Merge http://lia64.bkbits.net/linux-ia64-release-2.6.10

into ppc970.osdl.org:/home/torvalds/v2.6/linux

arch/ia64/ia32/elfcore32.h

@@ -9,6 +9,7 @@
 #define _ELFCORE32_H_
 
 #include <asm/intrinsics.h>
+#include <asm/uaccess.h>
 
 #define USE_ELF_CORE_DUMP 1
 

arch/ia64/kernel/efi.c

@@ -28,6 +28,7 @@
 
 #include <asm/io.h>
 #include <asm/kregs.h>
+#include <asm/meminit.h>
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/mca.h>
@@ -324,12 +325,12 @@ efi_memmap_walk (efi_freemem_callback_t callback, void *arg)
 		 * [granule_addr - first_non_wb_addr) is guaranteed to
 		 * be contiguous WB memory.
 		 */
-		granule_addr = md->phys_addr & ~(IA64_GRANULE_SIZE - 1);
+		granule_addr = GRANULEROUNDDOWN(md->phys_addr);
 		first_non_wb_addr = max(first_non_wb_addr, granule_addr);
 
 		if (first_non_wb_addr < md->phys_addr) {
 			trim_bottom(md, granule_addr + IA64_GRANULE_SIZE);
-			granule_addr = md->phys_addr & ~(IA64_GRANULE_SIZE - 1);
+			granule_addr = GRANULEROUNDDOWN(md->phys_addr);
 			first_non_wb_addr = max(first_non_wb_addr, granule_addr);
 		}
 
@@ -343,24 +344,36 @@ efi_memmap_walk (efi_freemem_callback_t callback, void *arg)
 				break;		/* non-WB or hole */
 		}
 
-		last_granule_addr = first_non_wb_addr & ~(IA64_GRANULE_SIZE - 1);
+		last_granule_addr = GRANULEROUNDDOWN(first_non_wb_addr);
 		if (last_granule_addr < md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT))
 			trim_top(md, last_granule_addr);
 
 		if (is_available_memory(md)) {
-			if (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT) > max_addr) {
-				if (md->phys_addr > max_addr)
+			if (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT) >= max_addr) {
+				if (md->phys_addr >= max_addr)
 					continue;
 				md->num_pages = (max_addr - md->phys_addr) >> EFI_PAGE_SHIFT;
+				first_non_wb_addr = max_addr;
 			}
 
 			if (total_mem >= mem_limit)
 				continue;
-			total_mem += (md->num_pages << EFI_PAGE_SHIFT);
-			if (total_mem > mem_limit) {
-				md->num_pages -= ((total_mem - mem_limit) >> EFI_PAGE_SHIFT);
-				max_addr = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT);
+
+			if (total_mem + (md->num_pages << EFI_PAGE_SHIFT) > mem_limit) {
+				unsigned long limit_addr = md->phys_addr;
+
+				limit_addr += mem_limit - total_mem;
+				limit_addr = GRANULEROUNDDOWN(limit_addr);
+
+				if (md->phys_addr > limit_addr)
+					continue;
+
+				md->num_pages = (limit_addr - md->phys_addr) >>
+				                EFI_PAGE_SHIFT;
+				first_non_wb_addr = max_addr = md->phys_addr +
+				              (md->num_pages << EFI_PAGE_SHIFT);
 			}
+			total_mem += (md->num_pages << EFI_PAGE_SHIFT);
 
 			if (md->num_pages == 0)
 				continue;
@@ -495,13 +508,13 @@ efi_init (void)
 	for (cp = saved_command_line; *cp; ) {
 		if (memcmp(cp, "mem=", 4) == 0) {
 			cp += 4;
-			mem_limit = memparse(cp, &end) - 2;
+			mem_limit = memparse(cp, &end);
 			if (end != cp)
 				break;
 			cp = end;
 		} else if (memcmp(cp, "max_addr=", 9) == 0) {
 			cp += 9;
-			max_addr = memparse(cp, &end) - 1;
+			max_addr = GRANULEROUNDDOWN(memparse(cp, &end));
 			if (end != cp)
 				break;
 			cp = end;

arch/ia64/kernel/mca_asm.S

@@ -868,7 +868,7 @@ end_os_mca_restore:
 
 
 GLOBAL_ENTRY(ia64_monarch_init_handler)
-
+	.prologue
 	// stash the information the SAL passed to os
 	SAL_TO_OS_MCA_HANDOFF_STATE_SAVE(r2)
 	;;
@@ -907,6 +907,7 @@ IVirtual_Switch:
 	adds out0=16,sp				// out0 = pointer to pt_regs
 	;;
 	DO_SAVE_SWITCH_STACK
+	.body
 	adds out1=16,sp				// out0 = pointer to switch_stack
 
 	br.call.sptk.many rp=ia64_init_handler

arch/ia64/kernel/perfmon.c

@@ -1638,7 +1638,7 @@ pfm_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned lon
 }
 
 /*
- * context is locked when coming here and interrupts are disabled
+ * interrupt cannot be masked when coming here
  */
 static inline int
 pfm_do_fasync(int fd, struct file *filp, pfm_context_t *ctx, int on)
@@ -2270,7 +2270,7 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon
 	 * 	return -ENOMEM;
 	 */
 	if (size > task->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
-		return -EAGAIN;
+		return -ENOMEM;
 
 	/*
 	 * We do the easy to undo allocations first.
@@ -2295,10 +2295,6 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon
 
 	/*
 	 * partially initialize the vma for the sampling buffer
-	 *
-	 * The VM_DONTCOPY flag is very important as it ensures that the mapping
-	 * will never be inherited for any child process (via fork()) which is always
-	 * what we want.
 	 */
 	vma->vm_mm	     = mm;
 	vma->vm_flags	     = VM_READ| VM_MAYREAD |VM_RESERVED;
@@ -2328,6 +2324,7 @@ pfm_smpl_buffer_alloc(struct task_struct *task, pfm_context_t *ctx, unsigned lon
 		goto error;
 	}
 	vma->vm_end = vma->vm_start + size;
+	vma->vm_pgoff = vma->vm_start >> PAGE_SHIFT;
 
 	DPRINT(("aligned size=%ld, hdr=%p mapped @0x%lx\n", size, ctx->ctx_smpl_hdr, vma->vm_start));
 
@@ -3060,11 +3057,12 @@ pfm_write_pmcs(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 #endif
 		}
 
