Merge refs/heads/release from master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6

arch/ia64/Kconfig

@@ -383,6 +383,12 @@ source "drivers/acpi/Kconfig"
 
 endif
 
+if PM
+
+source "arch/ia64/kernel/cpufreq/Kconfig"
+
+endif
+
 endmenu
 
 if !IA64_HP_SIM

arch/ia64/hp/sim/boot/fw-emu.c

@@ -237,17 +237,6 @@ sal_emulator (long index, unsigned long in1, unsigned long in2,
 	return ((struct sal_ret_values) {status, r9, r10, r11});
 }
 
-
-/*
- * This is here to work around a bug in egcs-1.1.1b that causes the
- * compiler to crash (seems like a bug in the new alias analysis code.
- */
-void *
-id (long addr)
-{
-	return (void *) addr;
-}
-
 struct ia64_boot_param *
 sys_fw_init (const char *args, int arglen)
 {

arch/ia64/ia32/ia32_signal.c

@@ -29,7 +29,6 @@
 #include <asm/uaccess.h>
 #include <asm/rse.h>
 #include <asm/sigcontext.h>
-#include <asm/segment.h>
 
 #include "ia32priv.h"
 

arch/ia64/kernel/Makefile

@@ -20,6 +20,7 @@ obj-$(CONFIG_SMP)		+= smp.o smpboot.o domain.o
 obj-$(CONFIG_NUMA)		+= numa.o
 obj-$(CONFIG_PERFMON)		+= perfmon_default_smpl.o
 obj-$(CONFIG_IA64_CYCLONE)	+= cyclone.o
+obj-$(CONFIG_CPU_FREQ)		+= cpufreq/
 obj-$(CONFIG_IA64_MCA_RECOVERY)	+= mca_recovery.o
 obj-$(CONFIG_KPROBES)		+= kprobes.o jprobes.o
 obj-$(CONFIG_IA64_UNCACHED_ALLOCATOR)	+= uncached.o

arch/ia64/kernel/cpufreq/Kconfig

+
+#
+# CPU Frequency scaling
+#
+
+menu "CPU Frequency scaling"
+
+source "drivers/cpufreq/Kconfig"
+
+if CPU_FREQ
+
+comment "CPUFreq processor drivers"
+
+config IA64_ACPI_CPUFREQ
+	tristate "ACPI Processor P-States driver"
+	select CPU_FREQ_TABLE
+	depends on ACPI_PROCESSOR
+	help
+	This driver adds a CPUFreq driver which utilizes the ACPI
+	Processor Performance States.
+
+	For details, take a look at <file:Documentation/cpu-freq/>.
+
+	If in doubt, say N.
+
+endif   # CPU_FREQ
+
+endmenu
+

arch/ia64/kernel/cpufreq/Makefile

+obj-$(CONFIG_IA64_ACPI_CPUFREQ)		+= acpi-cpufreq.o

arch/ia64/kernel/cpufreq/acpi-cpufreq.c

+/*
+ * arch/ia64/kernel/cpufreq/acpi-cpufreq.c
+ * This file provides the ACPI based P-state support. This
+ * module works with generic cpufreq infrastructure. Most of
+ * the code is based on i386 version
+ * (arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c)
+ *
+ * Copyright (C) 2005 Intel Corp
+ *      Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
+ */
+
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/cpufreq.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
+#include <asm/pal.h>
+
+#include <linux/acpi.h>
+#include <acpi/processor.h>
+
+#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "acpi-cpufreq", msg)
+
+MODULE_AUTHOR("Venkatesh Pallipadi");
+MODULE_DESCRIPTION("ACPI Processor P-States Driver");
+MODULE_LICENSE("GPL");
+
+
+struct cpufreq_acpi_io {
+	struct acpi_processor_performance	acpi_data;
+	struct cpufreq_frequency_table		*freq_table;
+	unsigned int				resume;
+};
+
+static struct cpufreq_acpi_io	*acpi_io_data[NR_CPUS];
+
+static struct cpufreq_driver acpi_cpufreq_driver;
+
+
+static int
+processor_set_pstate (
+	u32	value)
+{
+	s64 retval;
+
+	dprintk("processor_set_pstate\n");
+
+	retval = ia64_pal_set_pstate((u64)value);
+
+	if (retval) {
+		dprintk("Failed to set freq to 0x%x, with error 0x%x\n",
+		        value, retval);
+		return -ENODEV;
+	}
+	return (int)retval;
+}
+
+
+static int
+processor_get_pstate (
+	u32	*value)
+{
+	u64	pstate_index = 0;
+	s64 	retval;
+
+	dprintk("processor_get_pstate\n");
+
+	retval = ia64_pal_get_pstate(&pstate_index);
+	*value = (u32) pstate_index;
+
+	if (retval)
+		dprintk("Failed to get current freq with "
+		        "error 0x%x, idx 0x%x\n", retval, *value);
+
+	return (int)retval;
+}
+
+
+/* To be used only after data->acpi_data is initialized */
+static unsigned
+extract_clock (
+	struct cpufreq_acpi_io *data,
+	unsigned value,
+	unsigned int cpu)
+{
+	unsigned long i;
+
+	dprintk("extract_clock\n");
+
+	for (i = 0; i < data->acpi_data.state_count; i++) {
+		if (value >= data->acpi_data.states[i].control)
+			return data->acpi_data.states[i].core_frequency;
+	}
+	return data->acpi_data.states[i-1].core_frequency;
+}
+
+
+static unsigned int
+processor_get_freq (
+	struct cpufreq_acpi_io	*data,
+	unsigned int		cpu)
+{
+	int			ret = 0;
+	u32			value = 0;
+	cpumask_t		saved_mask;
+	unsigned long 		clock_freq;
+
+	dprintk("processor_get_freq\n");
+
+	saved_mask = current->cpus_allowed;
+	set_cpus_allowed(current, cpumask_of_cpu(cpu));
+	if (smp_processor_id() != cpu) {
+		ret = -EAGAIN;
+		goto migrate_end;
+	}
+
+	/*
+	 * processor_get_pstate gets the average frequency since the
+	 * last get. So, do two PAL_get_freq()...
+	 */
+	ret = processor_get_pstate(&value);
+	ret = processor_get_pstate(&value);
+
+	if (ret) {
+		set_cpus_allowed(current, saved_mask);
+		printk(KERN_WARNING "get performance failed with error %d\n",
+		       ret);
+		ret = -EAGAIN;
+		goto migrate_end;
+	}
+	clock_freq = extract_clock(data, value, cpu);
+	ret = (clock_freq*1000);
+
+migrate_end:
+	set_cpus_allowed(current, saved_mask);
+	return ret;
+}
+
+
+static int
+processor_set_freq (
+	struct cpufreq_acpi_io	*data,
+	unsigned int		cpu,
+	int			state)
+{
+	int			ret = 0;
+	u32			value = 0;
+	struct cpufreq_freqs    cpufreq_freqs;
+	cpumask_t		saved_mask;
+	int			retval;
+
+	dprintk("processor_set_freq\n");
+
+	saved_mask = current->cpus_allowed;
+	set_cpus_allowed(current, cpumask_of_cpu(cpu));
+	if (smp_processor_id() != cpu) {
+		retval = -EAGAIN;
+		goto migrate_end;
+	}
+
+	if (state == data->acpi_data.state) {
+		if (unlikely(data->resume)) {
+			dprintk("Called after resume, resetting to P%d\n", state);
+			data->resume = 0;
+		} else {
+			dprintk("Already at target state (P%d)\n", state);
+			retval = 0;
+			goto migrate_end;
+		}
+	}
+
+	dprintk("Transitioning from P%d to P%d\n",
+		data->acpi_data.state, state);
+
+	/* cpufreq frequency struct */
+	cpufreq_freqs.cpu = cpu;
+	cpufreq_freqs.old = data->freq_table[data->acpi_data.state].frequency;
+	cpufreq_freqs.new = data->freq_table[state].frequency;
+
+	/* notify cpufreq */
+	cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_PRECHANGE);
+
+	/*
+	 * First we write the target state's 'control' value to the
+	 * control_register.
+	 */
+
+	value = (u32) data->acpi_data.states[state].control;
+
+	dprintk("Transitioning to state: 0x%08x\n", value);
+
+	ret = processor_set_pstate(value);
+	if (ret) {
+		unsigned int tmp = cpufreq_freqs.new;
+		cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_POSTCHANGE);
+		cpufreq_freqs.new = cpufreq_freqs.old;
+		cpufreq_freqs.old = tmp;
+		cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_PRECHANGE);
+		cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_POSTCHANGE);
+		printk(KERN_WARNING "Transition failed with error %d\n", ret);
+		retval = -ENODEV;
+		goto migrate_end;
+	}
+
+	cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_POSTCHANGE);
+
+	data->acpi_data.state = state;
+
+	retval = 0;
+
+migrate_end:
+	set_cpus_allowed(current, saved_mask);
+	return (retval);
+}
+
+
+static unsigned int
+acpi_cpufreq_get (
+	unsigned int		cpu)
+{
+	struct cpufreq_acpi_io *data = acpi_io_data[cpu];
+
+	dprintk("acpi_cpufreq_get\n");
+
+	return processor_get_freq(data, cpu);
+}
+
+
+static int
+acpi_cpufreq_target (
+	struct cpufreq_policy   *policy,
+	unsigned int target_freq,
+	unsigned int relation)
+{
+	struct cpufreq_acpi_io *data = acpi_io_data[policy->cpu];
+	unsigned int next_state = 0;
+	unsigned int result = 0;
+
+	dprintk("acpi_cpufreq_setpolicy\n");
+
+	result = cpufreq_frequency_table_target(policy,
+			data->freq_table, target_freq, relation, &next_state);
+	if (result)
+		return (result);
+
+	result = processor_set_freq(data, policy->cpu, next_state);
+
+	return (result);
+}
+
+
+static int
+acpi_cpufreq_verify (
+	struct cpufreq_policy   *policy)
+{
+	unsigned int result = 0;
+	struct cpufreq_acpi_io *data = acpi_io_data[policy->cpu];
+
+	dprintk("acpi_cpufreq_verify\n");
+
+	result = cpufreq_frequency_table_verify(policy,
+			data->freq_table);
+
+	return (result);
+}
+
+
+/*
+ * processor_init_pdc - let BIOS know about the SMP capabilities
+ * of this driver
+ * @perf: processor-specific acpi_io_data struct
+ * @cpu: CPU being initialized
+ *
+ * To avoid issues with legacy OSes, some BIOSes require to be informed of
+ * the SMP capabilities of OS P-state driver. Here we set the bits in _PDC
+ * accordingly. Actual call to _PDC is done in driver/acpi/processor.c
+ */
+static void
+processor_init_pdc (
+		struct acpi_processor_performance *perf,
+		unsigned int cpu,
+		struct acpi_object_list *obj_list
+		)
+{
+	union acpi_object *obj;
+	u32 *buf;
+
+	dprintk("processor_init_pdc\n");
+
+	perf->pdc = NULL;
+	/* Initialize pdc. It will be used later. */
+	if (!obj_list)
+		return;
+
+	if (!(obj_list->count && obj_list->pointer))
+		return;
+
+	obj = obj_list->pointer;
+	if ((obj->buffer.length == 12) && obj->buffer.pointer) {
+		buf = (u32 *)obj->buffer.pointer;
+       		buf[0] = ACPI_PDC_REVISION_ID;
+       		buf[1] = 1;
+       		buf[2] = ACPI_PDC_EST_CAPABILITY_SMP;
+		perf->pdc = obj_list;
+	}
+	return;
+}
+
+
+static int
+acpi_cpufreq_cpu_init (
+	struct cpufreq_policy   *policy)
+{
+	unsigned int		i;
+	unsigned int		cpu = policy->cpu;
+	struct cpufreq_acpi_io	*data;
+	unsigned int		result = 0;
+
+	union acpi_object		arg0 = {ACPI_TYPE_BUFFER};
+	u32				arg0_buf[3];
+	struct acpi_object_list 	arg_list = {1, &arg0};
+
+	dprintk("acpi_cpufreq_cpu_init\n");
+	/* setup arg_list for _PDC settings */
+        arg0.buffer.length = 12;
+        arg0.buffer.pointer = (u8 *) arg0_buf;
+
+	data = kmalloc(sizeof(struct cpufreq_acpi_io), GFP_KERNEL);
+	if (!data)
+		return (-ENOMEM);
+
+	memset(data, 0, sizeof(struct cpufreq_acpi_io));
+
+	acpi_io_data[cpu] = data;
+
+	processor_init_pdc(&data->acpi_data, cpu, &arg_list);
+	result = acpi_processor_register_performance(&data->acpi_data, cpu);
+	data->acpi_data.pdc = NULL;
+
+	if (result)
+		goto err_free;
+
+	/* capability check */
+	if (data->acpi_data.state_count <= 1) {
+		dprintk("No P-States\n");
+		result = -ENODEV;
+		goto err_unreg;
+	}
+
+	if ((data->acpi_data.control_register.space_id !=
+					ACPI_ADR_SPACE_FIXED_HARDWARE) ||
+	    (data->acpi_data.status_register.space_id !=
+					ACPI_ADR_SPACE_FIXED_HARDWARE)) {
+		dprintk("Unsupported address space [%d, %d]\n",
+			(u32) (data->acpi_data.control_register.space_id),
+			(u32) (data->acpi_data.status_register.space_id));
+		result = -ENODEV;
+		goto err_unreg;
+	}
+
+	/* alloc freq_table */
+	data->freq_table = kmalloc(sizeof(struct cpufreq_frequency_table) *
+	                           (data->acpi_data.state_count + 1),
+	                           GFP_KERNEL);
+	if (!data->freq_table) {
+		result = -ENOMEM;
+		goto err_unreg;
+	}
+
+	/* detect transition latency */
+	policy->cpuinfo.transition_latency = 0;
+	for (i=0; i<data->acpi_data.state_count; i++) {
+		if ((data->acpi_data.states[i].transition_latency * 1000) >
+		    policy->cpuinfo.transition_latency) {
+			policy->cpuinfo.transition_latency =
+			    data->acpi_data.states[i].transition_latency * 1000;
+		}
+	}
+	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
+
+	policy->cur = processor_get_freq(data, policy->cpu);
+
+	/* table init */
+	for (i = 0; i <= data->acpi_data.state_count; i++)
+	{
+		data->freq_table[i].index = i;
+		if (i < data->acpi_data.state_count) {
+			data->freq_table[i].frequency =
+			      data->acpi_data.states[i].core_frequency * 1000;
+		} else {
+			data->freq_table[i].frequency = CPUFREQ_TABLE_END;
+		}
+	}
+
+	result = cpufreq_frequency_table_cpuinfo(policy, data->freq_table);
+	if (result) {
+		goto err_freqfree;
+	}
+
+	/* notify BIOS that we exist */
+	acpi_processor_notify_smm(THIS_MODULE);
+
+	printk(KERN_INFO "acpi-cpufreq: CPU%u - ACPI performance management "
+	       "activated.\n", cpu);
+
+	for (i = 0; i < data->acpi_data.state_count; i++)
+		dprintk("     %cP%d: %d MHz, %d mW, %d uS, %d uS, 0x%x 0x%x\n",
+			(i == data->acpi_data.state?'*':' '), i,
+			(u32) data->acpi_data.states[i].core_frequency,
+			(u32) data->acpi_data.states[i].power,
+			(u32) data->acpi_data.states[i].transition_latency,
+			(u32) data->acpi_data.states[i].bus_master_latency,
+			(u32) data->acpi_data.states[i].status,
+			(u32) data->acpi_data.states[i].control);
+
+	cpufreq_frequency_table_get_attr(data->freq_table, policy->cpu);
+
+	/* the first call to ->target() should result in us actually
+	 * writing something to the appropriate registers. */
+	data->resume = 1;
+
+	return (result);
+
+ err_freqfree:
+	kfree(data->freq_table);
+ err_unreg:
+	acpi_processor_unregister_performance(&data->acpi_data, cpu);
+ err_free:
+	kfree(data);
+	acpi_io_data[cpu] = NULL;
+
+	return (result);
+}
+
+
+static int
+acpi_cpufreq_cpu_exit (
+	struct cpufreq_policy   *policy)
+{
+	struct cpufreq_acpi_io *data = acpi_io_data[policy->cpu];
+
+	dprintk("acpi_cpufreq_cpu_exit\n");
+
+	if (data) {
+		cpufreq_frequency_table_put_attr(policy->cpu);
+		acpi_io_data[policy->cpu] = NULL;
+		acpi_processor_unregister_performance(&data->acpi_data,
+		                                      policy->cpu);
+		kfree(data);
+	}
+
+	return (0);
+}
+
+
+static struct freq_attr* acpi_cpufreq_attr[] = {
+	&cpufreq_freq_attr_scaling_available_freqs,
+	NULL,
+};
+
+
+static struct cpufreq_driver acpi_cpufreq_driver = {
+	.verify 	= acpi_cpufreq_verify,
+	.target 	= acpi_cpufreq_target,
+	.get 		= acpi_cpufreq_get,
+	.init		= acpi_cpufreq_cpu_init,
+	.exit		= acpi_cpufreq_cpu_exit,
+	.name		= "acpi-cpufreq",
+	.owner		= THIS_MODULE,
+	.attr           = acpi_cpufreq_attr,
+};
+
+
+static int __init
+acpi_cpufreq_init (void)
+{
+	dprintk("acpi_cpufreq_init\n");
+
+ 	return cpufreq_register_driver(&acpi_cpufreq_driver);
+}
+
+
+static void __exit
+acpi_cpufreq_exit (void)
+{
+	dprintk("acpi_cpufreq_exit\n");
+
+	cpufreq_unregister_driver(&acpi_cpufreq_driver);
+	return;
+}
+
+
+late_initcall(acpi_cpufreq_init);
+module_exit(acpi_cpufreq_exit);
+

arch/ia64/kernel/sys_ia64.c

@@ -35,7 +35,7 @@ arch_get_unmapped_area (struct file *filp, unsigned long addr, unsigned long len
 		return -ENOMEM;
 
