Merge agluck-lia64.sc.intel.com:/data/home/aegl/BK/Linus

into agluck-lia64.sc.intel.com:/data/home/aegl/BK/linux-ia64-release-2.6.9

arch/ia64/ia32/ia32_support.c

@@ -145,6 +145,9 @@ ia32_gdt_init (void)
 	int cpu = smp_processor_id();
 
 	ia32_shared_page[cpu] = alloc_page(GFP_KERNEL);
+	if (!ia32_shared_page[cpu])
+		panic("failed to allocate ia32_shared_page[%d]\n", cpu);
+
 	cpu_gdt_table[cpu] = page_address(ia32_shared_page[cpu]);
 
 	/* Copy from the boot cpu's GDT */
@@ -161,6 +164,9 @@ ia32_boot_gdt_init (void)
 	unsigned long ldt_size;
 
 	ia32_shared_page[0] = alloc_page(GFP_KERNEL);
+	if (!ia32_shared_page[0])
+		panic("failed to allocate ia32_shared_page[0]\n");
+
 	ia32_boot_gdt = page_address(ia32_shared_page[0]);
 	cpu_gdt_table[0] = ia32_boot_gdt;
 

arch/ia64/kernel/cyclone.c

+#include <linux/module.h>
 #include <linux/smp.h>
 #include <linux/time.h>
 #include <linux/errno.h>
+#include <asm/io.h>
 
 /* IBM Summit (EXA) Cyclone counter code*/
 #define CYCLONE_CBAR_ADDR 0xFEB00CD0

arch/ia64/kernel/head.S

@@ -781,8 +781,7 @@ GLOBAL_ENTRY(ia64_switch_mode_virt)
 
 	// going to virtual
 	//   - for code addresses, set upper bits of addr to KERNEL_START
-	//   - for stack addresses, set upper 3 bits to 0xe.... Dont change any of the
-	//     lower bits since we want it to stay identity mapped
+	//   - for stack addresses, copy from input argument
 	movl r18=KERNEL_START
 	dep r3=0,r3,KERNEL_TR_PAGE_SHIFT,64-KERNEL_TR_PAGE_SHIFT
 	dep r14=0,r14,KERNEL_TR_PAGE_SHIFT,64-KERNEL_TR_PAGE_SHIFT

arch/ia64/kernel/time.c

@@ -371,7 +371,7 @@ ia64_init_itm (void)
 		itc_drift = -1;
 
 	local_cpu_data->itm_delta = (itc_freq + HZ/2) / HZ;
-	printk(KERN_INFO "CPU %d: base freq=%lu.%03luMHz, ITC ratio=%lu/%lu, "
+	printk(KERN_DEBUG "CPU %d: base freq=%lu.%03luMHz, ITC ratio=%lu/%lu, "
 	       "ITC freq=%lu.%03luMHz+/-%ldppm\n", smp_processor_id(),
 	       platform_base_freq / 1000000, (platform_base_freq / 1000) % 1000,
 	       itc_ratio.num, itc_ratio.den, itc_freq / 1000000, (itc_freq / 1000) % 1000,

arch/ia64/lib/swiotlb.c

@@ -11,6 +11,7 @@
  * 03/05/07 davidm	Switch from PCI-DMA to generic device DMA API.
  * 00/12/13 davidm	Rename to swiotlb.c and add mark_clean() to avoid
  *			unnecessary i-cache flushing.
+ * 04/07/.. ak          Better overflow handling. Assorted fixes.
  */
 
 #include <linux/cache.h>
@@ -20,6 +21,7 @@
 #include <linux/spinlock.h>
 #include <linux/string.h>
 #include <linux/types.h>
+#include <linux/ctype.h>
 
 #include <asm/io.h>
 #include <asm/pci.h>
@@ -46,6 +48,8 @@
  */
 #define IO_TLB_SHIFT 11
 
+int swiotlb_force;
+
 /*
  * Used to do a quick range check in swiotlb_unmap_single and swiotlb_sync_single_*, to see
  * if the memory was in fact allocated by this API.
@@ -55,8 +59,16 @@ static char *io_tlb_start, *io_tlb_end;
 /*
  * The number of IO TLB blocks (in groups of 64) betweeen io_tlb_start and io_tlb_end.
  * This is command line adjustable via setup_io_tlb_npages.
+ * Default to 64MB.
+ */
+static unsigned long io_tlb_nslabs = 32768;
+
+/* 
+ * When the IOMMU overflows we return a fallback buffer. This sets the size.
  */
-static unsigned long io_tlb_nslabs = 1024;
+static unsigned long io_tlb_overflow = 32*1024;
+
+void *io_tlb_overflow_buffer; 
 
 /*
  * This is a free list describing the number of free entries available from each index
@@ -78,15 +90,19 @@ static spinlock_t io_tlb_lock = SPIN_LOCK_UNLOCKED;
 static int __init
 setup_io_tlb_npages (char *str)
 {
-	io_tlb_nslabs = simple_strtoul(str, NULL, 0) << (PAGE_SHIFT - IO_TLB_SHIFT);
-
-	/* avoid tail segment of size < IO_TLB_SEGSIZE */
-	io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE);
-
+	if (isdigit(*str)) { 
+		io_tlb_nslabs = simple_strtoul(str, &str, 0) << (PAGE_SHIFT - IO_TLB_SHIFT);
+		/* avoid tail segment of size < IO_TLB_SEGSIZE */
+		io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE);
+	}
+	if (*str == ',')
+		++str;
+	if (!strcmp(str, "force"))
+		swiotlb_force = 1;
 	return 1;
 }
 __setup("swiotlb=", setup_io_tlb_npages);
-
+/* make io_tlb_overflow tunable too? */
 
 /*
  * Statically reserve bounce buffer space and initialize bounce buffer data structures for
@@ -102,7 +118,7 @@ swiotlb_init (void)
 	 */
 	io_tlb_start = alloc_bootmem_low_pages(io_tlb_nslabs * (1 << IO_TLB_SHIFT));
 	if (!io_tlb_start)
-		BUG();
+		panic("Cannot allocate SWIOTLB buffer");
 	io_tlb_end = io_tlb_start + io_tlb_nslabs * (1 << IO_TLB_SHIFT);
 