-		DPRINT(("pmc[%u]=0x%lx loaded=%d access_pmu=%d all_pmcs=0x%lx used_pmds=0x%lx eventid=%ld smpl_pmds=0x%lx reset_pmds=0x%lx reloads_pmcs=0x%lx used_monitors=0x%lx ovfl_regs=0x%lx\n",
+		DPRINT(("pmc[%u]=0x%lx ld=%d apmu=%d flags=0x%lx all_pmcs=0x%lx used_pmds=0x%lx eventid=%ld smpl_pmds=0x%lx reset_pmds=0x%lx reloads_pmcs=0x%lx used_monitors=0x%lx ovfl_regs=0x%lx\n",
 			  cnum,
 			  value,
 			  is_loaded,
 			  can_access_pmu,
+			  flags,
 			  ctx->ctx_all_pmcs[0],
 			  ctx->ctx_used_pmds[0],
 			  ctx->ctx_pmds[cnum].eventid,
@@ -3240,8 +3238,8 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			}
 		}
 
-		DPRINT(("pmd[%u]=0x%lx loaded=%d access_pmu=%d, hw_value=0x%lx ctx_pmd=0x%lx  short_reset=0x%lx "
-			  "long_reset=0x%lx notify=%c used_pmds=0x%lx reset_pmds=0x%lx reload_pmds=0x%lx all_pmds=0x%lx ovfl_regs=0x%lx\n",
+		DPRINT(("pmd[%u]=0x%lx ld=%d apmu=%d, hw_value=0x%lx ctx_pmd=0x%lx  short_reset=0x%lx "
+			  "long_reset=0x%lx notify=%c seed=0x%lx mask=0x%lx used_pmds=0x%lx reset_pmds=0x%lx reload_pmds=0x%lx all_pmds=0x%lx ovfl_regs=0x%lx\n",
 			cnum,
 			value,
 			is_loaded,
@@ -3251,6 +3249,8 @@ pfm_write_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 			ctx->ctx_pmds[cnum].short_reset,
 			ctx->ctx_pmds[cnum].long_reset,
 			PMC_OVFL_NOTIFY(ctx, cnum) ? 'Y':'N',
+			ctx->ctx_pmds[cnum].seed,
+			ctx->ctx_pmds[cnum].mask,
 			ctx->ctx_used_pmds[0],
 			ctx->ctx_pmds[cnum].reset_pmds[0],
 			ctx->ctx_reload_pmds[0],
@@ -3328,7 +3328,7 @@ pfm_read_pmds(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	}
 	expert_mode = pfm_sysctl.expert_mode; 
 
-	DPRINT(("loaded=%d access_pmu=%d ctx_state=%d\n",
+	DPRINT(("ld=%d apmu=%d ctx_state=%d\n",
 		is_loaded,
 		can_access_pmu,
 		state));
@@ -3868,7 +3868,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 
 			ctx->ctx_ibrs[rnum] = dbreg.val;
 
-			DPRINT(("write ibr%u=0x%lx used_ibrs=0x%x is_loaded=%d access_pmu=%d\n",
+			DPRINT(("write ibr%u=0x%lx used_ibrs=0x%x ld=%d apmu=%d\n",
 				rnum, dbreg.val, ctx->ctx_used_ibrs[0], is_loaded, can_access_pmu));
 		} else {
 			CTX_USED_DBR(ctx, rnum);
@@ -3879,7 +3879,7 @@ pfm_write_ibr_dbr(int mode, pfm_context_t *ctx, void *arg, int count, struct pt_
 			}
 			ctx->ctx_dbrs[rnum] = dbreg.val;
 
-			DPRINT(("write dbr%u=0x%lx used_dbrs=0x%x is_loaded=%d access_pmu=%d\n",
+			DPRINT(("write dbr%u=0x%lx used_dbrs=0x%x ld=%d apmu=%d\n",
 				rnum, dbreg.val, ctx->ctx_used_dbrs[0], is_loaded, can_access_pmu));
 		}
 	}
@@ -5854,14 +5854,6 @@ pfm_save_regs(struct task_struct *task)
 		return;
 	}
 
-	/*
-	 * sanity check
-	 */
-	if (ctx->ctx_last_activation != GET_ACTIVATION()) {
-		pfm_unprotect_ctx_ctxsw(ctx, flags);
-		return;
-	}
-
 	/*
 	 * save current PSR: needed because we modify it
 	 */

arch/ia64/lib/io.c

@@ -73,6 +73,7 @@ EXPORT_SYMBOL(__ia64_memset_c_io);
 #undef __ia64_writew
 #undef __ia64_writel
 #undef __ia64_writeq
+#undef __ia64_mmiowb
 
 unsigned int
 __ia64_inb (unsigned long port)
@@ -158,4 +159,10 @@ __ia64_readq_relaxed (void __iomem *addr)
 	return ___ia64_readq (addr);
 }
 
+void
+__ia64_mmiowb(void)
+{
+	___ia64_mmiowb();
+}
+
 #endif /* CONFIG_IA64_GENERIC */

arch/ia64/mm/discontig.c

@@ -602,7 +602,7 @@ void call_pernode_memory(unsigned long start, unsigned long len, void *arg)
  * for each piece of usable memory and will setup these values for each node.
  * Very similar to build_maps().
  */
-static int count_node_pages(unsigned long start, unsigned long len, int node)
+static __init int count_node_pages(unsigned long start, unsigned long len, int node)
 {
 	unsigned long end = start + len;
 
@@ -627,7 +627,7 @@ static int count_node_pages(unsigned long start, unsigned long len, int node)
  * paging_init() sets up the page tables for each node of the system and frees
  * the bootmem allocator memory for general use.
  */
-void paging_init(void)
+void __init paging_init(void)
 {
 	unsigned long max_dma;
 	unsigned long zones_size[MAX_NR_ZONES];

arch/ia64/sn/kernel/idle.c

@@ -3,14 +3,13 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (c) 2001-2003 Silicon Graphics, Inc.  All rights reserved.
+ * Copyright (c) 2001-2004 Silicon Graphics, Inc.  All rights reserved.
  */
 
-#include <linux/config.h>
 #include <asm/sn/leds.h>
-#include <asm/sn/simulator.h>
 
-void snidle(int state) {
+void snidle(int state)
+{
 	if (state) {
 		if (pda->idle_flag == 0) {
 			/* 
@@ -19,11 +18,6 @@ void snidle(int state) {
 			set_led_bits(0, LED_CPU_ACTIVITY);
 		}
 
