	int slot2;
+	unsigned long msg_res;
+	unsigned long mmr = 0;
+	struct bau_payload_queue_entry *msg;
+	struct bau_payload_queue_entry *msg2;
+	struct ptc_stats *stat;
+
+	msg = mdp->msg;
+	stat = bcp->statp;
+	stat->d_retries++;
+	/*
+	 * cancel any message from msg+1 to the retry itself
+	 */
+	for (msg2 = msg+1, i = 0; i < DEST_Q_SIZE; msg2++, i++) {
+		if (msg2 > mdp->va_queue_last)
+			msg2 = mdp->va_queue_first;
+		if (msg2 == msg)
+			break;
+
+		/* same conditions for cancellation as uv_do_reset */
+		if ((msg2->replied_to == 0) && (msg2->canceled == 0) &&
+		    (msg2->sw_ack_vector) && ((msg2->sw_ack_vector &
+			msg->sw_ack_vector) == 0) &&
+		    (msg2->sending_cpu == msg->sending_cpu) &&
+		    (msg2->msg_type != MSG_NOOP)) {
+			slot2 = msg2 - mdp->va_queue_first;
+			mmr = uv_read_local_mmr
+				(UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE);
+			msg_res = msg2->sw_ack_vector;
+			/*
+			 * This is a message retry; clear the resources held
+			 * by the previous message only if they timed out.
+			 * If it has not timed out we have an unexpected
+			 * situation to report.
+			 */
+			if (mmr & (msg_res << UV_SW_ACK_NPENDING)) {
+				/*
+				 * is the resource timed out?
+				 * make everyone ignore the cancelled message.
+				 */
+				msg2->canceled = 1;
+				stat->d_canceled++;
+				cancel_count++;
+				uv_write_local_mmr(
+				    UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS,
+					(msg_res << UV_SW_ACK_NPENDING) |
+					 msg_res);
+			}
+		}
+	}
+	if (!cancel_count)
+		stat->d_nocanceled++;
+}
+
+/*
+ * Do all the things a cpu should do for a TLB shootdown message.
+ * Other cpu's may come here at the same time for this message.
+ */
+static void uv_bau_process_message(struct msg_desc *mdp,
+				   struct bau_control *bcp)
+{
+	int msg_ack_count;
+	short socket_ack_count = 0;
+	struct ptc_stats *stat;
+	struct bau_payload_queue_entry *msg;
+	struct bau_control *smaster = bcp->socket_master;
+
+	/*
+	 * This must be a normal message, or retry of a normal message
+	 */
+	msg = mdp->msg;
+	stat = bcp->statp;
+	if (msg->address == TLB_FLUSH_ALL) {
+		local_flush_tlb();
+		stat->d_alltlb++;
+	} else {
+		__flush_tlb_one(msg->address);
+		stat->d_onetlb++;
+	}
+	stat->d_requestee++;
+
+	/*
+	 * One cpu on each uvhub has the additional job on a RETRY
+	 * of releasing the resource held by the message that is
+	 * being retried.  That message is identified by sending
+	 * cpu number.
+	 */
+	if (msg->msg_type == MSG_RETRY && bcp == bcp->uvhub_master)
+		uv_bau_process_retry_msg(mdp, bcp);
+
+	/*
+	 * This is a sw_ack message, so we have to reply to it.
+	 * Count each responding cpu on the socket. This avoids
+	 * pinging the count's cache line back and forth between
+	 * the sockets.
+	 */
+	socket_ack_count = atomic_add_short_return(1, (struct atomic_short *)
+			&smaster->socket_acknowledge_count[mdp->msg_slot]);
+	if (socket_ack_count == bcp->cpus_in_socket) {
+		/*
+		 * Both sockets dump their completed count total into
+		 * the message's count.
+		 */
+		smaster->socket_acknowledge_count[mdp->msg_slot] = 0;
+		msg_ack_count = atomic_add_short_return(socket_ack_count,
+				(struct atomic_short *)&msg->acknowledge_count);
+
+		if (msg_ack_count == bcp->cpus_in_uvhub) {
+			/*
+			 * All cpus in uvhub saw it; reply
+			 */
+			uv_reply_to_message(mdp, bcp);
+		}
+	}
+
+	return;
+}
+
+/*
+ * Determine the first cpu on a uvhub.
+ */
+static int uvhub_to_first_cpu(int uvhub)
+{
+	int cpu;
+	for_each_present_cpu(cpu)
+		if (uvhub == uv_cpu_to_blade_id(cpu))
+			return cpu;
+	return -1;
+}
+
+/*
+ * Last resort when we get a large number of destination timeouts is
+ * to clear resources held by a given cpu.
+ * Do this with IPI so that all messages in the BAU message queue
+ * can be identified by their nonzero sw_ack_vector field.
+ *
+ * This is entered for a single cpu on the uvhub.
+ * The sender want's this uvhub to free a specific message's
+ * sw_ack resources.
+ */
+static void
+uv_do_reset(void *ptr)
+{
+	int i;
+	int slot;
+	int count = 0;
+	unsigned long mmr;
+	unsigned long msg_res;
+	struct bau_control *bcp;
+	struct reset_args *rap;
+	struct bau_payload_queue_entry *msg;
+	struct ptc_stats *stat;
+
+	bcp = &per_cpu(bau_control, smp_processor_id());
+	rap = (struct reset_args *)ptr;
+	stat = bcp->statp;
+	stat->d_resets++;
+
+	/*
+	 * We're looking for the given sender, and
+	 * will free its sw_ack resource.
+	 * If all cpu's finally responded after the timeout, its
+	 * message 'replied_to' was set.
+	 */
+	for (msg = bcp->va_queue_first, i = 0; i < DEST_Q_SIZE; msg++, i++) {
+		/* uv_do_reset: same conditions for cancellation as
+		   uv_bau_process_retry_msg() */
+		if ((msg->replied_to == 0) &&
+		    (msg->canceled == 0) &&
+		    (msg->sending_cpu == rap->sender) &&
+		    (msg->sw_ack_vector) &&
+		    (msg->msg_type != MSG_NOOP)) {
+			/*
+			 * make everyone else ignore this message
+			 */
+			msg->canceled = 1;
+			slot = msg - bcp->va_queue_first;
+			count++;
+			/*
+			 * only reset the resource if it is still pending
+			 */
+			mmr = uv_read_local_mmr
+					(UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE);
+			msg_res = msg->sw_ack_vector;
+			if (mmr & msg_res) {
+				stat->d_rcanceled++;
+				uv_write_local_mmr(
+				    UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS,
+					(msg_res << UV_SW_ACK_NPENDING) |
+					 msg_res);
+			}
+		}
+	}
+	return;
+}
+
+/*
+ * Use IPI to get all target uvhubs to release resources held by
+ * a given sending cpu number.
+ */
+static void uv_reset_with_ipi(struct bau_target_uvhubmask *distribution,
+			      int sender)
+{
+	int uvhub;
+	int cpu;
+	cpumask_t mask;
+	struct reset_args reset_args;
+
+	reset_args.sender = sender;
+
+	cpus_clear(mask);
+	/* find a single cpu for each uvhub in this distribution mask */
+	for (uvhub = 0;
+		    uvhub < sizeof(struct bau_target_uvhubmask) * BITSPERBYTE;
+		    uvhub++) {
+		if (!bau_uvhub_isset(uvhub, distribution))
+			continue;
+		/* find a cpu for this uvhub */
+		cpu = uvhub_to_first_cpu(uvhub);
+		cpu_set(cpu, mask);
+	}
+	/* IPI all cpus; Preemption is already disabled */
+	smp_call_function_many(&mask, uv_do_reset, (void *)&reset_args, 1);
+	return;
+}
+
+static inline unsigned long
+cycles_2_us(unsigned long long cyc)
+{
+	unsigned long long ns;
+	unsigned long us;
+	ns =  (cyc * per_cpu(cyc2ns, smp_processor_id()))
+						>> CYC2NS_SCALE_FACTOR;
+	us = ns / 1000;
+	return us;
+}
+
+/*
+ * wait for all cpus on this hub to finish their sends and go quiet
+ * leaves uvhub_quiesce set so that no new broadcasts are started by
+ * bau_flush_send_and_wait()
+ */
+static inline void
+quiesce_local_uvhub(struct bau_control *hmaster)
+{
+	atomic_add_short_return(1, (struct atomic_short *)
+		 &hmaster->uvhub_quiesce);
+}
+
+/*
+ * mark this quiet-requestor as done
+ */
+static inline void
+end_uvhub_quiesce(struct bau_control *hmaster)
+{
+	atomic_add_short_return(-1, (struct atomic_short *)
+		&hmaster->uvhub_quiesce);
+}
+
+/*
+ * Wait for completion of a broadcast software ack message
+ * return COMPLETE, RETRY(PLUGGED or TIMEOUT) or GIVEUP
+ */
+static int uv_wait_completion(struct bau_desc *bau_desc,
+	unsigned long mmr_offset, int right_shift, int this_cpu,
+	struct bau_control *bcp, struct bau_control *smaster, long try)
+{
+	unsigned long descriptor_status;
+	cycles_t ttime;
+	struct ptc_stats *stat = bcp->statp;
+	struct bau_control *hmaster;
+
+	hmaster = bcp->uvhub_master;
+
+	/* spin on the status MMR, waiting for it to go idle */
+	while ((descriptor_status = (((unsigned long)
+		uv_read_local_mmr(mmr_offset) >>
+			right_shift) & UV_ACT_STATUS_MASK)) !=
+			DESC_STATUS_IDLE) {
+		/*
+		 * Our software ack messages may be blocked because there are
+		 * no swack resources available.  As long as none of them
+		 * has timed out hardware will NACK our message and its
+		 * state will stay IDLE.
+		 */
+		if (descriptor_status == DESC_STATUS_SOURCE_TIMEOUT) {
+			stat->s_stimeout++;
+			return FLUSH_GIVEUP;
+		} else if (descriptor_status ==
+					DESC_STATUS_DESTINATION_TIMEOUT) {
+			stat->s_dtimeout++;
+			ttime = get_cycles();
+
+			/*
+			 * Our retries may be blocked by all destination
+			 * swack resources being consumed, and a timeout
+			 * pending.  In that case hardware returns the
+			 * ERROR that looks like a destination timeout.
+			 */
+			if (cycles_2_us(ttime - bcp->send_message) <
+							timeout_us) {
+				bcp->conseccompletes = 0;
+				return FLUSH_RETRY_PLUGGED;
+			}
+
+			bcp->conseccompletes = 0;
+			return FLUSH_RETRY_TIMEOUT;
+		} else {
+			/*
+			 * descriptor_status is still BUSY
+			 */
+			cpu_relax();
+		}
+	}
+	bcp->conseccompletes++;
+	return FLUSH_COMPLETE;
+}
+
+static inline cycles_t
+sec_2_cycles(unsigned long sec)
+{
+	unsigned long ns;
+	cycles_t cyc;
+
+	ns = sec * 1000000000;
+	cyc = (ns << CYC2NS_SCALE_FACTOR)/(per_cpu(cyc2ns, smp_processor_id()));
+	return cyc;
+}
+
+/*
+ * conditionally add 1 to *v, unless *v is >= u
+ * return 0 if we cannot add 1 to *v because it is >= u
+ * return 1 if we can add 1 to *v because it is < u
+ * the add is atomic
+ *
+ * This is close to atomic_add_unless(), but this allows the 'u' value
+ * to be lowered below the current 'v'.  atomic_add_unless can only stop
+ * on equal.
+ */
+static inline int atomic_inc_unless_ge(spinlock_t *lock, atomic_t *v, int u)
+{
+	spin_lock(lock);
+	if (atomic_read(v) >= u) {
+		spin_unlock(lock);
+		return 0;
+	}
+	atomic_inc(v);
+	spin_unlock(lock);
+	return 1;
+}
+
+/*
+ * Our retries are blocked by all destination swack resources being
+ * in use, and a timeout is pending. In that case hardware immediately
+ * returns the ERROR that looks like a destination timeout.
+ */
+static void
+destination_plugged(struct bau_desc *bau_desc, struct bau_control *bcp,
+			struct bau_control *hmaster, struct ptc_stats *stat)
+{
+	udelay(bcp->plugged_delay);
+	bcp->plugged_tries++;
+	if (bcp->plugged_tries >= bcp->plugsb4reset) {
+		bcp->plugged_tries = 0;
+		quiesce_local_uvhub(hmaster);
+		spin_lock(&hmaster->queue_lock);
+		uv_reset_with_ipi(&bau_desc->distribution, bcp->cpu);
+		spin_unlock(&hmaster->queue_lock);
+		end_uvhub_quiesce(hmaster);
+		bcp->ipi_attempts++;
+		stat->s_resets_plug++;
+	}
+}
+
+static void
+destination_timeout(struct bau_desc *bau_desc, struct bau_control *bcp,
+			struct bau_control *hmaster, struct ptc_stats *stat)
+{
+	hmaster->max_bau_concurrent = 1;
+	bcp->timeout_tries++;
+	if (bcp->timeout_tries >= bcp->timeoutsb4reset) {
+		bcp->timeout_tries = 0;
+		quiesce_local_uvhub(hmaster);
+		spin_lock(&hmaster->queue_lock);
+		uv_reset_with_ipi(&bau_desc->distribution, bcp->cpu);
+		spin_unlock(&hmaster->queue_lock);
+		end_uvhub_quiesce(hmaster);
+		bcp->ipi_attempts++;
+		stat->s_resets_timeout++;
+	}
+}
+
+/*
+ * Completions are taking a very long time due to a congested numalink
+ * network.
