M7350v1_en_gpl

kernel/drivers/misc/sgi-xp/Makefile

@@ -0,0 +1,19 @@
+#
+# Makefile for SGI's XP devices.
+#
+
+obj-$(CONFIG_SGI_XP)		+= xp.o
+xp-y				:= xp_main.o
+xp-$(CONFIG_IA64_SGI_SN2)	+= xp_sn2.o xp_nofault.o
+xp-$(CONFIG_IA64_GENERIC)	+= xp_sn2.o xp_nofault.o
+xp-$(CONFIG_IA64_SGI_UV)	+= xp_uv.o
+xp-$(CONFIG_X86_64)		+= xp_uv.o
+
+obj-$(CONFIG_SGI_XP)		+= xpc.o
+xpc-y				:= xpc_main.o xpc_channel.o xpc_partition.o
+xpc-$(CONFIG_IA64_SGI_SN2)	+= xpc_sn2.o
+xpc-$(CONFIG_IA64_GENERIC)	+= xpc_sn2.o
+xpc-$(CONFIG_IA64_SGI_UV) 	+= xpc_uv.o
+xpc-$(CONFIG_X86_64)		+= xpc_uv.o
+
+obj-$(CONFIG_SGI_XP)		+= xpnet.o

kernel/drivers/misc/sgi-xp/xp.h

@@ -0,0 +1,358 @@
+/*
+ * 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) 2004-2008 Silicon Graphics, Inc. All rights reserved.
+ */
+
+/*
+ * External Cross Partition (XP) structures and defines.
+ */
+
+#ifndef _DRIVERS_MISC_SGIXP_XP_H
+#define _DRIVERS_MISC_SGIXP_XP_H
+
+#include <linux/mutex.h>
+
+#if defined CONFIG_X86_UV || defined CONFIG_IA64_SGI_UV
+#include <asm/uv/uv.h>
+#define is_uv()		is_uv_system()
+#endif
+
+#ifndef is_uv
+#define is_uv()		0
+#endif
+
+#if defined CONFIG_IA64
+#include <asm/sn/arch.h>	/* defines is_shub1() and is_shub2() */
+#define is_shub()	ia64_platform_is("sn2")
+#endif
+
+#ifndef is_shub1
+#define is_shub1()	0
+#endif
+
+#ifndef is_shub2
+#define is_shub2()	0
+#endif
+
+#ifndef is_shub
+#define is_shub()	0
+#endif
+
+#ifdef USE_DBUG_ON
+#define DBUG_ON(condition)	BUG_ON(condition)
+#else
+#define DBUG_ON(condition)
+#endif
+
+/*
+ * Define the maximum number of partitions the system can possibly support.
+ * It is based on the maximum number of hardware partitionable regions. The
+ * term 'region' in this context refers to the minimum number of nodes that
+ * can comprise an access protection grouping. The access protection is in
+ * regards to memory, IPI and IOI.
+ *
+ * The maximum number of hardware partitionable regions is equal to the
+ * maximum number of nodes in the entire system divided by the minimum number
+ * of nodes that comprise an access protection grouping.
+ */
+#define XP_MAX_NPARTITIONS_SN2	64
+#define XP_MAX_NPARTITIONS_UV	256
+
+/*
+ * XPC establishes channel connections between the local partition and any
+ * other partition that is currently up. Over these channels, kernel-level
+ * `users' can communicate with their counterparts on the other partitions.
+ *
+ * If the need for additional channels arises, one can simply increase
+ * XPC_MAX_NCHANNELS accordingly. If the day should come where that number
+ * exceeds the absolute MAXIMUM number of channels possible (eight), then one
+ * will need to make changes to the XPC code to accommodate for this.
+ *
+ * The absolute maximum number of channels possible is limited to eight for
+ * performance reasons on sn2 hardware. The internal cross partition structures
+ * require sixteen bytes per channel, and eight allows all of this
+ * interface-shared info to fit in one 128-byte cacheline.
+ */
+#define XPC_MEM_CHANNEL		0	/* memory channel number */
+#define	XPC_NET_CHANNEL		1	/* network channel number */
+
+#define XPC_MAX_NCHANNELS	2	/* max #of channels allowed */
+
+#if XPC_MAX_NCHANNELS > 8
+#error	XPC_MAX_NCHANNELS exceeds absolute MAXIMUM possible.