 #ifdef CONFIG_HUGETLB_PAGE
-	if (REGION_NUMBER(addr) == REGION_HPAGE)
+	if (REGION_NUMBER(addr) == RGN_HPAGE)
 		addr = 0;
 #endif
 	if (!addr)

arch/ia64/kernel/uncached.c

@@ -184,7 +184,7 @@ uncached_free_page(unsigned long maddr)
 {
 	int node;
 
-	node = nasid_to_cnodeid(NASID_GET(maddr));
+	node = paddr_to_nid(maddr - __IA64_UNCACHED_OFFSET);
 
 	dprintk(KERN_DEBUG "uncached_free_page(%lx) on node %i\n", maddr, node);
 
@@ -217,7 +217,7 @@ uncached_build_memmap(unsigned long start, unsigned long end, void *arg)
 
 	memset((char *)vstart, 0, length);
 
-	node = nasid_to_cnodeid(NASID_GET(start));
+	node = paddr_to_nid(start);
 
 	for (; vstart < vend ; vstart += PAGE_SIZE) {
 		dprintk(KERN_INFO "sticking %lx into the pool!\n", vstart);

arch/ia64/lib/Makefile

@@ -6,7 +6,7 @@ obj-y := io.o
 
 lib-y := __divsi3.o __udivsi3.o __modsi3.o __umodsi3.o			\
 	__divdi3.o __udivdi3.o __moddi3.o __umoddi3.o			\
-	bitop.o checksum.o clear_page.o csum_partial_copy.o copy_page.o	\
+	bitop.o checksum.o clear_page.o csum_partial_copy.o		\
 	clear_user.o strncpy_from_user.o strlen_user.o strnlen_user.o	\
 	flush.o ip_fast_csum.o do_csum.o				\
 	memset.o strlen.o swiotlb.o

arch/ia64/lib/swiotlb.c

@@ -93,8 +93,7 @@ static int __init
 setup_io_tlb_npages(char *str)
 {
 	if (isdigit(*str)) {
-		io_tlb_nslabs = simple_strtoul(str, &str, 0) <<
-			(PAGE_SHIFT - IO_TLB_SHIFT);
+		io_tlb_nslabs = simple_strtoul(str, &str, 0);
 		/* avoid tail segment of size < IO_TLB_SEGSIZE */
 		io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE);
 	}
@@ -117,7 +116,7 @@ swiotlb_init_with_default_size (size_t default_size)
 	unsigned long i;
 
 	if (!io_tlb_nslabs) {
-		io_tlb_nslabs = (default_size >> PAGE_SHIFT);
+		io_tlb_nslabs = (default_size >> IO_TLB_SHIFT);
 		io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE);
 	}
 

arch/ia64/mm/hugetlbpage.c

@@ -76,7 +76,7 @@ int is_aligned_hugepage_range(unsigned long addr, unsigned long len)
 		return -EINVAL;
 	if (addr & ~HPAGE_MASK)
 		return -EINVAL;
-	if (REGION_NUMBER(addr) != REGION_HPAGE)
+	if (REGION_NUMBER(addr) != RGN_HPAGE)
 		return -EINVAL;
 
 	return 0;
@@ -87,7 +87,7 @@ struct page *follow_huge_addr(struct mm_struct *mm, unsigned long addr, int writ
 	struct page *page;
 	pte_t *ptep;
 
-	if (REGION_NUMBER(addr) != REGION_HPAGE)
+	if (REGION_NUMBER(addr) != RGN_HPAGE)
 		return ERR_PTR(-EINVAL);
 
 	ptep = huge_pte_offset(mm, addr);
@@ -142,8 +142,8 @@ unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, u
 		return -ENOMEM;
 	if (len & ~HPAGE_MASK)
 		return -EINVAL;
-	/* This code assumes that REGION_HPAGE != 0. */
-	if ((REGION_NUMBER(addr) != REGION_HPAGE) || (addr & (HPAGE_SIZE - 1)))
+	/* This code assumes that RGN_HPAGE != 0. */
+	if ((REGION_NUMBER(addr) != RGN_HPAGE) || (addr & (HPAGE_SIZE - 1)))
 		addr = HPAGE_REGION_BASE;
 	else
 		addr = ALIGN(addr, HPAGE_SIZE);

arch/ia64/pci/pci.c

@@ -24,7 +24,6 @@
 
 #include <asm/machvec.h>
 #include <asm/page.h>
-#include <asm/segment.h>
 #include <asm/system.h>
 #include <asm/io.h>
 #include <asm/sal.h>

arch/ia64/sn/include/tio.h

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #ifndef _ASM_IA64_SN_TIO_H
@@ -26,6 +26,10 @@
 #define TIO_ITTE_VALID_MASK	0x1
 #define TIO_ITTE_VALID_SHIFT	16
 
+#define TIO_ITTE_WIDGET(itte) \
+	(((itte) >> TIO_ITTE_WIDGET_SHIFT) & TIO_ITTE_WIDGET_MASK)
+#define TIO_ITTE_VALID(itte) \
+	(((itte) >> TIO_ITTE_VALID_SHIFT) & TIO_ITTE_VALID_MASK)
 
 #define TIO_ITTE_PUT(nasid, bigwin, widget, addr, valid) \
         REMOTE_HUB_S((nasid), TIO_ITTE(bigwin), \

arch/ia64/sn/include/xtalk/hubdev.h

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1992 - 1997, 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 1992 - 1997, 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 #ifndef _ASM_IA64_SN_XTALK_HUBDEV_H
 #define _ASM_IA64_SN_XTALK_HUBDEV_H
@@ -16,6 +16,9 @@
 #define IIO_ITTE_WIDGET_MASK    ((1<<IIO_ITTE_WIDGET_BITS)-1)
 #define IIO_ITTE_WIDGET_SHIFT   8
 
+#define IIO_ITTE_WIDGET(itte)	\
+	(((itte) >> IIO_ITTE_WIDGET_SHIFT) & IIO_ITTE_WIDGET_MASK)
+
 /*
  * Use the top big window as a surrogate for the first small window
  */
@@ -34,7 +37,8 @@ struct sn_flush_device_list {
 	unsigned long sfdl_force_int_addr;
 	unsigned long sfdl_flush_value;
 	volatile unsigned long *sfdl_flush_addr;
-	uint64_t sfdl_persistent_busnum;
+	uint32_t sfdl_persistent_busnum;
+	uint32_t sfdl_persistent_segment;
 	struct pcibus_info *sfdl_pcibus_info;
 	spinlock_t sfdl_flush_lock;
 };
@@ -58,7 +62,8 @@ struct hubdev_info {
 
 	void				*hdi_nodepda;
 	void				*hdi_node_vertex;
-	void				*hdi_xtalk_vertex;
+	uint32_t			max_segment_number;
+	uint32_t			max_pcibus_number;
 };
 
 extern void hubdev_init_node(nodepda_t *, cnodeid_t);

arch/ia64/sn/kernel/bte.c

@@ -29,16 +29,30 @@
 
 /* two interfaces on two btes */
 #define MAX_INTERFACES_TO_TRY		4
+#define MAX_NODES_TO_TRY		2
 
 static struct bteinfo_s *bte_if_on_node(nasid_t nasid, int interface)
 {
 	nodepda_t *tmp_nodepda;
 
+	if (nasid_to_cnodeid(nasid) == -1)
+		return (struct bteinfo_s *)NULL;;
+
 	tmp_nodepda = NODEPDA(nasid_to_cnodeid(nasid));
 	return &tmp_nodepda->bte_if[interface];
 
 }
 
+static inline void bte_start_transfer(struct bteinfo_s *bte, u64 len, u64 mode)
+{
+	if (is_shub2()) {
+		BTE_CTRL_STORE(bte, (IBLS_BUSY | ((len) | (mode) << 24)));
+	} else {
+		BTE_LNSTAT_STORE(bte, len);
+		BTE_CTRL_STORE(bte, mode);
+	}
+}
+
 /************************************************************************
  * Block Transfer Engine copy related functions.
  *
@@ -67,13 +81,15 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 {
 	u64 transfer_size;
 	u64 transfer_stat;
+	u64 notif_phys_addr;
 	struct bteinfo_s *bte;
 	bte_result_t bte_status;
 	unsigned long irq_flags;
 	unsigned long itc_end = 0;
-	struct bteinfo_s *btes_to_try[MAX_INTERFACES_TO_TRY];
-	int bte_if_index;
-	int bte_pri, bte_sec;
+	int nasid_to_try[MAX_NODES_TO_TRY];
+	int my_nasid = get_nasid();
+	int bte_if_index, nasid_index;
+	int bte_first, btes_per_node = BTES_PER_NODE;
 
 	BTE_PRINTK(("bte_copy(0x%lx, 0x%lx, 0x%lx, 0x%lx, 0x%p)\n",
 		    src, dest, len, mode, notification));
@@ -86,36 +102,26 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 		 (src & L1_CACHE_MASK) || (dest & L1_CACHE_MASK));
 	BUG_ON(!(len < ((BTE_LEN_MASK + 1) << L1_CACHE_SHIFT)));
 
-	/* CPU 0 (per node) tries bte0 first, CPU 1 try bte1 first */
-	if (cpuid_to_subnode(smp_processor_id()) == 0) {
-		bte_pri = 0;
-		bte_sec = 1;
-	} else {
-		bte_pri = 1;
-		bte_sec = 0;
-	}
+	/*
+	 * Start with interface corresponding to cpu number
+	 */
+	bte_first = raw_smp_processor_id() % btes_per_node;
 
 	if (mode & BTE_USE_DEST) {
 		/* try remote then local */
-		btes_to_try[0] = bte_if_on_node(NASID_GET(dest), bte_pri);
-		btes_to_try[1] = bte_if_on_node(NASID_GET(dest), bte_sec);
+		nasid_to_try[0] = NASID_GET(dest);
 		if (mode & BTE_USE_ANY) {
-			btes_to_try[2] = bte_if_on_node(get_nasid(), bte_pri);
-			btes_to_try[3] = bte_if_on_node(get_nasid(), bte_sec);
+			nasid_to_try[1] = my_nasid;
 		} else {
-			btes_to_try[2] = NULL;
-			btes_to_try[3] = NULL;
+			nasid_to_try[1] = (int)NULL;
 		}
 	} else {
 		/* try local then remote */
-		btes_to_try[0] = bte_if_on_node(get_nasid(), bte_pri);
-		btes_to_try[1] = bte_if_on_node(get_nasid(), bte_sec);
+		nasid_to_try[0] = my_nasid;
 		if (mode & BTE_USE_ANY) {
-			btes_to_try[2] = bte_if_on_node(NASID_GET(dest), bte_pri);
-			btes_to_try[3] = bte_if_on_node(NASID_GET(dest), bte_sec);
+			nasid_to_try[1] = NASID_GET(dest);
 		} else {
-			btes_to_try[2] = NULL;
-			btes_to_try[3] = NULL;
+			nasid_to_try[1] = (int)NULL;
 		}
 	}
 
@@ -123,11 +129,12 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 	do {
 		local_irq_save(irq_flags);
 
-		bte_if_index = 0;
+		bte_if_index = bte_first;
+		nasid_index = 0;
 
 		/* Attempt to lock one of the BTE interfaces. */
-		while (bte_if_index < MAX_INTERFACES_TO_TRY) {
-			bte = btes_to_try[bte_if_index++];
+		while (nasid_index < MAX_NODES_TO_TRY) {
+			bte = bte_if_on_node(nasid_to_try[nasid_index],bte_if_index);
 
 			if (bte == NULL) {
 				continue;
@@ -143,6 +150,15 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 					break;
 				}
 			}
+
+			bte_if_index = (bte_if_index + 1) % btes_per_node; /* Next interface */
+			if (bte_if_index == bte_first) {
+				/*
+				 * We've tried all interfaces on this node
+				 */
+				nasid_index++;
+			}
+
 			bte = NULL;
 		}
 
@@ -169,7 +185,13 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 
 	/* Initialize the notification to a known value. */
 	*bte->most_rcnt_na = BTE_WORD_BUSY;
+	notif_phys_addr = TO_PHYS(ia64_tpa((unsigned long)bte->most_rcnt_na));
 
+	if (is_shub2()) {
+		src = SH2_TIO_PHYS_TO_DMA(src);
+		dest = SH2_TIO_PHYS_TO_DMA(dest);
+		notif_phys_addr = SH2_TIO_PHYS_TO_DMA(notif_phys_addr);
+	}
 	/* Set the source and destination registers */
 	BTE_PRINTKV(("IBSA = 0x%lx)\n", (TO_PHYS(src))));
 	BTE_SRC_STORE(bte, TO_PHYS(src));
@@ -177,14 +199,12 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 	BTE_DEST_STORE(bte, TO_PHYS(dest));
 
 	/* Set the notification register */
-	BTE_PRINTKV(("IBNA = 0x%lx)\n",
-		     TO_PHYS(ia64_tpa((unsigned long)bte->most_rcnt_na))));
-	BTE_NOTIF_STORE(bte,
-			TO_PHYS(ia64_tpa((unsigned long)bte->most_rcnt_na)));
+	BTE_PRINTKV(("IBNA = 0x%lx)\n", notif_phys_addr));
+	BTE_NOTIF_STORE(bte, notif_phys_addr);
 
 	/* Initiate the transfer */
 	BTE_PRINTK(("IBCT = 0x%lx)\n", BTE_VALID_MODE(mode)));
-	BTE_START_TRANSFER(bte, transfer_size, BTE_VALID_MODE(mode));
+	bte_start_transfer(bte, transfer_size, BTE_VALID_MODE(mode));
 
 	itc_end = ia64_get_itc() + (40000000 * local_cpu_data->cyc_per_usec);
 
@@ -195,6 +215,7 @@ bte_result_t bte_copy(u64 src, u64 dest, u64 len, u64 mode, void *notification)
 	}
 
 	while ((transfer_stat = *bte->most_rcnt_na) == BTE_WORD_BUSY) {
+		cpu_relax();
 		if (ia64_get_itc() > itc_end) {
 			BTE_PRINTK(("BTE timeout nasid 0x%x bte%d IBLS = 0x%lx na 0x%lx\n",
 				NASID_GET(bte->bte_base_addr), bte->bte_num,

arch/ia64/sn/kernel/huberror.c

@@ -76,7 +76,7 @@ void hubiio_crb_free(struct hubdev_info *hubdev_info, int crbnum)
 	 */
 	REMOTE_HUB_S(hubdev_info->hdi_nasid, IIO_ICDR, (IIO_ICDR_PND | crbnum));
 	while (REMOTE_HUB_L(hubdev_info->hdi_nasid, IIO_ICDR) & IIO_ICDR_PND)
-		udelay(1);
+		cpu_relax();
 
 }
 

arch/ia64/sn/kernel/io_init.c

@@ -18,6 +18,7 @@
 #include <asm/sn/simulator.h>
 #include <asm/sn/sn_sal.h>
 #include <asm/sn/tioca_provider.h>
+#include <asm/sn/tioce_provider.h>
 #include "xtalk/hubdev.h"
 #include "xtalk/xwidgetdev.h"
 
@@ -44,6 +45,9 @@ int sn_ioif_inited = 0;		/* SN I/O infrastructure initialized? */
 
 struct sn_pcibus_provider *sn_pci_provider[PCIIO_ASIC_MAX_TYPES];	/* indexed by asic type */
 
+static int max_segment_number = 0; /* Default highest segment number */
+static int max_pcibus_number = 255; /* Default highest pci bus number */
+
 /*
  * Hooks and struct for unsupported pci providers
  */
@@ -157,13 +161,28 @@ static void sn_fixup_ionodes(void)
 	uint64_t nasid;
 	int i, widget;
 
+	/*
+	 * Get SGI Specific HUB chipset information.
+	 * Inform Prom that this kernel can support domain bus numbering.
+	 */
 	for (i = 0; i < numionodes; i++) {
 		hubdev = (struct hubdev_info *)(NODEPDA(i)->pdinfo);
 		nasid = cnodeid_to_nasid(i);
+		hubdev->max_segment_number = 0xffffffff;
+		hubdev->max_pcibus_number = 0xff;
 		status = sal_get_hubdev_info(nasid, (uint64_t) __pa(hubdev));
 		if (status)
 			continue;
 
+		/* Save the largest Domain and pcibus numbers found. */
+		if (hubdev->max_segment_number) {
+			/*
+			 * Dealing with a Prom that supports segments.
+			 */
+			max_segment_number = hubdev->max_segment_number;
+			max_pcibus_number = hubdev->max_pcibus_number;
+		}
+
 		/* Attach the error interrupt handlers */
 		if (nasid & 1)
 			ice_error_init(hubdev);
@@ -230,7 +249,7 @@ void sn_pci_unfixup_slot(struct pci_dev *dev)
 void sn_pci_fixup_slot(struct pci_dev *dev)
 {
 	int idx;
-	int segment = 0;
+	int segment = pci_domain_nr(dev->bus);
 	int status = 0;
 	struct pcibus_bussoft *bs;
  	struct pci_bus *host_pci_bus;
@@ -283,9 +302,9 @@ void sn_pci_fixup_slot(struct pci_dev *dev)
  	 * PCI host_pci_dev struct and set up host bus linkages
  	 */
 