+ */
+static void
+disable_for_congestion(struct bau_control *bcp, struct ptc_stats *stat)
+{
+	int tcpu;
+	struct bau_control *tbcp;
+
+	/* let only one cpu do this disabling */
+	spin_lock(&disable_lock);
+	if (!baudisabled && bcp->period_requests &&
+	    ((bcp->period_time / bcp->period_requests) > congested_cycles)) {
+		/* it becomes this cpu's job to turn on the use of the
+		   BAU again */
+		baudisabled = 1;
+		bcp->set_bau_off = 1;
+		bcp->set_bau_on_time = get_cycles() +
+			sec_2_cycles(bcp->congested_period);
+		stat->s_bau_disabled++;
+		for_each_present_cpu(tcpu) {
+			tbcp = &per_cpu(bau_control, tcpu);
+				tbcp->baudisabled = 1;
+		}
+	}
+	spin_unlock(&disable_lock);
+}
+
+/**
+ * uv_flush_send_and_wait
+ *
+ * Send a broadcast and wait for it to complete.
+ *
+ * The flush_mask contains the cpus the broadcast is to be sent to including
+ * cpus that are on the local uvhub.
+ *
+ * Returns 0 if all flushing represented in the mask was done.
+ * Returns 1 if it gives up entirely and the original cpu mask is to be
+ * returned to the kernel.
+ */
+int uv_flush_send_and_wait(struct bau_desc *bau_desc,
+			   struct cpumask *flush_mask, struct bau_control *bcp)
+{
+	int right_shift;
+	int completion_status = 0;
+	int seq_number = 0;
+	long try = 0;
+	int cpu = bcp->uvhub_cpu;
+	int this_cpu = bcp->cpu;
+	unsigned long mmr_offset;
+	unsigned long index;
+	cycles_t time1;
+	cycles_t time2;
+	cycles_t elapsed;
+	struct ptc_stats *stat = bcp->statp;
+	struct bau_control *smaster = bcp->socket_master;
+	struct bau_control *hmaster = bcp->uvhub_master;
+
+	if (!atomic_inc_unless_ge(&hmaster->uvhub_lock,
+			&hmaster->active_descriptor_count,
+			hmaster->max_bau_concurrent)) {
+		stat->s_throttles++;
+		do {
+			cpu_relax();
+		} while (!atomic_inc_unless_ge(&hmaster->uvhub_lock,
+			&hmaster->active_descriptor_count,
+			hmaster->max_bau_concurrent));
+	}
+	while (hmaster->uvhub_quiesce)
+		cpu_relax();
+
+	if (cpu < UV_CPUS_PER_ACT_STATUS) {
+		mmr_offset = UVH_LB_BAU_SB_ACTIVATION_STATUS_0;
+		right_shift = cpu * UV_ACT_STATUS_SIZE;
+	} else {
+		mmr_offset = UVH_LB_BAU_SB_ACTIVATION_STATUS_1;
+		right_shift =
+		    ((cpu - UV_CPUS_PER_ACT_STATUS) * UV_ACT_STATUS_SIZE);
+	}
+	time1 = get_cycles();
+	do {
+		if (try == 0) {
+			bau_desc->header.msg_type = MSG_REGULAR;
+			seq_number = bcp->message_number++;
+		} else {
+			bau_desc->header.msg_type = MSG_RETRY;
+			stat->s_retry_messages++;
+		}
+		bau_desc->header.sequence = seq_number;
+		index = (1UL << UVH_LB_BAU_SB_ACTIVATION_CONTROL_PUSH_SHFT) |
+			bcp->uvhub_cpu;
+		bcp->send_message = get_cycles();
+		uv_write_local_mmr(UVH_LB_BAU_SB_ACTIVATION_CONTROL, index);
+		try++;
+		completion_status = uv_wait_completion(bau_desc, mmr_offset,
+			right_shift, this_cpu, bcp, smaster, try);
+
+		if (completion_status == FLUSH_RETRY_PLUGGED) {
+			destination_plugged(bau_desc, bcp, hmaster, stat);
+		} else if (completion_status == FLUSH_RETRY_TIMEOUT) {
+			destination_timeout(bau_desc, bcp, hmaster, stat);
+		}
+		if (bcp->ipi_attempts >= bcp->ipi_reset_limit) {
+			bcp->ipi_attempts = 0;
+			completion_status = FLUSH_GIVEUP;
+			break;
+		}
+		cpu_relax();
+	} while ((completion_status == FLUSH_RETRY_PLUGGED) ||
+		 (completion_status == FLUSH_RETRY_TIMEOUT));
+	time2 = get_cycles();
+	bcp->plugged_tries = 0;
+	bcp->timeout_tries = 0;
+	if ((completion_status == FLUSH_COMPLETE) &&
+	    (bcp->conseccompletes > bcp->complete_threshold) &&
+	    (hmaster->max_bau_concurrent <
+					hmaster->max_bau_concurrent_constant))
+			hmaster->max_bau_concurrent++;
+	while (hmaster->uvhub_quiesce)
+		cpu_relax();
+	atomic_dec(&hmaster->active_descriptor_count);
+	if (time2 > time1) {
+		elapsed = time2 - time1;
+		stat->s_time += elapsed;
+		if ((completion_status == FLUSH_COMPLETE) && (try == 1)) {
+			bcp->period_requests++;
+			bcp->period_time += elapsed;
+			if ((elapsed > congested_cycles) &&
+			    (bcp->period_requests > bcp->congested_reps)) {
+				disable_for_congestion(bcp, stat);
+			}
+		}
+	} else
+		stat->s_requestor--;
+	if (completion_status == FLUSH_COMPLETE && try > 1)
+		stat->s_retriesok++;
+	else if (completion_status == FLUSH_GIVEUP) {
+		stat->s_giveup++;
+		return 1;
+	}
+	return 0;
+}
+
+/**
+ * uv_flush_tlb_others - globally purge translation cache of a virtual
+ * address or all TLB's
+ * @cpumask: mask of all cpu's in which the address is to be removed
+ * @mm: mm_struct containing virtual address range
+ * @va: virtual address to be removed (or TLB_FLUSH_ALL for all TLB's on cpu)
+ * @cpu: the current cpu
+ *
+ * This is the entry point for initiating any UV global TLB shootdown.
+ *
+ * Purges the translation caches of all specified processors of the given
+ * virtual address, or purges all TLB's on specified processors.
+ *
+ * The caller has derived the cpumask from the mm_struct.  This function
+ * is called only if there are bits set in the mask. (e.g. flush_tlb_page())
+ *
+ * The cpumask is converted into a uvhubmask of the uvhubs containing
+ * those cpus.
+ *
+ * Note that this function should be called with preemption disabled.
+ *
+ * Returns NULL if all remote flushing was done.
+ * Returns pointer to cpumask if some remote flushing remains to be
+ * done.  The returned pointer is valid till preemption is re-enabled.
+ */
+const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
+					  struct mm_struct *mm,
+					  unsigned long va, unsigned int cpu)
+{
+	int tcpu;
+	int uvhub;
+	int locals = 0;
+	int remotes = 0;
+	int hubs = 0;
+	struct bau_desc *bau_desc;
+	struct cpumask *flush_mask;
+	struct ptc_stats *stat;
+	struct bau_control *bcp;
+	struct bau_control *tbcp;
+
+	/* kernel was booted 'nobau' */
+	if (nobau)
+		return cpumask;
+
+	bcp = &per_cpu(bau_control, cpu);
+	stat = bcp->statp;
+
+	/* bau was disabled due to slow response */
+	if (bcp->baudisabled) {
+		/* the cpu that disabled it must re-enable it */
+		if (bcp->set_bau_off) {
+			if (get_cycles() >= bcp->set_bau_on_time) {
+				stat->s_bau_reenabled++;
+				baudisabled = 0;
+				for_each_present_cpu(tcpu) {
+					tbcp = &per_cpu(bau_control, tcpu);
+					tbcp->baudisabled = 0;
+					tbcp->period_requests = 0;
+					tbcp->period_time = 0;
+				}
+			}
+		}
+		return cpumask;
+	}
+
+	/*
+	 * Each sending cpu has a per-cpu mask which it fills from the caller's
+	 * cpu mask.  All cpus are converted to uvhubs and copied to the
+	 * activation descriptor.
+	 */
+	flush_mask = (struct cpumask *)per_cpu(uv_flush_tlb_mask, cpu);
+	/* don't actually do a shootdown of the local cpu */
+	cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu));
+	if (cpu_isset(cpu, *cpumask))
+		stat->s_ntargself++;
+
+	bau_desc = bcp->descriptor_base;
+	bau_desc += UV_ITEMS_PER_DESCRIPTOR * bcp->uvhub_cpu;
+	bau_uvhubs_clear(&bau_desc->distribution, UV_DISTRIBUTION_SIZE);
+
+	/* cpu statistics */
+	for_each_cpu(tcpu, flush_mask) {
+		uvhub = uv_cpu_to_blade_id(tcpu);
+		bau_uvhub_set(uvhub, &bau_desc->distribution);
+		if (uvhub == bcp->uvhub)
+			locals++;
+		else
+			remotes++;
+	}
+	if ((locals + remotes) == 0)
+		return NULL;
+	stat->s_requestor++;
+	stat->s_ntargcpu += remotes + locals;
+	stat->s_ntargremotes += remotes;
+	stat->s_ntarglocals += locals;
+	remotes = bau_uvhub_weight(&bau_desc->distribution);
+
+	/* uvhub statistics */
+	hubs = bau_uvhub_weight(&bau_desc->distribution);
+	if (locals) {
+		stat->s_ntarglocaluvhub++;
+		stat->s_ntargremoteuvhub += (hubs - 1);
+	} else
+		stat->s_ntargremoteuvhub += hubs;
+	stat->s_ntarguvhub += hubs;
+	if (hubs >= 16)
+		stat->s_ntarguvhub16++;
+	else if (hubs >= 8)
+		stat->s_ntarguvhub8++;
+	else if (hubs >= 4)
+		stat->s_ntarguvhub4++;
+	else if (hubs >= 2)
+		stat->s_ntarguvhub2++;
+	else
+		stat->s_ntarguvhub1++;
+
+	bau_desc->payload.address = va;
+	bau_desc->payload.sending_cpu = cpu;
+
+	/*
+	 * uv_flush_send_and_wait returns 0 if all cpu's were messaged,
+	 * or 1 if it gave up and the original cpumask should be returned.
+	 */
+	if (!uv_flush_send_and_wait(bau_desc, flush_mask, bcp))
+		return NULL;
+	else
+		return cpumask;
+}
+
+/*
+ * The BAU message interrupt comes here. (registered by set_intr_gate)
+ * See entry_64.S
+ *
+ * We received a broadcast assist message.
+ *
+ * Interrupts are disabled; this interrupt could represent
+ * the receipt of several messages.
+ *
+ * All cores/threads on this hub get this interrupt.
+ * The last one to see it does the software ack.
+ * (the resource will not be freed until noninterruptable cpus see this
+ *  interrupt; hardware may timeout the s/w ack and reply ERROR)
+ */
+void uv_bau_message_interrupt(struct pt_regs *regs)
+{
+	int count = 0;
+	cycles_t time_start;
+	struct bau_payload_queue_entry *msg;
+	struct bau_control *bcp;
+	struct ptc_stats *stat;
+	struct msg_desc msgdesc;
+
+	time_start = get_cycles();
+	bcp = &per_cpu(bau_control, smp_processor_id());
+	stat = bcp->statp;
+	msgdesc.va_queue_first = bcp->va_queue_first;
+	msgdesc.va_queue_last = bcp->va_queue_last;
+	msg = bcp->bau_msg_head;
+	while (msg->sw_ack_vector) {
+		count++;
+		msgdesc.msg_slot = msg - msgdesc.va_queue_first;
+		msgdesc.sw_ack_slot = ffs(msg->sw_ack_vector) - 1;
+		msgdesc.msg = msg;
+		uv_bau_process_message(&msgdesc, bcp);
+		msg++;
+		if (msg > msgdesc.va_queue_last)
+			msg = msgdesc.va_queue_first;
+		bcp->bau_msg_head = msg;
+	}
+	stat->d_time += (get_cycles() - time_start);
+	if (!count)
+		stat->d_nomsg++;
+	else if (count > 1)
+		stat->d_multmsg++;
+	ack_APIC_irq();
+}
+
+/*
+ * uv_enable_timeouts
+ *
+ * Each target uvhub (i.e. a uvhub that has no cpu's) needs to have
+ * shootdown message timeouts enabled.  The timeout does not cause
+ * an interrupt, but causes an error message to be returned to
+ * the sender.