+#endif
+
+/*
+ * Define macro, XPC_MSG_SIZE(), is provided for the user
+ * that wants to fit as many msg entries as possible in a given memory size
+ * (e.g. a memory page).
+ */
+#define XPC_MSG_MAX_SIZE	128
+#define XPC_MSG_HDR_MAX_SIZE	16
+#define XPC_MSG_PAYLOAD_MAX_SIZE (XPC_MSG_MAX_SIZE - XPC_MSG_HDR_MAX_SIZE)
+
+#define XPC_MSG_SIZE(_payload_size) \
+				ALIGN(XPC_MSG_HDR_MAX_SIZE + (_payload_size), \
+				      is_uv() ? 64 : 128)
+
+
+/*
+ * Define the return values and values passed to user's callout functions.
+ * (It is important to add new value codes at the end just preceding
+ * xpUnknownReason, which must have the highest numerical value.)
+ */
+enum xp_retval {
+	xpSuccess = 0,
+
+	xpNotConnected,		/*  1: channel is not connected */
+	xpConnected,		/*  2: channel connected (opened) */
+	xpRETIRED1,		/*  3: (formerly xpDisconnected) */
+
+	xpMsgReceived,		/*  4: message received */
+	xpMsgDelivered,		/*  5: message delivered and acknowledged */
+
+	xpRETIRED2,		/*  6: (formerly xpTransferFailed) */
+
+	xpNoWait,		/*  7: operation would require wait */
+	xpRetry,		/*  8: retry operation */
+	xpTimeout,		/*  9: timeout in xpc_allocate_msg_wait() */
+	xpInterrupted,		/* 10: interrupted wait */
+
+	xpUnequalMsgSizes,	/* 11: message size disparity between sides */
+	xpInvalidAddress,	/* 12: invalid address */
+
+	xpNoMemory,		/* 13: no memory available for XPC structures */
+	xpLackOfResources,	/* 14: insufficient resources for operation */
+	xpUnregistered,		/* 15: channel is not registered */
+	xpAlreadyRegistered,	/* 16: channel is already registered */
+
+	xpPartitionDown,	/* 17: remote partition is down */
+	xpNotLoaded,		/* 18: XPC module is not loaded */
+	xpUnloading,		/* 19: this side is unloading XPC module */
+
+	xpBadMagic,		/* 20: XPC MAGIC string not found */
+
+	xpReactivating,		/* 21: remote partition was reactivated */
+
+	xpUnregistering,	/* 22: this side is unregistering channel */
+	xpOtherUnregistering,	/* 23: other side is unregistering channel */
+
+	xpCloneKThread,		/* 24: cloning kernel thread */
+	xpCloneKThreadFailed,	/* 25: cloning kernel thread failed */
+
+	xpNoHeartbeat,		/* 26: remote partition has no heartbeat */
+
+	xpPioReadError,		/* 27: PIO read error */
+	xpPhysAddrRegFailed,	/* 28: registration of phys addr range failed */
+
+	xpRETIRED3,		/* 29: (formerly xpBteDirectoryError) */
+	xpRETIRED4,		/* 30: (formerly xpBtePoisonError) */
+	xpRETIRED5,		/* 31: (formerly xpBteWriteError) */
+	xpRETIRED6,		/* 32: (formerly xpBteAccessError) */
+	xpRETIRED7,		/* 33: (formerly xpBtePWriteError) */
+	xpRETIRED8,		/* 34: (formerly xpBtePReadError) */
+	xpRETIRED9,		/* 35: (formerly xpBteTimeOutError) */
+	xpRETIRED10,		/* 36: (formerly xpBteXtalkError) */
+	xpRETIRED11,		/* 37: (formerly xpBteNotAvailable) */
+	xpRETIRED12,		/* 38: (formerly xpBteUnmappedError) */
+
+	xpBadVersion,		/* 39: bad version number */
+	xpVarsNotSet,		/* 40: the XPC variables are not set up */
+	xpNoRsvdPageAddr,	/* 41: unable to get rsvd page's phys addr */
+	xpInvalidPartid,	/* 42: invalid partition ID */
+	xpLocalPartid,		/* 43: local partition ID */
+
+	xpOtherGoingDown,	/* 44: other side going down, reason unknown */
+	xpSystemGoingDown,	/* 45: system is going down, reason unknown */
+	xpSystemHalt,		/* 46: system is being halted */
+	xpSystemReboot,		/* 47: system is being rebooted */
+	xpSystemPoweroff,	/* 48: system is being powered off */
+
+	xpDisconnecting,	/* 49: channel disconnecting (closing) */
+
+	xpOpenCloseError,	/* 50: channel open/close protocol error */
+
+	xpDisconnected,		/* 51: channel disconnected (closed) */