 	/*
@@ -115,11 +131,23 @@ swiotlb_init (void)
  		io_tlb_list[i] = IO_TLB_SEGSIZE - OFFSET(i, IO_TLB_SEGSIZE);
 	io_tlb_index = 0;
 	io_tlb_orig_addr = alloc_bootmem(io_tlb_nslabs * sizeof(char *));
-
-	printk(KERN_INFO "Placing software IO TLB between 0x%p - 0x%p\n",
-	       (void *) io_tlb_start, (void *) io_tlb_end);
+	
+	/* 
+	 * Get the overflow emergency buffer 
+	 */
+	io_tlb_overflow_buffer = alloc_bootmem_low(io_tlb_overflow); 
+	printk(KERN_INFO "Placing software IO TLB between 0x%lx - 0x%lx\n",
+	       virt_to_phys(io_tlb_start), virt_to_phys(io_tlb_end));
 }
 
+static inline int address_needs_mapping(struct device *hwdev, dma_addr_t addr)
+{ 
+	dma_addr_t mask = 0xffffffff; 
+	if (hwdev && hwdev->dma_mask) 
+		mask = *hwdev->dma_mask; 
+	return (addr & ~mask) != 0; 		
+} 
+
 /*
  * Allocates bounce buffer and returns its kernel virtual address.
  */
@@ -184,11 +212,8 @@ map_single (struct device *hwdev, char *buffer, size_t size, int dir)
 				index = 0;
 		} while (index != wrap);
 
-		/*
-		 * XXX What is a suitable recovery mechanism here?  We cannot
-		 * sleep because we are called from with in interrupts!
-		 */
-		panic("map_single: could not allocate software IO TLB (%ld bytes)", size);
+		spin_unlock_irqrestore(&io_tlb_lock, flags);
+		return NULL;
 	}
   found:
 	spin_unlock_irqrestore(&io_tlb_lock, flags);
@@ -285,7 +310,7 @@ swiotlb_alloc_coherent (struct device *hwdev, size_t size, dma_addr_t *dma_handl
 
 	memset(ret, 0, size);
 	dev_addr = virt_to_phys(ret);
-	if (hwdev && hwdev->dma_mask && (dev_addr & ~*hwdev->dma_mask) != 0)
+	if (address_needs_mapping(hwdev,dev_addr))
 		panic("swiotlb_alloc_consistent: allocated memory is out of range for device");
 	*dma_handle = dev_addr;
 	return ret;
@@ -297,6 +322,28 @@ swiotlb_free_coherent (struct device *hwdev, size_t size, void *vaddr, dma_addr_
 	free_pages((unsigned long) vaddr, get_order(size));
 }
 
+static void swiotlb_full(struct device *dev, size_t size, int dir, int do_panic)
+{
+	/* 
+	 * Ran out of IOMMU space for this operation. This is very bad.
+	 * Unfortunately the drivers cannot handle this operation properly.
+	 * unless they check for pci_dma_mapping_error (most don't)
+	 * When the mapping is small enough return a static buffer to limit
+	 * the damage, or panic when the transfer is too big. 
+	 */ 
+	
+	printk(KERN_ERR 
+  "PCI-DMA: Out of SW-IOMMU space for %lu bytes at device %s\n",
+	       size, dev ? dev->bus_id : "?");
+
+	if (size > io_tlb_overflow && do_panic) {
+		if (dir == PCI_DMA_FROMDEVICE || dir == PCI_DMA_BIDIRECTIONAL)
+			panic("PCI-DMA: Memory would be corrupted\n");
+		if (dir == PCI_DMA_TODEVICE || dir == PCI_DMA_BIDIRECTIONAL) 
+			panic("PCI-DMA: Random memory would be DMAed\n"); 
+	} 
+} 
+
 /*
  * Map a single buffer of the indicated size for DMA in streaming mode.  The PCI address
  * to use is returned.
@@ -308,13 +355,14 @@ dma_addr_t
 swiotlb_map_single (struct device *hwdev, void *ptr, size_t size, int dir)
 {
 	unsigned long dev_addr = virt_to_phys(ptr);
+	void *map; 
 
 	if (dir == DMA_NONE)
 		BUG();
 	/*
 	 * Check if the PCI device can DMA to ptr... if so, just return ptr
 	 */
-	if (hwdev && hwdev->dma_mask && (dev_addr & ~*hwdev->dma_mask) == 0)
+	if (!address_needs_mapping(hwdev, dev_addr) && !swiotlb_force) 
 		/*
 		 * Device is bit capable of DMA'ing to the buffer... just return the PCI
 		 * address of ptr
@@ -324,12 +372,18 @@ swiotlb_map_single (struct device *hwdev, void *ptr, size_t size, int dir)
 	/*
 	 * get a bounce buffer:
 	 */
-	dev_addr = virt_to_phys(map_single(hwdev, ptr, size, dir));
+	map = map_single(hwdev, ptr, size, dir);
+	if (!map) { 
+		swiotlb_full(hwdev, size, dir, 1);		
+		map = io_tlb_overflow_buffer; 
+	}
+
+	dev_addr = virt_to_phys(map);
 
 	/*
 	 * Ensure that the address returned is DMA'ble:
 	 */
-	if (hwdev && hwdev->dma_mask && (dev_addr & ~*hwdev->dma_mask) != 0)
+	if (address_needs_mapping(hwdev, dev_addr))
 		panic("map_single: bounce buffer is not DMA'ble");
 
 	return dev_addr;
@@ -437,9 +491,17 @@ swiotlb_map_sg (struct device *hwdev, struct scatterlist *sg, int nelems, int di
 	for (i = 0; i < nelems; i++, sg++) {
 		addr = SG_ENT_VIRT_ADDRESS(sg);
 		dev_addr = virt_to_phys(addr);
-		if (hwdev && hwdev->dma_mask && (dev_addr & ~*hwdev->dma_mask) != 0)
-			sg->dma_address = (dma_addr_t) map_single(hwdev, addr, sg->length, dir);
-		else
+		if (swiotlb_force || address_needs_mapping(hwdev, dev_addr)) {
+			sg->dma_address = (dma_addr_t) virt_to_phys(map_single(hwdev, addr, sg->length, dir));
+			if (!sg->dma_address) {
+				/* Don't panic here, we expect pci_map_sg users
+				   to do proper error handling. */
+				swiotlb_full(hwdev, sg->length, dir, 0); 
+				swiotlb_unmap_sg(hwdev, sg - i, i, dir);
+				sg[0].dma_length = 0;
+				return 0;
+			}
+		} else
 			sg->dma_address = dev_addr;
 		sg->dma_length = sg->length;
 	}
@@ -460,7 +522,7 @@ swiotlb_unmap_sg (struct device *hwdev, struct scatterlist *sg, int nelems, int
 
 	for (i = 0; i < nelems; i++, sg++)
 		if (sg->dma_address != SG_ENT_PHYS_ADDRESS(sg))
-			unmap_single(hwdev, (void *) sg->dma_address, sg->dma_length, dir);
+			unmap_single(hwdev, (void *) phys_to_virt(sg->dma_address), sg->dma_length, dir);
 		else if (dir == DMA_FROM_DEVICE)
 			mark_clean(SG_ENT_VIRT_ADDRESS(sg), sg->dma_length);
 }
@@ -501,7 +563,7 @@ swiotlb_sync_sg_for_device (struct device *hwdev, struct scatterlist *sg, int ne
 int
 swiotlb_dma_mapping_error (dma_addr_t dma_addr)
 {
-	return 0;
+	return (dma_addr == virt_to_phys(io_tlb_overflow_buffer));
 }
 
