24 files changed, 7149 insertions, 0 deletions
diff --git a/include/asm-ia64/sn/addrs.h b/include/asm-ia64/sn/addrs.h
new file mode 100644
index 00000000000..c916bd22767
--- /dev/null
+++ b/include/asm-ia64/sn/addrs.h
@@ -0,0 +1,238 @@
+/*
+ * 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) 1992-1999,2001-2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+#ifndef _ASM_IA64_SN_ADDRS_H
+#define _ASM_IA64_SN_ADDRS_H
+
+#include <asm/percpu.h>
+#include <asm/sn/types.h>
+#include <asm/sn/arch.h>
+#include <asm/sn/pda.h>
+
+/*
+ *  Memory/SHUB Address Format:
+ *  +-+---------+--+--------------+
+ *  |0|  NASID  |AS| NodeOffset   |
+ *  +-+---------+--+--------------+
+ *
+ *  NASID: (low NASID bit is 0) Memory and SHUB MMRs
+ *   AS: 2-bit Address Space Identifier. Used only if low NASID bit is 0
+ *     00: Local Resources and MMR space
+ *           Top bit of NodeOffset
+ *               0: Local resources space
+ *                  node id:
+ *                        0: IA64/NT compatibility space
+ *                        2: Local MMR Space
+ *                        4: Local memory, regardless of local node id
+ *               1: Global MMR space
+ *     01: GET space.
+ *     10: AMO space.
+ *     11: Cacheable memory space.
+ *
+ *   NodeOffset: byte offset
+ *
+ *
+ *  TIO address format:
+ *  +-+----------+--+--------------+
+ *  |0|  NASID   |AS| Nodeoffset   |
+ *  +-+----------+--+--------------+
+ *
+ *  NASID: (low NASID bit is 1) TIO
+ *   AS: 2-bit Chiplet Identifier
+ *     00: TIO LB (Indicates TIO MMR access.)
+ *     01: TIO ICE (indicates coretalk space access.)
+ * 
+ *   NodeOffset: top bit must be set.
+ *
+ *
+ * Note that in both of the above address formats, the low
+ * NASID bit indicates if the reference is to the SHUB or TIO MMRs.
+ */
+
+
+/*
+ * Define basic shift & mask constants for manipulating NASIDs and AS values.
+ */
+#define NASID_BITMASK		(sn_hub_info->nasid_bitmask)
+#define NASID_SHIFT		(sn_hub_info->nasid_shift)
+#define AS_SHIFT		(sn_hub_info->as_shift)
+#define AS_BITMASK		0x3UL
+
+#define NASID_MASK              ((u64)NASID_BITMASK << NASID_SHIFT)
+#define AS_MASK			((u64)AS_BITMASK << AS_SHIFT)
+#define REGION_BITS		0xe000000000000000UL
+
+
+/*
+ * AS values. These are the same on both SHUB1 & SHUB2.
+ */
+#define AS_GET_VAL		1UL
+#define AS_AMO_VAL		2UL
+#define AS_CAC_VAL		3UL
+#define AS_GET_SPACE		(AS_GET_VAL << AS_SHIFT)
+#define AS_AMO_SPACE		(AS_AMO_VAL << AS_SHIFT)
+#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 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)
+
+/*
+ * Physical mode addresses
+ */
+#define GLOBAL_PHYS_MMR_SPACE	(UNCACHED_PHYS | GLOBAL_MMR_OFFSET)
+
+
+/*
+ * Clear region & AS bits.
+ */
+#define TO_PHYS_MASK		(~(REGION_BITS | AS_MASK))
+
+
+/*
+ * Misc NASID manipulation.
+ */
+#define NASID_SPACE(n)		((u64)(n) << NASID_SHIFT)
+#define REMOTE_ADDR(n,a)	(NASID_SPACE(n) | (a))
+#define NODE_OFFSET(x)		((x) & (NODE_ADDRSPACE_SIZE - 1))
+#define NODE_ADDRSPACE_SIZE     (1UL << AS_SHIFT)
+#define NASID_GET(x)		(int) (((u64) (x) >> NASID_SHIFT) & NASID_BITMASK)
+#define LOCAL_MMR_ADDR(a)	(LOCAL_MMR_SPACE | (a))
+#define GLOBAL_MMR_ADDR(n,a)	(GLOBAL_MMR_SPACE | REMOTE_ADDR(n,a))
+#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)))
+
+
+/* non-II mmr's start at top of big window space (4G) */
+#define BWIN_TOP		0x0000000100000000UL
+
+/*
+ * general address defines
+ */
+#define CAC_BASE		(CACHED   | AS_CAC_SPACE)
+#define AMO_BASE		(UNCACHED | AS_AMO_SPACE)
+#define GET_BASE		(CACHED   | AS_GET_SPACE)
+
+/*
+ * Convert Memory addresses between various addressing modes.