+
+	xpBteCopyError,		/* 52: bte_copy() returned error */
+	xpSalError,		/* 53: sn SAL error */
+	xpRsvdPageNotSet,	/* 54: the reserved page is not set up */
+	xpPayloadTooBig,	/* 55: payload too large for message slot */
+
+	xpUnsupported,		/* 56: unsupported functionality or resource */
+	xpNeedMoreInfo,		/* 57: more info is needed by SAL */
+
+	xpGruCopyError,		/* 58: gru_copy_gru() returned error */
+	xpGruSendMqError,	/* 59: gru send message queue related error */
+
+	xpBadChannelNumber,	/* 60: invalid channel number */
+	xpBadMsgType,		/* 61: invalid message type */
+	xpBiosError,		/* 62: BIOS error */
+
+	xpUnknownReason		/* 63: unknown reason - must be last in enum */
+};
+
+/*
+ * Define the callout function type used by XPC to update the user on
+ * connection activity and state changes via the user function registered
+ * by xpc_connect().
+ *
+ * Arguments:
+ *
+ *	reason - reason code.
+ *	partid - partition ID associated with condition.
+ *	ch_number - channel # associated with condition.
+ *	data - pointer to optional data.
+ *	key - pointer to optional user-defined value provided as the "key"
+ *	      argument to xpc_connect().
+ *
+ * A reason code of xpConnected indicates that a connection has been
+ * established to the specified partition on the specified channel. The data
+ * argument indicates the max number of entries allowed in the message queue.
+ *
+ * A reason code of xpMsgReceived indicates that a XPC message arrived from
+ * the specified partition on the specified channel. The data argument
+ * specifies the address of the message's payload. The user must call
+ * xpc_received() when finished with the payload.
+ *
+ * All other reason codes indicate failure. The data argmument is NULL.
+ * When a failure reason code is received, one can assume that the channel
+ * is not connected.
+ */
+typedef void (*xpc_channel_func) (enum xp_retval reason, short partid,
+				  int ch_number, void *data, void *key);
+
+/*
+ * Define the callout function type used by XPC to notify the user of
+ * messages received and delivered via the user function registered by
+ * xpc_send_notify().
+ *
+ * Arguments:
+ *
+ *	reason - reason code.
+ *	partid - partition ID associated with condition.
+ *	ch_number - channel # associated with condition.
+ *	key - pointer to optional user-defined value provided as the "key"
+ *	      argument to xpc_send_notify().
+ *
+ * A reason code of xpMsgDelivered indicates that the message was delivered
+ * to the intended recipient and that they have acknowledged its receipt by
+ * calling xpc_received().
+ *
+ * All other reason codes indicate failure.
+ *
+ * NOTE: The user defined function must be callable by an interrupt handler
+ *       and thus cannot block.
+ */
+typedef void (*xpc_notify_func) (enum xp_retval reason, short partid,
+				 int ch_number, void *key);
+
+/*
+ * The following is a registration entry. There is a global array of these,
+ * one per channel. It is used to record the connection registration made
+ * by the users of XPC. As long as a registration entry exists, for any
+ * partition that comes up, XPC will attempt to establish a connection on
+ * that channel. Notification that a connection has been made will occur via
+ * the xpc_channel_func function.
+ *
+ * The 'func' field points to the function to call when aynchronous
+ * notification is required for such events as: a connection established/lost,
+ * or an incoming message received, or an error condition encountered. A
+ * non-NULL 'func' field indicates that there is an active registration for
+ * the channel.