+ */
+static void uv_enable_timeouts(void)
+{
+	int uvhub;
+	int nuvhubs;
+	int pnode;
+	unsigned long mmr_image;
+
+	nuvhubs = uv_num_possible_blades();
+
+	for (uvhub = 0; uvhub < nuvhubs; uvhub++) {
+		if (!uv_blade_nr_possible_cpus(uvhub))
+			continue;
+
+		pnode = uv_blade_to_pnode(uvhub);
+		mmr_image =
+		    uv_read_global_mmr64(pnode, UVH_LB_BAU_MISC_CONTROL);
+		/*
+		 * Set the timeout period and then lock it in, in three
+		 * steps; captures and locks in the period.
+		 *
+		 * To program the period, the SOFT_ACK_MODE must be off.
+		 */
+		mmr_image &= ~((unsigned long)1 <<
+		    UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_SHFT);
+		uv_write_global_mmr64
+		    (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
+		/*
+		 * Set the 4-bit period.
+		 */
+		mmr_image &= ~((unsigned long)0xf <<
+		     UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHFT);
+		mmr_image |= (UV_INTD_SOFT_ACK_TIMEOUT_PERIOD <<
+		     UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHFT);
+		uv_write_global_mmr64
+		    (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
+		/*
+		 * Subsequent reversals of the timebase bit (3) cause an
+		 * immediate timeout of one or all INTD resources as
+		 * indicated in bits 2:0 (7 causes all of them to timeout).
+		 */
+		mmr_image |= ((unsigned long)1 <<
+		    UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_SHFT);
+		uv_write_global_mmr64
+		    (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
+	}
+}
+
+static void *uv_ptc_seq_start(struct seq_file *file, loff_t *offset)
+{
+	if (*offset < num_possible_cpus())
+		return offset;
+	return NULL;
+}
+
+static void *uv_ptc_seq_next(struct seq_file *file, void *data, loff_t *offset)
+{
+	(*offset)++;
+	if (*offset < num_possible_cpus())
+		return offset;
+	return NULL;
+}
+
+static void uv_ptc_seq_stop(struct seq_file *file, void *data)
+{
+}
+
+static inline unsigned long long
+microsec_2_cycles(unsigned long microsec)
+{
+	unsigned long ns;
+	unsigned long long cyc;
+
+	ns = microsec * 1000;
+	cyc = (ns << CYC2NS_SCALE_FACTOR)/(per_cpu(cyc2ns, smp_processor_id()));
+	return cyc;
+}
+
+/*
+ * Display the statistics thru /proc.
+ * 'data' points to the cpu number
+ */
+static int uv_ptc_seq_show(struct seq_file *file, void *data)
+{
+	struct ptc_stats *stat;
+	int cpu;
+
+	cpu = *(loff_t *)data;
+
+	if (!cpu) {
+		seq_printf(file,
+			"# cpu sent stime self locals remotes ncpus localhub ");
+		seq_printf(file,
+			"remotehub numuvhubs numuvhubs16 numuvhubs8 ");
+		seq_printf(file,
+			"numuvhubs4 numuvhubs2 numuvhubs1 dto ");
+		seq_printf(file,
+			"retries rok resetp resett giveup sto bz throt ");
+		seq_printf(file,
+			"sw_ack recv rtime all ");
+		seq_printf(file,
+			"one mult none retry canc nocan reset rcan ");
+		seq_printf(file,
+			"disable enable\n");
+	}
+	if (cpu < num_possible_cpus() && cpu_online(cpu)) {
+		stat = &per_cpu(ptcstats, cpu);
+		/* source side statistics */
+		seq_printf(file,
+			"cpu %d %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld ",
+			   cpu, stat->s_requestor, cycles_2_us(stat->s_time),
+			   stat->s_ntargself, stat->s_ntarglocals,
+			   stat->s_ntargremotes, stat->s_ntargcpu,
+			   stat->s_ntarglocaluvhub, stat->s_ntargremoteuvhub,
+			   stat->s_ntarguvhub, stat->s_ntarguvhub16);
+		seq_printf(file, "%ld %ld %ld %ld %ld ",
+			   stat->s_ntarguvhub8, stat->s_ntarguvhub4,
+			   stat->s_ntarguvhub2, stat->s_ntarguvhub1,
+			   stat->s_dtimeout);
+		seq_printf(file, "%ld %ld %ld %ld %ld %ld %ld %ld ",
+			   stat->s_retry_messages, stat->s_retriesok,
+			   stat->s_resets_plug, stat->s_resets_timeout,
+			   stat->s_giveup, stat->s_stimeout,
+			   stat->s_busy, stat->s_throttles);
+
+		/* destination side statistics */
+		seq_printf(file,
+			   "%lx %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld ",
+			   uv_read_global_mmr64(uv_cpu_to_pnode(cpu),
+					UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE),
+			   stat->d_requestee, cycles_2_us(stat->d_time),
+			   stat->d_alltlb, stat->d_onetlb, stat->d_multmsg,
+			   stat->d_nomsg, stat->d_retries, stat->d_canceled,
+			   stat->d_nocanceled, stat->d_resets,
+			   stat->d_rcanceled);
+		seq_printf(file, "%ld %ld\n",
+			stat->s_bau_disabled, stat->s_bau_reenabled);
+	}
+
+	return 0;
+}
+
+/*
+ * Display the tunables thru debugfs
+ */
+static ssize_t tunables_read(struct file *file, char __user *userbuf,
+						size_t count, loff_t *ppos)
+{
+	char *buf;
+	int ret;
+
+	buf = kasprintf(GFP_KERNEL, "%s %s %s\n%d %d %d %d %d %d %d %d %d\n",
+		"max_bau_concurrent plugged_delay plugsb4reset",
+		"timeoutsb4reset ipi_reset_limit complete_threshold",
+		"congested_response_us congested_reps congested_period",
+		max_bau_concurrent, plugged_delay, plugsb4reset,
+		timeoutsb4reset, ipi_reset_limit, complete_threshold,
+		congested_response_us, congested_reps, congested_period);
+
+	if (!buf)
+		return -ENOMEM;
+
+	ret = simple_read_from_buffer(userbuf, count, ppos, buf, strlen(buf));
+	kfree(buf);
+	return ret;
+}
+
+/*
+ * -1: resetf the statistics
+ *  0: display meaning of the statistics
+ */
+static ssize_t uv_ptc_proc_write(struct file *file, const char __user *user,
+				 size_t count, loff_t *data)
+{
+	int cpu;
+	long input_arg;
+	char optstr[64];
+	struct ptc_stats *stat;
+
+	if (count == 0 || count > sizeof(optstr))
+		return -EINVAL;
+	if (copy_from_user(optstr, user, count))
+		return -EFAULT;
+	optstr[count - 1] = '\0';
+	if (strict_strtol(optstr, 10, &input_arg) < 0) {
+		printk(KERN_DEBUG "%s is invalid\n", optstr);
+		return -EINVAL;
+	}
+
+	if (input_arg == 0) {
+		printk(KERN_DEBUG "# cpu:      cpu number\n");
+		printk(KERN_DEBUG "Sender statistics:\n");
+		printk(KERN_DEBUG
+		"sent:     number of shootdown messages sent\n");
+		printk(KERN_DEBUG
+		"stime:    time spent sending messages\n");
+		printk(KERN_DEBUG
+		"numuvhubs: number of hubs targeted with shootdown\n");
+		printk(KERN_DEBUG
+		"numuvhubs16: number times 16 or more hubs targeted\n");
+		printk(KERN_DEBUG
+		"numuvhubs8: number times 8 or more hubs targeted\n");
+		printk(KERN_DEBUG
+		"numuvhubs4: number times 4 or more hubs targeted\n");
+		printk(KERN_DEBUG
+		"numuvhubs2: number times 2 or more hubs targeted\n");
+		printk(KERN_DEBUG
+		"numuvhubs1: number times 1 hub targeted\n");
+		printk(KERN_DEBUG
+		"numcpus:  number of cpus targeted with shootdown\n");
+		printk(KERN_DEBUG
+		"dto:      number of destination timeouts\n");
+		printk(KERN_DEBUG
+		"retries:  destination timeout retries sent\n");
+		printk(KERN_DEBUG
+		"rok:   :  destination timeouts successfully retried\n");
+		printk(KERN_DEBUG
+		"resetp:   ipi-style resource resets for plugs\n");
+		printk(KERN_DEBUG
+		"resett:   ipi-style resource resets for timeouts\n");
+		printk(KERN_DEBUG
+		"giveup:   fall-backs to ipi-style shootdowns\n");
+		printk(KERN_DEBUG
+		"sto:      number of source timeouts\n");
+		printk(KERN_DEBUG
+		"bz:       number of stay-busy's\n");
+		printk(KERN_DEBUG
+		"throt:    number times spun in throttle\n");
+		printk(KERN_DEBUG "Destination side statistics:\n");
+		printk(KERN_DEBUG
+		"sw_ack:   image of UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE\n");
+		printk(KERN_DEBUG
+		"recv:     shootdown messages received\n");
+		printk(KERN_DEBUG
+		"rtime:    time spent processing messages\n");
+		printk(KERN_DEBUG
+		"all:      shootdown all-tlb messages\n");
+		printk(KERN_DEBUG
+		"one:      shootdown one-tlb messages\n");
+		printk(KERN_DEBUG
+		"mult:     interrupts that found multiple messages\n");
+		printk(KERN_DEBUG
+		"none:     interrupts that found no messages\n");
+		printk(KERN_DEBUG
+		"retry:    number of retry messages processed\n");
+		printk(KERN_DEBUG
+		"canc:     number messages canceled by retries\n");
+		printk(KERN_DEBUG
+		"nocan:    number retries that found nothing to cancel\n");
+		printk(KERN_DEBUG
+		"reset:    number of ipi-style reset requests processed\n");
+		printk(KERN_DEBUG
+		"rcan:     number messages canceled by reset requests\n");
+		printk(KERN_DEBUG
+		"disable:  number times use of the BAU was disabled\n");
+		printk(KERN_DEBUG
+		"enable:   number times use of the BAU was re-enabled\n");
+	} else if (input_arg == -1) {
+		for_each_present_cpu(cpu) {
+			stat = &per_cpu(ptcstats, cpu);
+			memset(stat, 0, sizeof(struct ptc_stats));
+		}
+	}
+
+	return count;
+}
+
+static int local_atoi(const char *name)
+{
+	int val = 0;
+
+	for (;; name++) {
+		switch (*name) {
+		case '0' ... '9':
+			val = 10*val+(*name-'0');
+			break;
+		default:
+			return val;
+		}
+	}
+}
+
+/*
+ * set the tunables
+ * 0 values reset them to defaults
+ */
+static ssize_t tunables_write(struct file *file, const char __user *user,
+				 size_t count, loff_t *data)
+{
+	int cpu;
+	int cnt = 0;
+	int val;
+	char *p;
+	char *q;
+	char instr[64];
+	struct bau_control *bcp;
+
+	if (count == 0 || count > sizeof(instr)-1)
+		return -EINVAL;
+	if (copy_from_user(instr, user, count))
+		return -EFAULT;
+
+	instr[count] = '\0';
+	/* count the fields */
+	p = instr + strspn(instr, WHITESPACE);
+	q = p;
+	for (; *p; p = q + strspn(q, WHITESPACE)) {
+		q = p + strcspn(p, WHITESPACE);
+		cnt++;
+		if (q == p)
+			break;
+	}
+	if (cnt != 9) {
+		printk(KERN_INFO "bau tunable error: should be 9 numbers\n");
+		return -EINVAL;
+	}
+
+	p = instr + strspn(instr, WHITESPACE);
+	q = p;
+	for (cnt = 0; *p; p = q + strspn(q, WHITESPACE), cnt++) {
+		q = p + strcspn(p, WHITESPACE);
+		val = local_atoi(p);
+		switch (cnt) {
+		case 0:
+			if (val == 0) {
+				max_bau_concurrent = MAX_BAU_CONCURRENT;
+				max_bau_concurrent_constant =
+							MAX_BAU_CONCURRENT;
+				continue;
+			}
+			bcp = &per_cpu(bau_control, smp_processor_id());
+			if (val < 1 || val > bcp->cpus_in_uvhub) {
+				printk(KERN_DEBUG
+				"Error: BAU max concurrent %d is invalid\n",
+				val);
+				return -EINVAL;
+			}
+			max_bau_concurrent = val;
+			max_bau_concurrent_constant = val;
+			continue;
+		case 1:
+			if (val == 0)
+				plugged_delay = PLUGGED_DELAY;
+			else
+				plugged_delay = val;
+			continue;
+		case 2:
+			if (val == 0)
+				plugsb4reset = PLUGSB4RESET;
+			else
+				plugsb4reset = val;