 /*

arch/ia64/sn/fakeprom/Makefile

-# arch/ia64/sn/fakeprom/Makefile
-#
-# This file is subject to the terms and conditions of the GNU General Public
-# License.  See the file "COPYING" in the main directory of this archive
-# for more details.
-#
-#  Copyright (c) 2000-2003 Silicon Graphics, Inc.  All rights reserved.
-#
-# Medusa fake PROM support
-#
-
-EXTRA_TARGETS := fpromasm.o main.o fw-emu.o fpmem.o klgraph_init.o \
-		 fprom vmlinux.sym
-
-OBJS := $(obj)/fpromasm.o $(obj)/main.o $(obj)/fw-emu.o $(obj)/fpmem.o \
-	$(obj)/klgraph_init.o
-
-LDFLAGS_fprom = -static -T
-
-.PHONY: fprom
-
-fprom: $(obj)/fprom
-
-$(obj)/fprom: $(src)/fprom.lds $(OBJS) arch/ia64/lib/lib.a FORCE
-	$(call if_changed,ld)
-
-$(obj)/vmlinux.sym: $(src)/make_textsym System.map
-	$(src)/make_textsym vmlinux > vmlinux.sym
-	$(call cmd,cptotop)

arch/ia64/sn/fakeprom/README

-/*
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (c) 2002-2003 Silicon Graphics, Inc.  All Rights Reserved.
- */
-
-This directory contains the files required to build
-the fake PROM image that is currently being used to
-boot IA64 kernels running under the SGI Medusa kernel.
-
-The FPROM currently provides the following functions:
-
-	- PAL emulation for all PAL calls we've made so far.
-	- SAL emulation for all SAL calls we've made so far.
-	- EFI emulation for all EFI calls we've made so far.
-	- builds the "ia64_bootparam" structure that is
-	  passed to the kernel from SAL. This structure 
-	  shows the cpu & memory configurations.
-	- supports medusa boottime options for changing
-	  the number of cpus present
-	- supports medusa boottime options for changing
-	  the memory configuration.
-
-
-
-At some point, this fake PROM will be replaced by the
-real PROM.
-
-
-
-
-To build a fake PROM, cd to this directory & type:
-
-	make
-
-This will (or should) build a fake PROM named "fprom".
-
-
-
-
-Use this fprom image when booting the Medusa simulator. The
-control file used to boot Medusa should include the 
-following lines:
-
-	load fprom
-	load vmlinux
-	sr pc 0x100000
-	sr g 9 <address of kernel _start function> #(currently 0xe000000000520000)
-
-NOTE: There is a script "runsim" in this directory that can be used to
-simplify setting up an environment for running under Medusa.
-
-
-
-
-The following parameters may be passed to the fake PROM to
-control the PAL/SAL/EFI parameters passed to the kernel:
-
-	GR[8] = # of cpus
-	GR[9] = address of primary entry point into the kernel
-	GR[20] = memory configuration for node 0
-	GR[21] = memory configuration for node 1
-	GR[22] = memory configuration for node 2
-	GR[23] = memory configuration for node 3
-
-
-Registers GR[20] - GR[23] contain information to specify the
-amount of memory present on nodes 0-3.
-
-  - if nothing is specified (all registers are 0), the configuration
-    defaults to 8 MB on node 0.
-
-  - a mem config entry for node N is passed in GR[20+N]
-
-  - a mem config entry consists of 8 hex digits. Each digit gives the
-    amount of physical memory available on the node starting at
-    1GB*<dn>, where dn is the digit number. The amount of memory
-    is 8MB*2**<d>. (If <d> = 0, the memory size is 0).
-
-    SN1 doesn't support dimms this small but small memory systems 
-    boot faster on Medusa.
-
-
-
-An example helps a lot. The following specifies that node 0 has
-physical memory 0 to 8MB and 1GB to 1GB+32MB, and that node 1 has
-64MB starting at address 0 of the node which is 8GB.
-
-      gr[20] = 0x21           # 0 to 8MB, 1GB to 1GB+32MB
-      gr[21] = 0x4            # 8GB to 8GB+64MB
-