+ */
+struct xpc_registration {
+	struct mutex mutex;
+	xpc_channel_func func;	/* function to call */
+	void *key;		/* pointer to user's key */
+	u16 nentries;		/* #of msg entries in local msg queue */
+	u16 entry_size;		/* message queue's message entry size */
+	u32 assigned_limit;	/* limit on #of assigned kthreads */
+	u32 idle_limit;		/* limit on #of idle kthreads */
+} ____cacheline_aligned;
+
+#define XPC_CHANNEL_REGISTERED(_c)	(xpc_registrations[_c].func != NULL)
+
+/* the following are valid xpc_send() or xpc_send_notify() flags */
+#define XPC_WAIT	0	/* wait flag */
+#define XPC_NOWAIT	1	/* no wait flag */
+
+struct xpc_interface {
+	void (*connect) (int);
+	void (*disconnect) (int);
+	enum xp_retval (*send) (short, int, u32, void *, u16);
+	enum xp_retval (*send_notify) (short, int, u32, void *, u16,
+					xpc_notify_func, void *);
+	void (*received) (short, int, void *);
+	enum xp_retval (*partid_to_nasids) (short, void *);
+};
+
+extern struct xpc_interface xpc_interface;
+
+extern void xpc_set_interface(void (*)(int),
+			      void (*)(int),
+			      enum xp_retval (*)(short, int, u32, void *, u16),
+			      enum xp_retval (*)(short, int, u32, void *, u16,
+						 xpc_notify_func, void *),
+			      void (*)(short, int, void *),
+			      enum xp_retval (*)(short, void *));
+extern void xpc_clear_interface(void);
+
+extern enum xp_retval xpc_connect(int, xpc_channel_func, void *, u16,
+				   u16, u32, u32);
+extern void xpc_disconnect(int);
+
+static inline enum xp_retval
+xpc_send(short partid, int ch_number, u32 flags, void *payload,
+	 u16 payload_size)
+{
+	return xpc_interface.send(partid, ch_number, flags, payload,
+				  payload_size);
+}
+
+static inline enum xp_retval
+xpc_send_notify(short partid, int ch_number, u32 flags, void *payload,
+		u16 payload_size, xpc_notify_func func, void *key)
+{
+	return xpc_interface.send_notify(partid, ch_number, flags, payload,
+					 payload_size, func, key);
+}
+
+static inline void
+xpc_received(short partid, int ch_number, void *payload)
+{
+	return xpc_interface.received(partid, ch_number, payload);
+}
+
+static inline enum xp_retval
+xpc_partid_to_nasids(short partid, void *nasids)
+{
+	return xpc_interface.partid_to_nasids(partid, nasids);
+}
+
+extern short xp_max_npartitions;
+extern short xp_partition_id;
+extern u8 xp_region_size;
+
+extern unsigned long (*xp_pa) (void *);
+extern unsigned long (*xp_socket_pa) (unsigned long);
+extern enum xp_retval (*xp_remote_memcpy) (unsigned long, const unsigned long,
+		       size_t);
+extern int (*xp_cpu_to_nasid) (int);
+extern enum xp_retval (*xp_expand_memprotect) (unsigned long, unsigned long);
+extern enum xp_retval (*xp_restrict_memprotect) (unsigned long, unsigned long);
+
+extern u64 xp_nofault_PIOR_target;
+extern int xp_nofault_PIOR(void *);
+extern int xp_error_PIOR(void);
+
+extern struct device *xp;
+extern enum xp_retval xp_init_sn2(void);
+extern enum xp_retval xp_init_uv(void);
+extern void xp_exit_sn2(void);
+extern void xp_exit_uv(void);
+
+#endif /* _DRIVERS_MISC_SGIXP_XP_H */

kernel/drivers/misc/sgi-xp/xp_main.c

@@ -0,0 +1,286 @@
+/*
+ * 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) 2004-2008 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+/*
+ * Cross Partition (XP) base.
+ *
+ *	XP provides a base from which its users can interact
+ *	with XPC, yet not be dependent on XPC.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/device.h>
+#include "xp.h"
+
+/* define the XP debug device structures to be used with dev_dbg() et al */
+
+struct device_driver xp_dbg_name = {
+	.name = "xp"
+};
+
+struct device xp_dbg_subname = {
+	.init_name = "",		/* set to "" */
+	.driver = &xp_dbg_name
+};
+
+struct device *xp = &xp_dbg_subname;
+
+/* max #of partitions possible */
+short xp_max_npartitions;
+EXPORT_SYMBOL_GPL(xp_max_npartitions);
+
+short xp_partition_id;
+EXPORT_SYMBOL_GPL(xp_partition_id);
+
+u8 xp_region_size;
+EXPORT_SYMBOL_GPL(xp_region_size);
+
+unsigned long (*xp_pa) (void *addr);
+EXPORT_SYMBOL_GPL(xp_pa);
+
+unsigned long (*xp_socket_pa) (unsigned long gpa);
+EXPORT_SYMBOL_GPL(xp_socket_pa);
+
+enum xp_retval (*xp_remote_memcpy) (unsigned long dst_gpa,
+				    const unsigned long src_gpa, size_t len);
+EXPORT_SYMBOL_GPL(xp_remote_memcpy);
+
+int (*xp_cpu_to_nasid) (int cpuid);
+EXPORT_SYMBOL_GPL(xp_cpu_to_nasid);
+
+enum xp_retval (*xp_expand_memprotect) (unsigned long phys_addr,
+					unsigned long size);
+EXPORT_SYMBOL_GPL(xp_expand_memprotect);
+enum xp_retval (*xp_restrict_memprotect) (unsigned long phys_addr,
+					  unsigned long size);
+EXPORT_SYMBOL_GPL(xp_restrict_memprotect);
+
+/*
+ * xpc_registrations[] keeps track of xpc_connect()'s done by the kernel-level
+ * users of XPC.