-#ifdef CONFIG_IA64_SGI_SN_SIM
-		if (IS_RUNNING_ON_SIMULATOR())
-			SIMULATOR_SLEEP();
-#endif
-
 		pda->idle_flag = 1;
 	} else {
 		/* 

arch/ia64/sn/kernel/io_init.c

@@ -204,10 +204,12 @@ static void sn_pci_fixup_slot(struct pci_dev *dev)
 	SN_PCIDEV_INFO(dev) = kmalloc(sizeof(struct pcidev_info), GFP_KERNEL);
 	if (SN_PCIDEV_INFO(dev) <= 0)
 		BUG();		/* Cannot afford to run out of memory */
+	memset(SN_PCIDEV_INFO(dev), 0, sizeof(struct pcidev_info));
 
 	sn_irq_info = kmalloc(sizeof(struct sn_irq_info), GFP_KERNEL);
 	if (sn_irq_info <= 0)
 		BUG();		/* Cannot afford to run out of memory */
+	memset(sn_irq_info, 0, sizeof(struct sn_irq_info));
 
 	/* Call to retrieve pci device information needed by kernel. */
 	status = sal_get_pcidev_info((u64) segment, (u64) dev->bus->number, 
@@ -248,7 +250,7 @@ static void sn_pci_fixup_slot(struct pci_dev *dev)
 	SN_PCIDEV_INFO(dev)->pdi_pcibus_info = SN_PCIBUS_BUSSOFT(dev->bus);
 
 	/* Only set up IRQ stuff if this device has a host bus context */
-	if (SN_PCIDEV_BUSSOFT(dev)) {
+	if (SN_PCIDEV_BUSSOFT(dev) && sn_irq_info->irq_irq) {
 		SN_PCIDEV_INFO(dev)->pdi_sn_irq_info = sn_irq_info;
 		dev->irq = SN_PCIDEV_INFO(dev)->pdi_sn_irq_info->irq_irq;
 		sn_irq_fixup(dev, sn_irq_info);

arch/ia64/sn/kernel/iomv.c

@@ -9,6 +9,7 @@
 #include <linux/module.h>
 #include <asm/io.h>
 #include <asm/delay.h>
+#include <asm/sn/nodepda.h>
 #include <asm/sn/simulator.h>
 #include <asm/sn/pda.h>
 #include <asm/sn/sn_cpuid.h>

arch/ia64/sn/kernel/irq.c

@@ -79,7 +79,7 @@ static void sn_ack_irq(unsigned int irq)
 	int nasid;
 
 	irq = irq & 0xff;
-	nasid = smp_physical_node_id();
+	nasid = get_nasid();
 	event_occurred =
 	    HUB_L((uint64_t *) GLOBAL_MMR_ADDR(nasid, SH_EVENT_OCCURRED));
 	if (event_occurred & SH_EVENT_OCCURRED_UART_INT_MASK) {
@@ -109,7 +109,7 @@ static void sn_end_irq(unsigned int irq)
 
 	ivec = irq & 0xff;
 	if (ivec == SGI_UART_VECTOR) {
-		nasid = smp_physical_node_id();
+		nasid = get_nasid();
 		event_occurred = HUB_L((uint64_t *) GLOBAL_MMR_ADDR
 				       (nasid, SH_EVENT_OCCURRED));
 		/* If the UART bit is set here, we may have received an 
@@ -141,8 +141,8 @@ static void sn_set_affinity_irq(unsigned int irq, cpumask_t mask)
 
 	cpuid = first_cpu(mask);
 	cpuphys = cpu_physical_id(cpuid);
-	t_nasid = cpu_physical_id_to_nasid(cpuphys);
-	t_slice = cpu_physical_id_to_slice(cpuphys);
+	t_nasid = cpuid_to_nasid(cpuid);
+	t_slice = cpuid_to_slice(cpuid);
 
 	while (sn_irq_info) {
 		int status;

arch/ia64/sn/kernel/mca.c

@@ -94,15 +94,11 @@ static int
 sn_platform_plat_specific_err_print(const u8 * sect_header, u8 ** oemdata,
 				    u64 * oemdata_size)
 {
-	sal_log_plat_specific_err_info_t *psei =
-	    (sal_log_plat_specific_err_info_t *) sect_header;
-	if (!psei->valid.oem_data)
-		return 0;
 	down(&sn_oemdata_mutex);
 	sn_oemdata = oemdata;
 	sn_oemdata_size = oemdata_size;
 	sn_oemdata_bufsize = 0;
-	ia64_sn_plat_specific_err_print(print_hook, (char *)psei);
+	ia64_sn_plat_specific_err_print(print_hook, (char *)sect_header);
 	up(&sn_oemdata_mutex);
 	return 0;
 }
@@ -110,18 +106,24 @@ sn_platform_plat_specific_err_print(const u8 * sect_header, u8 ** oemdata,
 /* Callback when userspace salinfo wants to decode oem data via the platform
  * kernel and/or prom.
  */
-int sn_salinfo_platform_oemdata(const u8 * sect_header, u8 ** oemdata,
-				u64 * oemdata_size)
+int sn_salinfo_platform_oemdata(const u8 *sect_header, u8 **oemdata, u64 *oemdata_size)
 {
-	efi_guid_t guid = *(efi_guid_t *) sect_header;
+	efi_guid_t guid = *(efi_guid_t *)sect_header;
+	int valid = 0;
 	*oemdata_size = 0;
 	vfree(*oemdata);
 	*oemdata = NULL;
-	if (efi_guidcmp(guid, SAL_PLAT_SPECIFIC_ERR_SECT_GUID) == 0 ||
-	    efi_guidcmp(guid, SAL_PLAT_MEM_DEV_ERR_SECT_GUID) == 0)
-		return sn_platform_plat_specific_err_print(sect_header, oemdata,
-							   oemdata_size);
-	return 0;
+	if (efi_guidcmp(guid, SAL_PLAT_SPECIFIC_ERR_SECT_GUID) == 0) {
+		sal_log_plat_specific_err_info_t *psei = (sal_log_plat_specific_err_info_t *)sect_header;
+		valid = psei->valid.oem_data;
+	} else if (efi_guidcmp(guid, SAL_PLAT_MEM_DEV_ERR_SECT_GUID) == 0) {
+		sal_log_mem_dev_err_info_t *mdei = (sal_log_mem_dev_err_info_t *)sect_header;
+		valid = mdei->valid.oem_data;
+	}
+	if (valid)
+		return sn_platform_plat_specific_err_print(sect_header, oemdata, oemdata_size);
+	else
+		return 0;
 }
 
 static int __init sn_salinfo_init(void)

arch/ia64/sn/kernel/setup.c

@@ -127,6 +127,19 @@ extern char drive_info[4 * 16];
 char drive_info[4 * 16];
 #endif
 