arch/ia64/sn/fakeprom/fpmem.c

-/* 
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 2000-2003 Silicon Graphics, Inc.  All rights reserved.
- */
-
-
-
-/*
- * FPROM EFI memory descriptor build routines
- *
- * 	- Routines to build the EFI memory descriptor map
- * 	- Should also be usable by the SGI prom to convert
- * 	  klconfig to efi_memmap
- */
-
-#include <linux/config.h>
-#include <linux/efi.h>
-#include "fpmem.h"
-
-/*
- * args points to a layout in memory like this
- *
- *		32 bit		32 bit
- *
- * 		numnodes	numcpus
- *
- *		16 bit   16 bit		   32 bit
- *		nasid0	cpuconf		membankdesc0
- *		nasid1	cpuconf		membankdesc1
- *			   .
- *			   .
- *			   .
- *			   .
- *			   .
- */
-
-sn_memmap_t	*sn_memmap ;
-sn_config_t	*sn_config ;
-
-/*
- * There is a hole in the node 0 address space. Dont put it
- * in the memory map
- */
-#define NODE0_HOLE_SIZE         (20*MB)
-#define NODE0_HOLE_END          (4UL*GB)
-
-#define	MB			(1024*1024)
-#define GB			(1024*MB)
-#define KERNEL_SIZE		(4*MB)
-#define PROMRESERVED_SIZE	(1*MB)
-
-#ifdef SGI_SN2
-#define PHYS_ADDRESS(_n, _x)		(((long)_n<<38) | (long)_x | 0x3000000000UL)
-#define MD_BANK_SHFT 34
-#endif
-
-/*
- * For SN, this may not take an arg and gets the numnodes from 
- * the prom variable or by traversing klcfg or promcfg
- */
-int
-GetNumNodes(void)
-{
-	return sn_config->nodes;
-}
-
-int
-GetNumCpus(void)
-{
-	return sn_config->cpus;
-}
-
-/* For SN, get the index th nasid */
-
-int
-GetNasid(int index)
-{
-	return sn_memmap[index].nasid ;
-}
-
-node_memmap_t
-GetMemBankInfo(int index)
-{
-	return sn_memmap[index].node_memmap ;
-}
-
-int
-IsCpuPresent(int cnode, int cpu)
-{
-	return  sn_memmap[cnode].cpuconfig & (1UL<<cpu);
-}
-
-
-/*
- * Made this into an explicit case statement so that
- * we can assign specific properties to banks like bank0
- * actually disabled etc.
- */
-
-#ifdef SGI_SN2
-int
-IsBankPresent(int index, node_memmap_t nmemmap)
-{
-	switch (index) {
-		case 0:return BankPresent(nmemmap.b0size);
-		case 1:return BankPresent(nmemmap.b1size);
-		case 2:return BankPresent(nmemmap.b2size);
-		case 3:return BankPresent(nmemmap.b3size);
-		default:return -1 ;
-	}
-}
-
-int
-GetBankSize(int index, node_memmap_t nmemmap)
-{
-	/*
-	 * Add 2 because there are 4 dimms per bank.
-	 */
-        switch (index) {
-                case 0:return 2 + ((long)nmemmap.b0size + nmemmap.b0dou);
-                case 1:return 2 + ((long)nmemmap.b1size + nmemmap.b1dou);
-                case 2:return 2 + ((long)nmemmap.b2size + nmemmap.b2dou);
-                case 3:return 2 + ((long)nmemmap.b3size + nmemmap.b3dou);
-                default:return -1 ;
-        }
-}
-
-#endif
-
-void
-build_mem_desc(efi_memory_desc_t *md, int type, long paddr, long numbytes, long attr)
-{
-        md->type = type;
-        md->phys_addr = paddr;
-        md->virt_addr = 0;
-        md->num_pages = numbytes >> 12;
-        md->attribute = attr;
-}
-
-int
-build_efi_memmap(void *md, int mdsize)
-{
-	int		numnodes = GetNumNodes() ;
-	int		cnode,bank ;
-	int		nasid ;
-	node_memmap_t	membank_info ;
-	int		bsize;
-	int		count = 0 ;
-	long		paddr, hole, numbytes;
-
-
-	for (cnode=0;cnode<numnodes;cnode++) {
-		nasid = GetNasid(cnode) ;
-		membank_info = GetMemBankInfo(cnode) ;
-		for (bank=0;bank<MD_BANKS_PER_NODE;bank++) {
-			if (IsBankPresent(bank, membank_info)) {
-				bsize = GetBankSize(bank, membank_info) ;
-                                paddr = PHYS_ADDRESS(nasid, (long)bank<<MD_BANK_SHFT);
-                                numbytes = BankSizeBytes(bsize);
-#ifdef SGI_SN2
-				/* 
-				 * Ignore directory.
-				 * Shorten memory chunk by 1 page - makes a better
-				 * testcase & is more like the real PROM.
-				 */
-				numbytes = numbytes * 31 / 32;
-#endif
-				/*
-				 * Only emulate the memory prom grabs
-				 * if we have lots of memory, to allow
-				 * us to simulate smaller memory configs than
-				 * we can actually run on h/w.  Otherwise,
-				 * linux throws away a whole "granule".
-				 */
-				if (cnode == 0 && bank == 0 &&
-				    numbytes > 128*1024*1024) {
-					numbytes -= 1000;
-				}
-
-                                /*
-                                 * Check for the node 0 hole. Since banks cant
-                                 * span the hole, we only need to check if the end of
-                                 * the range is the end of the hole.
-                                 */
-                                if (paddr+numbytes == NODE0_HOLE_END)
-                                        numbytes -= NODE0_HOLE_SIZE;
-                                /*
-                                 * UGLY hack - we must skip overr the kernel and
-                                 * PROM runtime services but we dont exactly where it is.
-                                 * So lets just reserve:
-				 *	node 0
-				 *		0-1MB for PAL
-				 *		1-4MB for SAL
-				 *	node 1-N
-				 *		0-1 for SAL
-                                 */
-                                if (bank == 0) {
-					if (cnode == 0) {
-						hole = 2*1024*1024;
-						build_mem_desc(md, EFI_PAL_CODE, paddr, hole, EFI_MEMORY_WB|EFI_MEMORY_WB);
-						numbytes -= hole;
-                                        	paddr += hole;
-			        		count++ ;
-                                        	md += mdsize;
-						hole = 1*1024*1024;
-						build_mem_desc(md, EFI_CONVENTIONAL_MEMORY, paddr, hole, EFI_MEMORY_UC);
-						numbytes -= hole;
-                                        	paddr += hole;
-			        		count++ ;
-                                        	md += mdsize;
-						hole = 1*1024*1024;
-						build_mem_desc(md, EFI_RUNTIME_SERVICES_DATA, paddr, hole, EFI_MEMORY_WB|EFI_MEMORY_WB);
-						numbytes -= hole;
-                                        	paddr += hole;
-			        		count++ ;
-                                        	md += mdsize;
-					} else {
-						hole = 2*1024*1024;
-                                        	build_mem_desc(md, EFI_RUNTIME_SERVICES_DATA, paddr, hole, EFI_MEMORY_WB|EFI_MEMORY_WB);
-						numbytes -= hole;
-                                        	paddr += hole;
-			        		count++ ;
-                                        	md += mdsize;
-						hole = 2*1024*1024;
-                                        	build_mem_desc(md, EFI_RUNTIME_SERVICES_DATA, paddr, hole, EFI_MEMORY_UC);
-						numbytes -= hole;
-                                        	paddr += hole;
-			        		count++ ;
-                                        	md += mdsize;
-					}
-                                }
-                                build_mem_desc(md, EFI_CONVENTIONAL_MEMORY, paddr, numbytes, EFI_MEMORY_WB|EFI_MEMORY_WB);
-
-			        md += mdsize ;
-			        count++ ;
-			}
-		}
-	}
-	return count ;
-}
-
-void
-build_init(unsigned long args)
-{
-	sn_config = (sn_config_t *) (args);	
-	sn_memmap = (sn_memmap_t *)(args + 8) ; /* SN equiv for this is */
-						/* init to klconfig start */
-}