- 	bus_no = SN_PCIDEV_INFO(dev)->pdi_slot_host_handle >> 32;
+ 	bus_no = (SN_PCIDEV_INFO(dev)->pdi_slot_host_handle >> 32) & 0xff;
  	devfn = SN_PCIDEV_INFO(dev)->pdi_slot_host_handle & 0xffffffff;
- 	host_pci_bus = pci_find_bus(pci_domain_nr(dev->bus), bus_no);
+ 	host_pci_bus = pci_find_bus(segment, bus_no);
  	host_pci_dev = pci_get_slot(host_pci_bus, devfn);
 
 	SN_PCIDEV_INFO(dev)->host_pci_dev = host_pci_dev;
@@ -333,6 +352,7 @@ void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus)
 	prom_bussoft_ptr = __va(prom_bussoft_ptr);
 
  	controller = kcalloc(1,sizeof(struct pci_controller), GFP_KERNEL);
+	controller->segment = segment;
  	if (!controller)
  		BUG();
 
@@ -390,7 +410,7 @@ void sn_pci_controller_fixup(int segment, int busnum, struct pci_bus *bus)
 	if (controller->node >= num_online_nodes()) {
 		struct pcibus_bussoft *b = SN_PCIBUS_BUSSOFT(bus);
 
-		printk(KERN_WARNING "Device ASIC=%u XID=%u PBUSNUM=%lu"
+		printk(KERN_WARNING "Device ASIC=%u XID=%u PBUSNUM=%u"
 				    "L_IO=%lx L_MEM=%lx BASE=%lx\n",
 			b->bs_asic_type, b->bs_xid, b->bs_persist_busnum,
 			b->bs_legacy_io, b->bs_legacy_mem, b->bs_base);
@@ -445,6 +465,7 @@ void sn_bus_free_sysdata(void)
 static int __init sn_pci_init(void)
 {
 	int i = 0;
+	int j = 0;
 	struct pci_dev *pci_dev = NULL;
 	extern void sn_init_cpei_timer(void);
 #ifdef CONFIG_PROC_FS
@@ -464,6 +485,7 @@ static int __init sn_pci_init(void)
 
 	pcibr_init_provider();
 	tioca_init_provider();
+	tioce_init_provider();
 
 	/*
 	 * This is needed to avoid bounce limit checks in the blk layer
@@ -479,8 +501,9 @@ static int __init sn_pci_init(void)
 #endif
 
 	/* busses are not known yet ... */
-	for (i = 0; i < PCI_BUSES_TO_SCAN; i++)
-		sn_pci_controller_fixup(0, i, NULL);
+	for (i = 0; i <= max_segment_number; i++)
+		for (j = 0; j <= max_pcibus_number; j++)
+			sn_pci_controller_fixup(i, j, NULL);
 
 	/*
 	 * Generic Linux PCI Layer has created the pci_bus and pci_dev 

arch/ia64/sn/kernel/irq.c

@@ -5,7 +5,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  */
 
 #include <linux/irq.h>
@@ -76,16 +76,14 @@ static void sn_enable_irq(unsigned int irq)
 
 static void sn_ack_irq(unsigned int irq)
 {
-	uint64_t event_occurred, mask = 0;
-	int nasid;
+	u64 event_occurred, mask = 0;
 
 	irq = irq & 0xff;
-	nasid = get_nasid();
 	event_occurred =
-	    HUB_L((uint64_t *) GLOBAL_MMR_ADDR(nasid, SH_EVENT_OCCURRED));
+	    HUB_L((u64*)LOCAL_MMR_ADDR(SH_EVENT_OCCURRED));
 	mask = event_occurred & SH_ALL_INT_MASK;
-	HUB_S((uint64_t *) GLOBAL_MMR_ADDR(nasid, SH_EVENT_OCCURRED_ALIAS),
-		 mask);
+	HUB_S((u64*)LOCAL_MMR_ADDR(SH_EVENT_OCCURRED_ALIAS),
+	      mask);
 	__set_bit(irq, (volatile void *)pda->sn_in_service_ivecs);
 
 	move_irq(irq);
@@ -93,15 +91,12 @@ static void sn_ack_irq(unsigned int irq)
 
 static void sn_end_irq(unsigned int irq)
 {
-	int nasid;
 	int ivec;
-	uint64_t event_occurred;
+	u64 event_occurred;
 
 	ivec = irq & 0xff;
 	if (ivec == SGI_UART_VECTOR) {
-		nasid = get_nasid();
-		event_occurred = HUB_L((uint64_t *) GLOBAL_MMR_ADDR
-				       (nasid, SH_EVENT_OCCURRED));
+		event_occurred = HUB_L((u64*)LOCAL_MMR_ADDR (SH_EVENT_OCCURRED));
 		/* If the UART bit is set here, we may have received an
 		 * interrupt from the UART that the driver missed.  To
 		 * make sure, we IPI ourselves to force us to look again.
@@ -132,6 +127,7 @@ static void sn_set_affinity_irq(unsigned int irq, cpumask_t mask)
 		int local_widget, status;
 		nasid_t local_nasid;
 		struct sn_irq_info *new_irq_info;
+		struct sn_pcibus_provider *pci_provider;
 
 		new_irq_info = kmalloc(sizeof(struct sn_irq_info), GFP_ATOMIC);
 		if (new_irq_info == NULL)
@@ -171,8 +167,9 @@ static void sn_set_affinity_irq(unsigned int irq, cpumask_t mask)
 		new_irq_info->irq_cpuid = cpuid;
 		register_intr_pda(new_irq_info);
 
-		if (IS_PCI_BRIDGE_ASIC(new_irq_info->irq_bridge_type))
-			pcibr_change_devices_irq(new_irq_info);
+		pci_provider = sn_pci_provider[new_irq_info->irq_bridge_type];
+		if (pci_provider && pci_provider->target_interrupt)
+			(pci_provider->target_interrupt)(new_irq_info);
 
 		spin_lock(&sn_irq_info_lock);
 		list_replace_rcu(&sn_irq_info->list, &new_irq_info->list);
@@ -317,6 +314,16 @@ void sn_irq_unfixup(struct pci_dev *pci_dev)
 	pci_dev_put(pci_dev);
 }
 
+static inline void
+sn_call_force_intr_provider(struct sn_irq_info *sn_irq_info)
+{
+	struct sn_pcibus_provider *pci_provider;
+
+	pci_provider = sn_pci_provider[sn_irq_info->irq_bridge_type];
+	if (pci_provider && pci_provider->force_interrupt)
+		(*pci_provider->force_interrupt)(sn_irq_info);
+}
+
 static void force_interrupt(int irq)
 {
 	struct sn_irq_info *sn_irq_info;
@@ -325,11 +332,9 @@ static void force_interrupt(int irq)
 		return;
 
 	rcu_read_lock();
-	list_for_each_entry_rcu(sn_irq_info, sn_irq_lh[irq], list) {
-		if (IS_PCI_BRIDGE_ASIC(sn_irq_info->irq_bridge_type) &&
-		    (sn_irq_info->irq_bridge != NULL))
-			pcibr_force_interrupt(sn_irq_info);
-	}
+	list_for_each_entry_rcu(sn_irq_info, sn_irq_lh[irq], list)
+		sn_call_force_intr_provider(sn_irq_info);
+
 	rcu_read_unlock();
 }
 
@@ -351,6 +356,14 @@ static void sn_check_intr(int irq, struct sn_irq_info *sn_irq_info)
 	struct pcidev_info *pcidev_info;
 	struct pcibus_info *pcibus_info;
 
+	/*
+	 * Bridge types attached to TIO (anything but PIC) do not need this WAR
+	 * since they do not target Shub II interrupt registers.  If that
+	 * ever changes, this check needs to accomodate.
+	 */
+	if (sn_irq_info->irq_bridge_type != PCIIO_ASIC_TYPE_PIC)
+		return;
+
 	pcidev_info = (struct pcidev_info *)sn_irq_info->irq_pciioinfo;
 	if (!pcidev_info)
 		return;
@@ -377,16 +390,12 @@ static void sn_check_intr(int irq, struct sn_irq_info *sn_irq_info)
 		break;
 	}
 	if (!test_bit(irr_bit, &irr_reg)) {
-		if (!test_bit(irq, pda->sn_soft_irr)) {
-			if (!test_bit(irq, pda->sn_in_service_ivecs)) {
-				regval &= 0xff;
-				if (sn_irq_info->irq_int_bit & regval &
-				    sn_irq_info->irq_last_intr) {
-					regval &=
-					    ~(sn_irq_info->
-					      irq_int_bit & regval);
-					pcibr_force_interrupt(sn_irq_info);
-				}
+		if (!test_bit(irq, pda->sn_in_service_ivecs)) {
+			regval &= 0xff;
+			if (sn_irq_info->irq_int_bit & regval &
+			    sn_irq_info->irq_last_intr) {
+				regval &= ~(sn_irq_info->irq_int_bit & regval);
+				sn_call_force_intr_provider(sn_irq_info);
 			}
 		}
 	}
@@ -404,13 +413,7 @@ void sn_lb_int_war_check(void)
 	rcu_read_lock();
 	for (i = pda->sn_first_irq; i <= pda->sn_last_irq; i++) {
 		list_for_each_entry_rcu(sn_irq_info, sn_irq_lh[i], list) {
-			/*
-			 * Only call for PCI bridges that are fully
-			 * initialized.
-			 */
-			if (IS_PCI_BRIDGE_ASIC(sn_irq_info->irq_bridge_type) &&
-			    (sn_irq_info->irq_bridge != NULL))
-				sn_check_intr(i, sn_irq_info);
+			sn_check_intr(i, sn_irq_info);
 		}
 	}
 	rcu_read_unlock();

arch/ia64/sn/kernel/setup.c

@@ -80,8 +80,6 @@ EXPORT_PER_CPU_SYMBOL(__sn_cnodeid_to_nasid);
 DEFINE_PER_CPU(struct nodepda_s *, __sn_nodepda);
 EXPORT_PER_CPU_SYMBOL(__sn_nodepda);
 
-partid_t sn_partid = -1;
-EXPORT_SYMBOL(sn_partid);
 char sn_system_serial_number_string[128];
 EXPORT_SYMBOL(sn_system_serial_number_string);
 u64 sn_partition_serial_number;
@@ -403,6 +401,7 @@ static void __init sn_init_pdas(char **cmdline_p)
 		memset(nodepdaindr[cnode], 0, sizeof(nodepda_t));
 		memset(nodepdaindr[cnode]->phys_cpuid, -1,
 		    sizeof(nodepdaindr[cnode]->phys_cpuid));
+		spin_lock_init(&nodepdaindr[cnode]->ptc_lock);
 	}
 
 	/*
@@ -532,8 +531,8 @@ void __init sn_cpu_init(void)
 	 */
 	{
 		u64 pio1[] = {SH1_PIO_WRITE_STATUS_0, 0, SH1_PIO_WRITE_STATUS_1, 0};
-		u64 pio2[] = {SH2_PIO_WRITE_STATUS_0, SH2_PIO_WRITE_STATUS_1,
-			SH2_PIO_WRITE_STATUS_2, SH2_PIO_WRITE_STATUS_3};
+		u64 pio2[] = {SH2_PIO_WRITE_STATUS_0, SH2_PIO_WRITE_STATUS_2,
+			SH2_PIO_WRITE_STATUS_1, SH2_PIO_WRITE_STATUS_3};
 		u64 *pio;
 		pio = is_shub1() ? pio1 : pio2;
 		pda->pio_write_status_addr = (volatile unsigned long *) LOCAL_MMR_ADDR(pio[slice]);

arch/ia64/sn/kernel/sn2/ptc_deadlock.S

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #include <asm/types.h>
@@ -11,7 +11,7 @@
 
 #define DEADLOCKBIT	SH_PIO_WRITE_STATUS_WRITE_DEADLOCK_SHFT
 #define WRITECOUNTMASK	SH_PIO_WRITE_STATUS_PENDING_WRITE_COUNT_MASK
-#define ALIAS_OFFSET	(SH1_PIO_WRITE_STATUS_0_ALIAS-SH1_PIO_WRITE_STATUS_0)
+#define ALIAS_OFFSET	8
 
 
 	.global	sn2_ptc_deadlock_recovery_core
@@ -36,13 +36,15 @@ sn2_ptc_deadlock_recovery_core:
 	extr.u	piowcphy=piowc,0,61;;	// Convert piowc to uncached physical address
 	dep	piowcphy=-1,piowcphy,63,1
 	movl	mask=WRITECOUNTMASK
+	mov	r8=r0
 
 1:
 	add	scr2=ALIAS_OFFSET,piowc	// Address of WRITE_STATUS alias register 
-	mov	scr1=7;;		// Clear DEADLOCK, WRITE_ERROR, MULTI_WRITE_ERROR
-	st8.rel	[scr2]=scr1;;
+	;;
+	ld8.acq	scr1=[scr2];;
 
 5:	ld8.acq	scr1=[piowc];;		// Wait for PIOs to complete.
+	hint	@pause
 	and	scr2=scr1,mask;;	// mask of writecount bits
 	cmp.ne	p6,p0=zeroval,scr2
 (p6)	br.cond.sptk 5b
@@ -57,6 +59,7 @@ sn2_ptc_deadlock_recovery_core:
 	st8.rel [ptc0]=data0		// Write PTC0 & wait for completion.
 
 5:	ld8.acq	scr1=[piowcphy];;	// Wait for PIOs to complete.
+	hint	@pause
 	and	scr2=scr1,mask;;	// mask of writecount bits
 	cmp.ne	p6,p0=zeroval,scr2
 (p6)	br.cond.sptk 5b;;
@@ -67,6 +70,7 @@ sn2_ptc_deadlock_recovery_core:
 (p7)	st8.rel [ptc1]=data1;;		// Now write PTC1.
 
 5:	ld8.acq	scr1=[piowcphy];;	// Wait for PIOs to complete.
+	hint	@pause
 	and	scr2=scr1,mask;;	// mask of writecount bits
 	cmp.ne	p6,p0=zeroval,scr2
 (p6)	br.cond.sptk 5b
@@ -77,6 +81,7 @@ sn2_ptc_deadlock_recovery_core:
 	srlz.i;;
 	////////////// END   PHYSICAL MODE ////////////////////
 
+(p8)	add	r8=1,r8
 (p8)	br.cond.spnt 1b;;		// Repeat if DEADLOCK occurred.
 
 	br.ret.sptk	rp

arch/ia64/sn/kernel/sn2/sn2_smp.c

@@ -5,7 +5,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #include <linux/init.h>
@@ -20,6 +20,8 @@
 #include <linux/module.h>
 #include <linux/bitops.h>
 #include <linux/nodemask.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
 
 #include <asm/processor.h>
 #include <asm/irq.h>
@@ -39,12 +41,120 @@
 #include <asm/sn/nodepda.h>
 #include <asm/sn/rw_mmr.h>
 
-void sn2_ptc_deadlock_recovery(volatile unsigned long *, unsigned long data0, 
-	volatile unsigned long *, unsigned long data1);
+DEFINE_PER_CPU(struct ptc_stats, ptcstats);
+DECLARE_PER_CPU(struct ptc_stats, ptcstats);
 
 static  __cacheline_aligned DEFINE_SPINLOCK(sn2_global_ptc_lock);
 