+			continue;
+		case 3:
+			if (val == 0)
+				timeoutsb4reset = TIMEOUTSB4RESET;
+			else
+				timeoutsb4reset = val;
+			continue;
+		case 4:
+			if (val == 0)
+				ipi_reset_limit = IPI_RESET_LIMIT;
+			else
+				ipi_reset_limit = val;
+			continue;
+		case 5:
+			if (val == 0)
+				complete_threshold = COMPLETE_THRESHOLD;
+			else
+				complete_threshold = val;
+			continue;
+		case 6:
+			if (val == 0)
+				congested_response_us = CONGESTED_RESPONSE_US;
+			else
+				congested_response_us = val;
+			continue;
+		case 7:
+			if (val == 0)
+				congested_reps = CONGESTED_REPS;
+			else
+				congested_reps = val;
+			continue;
+		case 8:
+			if (val == 0)
+				congested_period = CONGESTED_PERIOD;
+			else
+				congested_period = val;
+			continue;
+		}
+		if (q == p)
+			break;
+	}
+	for_each_present_cpu(cpu) {
+		bcp = &per_cpu(bau_control, cpu);
+		bcp->max_bau_concurrent = max_bau_concurrent;
+		bcp->max_bau_concurrent_constant = max_bau_concurrent;
+		bcp->plugged_delay = plugged_delay;
+		bcp->plugsb4reset = plugsb4reset;
+		bcp->timeoutsb4reset = timeoutsb4reset;
+		bcp->ipi_reset_limit = ipi_reset_limit;
+		bcp->complete_threshold = complete_threshold;
+		bcp->congested_response_us = congested_response_us;
+		bcp->congested_reps = congested_reps;
+		bcp->congested_period = congested_period;
+	}
+	return count;
+}
+
+static const struct seq_operations uv_ptc_seq_ops = {
+	.start		= uv_ptc_seq_start,
+	.next		= uv_ptc_seq_next,
+	.stop		= uv_ptc_seq_stop,
+	.show		= uv_ptc_seq_show
+};
+
+static int uv_ptc_proc_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &uv_ptc_seq_ops);
+}
+
+static int tunables_open(struct inode *inode, struct file *file)
+{
+	return 0;
+}
+
+static const struct file_operations proc_uv_ptc_operations = {
+	.open		= uv_ptc_proc_open,
+	.read		= seq_read,
+	.write		= uv_ptc_proc_write,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static const struct file_operations tunables_fops = {
+	.open		= tunables_open,
+	.read		= tunables_read,
+	.write		= tunables_write,
+	.llseek		= default_llseek,
+};
+
+static int __init uv_ptc_init(void)
+{
+	struct proc_dir_entry *proc_uv_ptc;
+
+	if (!is_uv_system())
+		return 0;
+
+	proc_uv_ptc = proc_create(UV_PTC_BASENAME, 0444, NULL,
+				  &proc_uv_ptc_operations);
+	if (!proc_uv_ptc) {
+		printk(KERN_ERR "unable to create %s proc entry\n",
+		       UV_PTC_BASENAME);
+		return -EINVAL;
+	}
+
+	tunables_dir = debugfs_create_dir(UV_BAU_TUNABLES_DIR, NULL);
+	if (!tunables_dir) {
+		printk(KERN_ERR "unable to create debugfs directory %s\n",
+		       UV_BAU_TUNABLES_DIR);
+		return -EINVAL;
+	}
+	tunables_file = debugfs_create_file(UV_BAU_TUNABLES_FILE, 0600,
+			tunables_dir, NULL, &tunables_fops);
+	if (!tunables_file) {
+		printk(KERN_ERR "unable to create debugfs file %s\n",
+		       UV_BAU_TUNABLES_FILE);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/*
+ * initialize the sending side's sending buffers
+ */
+static void
+uv_activation_descriptor_init(int node, int pnode)
+{
+	int i;
+	int cpu;
+	unsigned long pa;
+	unsigned long m;
+	unsigned long n;
+	struct bau_desc *bau_desc;
+	struct bau_desc *bd2;
+	struct bau_control *bcp;
+
+	/*
+	 * each bau_desc is 64 bytes; there are 8 (UV_ITEMS_PER_DESCRIPTOR)
+	 * per cpu; and up to 32 (UV_ADP_SIZE) cpu's per uvhub
+	 */
+	bau_desc = kmalloc_node(sizeof(struct bau_desc) * UV_ADP_SIZE
+				* UV_ITEMS_PER_DESCRIPTOR, GFP_KERNEL, node);
+	BUG_ON(!bau_desc);
+
+	pa = uv_gpa(bau_desc); /* need the real nasid*/
+	n = pa >> uv_nshift;
+	m = pa & uv_mmask;
+
+	uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE,
+			      (n << UV_DESC_BASE_PNODE_SHIFT | m));
+
+	/*
+	 * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each
+	 * cpu even though we only use the first one; one descriptor can
+	 * describe a broadcast to 256 uv hubs.
+	 */
+	for (i = 0, bd2 = bau_desc; i < (UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR);
+		i++, bd2++) {
+		memset(bd2, 0, sizeof(struct bau_desc));
+		bd2->header.sw_ack_flag = 1;
+		/*
+		 * base_dest_nodeid is the nasid (pnode<<1) of the first uvhub
+		 * in the partition. The bit map will indicate uvhub numbers,
+		 * which are 0-N in a partition. Pnodes are unique system-wide.
+		 */
+		bd2->header.base_dest_nodeid = uv_partition_base_pnode << 1;
+		bd2->header.dest_subnodeid = 0x10; /* the LB */
+		bd2->header.command = UV_NET_ENDPOINT_INTD;
+		bd2->header.int_both = 1;
+		/*
+		 * all others need to be set to zero:
+		 *   fairness chaining multilevel count replied_to
+		 */
+	}
+	for_each_present_cpu(cpu) {
+		if (pnode != uv_blade_to_pnode(uv_cpu_to_blade_id(cpu)))
+			continue;
+		bcp = &per_cpu(bau_control, cpu);
+		bcp->descriptor_base = bau_desc;
+	}
+}
+
+/*
+ * initialize the destination side's receiving buffers
+ * entered for each uvhub in the partition
+ * - node is first node (kernel memory notion) on the uvhub
+ * - pnode is the uvhub's physical identifier
+ */
+static void
+uv_payload_queue_init(int node, int pnode)
+{
+	int pn;
+	int cpu;
+	char *cp;
+	unsigned long pa;
+	struct bau_payload_queue_entry *pqp;
+	struct bau_payload_queue_entry *pqp_malloc;
+	struct bau_control *bcp;
+
+	pqp = kmalloc_node((DEST_Q_SIZE + 1)
+			   * sizeof(struct bau_payload_queue_entry),
+			   GFP_KERNEL, node);
+	BUG_ON(!pqp);
+	pqp_malloc = pqp;
+
+	cp = (char *)pqp + 31;
+	pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5);
+
+	for_each_present_cpu(cpu) {
+		if (pnode != uv_cpu_to_pnode(cpu))
+			continue;
+		/* for every cpu on this pnode: */
+		bcp = &per_cpu(bau_control, cpu);
+		bcp->va_queue_first = pqp;
+		bcp->bau_msg_head = pqp;
+		bcp->va_queue_last = pqp + (DEST_Q_SIZE - 1);
+	}
+	/*
+	 * need the pnode of where the memory was really allocated
+	 */
+	pa = uv_gpa(pqp);
+	pn = pa >> uv_nshift;
+	uv_write_global_mmr64(pnode,
+			      UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST,
+			      ((unsigned long)pn << UV_PAYLOADQ_PNODE_SHIFT) |
+			      uv_physnodeaddr(pqp));
+	uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL,
+			      uv_physnodeaddr(pqp));
+	uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_LAST,
+			      (unsigned long)
+			      uv_physnodeaddr(pqp + (DEST_Q_SIZE - 1)));
+	/* in effect, all msg_type's are set to MSG_NOOP */
+	memset(pqp, 0, sizeof(struct bau_payload_queue_entry) * DEST_Q_SIZE);
+}
+
+/*
+ * Initialization of each UV hub's structures
+ */
+static void __init uv_init_uvhub(int uvhub, int vector)
+{
+	int node;
+	int pnode;
+	unsigned long apicid;
+
+	node = uvhub_to_first_node(uvhub);
+	pnode = uv_blade_to_pnode(uvhub);
+	uv_activation_descriptor_init(node, pnode);
+	uv_payload_queue_init(node, pnode);
+	/*
+	 * the below initialization can't be in firmware because the
+	 * messaging IRQ will be determined by the OS
+	 */
+	apicid = uvhub_to_first_apicid(uvhub);
+	uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG,
+				      ((apicid << 32) | vector));
+}
+
+/*
+ * We will set BAU_MISC_CONTROL with a timeout period.
+ * But the BIOS has set UVH_AGING_PRESCALE_SEL and UVH_TRANSACTION_TIMEOUT.
+ * So the destination timeout period has be be calculated from them.
+ */
+static int
+calculate_destination_timeout(void)
+{
+	unsigned long mmr_image;
+	int mult1;
+	int mult2;
+	int index;
+	int base;
+	int ret;
+	unsigned long ts_ns;
+
+	mult1 = UV_INTD_SOFT_ACK_TIMEOUT_PERIOD & BAU_MISC_CONTROL_MULT_MASK;
+	mmr_image = uv_read_local_mmr(UVH_AGING_PRESCALE_SEL);
+	index = (mmr_image >> BAU_URGENCY_7_SHIFT) & BAU_URGENCY_7_MASK;
+	mmr_image = uv_read_local_mmr(UVH_TRANSACTION_TIMEOUT);
+	mult2 = (mmr_image >> BAU_TRANS_SHIFT) & BAU_TRANS_MASK;
+	base = timeout_base_ns[index];
+	ts_ns = base * mult1 * mult2;
+	ret = ts_ns / 1000;
+	return ret;
+}
+
+/*
+ * initialize the bau_control structure for each cpu
+ */
+static void __init uv_init_per_cpu(int nuvhubs)
+{
+	int i;
+	int cpu;
+	int pnode;
+	int uvhub;
+	int have_hmaster;
+	short socket = 0;
+	unsigned short socket_mask;
+	unsigned char *uvhub_mask;
+	struct bau_control *bcp;
+	struct uvhub_desc *bdp;
+	struct socket_desc *sdp;
+	struct bau_control *hmaster = NULL;
+	struct bau_control *smaster = NULL;
+	struct socket_desc {
+		short num_cpus;
+		short cpu_number[16];
+	};
+	struct uvhub_desc {
+		unsigned short socket_mask;
+		short num_cpus;
+		short uvhub;
+		short pnode;
+		struct socket_desc socket[2];
+	};
+	struct uvhub_desc *uvhub_descs;
+
+	timeout_us = calculate_destination_timeout();
+
+	uvhub_descs = kmalloc(nuvhubs * sizeof(struct uvhub_desc), GFP_KERNEL);
+	memset(uvhub_descs, 0, nuvhubs * sizeof(struct uvhub_desc));
+	uvhub_mask = kzalloc((nuvhubs+7)/8, GFP_KERNEL);
+	for_each_present_cpu(cpu) {
+		bcp = &per_cpu(bau_control, cpu);
+		memset(bcp, 0, sizeof(struct bau_control));
+		pnode = uv_cpu_hub_info(cpu)->pnode;
+		uvhub = uv_cpu_hub_info(cpu)->numa_blade_id;
+		*(uvhub_mask + (uvhub/8)) |= (1 << (uvhub%8));
+		bdp = &uvhub_descs[uvhub];
+		bdp->num_cpus++;
+		bdp->uvhub = uvhub;
+		bdp->pnode = pnode;
+		/* kludge: 'assuming' one node per socket, and assuming that
+		   disabling a socket just leaves a gap in node numbers */
+		socket = (cpu_to_node(cpu) & 1);
+		bdp->socket_mask |= (1 << socket);
+		sdp = &bdp->socket[socket];
+		sdp->cpu_number[sdp->num_cpus] = cpu;
+		sdp->num_cpus++;
+	}
+	for (uvhub = 0; uvhub < nuvhubs; uvhub++) {
+		if (!(*(uvhub_mask + (uvhub/8)) & (1 << (uvhub%8))))
+			continue;
+		have_hmaster = 0;
+		bdp = &uvhub_descs[uvhub];
+		socket_mask = bdp->socket_mask;
+		socket = 0;
+		while (socket_mask) {
+			if (!(socket_mask & 1))
+				goto nextsocket;
+			sdp = &bdp->socket[socket];
+			for (i = 0; i < sdp->num_cpus; i++) {
+				cpu = sdp->cpu_number[i];
+				bcp = &per_cpu(bau_control, cpu);
+				bcp->cpu = cpu;
+				if (i == 0) {
+					smaster = bcp;
+					if (!have_hmaster) {
+						have_hmaster++;
+						hmaster = bcp;
+					}
+				}
+				bcp->cpus_in_uvhub = bdp->num_cpus;
+				bcp->cpus_in_socket = sdp->num_cpus;
+				bcp->socket_master = smaster;
+				bcp->uvhub = bdp->uvhub;
+				bcp->uvhub_master = hmaster;
+				bcp->uvhub_cpu = uv_cpu_hub_info(cpu)->
+						blade_processor_id;
+			}
+nextsocket:
+			socket++;
+			socket_mask = (socket_mask >> 1);