arch/ia64/sn/fakeprom/fpmem.h

-/* 
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 2000-2003 Silicon Graphics, Inc.  All rights reserved.
- */
-
-#include <linux/config.h>
-
-/*
- * Structure of the mem config of the node as a SN MI reg
- * Medusa supports this reg config.
- *
- * BankSize nibble to bank size mapping
- *
- *      1 - 64 MB
- *      2 - 128 MB
- *      3 - 256 MB
- *      4 - 512 MB
- *      5 - 1024 MB (1GB)
- */
-
-#define MBSHIFT				20
-
-#ifdef SGI_SN2
-typedef struct node_memmap_s
-{
-        unsigned int    b0size  :3,     /* 0-2   bank 0 size */
-                        b0dou   :1,     /* 3     bank 0 is 2-sided */
-                        ena0    :1,     /* 4     bank 0 enabled */
-                        r0      :3,     /* 5-7   reserved */
-        		b1size  :3,     /* 8-10  bank 1 size */
-                        b1dou   :1,     /* 11    bank 1 is 2-sided */
-                        ena1    :1,     /* 12    bank 1 enabled */
-                        r1      :3,     /* 13-15 reserved */
-        		b2size  :3,     /* 16-18 bank 2 size */
-                        b2dou   :1,     /* 19    bank 1 is 2-sided */
-                        ena2    :1,     /* 20    bank 2 enabled */
-                        r2      :3,     /* 21-23 reserved */
-        		b3size  :3,     /* 24-26 bank 3 size */
-                        b3dou   :1,     /* 27    bank 3 is 2-sided */
-                        ena3    :1,     /* 28    bank 3 enabled */
-                        r3      :3;     /* 29-31 reserved */
-} node_memmap_t ;
-
-#define SN2_BANK_SIZE_SHIFT		(MBSHIFT+6)     /* 64 MB */
-#define BankPresent(bsize)		(bsize<6)
-#define BankSizeBytes(bsize)            (BankPresent(bsize) ? 1UL<<((bsize)+SN2_BANK_SIZE_SHIFT) : 0)
-#define MD_BANKS_PER_NODE 4
-#define MD_BANKSIZE			(1UL << 34)
-#endif
-
-typedef struct sn_memmap_s
-{
-	short		nasid ;
-	short		cpuconfig;
-	node_memmap_t 	node_memmap ;
-} sn_memmap_t ;
-
-typedef struct sn_config_s
-{
-	int		cpus;
-	int		nodes;
-	sn_memmap_t	memmap[1];		/* start of array */
-} sn_config_t;
-
-
-
-extern void build_init(unsigned long);
-extern int build_efi_memmap(void *, int);
-extern int GetNumNodes(void);
-extern int GetNumCpus(void);
-extern int IsCpuPresent(int, int);
-extern int GetNasid(int);

arch/ia64/sn/fakeprom/fprom.lds

-/*
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (c) 2002-2003 Silicon Graphics, Inc.  All Rights Reserved.
- */
-
-OUTPUT_FORMAT("elf64-ia64-little")
-OUTPUT_ARCH(ia64)
-ENTRY(_start)
-SECTIONS
-{
-  v = 0x0000000000000000 ;	/* this symbol is here to make debugging with kdb easier... */
-
-  . = (0x000000000000000  + 0x100000) ;
-
-  _text = .;
-  .text : AT(ADDR(.text) - 0x0000000000000000 )
-    {
-	*(__ivt_section)
-	/* these are not really text pages, but the zero page needs to be in a fixed location: */
-	*(__special_page_section)
-	__start_gate_section = .;
-	*(__gate_section)
-	__stop_gate_section = .;
-	*(.text)
-    }
-
-  /* Global data */
-  _data = .;
-
-  .rodata : AT(ADDR(.rodata) - 0x0000000000000000 )
-	{ *(.rodata) *(.rodata.*) }
-  .opd : AT(ADDR(.opd) - 0x0000000000000000 )
-	{ *(.opd) }
-  .data : AT(ADDR(.data) - 0x0000000000000000 )
-	{ *(.data) *(.gnu.linkonce.d*) CONSTRUCTORS }
-
-  __gp = ALIGN (8) + 0x200000;
-
-  .got : AT(ADDR(.got) - 0x0000000000000000 )
-	{ *(.got.plt) *(.got) }
-  /* We want the small data sections together, so single-instruction offsets
-     can access them all, and initialized data all before uninitialized, so
-     we can shorten the on-disk segment size.  */
-  .sdata : AT(ADDR(.sdata) - 0x0000000000000000 )
-	{ *(.sdata) }
-  _edata  =  .;
-  _bss = .;
-  .sbss : AT(ADDR(.sbss) - 0x0000000000000000 )
-	{ *(.sbss) *(.scommon) }
-  .bss : AT(ADDR(.bss) - 0x0000000000000000 )
-	{ *(.bss) *(COMMON) }
-  . = ALIGN(64 / 8);
-  _end = .;
-
-  /* Sections to be discarded */
-  /DISCARD/ : {
-	*(.text.exit)
-	*(.data.exit)
-	}
-
-  /* Stabs debugging sections.  */
-  .stab 0 : { *(.stab) }
-  .stabstr 0 : { *(.stabstr) }
-  .stab.excl 0 : { *(.stab.excl) }
-  .stab.exclstr 0 : { *(.stab.exclstr) }
-  .stab.index 0 : { *(.stab.index) }
-  .stab.indexstr 0 : { *(.stab.indexstr) }
-  /* DWARF debug sections.
-     Symbols in the DWARF debugging sections are relative to the beginning
-     of the section so we begin them at 0.  */
-  /* DWARF 1 */
-  .debug          0 : { *(.debug) }
-  .line           0 : { *(.line) }
-  /* GNU DWARF 1 extensions */
-  .debug_srcinfo  0 : { *(.debug_srcinfo) }
-  .debug_sfnames  0 : { *(.debug_sfnames) }
-  /* DWARF 1.1 and DWARF 2 */
-  .debug_aranges  0 : { *(.debug_aranges) }
-  .debug_pubnames 0 : { *(.debug_pubnames) }
-  /* DWARF 2 */
-  .debug_info     0 : { *(.debug_info) }
-  .debug_abbrev   0 : { *(.debug_abbrev) }
-  .debug_line     0 : { *(.debug_line) }
-  .debug_frame    0 : { *(.debug_frame) }
-  .debug_str      0 : { *(.debug_str) }
-  .debug_loc      0 : { *(.debug_loc) }
-  .debug_macinfo  0 : { *(.debug_macinfo) }
-  /* SGI/MIPS DWARF 2 extensions */
-  .debug_weaknames 0 : { *(.debug_weaknames) }
-  .debug_funcnames 0 : { *(.debug_funcnames) }
-  .debug_typenames 0 : { *(.debug_typenames) }
-  .debug_varnames  0 : { *(.debug_varnames) }
-  /* These must appear regardless of  .  */
-  /* Discard them for now since Intel SoftSDV cannot handle them.
-  .comment 0 : { *(.comment) }
-  .note 0 : { *(.note) }
-  */
-  /DISCARD/ : { *(.comment) }
-  /DISCARD/ : { *(.note) }
-}