-static unsigned long sn2_ptc_deadlock_count;
+void sn2_ptc_deadlock_recovery(short *, short, int, volatile unsigned long *, unsigned long data0,
+	volatile unsigned long *, unsigned long data1);
+
+#ifdef DEBUG_PTC
+/*
+ * ptctest:
+ *
+ * 	xyz - 3 digit hex number:
+ * 		x - Force PTC purges to use shub:
+ * 			0 - no force
+ * 			1 - force
+ * 		y - interupt enable
+ * 			0 - disable interrupts
+ * 			1 - leave interuupts enabled
+ * 		z - type of lock:
+ * 			0 - global lock
+ * 			1 - node local lock
+ * 			2 - no lock
+ *
+ *   	Note: on shub1, only ptctest == 0 is supported. Don't try other values!
+ */
+
+static unsigned int sn2_ptctest = 0;
+
+static int __init ptc_test(char *str)
+{
+	get_option(&str, &sn2_ptctest);
+	return 1;
+}
+__setup("ptctest=", ptc_test);
+
+static inline int ptc_lock(unsigned long *flagp)
+{
+	unsigned long opt = sn2_ptctest & 255;
+
+	switch (opt) {
+	case 0x00:
+		spin_lock_irqsave(&sn2_global_ptc_lock, *flagp);
+		break;
+	case 0x01:
+		spin_lock_irqsave(&sn_nodepda->ptc_lock, *flagp);
+		break;
+	case 0x02:
+		local_irq_save(*flagp);
+		break;
+	case 0x10:
+		spin_lock(&sn2_global_ptc_lock);
+		break;
+	case 0x11:
+		spin_lock(&sn_nodepda->ptc_lock);
+		break;
+	case 0x12:
+		break;
+	default:
+		BUG();
+	}
+	return opt;
+}
+
+static inline void ptc_unlock(unsigned long flags, int opt)
+{
+	switch (opt) {
+	case 0x00:
+		spin_unlock_irqrestore(&sn2_global_ptc_lock, flags);
+		break;
+	case 0x01:
+		spin_unlock_irqrestore(&sn_nodepda->ptc_lock, flags);
+		break;
+	case 0x02:
+		local_irq_restore(flags);
+		break;
+	case 0x10:
+		spin_unlock(&sn2_global_ptc_lock);
+		break;
+	case 0x11:
+		spin_unlock(&sn_nodepda->ptc_lock);
+		break;
+	case 0x12:
+		break;
+	default:
+		BUG();
+	}
+}
+#else
+
+#define sn2_ptctest	0
+
+static inline int ptc_lock(unsigned long *flagp)
+{
+	spin_lock_irqsave(&sn2_global_ptc_lock, *flagp);
+	return 0;
+}
+
+static inline void ptc_unlock(unsigned long flags, int opt)
+{
+	spin_unlock_irqrestore(&sn2_global_ptc_lock, flags);
+}
+#endif
+
+struct ptc_stats {
+	unsigned long ptc_l;
+	unsigned long change_rid;
+	unsigned long shub_ptc_flushes;
+	unsigned long nodes_flushed;
+	unsigned long deadlocks;
+	unsigned long lock_itc_clocks;
+	unsigned long shub_itc_clocks;
+	unsigned long shub_itc_clocks_max;
+};
 
 static inline unsigned long wait_piowc(void)
 {
@@ -89,9 +199,9 @@ void
 sn2_global_tlb_purge(unsigned long start, unsigned long end,
 		     unsigned long nbits)
 {
-	int i, shub1, cnode, mynasid, cpu, lcpu = 0, nasid, flushed = 0;
+	int i, opt, shub1, cnode, mynasid, cpu, lcpu = 0, nasid, flushed = 0;
 	volatile unsigned long *ptc0, *ptc1;
-	unsigned long flags = 0, data0 = 0, data1 = 0;
+	unsigned long itc, itc2, flags, data0 = 0, data1 = 0;
 	struct mm_struct *mm = current->active_mm;
 	short nasids[MAX_NUMNODES], nix;
 	nodemask_t nodes_flushed;
@@ -114,16 +224,19 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
 			start += (1UL << nbits);
 		} while (start < end);
 		ia64_srlz_i();
+		__get_cpu_var(ptcstats).ptc_l++;
 		preempt_enable();
 		return;
 	}
 
 	if (atomic_read(&mm->mm_users) == 1) {
 		flush_tlb_mm(mm);
+		__get_cpu_var(ptcstats).change_rid++;
 		preempt_enable();
 		return;
 	}
 
+	itc = ia64_get_itc();
 	nix = 0;
 	for_each_node_mask(cnode, nodes_flushed)
 		nasids[nix++] = cnodeid_to_nasid(cnode);
@@ -148,7 +261,12 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
 
 	mynasid = get_nasid();
 
-	spin_lock_irqsave(&sn2_global_ptc_lock, flags);
+	itc = ia64_get_itc();
+	opt = ptc_lock(&flags);
+	itc2 = ia64_get_itc();
+	__get_cpu_var(ptcstats).lock_itc_clocks += itc2 - itc;
+	__get_cpu_var(ptcstats).shub_ptc_flushes++;
+	__get_cpu_var(ptcstats).nodes_flushed += nix;
 
 	do {
 		if (shub1)
@@ -157,7 +275,7 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
 			data0 = (data0 & ~SH2_PTC_ADDR_MASK) | (start & SH2_PTC_ADDR_MASK);
 		for (i = 0; i < nix; i++) {
 			nasid = nasids[i];
-			if (unlikely(nasid == mynasid)) {
+			if ((!(sn2_ptctest & 3)) && unlikely(nasid == mynasid)) {
 				ia64_ptcga(start, nbits << 2);
 				ia64_srlz_i();
 			} else {
@@ -169,18 +287,22 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
 				flushed = 1;
 			}
 		}
-
 		if (flushed
 		    && (wait_piowc() &
-			SH_PIO_WRITE_STATUS_WRITE_DEADLOCK_MASK)) {
-			sn2_ptc_deadlock_recovery(ptc0, data0, ptc1, data1);
+				(SH_PIO_WRITE_STATUS_WRITE_DEADLOCK_MASK))) {
+			sn2_ptc_deadlock_recovery(nasids, nix, mynasid, ptc0, data0, ptc1, data1);
 		}
 
 		start += (1UL << nbits);
 
 	} while (start < end);
 
-	spin_unlock_irqrestore(&sn2_global_ptc_lock, flags);
+	itc2 = ia64_get_itc() - itc2;
+	__get_cpu_var(ptcstats).shub_itc_clocks += itc2;
+	if (itc2 > __get_cpu_var(ptcstats).shub_itc_clocks_max)
+		__get_cpu_var(ptcstats).shub_itc_clocks_max = itc2;
+
+	ptc_unlock(flags, opt);
 
 	preempt_enable();
 }
@@ -192,31 +314,29 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
  * TLB flush transaction.  The recovery sequence is somewhat tricky & is
  * coded in assembly language.
  */
-void sn2_ptc_deadlock_recovery(volatile unsigned long *ptc0, unsigned long data0,
+void sn2_ptc_deadlock_recovery(short *nasids, short nix, int mynasid, volatile unsigned long *ptc0, unsigned long data0,
 	volatile unsigned long *ptc1, unsigned long data1)
 {
 	extern void sn2_ptc_deadlock_recovery_core(volatile unsigned long *, unsigned long,
 	        volatile unsigned long *, unsigned long, volatile unsigned long *, unsigned long);
-	int cnode, mycnode, nasid;
-	volatile unsigned long *piows;
-	volatile unsigned long zeroval;
+	short nasid, i;
+	unsigned long *piows, zeroval;
 
-	sn2_ptc_deadlock_count++;
+	__get_cpu_var(ptcstats).deadlocks++;
 
-	piows = pda->pio_write_status_addr;
+	piows = (unsigned long *) pda->pio_write_status_addr;
 	zeroval = pda->pio_write_status_val;
 
-	mycnode = numa_node_id();
-
-	for_each_online_node(cnode) {
-		if (is_headless_node(cnode) || cnode == mycnode)
+	for (i=0; i < nix; i++) {
+		nasid = nasids[i];
+		if (!(sn2_ptctest & 3) && nasid == mynasid)
 			continue;
-		nasid = cnodeid_to_nasid(cnode);
 		ptc0 = CHANGE_NASID(nasid, ptc0);
 		if (ptc1)
 			ptc1 = CHANGE_NASID(nasid, ptc1);
 		sn2_ptc_deadlock_recovery_core(ptc0, data0, ptc1, data1, piows, zeroval);
 	}
+
 }
 
 /**
@@ -293,3 +413,93 @@ void sn2_send_IPI(int cpuid, int vector, int delivery_mode, int redirect)
 
 	sn_send_IPI_phys(nasid, physid, vector, delivery_mode);
 }
+
+#ifdef CONFIG_PROC_FS
+
+#define PTC_BASENAME	"sgi_sn/ptc_statistics"
+
+static void *sn2_ptc_seq_start(struct seq_file *file, loff_t * offset)
+{
+	if (*offset < NR_CPUS)
+		return offset;
+	return NULL;
+}
+
+static void *sn2_ptc_seq_next(struct seq_file *file, void *data, loff_t * offset)
+{
+	(*offset)++;
+	if (*offset < NR_CPUS)
+		return offset;
+	return NULL;
+}
+
+static void sn2_ptc_seq_stop(struct seq_file *file, void *data)
+{
+}
+
+static int sn2_ptc_seq_show(struct seq_file *file, void *data)
+{
+	struct ptc_stats *stat;
+	int cpu;
+
+	cpu = *(loff_t *) data;
+
+	if (!cpu) {
+		seq_printf(file, "# ptc_l change_rid shub_ptc_flushes shub_nodes_flushed deadlocks lock_nsec shub_nsec shub_nsec_max\n");
+		seq_printf(file, "# ptctest %d\n", sn2_ptctest);
+	}
+
+	if (cpu < NR_CPUS && cpu_online(cpu)) {
+		stat = &per_cpu(ptcstats, cpu);
+		seq_printf(file, "cpu %d %ld %ld %ld %ld %ld %ld %ld %ld\n", cpu, stat->ptc_l,
+				stat->change_rid, stat->shub_ptc_flushes, stat->nodes_flushed,
+				stat->deadlocks,
+				1000 * stat->lock_itc_clocks / per_cpu(cpu_info, cpu).cyc_per_usec,
+				1000 * stat->shub_itc_clocks / per_cpu(cpu_info, cpu).cyc_per_usec,
+				1000 * stat->shub_itc_clocks_max / per_cpu(cpu_info, cpu).cyc_per_usec);
+	}
+
+	return 0;
+}
+
+static struct seq_operations sn2_ptc_seq_ops = {
+	.start = sn2_ptc_seq_start,
+	.next = sn2_ptc_seq_next,
+	.stop = sn2_ptc_seq_stop,
+	.show = sn2_ptc_seq_show
+};
+
+int sn2_ptc_proc_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &sn2_ptc_seq_ops);
+}
+
+static struct file_operations proc_sn2_ptc_operations = {
+	.open = sn2_ptc_proc_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = seq_release,
+};
+
+static struct proc_dir_entry *proc_sn2_ptc;
+
+static int __init sn2_ptc_init(void)
+{
+	if (!(proc_sn2_ptc = create_proc_entry(PTC_BASENAME, 0444, NULL))) {
+		printk(KERN_ERR "unable to create %s proc entry", PTC_BASENAME);
+		return -EINVAL;
+	}
+	proc_sn2_ptc->proc_fops = &proc_sn2_ptc_operations;
+	spin_lock_init(&sn2_global_ptc_lock);
+	return 0;
+}
+
+static void __exit sn2_ptc_exit(void)
+{
+	remove_proc_entry(PTC_BASENAME, NULL);
+}
+
+module_init(sn2_ptc_init);
+module_exit(sn2_ptc_exit);
+#endif /* CONFIG_PROC_FS */
+

arch/ia64/sn/kernel/sn2/sn_hwperf.c

@@ -36,7 +36,6 @@
 #include <asm/topology.h>
 #include <asm/smp.h>
 #include <asm/semaphore.h>
-#include <asm/segment.h>
 #include <asm/uaccess.h>
 #include <asm/sal.h>
 #include <asm/sn/io.h>
@@ -59,7 +58,7 @@ static int sn_hwperf_enum_objects(int *nobj, struct sn_hwperf_object_info **ret)
 	struct sn_hwperf_object_info *objbuf = NULL;
 
 	if ((e = sn_hwperf_init()) < 0) {
-		printk("sn_hwperf_init failed: err %d\n", e);
+		printk(KERN_ERR "sn_hwperf_init failed: err %d\n", e);
 		goto out;
 	}
 
@@ -111,7 +110,7 @@ static int sn_hwperf_geoid_to_cnode(char *location)
 	if (sn_hwperf_location_to_bpos(location, &rack, &bay, &slot, &slab))
 		return -1;
 
-	for (cnode = 0; cnode < numionodes; cnode++) {
+	for_each_node(cnode) {
 		geoid = cnodeid_get_geoid(cnode);
 		module_id = geo_module(geoid);
 		this_rack = MODULE_GET_RACK(module_id);
@@ -124,11 +123,13 @@ static int sn_hwperf_geoid_to_cnode(char *location)
 		}
 	}
 
-	return cnode < numionodes ? cnode : -1;
+	return node_possible(cnode) ? cnode : -1;
 }
 
 static int sn_hwperf_obj_to_cnode(struct sn_hwperf_object_info * obj)
 {
+	if (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj))
+		BUG();
 	if (!obj->sn_hwp_this_part)
 		return -1;
 	return sn_hwperf_geoid_to_cnode(obj->location);
@@ -174,31 +175,199 @@ static const char *sn_hwperf_get_slabname(struct sn_hwperf_object_info *obj,
 	return slabname;
 }
 
-static void print_pci_topology(struct seq_file *s,
-	struct sn_hwperf_object_info *obj, int *ordinal,
-	u64 rack, u64 bay, u64 slot, u64 slab)
+static void print_pci_topology(struct seq_file *s)
+{
+	char *p;
+	size_t sz;
+	int e;
+
+	for (sz = PAGE_SIZE; sz < 16 * PAGE_SIZE; sz += PAGE_SIZE) {
+		if (!(p = (char *)kmalloc(sz, GFP_KERNEL)))
+			break;
+		e = ia64_sn_ioif_get_pci_topology(__pa(p), sz);
+		if (e == SALRET_OK)
+			seq_puts(s, p);
+		kfree(p);
+		if (e == SALRET_OK || e == SALRET_NOT_IMPLEMENTED)
+			break;
+	}
+}
+
+static inline int sn_hwperf_has_cpus(cnodeid_t node)
+{
+	return node_online(node) && nr_cpus_node(node);
+}
+
+static inline int sn_hwperf_has_mem(cnodeid_t node)
+{
+	return node_online(node) && NODE_DATA(node)->node_present_pages;
+}
+
+static struct sn_hwperf_object_info *
+sn_hwperf_findobj_id(struct sn_hwperf_object_info *objbuf,
+	int nobj, int id)
 {
-	char *p1;
-	char *p2;
-	char *pg;
-
-	if (!(pg = (char *)get_zeroed_page(GFP_KERNEL)))
-		return; /* ignore */
-	if (ia64_sn_ioif_get_pci_topology(rack, bay, slot, slab,
-		__pa(pg), PAGE_SIZE) == SN_HWPERF_OP_OK) {
-		for (p1=pg; *p1 && p1 < pg + PAGE_SIZE;) {
-			if (!(p2 = strchr(p1, '\n')))
+	int i;
+	struct sn_hwperf_object_info *p = objbuf;
+
+	for (i=0; i < nobj; i++, p++) {
+		if (p->id == id)
+			return p;
+	}
+
+	return NULL;
+
+}
+
+static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objbuf,
+	int nobj, cnodeid_t node, cnodeid_t *near_mem_node, cnodeid_t *near_cpu_node)
+{
+	int e;
+	struct sn_hwperf_object_info *nodeobj = NULL;
+	struct sn_hwperf_object_info *op;
+	struct sn_hwperf_object_info *dest;
+	struct sn_hwperf_object_info *router;
+	struct sn_hwperf_port_info ptdata[16];
+	int sz, i, j;
+	cnodeid_t c;
+	int found_mem = 0;
+	int found_cpu = 0;
+
+	if (!node_possible(node))
+		return -EINVAL;
+
+	if (sn_hwperf_has_cpus(node)) {
+		if (near_cpu_node)
+			*near_cpu_node = node;
+		found_cpu++;
+	}
+
+	if (sn_hwperf_has_mem(node)) {
+		if (near_mem_node)
+			*near_mem_node = node;
+		found_mem++;
+	}
+
+	if (found_cpu && found_mem)
+		return 0; /* trivially successful */
+
+	/* find the argument node object */
+	for (i=0, op=objbuf; i < nobj; i++, op++) {
+		if (!SN_HWPERF_IS_NODE(op) && !SN_HWPERF_IS_IONODE(op))
+			continue;
+		if (node == sn_hwperf_obj_to_cnode(op)) {
+			nodeobj = op;
+			break;
+		}
+	}
+	if (!nodeobj) {
+		e = -ENOENT;
+		goto err;
+	}
+
+	/* get it's interconnect topology */
+	sz = op->ports * sizeof(struct sn_hwperf_port_info);
+	if (sz > sizeof(ptdata))
+		BUG();
+	e = ia64_sn_hwperf_op(sn_hwperf_master_nasid,
+			      SN_HWPERF_ENUM_PORTS, nodeobj->id, sz,
+			      (u64)&ptdata, 0, 0, NULL);
+	if (e != SN_HWPERF_OP_OK) {
+		e = -EINVAL;
+		goto err;
+	}
+
+	/* find nearest node with cpus and nearest memory */
+	for (router=NULL, j=0; j < op->ports; j++) {
+		dest = sn_hwperf_findobj_id(objbuf, nobj, ptdata[j].conn_id);
+		if (!dest || SN_HWPERF_FOREIGN(dest) ||
+		    !SN_HWPERF_IS_NODE(dest) || SN_HWPERF_IS_IONODE(dest)) {
+			continue;
+		}
+		c = sn_hwperf_obj_to_cnode(dest);
+		if (!found_cpu && sn_hwperf_has_cpus(c)) {
+			if (near_cpu_node)
+				*near_cpu_node = c;
+			found_cpu++;
+		}
+		if (!found_mem && sn_hwperf_has_mem(c)) {
+			if (near_mem_node)
+				*near_mem_node = c;
+			found_mem++;
+		}
+		if (SN_HWPERF_IS_ROUTER(dest))
+			router = dest;
+	}
+
+	if (router && (!found_cpu || !found_mem)) {
+		/* search for a node connected to the same router */
+		sz = router->ports * sizeof(struct sn_hwperf_port_info);
+		if (sz > sizeof(ptdata))
+			BUG();
+		e = ia64_sn_hwperf_op(sn_hwperf_master_nasid,
+				      SN_HWPERF_ENUM_PORTS, router->id, sz,
+				      (u64)&ptdata, 0, 0, NULL);
+		if (e != SN_HWPERF_OP_OK) {
+			e = -EINVAL;
+			goto err;
+		}
+		for (j=0; j < router->ports; j++) {
+			dest = sn_hwperf_findobj_id(objbuf, nobj,
+				ptdata[j].conn_id);
+			if (!dest || dest->id == node ||
+			    SN_HWPERF_FOREIGN(dest) ||
+			    !SN_HWPERF_IS_NODE(dest) ||
+			    SN_HWPERF_IS_IONODE(dest)) {
+				continue;
+			}
+			c = sn_hwperf_obj_to_cnode(dest);
+			if (!found_cpu && sn_hwperf_has_cpus(c)) {
+				if (near_cpu_node)
+					*near_cpu_node = c;
+				found_cpu++;
+			}
+			if (!found_mem && sn_hwperf_has_mem(c)) {
+				if (near_mem_node)
+					*near_mem_node = c;
+				found_mem++;
+			}
+			if (found_cpu && found_mem)
+				break;
+		}
+	}
+
+	if (!found_cpu || !found_mem) {
+		/* resort to _any_ node with CPUs and memory */
+		for (i=0, op=objbuf; i < nobj; i++, op++) {
+			if (SN_HWPERF_FOREIGN(op) ||
+			    SN_HWPERF_IS_IONODE(op) ||
+			    !SN_HWPERF_IS_NODE(op)) {
+				continue;
+			}
+			c = sn_hwperf_obj_to_cnode(op);
+			if (!found_cpu && sn_hwperf_has_cpus(c)) {
+				if (near_cpu_node)
+					*near_cpu_node = c;
+				found_cpu++;
+			}
+			if (!found_mem && sn_hwperf_has_mem(c)) {
+				if (near_mem_node)
+					*near_mem_node = c;
+				found_mem++;
+			}
+			if (found_cpu && found_mem)
 				break;
-			*p2 = '\0';
-			seq_printf(s, "pcibus %d %s-%s\n",
-				*ordinal, obj->location, p1);
-			(*ordinal)++;
-			p1 = p2 + 1;
 		}
 	}
-	free_page((unsigned long)pg);
+
+	if (!found_cpu || !found_mem)
+		e = -ENODATA;
+
+err:
+	return e;
 }
 