+ */
+struct xpc_registration xpc_registrations[XPC_MAX_NCHANNELS];
+EXPORT_SYMBOL_GPL(xpc_registrations);
+
+/*
+ * Initialize the XPC interface to indicate that XPC isn't loaded.
+ */
+static enum xp_retval
+xpc_notloaded(void)
+{
+	return xpNotLoaded;
+}
+
+struct xpc_interface xpc_interface = {
+	(void (*)(int))xpc_notloaded,
+	(void (*)(int))xpc_notloaded,
+	(enum xp_retval(*)(short, int, u32, void *, u16))xpc_notloaded,
+	(enum xp_retval(*)(short, int, u32, void *, u16, xpc_notify_func,
+			   void *))xpc_notloaded,
+	(void (*)(short, int, void *))xpc_notloaded,
+	(enum xp_retval(*)(short, void *))xpc_notloaded
+};
+EXPORT_SYMBOL_GPL(xpc_interface);
+
+/*
+ * XPC calls this when it (the XPC module) has been loaded.
+ */
+void
+xpc_set_interface(void (*connect) (int),
+		  void (*disconnect) (int),
+		  enum xp_retval (*send) (short, int, u32, void *, u16),
+		  enum xp_retval (*send_notify) (short, int, u32, void *, u16,
+						  xpc_notify_func, void *),
+		  void (*received) (short, int, void *),
+		  enum xp_retval (*partid_to_nasids) (short, void *))
+{
+	xpc_interface.connect = connect;
+	xpc_interface.disconnect = disconnect;
+	xpc_interface.send = send;
+	xpc_interface.send_notify = send_notify;
+	xpc_interface.received = received;
+	xpc_interface.partid_to_nasids = partid_to_nasids;
+}
+EXPORT_SYMBOL_GPL(xpc_set_interface);
+
+/*
+ * XPC calls this when it (the XPC module) is being unloaded.
+ */
+void
+xpc_clear_interface(void)
+{
+	xpc_interface.connect = (void (*)(int))xpc_notloaded;
+	xpc_interface.disconnect = (void (*)(int))xpc_notloaded;
+	xpc_interface.send = (enum xp_retval(*)(short, int, u32, void *, u16))
+	    xpc_notloaded;
+	xpc_interface.send_notify = (enum xp_retval(*)(short, int, u32, void *,
+						       u16, xpc_notify_func,
+						       void *))xpc_notloaded;
+	xpc_interface.received = (void (*)(short, int, void *))
+	    xpc_notloaded;
+	xpc_interface.partid_to_nasids = (enum xp_retval(*)(short, void *))
+	    xpc_notloaded;
+}
+EXPORT_SYMBOL_GPL(xpc_clear_interface);
+
+/*
+ * Register for automatic establishment of a channel connection whenever
+ * a partition comes up.
+ *
+ * Arguments:
+ *
+ *	ch_number - channel # to register for connection.
+ *	func - function to call for asynchronous notification of channel
+ *	       state changes (i.e., connection, disconnection, error) and
+ *	       the arrival of incoming messages.
+ *      key - pointer to optional user-defined value that gets passed back
+ *	      to the user on any callouts made to func.
+ *	payload_size - size in bytes of the XPC message's payload area which
+ *		       contains a user-defined message. The user should make
+ *		       this large enough to hold their largest message.
+ *	nentries - max #of XPC message entries a message queue can contain.