arch/ia64/sn/fakeprom/main.c

-/* 
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 2000-2003 Silicon Graphics, Inc.  All rights reserved.
- */
-
-
-
-#include <linux/config.h>
-#include <linux/types.h>
-#include <asm/bitops.h>
-
-extern void klgraph_init(void);
-void bedrock_init(int);
-void synergy_init(int, int);
-void sys_fw_init (const char *args, int arglen, int bsp);
-
-volatile int	bootmaster=0;		/* Used to pick bootmaster */
-volatile int	nasidmaster[128]={0};	/* Used to pick node/synergy masters */
-int		init_done=0;
-extern int	bsp_lid;
-
-#define get_bit(b,p)	(((*p)>>(b))&1)
-
-int
-fmain(int lid, int bsp) {
-	int	syn, nasid, cpu;
-
-	/*
-	 * First lets figure out who we are. This is done from the
-	 * LID passed to us.
-	 */
-	nasid = (lid>>16)&0xfff;
-	cpu = (lid>>28)&3;
-	syn = 0;
-	
-	/*
-	 * Now pick a nasid master to initialize Bedrock registers.
-	 */
-	if (test_and_set_bit(8, &nasidmaster[nasid]) == 0) {
-		bedrock_init(nasid);
-		test_and_set_bit(9, &nasidmaster[nasid]);
-	} else
-		while (get_bit(9, &nasidmaster[nasid]) == 0);
-	
-
-	/*
-	 * Now pick a BSP & finish init.
-	 */
-	if (test_and_set_bit(0, &bootmaster) == 0) {
-		sys_fw_init(0, 0, bsp);
-		test_and_set_bit(1, &bootmaster);
-	} else
-		while (get_bit(1, &bootmaster) == 0);
-
-	return (lid == bsp_lid);
-}
-
-
-void
-bedrock_init(int nasid)
-{
-	nasid = nasid;		/* to quiet gcc */
-#if 0
-	/*
-	 * Undef if you need fprom to generate a 1 node klgraph
-	 * information .. only works for 1 node for nasid 0.
-	 */
-	klgraph_init();
-#endif
-}
-
-
-void
-synergy_init(int nasid, int syn)
-{
-	long	*base;
-	long	off;
-
-	/*
-	 * Enable all FSB flashed interrupts.
-	 * I'd really like defines for this......
-	 */
-	base = (long*)0x80000e0000000000LL;		/* base of synergy regs */
-	for (off = 0x2a0; off < 0x2e0; off+=8)		/* offset for VEC_MASK_{0-3}_A/B */
-		*(base+off/8) = -1LL;
-
-	/*
-	 * Set the NASID in the FSB_CONFIG register.
-	 */
-	base = (long*)0x80000e0000000450LL;
-	*base = (long)((nasid<<16)|(syn<<9));
-}
-
-
-/* Why isnt there a bcopy/memcpy in lib64.a */
-
-void* 
-memcpy(void * dest, const void *src, size_t count)
-{
-	char *s, *se, *d;
-
-	for(d=dest, s=(char*)src, se=s+count; s<se; s++, d++)
-		*d = *s;
-	return dest;
-}