+/*
+ * Get nasid of current cpu early in boot before nodepda is initialized
+ */
+static int
+boot_get_nasid(void)
+{
+	int nasid;
+
+	if (ia64_sn_get_sapic_info(get_sapicid(), &nasid, NULL, NULL))
+		BUG();
+	return nasid;
+}
+
 /*
  * This routine can only be used during init, since
  * smp_boot_data is an init data structure.
@@ -197,7 +210,7 @@ void __init early_sn_setup(void)
 
 	if (IS_RUNNING_ON_SIMULATOR()) {
 		master_node_bedrock_address = (u64 __iomem *)
-			REMOTE_HUB(get_nasid(), SH_JUNK_BUS_UART0);
+			REMOTE_HUB(boot_get_nasid(), SH_JUNK_BUS_UART0);
 		printk(KERN_DEBUG "early_sn_setup: setting "
 		       "master_node_bedrock_address to 0x%p\n",
 		       master_node_bedrock_address);
@@ -297,7 +310,7 @@ void __init sn_setup(char **cmdline_p)
 		panic("PROM version too old\n");
 	}
 
-	master_nasid = get_nasid();
+	master_nasid = boot_get_nasid();
 
 	status =
 	    ia64_sal_freq_base(SAL_FREQ_BASE_REALTIME_CLOCK, &ticks_per_sec,
@@ -314,7 +327,7 @@ void __init sn_setup(char **cmdline_p)
 
 	if (IS_RUNNING_ON_SIMULATOR()) {
 		master_node_bedrock_address = (u64 __iomem *)
-			REMOTE_HUB(get_nasid(), SH_JUNK_BUS_UART0);
+			REMOTE_HUB(boot_get_nasid(), SH_JUNK_BUS_UART0);
 		printk(KERN_DEBUG "sn_setup: setting "
 		       "master_node_bedrock_address to 0x%p\n",
 		       master_node_bedrock_address);
@@ -372,6 +385,8 @@ void __init sn_init_pdas(char **cmdline_p)
 		nodepdaindr[cnode] =
 		    alloc_bootmem_node(NODE_DATA(cnode), sizeof(nodepda_t));
 		memset(nodepdaindr[cnode], 0, sizeof(nodepda_t));
+		memset(nodepdaindr[cnode]->phys_cpuid, -1, 
+		    sizeof(nodepdaindr[cnode]->phys_cpuid));
 	}
 
 	/*
@@ -422,8 +437,10 @@ void __init sn_cpu_init(void)
 	int cpuid;
 	int cpuphyid;
 	int nasid;
+	int subnode;
 	int slice;
 	int cnode;
+	int i;
 	static int wars_have_been_checked;
 
 	/*
@@ -434,10 +451,20 @@ void __init sn_cpu_init(void)
 		return;
 
 	cpuid = smp_processor_id();
-	cpuphyid = ((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff);
-	nasid = cpu_physical_id_to_nasid(cpuphyid);
+	cpuphyid = get_sapicid();
+
+	if (ia64_sn_get_sapic_info(cpuphyid, &nasid, &subnode, &slice))
+		BUG();
+
+	for (i=0; i < NR_NODES; i++) {
+		if (nodepdaindr[i]) {
+			nodepdaindr[i]->phys_cpuid[cpuid].nasid = nasid;
+			nodepdaindr[i]->phys_cpuid[cpuid].slice = slice;
+			nodepdaindr[i]->phys_cpuid[cpuid].subnode = subnode;
+		}
+	}
+
 	cnode = nasid_to_cnodeid(nasid);
-	slice = cpu_physical_id_to_slice(cpuphyid);
 
 	memset(pda, 0, sizeof(pda));
 	pda->p_nodepda = nodepdaindr[cnode];
@@ -575,3 +602,15 @@ static void __init scan_for_ionodes(void)
 	}
 
 }
+
+int
+nasid_slice_to_cpuid(int nasid, int slice)
+{
+	long cpu;
+	
+	for (cpu=0; cpu < NR_CPUS; cpu++) 
+		if (nodepda->phys_cpuid[cpu].nasid == nasid && nodepda->phys_cpuid[cpu].slice == slice)
+			return cpu;
+
+	return -1;
+}

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

@@ -32,6 +32,7 @@
 #include <asm/hw_irq.h>
 #include <asm/current.h>
 #include <asm/sn/sn_cpuid.h>
+#include <asm/sn/sn_sal.h>
 #include <asm/sn/addrs.h>
 #include <asm/sn/shub_mmr.h>
 #include <asm/sn/nodepda.h>
@@ -136,7 +137,7 @@ sn2_global_tlb_purge(unsigned long start, unsigned long end,
 	ptc0 = (long *)GLOBAL_MMR_PHYS_ADDR(0, SH_PTC_0);
 	ptc1 = (long *)GLOBAL_MMR_PHYS_ADDR(0, SH_PTC_1);
 
-	mynasid = smp_physical_node_id();
+	mynasid = get_nasid();
 
 	spin_lock_irqsave(&sn2_global_ptc_lock, flags);
 
@@ -205,6 +206,7 @@ void sn2_ptc_deadlock_recovery(unsigned long data0, unsigned long data1)
 
 /**
  * sn_send_IPI_phys - send an IPI to a Nasid and slice
+ * @nasid: nasid to receive the interrupt (may be outside partition)
  * @physid: physical cpuid to receive the interrupt.
  * @vector: command to send
  * @delivery_mode: delivery mechanism
@@ -219,15 +221,12 @@ void sn2_ptc_deadlock_recovery(unsigned long data0, unsigned long data1)
  * %IA64_IPI_DM_NMI - pend an NMI
  * %IA64_IPI_DM_INIT - pend an INIT interrupt
  */
-void sn_send_IPI_phys(long physid, int vector, int delivery_mode)
+void sn_send_IPI_phys(int nasid, long physid, int vector, int delivery_mode)
 {
-	long nasid, slice, val;
+	long val;
 	unsigned long flags = 0;
 	volatile long *p;
 
-	nasid = cpu_physical_id_to_nasid(physid);
-	slice = cpu_physical_id_to_slice(physid);
-
 	p = (long *)GLOBAL_MMR_PHYS_ADDR(nasid, SH_IPI_INT);
 	val = (1UL << SH_IPI_INT_SEND_SHFT) |
 	    (physid << SH_IPI_INT_PID_SHFT) |
@@ -268,8 +267,14 @@ EXPORT_SYMBOL(sn_send_IPI_phys);
 void sn2_send_IPI(int cpuid, int vector, int delivery_mode, int redirect)
 {
 	long physid;
+	int nasid;
 
 	physid = cpu_physical_id(cpuid);
+	nasid = cpuid_to_nasid(cpuid);
+
+	/* the following is used only when starting cpus at boot time */
+	if (unlikely(nasid == -1))
+		ia64_sn_get_sapic_info(physid, &nasid, NULL, NULL);
 