+
 static int sn_topology_show(struct seq_file *s, void *d)
 {
 	int sz;
@@ -215,7 +384,6 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	struct sn_hwperf_object_info *p;
 	struct sn_hwperf_object_info *obj = d;	/* this object */
 	struct sn_hwperf_object_info *objs = s->private; /* all objects */
-	int rack, bay, slot, slab;
 	u8 shubtype;
 	u8 system_size;
 	u8 sharing_size;
@@ -225,7 +393,6 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	u8 region_size;
 	u16 nasid_mask;
 	int nasid_msb;
-	int pci_bus_ordinal = 0;
 
 	if (obj == objs) {
 		seq_printf(s, "# sn_topology version 2\n");
@@ -253,6 +420,8 @@ static int sn_topology_show(struct seq_file *s, void *d)
 			shubtype ? "shub2" : "shub1", 
 			(u64)nasid_mask << nasid_shift, nasid_msb, nasid_shift,
 			system_size, sharing_size, coher, region_size);
+
+		print_pci_topology(s);
 	}
 
 	if (SN_HWPERF_FOREIGN(obj)) {
@@ -272,11 +441,24 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	if (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj))
 		seq_putc(s, '\n');
 	else {
+		cnodeid_t near_mem = -1;
+		cnodeid_t near_cpu = -1;
+
 		seq_printf(s, ", nasid 0x%x", cnodeid_to_nasid(ordinal));
-		for (i=0; i < numionodes; i++) {
-			seq_printf(s, i ? ":%d" : ", dist %d",
-				node_distance(ordinal, i));
+
+		if (sn_hwperf_get_nearest_node_objdata(objs, sn_hwperf_obj_cnt,
+			ordinal, &near_mem, &near_cpu) == 0) {
+			seq_printf(s, ", near_mem_nodeid %d, near_cpu_nodeid %d",
+				near_mem, near_cpu);
+		}
+
+		if (!SN_HWPERF_IS_IONODE(obj)) {
+			for_each_online_node(i) {
+				seq_printf(s, i ? ":%d" : ", dist %d",
+					node_distance(ordinal, i));
+			}
 		}
+
 		seq_putc(s, '\n');
 
 		/*
@@ -300,17 +482,6 @@ static int sn_topology_show(struct seq_file *s, void *d)
 				seq_putc(s, '\n');
 			}
 		}
-
-		/*
-		 * PCI busses attached to this node, if any
-		 */
-		if (sn_hwperf_location_to_bpos(obj->location,
-			&rack, &bay, &slot, &slab)) {
-			/* export pci bus info */
-			print_pci_topology(s, obj, &pci_bus_ordinal,
-				rack, bay, slot, slab);
-
-		}
 	}
 
 	if (obj->ports) {
@@ -572,6 +743,8 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 		if ((r = sn_hwperf_enum_objects(&nobj, &objs)) == 0) {
 			memset(p, 0, a.sz);
 			for (i = 0; i < nobj; i++) {
+				if (!SN_HWPERF_IS_NODE(objs + i))
+					continue;
 				node = sn_hwperf_obj_to_cnode(objs + i);
 				for_each_online_cpu(j) {
 					if (node != cpu_to_node(j))
@@ -598,7 +771,7 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 
 	case SN_HWPERF_GET_NODE_NASID:
 		if (a.sz != sizeof(u64) ||
-		   (node = a.arg) < 0 || node >= numionodes) {
+		   (node = a.arg) < 0 || !node_possible(node)) {
 			r = -EINVAL;
 			goto error;
 		}
@@ -627,6 +800,14 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 				vfree(objs);
 				goto error;
 			}
+
+			if (!SN_HWPERF_IS_NODE(objs + i) &&
+			    !SN_HWPERF_IS_IONODE(objs + i)) {
+			    	r = -ENOENT;
+				vfree(objs);
+				goto error;
+			}
+
 			*(u64 *)p = (u64)sn_hwperf_obj_to_cnode(objs + i);
 			vfree(objs);
 		}
@@ -692,6 +873,7 @@ static int sn_hwperf_init(void)
 
 	/* single threaded, once-only initialization */
 	down(&sn_hwperf_init_mutex);
+
 	if (sn_hwperf_salheap) {
 		up(&sn_hwperf_init_mutex);
 		return e;
@@ -742,19 +924,6 @@ static int sn_hwperf_init(void)
 		sn_hwperf_salheap = NULL;
 		sn_hwperf_obj_cnt = 0;
 	}
-
-	if (!e) {
-		/*
-		 * Register a dynamic misc device for ioctl. Platforms
-		 * supporting hotplug will create /dev/sn_hwperf, else
-		 * user can to look up the minor number in /proc/misc.
-		 */
-		if ((e = misc_register(&sn_hwperf_dev)) != 0) {
-			printk(KERN_ERR "sn_hwperf_init: misc register "
-			       "for \"sn_hwperf\" failed, err %d\n", e);
-		}
-	}
-
 	up(&sn_hwperf_init_mutex);
 	return e;
 }
@@ -782,3 +951,41 @@ int sn_topology_release(struct inode *inode, struct file *file)
 	vfree(seq->private);
 	return seq_release(inode, file);
 }
+
+int sn_hwperf_get_nearest_node(cnodeid_t node,
+	cnodeid_t *near_mem_node, cnodeid_t *near_cpu_node)
+{
+	int e;
+	int nobj;
+	struct sn_hwperf_object_info *objbuf;
+
+	if ((e = sn_hwperf_enum_objects(&nobj, &objbuf)) == 0) {
+		e = sn_hwperf_get_nearest_node_objdata(objbuf, nobj,
+			node, near_mem_node, near_cpu_node);
+		vfree(objbuf);
+	}
+
+	return e;
+}
+
+static int __devinit sn_hwperf_misc_register_init(void)
+{
+	int e;
+
+	sn_hwperf_init();
+
+	/*
+	 * Register a dynamic misc device for hwperf ioctls. Platforms
+	 * supporting hotplug will create /dev/sn_hwperf, else user
+	 * can to look up the minor number in /proc/misc.
+	 */
+	if ((e = misc_register(&sn_hwperf_dev)) != 0) {
+		printk(KERN_ERR "sn_hwperf_misc_register_init: failed to "
+		"register misc device for \"%s\"\n", sn_hwperf_dev.name);
+	}
+
+	return e;
+}
+
+device_initcall(sn_hwperf_misc_register_init); /* after misc_init() */
+EXPORT_SYMBOL(sn_hwperf_get_nearest_node);

arch/ia64/sn/kernel/sn2/sn_proc_fs.c

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 #include <linux/config.h>
 #include <asm/uaccess.h>
@@ -15,7 +15,7 @@
 
 static int partition_id_show(struct seq_file *s, void *p)
 {
-	seq_printf(s, "%d\n", sn_local_partid());
+	seq_printf(s, "%d\n", sn_partition_id);
 	return 0;
 }
 

arch/ia64/sn/kernel/sn2/timer_interrupt.c

 /*
  *
  *
- * Copyright (c) 2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2005 Silicon Graphics, Inc.  All Rights Reserved.
  * 
  * This program is free software; you can redistribute it and/or modify it 
  * under the terms of version 2 of the GNU General Public License 
@@ -50,14 +50,16 @@ void sn_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
 			     LED_CPU_HEARTBEAT, LED_CPU_HEARTBEAT);
 	}
 
-	if (enable_shub_wars_1_1()) {
-		/* Bugfix code for SHUB 1.1 */
-		if (pda->pio_shub_war_cam_addr)
-			*pda->pio_shub_war_cam_addr = 0x8000000000000010UL;
+	if (is_shub1()) {
+		if (enable_shub_wars_1_1()) {
+			/* Bugfix code for SHUB 1.1 */
+			if (pda->pio_shub_war_cam_addr)
+				*pda->pio_shub_war_cam_addr = 0x8000000000000010UL;
+		}
+		if (pda->sn_lb_int_war_ticks == 0)
+			sn_lb_int_war_check();
+		pda->sn_lb_int_war_ticks++;
+		if (pda->sn_lb_int_war_ticks >= SN_LB_INT_WAR_INTERVAL)
+			pda->sn_lb_int_war_ticks = 0;
 	}
-	if (pda->sn_lb_int_war_ticks == 0)
-		sn_lb_int_war_check();
-	pda->sn_lb_int_war_ticks++;
-	if (pda->sn_lb_int_war_ticks >= SN_LB_INT_WAR_INTERVAL)
-		pda->sn_lb_int_war_ticks = 0;
 }

arch/ia64/sn/pci/Makefile

@@ -7,4 +7,4 @@
 #
 # Makefile for the sn pci general routines.
 
-obj-y := pci_dma.o tioca_provider.o pcibr/ 
+obj-y := pci_dma.o tioca_provider.o tioce_provider.o pcibr/

arch/ia64/sn/pci/pcibr/pcibr_dma.c

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2001-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2001-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #include <linux/types.h>
@@ -215,8 +215,8 @@ void sn_dma_flush(uint64_t addr)
 	int is_tio;
 	int wid_num;
 	int i, j;
-	int bwin;
 	uint64_t flags;
+	uint64_t itte;
 	struct hubdev_info *hubinfo;
 	volatile struct sn_flush_device_list *p;
 	struct sn_flush_nasid_entry *flush_nasid_list;
@@ -233,31 +233,36 @@ void sn_dma_flush(uint64_t addr)
 	if (!hubinfo) {
 		BUG();
 	}
-	is_tio = (nasid & 1);
-	if (is_tio) {
-		wid_num = TIO_SWIN_WIDGETNUM(addr);
-		bwin = TIO_BWIN_WINDOWNUM(addr);
-	} else {
-		wid_num = SWIN_WIDGETNUM(addr);
-		bwin = BWIN_WINDOWNUM(addr);
-	}
 
 	flush_nasid_list = &hubinfo->hdi_flush_nasid_list;
 	if (flush_nasid_list->widget_p == NULL)
 		return;
-	if (bwin > 0) {
-		uint64_t itte = flush_nasid_list->iio_itte[bwin];
 
-		if (is_tio) {
-			wid_num = (itte >> TIO_ITTE_WIDGET_SHIFT) &
-			    TIO_ITTE_WIDGET_MASK;
-		} else {
-			wid_num = (itte >> IIO_ITTE_WIDGET_SHIFT) &
-			    IIO_ITTE_WIDGET_MASK;
-		}
+	is_tio = (nasid & 1);
+	if (is_tio) {
+		int itte_index;
+
+		if (TIO_HWIN(addr))
+			itte_index = 0;
+		else if (TIO_BWIN_WINDOWNUM(addr))
+			itte_index = TIO_BWIN_WINDOWNUM(addr);
+		else
+			itte_index = -1;
+
+		if (itte_index >= 0) {
+			itte = flush_nasid_list->iio_itte[itte_index];
+			if (! TIO_ITTE_VALID(itte))
+				return;
+			wid_num = TIO_ITTE_WIDGET(itte);
+		} else
+			wid_num = TIO_SWIN_WIDGETNUM(addr);
+	} else {
+		if (BWIN_WINDOWNUM(addr)) {
+			itte = flush_nasid_list->iio_itte[BWIN_WINDOWNUM(addr)];
+			wid_num = IIO_ITTE_WIDGET(itte);
+		} else
+			wid_num = SWIN_WIDGETNUM(addr);
 	}
-	if (flush_nasid_list->widget_p == NULL)
-		return;
 	if (flush_nasid_list->widget_p[wid_num] == NULL)
 		return;
 	p = &flush_nasid_list->widget_p[wid_num][0];
@@ -283,10 +288,16 @@ void sn_dma_flush(uint64_t addr)
 	/*
 	 * For TIOCP use the Device(x) Write Request Buffer Flush Bridge
 	 * register since it ensures the data has entered the coherence
-	 * domain, unlike PIC
+	 * domain, unlike PIC.
 	 */
 	if (is_tio) {
-		uint32_t tio_id = REMOTE_HUB_L(nasid, TIO_NODE_ID);
+		/*
+	 	 * Note:  devices behind TIOCE should never be matched in the
+		 * above code, and so the following code is PIC/CP centric.
+		 * If CE ever needs the sn_dma_flush mechanism, we will have
+		 * to account for that here and in tioce_bus_fixup().
+	 	 */
+		uint32_t tio_id = HUB_L(TIO_IOSPACE_ADDR(nasid, TIO_NODE_ID));
 		uint32_t revnum = XWIDGET_PART_REV_NUM(tio_id);
 
 		/* TIOCP BRINGUP WAR (PV907516): Don't write buffer flush reg */
@@ -306,7 +317,8 @@ void sn_dma_flush(uint64_t addr)
 		*(volatile uint32_t *)(p->sfdl_force_int_addr) = 1;
 
 		/* wait for the interrupt to come back. */
-		while (*(p->sfdl_flush_addr) != 0x10f) ;
+		while (*(p->sfdl_flush_addr) != 0x10f)
+			cpu_relax();
 
 		/* okay, everything is synched up. */
 		spin_unlock_irqrestore((spinlock_t *)&p->sfdl_flush_lock, flags);

arch/ia64/sn/pci/pcibr/pcibr_provider.c

@@ -15,6 +15,7 @@
 #include <asm/sn/pcibus_provider_defs.h>
 #include <asm/sn/pcidev.h>
 #include <asm/sn/sn_sal.h>
+#include <asm/sn/sn2/sn_hwperf.h>
 #include "xtalk/xwidgetdev.h"
 #include "xtalk/hubdev.h"
 
@@ -60,7 +61,7 @@ static int sal_pcibr_error_interrupt(struct pcibus_info *soft)
 	ret_stuff.status = 0;
 	ret_stuff.v0 = 0;
 
-	segment = 0;
+	segment = soft->pbi_buscommon.bs_persist_segment;
 	busnum = soft->pbi_buscommon.bs_persist_busnum;
 	SAL_CALL_NOLOCK(ret_stuff,
 			(u64) SN_SAL_IOIF_ERROR_INTERRUPT,
@@ -88,6 +89,7 @@ void *
 pcibr_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *controller)
 {
 	int nasid, cnode, j;
+	cnodeid_t near_cnode;
 	struct hubdev_info *hubdev_info;
 	struct pcibus_info *soft;
 	struct sn_flush_device_list *sn_flush_device_list;
@@ -115,7 +117,7 @@ pcibr_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont
 	/*
 	 * register the bridge's error interrupt handler
 	 */
-	if (request_irq(SGI_PCIBR_ERROR, (void *)pcibr_error_intr_handler,
+	if (request_irq(SGI_PCIASIC_ERROR, (void *)pcibr_error_intr_handler,
 			SA_SHIRQ, "PCIBR error", (void *)(soft))) {
 		printk(KERN_WARNING
 		       "pcibr cannot allocate interrupt for error handler\n");
@@ -142,9 +144,12 @@ pcibr_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont
 			     j++, sn_flush_device_list++) {
 				if (sn_flush_device_list->sfdl_slot == -1)
 					continue;
-				if (sn_flush_device_list->
-				    sfdl_persistent_busnum ==
-				    soft->pbi_buscommon.bs_persist_busnum)
+				if ((sn_flush_device_list->
+				     sfdl_persistent_segment ==
+				     soft->pbi_buscommon.bs_persist_segment) &&
+				     (sn_flush_device_list->
+				     sfdl_persistent_busnum ==
+				     soft->pbi_buscommon.bs_persist_busnum))
 					sn_flush_device_list->sfdl_pcibus_info =
 					    soft;
 			}
@@ -158,12 +163,18 @@ pcibr_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont
 	memset(soft->pbi_int_ate_resource.ate, 0,
  	       (soft->pbi_int_ate_size * sizeof(uint64_t)));
 