+ */
+#define TO_PHYS(x)		(TO_PHYS_MASK & (x))
+#define TO_CAC(x)		(CAC_BASE     | TO_PHYS(x))
+#define TO_AMO(x)		(AMO_BASE     | TO_PHYS(x))
+#define TO_GET(x)		(GET_BASE     | TO_PHYS(x))
+
+
+/*
+ * Covert from processor physical address to II/TIO physical address:
+ *	II - squeeze out the AS bits
+ *	TIO- requires a chiplet id in bits 38-39.  For DMA to memory,
+ *           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.
+ */
+#define PHYS_TO_TIODMA(x)	( (((u64)(x) & NASID_MASK) << 2) | NODE_OFFSET(x))
+#define PHYS_TO_DMA(x)          ( (((u64)(x) & NASID_MASK) >> 2) | NODE_OFFSET(x))
+
+
+/*
+ * The following definitions pertain to the IO special address
+ * space.  They define the location of the big and little windows
+ * of any given node.
+ */
+#define BWIN_SIZE_BITS			29	/* big window size: 512M */
+#define TIO_BWIN_SIZE_BITS		30	/* big window size: 1G */
+#define NODE_SWIN_BASE(n, w)		((w == 0) ? NODE_BWIN_BASE((n), SWIN0_BIGWIN) \
+		: RAW_NODE_SWIN_BASE(n, w))
+#define NODE_IO_BASE(n)			(GLOBAL_MMR_SPACE | 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 TIO_BWIN_WINDOW_SELECT_MASK	0x7
+#define TIO_BWIN_WINDOWNUM(x)		(((x) >> TIO_BWIN_SIZE_BITS) & TIO_BWIN_WINDOW_SELECT_MASK)
+
+
+
+/*
+ * The following definitions pertain to the IO special address
+ * space.  They define the location of the big and little windows
+ * of any given node.
+ */
+
+#define SWIN_SIZE_BITS			24
+#define	SWIN_WIDGET_MASK		0xF
+
+#define TIO_SWIN_SIZE_BITS		28
+#define TIO_SWIN_SIZE			(1UL << TIO_SWIN_SIZE_BITS)
+#define TIO_SWIN_WIDGET_MASK		0x3
+
+/*
+ * Convert smallwindow address to xtalk address.
+ *
+ * 'addr' can be physical or virtual address, but will be converted
+ * to Xtalk address in the range 0 -> SWINZ_SIZEMASK
+ */
+#define	SWIN_WIDGETNUM(x)		(((x)  >> SWIN_SIZE_BITS) & SWIN_WIDGET_MASK)
+#define TIO_SWIN_WIDGETNUM(x)		(((x)  >> TIO_SWIN_SIZE_BITS) & TIO_SWIN_WIDGET_MASK)
+
+
+/*
+ * The following macros produce the correct base virtual address for
+ * the hub registers. The REMOTE_HUB_* macro produce
+ * the address for the specified hub's registers.  The intent is
+ * that the appropriate PI, MD, NI, or II register would be substituted
+ * for x.
+ *
+ *   WARNING:
+ *	When certain Hub chip workaround are defined, it's not sufficient
+ *	to dereference the *_HUB_ADDR() macros.  You should instead use
+ *	HUB_L() and HUB_S() if you must deal with pointers to hub registers.
+ *	Otherwise, the recommended approach is to use *_HUB_L() and *_HUB_S().
+ *	They're always safe.