+ *		   The actual number, which is determined when a connection
+ * 		   is established and may be less then requested, will be
+ *		   passed to the user via the xpConnected callout.
+ *	assigned_limit - max number of kthreads allowed to be processing
+ * 			 messages (per connection) at any given instant.
+ *	idle_limit - max number of kthreads allowed to be idle at any given
+ * 		     instant.
+ */
+enum xp_retval
+xpc_connect(int ch_number, xpc_channel_func func, void *key, u16 payload_size,
+	    u16 nentries, u32 assigned_limit, u32 idle_limit)
+{
+	struct xpc_registration *registration;
+
+	DBUG_ON(ch_number < 0 || ch_number >= XPC_MAX_NCHANNELS);
+	DBUG_ON(payload_size == 0 || nentries == 0);
+	DBUG_ON(func == NULL);
+	DBUG_ON(assigned_limit == 0 || idle_limit > assigned_limit);
+
+	if (XPC_MSG_SIZE(payload_size) > XPC_MSG_MAX_SIZE)
+		return xpPayloadTooBig;
+
+	registration = &xpc_registrations[ch_number];
+
+	if (mutex_lock_interruptible(&registration->mutex) != 0)
+		return xpInterrupted;
+
+	/* if XPC_CHANNEL_REGISTERED(ch_number) */
+	if (registration->func != NULL) {
+		mutex_unlock(&registration->mutex);
+		return xpAlreadyRegistered;
+	}
+
+	/* register the channel for connection */
+	registration->entry_size = XPC_MSG_SIZE(payload_size);
+	registration->nentries = nentries;
+	registration->assigned_limit = assigned_limit;
+	registration->idle_limit = idle_limit;
+	registration->key = key;
+	registration->func = func;
+
+	mutex_unlock(&registration->mutex);
+
+	xpc_interface.connect(ch_number);
+
+	return xpSuccess;
+}
+EXPORT_SYMBOL_GPL(xpc_connect);
+
+/*
+ * Remove the registration for automatic connection of the specified channel
+ * when a partition comes up.
+ *
+ * Before returning this xpc_disconnect() will wait for all connections on the
+ * specified channel have been closed/torndown. So the caller can be assured
+ * that they will not be receiving any more callouts from XPC to their
+ * function registered via xpc_connect().
+ *
+ * Arguments:
+ *
+ *	ch_number - channel # to unregister.
+ */
+void
+xpc_disconnect(int ch_number)
+{
+	struct xpc_registration *registration;
+
+	DBUG_ON(ch_number < 0 || ch_number >= XPC_MAX_NCHANNELS);
+
+	registration = &xpc_registrations[ch_number];
+
+	/*
+	 * We've decided not to make this a down_interruptible(), since we
+	 * figured XPC's users will just turn around and call xpc_disconnect()
+	 * again anyways, so we might as well wait, if need be.
+	 */
+	mutex_lock(&registration->mutex);
+
+	/* if !XPC_CHANNEL_REGISTERED(ch_number) */
+	if (registration->func == NULL) {
+		mutex_unlock(&registration->mutex);
+		return;
+	}
+
+	/* remove the connection registration for the specified channel */
+	registration->func = NULL;
+	registration->key = NULL;
+	registration->nentries = 0;
+	registration->entry_size = 0;
+	registration->assigned_limit = 0;
+	registration->idle_limit = 0;
+
+	xpc_interface.disconnect(ch_number);
+
+	mutex_unlock(&registration->mutex);
+
+	return;
+}
+EXPORT_SYMBOL_GPL(xpc_disconnect);
+
+int __init
+xp_init(void)
+{
+	enum xp_retval ret;
+	int ch_number;
+
+	/* initialize the connection registration mutex */
+	for (ch_number = 0; ch_number < XPC_MAX_NCHANNELS; ch_number++)
+		mutex_init(&xpc_registrations[ch_number].mutex);
+
+	if (is_shub())
+		ret = xp_init_sn2();
+	else if (is_uv())
+		ret = xp_init_uv();
+	else
+		ret = 0;
+
+	if (ret != xpSuccess)
+		return ret;
+
+	return 0;
+}
+
+module_init(xp_init);
+
+void __exit
+xp_exit(void)
+{
+	if (is_shub())
+		xp_exit_sn2();
+	else if (is_uv())
+		xp_exit_uv();
+}
+
+module_exit(xp_exit);
+
+MODULE_AUTHOR("Silicon Graphics, Inc.");
+MODULE_DESCRIPTION("Cross Partition (XP) base");
+MODULE_LICENSE("GPL");

kernel/drivers/misc/sgi-xp/xp_nofault.S

@@ -0,0 +1,35 @@
+/*
+ * 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) 2004-2008 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+/*
+ * The xp_nofault_PIOR function takes a pointer to a remote PIO register
+ * and attempts to load and consume a value from it.  This function
+ * will be registered as a nofault code block.  In the event that the
+ * PIO read fails, the MCA handler will force the error to look
+ * corrected and vector to the xp_error_PIOR which will return an error.