+		}
+	}
+	kfree(uvhub_descs);
+	kfree(uvhub_mask);
+	for_each_present_cpu(cpu) {
+		bcp = &per_cpu(bau_control, cpu);
+		bcp->baudisabled = 0;
+		bcp->statp = &per_cpu(ptcstats, cpu);
+		/* time interval to catch a hardware stay-busy bug */
+		bcp->timeout_interval = microsec_2_cycles(2*timeout_us);
+		bcp->max_bau_concurrent = max_bau_concurrent;
+		bcp->max_bau_concurrent_constant = max_bau_concurrent;
+		bcp->plugged_delay = plugged_delay;
+		bcp->plugsb4reset = plugsb4reset;
+		bcp->timeoutsb4reset = timeoutsb4reset;
+		bcp->ipi_reset_limit = ipi_reset_limit;
+		bcp->complete_threshold = complete_threshold;
+		bcp->congested_response_us = congested_response_us;
+		bcp->congested_reps = congested_reps;
+		bcp->congested_period = congested_period;
+	}
+}
+
+/*
+ * Initialization of BAU-related structures
+ */
+static int __init uv_bau_init(void)
+{
+	int uvhub;
+	int pnode;
+	int nuvhubs;
+	int cur_cpu;
+	int vector;
+	unsigned long mmr;
+
+	if (!is_uv_system())
+		return 0;
+
+	if (nobau)
+		return 0;
+
+	for_each_possible_cpu(cur_cpu)
+		zalloc_cpumask_var_node(&per_cpu(uv_flush_tlb_mask, cur_cpu),
+				       GFP_KERNEL, cpu_to_node(cur_cpu));
+
+	uv_nshift = uv_hub_info->m_val;
+	uv_mmask = (1UL << uv_hub_info->m_val) - 1;
+	nuvhubs = uv_num_possible_blades();
+	spin_lock_init(&disable_lock);
+	congested_cycles = microsec_2_cycles(congested_response_us);
+
+	uv_init_per_cpu(nuvhubs);
+
+	uv_partition_base_pnode = 0x7fffffff;
+	for (uvhub = 0; uvhub < nuvhubs; uvhub++)
+		if (uv_blade_nr_possible_cpus(uvhub) &&
+			(uv_blade_to_pnode(uvhub) < uv_partition_base_pnode))
+			uv_partition_base_pnode = uv_blade_to_pnode(uvhub);
+
+	vector = UV_BAU_MESSAGE;
+	for_each_possible_blade(uvhub)
+		if (uv_blade_nr_possible_cpus(uvhub))
+			uv_init_uvhub(uvhub, vector);
+
+	uv_enable_timeouts();
+	alloc_intr_gate(vector, uv_bau_message_intr1);
+
+	for_each_possible_blade(uvhub) {
+		if (uv_blade_nr_possible_cpus(uvhub)) {
+			pnode = uv_blade_to_pnode(uvhub);
+			/* INIT the bau */
+			uv_write_global_mmr64(pnode,
+					UVH_LB_BAU_SB_ACTIVATION_CONTROL,
+					((unsigned long)1 << 63));
+			mmr = 1; /* should be 1 to broadcast to both sockets */
+			uv_write_global_mmr64(pnode, UVH_BAU_DATA_BROADCAST,
+						mmr);
+		}
+	}
+
+	return 0;
+}
+core_initcall(uv_bau_init);
+fs_initcall(uv_ptc_init);
diff --git a/arch/x86/platform/uv/uv_irq.c b/arch/x86/platform/uv/uv_irq.c
new file mode 100644
index 00000000000..7b24460917d
--- /dev/null
+++ b/arch/x86/platform/uv/uv_irq.c
@@ -0,0 +1,285 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * SGI UV IRQ functions
+ *
+ * Copyright (C) 2008 Silicon Graphics, Inc. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/rbtree.h>
+#include <linux/slab.h>
+#include <linux/irq.h>
+
+#include <asm/apic.h>
+#include <asm/uv/uv_irq.h>
+#include <asm/uv/uv_hub.h>
+
+/* MMR offset and pnode of hub sourcing interrupts for a given irq */
+struct uv_irq_2_mmr_pnode{
+	struct rb_node		list;
+	unsigned long		offset;
+	int			pnode;
+	int			irq;
+};
+
+static spinlock_t		uv_irq_lock;
+static struct rb_root		uv_irq_root;
+
+static int uv_set_irq_affinity(struct irq_data *, const struct cpumask *, bool);
+
+static void uv_noop(struct irq_data *data) { }
+
+static void uv_ack_apic(struct irq_data *data)
+{
+	ack_APIC_irq();
+}
+
+static struct irq_chip uv_irq_chip = {
+	.name			= "UV-CORE",
+	.irq_mask		= uv_noop,
+	.irq_unmask		= uv_noop,
+	.irq_eoi		= uv_ack_apic,
+	.irq_set_affinity	= uv_set_irq_affinity,
+};
+
+/*
+ * Add offset and pnode information of the hub sourcing interrupts to the
+ * rb tree for a specific irq.
+ */
+static int uv_set_irq_2_mmr_info(int irq, unsigned long offset, unsigned blade)
+{
+	struct rb_node **link = &uv_irq_root.rb_node;
+	struct rb_node *parent = NULL;
+	struct uv_irq_2_mmr_pnode *n;
+	struct uv_irq_2_mmr_pnode *e;
+	unsigned long irqflags;
+
+	n = kmalloc_node(sizeof(struct uv_irq_2_mmr_pnode), GFP_KERNEL,
+				uv_blade_to_memory_nid(blade));
+	if (!n)
+		return -ENOMEM;
+
+	n->irq = irq;
+	n->offset = offset;
+	n->pnode = uv_blade_to_pnode(blade);
+	spin_lock_irqsave(&uv_irq_lock, irqflags);
+	/* Find the right place in the rbtree: */
+	while (*link) {
+		parent = *link;
+		e = rb_entry(parent, struct uv_irq_2_mmr_pnode, list);
+
+		if (unlikely(irq == e->irq)) {
+			/* irq entry exists */
+			e->pnode = uv_blade_to_pnode(blade);
+			e->offset = offset;
+			spin_unlock_irqrestore(&uv_irq_lock, irqflags);
+			kfree(n);
+			return 0;
+		}
+
+		if (irq < e->irq)
+			link = &(*link)->rb_left;
+		else
+			link = &(*link)->rb_right;
+	}
+
+	/* Insert the node into the rbtree. */
+	rb_link_node(&n->list, parent, link);
+	rb_insert_color(&n->list, &uv_irq_root);
+
+	spin_unlock_irqrestore(&uv_irq_lock, irqflags);
+	return 0;
+}
+
+/* Retrieve offset and pnode information from the rb tree for a specific irq */
+int uv_irq_2_mmr_info(int irq, unsigned long *offset, int *pnode)
+{
+	struct uv_irq_2_mmr_pnode *e;
+	struct rb_node *n;
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&uv_irq_lock, irqflags);
+	n = uv_irq_root.rb_node;
+	while (n) {
+		e = rb_entry(n, struct uv_irq_2_mmr_pnode, list);
+
+		if (e->irq == irq) {
+			*offset = e->offset;
+			*pnode = e->pnode;
+			spin_unlock_irqrestore(&uv_irq_lock, irqflags);
+			return 0;
+		}
+
+		if (irq < e->irq)
+			n = n->rb_left;
+		else
+			n = n->rb_right;
+	}
+	spin_unlock_irqrestore(&uv_irq_lock, irqflags);
+	return -1;
+}
+
+/*
+ * Re-target the irq to the specified CPU and enable the specified MMR located
+ * on the specified blade to allow the sending of MSIs to the specified CPU.
+ */
+static int
+arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
+		       unsigned long mmr_offset, int limit)
+{
+	const struct cpumask *eligible_cpu = cpumask_of(cpu);
+	struct irq_cfg *cfg = get_irq_chip_data(irq);
+	unsigned long mmr_value;
+	struct uv_IO_APIC_route_entry *entry;
+	int mmr_pnode, err;
+
+	BUILD_BUG_ON(sizeof(struct uv_IO_APIC_route_entry) !=
+			sizeof(unsigned long));
+
+	err = assign_irq_vector(irq, cfg, eligible_cpu);
+	if (err != 0)
+		return err;
+
+	if (limit == UV_AFFINITY_CPU)
+		irq_set_status_flags(irq, IRQ_NO_BALANCING);
+	else
+		irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
+
+	set_irq_chip_and_handler_name(irq, &uv_irq_chip, handle_percpu_irq,
+				      irq_name);
+
+	mmr_value = 0;
+	entry = (struct uv_IO_APIC_route_entry *)&mmr_value;
+	entry->vector		= cfg->vector;
+	entry->delivery_mode	= apic->irq_delivery_mode;
+	entry->dest_mode	= apic->irq_dest_mode;
+	entry->polarity		= 0;
+	entry->trigger		= 0;
+	entry->mask		= 0;
+	entry->dest		= apic->cpu_mask_to_apicid(eligible_cpu);
+
+	mmr_pnode = uv_blade_to_pnode(mmr_blade);
+	uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value);
+
+	if (cfg->move_in_progress)
+		send_cleanup_vector(cfg);
+
+	return irq;
+}
+
+/*
+ * Disable the specified MMR located on the specified blade so that MSIs are
+ * longer allowed to be sent.
+ */
+static void arch_disable_uv_irq(int mmr_pnode, unsigned long mmr_offset)
+{
+	unsigned long mmr_value;
+	struct uv_IO_APIC_route_entry *entry;
+
+	BUILD_BUG_ON(sizeof(struct uv_IO_APIC_route_entry) !=
+			sizeof(unsigned long));
+
+	mmr_value = 0;
+	entry = (struct uv_IO_APIC_route_entry *)&mmr_value;
+	entry->mask = 1;
+
+	uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value);
+}
+
+static int
+uv_set_irq_affinity(struct irq_data *data, const struct cpumask *mask,
+		    bool force)
+{
+	struct irq_cfg *cfg = data->chip_data;
+	unsigned int dest;
+	unsigned long mmr_value, mmr_offset;
+	struct uv_IO_APIC_route_entry *entry;
+	int mmr_pnode;
+
+	if (__ioapic_set_affinity(data, mask, &dest))
+		return -1;
+
+	mmr_value = 0;
+	entry = (struct uv_IO_APIC_route_entry *)&mmr_value;
+
+	entry->vector		= cfg->vector;
+	entry->delivery_mode	= apic->irq_delivery_mode;
+	entry->dest_mode	= apic->irq_dest_mode;
+	entry->polarity		= 0;
+	entry->trigger		= 0;
+	entry->mask		= 0;
+	entry->dest		= dest;
+
+	/* Get previously stored MMR and pnode of hub sourcing interrupts */
+	if (uv_irq_2_mmr_info(data->irq, &mmr_offset, &mmr_pnode))
+		return -1;
+
+	uv_write_global_mmr64(mmr_pnode, mmr_offset, mmr_value);
+
+	if (cfg->move_in_progress)
+		send_cleanup_vector(cfg);
+
+	return 0;
+}
+
+/*
+ * Set up a mapping of an available irq and vector, and enable the specified
+ * MMR that defines the MSI that is to be sent to the specified CPU when an
+ * interrupt is raised.
+ */
+int uv_setup_irq(char *irq_name, int cpu, int mmr_blade,
+		 unsigned long mmr_offset, int limit)
+{
+	int irq, ret;
+
+	irq = create_irq_nr(NR_IRQS_LEGACY, uv_blade_to_memory_nid(mmr_blade));
+
+	if (irq <= 0)
+		return -EBUSY;
+
+	ret = arch_enable_uv_irq(irq_name, irq, cpu, mmr_blade, mmr_offset,
+		limit);
+	if (ret == irq)
+		uv_set_irq_2_mmr_info(irq, mmr_offset, mmr_blade);
+	else
+		destroy_irq(irq);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(uv_setup_irq);
+
+/*
+ * Tear down a mapping of an irq and vector, and disable the specified MMR that
+ * defined the MSI that was to be sent to the specified CPU when an interrupt
+ * was raised.
+ *
+ * Set mmr_blade and mmr_offset to what was passed in on uv_setup_irq().
+ */
+void uv_teardown_irq(unsigned int irq)
+{
+	struct uv_irq_2_mmr_pnode *e;
+	struct rb_node *n;
+	unsigned long irqflags;
+
+	spin_lock_irqsave(&uv_irq_lock, irqflags);
+	n = uv_irq_root.rb_node;
+	while (n) {
+		e = rb_entry(n, struct uv_irq_2_mmr_pnode, list);
+		if (e->irq == irq) {
+			arch_disable_uv_irq(e->pnode, e->offset);
+			rb_erase(n, &uv_irq_root);
+			kfree(e);
+			break;
+		}
+		if (irq < e->irq)
+			n = n->rb_left;
+		else
+			n = n->rb_right;
+	}
+	spin_unlock_irqrestore(&uv_irq_lock, irqflags);
+	destroy_irq(irq);
+}
+EXPORT_SYMBOL_GPL(uv_teardown_irq);
diff --git a/arch/x86/platform/uv/uv_sysfs.c b/arch/x86/platform/uv/uv_sysfs.c
new file mode 100644
index 00000000000..309c70fb775
--- /dev/null
+++ b/arch/x86/platform/uv/uv_sysfs.c
@@ -0,0 +1,76 @@
+/*
+ * This file supports the /sys/firmware/sgi_uv interfaces for SGI UV.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ *
+ *  Copyright (c) 2008 Silicon Graphics, Inc.  All Rights Reserved.