+ */
+#define REMOTE_HUB_ADDR(n,x)						\
+	((n & 1) ?							\
+	/* TIO: */							\
+	((volatile u64 *)(GLOBAL_MMR_ADDR(n,x)))			\
+	: /* SHUB: */							\
+	(((x) & BWIN_TOP) ? ((volatile u64 *)(GLOBAL_MMR_ADDR(n,x)))\
+	: ((volatile u64 *)(NODE_SWIN_BASE(n,1) + 0x800000 + (x)))))
+
+
+
+#define HUB_L(x)			(*((volatile typeof(*x) *)x))
+#define	HUB_S(x,d)			(*((volatile typeof(*x) *)x) = (d))
+
+#define REMOTE_HUB_L(n, a)		HUB_L(REMOTE_HUB_ADDR((n), (a)))
+#define REMOTE_HUB_S(n, a, d)		HUB_S(REMOTE_HUB_ADDR((n), (a)), (d))
+
+
+#endif /* _ASM_IA64_SN_ADDRS_H */
diff --git a/include/asm-ia64/sn/arch.h b/include/asm-ia64/sn/arch.h
new file mode 100644
index 00000000000..7c349f07916
--- /dev/null
+++ b/include/asm-ia64/sn/arch.h
@@ -0,0 +1,52 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * SGI specific setup.
+ *
+ * Copyright (C) 1995-1997,1999,2001-2004 Silicon Graphics, Inc.  All rights reserved.
+ * Copyright (C) 1999 Ralf Baechle (ralf@gnu.org)
+ */
+#ifndef _ASM_IA64_SN_ARCH_H
+#define _ASM_IA64_SN_ARCH_H
+
+#include <asm/types.h>
+#include <asm/percpu.h>
+#include <asm/sn/types.h>
+#include <asm/sn/sn_cpuid.h>
+
+/*
+ * The following defines attributes of the HUB chip. These attributes are
+ * frequently referenced. They are kept in the per-cpu data areas of each cpu.
+ * They are kept together in a struct to minimize cache misses.
+ */
+struct sn_hub_info_s {
+	u8 shub2;
+	u8 nasid_shift;
+	u8 as_shift;
+	u8 shub_1_1_found;
+	u16 nasid_bitmask;
+};
+DECLARE_PER_CPU(struct sn_hub_info_s, __sn_hub_info);
+#define sn_hub_info 	(&__get_cpu_var(__sn_hub_info))
+#define is_shub2()	(sn_hub_info->shub2)
+#define is_shub1()	(sn_hub_info->shub2 == 0)
+
+/*
+ * Use this macro to test if shub 1.1 wars should be enabled
+ */
+#define enable_shub_wars_1_1()	(sn_hub_info->shub_1_1_found)
+
+
+/*
+ * This is the maximum number of nodes that can be part of a kernel.
+ * Effectively, it's the maximum number of compact node ids (cnodeid_t).
+ * This is not necessarily the same as MAX_NASIDS.
+ */
+#define MAX_COMPACT_NODES	2048
+#define CPUS_PER_NODE		4
+
+extern void sn_flush_all_caches(long addr, long bytes);
+
+#endif /* _ASM_IA64_SN_ARCH_H */
diff --git a/include/asm-ia64/sn/bte.h b/include/asm-ia64/sn/bte.h
new file mode 100644
index 00000000000..0ec27f99c18
--- /dev/null
+++ b/include/asm-ia64/sn/bte.h
@@ -0,0 +1,148 @@
+/*
+ * 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) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+
+#ifndef _ASM_IA64_SN_BTE_H
+#define _ASM_IA64_SN_BTE_H
+
+#include <linux/timer.h>
+#include <linux/spinlock.h>
+#include <linux/cache.h>
+#include <asm/sn/types.h>
+
+
+/* #define BTE_DEBUG */
+/* #define BTE_DEBUG_VERBOSE */
+
+#ifdef BTE_DEBUG
+#  define BTE_PRINTK(x) printk x	/* Terse */
+#  ifdef BTE_DEBUG_VERBOSE
+#    define BTE_PRINTKV(x) printk x	/* Verbose */
+#  else
+#    define BTE_PRINTKV(x)
+#  endif /* BTE_DEBUG_VERBOSE */
+#else
+#  define BTE_PRINTK(x)
+#  define BTE_PRINTKV(x)
+#endif	/* BTE_DEBUG */
+
+
+/* BTE status register only supports 16 bits for length field */
+#define BTE_LEN_BITS (16)
+#define BTE_LEN_MASK ((1 << BTE_LEN_BITS) - 1)
+#define BTE_MAX_XFER ((1 << BTE_LEN_BITS) * L1_CACHE_BYTES)
+
+
+/* Define hardware */
+#define BTES_PER_NODE 2
+
+
+/* Define hardware modes */
+#define BTE_NOTIFY (IBCT_NOTIFY)
+#define BTE_NORMAL BTE_NOTIFY
+#define BTE_ZERO_FILL (BTE_NOTIFY | IBCT_ZFIL_MODE)
+/* Use a reserved bit to let the caller specify a wait for any BTE */
+#define BTE_WACQUIRE (0x4000)
+/* Use the BTE on the node with the destination memory */
+#define BTE_USE_DEST (BTE_WACQUIRE << 1)
+/* Use any available BTE interface on any node for the transfer */
+#define BTE_USE_ANY (BTE_USE_DEST << 1)
+/* macro to force the IBCT0 value valid */
+#define BTE_VALID_MODE(x) ((x) & (IBCT_NOTIFY | IBCT_ZFIL_MODE))
+
+#define BTE_ACTIVE		(IBLS_BUSY | IBLS_ERROR)
+#define BTE_WORD_AVAILABLE	(IBLS_BUSY << 1)
+#define BTE_WORD_BUSY		(~BTE_WORD_AVAILABLE)
+
+/*
+ * Some macros to simplify reading.