arch/ia64/sn/fakeprom/make_textsym

-#!/bin/sh
-#
-# Build a textsym file for use in the Arium ITP probe.
-#
-#
-# This file is subject to the terms and conditions of the GNU General Public
-# License.  See the file "COPYING" in the main directory of this archive
-# for more details.
-#
-# Copyright (c) 2001-2003 Silicon Graphics, Inc.  All rights reserved.
-#
-
-help() {
-cat <<END
-Build a WinDD "symtxt" file for use with the Arium ECM-30 probe.
-
-	Usage: $0 [<vmlinux file> [<output file>]]
-		If no input file is specified, it defaults to vmlinux.
-		If no output file name is specified, it defaults to "textsym".
-END
-exit 1
-}
-
-err () {
-	echo "ERROR - $*" >&2
-	exit 1
-}
-
-
-OPTS="H"
-while getopts "$OPTS" c ; do
-	case $c in
-		H)  help;;
-		\?) help;;
-	esac
-
-done
-shift `expr $OPTIND - 1`
-
-#OBJDUMP=/usr/bin/ia64-linux-objdump
-LINUX=${1:-vmlinux}
-TEXTSYM=${2:-${LINUX}.sym}
-TMPSYM=${2:-${LINUX}.sym.tmp}
-trap "/bin/rm -f $TMPSYM" 0
-
-[ -f $VMLINUX ] || help
-
-$OBJDUMP -t $LINUX | egrep -v '__ks' | sort > $TMPSYM
-SN1=`egrep "dig_setup|Synergy_da_indr" $TMPSYM|wc -l`
-
-# Dataprefix and textprefix correspond to the VGLOBAL_BASE and VPERNODE_BASE.
-# Eventually, these values should be:
-#	dataprefix	ffffffff
-#	textprefix	fffffffe
-# but right now they're still changing, so make them dynamic.
-dataprefix=`awk ' / \.data	/ { print substr($1, 0, 8) ; exit ; }' $TMPSYM`
-textprefix=`awk ' / \.text	/ { print substr($1, 0, 8) ; exit ; }' $TMPSYM`
-
-# pipe everything thru sort
-echo "TEXTSYM V1.0"
-(cat <<END 
-GLOBAL | ${textprefix}00000000 | CODE | VEC_VHPT_Translation_0000
-GLOBAL | ${textprefix}00000400 | CODE | VEC_ITLB_0400
-GLOBAL | ${textprefix}00000800 | CODE | VEC_DTLB_0800
-GLOBAL | ${textprefix}00000c00 | CODE | VEC_Alt_ITLB_0c00
-GLOBAL | ${textprefix}00001000 | CODE | VEC_Alt_DTLB_1000
-GLOBAL | ${textprefix}00001400 | CODE | VEC_Data_nested_TLB_1400
-GLOBAL | ${textprefix}00001800 | CODE | VEC_Instruction_Key_Miss_1800
-GLOBAL | ${textprefix}00001c00 | CODE | VEC_Data_Key_Miss_1c00
-GLOBAL | ${textprefix}00002000 | CODE | VEC_Dirty-bit_2000
-GLOBAL | ${textprefix}00002400 | CODE | VEC_Instruction_Access-bit_2400
-GLOBAL | ${textprefix}00002800 | CODE | VEC_Data_Access-bit_2800
-GLOBAL | ${textprefix}00002c00 | CODE | VEC_Break_instruction_2c00
-GLOBAL | ${textprefix}00003000 | CODE | VEC_External_Interrupt_3000
-GLOBAL | ${textprefix}00003400 | CODE | VEC_Reserved_3400
-GLOBAL | ${textprefix}00003800 | CODE | VEC_Reserved_3800
-GLOBAL | ${textprefix}00003c00 | CODE | VEC_Reserved_3c00
-GLOBAL | ${textprefix}00004000 | CODE | VEC_Reserved_4000
-GLOBAL | ${textprefix}00004400 | CODE | VEC_Reserved_4400
-GLOBAL | ${textprefix}00004800 | CODE | VEC_Reserved_4800
-GLOBAL | ${textprefix}00004c00 | CODE | VEC_Reserved_4c00
-GLOBAL | ${textprefix}00005000 | CODE | VEC_Page_Not_Present_5000
-GLOBAL | ${textprefix}00005100 | CODE | VEC_Key_Permission_5100
-GLOBAL | ${textprefix}00005200 | CODE | VEC_Instruction_Access_Rights_5200
-GLOBAL | ${textprefix}00005300 | CODE | VEC_Data_Access_Rights_5300
-GLOBAL | ${textprefix}00005400 | CODE | VEC_General_Exception_5400
-GLOBAL | ${textprefix}00005500 | CODE | VEC_Disabled_FP-Register_5500
-GLOBAL | ${textprefix}00005600 | CODE | VEC_Nat_Consumption_5600
-GLOBAL | ${textprefix}00005700 | CODE | VEC_Speculation_5700
-GLOBAL | ${textprefix}00005800 | CODE | VEC_Reserved_5800
-GLOBAL | ${textprefix}00005900 | CODE | VEC_Debug_5900
-GLOBAL | ${textprefix}00005a00 | CODE | VEC_Unaligned_Reference_5a00
-GLOBAL | ${textprefix}00005b00 | CODE | VEC_Unsupported_Data_Reference_5b00
-GLOBAL | ${textprefix}00005c00 | CODE | VEC_Floating-Point_Fault_5c00
-GLOBAL | ${textprefix}00005d00 | CODE | VEC_Floating_Point_Trap_5d00
-GLOBAL | ${textprefix}00005e00 | CODE | VEC_Lower_Privilege_Tranfer_Trap_5e00
-GLOBAL | ${textprefix}00005f00 | CODE | VEC_Taken_Branch_Trap_5f00
-GLOBAL | ${textprefix}00006000 | CODE | VEC_Single_Step_Trap_6000
-GLOBAL | ${textprefix}00006100 | CODE | VEC_Reserved_6100
-GLOBAL | ${textprefix}00006200 | CODE | VEC_Reserved_6200
-GLOBAL | ${textprefix}00006300 | CODE | VEC_Reserved_6300
-GLOBAL | ${textprefix}00006400 | CODE | VEC_Reserved_6400
-GLOBAL | ${textprefix}00006500 | CODE | VEC_Reserved_6500
-GLOBAL | ${textprefix}00006600 | CODE | VEC_Reserved_6600
-GLOBAL | ${textprefix}00006700 | CODE | VEC_Reserved_6700
-GLOBAL | ${textprefix}00006800 | CODE | VEC_Reserved_6800
-GLOBAL | ${textprefix}00006900 | CODE | VEC_IA-32_Exeception_6900
-GLOBAL | ${textprefix}00006a00 | CODE | VEC_IA-32_Intercept_6a00
-GLOBAL | ${textprefix}00006b00 | CODE | VEC_IA-32_Interrupt_6b00
-GLOBAL | ${textprefix}00006c00 | CODE | VEC_Reserved_6c00
-GLOBAL | ${textprefix}00006d00 | CODE | VEC_Reserved_6d00
-GLOBAL | ${textprefix}00006e00 | CODE | VEC_Reserved_6e00
-GLOBAL | ${textprefix}00006f00 | CODE | VEC_Reserved_6f00
-GLOBAL | ${textprefix}00007000 | CODE | VEC_Reserved_7000
-GLOBAL | ${textprefix}00007100 | CODE | VEC_Reserved_7100
-GLOBAL | ${textprefix}00007200 | CODE | VEC_Reserved_7200
-GLOBAL | ${textprefix}00007300 | CODE | VEC_Reserved_7300
-GLOBAL | ${textprefix}00007400 | CODE | VEC_Reserved_7400
-GLOBAL | ${textprefix}00007500 | CODE | VEC_Reserved_7500
-GLOBAL | ${textprefix}00007600 | CODE | VEC_Reserved_7600
-GLOBAL | ${textprefix}00007700 | CODE | VEC_Reserved_7700
-GLOBAL | ${textprefix}00007800 | CODE | VEC_Reserved_7800
-GLOBAL | ${textprefix}00007900 | CODE | VEC_Reserved_7900
-GLOBAL | ${textprefix}00007a00 | CODE | VEC_Reserved_7a00
-GLOBAL | ${textprefix}00007b00 | CODE | VEC_Reserved_7b00
-GLOBAL | ${textprefix}00007c00 | CODE | VEC_Reserved_7c00
-GLOBAL | ${textprefix}00007d00 | CODE | VEC_Reserved_7d00
-GLOBAL | ${textprefix}00007e00 | CODE | VEC_Reserved_7e00
-GLOBAL | ${textprefix}00007f00 | CODE | VEC_Reserved_7f00
-END
-
-
-awk '
-/ _start$/ {start=1}
-/ start_ap$/ {start=1}
-/__start_gate_section/ {start=1}
-/^'${dataprefix}\|${textprefix}'/ {
-	if ($4 == ".kdb")
-		next
-	if (start && substr($NF,1,1) != "0") {
-		type = substr($0,26,5)
-		if (type == ".text")
-			printf "GLOBAL | %s | CODE | %s\n", $1, $NF
-		else {
-			n = 0
-			s = $(NF-1)
-			while (length(s) > 0) {
-				n = n*16 + (index("0123456789abcdef", substr(s,1,1)) - 1)
-				s = substr(s,2)
-			}
-			printf "GLOBAL | %s | DATA | %s | %d\n", $1, $NF, n
-		}
-	}
-	if($NF == "_end") 
-		exit
-
-}
-' $TMPSYM ) | egrep -v " __device| __vendor" | awk -v sn1="$SN1" '
-/GLOBAL/ {
-	print $0
-	if (sn1 != 0) {
-		/* 32 bits of sn1 physical addrs, */
-		print substr($0,1,9) "04" substr($0,20,16) "Phy_" substr($0,36)
-	} else {
-		/* 38 bits of sn2 physical addrs, need addr space bits */
-		print substr($0,1,9) "3004" substr($0,20,16) "Phy_" substr($0,36)
-	}
-
-} ' | sort -k3
-
-N=`wc -l $TEXTSYM|awk '{print $1}'`
-echo "Generated TEXTSYM file" >&2
-echo "  $LINUX --> $TEXTSYM" >&2
-echo "  Found $N symbols" >&2