+ *  Copyright (c) Russ Anderson
+ */
+
+#include <linux/sysdev.h>
+#include <asm/uv/bios.h>
+#include <asm/uv/uv.h>
+
+struct kobject *sgi_uv_kobj;
+
+static ssize_t partition_id_show(struct kobject *kobj,
+			struct kobj_attribute *attr, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%ld\n", sn_partition_id);
+}
+
+static ssize_t coherence_id_show(struct kobject *kobj,
+			struct kobj_attribute *attr, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%ld\n", partition_coherence_id());
+}
+
+static struct kobj_attribute partition_id_attr =
+	__ATTR(partition_id, S_IRUGO, partition_id_show, NULL);
+
+static struct kobj_attribute coherence_id_attr =
+	__ATTR(coherence_id, S_IRUGO, coherence_id_show, NULL);
+
+
+static int __init sgi_uv_sysfs_init(void)
+{
+	unsigned long ret;
+
+	if (!is_uv_system())
+		return -ENODEV;
+
+	if (!sgi_uv_kobj)
+		sgi_uv_kobj = kobject_create_and_add("sgi_uv", firmware_kobj);
+	if (!sgi_uv_kobj) {
+		printk(KERN_WARNING "kobject_create_and_add sgi_uv failed\n");
+		return -EINVAL;
+	}
+
+	ret = sysfs_create_file(sgi_uv_kobj, &partition_id_attr.attr);
+	if (ret) {
+		printk(KERN_WARNING "sysfs_create_file partition_id failed\n");
+		return ret;
+	}
+
+	ret = sysfs_create_file(sgi_uv_kobj, &coherence_id_attr.attr);
+	if (ret) {
+		printk(KERN_WARNING "sysfs_create_file coherence_id failed\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+device_initcall(sgi_uv_sysfs_init);
diff --git a/arch/x86/platform/uv/uv_time.c b/arch/x86/platform/uv/uv_time.c
new file mode 100644
index 00000000000..56e421bc379
--- /dev/null
+++ b/arch/x86/platform/uv/uv_time.c
@@ -0,0 +1,423 @@
+/*
+ * SGI RTC clock/timer routines.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ *
+ *  Copyright (c) 2009 Silicon Graphics, Inc.  All Rights Reserved.
+ *  Copyright (c) Dimitri Sivanich
+ */
+#include <linux/clockchips.h>
+#include <linux/slab.h>
+
+#include <asm/uv/uv_mmrs.h>
+#include <asm/uv/uv_hub.h>
+#include <asm/uv/bios.h>
+#include <asm/uv/uv.h>
+#include <asm/apic.h>
+#include <asm/cpu.h>
+
+#define RTC_NAME		"sgi_rtc"
+
+static cycle_t uv_read_rtc(struct clocksource *cs);
+static int uv_rtc_next_event(unsigned long, struct clock_event_device *);
+static void uv_rtc_timer_setup(enum clock_event_mode,
+				struct clock_event_device *);
+
+static struct clocksource clocksource_uv = {
+	.name		= RTC_NAME,
+	.rating		= 400,
+	.read		= uv_read_rtc,
+	.mask		= (cycle_t)UVH_RTC_REAL_TIME_CLOCK_MASK,
+	.shift		= 10,
+	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
+};
+
+static struct clock_event_device clock_event_device_uv = {
+	.name		= RTC_NAME,
+	.features	= CLOCK_EVT_FEAT_ONESHOT,
+	.shift		= 20,
+	.rating		= 400,
+	.irq		= -1,
+	.set_next_event	= uv_rtc_next_event,
+	.set_mode	= uv_rtc_timer_setup,
+	.event_handler	= NULL,
+};
+
+static DEFINE_PER_CPU(struct clock_event_device, cpu_ced);
+
+/* There is one of these allocated per node */
+struct uv_rtc_timer_head {
+	spinlock_t	lock;
+	/* next cpu waiting for timer, local node relative: */
+	int		next_cpu;
+	/* number of cpus on this node: */
+	int		ncpus;
+	struct {
+		int	lcpu;		/* systemwide logical cpu number */
+		u64	expires;	/* next timer expiration for this cpu */
+	} cpu[1];
+};
+
+/*
+ * Access to uv_rtc_timer_head via blade id.
+ */
+static struct uv_rtc_timer_head		**blade_info __read_mostly;
+
+static int				uv_rtc_evt_enable;
+
+/*
+ * Hardware interface routines
+ */
+
+/* Send IPIs to another node */
+static void uv_rtc_send_IPI(int cpu)
+{
+	unsigned long apicid, val;
+	int pnode;
+
+	apicid = cpu_physical_id(cpu);
+	pnode = uv_apicid_to_pnode(apicid);
+	val = (1UL << UVH_IPI_INT_SEND_SHFT) |
+	      (apicid << UVH_IPI_INT_APIC_ID_SHFT) |
+	      (X86_PLATFORM_IPI_VECTOR << UVH_IPI_INT_VECTOR_SHFT);
+
+	uv_write_global_mmr64(pnode, UVH_IPI_INT, val);
+}
+
+/* Check for an RTC interrupt pending */
+static int uv_intr_pending(int pnode)
+{
+	return uv_read_global_mmr64(pnode, UVH_EVENT_OCCURRED0) &
+		UVH_EVENT_OCCURRED0_RTC1_MASK;
+}
+
+/* Setup interrupt and return non-zero if early expiration occurred. */
+static int uv_setup_intr(int cpu, u64 expires)
+{
+	u64 val;
+	int pnode = uv_cpu_to_pnode(cpu);
+
+	uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG,
+		UVH_RTC1_INT_CONFIG_M_MASK);
+	uv_write_global_mmr64(pnode, UVH_INT_CMPB, -1L);
+
+	uv_write_global_mmr64(pnode, UVH_EVENT_OCCURRED0_ALIAS,
+		UVH_EVENT_OCCURRED0_RTC1_MASK);
+
+	val = (X86_PLATFORM_IPI_VECTOR << UVH_RTC1_INT_CONFIG_VECTOR_SHFT) |
+		((u64)cpu_physical_id(cpu) << UVH_RTC1_INT_CONFIG_APIC_ID_SHFT);
+
+	/* Set configuration */
+	uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG, val);
+	/* Initialize comparator value */
+	uv_write_global_mmr64(pnode, UVH_INT_CMPB, expires);
+
+	if (uv_read_rtc(NULL) <= expires)
+		return 0;
+
+	return !uv_intr_pending(pnode);
+}
+
+/*
+ * Per-cpu timer tracking routines
+ */
+
+static __init void uv_rtc_deallocate_timers(void)
+{
+	int bid;
+
+	for_each_possible_blade(bid) {
+		kfree(blade_info[bid]);
+	}
+	kfree(blade_info);
+}
+
+/* Allocate per-node list of cpu timer expiration times. */
+static __init int uv_rtc_allocate_timers(void)
+{
+	int cpu;
+
+	blade_info = kmalloc(uv_possible_blades * sizeof(void *), GFP_KERNEL);
+	if (!blade_info)
+		return -ENOMEM;
+	memset(blade_info, 0, uv_possible_blades * sizeof(void *));
+
+	for_each_present_cpu(cpu) {
+		int nid = cpu_to_node(cpu);
+		int bid = uv_cpu_to_blade_id(cpu);
+		int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+		struct uv_rtc_timer_head *head = blade_info[bid];
+
+		if (!head) {
+			head = kmalloc_node(sizeof(struct uv_rtc_timer_head) +
+				(uv_blade_nr_possible_cpus(bid) *
+					2 * sizeof(u64)),
+				GFP_KERNEL, nid);
+			if (!head) {
+				uv_rtc_deallocate_timers();
+				return -ENOMEM;
+			}
+			spin_lock_init(&head->lock);
+			head->ncpus = uv_blade_nr_possible_cpus(bid);
+			head->next_cpu = -1;
+			blade_info[bid] = head;
+		}
+
+		head->cpu[bcpu].lcpu = cpu;
+		head->cpu[bcpu].expires = ULLONG_MAX;
+	}
+
+	return 0;
+}
+
+/* Find and set the next expiring timer.  */
+static void uv_rtc_find_next_timer(struct uv_rtc_timer_head *head, int pnode)
+{
+	u64 lowest = ULLONG_MAX;
+	int c, bcpu = -1;
+
+	head->next_cpu = -1;
+	for (c = 0; c < head->ncpus; c++) {
+		u64 exp = head->cpu[c].expires;
+		if (exp < lowest) {
+			bcpu = c;
+			lowest = exp;
+		}
+	}
+	if (bcpu >= 0) {
+		head->next_cpu = bcpu;
+		c = head->cpu[bcpu].lcpu;
+		if (uv_setup_intr(c, lowest))
+			/* If we didn't set it up in time, trigger */
+			uv_rtc_send_IPI(c);
+	} else {
+		uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG,
+			UVH_RTC1_INT_CONFIG_M_MASK);
+	}
+}
+
+/*
+ * Set expiration time for current cpu.
+ *
+ * Returns 1 if we missed the expiration time.
+ */
+static int uv_rtc_set_timer(int cpu, u64 expires)
+{
+	int pnode = uv_cpu_to_pnode(cpu);
+	int bid = uv_cpu_to_blade_id(cpu);
+	struct uv_rtc_timer_head *head = blade_info[bid];
+	int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+	u64 *t = &head->cpu[bcpu].expires;
+	unsigned long flags;
+	int next_cpu;
+
+	spin_lock_irqsave(&head->lock, flags);
+
+	next_cpu = head->next_cpu;
+	*t = expires;
+
+	/* Will this one be next to go off? */
+	if (next_cpu < 0 || bcpu == next_cpu ||
+			expires < head->cpu[next_cpu].expires) {
+		head->next_cpu = bcpu;
+		if (uv_setup_intr(cpu, expires)) {
+			*t = ULLONG_MAX;
+			uv_rtc_find_next_timer(head, pnode);
+			spin_unlock_irqrestore(&head->lock, flags);
+			return -ETIME;
+		}
+	}
+
+	spin_unlock_irqrestore(&head->lock, flags);
+	return 0;
+}
+
+/*
+ * Unset expiration time for current cpu.
+ *
+ * Returns 1 if this timer was pending.
+ */
+static int uv_rtc_unset_timer(int cpu, int force)
+{
+	int pnode = uv_cpu_to_pnode(cpu);
+	int bid = uv_cpu_to_blade_id(cpu);
+	struct uv_rtc_timer_head *head = blade_info[bid];
+	int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+	u64 *t = &head->cpu[bcpu].expires;
+	unsigned long flags;
+	int rc = 0;
+
+	spin_lock_irqsave(&head->lock, flags);
+
+	if ((head->next_cpu == bcpu && uv_read_rtc(NULL) >= *t) || force)
+		rc = 1;
+
+	if (rc) {
+		*t = ULLONG_MAX;
+		/* Was the hardware setup for this timer? */
+		if (head->next_cpu == bcpu)
+			uv_rtc_find_next_timer(head, pnode);
+	}
+
+	spin_unlock_irqrestore(&head->lock, flags);
+
+	return rc;
+}
+
+
+/*
+ * Kernel interface routines.
+ */
+
+/*
+ * Read the RTC.
+ *
+ * Starting with HUB rev 2.0, the UV RTC register is replicated across all
+ * cachelines of it's own page.  This allows faster simultaneous reads
+ * from a given socket.