+ * Start with macros to locate the BTE control registers.
+ */
+#define BTE_LNSTAT_LOAD(_bte)						\
+			HUB_L(_bte->bte_base_addr)
+#define BTE_LNSTAT_STORE(_bte, _x)					\
+			HUB_S(_bte->bte_base_addr, (_x))
+#define BTE_SRC_STORE(_bte, _x)						\
+			HUB_S(_bte->bte_base_addr + (BTEOFF_SRC/8), (_x))
+#define BTE_DEST_STORE(_bte, _x)					\
+			HUB_S(_bte->bte_base_addr + (BTEOFF_DEST/8), (_x))
+#define BTE_CTRL_STORE(_bte, _x)					\
+			HUB_S(_bte->bte_base_addr + (BTEOFF_CTRL/8), (_x))
+#define BTE_NOTIF_STORE(_bte, _x)					\
+			HUB_S(_bte->bte_base_addr + (BTEOFF_NOTIFY/8), (_x))
+
+
+/* Possible results from bte_copy and bte_unaligned_copy */
+/* The following error codes map into the BTE hardware codes
+ * IIO_ICRB_ECODE_* (in shubio.h). The hardware uses
+ * an error code of 0 (IIO_ICRB_ECODE_DERR), but we want zero
+ * to mean BTE_SUCCESS, so add one (BTEFAIL_OFFSET) to the error
+ * codes to give the following error codes.
+ */
+#define BTEFAIL_OFFSET	1
+
+typedef enum {
+	BTE_SUCCESS,		/* 0 is success */
+	BTEFAIL_DIR,		/* Directory error due to IIO access*/
+	BTEFAIL_POISON,		/* poison error on IO access (write to poison page) */
+	BTEFAIL_WERR,		/* Write error (ie WINV to a Read only line) */
+	BTEFAIL_ACCESS,		/* access error (protection violation) */
+	BTEFAIL_PWERR,		/* Partial Write Error */
+	BTEFAIL_PRERR,		/* Partial Read Error */
+	BTEFAIL_TOUT,		/* CRB Time out */
+	BTEFAIL_XTERR,		/* Incoming xtalk pkt had error bit */
+	BTEFAIL_NOTAVAIL,	/* BTE not available */
+} bte_result_t;
+
+
+/*
+ * Structure defining a bte.  An instance of this
+ * structure is created in the nodepda for each
+ * bte on that node (as defined by BTES_PER_NODE)
+ * This structure contains everything necessary
+ * to work with a BTE.
+ */
+struct bteinfo_s {
+	volatile u64 notify ____cacheline_aligned;
+	u64 *bte_base_addr ____cacheline_aligned;
+	spinlock_t spinlock;
+	cnodeid_t bte_cnode;	/* cnode                            */
+	int bte_error_count;	/* Number of errors encountered     */
+	int bte_num;		/* 0 --> BTE0, 1 --> BTE1           */
+	int cleanup_active;	/* Interface is locked for cleanup  */
+	volatile bte_result_t bh_error;	/* error while processing   */
+	volatile u64 *most_rcnt_na;
+};
+
+
+/*
+ * Function prototypes (functions defined in bte.c, used elsewhere)
+ */
+extern bte_result_t bte_copy(u64, u64, u64, u64, void *);
+extern bte_result_t bte_unaligned_copy(u64, u64, u64, u64);
+extern void bte_error_handler(unsigned long);
+
+#define bte_zero(dest, len, mode, notification) \
+	bte_copy(0, dest, len, ((mode) | BTE_ZERO_FILL), notification)
+
+/*
+ * The following is the prefered way of calling bte_unaligned_copy
+ * If the copy is fully cache line aligned, then bte_copy is
+ * used instead.  Since bte_copy is inlined, this saves a call
+ * stack.  NOTE: bte_copy is called synchronously and does block
+ * until the transfer is complete.  In order to get the asynch
+ * version of bte_copy, you must perform this check yourself.