-	if (prom_bussoft->bs_asic_type == PCIIO_ASIC_TYPE_TIOCP)
-		/*
-		 * TIO PCI Bridge with no closest node information.
-		 * FIXME: Find another way to determine the closest node
-		 */
-		controller->node = -1;
+	if (prom_bussoft->bs_asic_type == PCIIO_ASIC_TYPE_TIOCP) {
+		/* TIO PCI Bridge: find nearest node with CPUs */
+		int e = sn_hwperf_get_nearest_node(cnode, NULL, &near_cnode);
+
+		if (e < 0) {
+			near_cnode = (cnodeid_t)-1; /* use any node */
+			printk(KERN_WARNING "pcibr_bus_fixup: failed to find "
+				"near node with CPUs to TIO node %d, err=%d\n",
+				cnode, e);
+		}
+		controller->node = near_cnode;
+	}
 	else
 		controller->node = cnode;
 	return soft;
@@ -175,6 +186,9 @@ void pcibr_force_interrupt(struct sn_irq_info *sn_irq_info)
 	struct pcibus_info *pcibus_info;
 	int bit = sn_irq_info->irq_int_bit;
 
+	if (! sn_irq_info->irq_bridge)
+		return;
+
 	pcidev_info = (struct pcidev_info *)sn_irq_info->irq_pciioinfo;
 	if (pcidev_info) {
 		pcibus_info =
@@ -184,7 +198,7 @@ void pcibr_force_interrupt(struct sn_irq_info *sn_irq_info)
 	}
 }
 
-void pcibr_change_devices_irq(struct sn_irq_info *sn_irq_info)
+void pcibr_target_interrupt(struct sn_irq_info *sn_irq_info)
 {
 	struct pcidev_info *pcidev_info;
 	struct pcibus_info *pcibus_info;
@@ -219,6 +233,8 @@ struct sn_pcibus_provider pcibr_provider = {
 	.dma_map_consistent = pcibr_dma_map_consistent,
 	.dma_unmap = pcibr_dma_unmap,
 	.bus_fixup = pcibr_bus_fixup,
+	.force_interrupt = pcibr_force_interrupt,
+	.target_interrupt = pcibr_target_interrupt
 };
 
 int

arch/ia64/sn/pci/tioca_provider.c

@@ -559,7 +559,7 @@ tioca_error_intr_handler(int irq, void *arg, struct pt_regs *pt)
 	ret_stuff.status = 0;
 	ret_stuff.v0 = 0;
 
-	segment = 0;
+	segment = soft->ca_common.bs_persist_segment;
 	busnum = soft->ca_common.bs_persist_busnum;
 
 	SAL_CALL_NOLOCK(ret_stuff,
@@ -622,7 +622,8 @@ tioca_bus_fixup(struct pcibus_bussoft *prom_bussoft, struct pci_controller *cont
 	    nasid_to_cnodeid(tioca_common->ca_closest_nasid);
 	tioca_common->ca_kernel_private = (uint64_t) tioca_kern;
 
-	bus = pci_find_bus(0, tioca_common->ca_common.bs_persist_busnum);
+	bus = pci_find_bus(tioca_common->ca_common.bs_persist_segment,
+		tioca_common->ca_common.bs_persist_busnum);
 	BUG_ON(!bus);
 	tioca_kern->ca_devices = &bus->devices;
 
@@ -656,6 +657,8 @@ static struct sn_pcibus_provider tioca_pci_interfaces = {
 	.dma_map_consistent = tioca_dma_map,
 	.dma_unmap = tioca_dma_unmap,
 	.bus_fixup = tioca_bus_fixup,
+	.force_interrupt = NULL,
+	.target_interrupt = NULL
 };
 
 /**

drivers/char/snsc_event.c

@@ -19,6 +19,7 @@
 #include <linux/sched.h>
 #include <linux/byteorder/generic.h>
 #include <asm/sn/sn_sal.h>
+#include <asm/unaligned.h>
 #include "snsc.h"
 
 static struct subch_data_s *event_sd;
@@ -62,13 +63,16 @@ static int
 scdrv_parse_event(char *event, int *src, int *code, int *esp_code, char *desc)
 {
 	char *desc_end;
+	__be32 from_buf;
 
 	/* record event source address */
-	*src = be32_to_cpup((__be32 *)event);
+	from_buf = get_unaligned((__be32 *)event);
+	*src = be32_to_cpup(&from_buf);
 	event += 4; 			/* move on to event code */
 
 	/* record the system controller's event code */
-	*code = be32_to_cpup((__be32 *)event);
+	from_buf = get_unaligned((__be32 *)event);
+	*code = be32_to_cpup(&from_buf);
 	event += 4;			/* move on to event arguments */
 
 	/* how many arguments are in the packet? */
@@ -82,7 +86,8 @@ scdrv_parse_event(char *event, int *src, int *code, int *esp_code, char *desc)
 		/* not an integer argument, so give up */
 		return -1;
 	}
-	*esp_code = be32_to_cpup((__be32 *)event);
+	from_buf = get_unaligned((__be32 *)event);
+	*esp_code = be32_to_cpup(&from_buf);
 	event += 4;
 
 	/* parse out the event description */

include/asm-ia64/acpi.h

@@ -116,6 +116,11 @@ extern int __initdata nid_to_pxm_map[MAX_NUMNODES];
 
 extern u16 ia64_acpiid_to_sapicid[];
 
+/*
+ * Refer Intel ACPI _PDC support document for bit definitions
+ */
+#define ACPI_PDC_EST_CAPABILITY_SMP     0x8
+
 #endif /*__KERNEL__*/
 
 #endif /*_ASM_ACPI_H*/

include/asm-ia64/fcntl.h

@@ -81,6 +81,7 @@ struct flock {
 
 #define F_LINUX_SPECIFIC_BASE	1024
 
-#define force_o_largefile() ( ! (current->personality & PER_LINUX32) )
+#define force_o_largefile()	\
+		(personality(current->personality) != PER_LINUX32)
 
 #endif /* _ASM_IA64_FCNTL_H */

include/asm-ia64/io.h

@@ -23,7 +23,7 @@
 #define __SLOW_DOWN_IO	do { } while (0)
 #define SLOW_DOWN_IO	do { } while (0)
 
-#define __IA64_UNCACHED_OFFSET	0xc000000000000000UL	/* region 6 */
+#define __IA64_UNCACHED_OFFSET	RGN_BASE(RGN_UNCACHED)
 
 /*
  * The legacy I/O space defined by the ia64 architecture supports only 65536 ports, but
@@ -41,7 +41,7 @@
 #define IO_SPACE_BASE(space)		((space) << IO_SPACE_BITS)
 #define IO_SPACE_PORT(port)		((port) & (IO_SPACE_SIZE - 1))
 
-#define IO_SPACE_SPARSE_ENCODING(p)	((((p) >> 2) << 12) | (p & 0xfff))
+#define IO_SPACE_SPARSE_ENCODING(p)	((((p) >> 2) << 12) | ((p) & 0xfff))
 
 struct io_space {
 	unsigned long mmio_base;	/* base in MMIO space */

include/asm-ia64/mmu.h

@@ -2,10 +2,12 @@
 #define __MMU_H
 
 /*
- * Type for a context number.  We declare it volatile to ensure proper ordering when it's
- * accessed outside of spinlock'd critical sections (e.g., as done in activate_mm() and
- * init_new_context()).
+ * Type for a context number.  We declare it volatile to ensure proper
+ * ordering when it's accessed outside of spinlock'd critical sections
+ * (e.g., as done in activate_mm() and init_new_context()).
  */
 typedef volatile unsigned long mm_context_t;
 
+typedef unsigned long nv_mm_context_t;
+
 #endif

include/asm-ia64/mmu_context.h

@@ -19,6 +19,7 @@
 
 #define ia64_rid(ctx,addr)	(((ctx) << 3) | (addr >> 61))
 
+# include <asm/page.h>
 # ifndef __ASSEMBLY__
 
 #include <linux/compiler.h>
@@ -55,34 +56,46 @@ static inline void
 delayed_tlb_flush (void)
 {
 	extern void local_flush_tlb_all (void);
+	unsigned long flags;
 
 	if (unlikely(__ia64_per_cpu_var(ia64_need_tlb_flush))) {
-		local_flush_tlb_all();
-		__ia64_per_cpu_var(ia64_need_tlb_flush) = 0;
+		spin_lock_irqsave(&ia64_ctx.lock, flags);
+		{
+			if (__ia64_per_cpu_var(ia64_need_tlb_flush)) {
+				local_flush_tlb_all();
+				__ia64_per_cpu_var(ia64_need_tlb_flush) = 0;
+			}
+		}
+		spin_unlock_irqrestore(&ia64_ctx.lock, flags);
 	}
 }
 
-static inline mm_context_t
+static inline nv_mm_context_t
 get_mmu_context (struct mm_struct *mm)
 {
 	unsigned long flags;
-	mm_context_t context = mm->context;
-
-	if (context)
-		return context;
-
-	spin_lock_irqsave(&ia64_ctx.lock, flags);
-	{
-		/* re-check, now that we've got the lock: */
-		context = mm->context;
-		if (context == 0) {
-			cpus_clear(mm->cpu_vm_mask);
-			if (ia64_ctx.next >= ia64_ctx.limit)
-				wrap_mmu_context(mm);
-			mm->context = context = ia64_ctx.next++;
+	nv_mm_context_t context = mm->context;
+
+	if (unlikely(!context)) {
+		spin_lock_irqsave(&ia64_ctx.lock, flags);
+		{
+			/* re-check, now that we've got the lock: */
+			context = mm->context;
+			if (context == 0) {
+				cpus_clear(mm->cpu_vm_mask);
+				if (ia64_ctx.next >= ia64_ctx.limit)
+					wrap_mmu_context(mm);
+				mm->context = context = ia64_ctx.next++;
+			}
 		}
+		spin_unlock_irqrestore(&ia64_ctx.lock, flags);
 	}
-	spin_unlock_irqrestore(&ia64_ctx.lock, flags);
+	/*
+	 * Ensure we're not starting to use "context" before any old
+	 * uses of it are gone from our TLB.
+	 */
+	delayed_tlb_flush();
+
 	return context;
 }
 
@@ -104,13 +117,13 @@ destroy_context (struct mm_struct *mm)
 }
 
 static inline void
-reload_context (mm_context_t context)
+reload_context (nv_mm_context_t context)
 {
 	unsigned long rid;
 	unsigned long rid_incr = 0;
 	unsigned long rr0, rr1, rr2, rr3, rr4, old_rr4;
 
-	old_rr4 = ia64_get_rr(0x8000000000000000UL);
+	old_rr4 = ia64_get_rr(RGN_BASE(RGN_HPAGE));
 	rid = context << 3;	/* make space for encoding the region number */
 	rid_incr = 1 << 8;
 
@@ -122,6 +135,10 @@ reload_context (mm_context_t context)
 	rr4 = rr0 + 4*rid_incr;
 #ifdef  CONFIG_HUGETLB_PAGE
 	rr4 = (rr4 & (~(0xfcUL))) | (old_rr4 & 0xfc);
+
+#  if RGN_HPAGE != 4
+#    error "reload_context assumes RGN_HPAGE is 4"
+#  endif
 #endif
 
 	ia64_set_rr(0x0000000000000000UL, rr0);
@@ -138,7 +155,7 @@ reload_context (mm_context_t context)
 static inline void
 activate_context (struct mm_struct *mm)
 {
-	mm_context_t context;
+	nv_mm_context_t context;
 
 	do {
 		context = get_mmu_context(mm);
@@ -157,8 +174,6 @@ activate_context (struct mm_struct *mm)
 static inline void
 activate_mm (struct mm_struct *prev, struct mm_struct *next)
 {
-	delayed_tlb_flush();
-
 	/*
 	 * We may get interrupts here, but that's OK because interrupt handlers cannot
 	 * touch user-space.

include/asm-ia64/page.h

@@ -12,6 +12,19 @@
 #include <asm/intrinsics.h>
 #include <asm/types.h>
 
+/*
+ * The top three bits of an IA64 address are its Region Number.
+ * Different regions are assigned to different purposes.
+ */
+#define RGN_SHIFT	(61)
+#define RGN_BASE(r)	(__IA64_UL_CONST(r)<<RGN_SHIFT)
+#define RGN_BITS	(RGN_BASE(-1))
+
+#define RGN_KERNEL	7	/* Identity mapped region */
+#define RGN_UNCACHED    6	/* Identity mapped I/O region */
+#define RGN_GATE	5	/* Gate page, Kernel text, etc */
+#define RGN_HPAGE	4	/* For Huge TLB pages */
+
 /*
  * PAGE_SHIFT determines the actual kernel page size.
  */
@@ -36,10 +49,9 @@
 
 #define RGN_MAP_LIMIT	((1UL << (4*PAGE_SHIFT - 12)) - PAGE_SIZE)	/* per region addr limit */
 
+
 #ifdef CONFIG_HUGETLB_PAGE
-# define REGION_HPAGE		(4UL)	/* note: this is hardcoded in reload_context()!*/
-# define REGION_SHIFT		61
-# define HPAGE_REGION_BASE	(REGION_HPAGE << REGION_SHIFT)
+# define HPAGE_REGION_BASE	RGN_BASE(RGN_HPAGE)
 # define HPAGE_SHIFT		hpage_shift
 # define HPAGE_SHIFT_DEFAULT	28	/* check ia64 SDM for architecture supported size */
 # define HPAGE_SIZE		(__IA64_UL_CONST(1) << HPAGE_SHIFT)
@@ -130,16 +142,13 @@ typedef union ia64_va {
 #define REGION_NUMBER(x)	({ia64_va _v; _v.l = (long) (x); _v.f.reg;})
 #define REGION_OFFSET(x)	({ia64_va _v; _v.l = (long) (x); _v.f.off;})
 
-#define REGION_SIZE		REGION_NUMBER(1)
-#define REGION_KERNEL		7
-
 #ifdef CONFIG_HUGETLB_PAGE
 # define htlbpage_to_page(x)	(((unsigned long) REGION_NUMBER(x) << 61)			\
 				 | (REGION_OFFSET(x) >> (HPAGE_SHIFT-PAGE_SHIFT)))
 # define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
 # define is_hugepage_only_range(mm, addr, len)		\
-	 (REGION_NUMBER(addr) == REGION_HPAGE &&	\
-	  REGION_NUMBER((addr)+(len)-1) == REGION_HPAGE)
+	 (REGION_NUMBER(addr) == RGN_HPAGE &&	\
+	  REGION_NUMBER((addr)+(len)-1) == RGN_HPAGE)
 extern unsigned int hpage_shift;
 #endif
 
@@ -197,7 +206,7 @@ get_order (unsigned long size)
 # define __pgprot(x)	(x)
 #endif /* !STRICT_MM_TYPECHECKS */
 
-#define PAGE_OFFSET			__IA64_UL_CONST(0xe000000000000000)
+#define PAGE_OFFSET			RGN_BASE(RGN_KERNEL)
 
 #define VM_DATA_DEFAULT_FLAGS		(VM_READ | VM_WRITE |					\
 					 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC |		\

include/asm-ia64/pal.h

@@ -75,6 +75,8 @@
 #define PAL_CACHE_READ		259	/* read tag & data of cacheline for diagnostic testing */
 #define PAL_CACHE_WRITE		260	/* write tag & data of cacheline for diagnostic testing */
 #define PAL_VM_TR_READ		261	/* read contents of translation register */
+#define PAL_GET_PSTATE		262	/* get the current P-state */
+#define PAL_SET_PSTATE		263	/* set the P-state */
 
 #ifndef __ASSEMBLY__
 
@@ -1111,6 +1113,25 @@ ia64_pal_halt_info (pal_power_mgmt_info_u_t *power_buf)
 	return iprv.status;
 }
 
+/* Get the current P-state information */
+static inline s64
+ia64_pal_get_pstate (u64 *pstate_index)
+{
+	struct ia64_pal_retval iprv;
+	PAL_CALL_STK(iprv, PAL_GET_PSTATE, 0, 0, 0);
+	*pstate_index = iprv.v0;
+	return iprv.status;
+}
+
+/* Set the P-state */
+static inline s64
+ia64_pal_set_pstate (u64 pstate_index)
+{
+	struct ia64_pal_retval iprv;
+	PAL_CALL_STK(iprv, PAL_SET_PSTATE, pstate_index, 0, 0);
+	return iprv.status;
+}
+
 /* Cause the processor to enter LIGHT HALT state, where prefetching and execution are
  * suspended, but cache and TLB coherency is maintained.
  */

include/asm-ia64/pgtable.h

@@ -204,21 +204,18 @@ ia64_phys_addr_valid (unsigned long addr)
 #define set_pte(ptep, pteval)	(*(ptep) = (pteval))
 #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
 
-#define RGN_SIZE	(1UL << 61)
-#define RGN_KERNEL	7
-
-#define VMALLOC_START		0xa000000200000000UL
+#define VMALLOC_START		(RGN_BASE(RGN_GATE) + 0x200000000UL)
 #ifdef CONFIG_VIRTUAL_MEM_MAP
-# define VMALLOC_END_INIT	(0xa000000000000000UL + (1UL << (4*PAGE_SHIFT - 9)))
+# define VMALLOC_END_INIT	(RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
 # define VMALLOC_END		vmalloc_end
   extern unsigned long vmalloc_end;
 #else
-# define VMALLOC_END		(0xa000000000000000UL + (1UL << (4*PAGE_SHIFT - 9)))
+# define VMALLOC_END		(RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
 #endif
 
 /* fs/proc/kcore.c */
-#define	kc_vaddr_to_offset(v) ((v) - 0xa000000000000000UL)
-#define	kc_offset_to_vaddr(o) ((o) + 0xa000000000000000UL)
+#define	kc_vaddr_to_offset(v) ((v) - RGN_BASE(RGN_GATE))
+#define	kc_offset_to_vaddr(o) ((o) + RGN_BASE(RGN_GATE))
 