arch/ia64/sn/io/machvec/pci_bus_cvlink.c

@@ -290,6 +290,7 @@ sn_pci_fixup_slot(struct pci_dev *dev)
 			addr |= __IA64_UNCACHED_OFFSET;
 			dev->resource[idx].start = addr;
 			dev->resource[idx].end = addr + size;
+			dev->resource[idx].parent = &ioport_resource;
 		}
 
 		if (dev->resource[idx].flags & IORESOURCE_IO)
@@ -322,6 +323,7 @@ sn_pci_fixup_slot(struct pci_dev *dev)
 			addr |= __IA64_UNCACHED_OFFSET;
 			dev->resource[idx].start = addr;
 			dev->resource[idx].end = addr + size;
+			dev->resource[idx].parent = &iomem_resource;
 		}
 
 		if (dev->resource[idx].flags & IORESOURCE_MEM)
@@ -351,6 +353,7 @@ sn_pci_fixup_slot(struct pci_dev *dev)
                         addr |= __IA64_UNCACHED_OFFSET;
                         dev->resource[PCI_ROM_RESOURCE].start = addr;
                         dev->resource[PCI_ROM_RESOURCE].end = addr + size;
+			dev->resource[idx].parent = &iomem_resource;
                         if (dev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_MEM)
                                 cmd |= PCI_COMMAND_MEMORY;
                 }

arch/ia64/sn/kernel/bte.c

@@ -421,9 +421,10 @@ bte_init_node(nodepda_t * mynodepda, cnodeid_t cnode)
 	mynodepda->bte_recovery_timer.data = (unsigned long) mynodepda;
 
 	for (i = 0; i < BTES_PER_NODE; i++) {
-		(u64) mynodepda->bte_if[i].bte_base_addr =
-		    REMOTE_HUB_ADDR(cnodeid_to_nasid(cnode),
-			(i == 0 ? IIO_IBLS0 : IIO_IBLS1));
+		/* Which link status register should we use? */
+		unsigned long link_status = (i == 0 ? IIO_IBLS0 : IIO_IBLS1);
+		mynodepda->bte_if[i].bte_base_addr = (u64 *)
+			REMOTE_HUB_ADDR(cnodeid_to_nasid(cnode), link_status);
 
 		/*
 		 * Initialize the notification and spinlock

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) 2000-2003 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
  */
 #include <linux/config.h>
 #include <asm/uaccess.h>
@@ -118,11 +118,33 @@ register_sn_force_interrupt(void) {
 	}
 }
 
+static int coherence_id_read_proc(char *page, char **start, off_t off,
+		int count, int *eof, void *data) {
+	return sprintf(page, "%d\n", cpuid_to_coherence_id(smp_processor_id()));
+}
+
+void
+register_sn_coherence_id(void) {
+	struct proc_dir_entry *entry;
+
+	if (!sgi_proc_dir) {
+		sgi_proc_dir = proc_mkdir("sgi_sn", 0);
+	}
+	entry = create_proc_entry("coherence_id", 0444, sgi_proc_dir);
+	if (entry) {
+		entry->nlink = 1;
+		entry->data = 0;
+		entry->read_proc = coherence_id_read_proc;
+		entry->write_proc = NULL;
+	}
+}
+
 void
 register_sn_procfs(void) {
 	register_sn_partition_id();
 	register_sn_serial_numbers();
 	register_sn_force_interrupt();
+	register_sn_coherence_id();
 }
 
 #endif /* CONFIG_PROC_FS */

include/asm-ia64/acpi.h

@@ -96,7 +96,6 @@ static inline void disable_acpi(void) { }
 
 const char *acpi_get_sysname (void);
 int acpi_request_vector (u32 int_type);
-int acpi_register_irq (u32 gsi, u32 polarity, u32 trigger);
 int acpi_gsi_to_irq (u32 gsi, unsigned int *irq);
 
 #ifdef CONFIG_ACPI_NUMA

include/asm-ia64/sal.h

@@ -364,7 +364,7 @@ typedef struct sal_processor_static_info {
 	u64 cr[128];
 	u64 ar[128];
 	u64 rr[8];
-	struct ia64_fpreg fr[128];
+	struct ia64_fpreg __attribute__ ((packed)) fr[128];
 } sal_processor_static_info_t;
 
 struct sal_cpuid_info {

include/asm-ia64/sn/sn_cpuid.h

@@ -4,7 +4,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2000-2003 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2000-2004 Silicon Graphics, Inc. All rights reserved.
  */
 
 
@@ -93,6 +93,7 @@
  */
 #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)	(cpu_physical_id_to_nasid(cpi) >> 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)
@@ -172,6 +173,11 @@ extern short physical_node_map[];			/* indexed by nasid to get cnode */
 
 #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))
 
 #endif /* _ASM_IA64_SN_SN_CPUID_H */