-	sn_send_IPI_phys(physid, vector, delivery_mode);
+	sn_send_IPI_phys(nasid, physid, vector, delivery_mode);
 }

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

@@ -99,7 +99,7 @@ static int sn_hwperf_geoid_to_cnode(char *location)
 		this_slot = MODULE_GET_BPOS(module_id);
 		this_slab = geo_slab(geoid);
 		if (rack == this_rack && slot == this_slot && slab == this_slab)
-				break;
+			break;
 	}
 
 	return cnode < numionodes ? cnode : -1;
@@ -121,41 +121,35 @@ static int sn_hwperf_generic_ordinal(struct sn_hwperf_object_info *obj,
 	for (ordinal=0, p=objs; p != obj; p++) {
 		if (SN_HWPERF_FOREIGN(p))
 			continue;
-		if (p->location[3] == obj->location[3])
+		if (SN_HWPERF_SAME_OBJTYPE(p, obj))
 			ordinal++;
 	}
 
 	return ordinal;
 }
 
-static struct {
-        char	*brick_chars;
-        char	*brick_name;
-} brick_names[] = {
-        {"c^jbf",   	"node"		},
-        {"r",   	"router"	},	
-	{NULL,		"?-brick"	}
-};
+static const char *slabname_node =	"node"; /* SHub asic */
+static const char *slabname_ionode =	"ionode"; /* TIO asic */
+static const char *slabname_router =	"router"; /* NL3R or NL4R */
+static const char *slabname_other =	"other"; /* unknown asic */
 
-static char *sn_hwperf_get_brickname(struct sn_hwperf_object_info *obj,
+static const char *sn_hwperf_get_slabname(struct sn_hwperf_object_info *obj,
 			struct sn_hwperf_object_info *objs, int *ordinal)
 {
-	int i;
+	int isnode;
+	const char *slabname = slabname_other;
 
-	for (i=0; brick_names[i].brick_chars; i++) {
-		if (strchr(brick_names[i].brick_chars, obj->location[3]))
-			break;
-	}
-
-	if (strcmp(brick_names[i].brick_name, "node") == 0)
+	if ((isnode = SN_HWPERF_IS_NODE(obj)) || SN_HWPERF_IS_IONODE(obj)) {
+	    	slabname = isnode ? slabname_node : slabname_ionode;
 		*ordinal = sn_hwperf_obj_to_cnode(obj);
+	}
 	else {
 		*ordinal = sn_hwperf_generic_ordinal(obj, objs);
-		if (!brick_names[i].brick_chars)
-			brick_names[i].brick_name[0] = obj->location[3];
+		if (SN_HWPERF_IS_ROUTER(obj))
+			slabname = slabname_router;
 	}
 
-	return brick_names[i].brick_name;
+	return slabname;
 }
 
 static int sn_topology_show(struct seq_file *s, void *d)
@@ -165,7 +159,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	int e;
 	int i;
 	int j;
-	const char *brickname;
+	const char *slabname;
 	int ordinal;
 	cpumask_t cpumask;
 	char slice;
@@ -191,11 +185,11 @@ static int sn_topology_show(struct seq_file *s, void *d)
 			obj->name[i] = '_';
 	}
 
-	brickname = sn_hwperf_get_brickname(obj, objs, &ordinal);
-	seq_printf(s, "%s %d %s %s asic %s", brickname, ordinal, obj->location,
+	slabname = sn_hwperf_get_slabname(obj, objs, &ordinal);
+	seq_printf(s, "%s %d %s %s asic %s", slabname, ordinal, obj->location,
 		obj->sn_hwp_this_part ? "local" : "shared", obj->name);
 
-	if (strcmp(brickname, "node") != 0)
+	if (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj))
 		seq_putc(s, '\n');
 	else {
 		seq_printf(s, ", nasid 0x%x", cnodeid_to_nasid(ordinal));
@@ -206,7 +200,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 		seq_putc(s, '\n');
 
 		/*
-		 * CPUs on this node
+		 * CPUs on this node, if any
 		 */
 		cpumask = node_to_cpumask(ordinal);
 		for_each_online_cpu(i) {
@@ -278,9 +272,8 @@ static int sn_topology_show(struct seq_file *s, void *d)
 			 */
 			seq_printf(s, " endpoint %s-%d, protocol %s\n",
 				p->location, ptdata[pt].conn_port,
-				strcmp(obj->name, "NL3Router") == 0 ||
-				strcmp(p->name, "NL3Router") == 0 ?
-				"LLP3" : "LLP4");
+				(SN_HWPERF_IS_NL3ROUTER(obj) ||
+				SN_HWPERF_IS_NL3ROUTER(p)) ?  "LLP3" : "LLP4");
 		}
 		vfree(ptdata);
 	}

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

@@ -80,7 +80,8 @@ static int sn_force_interrupt_open(struct inode *inode, struct file *file)
 
 static int coherence_id_show(struct seq_file *s, void *p)
 {
-	seq_printf(s, "%d\n", cpuid_to_coherence_id(smp_processor_id()));
+	seq_printf(s, "%d\n", partition_coherence_id());
+
 	return 0;
 }
 

drivers/ide/pci/sgiioc4.c

@@ -368,7 +368,7 @@ sgiioc4_INB(unsigned long port)
 }
 
 /* Creates a dma map for the scatter-gather list entries */
-static void __init
+static void __devinit
 ide_dma_sgiioc4(ide_hwif_t * hwif, unsigned long dma_base)
 {
 	int num_ports = sizeof (ioc4_dma_regs_t);
@@ -579,7 +579,7 @@ static int sgiioc4_ide_dma_setup(ide_drive_t *drive)
 	return 0;
 }
 
-static void __init
+static void __devinit
 ide_init_sgiioc4(ide_hwif_t * hwif)
 {
 	hwif->mmio = 2;
@@ -614,7 +614,7 @@ ide_init_sgiioc4(ide_hwif_t * hwif)
 	hwif->INB = &sgiioc4_INB;
 }
 
-static int __init
+static int __devinit
 sgiioc4_ide_setup_pci_device(struct pci_dev *dev, ide_pci_device_t * d)
 {
 	unsigned long base, ctl, dma_base, irqport;
@@ -677,7 +677,7 @@ sgiioc4_ide_setup_pci_device(struct pci_dev *dev, ide_pci_device_t * d)
 	return 0;
 }
 
-static unsigned int __init
+static unsigned int __devinit
 pci_init_sgiioc4(struct pci_dev *dev, ide_pci_device_t * d)
 {
 	unsigned int class_rev;
@@ -730,13 +730,13 @@ static struct pci_device_id sgiioc4_pci_tbl[] = {
 };
 MODULE_DEVICE_TABLE(pci, sgiioc4_pci_tbl);
 