+ */
+#define BTE_UNALIGNED_COPY(src, dest, len, mode)                        \
+	(((len & L1_CACHE_MASK) || (src & L1_CACHE_MASK) ||             \
+	  (dest & L1_CACHE_MASK)) ?                                     \
+	 bte_unaligned_copy(src, dest, len, mode) :              	\
+	 bte_copy(src, dest, len, mode, NULL))
+
+
+#endif	/* _ASM_IA64_SN_BTE_H */
diff --git a/include/asm-ia64/sn/clksupport.h b/include/asm-ia64/sn/clksupport.h
new file mode 100644
index 00000000000..d340c365a82
--- /dev/null
+++ b/include/asm-ia64/sn/clksupport.h
@@ -0,0 +1,28 @@
+/*
+ * 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) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+/*
+ * This file contains definitions for accessing a platform supported high resolution
+ * clock. The clock is monitonically increasing and can be accessed from any node
+ * in the system. The clock is synchronized across nodes - all nodes see the
+ * same value.
+ * 
+ *	RTC_COUNTER_ADDR - contains the address of the counter 
+ *
+ */
+
+#ifndef _ASM_IA64_SN_CLKSUPPORT_H
+#define _ASM_IA64_SN_CLKSUPPORT_H
+
+extern unsigned long sn_rtc_cycles_per_second;
+
+#define RTC_COUNTER_ADDR	((long *)LOCAL_MMR_ADDR(SH_RTC))
+
+#define rtc_time()		(*RTC_COUNTER_ADDR)
+
+#endif /* _ASM_IA64_SN_CLKSUPPORT_H */
diff --git a/include/asm-ia64/sn/fetchop.h b/include/asm-ia64/sn/fetchop.h
new file mode 100644
index 00000000000..5f4ad8f4b5d
--- /dev/null
+++ b/include/asm-ia64/sn/fetchop.h
@@ -0,0 +1,85 @@
+/*
+ *
+ * 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) 2001-2004 Silicon Graphics, Inc.  All rights reserved.
+ */
+
+#ifndef _ASM_IA64_SN_FETCHOP_H
+#define _ASM_IA64_SN_FETCHOP_H
+
+#include <linux/config.h>
+
+#define FETCHOP_BASENAME	"sgi_fetchop"
+#define FETCHOP_FULLNAME	"/dev/sgi_fetchop"
+
+
+
+#define FETCHOP_VAR_SIZE 64 /* 64 byte per fetchop variable */
+
+#define FETCHOP_LOAD		0
+#define FETCHOP_INCREMENT	8
+#define FETCHOP_DECREMENT	16
+#define FETCHOP_CLEAR		24
+
+#define FETCHOP_STORE		0
+#define FETCHOP_AND		24
+#define FETCHOP_OR		32
+
+#define FETCHOP_CLEAR_CACHE	56
+
+#define FETCHOP_LOAD_OP(addr, op) ( \
+         *(volatile long *)((char*) (addr) + (op)))
+
+#define FETCHOP_STORE_OP(addr, op, x) ( \
+         *(volatile long *)((char*) (addr) + (op)) = (long) (x))
+
+#ifdef __KERNEL__
+
+/*
+ * Convert a region 6 (kaddr) address to the address of the fetchop variable
+ */
+#define FETCHOP_KADDR_TO_MSPEC_ADDR(kaddr)	TO_MSPEC(kaddr)
+
+
+/*
+ * Each Atomic Memory Operation (AMO formerly known as fetchop)
+ * variable is 64 bytes long.  The first 8 bytes are used.  The
+ * remaining 56 bytes are unaddressable due to the operation taking
+ * that portion of the address.