+ */
+static cycle_t uv_read_rtc(struct clocksource *cs)
+{
+	unsigned long offset;
+
+	if (uv_get_min_hub_revision_id() == 1)
+		offset = 0;
+	else
+		offset = (uv_blade_processor_id() * L1_CACHE_BYTES) % PAGE_SIZE;
+
+	return (cycle_t)uv_read_local_mmr(UVH_RTC | offset);
+}
+
+/*
+ * Program the next event, relative to now
+ */
+static int uv_rtc_next_event(unsigned long delta,
+			     struct clock_event_device *ced)
+{
+	int ced_cpu = cpumask_first(ced->cpumask);
+
+	return uv_rtc_set_timer(ced_cpu, delta + uv_read_rtc(NULL));
+}
+
+/*
+ * Setup the RTC timer in oneshot mode
+ */
+static void uv_rtc_timer_setup(enum clock_event_mode mode,
+			       struct clock_event_device *evt)
+{
+	int ced_cpu = cpumask_first(evt->cpumask);
+
+	switch (mode) {
+	case CLOCK_EVT_MODE_PERIODIC:
+	case CLOCK_EVT_MODE_ONESHOT:
+	case CLOCK_EVT_MODE_RESUME:
+		/* Nothing to do here yet */
+		break;
+	case CLOCK_EVT_MODE_UNUSED:
+	case CLOCK_EVT_MODE_SHUTDOWN:
+		uv_rtc_unset_timer(ced_cpu, 1);
+		break;
+	}
+}
+
+static void uv_rtc_interrupt(void)
+{
+	int cpu = smp_processor_id();
+	struct clock_event_device *ced = &per_cpu(cpu_ced, cpu);
+
+	if (!ced || !ced->event_handler)
+		return;
+
+	if (uv_rtc_unset_timer(cpu, 0) != 1)
+		return;
+
+	ced->event_handler(ced);
+}
+
+static int __init uv_enable_evt_rtc(char *str)
+{
+	uv_rtc_evt_enable = 1;
+
+	return 1;
+}
+__setup("uvrtcevt", uv_enable_evt_rtc);
+
+static __init void uv_rtc_register_clockevents(struct work_struct *dummy)
+{
+	struct clock_event_device *ced = &__get_cpu_var(cpu_ced);
+
+	*ced = clock_event_device_uv;
+	ced->cpumask = cpumask_of(smp_processor_id());
+	clockevents_register_device(ced);
+}
+
+static __init int uv_rtc_setup_clock(void)
+{
+	int rc;
+
+	if (!is_uv_system())
+		return -ENODEV;
+
+	clocksource_uv.mult = clocksource_hz2mult(sn_rtc_cycles_per_second,
+				clocksource_uv.shift);
+
+	/* If single blade, prefer tsc */
+	if (uv_num_possible_blades() == 1)
+		clocksource_uv.rating = 250;
+
+	rc = clocksource_register(&clocksource_uv);
+	if (rc)
+		printk(KERN_INFO "UV RTC clocksource failed rc %d\n", rc);
+	else
+		printk(KERN_INFO "UV RTC clocksource registered freq %lu MHz\n",
+			sn_rtc_cycles_per_second/(unsigned long)1E6);
+
+	if (rc || !uv_rtc_evt_enable || x86_platform_ipi_callback)
+		return rc;
+
+	/* Setup and register clockevents */
+	rc = uv_rtc_allocate_timers();
+	if (rc)
+		goto error;
+
+	x86_platform_ipi_callback = uv_rtc_interrupt;
+
+	clock_event_device_uv.mult = div_sc(sn_rtc_cycles_per_second,
+				NSEC_PER_SEC, clock_event_device_uv.shift);
+
+	clock_event_device_uv.min_delta_ns = NSEC_PER_SEC /
+						sn_rtc_cycles_per_second;
+
+	clock_event_device_uv.max_delta_ns = clocksource_uv.mask *
+				(NSEC_PER_SEC / sn_rtc_cycles_per_second);
+
+	rc = schedule_on_each_cpu(uv_rtc_register_clockevents);
+	if (rc) {
+		x86_platform_ipi_callback = NULL;
+		uv_rtc_deallocate_timers();
+		goto error;
+	}
+
+	printk(KERN_INFO "UV RTC clockevents registered\n");
+
+	return 0;
+
+error:
+	clocksource_unregister(&clocksource_uv);
+	printk(KERN_INFO "UV RTC clockevents failed rc %d\n", rc);
+
+	return rc;
+}
+arch_initcall(uv_rtc_setup_clock);
diff --git a/arch/x86/platform/visws/Makefile b/arch/x86/platform/visws/Makefile
new file mode 100644
index 00000000000..91bc17ab2fd
--- /dev/null
+++ b/arch/x86/platform/visws/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_X86_VISWS)	+= visws_quirks.o
diff --git a/arch/x86/platform/visws/visws_quirks.c b/arch/x86/platform/visws/visws_quirks.c
new file mode 100644
index 00000000000..3371bd053b8
--- /dev/null
+++ b/arch/x86/platform/visws/visws_quirks.c
@@ -0,0 +1,614 @@
+/*
+ *  SGI Visual Workstation support and quirks, unmaintained.
+ *
+ *  Split out from setup.c by davej@suse.de
+ *
+ *	Copyright (C) 1999 Bent Hagemark, Ingo Molnar
+ *
+ *  SGI Visual Workstation interrupt controller
+ *
+ *  The Cobalt system ASIC in the Visual Workstation contains a "Cobalt" APIC
+ *  which serves as the main interrupt controller in the system.  Non-legacy
+ *  hardware in the system uses this controller directly.  Legacy devices
+ *  are connected to the PIIX4 which in turn has its 8259(s) connected to
+ *  a of the Cobalt APIC entry.
+ *
+ *  09/02/2000 - Updated for 2.4 by jbarnes@sgi.com
+ *
+ *  25/11/2002 - Updated for 2.5 by Andrey Panin <pazke@orbita1.ru>
+ */
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+
+#include <asm/visws/cobalt.h>
+#include <asm/visws/piix4.h>
+#include <asm/io_apic.h>
+#include <asm/fixmap.h>
+#include <asm/reboot.h>
+#include <asm/setup.h>
+#include <asm/apic.h>
+#include <asm/e820.h>
+#include <asm/time.h>
+#include <asm/io.h>
+
+#include <linux/kernel_stat.h>
+
+#include <asm/i8259.h>
+#include <asm/irq_vectors.h>
+#include <asm/visws/lithium.h>
+
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+
+extern int no_broadcast;
+
+char visws_board_type	= -1;
+char visws_board_rev	= -1;
+
+static void __init visws_time_init(void)
+{
+	printk(KERN_INFO "Starting Cobalt Timer system clock\n");
+
+	/* Set the countdown value */
+	co_cpu_write(CO_CPU_TIMEVAL, CO_TIME_HZ/HZ);
+
+	/* Start the timer */
+	co_cpu_write(CO_CPU_CTRL, co_cpu_read(CO_CPU_CTRL) | CO_CTRL_TIMERUN);
+
+	/* Enable (unmask) the timer interrupt */
+	co_cpu_write(CO_CPU_CTRL, co_cpu_read(CO_CPU_CTRL) & ~CO_CTRL_TIMEMASK);
+
+	setup_default_timer_irq();
+}
+
+/* Replaces the default init_ISA_irqs in the generic setup */
+static void __init visws_pre_intr_init(void);
+
+/* Quirk for machine specific memory setup. */
+
+#define MB (1024 * 1024)
+
+unsigned long sgivwfb_mem_phys;
+unsigned long sgivwfb_mem_size;
+EXPORT_SYMBOL(sgivwfb_mem_phys);
+EXPORT_SYMBOL(sgivwfb_mem_size);
+
+long long mem_size __initdata = 0;
+
+static char * __init visws_memory_setup(void)
+{
+	long long gfx_mem_size = 8 * MB;
+
+	mem_size = boot_params.alt_mem_k;
+
+	if (!mem_size) {
+		printk(KERN_WARNING "Bootloader didn't set memory size, upgrade it !\n");
+		mem_size = 128 * MB;
+	}
+
+	/*
+	 * this hardcodes the graphics memory to 8 MB
+	 * it really should be sized dynamically (or at least
+	 * set as a boot param)
+	 */
+	if (!sgivwfb_mem_size) {
+		printk(KERN_WARNING "Defaulting to 8 MB framebuffer size\n");
+		sgivwfb_mem_size = 8 * MB;
+	}
+
+	/*
+	 * Trim to nearest MB
+	 */
+	sgivwfb_mem_size &= ~((1 << 20) - 1);
+	sgivwfb_mem_phys = mem_size - gfx_mem_size;
+
+	e820_add_region(0, LOWMEMSIZE(), E820_RAM);
+	e820_add_region(HIGH_MEMORY, mem_size - sgivwfb_mem_size - HIGH_MEMORY, E820_RAM);
+	e820_add_region(sgivwfb_mem_phys, sgivwfb_mem_size, E820_RESERVED);
+
+	return "PROM";
+}
+
+static void visws_machine_emergency_restart(void)
+{
+	/*
+	 * Visual Workstations restart after this
+	 * register is poked on the PIIX4
+	 */
+	outb(PIIX4_RESET_VAL, PIIX4_RESET_PORT);
+}
+
+static void visws_machine_power_off(void)
+{
+	unsigned short pm_status;
+/*	extern unsigned int pci_bus0; */
+
+	while ((pm_status = inw(PMSTS_PORT)) & 0x100)
+		outw(pm_status, PMSTS_PORT);
+
+	outw(PM_SUSPEND_ENABLE, PMCNTRL_PORT);
+
+	mdelay(10);
+
+#define PCI_CONF1_ADDRESS(bus, devfn, reg) \
+	(0x80000000 | (bus << 16) | (devfn << 8) | (reg & ~3))
+
+/*	outl(PCI_CONF1_ADDRESS(pci_bus0, SPECIAL_DEV, SPECIAL_REG), 0xCF8); */
+	outl(PIIX_SPECIAL_STOP, 0xCFC);
+}
+
+static void __init visws_get_smp_config(unsigned int early)
+{
+}
+
+/*
+ * The Visual Workstation is Intel MP compliant in the hardware
+ * sense, but it doesn't have a BIOS(-configuration table).
+ * No problem for Linux.
+ */
+
+static void __init MP_processor_info(struct mpc_cpu *m)
+{
+	int ver, logical_apicid;
+	physid_mask_t apic_cpus;
+
+	if (!(m->cpuflag & CPU_ENABLED))
+		return;
+
+	logical_apicid = m->apicid;
+	printk(KERN_INFO "%sCPU #%d %u:%u APIC version %d\n",
+	       m->cpuflag & CPU_BOOTPROCESSOR ? "Bootup " : "",
+	       m->apicid, (m->cpufeature & CPU_FAMILY_MASK) >> 8,
+	       (m->cpufeature & CPU_MODEL_MASK) >> 4, m->apicver);
+
+	if (m->cpuflag & CPU_BOOTPROCESSOR)
+		boot_cpu_physical_apicid = m->apicid;
+
+	ver = m->apicver;
+	if ((ver >= 0x14 && m->apicid >= 0xff) || m->apicid >= 0xf) {
+		printk(KERN_ERR "Processor #%d INVALID. (Max ID: %d).\n",
+			m->apicid, MAX_APICS);
+		return;
+	}
+
+	apic->apicid_to_cpu_present(m->apicid, &apic_cpus);
+	physids_or(phys_cpu_present_map, phys_cpu_present_map, apic_cpus);
+	/*
+	 * Validate version
+	 */
+	if (ver == 0x0) {
+		printk(KERN_ERR "BIOS bug, APIC version is 0 for CPU#%d! "
+			"fixing up to 0x10. (tell your hw vendor)\n",
+			m->apicid);
+		ver = 0x10;
+	}
+	apic_version[m->apicid] = ver;
+}
+
+static void __init visws_find_smp_config(void)
+{
+	struct mpc_cpu *mp = phys_to_virt(CO_CPU_TAB_PHYS);
+	unsigned short ncpus = readw(phys_to_virt(CO_CPU_NUM_PHYS));
+
+	if (ncpus > CO_CPU_MAX) {
+		printk(KERN_WARNING "find_visws_smp: got cpu count of %d at %p\n",
+			ncpus, mp);
+
+		ncpus = CO_CPU_MAX;
+	}
+
+	if (ncpus > setup_max_cpus)
+		ncpus = setup_max_cpus;
+
+#ifdef CONFIG_X86_LOCAL_APIC
+	smp_found_config = 1;
+#endif
+	while (ncpus--)
+		MP_processor_info(mp++);
+
+	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
+}
+
+static void visws_trap_init(void);
+
+void __init visws_early_detect(void)
+{
+	int raw;
+
+	visws_board_type = (char)(inb_p(PIIX_GPI_BD_REG) & PIIX_GPI_BD_REG)
+							 >> PIIX_GPI_BD_SHIFT;
+
+	if (visws_board_type < 0)
+		return;
+
+	/*
+	 * Override the default platform setup functions
+	 */
+	x86_init.resources.memory_setup = visws_memory_setup;
+	x86_init.mpparse.get_smp_config = visws_get_smp_config;
+	x86_init.mpparse.find_smp_config = visws_find_smp_config;
+	x86_init.irqs.pre_vector_init = visws_pre_intr_init;
+	x86_init.irqs.trap_init = visws_trap_init;
+	x86_init.timers.timer_init = visws_time_init;
+	x86_init.pci.init = pci_visws_init;
+	x86_init.pci.init_irq = x86_init_noop;
+
+	/*
+	 * Install reboot quirks:
+	 */
+	pm_power_off			= visws_machine_power_off;
+	machine_ops.emergency_restart	= visws_machine_emergency_restart;
+
+	/*
+	 * Do not use broadcast IPIs:
+	 */
+	no_broadcast = 0;
+
+#ifdef CONFIG_X86_IO_APIC
+	/*
+	 * Turn off IO-APIC detection and initialization:
+	 */
+	skip_ioapic_setup		= 1;
+#endif
+
+	/*
+	 * Get Board rev.
+	 * First, we have to initialize the 307 part to allow us access
+	 * to the GPIO registers.  Let's map them at 0x0fc0 which is right
+	 * after the PIIX4 PM section.