-static struct pci_driver driver = {
+static struct pci_driver __devinitdata driver = {
 	.name = "SGI-IOC4_IDE",
 	.id_table = sgiioc4_pci_tbl,
 	.probe = sgiioc4_init_one,
 };
 
-static int
+static int __devinit
 sgiioc4_ide_init(void)
 {
 	return ide_pci_register_driver(&driver);

drivers/serial/sn_console.c

@@ -108,7 +108,6 @@ static struct sn_cons_port sal_console_port;
 /* Only used if USE_DYNAMIC_MINOR is set to 1 */
 static struct miscdevice misc;	/* used with misc_register for dynamic */
 
-extern u64 __iomem *master_node_bedrock_address;
 extern void early_sn_setup(void);
 
 #undef DEBUG
@@ -124,9 +123,6 @@ static int snt_hw_puts_raw(const char *, int);
 static int snt_hw_puts_buffered(const char *, int);
 static int snt_poll_getc(void);
 static int snt_poll_input_pending(void);
-static int snt_sim_puts(const char *, int);
-static int snt_sim_getc(void);
-static int snt_sim_input_pending(void);
 static int snt_intr_getc(void);
 static int snt_intr_input_pending(void);
 static void sn_transmit_chars(struct sn_cons_port *, int);
@@ -140,14 +136,6 @@ static struct sn_sal_ops poll_ops = {
 	.sal_input_pending = snt_poll_input_pending
 };
 
-/* A table for the simulator */
-static struct sn_sal_ops sim_ops = {
-	.sal_puts_raw = snt_sim_puts,
-	.sal_puts = snt_sim_puts,
-	.sal_getc = snt_sim_getc,
-	.sal_input_pending = snt_sim_input_pending
-};
-
 /* A table for interrupts enabled */
 static struct sn_sal_ops intr_ops = {
 	.sal_puts_raw = snt_hw_puts_raw,
@@ -194,53 +182,6 @@ static int snt_poll_input_pending(void)
 	return !status && input;
 }
 
-/* routines for running the console on the simulator */
-
-/**
- * snt_sim_puts - send to the console, used in simulator mode
- * @str: String to send
- * @count: length of string
- *
- */
-static int snt_sim_puts(const char *str, int count)
-{
-	int counter = count;
-
-#ifdef FLAG_DIRECT_CONSOLE_WRITES
-	/* This is an easy way to pre-pend the output to know whether the output
-	 * was done via sal or directly */
-	writeb('[', master_node_bedrock_address + (UART_TX << 3));
-	writeb('+', master_node_bedrock_address + (UART_TX << 3));
-	writeb(']', master_node_bedrock_address + (UART_TX << 3));
-	writeb(' ', master_node_bedrock_address + (UART_TX << 3));
-#endif				/* FLAG_DIRECT_CONSOLE_WRITES */
-	while (counter > 0) {
-		writeb(*str, master_node_bedrock_address + (UART_TX << 3));
-		counter--;
-		str++;
-	}
-	return count;
-}
-
-/**
- * snt_sim_getc - Get character from console in simulator mode
- *
- */
-static int snt_sim_getc(void)
-{
-	return readb(master_node_bedrock_address + (UART_RX << 3));
-}
-
-/**
- * snt_sim_input_pending - Check if there is input pending in simulator mode
- *
- */
-static int snt_sim_input_pending(void)
-{
-	return readb(master_node_bedrock_address +
-		     (UART_LSR << 3)) & UART_LSR_DR;
-}
-
 /* routines for an interrupt driven console (normal) */
 
 /**
@@ -491,11 +432,7 @@ static int sn_debug_printf(const char *fmt, ...)
 	printed_len = vsnprintf(printk_buf, sizeof(printk_buf), fmt, args);
 
 	if (!sal_console_port.sc_ops) {
-		if (IS_RUNNING_ON_SIMULATOR())
-			sal_console_port.sc_ops = &sim_ops;
-		else
-			sal_console_port.sc_ops = &poll_ops;
-
+		sal_console_port.sc_ops = &poll_ops;
 		early_sn_setup();
 	}
 	sal_console_port.sc_ops->sal_puts_raw(printk_buf, printed_len);
@@ -781,12 +718,8 @@ static void __init sn_sal_switch_to_asynch(struct sn_cons_port *port)
 	spin_lock_irqsave(&port->sc_port.lock, flags);
 
 	/* early_printk invocation may have done this for us */
-	if (!port->sc_ops) {
-		if (IS_RUNNING_ON_SIMULATOR())
-			port->sc_ops = &sim_ops;
-		else
-			port->sc_ops = &poll_ops;
-	}
+	if (!port->sc_ops)
+		port->sc_ops = &poll_ops;
 
 	/* we can't turn on the console interrupt (as request_irq
 	 * calls kmalloc, which isn't set up yet), so we rely on a
@@ -1155,11 +1088,7 @@ int __init sn_serial_console_early_setup(void)
 	if (!ia64_platform_is("sn2"))
 		return -1;
 
-	if (IS_RUNNING_ON_SIMULATOR())
-		sal_console_port.sc_ops = &sim_ops;
-	else
-		sal_console_port.sc_ops = &poll_ops;
-
+	sal_console_port.sc_ops = &poll_ops;
 	early_sn_setup();	/* Find SAL entry points */
 	register_console(&sal_console_early);
 

include/asm-ia64/io.h

@@ -107,7 +107,7 @@ extern int valid_phys_addr_range (unsigned long addr, size_t *count); /* efi.c *
 #define __ia64_mf_a()	ia64_mfa()
 
 /**
- * __ia64_mmiowb - I/O write barrier
+ * ___ia64_mmiowb - I/O write barrier
  *
  * Ensure ordering of I/O space writes.  This will make sure that writes
  * following the barrier will arrive after all previous writes.  For most
@@ -115,7 +115,7 @@ extern int valid_phys_addr_range (unsigned long addr, size_t *count); /* efi.c *
  *
  * See Documentation/DocBook/deviceiobook.tmpl for more information.
  */
-static inline void __ia64_mmiowb(void)
+static inline void ___ia64_mmiowb(void)
 {
 	ia64_mfa();
 }
@@ -162,6 +162,7 @@ __ia64_mk_io_addr (unsigned long port)
 #define __ia64_writew	___ia64_writew
 #define __ia64_writel	___ia64_writel
 #define __ia64_writeq	___ia64_writeq
+#define __ia64_mmiowb	___ia64_mmiowb
 