 /*
  * Conversion functions: convert page frame number (pfn) and a protection value to a page

include/asm-ia64/rwsem.h

@@ -3,6 +3,7 @@
  *
  * Copyright (C) 2003 Ken Chen <kenneth.w.chen@intel.com>
  * Copyright (C) 2003 Asit Mallick <asit.k.mallick@intel.com>
+ * Copyright (C) 2005 Christoph Lameter <clameter@sgi.com>
  *
  * Based on asm-i386/rwsem.h and other architecture implementation.
  *
@@ -11,9 +12,9 @@
  *
  * The lock count is initialized to 0 (no active and no waiting lockers).
  *
- * When a writer subtracts WRITE_BIAS, it'll get 0xffff0001 for the case
- * of an uncontended lock. Readers increment by 1 and see a positive value
- * when uncontended, negative if there are writers (and maybe) readers
+ * When a writer subtracts WRITE_BIAS, it'll get 0xffffffff00000001 for
+ * the case of an uncontended lock. Readers increment by 1 and see a positive
+ * value when uncontended, negative if there are writers (and maybe) readers
  * waiting (in which case it goes to sleep).
  */
 
@@ -29,7 +30,7 @@
  * the semaphore definition
  */
 struct rw_semaphore {
-	signed int		count;
+	signed long		count;
 	spinlock_t		wait_lock;
 	struct list_head	wait_list;
 #if RWSEM_DEBUG
@@ -37,10 +38,10 @@ struct rw_semaphore {
 #endif
 };
 
-#define RWSEM_UNLOCKED_VALUE		0x00000000
-#define RWSEM_ACTIVE_BIAS		0x00000001
-#define RWSEM_ACTIVE_MASK		0x0000ffff
-#define RWSEM_WAITING_BIAS		(-0x00010000)
+#define RWSEM_UNLOCKED_VALUE		__IA64_UL_CONST(0x0000000000000000)
+#define RWSEM_ACTIVE_BIAS		__IA64_UL_CONST(0x0000000000000001)
+#define RWSEM_ACTIVE_MASK		__IA64_UL_CONST(0x00000000ffffffff)
+#define RWSEM_WAITING_BIAS		-__IA64_UL_CONST(0x0000000100000000)
 #define RWSEM_ACTIVE_READ_BIAS		RWSEM_ACTIVE_BIAS
 #define RWSEM_ACTIVE_WRITE_BIAS		(RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
 
@@ -83,7 +84,7 @@ init_rwsem (struct rw_semaphore *sem)
 static inline void
 __down_read (struct rw_semaphore *sem)
 {
-	int result = ia64_fetchadd4_acq((unsigned int *)&sem->count, 1);
+	long result = ia64_fetchadd8_acq((unsigned long *)&sem->count, 1);
 
 	if (result < 0)
 		rwsem_down_read_failed(sem);
@@ -95,7 +96,7 @@ __down_read (struct rw_semaphore *sem)
 static inline void
 __down_write (struct rw_semaphore *sem)
 {
-	int old, new;
+	long old, new;
 
 	do {
 		old = sem->count;
@@ -112,7 +113,7 @@ __down_write (struct rw_semaphore *sem)
 static inline void
 __up_read (struct rw_semaphore *sem)
 {
-	int result = ia64_fetchadd4_rel((unsigned int *)&sem->count, -1);
+	long result = ia64_fetchadd8_rel((unsigned long *)&sem->count, -1);
 
 	if (result < 0 && (--result & RWSEM_ACTIVE_MASK) == 0)
 		rwsem_wake(sem);
@@ -124,7 +125,7 @@ __up_read (struct rw_semaphore *sem)
 static inline void
 __up_write (struct rw_semaphore *sem)
 {
-	int old, new;
+	long old, new;
 
 	do {
 		old = sem->count;
@@ -141,7 +142,7 @@ __up_write (struct rw_semaphore *sem)
 static inline int
 __down_read_trylock (struct rw_semaphore *sem)
 {
-	int tmp;
+	long tmp;
 	while ((tmp = sem->count) >= 0) {
 		if (tmp == cmpxchg_acq(&sem->count, tmp, tmp+1)) {
 			return 1;
@@ -156,7 +157,7 @@ __down_read_trylock (struct rw_semaphore *sem)
 static inline int
 __down_write_trylock (struct rw_semaphore *sem)
 {
-	int tmp = cmpxchg_acq(&sem->count, RWSEM_UNLOCKED_VALUE,
+	long tmp = cmpxchg_acq(&sem->count, RWSEM_UNLOCKED_VALUE,
 			      RWSEM_ACTIVE_WRITE_BIAS);
 	return tmp == RWSEM_UNLOCKED_VALUE;
 }
@@ -167,7 +168,7 @@ __down_write_trylock (struct rw_semaphore *sem)
 static inline void
 __downgrade_write (struct rw_semaphore *sem)
 {
-	int old, new;
+	long old, new;
 
 	do {
 		old = sem->count;
@@ -182,7 +183,7 @@ __downgrade_write (struct rw_semaphore *sem)
  * Implement atomic add functionality.  These used to be "inline" functions, but GCC v3.1
  * doesn't quite optimize this stuff right and ends up with bad calls to fetchandadd.
  */
-#define rwsem_atomic_add(delta, sem)	atomic_add(delta, (atomic_t *)(&(sem)->count))
-#define rwsem_atomic_update(delta, sem)	atomic_add_return(delta, (atomic_t *)(&(sem)->count))
+#define rwsem_atomic_add(delta, sem)	atomic64_add(delta, (atomic64_t *)(&(sem)->count))
+#define rwsem_atomic_update(delta, sem)	atomic64_add_return(delta, (atomic64_t *)(&(sem)->count))
 
 #endif /* _ASM_IA64_RWSEM_H */

include/asm-ia64/sn/addrs.h

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (c) 1992-1999,2001-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (c) 1992-1999,2001-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #ifndef _ASM_IA64_SN_ADDRS_H
@@ -65,7 +65,6 @@
 
 #define NASID_MASK              ((u64)NASID_BITMASK << NASID_SHIFT)
 #define AS_MASK			((u64)AS_BITMASK << AS_SHIFT)
-#define REGION_BITS		0xe000000000000000UL
 
 
 /*
@@ -79,38 +78,30 @@
 #define AS_CAC_SPACE		(AS_CAC_VAL << AS_SHIFT)
 
 
-/*
- * Base addresses for various address ranges.
- */
-#define CACHED			0xe000000000000000UL
-#define UNCACHED                0xc000000000000000UL
-#define UNCACHED_PHYS           0x8000000000000000UL
-
-
 /* 
  * Virtual Mode Local & Global MMR space.  
  */
 #define SH1_LOCAL_MMR_OFFSET	0x8000000000UL
 #define SH2_LOCAL_MMR_OFFSET	0x0200000000UL
 #define LOCAL_MMR_OFFSET	(is_shub2() ? SH2_LOCAL_MMR_OFFSET : SH1_LOCAL_MMR_OFFSET)
-#define LOCAL_MMR_SPACE		(UNCACHED | LOCAL_MMR_OFFSET)
-#define LOCAL_PHYS_MMR_SPACE	(UNCACHED_PHYS | LOCAL_MMR_OFFSET)
+#define LOCAL_MMR_SPACE		(__IA64_UNCACHED_OFFSET | LOCAL_MMR_OFFSET)
+#define LOCAL_PHYS_MMR_SPACE	(RGN_BASE(RGN_HPAGE) | LOCAL_MMR_OFFSET)
 
 #define SH1_GLOBAL_MMR_OFFSET	0x0800000000UL
 #define SH2_GLOBAL_MMR_OFFSET	0x0300000000UL
 #define GLOBAL_MMR_OFFSET	(is_shub2() ? SH2_GLOBAL_MMR_OFFSET : SH1_GLOBAL_MMR_OFFSET)
-#define GLOBAL_MMR_SPACE	(UNCACHED | GLOBAL_MMR_OFFSET)
+#define GLOBAL_MMR_SPACE	(__IA64_UNCACHED_OFFSET | GLOBAL_MMR_OFFSET)
 
 /*
  * Physical mode addresses
  */
-#define GLOBAL_PHYS_MMR_SPACE	(UNCACHED_PHYS | GLOBAL_MMR_OFFSET)
+#define GLOBAL_PHYS_MMR_SPACE	(RGN_BASE(RGN_HPAGE) | GLOBAL_MMR_OFFSET)
 
 
 /*
  * Clear region & AS bits.
  */
-#define TO_PHYS_MASK		(~(REGION_BITS | AS_MASK))
+#define TO_PHYS_MASK		(~(RGN_BITS | AS_MASK))
 
 
 /*
@@ -126,6 +117,7 @@
 #define GLOBAL_MMR_PHYS_ADDR(n,a) (GLOBAL_PHYS_MMR_SPACE | REMOTE_ADDR(n,a))
 #define GLOBAL_CAC_ADDR(n,a)	(CAC_BASE | REMOTE_ADDR(n,a))
 #define CHANGE_NASID(n,x)	((void *)(((u64)(x) & ~NASID_MASK) | NASID_SPACE(n)))
+#define IS_TIO_NASID(n)		((n) & 1)
 
 
 /* non-II mmr's start at top of big window space (4G) */
@@ -134,10 +126,10 @@
 /*
  * general address defines
  */
-#define CAC_BASE		(CACHED   | AS_CAC_SPACE)
-#define AMO_BASE		(UNCACHED | AS_AMO_SPACE)
-#define AMO_PHYS_BASE		(UNCACHED_PHYS | AS_AMO_SPACE)
-#define GET_BASE		(CACHED   | AS_GET_SPACE)
+#define CAC_BASE		(PAGE_OFFSET | AS_CAC_SPACE)
+#define AMO_BASE		(__IA64_UNCACHED_OFFSET | AS_AMO_SPACE)
+#define AMO_PHYS_BASE		(RGN_BASE(RGN_HPAGE) | AS_AMO_SPACE)
+#define GET_BASE		(PAGE_OFFSET | AS_GET_SPACE)
 
 /*
  * Convert Memory addresses between various addressing modes.
@@ -155,17 +147,35 @@
  *           the chiplet id is zero.  If we implement TIO-TIO dma, we might need
  *           to insert a chiplet id into this macro.  However, it is our belief
  *           right now that this chiplet id will be ICE, which is also zero.
- *           Nasid starts on bit 40.
  */
-#define PHYS_TO_TIODMA(x)	( (((u64)(NASID_GET(x))) << 40) | NODE_OFFSET(x))
-#define PHYS_TO_DMA(x)          ( (((u64)(x) & NASID_MASK) >> 2) | NODE_OFFSET(x))
+#define SH1_TIO_PHYS_TO_DMA(x) 						\
+	((((u64)(NASID_GET(x))) << 40) | NODE_OFFSET(x))
+
+#define SH2_NETWORK_BANK_OFFSET(x) 					\
+        ((u64)(x) & ((1UL << (sn_hub_info->nasid_shift - 4)) -1))
+
+#define SH2_NETWORK_BANK_SELECT(x) 					\
+        ((((u64)(x) & (0x3UL << (sn_hub_info->nasid_shift - 4)))	\
+        	>> (sn_hub_info->nasid_shift - 4)) << 36)
+
+#define SH2_NETWORK_ADDRESS(x) 						\
+	(SH2_NETWORK_BANK_OFFSET(x) | SH2_NETWORK_BANK_SELECT(x))
+
+#define SH2_TIO_PHYS_TO_DMA(x) 						\
+        (((u64)(NASID_GET(x)) << 40) | 	SH2_NETWORK_ADDRESS(x))
+
+#define PHYS_TO_TIODMA(x)						\
+	(is_shub1() ? SH1_TIO_PHYS_TO_DMA(x) : SH2_TIO_PHYS_TO_DMA(x))
+
+#define PHYS_TO_DMA(x)							\
+	((((u64)(x) & NASID_MASK) >> 2) | NODE_OFFSET(x))
 
 
 /*
  * Macros to test for address type.
  */
-#define IS_AMO_ADDRESS(x)	(((u64)(x) & (REGION_BITS | AS_MASK)) == AMO_BASE)
-#define IS_AMO_PHYS_ADDRESS(x)	(((u64)(x) & (REGION_BITS | AS_MASK)) == AMO_PHYS_BASE)
+#define IS_AMO_ADDRESS(x)	(((u64)(x) & (RGN_BITS | AS_MASK)) == AMO_BASE)
+#define IS_AMO_PHYS_ADDRESS(x)	(((u64)(x) & (RGN_BITS | AS_MASK)) == AMO_PHYS_BASE)
 
 
 /*
@@ -180,18 +190,20 @@
 #define TIO_SWIN_BASE(n, w) 		(TIO_IO_BASE(n) + \
 					    ((u64) (w) << TIO_SWIN_SIZE_BITS))
 #define NODE_IO_BASE(n)			(GLOBAL_MMR_SPACE | NASID_SPACE(n))
-#define TIO_IO_BASE(n)                  (UNCACHED | NASID_SPACE(n))
+#define TIO_IO_BASE(n)                  (__IA64_UNCACHED_OFFSET | NASID_SPACE(n))
 #define BWIN_SIZE			(1UL << BWIN_SIZE_BITS)
 #define NODE_BWIN_BASE0(n)		(NODE_IO_BASE(n) + BWIN_SIZE)
 #define NODE_BWIN_BASE(n, w)		(NODE_BWIN_BASE0(n) + ((u64) (w) << BWIN_SIZE_BITS))
 #define RAW_NODE_SWIN_BASE(n, w)	(NODE_IO_BASE(n) + ((u64) (w) << SWIN_SIZE_BITS))
 #define BWIN_WIDGET_MASK		0x7
 #define BWIN_WINDOWNUM(x)		(((x) >> BWIN_SIZE_BITS) & BWIN_WIDGET_MASK)
+#define SH1_IS_BIG_WINDOW_ADDR(x)	((x) & BWIN_TOP)
 
 #define TIO_BWIN_WINDOW_SELECT_MASK	0x7
 #define TIO_BWIN_WINDOWNUM(x)		(((x) >> TIO_BWIN_SIZE_BITS) & TIO_BWIN_WINDOW_SELECT_MASK)
 
-
+#define TIO_HWIN_SHIFT_BITS		33
+#define TIO_HWIN(x)			(NODE_OFFSET(x) >> TIO_HWIN_SHIFT_BITS)
 
 /*
  * The following definitions pertain to the IO special address
@@ -216,10 +228,6 @@
 #define TIO_SWIN_WIDGETNUM(x)		(((x)  >> TIO_SWIN_SIZE_BITS) & TIO_SWIN_WIDGET_MASK)
 
 
-#define TIO_IOSPACE_ADDR(n,x)					\
-	/* Move in the Chiplet ID for TIO Local Block MMR */	\
-	(REMOTE_ADDR(n,x) | 1UL << (NASID_SHIFT - 2))
-
 /*
  * The following macros produce the correct base virtual address for
  * the hub registers. The REMOTE_HUB_* macro produce
@@ -234,18 +242,40 @@
  *	Otherwise, the recommended approach is to use *_HUB_L() and *_HUB_S().
  *	They're always safe.
  */
+/* Shub1 TIO & MMR addressing macros */
+#define SH1_TIO_IOSPACE_ADDR(n,x)					\
+	GLOBAL_MMR_ADDR(n,x)
+
+#define SH1_REMOTE_BWIN_MMR(n,x)					\
+	GLOBAL_MMR_ADDR(n,x)
+
+#define SH1_REMOTE_SWIN_MMR(n,x)					\
+	(NODE_SWIN_BASE(n,1) + 0x800000UL + (x))
+
+#define SH1_REMOTE_MMR(n,x)						\
+	(SH1_IS_BIG_WINDOW_ADDR(x) ? SH1_REMOTE_BWIN_MMR(n,x) :		\
+	 	SH1_REMOTE_SWIN_MMR(n,x))
+
+/* Shub1 TIO & MMR addressing macros */
+#define SH2_TIO_IOSPACE_ADDR(n,x)					\
+	((__IA64_UNCACHED_OFFSET | REMOTE_ADDR(n,x) | 1UL << (NASID_SHIFT - 2)))
+
+#define SH2_REMOTE_MMR(n,x)						\
+	GLOBAL_MMR_ADDR(n,x)
+
+
+/* TIO & MMR addressing macros that work on both shub1 & shub2 */
+#define TIO_IOSPACE_ADDR(n,x)						\
+	((u64 *)(is_shub1() ? SH1_TIO_IOSPACE_ADDR(n,x) :		\
+		 SH2_TIO_IOSPACE_ADDR(n,x)))
+
+#define SH_REMOTE_MMR(n,x)						\
+	(is_shub1() ? SH1_REMOTE_MMR(n,x) : SH2_REMOTE_MMR(n,x))
+
 #define REMOTE_HUB_ADDR(n,x)						\
-	((n & 1) ?							\
-	/* TIO: */							\
-	(is_shub2() ?							\
-	/* TIO on Shub2 */						\
-	(volatile u64 *)(TIO_IOSPACE_ADDR(n,x))				\
-	: /* TIO on shub1 */						\
-	(volatile u64 *)(GLOBAL_MMR_ADDR(n,x)))				\
-									\
-	: /* SHUB1 and SHUB2 MMRs: */					\
-	(((x) & BWIN_TOP) ? ((volatile u64 *)(GLOBAL_MMR_ADDR(n,x)))	\
-	: ((volatile u64 *)(NODE_SWIN_BASE(n,1) + 0x800000 + (x)))))
+	(IS_TIO_NASID(n) ?  ((volatile u64*)TIO_IOSPACE_ADDR(n,x)) :	\
+	 ((volatile u64*)SH_REMOTE_MMR(n,x)))
+
 