+ * 
+ * NOTE: The AMO_t _MUST_ be placed in either the first or second half
+ * of the cache line.  The cache line _MUST NOT_ be used for anything
+ * other than additional AMO_t entries.  This is because there are two
+ * addresses which reference the same physical cache line.  One will
+ * be a cached entry with the memory type bits all set.  This address
+ * may be loaded into processor cache.  The AMO_t will be referenced
+ * uncached via the memory special memory type.  If any portion of the
+ * cached cache-line is modified, when that line is flushed, it will
+ * overwrite the uncached value in physical memory and lead to
+ * inconsistency.
+ */
+typedef struct {
+        u64 variable;
+        u64 unused[7];
+} AMO_t;
+
+
+/*
+ * The following APIs are externalized to the kernel to allocate/free pages of
+ * fetchop variables.
+ *	fetchop_kalloc_page	- Allocate/initialize 1 fetchop page on the
+ *				  specified cnode. 
+ *	fetchop_kfree_page	- Free a previously allocated fetchop page
+ */
+
+unsigned long fetchop_kalloc_page(int nid);
+void fetchop_kfree_page(unsigned long maddr);
+
+
+#endif /* __KERNEL__ */
+
+#endif /* _ASM_IA64_SN_FETCHOP_H */
+
diff --git a/include/asm-ia64/sn/geo.h b/include/asm-ia64/sn/geo.h
new file mode 100644
index 00000000000..f566343d25f
--- /dev/null
+++ b/include/asm-ia64/sn/geo.h
@@ -0,0 +1,124 @@
+/*
+ * 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) 1992 - 1997, 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+#ifndef _ASM_IA64_SN_GEO_H
+#define _ASM_IA64_SN_GEO_H
+
+/* The geoid_t implementation below is based loosely on the pcfg_t
+   implementation in sys/SN/promcfg.h. */
+
+/* Type declaractions */
+
+/* Size of a geoid_t structure (must be before decl. of geoid_u) */
+#define GEOID_SIZE	8	/* Would 16 be better?  The size can
+				   be different on different platforms. */
+
+#define MAX_SLABS	0xe	/* slabs per module */
+
+typedef unsigned char	geo_type_t;
+
+/* Fields common to all substructures */
+typedef struct geo_any_s {
+    moduleid_t	module;		/* The module (box) this h/w lives in */
+    geo_type_t	type;		/* What type of h/w is named by this geoid_t */
+    slabid_t	slab;		/* The logical assembly within the module */
+} geo_any_t;
+
+/* Additional fields for particular types of hardware */
+typedef struct geo_node_s {
+    geo_any_t	any;		/* No additional fields needed */
+} geo_node_t;
+
+typedef struct geo_rtr_s {
+    geo_any_t	any;		/* No additional fields needed */
+} geo_rtr_t;
+
+typedef struct geo_iocntl_s {
+    geo_any_t	any;		/* No additional fields needed */
+} geo_iocntl_t;
+
+typedef struct geo_pcicard_s {
+    geo_iocntl_t	any;
+    char		bus;	/* Bus/widget number */
+    char		slot;	/* PCI slot number */
+} geo_pcicard_t;
+
+/* Subcomponents of a node */
+typedef struct geo_cpu_s {
+    geo_node_t	node;
+    char	slice;		/* Which CPU on the node */
+} geo_cpu_t;
+
+typedef struct geo_mem_s {
+    geo_node_t	node;
+    char	membus;		/* The memory bus on the node */
+    char	memslot;	/* The memory slot on the bus */
+} geo_mem_t;
+
+
+typedef union geoid_u {
+    geo_any_t	any;
+    geo_node_t	node;
+    geo_iocntl_t	iocntl;
+    geo_pcicard_t	pcicard;
+    geo_rtr_t	rtr;
+    geo_cpu_t	cpu;
+    geo_mem_t	mem;
+    char	padsize[GEOID_SIZE];
+} geoid_t;
+
+
+/* Preprocessor macros */
+
+#define GEO_MAX_LEN	48	/* max. formatted length, plus some pad:
+				   module/001c07/slab/5/node/memory/2/slot/4 */
+
+/* Values for geo_type_t */
+#define GEO_TYPE_INVALID	0
+#define GEO_TYPE_MODULE		1
+#define GEO_TYPE_NODE		2
+#define GEO_TYPE_RTR		3
+#define GEO_TYPE_IOCNTL		4
+#define GEO_TYPE_IOCARD		5
+#define GEO_TYPE_CPU		6
+#define GEO_TYPE_MEM		7
+#define GEO_TYPE_MAX		(GEO_TYPE_MEM+1)
+
+/* Parameter for hwcfg_format_geoid_compt() */
+#define GEO_COMPT_MODULE	1
+#define GEO_COMPT_SLAB		2
+#define GEO_COMPT_IOBUS		3
+#define GEO_COMPT_IOSLOT	4
+#define GEO_COMPT_CPU		5
+#define GEO_COMPT_MEMBUS	6
+#define GEO_COMPT_MEMSLOT	7
+
+#define GEO_INVALID_STR		"<invalid>"
+
+#define INVALID_NASID           ((nasid_t)-1)
+#define INVALID_CNODEID         ((cnodeid_t)-1)
+#define INVALID_PNODEID         ((pnodeid_t)-1)
+#define INVALID_SLAB            (slabid_t)-1
+#define INVALID_MODULE          ((moduleid_t)-1)
+#define INVALID_PARTID          ((partid_t)-1)
+
+static inline slabid_t geo_slab(geoid_t g)
+{
+	return (g.any.type == GEO_TYPE_INVALID) ?