 /*
  * For the in/out routines, we need to do "mf.a" _after_ doing the I/O access to ensure

include/asm-ia64/pgtable.h

@@ -6,7 +6,7 @@
  * the IA-64 page table tree.
  *
  * This hopefully works with any (fixed) IA-64 page-size, as defined
- * in <asm/page.h> (currently 8192).
+ * in <asm/page.h>.
  *
  * Copyright (C) 1998-2004 Hewlett-Packard Co
  *	David Mosberger-Tang <davidm@hpl.hp.com>
@@ -309,15 +309,15 @@ pgd_index (unsigned long address)
 }
 
 /* The offset in the 1-level directory is given by the 3 region bits
-   (61..63) and the seven level-1 bits (33-39).  */
+   (61..63) and the level-1 bits.  */
 static inline pgd_t*
 pgd_offset (struct mm_struct *mm, unsigned long address)
 {
 	return mm->pgd + pgd_index(address);
 }
 
-/* In the kernel's mapped region we have a full 43 bit space available and completely
-   ignore the region number (since we know its in region number 5). */
+/* In the kernel's mapped region we completely ignore the region number
+   (since we know it's in region number 5). */
 #define pgd_offset_k(addr) \
 	(init_mm.pgd + (((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1)))
 

include/asm-ia64/sn/intr.h

@@ -43,7 +43,7 @@ struct sn_irq_info {
 	int		irq_share_cnt;	/* num devices sharing IRQ   */
 };
 
-extern void sn_send_IPI_phys(long, int, int);
+extern void sn_send_IPI_phys(int, long, int, int);
 
 #define CPU_VECTOR_TO_IRQ(cpuid,vector) (vector)
 

include/asm-ia64/sn/nodepda.h

@@ -30,6 +30,12 @@
  * This structure provides a convenient way of keeping together 
  * all per-node data structures. 
  */
+struct phys_cpuid {
+	short			nasid;
+	char			subnode;
+	char			slice;
+};
+
 struct nodepda_s {
 	void 		*pdinfo;	/* Platform-dependent per-node info */
 	spinlock_t		bist_lock;
@@ -46,6 +52,10 @@ struct nodepda_s {
 	 */
 	struct nodepda_s	*pernode_pdaindr[MAX_COMPACT_NODES]; 
 
+	/*
+	 * Array of physical cpu identifiers. Indexed by cpuid.
+	 */
+	struct phys_cpuid	phys_cpuid[NR_CPUS];
 };
 
 typedef struct nodepda_s nodepda_t;

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

@@ -41,7 +41,15 @@ struct sn_hwperf_object_info {
 #define sn_hwp_is_shared	f.fields.is_shared
 #define sn_hwp_flags		f.b.flags
 
-#define SN_HWPERF_FOREIGN(x)	(!(x)->sn_hwp_this_part && !(x)->sn_hwp_is_shared)
+/* macros for object classification */
+#define SN_HWPERF_IS_NODE(x)		((x) && strstr((x)->name, "SHub"))
+#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"))
+#define SN_HWPERF_FOREIGN(x)		((x) && !(x)->sn_hwp_this_part && !(x)->sn_hwp_is_shared)
+#define SN_HWPERF_SAME_OBJTYPE(x,y)	((SN_HWPERF_IS_NODE(x) && SN_HWPERF_IS_NODE(y)) ||\
+					(SN_HWPERF_IS_IONODE(x) && SN_HWPERF_IS_IONODE(y)) ||\
+					(SN_HWPERF_IS_ROUTER(x) && SN_HWPERF_IS_ROUTER(y)))
 
 /* numa port structure, SN_HWPERF_ENUM_PORTS returns an array of these */
 struct sn_hwperf_port_info {

include/asm-ia64/sn/sn_cpuid.h

@@ -7,6 +7,7 @@
  * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
  */
 
+
 #ifndef _ASM_IA64_SN_SN_CPUID_H
 #define _ASM_IA64_SN_SN_CPUID_H
 
@@ -35,9 +36,6 @@
  *		the boot cpu is 0.
  *			smp_processor_id() returns the cpuid of the current cpu.
  *
- *	   CPUNUM - On IA64, a cpunum and cpuid are the same. This is NOT true
- *		on other architectures like IA32.
- *
  * 	   CPU_PHYSICAL_ID (also known as HARD_PROCESSOR_ID)
  *		This is the same as 31:24 of the processor LID register
  *			hard_smp_processor_id()- cpu_physical_id of current processor
@@ -45,16 +43,16 @@
  *			cpu_logical_id(phy_id) - convert a <physical_cpuid> to a <cpuid> 
  *				* not real efficient - don't use in perf critical code
  *
- *         LID - processor defined register (see PRM V2).
+ *         SLICE - a number in the range of 0 - 3 (typically) that represents the
+ *		cpu number on a brick.
  *
- *           On SN2
- *		31:28 - id   Contains 0-3 to identify the cpu on the node
- *		27:16 - eid  Contains the NASID
+ *	   SUBNODE - (almost obsolete) the number of the FSB that a cpu is
+ *		connected to. This is also the same as the PI number. Usually 0 or 1.
  *
+ *	NOTE!!!: the value of the bits in the cpu physical id (SAPICid or LID) of a cpu has no 
+ *	significance. The SAPIC id (LID) is a 16-bit cookie that has meaning only to the PROM.
  *
  *
- * The following assumes the following mappings for LID register values:
- *
  * The macros convert between cpu physical ids & slice/nasid/cnodeid.
  * These terms are described below:
  *
@@ -83,19 +81,11 @@
  */
 
 #ifndef CONFIG_SMP
-#define cpu_logical_id(cpu)				0
+#define cpu_logical_id(cpu)			0
 #define cpu_physical_id(cpuid)			((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
 #endif
 
-/*
- * macros for some of these exist in sn/addrs.h & sn/arch.h, etc. However, 
- * trying #include these files here causes circular dependencies.
- */
-#define cpu_physical_id_to_nasid(cpi)		((cpi) &0xfff)
-#define cpu_physical_id_to_slice(cpi)		((cpi>>12) & 3)
-#define cpu_physical_id_to_coherence_id(cpi)	(((cpi) & 0x600) >> 9)
-#define get_nasid()				((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xfff)
-#define get_slice()				((ia64_getreg(_IA64_REG_CR_LID) >> 28) & 0xf)
+
 #define get_node_number(addr)			(((unsigned long)(addr)>>38) & 0x7ff)
 
 /*
@@ -103,43 +93,35 @@
  * 
  * NOTE: on non-MP systems, only cpuid 0 exists
  */
-#define id_eid_to_cpu_physical_id(id,eid)	 	(((id)<<8) | (eid))
-
-#define nasid_slice_to_cpuid(nasid,slice)		(cpu_logical_id(nasid_slice_to_cpu_physical_id((nasid),(slice))))
-
-#define nasid_slice_to_cpu_physical_id(nasid, slice)	(((slice)<<12) | (nasid))
 