 #define HUB_L(x)			(*((volatile typeof(*x) *)x))
 #define	HUB_S(x,d)			(*((volatile typeof(*x) *)x) = (d))

include/asm-ia64/sn/geo.h

@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1992 - 1997, 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 1992 - 1997, 2000-2005 Silicon Graphics, Inc. All rights reserved.
  */
 
 #ifndef _ASM_IA64_SN_GEO_H
@@ -108,7 +108,6 @@ typedef union geoid_u {
 #define INVALID_SLAB            (slabid_t)-1
 #define INVALID_SLOT            (slotid_t)-1
 #define INVALID_MODULE          ((moduleid_t)-1)
-#define INVALID_PARTID          ((partid_t)-1)
 
 static inline slabid_t geo_slab(geoid_t g)
 {

include/asm-ia64/sn/intr.h

@@ -12,13 +12,12 @@
 #include <linux/rcupdate.h>
 
 #define SGI_UART_VECTOR		(0xe9)
-#define SGI_PCIBR_ERROR		(0x33)
 
 /* Reserved IRQs : Note, not to exceed IA64_SN2_FIRST_DEVICE_VECTOR */
 #define SGI_XPC_ACTIVATE                (0x30)
 #define SGI_II_ERROR                    (0x31)
 #define SGI_XBOW_ERROR                  (0x32)
-#define SGI_PCIBR_ERROR                 (0x33)
+#define SGI_PCIASIC_ERROR               (0x33)
 #define SGI_ACPI_SCI_INT                (0x34)
 #define SGI_TIOCA_ERROR                 (0x35)
 #define SGI_TIO_ERROR                   (0x36)

include/asm-ia64/sn/nodepda.h

@@ -37,7 +37,6 @@ struct phys_cpuid {
 
 struct nodepda_s {
 	void 		*pdinfo;	/* Platform-dependent per-node info */
-	spinlock_t		bist_lock;
 
 	/*
 	 * The BTEs on this node are shared by the local cpus
@@ -55,6 +54,8 @@ struct nodepda_s {
 	 * Array of physical cpu identifiers. Indexed by cpuid.
 	 */
 	struct phys_cpuid	phys_cpuid[NR_CPUS];
+	spinlock_t		ptc_lock ____cacheline_aligned_in_smp;
+	spinlock_t		bist_lock;
 };
 
 typedef struct nodepda_s nodepda_t;

include/asm-ia64/sn/pcibus_provider_defs.h

@@ -18,8 +18,9 @@
 #define PCIIO_ASIC_TYPE_PIC	2
 #define PCIIO_ASIC_TYPE_TIOCP	3
 #define PCIIO_ASIC_TYPE_TIOCA	4
+#define PCIIO_ASIC_TYPE_TIOCE	5
 
-#define PCIIO_ASIC_MAX_TYPES	5
+#define PCIIO_ASIC_MAX_TYPES	6
 
 /*
  * Common pciio bus provider data.  There should be one of these as the
@@ -30,7 +31,8 @@
 struct pcibus_bussoft {
 	uint32_t		bs_asic_type;	/* chipset type */
 	uint32_t		bs_xid;		/* xwidget id */
-	uint64_t		bs_persist_busnum; /* Persistent Bus Number */
+	uint32_t		bs_persist_busnum; /* Persistent Bus Number */
+	uint32_t		bs_persist_segment; /* Segment Number */
 	uint64_t		bs_legacy_io;	/* legacy io pio addr */
 	uint64_t		bs_legacy_mem;	/* legacy mem pio addr */
 	uint64_t		bs_base;	/* widget base */
@@ -47,6 +49,8 @@ struct sn_pcibus_provider {
 	dma_addr_t	(*dma_map_consistent)(struct pci_dev *, unsigned long, size_t);
 	void		(*dma_unmap)(struct pci_dev *, dma_addr_t, int);
 	void *		(*bus_fixup)(struct pcibus_bussoft *, struct pci_controller *);
+ 	void		(*force_interrupt)(struct sn_irq_info *);
+ 	void		(*target_interrupt)(struct sn_irq_info *);
 };
 
 extern struct sn_pcibus_provider *sn_pci_provider[];

include/asm-ia64/sn/pda.h

@@ -39,7 +39,6 @@ typedef struct pda_s {
 	unsigned long pio_write_status_val;
 	volatile unsigned long *pio_shub_war_cam_addr;
 
-	unsigned long	sn_soft_irr[4];
 	unsigned long	sn_in_service_ivecs[4];
 	int		sn_lb_int_war_ticks;
 	int		sn_last_irq;

include/asm-ia64/sn/sn2/sn_hwperf.h

@@ -43,6 +43,7 @@ struct sn_hwperf_object_info {
 
 /* macros for object classification */
 #define SN_HWPERF_IS_NODE(x)		((x) && strstr((x)->name, "SHub"))
+#define SN_HWPERF_IS_NODE_SHUB2(x)	((x) && strstr((x)->name, "SHub 2."))
 #define SN_HWPERF_IS_IONODE(x)		((x) && strstr((x)->name, "TIO"))
 #define SN_HWPERF_IS_ROUTER(x)		((x) && strstr((x)->name, "Router"))
 #define SN_HWPERF_IS_NL3ROUTER(x)	((x) && strstr((x)->name, "NL3Router"))
@@ -214,6 +215,15 @@ struct sn_hwperf_ioctl_args {
  */
 #define SN_HWPERF_GET_NODE_NASID	(102|SN_HWPERF_OP_MEM_COPYOUT)
 
+/*
+ * Given a node id, determine the id of the nearest node with CPUs
+ * and the id of the nearest node that has memory. The argument
+ * node would normally be a "headless" node, e.g. an "IO node".
+ * Return 0 on success.
+ */
+extern int sn_hwperf_get_nearest_node(cnodeid_t node,
+	cnodeid_t *near_mem, cnodeid_t *near_cpu);
+
 /* return codes */
 #define SN_HWPERF_OP_OK			0
 #define SN_HWPERF_OP_NOMEM		1

include/asm-ia64/sn/sn_sal.h

@@ -55,7 +55,6 @@
 #define  SN_SAL_BUS_CONFIG		   	   0x02000037
 #define  SN_SAL_SYS_SERIAL_GET			   0x02000038
 #define  SN_SAL_PARTITION_SERIAL_GET		   0x02000039
-#define  SN_SAL_SYSCTL_PARTITION_GET		   0x0200003a
 #define  SN_SAL_SYSTEM_POWER_DOWN		   0x0200003b
 #define  SN_SAL_GET_MASTER_BASEIO_NASID		   0x0200003c
 #define  SN_SAL_COHERENCE                          0x0200003d
@@ -78,7 +77,8 @@
 
 #define SN_SAL_HUB_ERROR_INTERRUPT		   0x02000060
 #define SN_SAL_BTE_RECOVER			   0x02000061
-#define SN_SAL_IOIF_GET_PCI_TOPOLOGY	           0x02000062
+#define SN_SAL_RESERVED_DO_NOT_USE		   0x02000062
+#define SN_SAL_IOIF_GET_PCI_TOPOLOGY		   0x02000064
 
 /*
  * Service-specific constants
@@ -585,35 +585,6 @@ sn_partition_serial_number_val(void) {
 	return sn_partition_serial_number;
 }
 
-/*
- * Returns the partition id of the nasid passed in as an argument,
- * or INVALID_PARTID if the partition id cannot be retrieved.
- */
-static inline partid_t
-ia64_sn_sysctl_partition_get(nasid_t nasid)
-{
-	struct ia64_sal_retval ret_stuff;
-	ia64_sal_oemcall_nolock(&ret_stuff, SN_SAL_SYSCTL_PARTITION_GET, nasid,
-				0, 0, 0, 0, 0, 0);
-	if (ret_stuff.status != 0)
-	    return INVALID_PARTID;
-	return ((partid_t)ret_stuff.v0);
-}
-
-/*
- * Returns the partition id of the current processor.
- */
-
-extern partid_t sn_partid;
-
-static inline partid_t
-sn_local_partid(void) {
-	if (unlikely(sn_partid < 0)) {
-		sn_partid = ia64_sn_sysctl_partition_get(cpuid_to_nasid(smp_processor_id()));
-	}
-	return sn_partid;
-}
-
 /*
  * Returns the physical address of the partition's reserved page through
  * an iterative number of calls.
@@ -749,7 +720,8 @@ ia64_sn_power_down(void)
 {
 	struct ia64_sal_retval ret_stuff;
 	SAL_CALL(ret_stuff, SN_SAL_SYSTEM_POWER_DOWN, 0, 0, 0, 0, 0, 0, 0);
-	while(1);
+	while(1)
+		cpu_relax();
 	/* never returns */
 }
 
@@ -1018,24 +990,6 @@ ia64_sn_get_sn_info(int fc, u8 *shubtype, u16 *nasid_bitmask, u8 *nasid_shift,
 	ret_stuff.v2 = 0;
 	SAL_CALL_NOLOCK(ret_stuff, SN_SAL_GET_SN_INFO, fc, 0, 0, 0, 0, 0, 0);
 
-/***** BEGIN HACK - temp til old proms no longer supported ********/
-	if (ret_stuff.status == SALRET_NOT_IMPLEMENTED) {
-		int nasid = get_sapicid() & 0xfff;;
-#define SH_SHUB_ID_NODES_PER_BIT_MASK 0x001f000000000000UL                                               
-#define SH_SHUB_ID_NODES_PER_BIT_SHFT 48                                                               
-		if (shubtype) *shubtype = 0;
-		if (nasid_bitmask) *nasid_bitmask = 0x7ff;
-		if (nasid_shift) *nasid_shift = 38;
-		if (systemsize) *systemsize = 11;
-		if (sharing_domain_size) *sharing_domain_size = 9;
-		if (partid) *partid = ia64_sn_sysctl_partition_get(nasid);
-		if (coher) *coher = nasid >> 9;
-		if (reg) *reg = (HUB_L((u64 *) LOCAL_MMR_ADDR(SH1_SHUB_ID)) & SH_SHUB_ID_NODES_PER_BIT_MASK) >>
-			SH_SHUB_ID_NODES_PER_BIT_SHFT;
-		return 0;
-	}
-/***** END HACK *******/
-
 	if (ret_stuff.status < 0)
 		return ret_stuff.status;
 
@@ -1068,12 +1022,10 @@ ia64_sn_hwperf_op(nasid_t nasid, u64 opcode, u64 a0, u64 a1, u64 a2,
 }
 
 static inline int
-ia64_sn_ioif_get_pci_topology(u64 rack, u64 bay, u64 slot, u64 slab,
-			      u64 buf, u64 len)
+ia64_sn_ioif_get_pci_topology(u64 buf, u64 len)
 {
 	struct ia64_sal_retval rv;
-	SAL_CALL_NOLOCK(rv, SN_SAL_IOIF_GET_PCI_TOPOLOGY,
-		rack, bay, slot, slab, buf, len, 0);
+	SAL_CALL_NOLOCK(rv, SN_SAL_IOIF_GET_PCI_TOPOLOGY, buf, len, 0, 0, 0, 0, 0);
 	return (int) rv.status;
 }
 

include/asm-ia64/sn/tioce_provider.h

+/**************************************************************************
+ *             Copyright (C) 2005, Silicon Graphics, Inc.                 *
+ *									  *
+ *  These coded instructions, statements, and computer programs	 contain  *
+ *  unpublished	 proprietary  information of Silicon Graphics, Inc., and  *
+ *  are protected by Federal copyright law.  They  may	not be disclosed  *
+ *  to	third  parties	or copied or duplicated in any form, in whole or  *
+ *  in part, without the prior written consent of Silicon Graphics, Inc.  *
+ *									  *
+ **************************************************************************/
+
+#ifndef _ASM_IA64_SN_CE_PROVIDER_H
+#define _ASM_IA64_SN_CE_PROVIDER_H
+
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/tioce.h>
+
+/*
+ * Common TIOCE structure shared between the prom and kernel
+ *
+ * DO NOT CHANGE THIS STRUCT WITHOUT MAKING CORRESPONDING CHANGES TO THE
+ * PROM VERSION.
+ */
+struct tioce_common {
+	struct pcibus_bussoft	ce_pcibus;	/* common pciio header */
+
+	uint32_t		ce_rev;
+	uint64_t		ce_kernel_private;
+	uint64_t		ce_prom_private;
+};
+
+struct tioce_kernel {
+	struct tioce_common	*ce_common;
+	spinlock_t		ce_lock;
+	struct list_head	ce_dmamap_list;
+
+	uint64_t		ce_ate40_shadow[TIOCE_NUM_M40_ATES];
+	uint64_t		ce_ate3240_shadow[TIOCE_NUM_M3240_ATES];
+	uint32_t		ce_ate3240_pagesize;
+
+	uint8_t			ce_port1_secondary;
+
+	/* per-port resources */
+	struct {
+		int 		dirmap_refcnt;
+		uint64_t	dirmap_shadow;
+	} ce_port[TIOCE_NUM_PORTS];
+};
+
+struct tioce_dmamap {
+	struct list_head	ce_dmamap_list;	/* headed by tioce_kernel */
+	uint32_t		refcnt;
+
+	uint64_t		nbytes;		/* # bytes mapped */
+
+	uint64_t		ct_start;	/* coretalk start address */
+	uint64_t		pci_start;	/* bus start address */
+
+	uint64_t		*ate_hw;	/* hw ptr of first ate in map */
+	uint64_t		*ate_shadow;	/* shadow ptr of firat ate */
+	uint16_t		ate_count;	/* # ate's in the map */
+};
+
+extern int tioce_init_provider(void);
+
+#endif  /* __ASM_IA64_SN_CE_PROVIDER_H */

include/asm-ia64/spinlock.h

@@ -93,7 +93,15 @@ _raw_spin_lock_flags (spinlock_t *lock, unsigned long flags)
 # endif /* CONFIG_MCKINLEY */
 #endif
 }
+
 #define _raw_spin_lock(lock) _raw_spin_lock_flags(lock, 0)
+
+/* Unlock by doing an ordered store and releasing the cacheline with nta */
+static inline void _raw_spin_unlock(spinlock_t *x) {
+	barrier();
+	asm volatile ("st4.rel.nta [%0] = r0\n\t" :: "r"(x));
+}
+
 #else /* !ASM_SUPPORTED */
 #define _raw_spin_lock_flags(lock, flags) _raw_spin_lock(lock)
 # define _raw_spin_lock(x)								\
@@ -109,16 +117,16 @@ do {											\
 		} while (ia64_spinlock_val);						\
 	}										\
 } while (0)
+#define _raw_spin_unlock(x)	do { barrier(); ((spinlock_t *) x)->lock = 0; } while (0)
 #endif /* !ASM_SUPPORTED */
 
 #define spin_is_locked(x)	((x)->lock != 0)
-#define _raw_spin_unlock(x)	do { barrier(); ((spinlock_t *) x)->lock = 0; } while (0)
 #define _raw_spin_trylock(x)	(cmpxchg_acq(&(x)->lock, 0, 1) == 0)
 #define spin_unlock_wait(x)	do { barrier(); } while ((x)->lock)
 
 typedef struct {
-	volatile unsigned int read_counter	: 31;
-	volatile unsigned int write_lock	:  1;
+	volatile unsigned int read_counter	: 24;
+	volatile unsigned int write_lock	:  8;
 #ifdef CONFIG_PREEMPT
 	unsigned int break_lock;
 #endif
@@ -174,6 +182,13 @@ do {										\
 	(result == 0);								\
 })
 
+static inline void _raw_write_unlock(rwlock_t *x)
+{
+	u8 *y = (u8 *)x;
+	barrier();
+	asm volatile ("st1.rel.nta [%0] = r0\n\t" :: "r"(y+3) : "memory" );
+}
+
 #else /* !ASM_SUPPORTED */
 
 #define _raw_write_lock(l)								\
@@ -195,14 +210,14 @@ do {										\
 	(ia64_val == 0);						\
 })
 
+static inline void _raw_write_unlock(rwlock_t *x)
+{
+	barrier();
+	x->write_lock = 0;
+}
+
 #endif /* !ASM_SUPPORTED */
 
 #define _raw_read_trylock(lock) generic_raw_read_trylock(lock)
 
-#define _raw_write_unlock(x)								\
-({											\
-	smp_mb__before_clear_bit();	/* need barrier before releasing lock... */	\
-	clear_bit(31, (x));								\
-})
-
 #endif /*  _ASM_IA64_SPINLOCK_H */

include/asm-ia64/system.h

@@ -19,12 +19,13 @@
 #include <asm/pal.h>
 #include <asm/percpu.h>
 
-#define GATE_ADDR		__IA64_UL_CONST(0xa000000000000000)
+#define GATE_ADDR		RGN_BASE(RGN_GATE)
+
 /*
  * 0xa000000000000000+2*PERCPU_PAGE_SIZE
  * - 0xa000000000000000+3*PERCPU_PAGE_SIZE remain unmapped (guard page)
  */
-#define KERNEL_START		 __IA64_UL_CONST(0xa000000100000000)
+#define KERNEL_START		 (GATE_ADDR+0x100000000)
 #define PERCPU_ADDR		(-PERCPU_PAGE_SIZE)
 
 #ifndef __ASSEMBLY__