+	 */
+	outb_p(SIO_DEV_SEL, SIO_INDEX);
+	outb_p(SIO_GP_DEV, SIO_DATA);	/* Talk to GPIO regs. */
+
+	outb_p(SIO_DEV_MSB, SIO_INDEX);
+	outb_p(SIO_GP_MSB, SIO_DATA);	/* MSB of GPIO base address */
+
+	outb_p(SIO_DEV_LSB, SIO_INDEX);
+	outb_p(SIO_GP_LSB, SIO_DATA);	/* LSB of GPIO base address */
+
+	outb_p(SIO_DEV_ENB, SIO_INDEX);
+	outb_p(1, SIO_DATA);		/* Enable GPIO registers. */
+
+	/*
+	 * Now, we have to map the power management section to write
+	 * a bit which enables access to the GPIO registers.
+	 * What lunatic came up with this shit?
+	 */
+	outb_p(SIO_DEV_SEL, SIO_INDEX);
+	outb_p(SIO_PM_DEV, SIO_DATA);	/* Talk to GPIO regs. */
+
+	outb_p(SIO_DEV_MSB, SIO_INDEX);
+	outb_p(SIO_PM_MSB, SIO_DATA);	/* MSB of PM base address */
+
+	outb_p(SIO_DEV_LSB, SIO_INDEX);
+	outb_p(SIO_PM_LSB, SIO_DATA);	/* LSB of PM base address */
+
+	outb_p(SIO_DEV_ENB, SIO_INDEX);
+	outb_p(1, SIO_DATA);		/* Enable PM registers. */
+
+	/*
+	 * Now, write the PM register which enables the GPIO registers.
+	 */
+	outb_p(SIO_PM_FER2, SIO_PM_INDEX);
+	outb_p(SIO_PM_GP_EN, SIO_PM_DATA);
+
+	/*
+	 * Now, initialize the GPIO registers.
+	 * We want them all to be inputs which is the
+	 * power on default, so let's leave them alone.
+	 * So, let's just read the board rev!
+	 */
+	raw = inb_p(SIO_GP_DATA1);
+	raw &= 0x7f;	/* 7 bits of valid board revision ID. */
+
+	if (visws_board_type == VISWS_320) {
+		if (raw < 0x6) {
+			visws_board_rev = 4;
+		} else if (raw < 0xc) {
+			visws_board_rev = 5;
+		} else {
+			visws_board_rev = 6;
+		}
+	} else if (visws_board_type == VISWS_540) {
+			visws_board_rev = 2;
+		} else {
+			visws_board_rev = raw;
+		}
+
+	printk(KERN_INFO "Silicon Graphics Visual Workstation %s (rev %d) detected\n",
+	       (visws_board_type == VISWS_320 ? "320" :
+	       (visws_board_type == VISWS_540 ? "540" :
+		"unknown")), visws_board_rev);
+}
+
+#define A01234 (LI_INTA_0 | LI_INTA_1 | LI_INTA_2 | LI_INTA_3 | LI_INTA_4)
+#define BCD (LI_INTB | LI_INTC | LI_INTD)
+#define ALLDEVS (A01234 | BCD)
+
+static __init void lithium_init(void)
+{
+	set_fixmap(FIX_LI_PCIA, LI_PCI_A_PHYS);
+	set_fixmap(FIX_LI_PCIB, LI_PCI_B_PHYS);
+
+	if ((li_pcia_read16(PCI_VENDOR_ID) != PCI_VENDOR_ID_SGI) ||
+	    (li_pcia_read16(PCI_DEVICE_ID) != PCI_DEVICE_ID_SGI_LITHIUM)) {
+		printk(KERN_EMERG "Lithium hostbridge %c not found\n", 'A');
+/*		panic("This machine is not SGI Visual Workstation 320/540"); */
+	}
+
+	if ((li_pcib_read16(PCI_VENDOR_ID) != PCI_VENDOR_ID_SGI) ||
+	    (li_pcib_read16(PCI_DEVICE_ID) != PCI_DEVICE_ID_SGI_LITHIUM)) {
+		printk(KERN_EMERG "Lithium hostbridge %c not found\n", 'B');
+/*		panic("This machine is not SGI Visual Workstation 320/540"); */
+	}
+
+	li_pcia_write16(LI_PCI_INTEN, ALLDEVS);
+	li_pcib_write16(LI_PCI_INTEN, ALLDEVS);
+}
+
+static __init void cobalt_init(void)
+{
+	/*
+	 * On normal SMP PC this is used only with SMP, but we have to
+	 * use it and set it up here to start the Cobalt clock
+	 */
+	set_fixmap(FIX_APIC_BASE, APIC_DEFAULT_PHYS_BASE);
+	setup_local_APIC();
+	printk(KERN_INFO "Local APIC Version %#x, ID %#x\n",
+		(unsigned int)apic_read(APIC_LVR),
+		(unsigned int)apic_read(APIC_ID));
+
+	set_fixmap(FIX_CO_CPU, CO_CPU_PHYS);
+	set_fixmap(FIX_CO_APIC, CO_APIC_PHYS);
+	printk(KERN_INFO "Cobalt Revision %#lx, APIC ID %#lx\n",
+		co_cpu_read(CO_CPU_REV), co_apic_read(CO_APIC_ID));
+
+	/* Enable Cobalt APIC being careful to NOT change the ID! */
+	co_apic_write(CO_APIC_ID, co_apic_read(CO_APIC_ID) | CO_APIC_ENABLE);
+
+	printk(KERN_INFO "Cobalt APIC enabled: ID reg %#lx\n",
+		co_apic_read(CO_APIC_ID));
+}
+
+static void __init visws_trap_init(void)
+{
+	lithium_init();
+	cobalt_init();
+}
+
+/*
+ * IRQ controller / APIC support:
+ */
+
+static DEFINE_SPINLOCK(cobalt_lock);
+
+/*
+ * Set the given Cobalt APIC Redirection Table entry to point
+ * to the given IDT vector/index.
+ */
+static inline void co_apic_set(int entry, int irq)
+{
+	co_apic_write(CO_APIC_LO(entry), CO_APIC_LEVEL | (irq + FIRST_EXTERNAL_VECTOR));
+	co_apic_write(CO_APIC_HI(entry), 0);
+}
+
+/*
+ * Cobalt (IO)-APIC functions to handle PCI devices.
+ */
+static inline int co_apic_ide0_hack(void)
+{
+	extern char visws_board_type;
+	extern char visws_board_rev;
+
+	if (visws_board_type == VISWS_320 && visws_board_rev == 5)
+		return 5;
+	return CO_APIC_IDE0;
+}
+
+static int is_co_apic(unsigned int irq)
+{
+	if (IS_CO_APIC(irq))
+		return CO_APIC(irq);
+
+	switch (irq) {
+		case 0: return CO_APIC_CPU;
+		case CO_IRQ_IDE0: return co_apic_ide0_hack();
+		case CO_IRQ_IDE1: return CO_APIC_IDE1;
+		default: return -1;
+	}
+}
+
+
+/*
+ * This is the SGI Cobalt (IO-)APIC:
+ */
+static void enable_cobalt_irq(struct irq_data *data)
+{
+	co_apic_set(is_co_apic(data->irq), data->irq);
+}
+
+static void disable_cobalt_irq(struct irq_data *data)
+{
+	int entry = is_co_apic(data->irq);
+
+	co_apic_write(CO_APIC_LO(entry), CO_APIC_MASK);
+	co_apic_read(CO_APIC_LO(entry));
+}
+
+static void ack_cobalt_irq(struct irq_data *data)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cobalt_lock, flags);
+	disable_cobalt_irq(data);
+	apic_write(APIC_EOI, APIC_EIO_ACK);
+	spin_unlock_irqrestore(&cobalt_lock, flags);
+}
+
+static struct irq_chip cobalt_irq_type = {
+	.name		= "Cobalt-APIC",
+	.irq_enable	= enable_cobalt_irq,
+	.irq_disable	= disable_cobalt_irq,
+	.irq_ack	= ack_cobalt_irq,
+};
+
+
+/*
+ * This is the PIIX4-based 8259 that is wired up indirectly to Cobalt
+ * -- not the manner expected by the code in i8259.c.
+ *
+ * there is a 'master' physical interrupt source that gets sent to
+ * the CPU. But in the chipset there are various 'virtual' interrupts
+ * waiting to be handled. We represent this to Linux through a 'master'
+ * interrupt controller type, and through a special virtual interrupt-
+ * controller. Device drivers only see the virtual interrupt sources.
+ */
+static unsigned int startup_piix4_master_irq(struct irq_data *data)
+{
+	legacy_pic->init(0);
+	enable_cobalt_irq(data);
+}
+
+static void end_piix4_master_irq(struct irq_data *data)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cobalt_lock, flags);
+	enable_cobalt_irq(data);
+	spin_unlock_irqrestore(&cobalt_lock, flags);
+}
+
+static struct irq_chip piix4_master_irq_type = {
+	.name		= "PIIX4-master",
+	.irq_startup	= startup_piix4_master_irq,
+	.irq_ack	= ack_cobalt_irq,
+};
+
+static void pii4_mask(struct irq_data *data) { }
+
+static struct irq_chip piix4_virtual_irq_type = {
+	.name		= "PIIX4-virtual",
+	.mask		= pii4_mask,
+};
+
+/*
+ * PIIX4-8259 master/virtual functions to handle interrupt requests
+ * from legacy devices: floppy, parallel, serial, rtc.
+ *
+ * None of these get Cobalt APIC entries, neither do they have IDT
+ * entries. These interrupts are purely virtual and distributed from
+ * the 'master' interrupt source: CO_IRQ_8259.
+ *
+ * When the 8259 interrupts its handler figures out which of these
+ * devices is interrupting and dispatches to its handler.
+ *
+ * CAREFUL: devices see the 'virtual' interrupt only. Thus disable/
+ * enable_irq gets the right irq. This 'master' irq is never directly
+ * manipulated by any driver.
+ */
+static irqreturn_t piix4_master_intr(int irq, void *dev_id)
+{
+	unsigned long flags;
+	int realirq;
+
+	raw_spin_lock_irqsave(&i8259A_lock, flags);
+
+	/* Find out what's interrupting in the PIIX4 master 8259 */
+	outb(0x0c, 0x20);		/* OCW3 Poll command */
+	realirq = inb(0x20);
+
+	/*
+	 * Bit 7 == 0 means invalid/spurious
+	 */
+	if (unlikely(!(realirq & 0x80)))
+		goto out_unlock;
+
+	realirq &= 7;
+
+	if (unlikely(realirq == 2)) {
+		outb(0x0c, 0xa0);
+		realirq = inb(0xa0);
+
+		if (unlikely(!(realirq & 0x80)))
+			goto out_unlock;
+
+		realirq = (realirq & 7) + 8;
+	}
+
+	/* mask and ack interrupt */
+	cached_irq_mask |= 1 << realirq;
+	if (unlikely(realirq > 7)) {
+		inb(0xa1);
+		outb(cached_slave_mask, 0xa1);
+		outb(0x60 + (realirq & 7), 0xa0);
+		outb(0x60 + 2, 0x20);
+	} else {
+		inb(0x21);
+		outb(cached_master_mask, 0x21);
+		outb(0x60 + realirq, 0x20);
+	}
+
+	raw_spin_unlock_irqrestore(&i8259A_lock, flags);
+
+	/*
+	 * handle this 'virtual interrupt' as a Cobalt one now.
+	 */
+	generic_handle_irq(realirq);
+
+	return IRQ_HANDLED;
+
+out_unlock:
+	raw_spin_unlock_irqrestore(&i8259A_lock, flags);
+	return IRQ_NONE;
+}
+
+static struct irqaction master_action = {
+	.handler =	piix4_master_intr,
+	.name =		"PIIX4-8259",
+};
+
+static struct irqaction cascade_action = {
+	.handler = 	no_action,
+	.name =		"cascade",
+};
+
+static inline void set_piix4_virtual_irq_type(void)
+{
+	piix4_virtual_irq_type.enable =	i8259A_chip.unmask;
+	piix4_virtual_irq_type.disable = i8259A_chip.mask;
+	piix4_virtual_irq_type.unmask =	i8259A_chip.unmask;
+}
+
+static void __init visws_pre_intr_init(void)
+{
+	int i;
+
+	set_piix4_virtual_irq_type();
+
+	for (i = 0; i < CO_IRQ_APIC0 + CO_APIC_LAST + 1; i++) {
+		struct irq_chip *chip = NULL;
+
+		if (i == 0)
+			chip = &cobalt_irq_type;
+		else if (i == CO_IRQ_IDE0)
+			chip = &cobalt_irq_type;
+		else if (i == CO_IRQ_IDE1)
+			>chip = &cobalt_irq_type;
+		else if (i == CO_IRQ_8259)
+			chip = &piix4_master_irq_type;
+		else if (i < CO_IRQ_APIC0)
+			chip = &piix4_virtual_irq_type;
+		else if (IS_CO_APIC(i))
+			chip = &cobalt_irq_type;
+
+		if (chip)
+			set_irq_chip(i, chip);
+	}
+
+	setup_irq(CO_IRQ_8259, &master_action);
+	setup_irq(2, &cascade_action);
+}