+		INVALID_SLAB : g.any.slab;
+}
+
+static inline moduleid_t geo_module(geoid_t g)
+{
+	return (g.any.type == GEO_TYPE_INVALID) ?
+		INVALID_MODULE : g.any.module;
+}
+
+extern geoid_t cnodeid_get_geoid(cnodeid_t cnode);
+
+#endif /* _ASM_IA64_SN_GEO_H */
diff --git a/include/asm-ia64/sn/intr.h b/include/asm-ia64/sn/intr.h
new file mode 100644
index 00000000000..e51471fb086
--- /dev/null
+++ b/include/asm-ia64/sn/intr.h
@@ -0,0 +1,56 @@
+/*
+ * 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) 1992 - 1997, 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+#ifndef _ASM_IA64_SN_INTR_H
+#define _ASM_IA64_SN_INTR_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_ACPI_SCI_INT                (0x34)
+#define SGI_TIOCA_ERROR                 (0x35)
+#define SGI_TIO_ERROR                   (0x36)
+#define SGI_TIOCX_ERROR                 (0x37)
+#define SGI_MMTIMER_VECTOR              (0x38)
+#define SGI_XPC_NOTIFY                  (0xe7)
+
+#define IA64_SN2_FIRST_DEVICE_VECTOR    (0x3c)
+#define IA64_SN2_LAST_DEVICE_VECTOR     (0xe6)
+
+#define SN2_IRQ_RESERVED        (0x1)
+#define SN2_IRQ_CONNECTED       (0x2)
+#define SN2_IRQ_SHARED          (0x4)
+
+// The SN PROM irq struct
+struct sn_irq_info {
+	struct sn_irq_info *irq_next;	/* sharing irq list	     */
+	short		irq_nasid;	/* Nasid IRQ is assigned to  */
+	int		irq_slice;	/* slice IRQ is assigned to  */
+	int		irq_cpuid;	/* kernel logical cpuid	     */
+	int		irq_irq;	/* the IRQ number */
+	int		irq_int_bit;	/* Bridge interrupt pin */
+	uint64_t	irq_xtalkaddr;	/* xtalkaddr IRQ is sent to  */
+	int		irq_bridge_type;/* pciio asic type (pciio.h) */
+	void	       *irq_bridge;	/* bridge generating irq     */
+	void	       *irq_pciioinfo;	/* associated pciio_info_t   */
+	int		irq_last_intr;	/* For Shub lb lost intr WAR */
+	int		irq_cookie;	/* unique cookie 	     */
+	int		irq_flags;	/* flags */
+	int		irq_share_cnt;	/* num devices sharing IRQ   */
+};
+
+extern void sn_send_IPI_phys(int, long, int, int);
+
+#define CPU_VECTOR_TO_IRQ(cpuid,vector) (vector)
+
+#endif /* _ASM_IA64_SN_INTR_H */
diff --git a/include/asm-ia64/sn/io.h b/include/asm-ia64/sn/io.h
new file mode 100644
index 00000000000..42209733f6b
--- /dev/null
+++ b/include/asm-ia64/sn/io.h
@@ -0,0 +1,265 @@
+/* 
+ * 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) 2000-2004 Silicon Graphics, Inc. All rights reserved.