-/*
- * The following table/struct  is used for managing PTC coherency domains.
- */
-typedef struct {
-	u8	domain;
-	u8	reserved;
-	u16	sapicid;
-} sn_sapicid_info_t;
-
-extern sn_sapicid_info_t	sn_sapicid_info[];	/* indexed by cpuid */
 extern short physical_node_map[];			/* indexed by nasid to get cnode */
 
-
 /*
- * cpuid_to_slice  - convert a cpuid to the slice that it resides on
- *  There are 4 cpus per node. This function returns 0 .. 3)
+ * Macros for retrieving info about current cpu
  */
-#define cpuid_to_slice(cpuid)		(cpu_physical_id_to_slice(cpu_physical_id(cpuid)))
-
+#define get_nasid()			(nodepda->phys_cpuid[smp_processor_id()].nasid)
+#define get_subnode()			(nodepda->phys_cpuid[smp_processor_id()].subnode)
+#define get_slice()			(nodepda->phys_cpuid[smp_processor_id()].slice)
+#define get_cnode()			(nodepda->phys_cpuid[smp_processor_id()].cnode)
+#define get_sapicid()			((ia64_getreg(_IA64_REG_CR_LID) >> 16) & 0xffff)
 
 /*
- * cpuid_to_nasid  - convert a cpuid to the NASID that it resides on
+ * Macros for retrieving info about an arbitrary cpu
+ *	cpuid - logical cpu id
  */
-#define cpuid_to_nasid(cpuid)		(cpu_physical_id_to_nasid(cpu_physical_id(cpuid)))
+#define cpuid_to_nasid(cpuid)		(nodepda->phys_cpuid[cpuid].nasid)
+#define cpuid_to_subnode(cpuid)		(nodepda->phys_cpuid[cpuid].subnode)
+#define cpuid_to_slice(cpuid)		(nodepda->phys_cpuid[cpuid].slice)
+#define cpuid_to_cnodeid(cpuid)		(physical_node_map[cpuid_to_nasid(cpuid)])
 
 
 /*
- * cpuid_to_cnodeid  - convert a cpuid to the cnode that it resides on
+ * Dont use the following in performance critical code. They require scans
+ * of potentially large tables.
  */
-#define cpuid_to_cnodeid(cpuid)		(physical_node_map[cpuid_to_nasid(cpuid)])
-
+extern int nasid_slice_to_cpuid(int, int);
+#define nasid_slice_to_cpu_physical_id(nasid, slice)			\
+	cpu_physical_id(nasid_slice_to_cpuid(nasid, slice))
 
 /*
  * cnodeid_to_nasid - convert a cnodeid to a NASID
@@ -149,36 +131,15 @@ extern short physical_node_map[];			/* indexed by nasid to get cnode */
  */
 #define cnodeid_to_nasid(cnodeid)	pda->cnodeid_to_nasid_table[cnodeid]
  
-
 /*
  * nasid_to_cnodeid - convert a NASID to a cnodeid
  */
 #define nasid_to_cnodeid(nasid)		(physical_node_map[nasid])
 
-
-/*
- * cnode_slice_to_cpuid - convert a codeid & slice to a cpuid
- */
-
-#define cnode_slice_to_cpuid(cnodeid,slice) (nasid_slice_to_cpuid(cnodeid_to_nasid(cnodeid),(slice)))
- 
-
 /*
- * cpuid_to_subnode - convert a cpuid to the subnode it resides on.
- *   slice 0 & 1 are on subnode 0
- *   slice 2 & 3 are on subnode 1.
+ * partition_coherence_id - cget the coherence ID of the current partition
  */
-#define cpuid_to_subnode(cpuid)		((cpuid_to_slice(cpuid)<2) ? 0 : 1)
- 
-
-#define smp_physical_node_id()			(cpuid_to_nasid(smp_processor_id()))
-
-/*
- * cpuid_to_coherence_id - convert a cpuid to the coherence domain id it
- * resides on
- */
-#define cpuid_to_coherence_id(cpuid)    cpu_physical_id_to_coherence_id(cpu_physical_id(cpuid))
-
+#define partition_coherence_id()	(get_nasid() >> 9)
 
 #endif /* _ASM_IA64_SN_SN_CPUID_H */
 

include/asm-ia64/sn/sn_sal.h

@@ -31,6 +31,7 @@
 #define  SN_SAL_NO_FAULT_ZONE_VIRTUAL		   0x02000010
 #define  SN_SAL_NO_FAULT_ZONE_PHYSICAL		   0x02000011
 #define  SN_SAL_PRINT_ERROR			   0x02000012
+#define  SN_SAL_GET_SAPIC_INFO                     0x02009999	//ZZZZ fix
 #define  SN_SAL_SET_ERROR_HANDLING_FEATURES	   0x0200001a	// reentrant
 #define  SN_SAL_GET_FIT_COMPT			   0x0200001b	// reentrant
 #define  SN_SAL_CONSOLE_PUTC                       0x02000021
@@ -843,6 +844,37 @@ ia64_sn_irtr_init(nasid_t nasid, void *buf, int len)
 	return (int) rv.status;
 }
 
+/*
+ * Returns the nasid, subnode & slice corresponding to a SAPIC ID
+ */
+static inline u64
+ia64_sn_get_sapic_info(int sapicid, int *nasid, int *subnode, int *slice)
+{
+	struct ia64_sal_retval ret_stuff;
+
+	ret_stuff.status = 0;
+	ret_stuff.v0 = 0;
+	ret_stuff.v1 = 0;
+	ret_stuff.v2 = 0;
+	SAL_CALL_NOLOCK(ret_stuff, SN_SAL_GET_SAPIC_INFO, sapicid, 0, 0, 0, 0, 0, 0);
+
+/***** BEGIN HACK - temp til new proms available ********/
+	if (ret_stuff.status == SALRET_NOT_IMPLEMENTED) {
+		if (nasid) *nasid = sapicid & 0xfff;
+		if (subnode) *subnode = (sapicid >> 13) & 1;
+		if (slice) *slice = (sapicid >> 12) & 3;
+		return 0;
+	}
+/***** END HACK *******/
+
+	if (ret_stuff.status < 0)
+		return ret_stuff.status;
+
+	if (nasid) *nasid = (int) ret_stuff.v0;
+	if (subnode) *subnode = (int) ret_stuff.v1;
+	if (slice) *slice = (int) ret_stuff.v2;
+	return 0;
+}
 /*
  * This is the access point to the Altix PROM hardware performance
  * and status monitoring interface. For info on using this, see