+ */
+
+#ifndef _ASM_SN_IO_H
+#define _ASM_SN_IO_H
+#include <linux/compiler.h>
+#include <asm/intrinsics.h>
+
+extern void * sn_io_addr(unsigned long port) __attribute_const__; /* Forward definition */
+extern void __sn_mmiowb(void); /* Forward definition */
+
+extern int numionodes;
+
+#define __sn_mf_a()   ia64_mfa()
+
+extern void sn_dma_flush(unsigned long);
+
+#define __sn_inb ___sn_inb
+#define __sn_inw ___sn_inw
+#define __sn_inl ___sn_inl
+#define __sn_outb ___sn_outb
+#define __sn_outw ___sn_outw
+#define __sn_outl ___sn_outl
+#define __sn_readb ___sn_readb
+#define __sn_readw ___sn_readw
+#define __sn_readl ___sn_readl
+#define __sn_readq ___sn_readq
+#define __sn_readb_relaxed ___sn_readb_relaxed
+#define __sn_readw_relaxed ___sn_readw_relaxed
+#define __sn_readl_relaxed ___sn_readl_relaxed
+#define __sn_readq_relaxed ___sn_readq_relaxed
+
+/*
+ * The following routines are SN Platform specific, called when
+ * a reference is made to inX/outX set macros.  SN Platform
+ * inX set of macros ensures that Posted DMA writes on the
+ * Bridge is flushed.
+ *
+ * The routines should be self explainatory.
+ */
+
+static inline unsigned int
+___sn_inb (unsigned long port)
+{
+	volatile unsigned char *addr;
+	unsigned char ret = -1;
+
+	if ((addr = sn_io_addr(port))) {
+		ret = *addr;
+		__sn_mf_a();
+		sn_dma_flush((unsigned long)addr);
+	}
+	return ret;
+}
+
+static inline unsigned int
+___sn_inw (unsigned long port)
+{
+	volatile unsigned short *addr;
+	unsigned short ret = -1;
+
+	if ((addr = sn_io_addr(port))) {
+		ret = *addr;
+		__sn_mf_a();
+		sn_dma_flush((unsigned long)addr);
+	}
+	return ret;
+}
+
+static inline unsigned int
+___sn_inl (unsigned long port)
+{
+	volatile unsigned int *addr;
+	unsigned int ret = -1;
+
+	if ((addr = sn_io_addr(port))) {
+		ret = *addr;
+		__sn_mf_a();
+		sn_dma_flush((unsigned long)addr);
+	}
+	return ret;
+}
+
+static inline void
+___sn_outb (unsigned char val, unsigned long port)
+{
+	volatile unsigned char *addr;
+
+	if ((addr = sn_io_addr(port))) {
+		*addr = val;
+		__sn_mmiowb();
+	}
+}
+
+static inline void
+___sn_outw (unsigned short val, unsigned long port)
+{
+	volatile unsigned short *addr;
+
+	if ((addr = sn_io_addr(port))) {
+		*addr = val;
+		__sn_mmiowb();
+	}
+}
+
+static inline void
+___sn_outl (unsigned int val, unsigned long port)
+{
+	volatile unsigned int *addr;
+
+	if ((addr = sn_io_addr(port))) {
+		*addr = val;
+		__sn_mmiowb();
+	}
+}
+
+/*
+ * The following routines are SN Platform specific, called when 
+ * a reference is made to readX/writeX set macros.  SN Platform 
+ * readX set of macros ensures that Posted DMA writes on the 
+ * Bridge is flushed.
+ * 
+ * The routines should be self explainatory.
+ */
+
+static inline unsigned char
+___sn_readb (const volatile void __iomem *addr)
+{
+	unsigned char val;
+
+	val = *(volatile unsigned char __force *)addr;
+	__sn_mf_a();
+	sn_dma_flush((unsigned long)addr);
+        return val;
+}
+
+static inline unsigned short
+___sn_readw (const volatile void __iomem *addr)
+{
+	unsigned short val;
+
+	val = *(volatile unsigned short __force *)addr;
+	__sn_mf_a();
+	sn_dma_flush((unsigned long)addr);
+        return val;
+}
+
+static inline unsigned int
+___sn_readl (const volatile void __iomem *addr)
+{
+	unsigned int val;
+
+	val = *(volatile unsigned int __force *)addr;
+	__sn_mf_a();
+	sn_dma_flush((unsigned long)addr);
+        return val;
+}
+
+static inline unsigned long
+___sn_readq (const volatile vo