+ *
+ * The definition of "consumption" and the time it takes for an MCA
+ * to surface is processor implementation specific.  This code
+ * is sufficient on Itanium through the Montvale processor family.
+ * It may need to be adjusted for future processor implementations.
+ *
+ *	extern int xp_nofault_PIOR(void *remote_register);
+ */
+
+	.global xp_nofault_PIOR
+xp_nofault_PIOR:
+	mov	r8=r0			// Stage a success return value
+	ld8.acq	r9=[r32];;		// PIO Read the specified register
+	adds	r9=1,r9;;		// Add to force consumption
+	srlz.i;;			// Allow time for MCA to surface
+	br.ret.sptk.many b0;;		// Return success
+
+	.global xp_error_PIOR
+xp_error_PIOR:
+	mov	r8=1			// Return value of 1
+	br.ret.sptk.many b0;;		// Return failure

kernel/drivers/misc/sgi-xp/xp_sn2.c

@@ -0,0 +1,190 @@
+/*
+ * 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) 2008 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+/*
+ * Cross Partition (XP) sn2-based functions.
+ *
+ *      Architecture specific implementation of common functions.
+ */
+
+#include <linux/module.h>
+#include <linux/device.h>
+#include <asm/sn/bte.h>
+#include <asm/sn/sn_sal.h>
+#include "xp.h"
+
+/*
+ * The export of xp_nofault_PIOR needs to happen here since it is defined
+ * in drivers/misc/sgi-xp/xp_nofault.S. The target of the nofault read is
+ * defined here.
+ */
+EXPORT_SYMBOL_GPL(xp_nofault_PIOR);
+
+u64 xp_nofault_PIOR_target;
+EXPORT_SYMBOL_GPL(xp_nofault_PIOR_target);
+
+/*
+ * Register a nofault code region which performs a cross-partition PIO read.
+ * If the PIO read times out, the MCA handler will consume the error and
+ * return to a kernel-provided instruction to indicate an error. This PIO read
+ * exists because it is guaranteed to timeout if the destination is down
+ * (amo operations do not timeout on at least some CPUs on Shubs <= v1.2,
+ * which unfortunately we have to work around).
+ */
+static enum xp_retval
+xp_register_nofault_code_sn2(void)
+{
+	int ret;
+	u64 func_addr;
+	u64 err_func_addr;
+
+	func_addr = *(u64 *)xp_nofault_PIOR;
+	err_func_addr = *(u64 *)xp_error_PIOR;
+	ret = sn_register_nofault_code(func_addr, err_func_addr, err_func_addr,
+				       1, 1);
+	if (ret != 0) {
+		dev_err(xp, "can't register nofault code, error=%d\n", ret);
+		return xpSalError;
+	}
+	/*
+	 * Setup the nofault PIO read target. (There is no special reason why
+	 * SH_IPI_ACCESS was selected.)
+	 */
+	if (is_shub1())
+		xp_nofault_PIOR_target = SH1_IPI_ACCESS;
+	else if (is_shub2())
+		xp_nofault_PIOR_target = SH2_IPI_ACCESS0;
+
+	return xpSuccess;
+}
+
+static void
+xp_unregister_nofault_code_sn2(void)
+{
+	u64 func_addr = *(u64 *)xp_nofault_PIOR;
+	u64 err_func_addr = *(u64 *)xp_error_PIOR;
+
+	/* unregister the PIO read nofault code region */
+	(void)sn_register_nofault_code(func_addr, err_func_addr,
+				       err_func_addr, 1, 0);
+}
+
+/*
+ * Convert a virtual memory address to a physical memory address.
+ */
+static unsigned long
+xp_pa_sn2(void *addr)
+{
+	return __pa(addr);
+}
+
+/*
+ * Convert a global physical to a socket physical address.
+ */
+static unsigned long
+xp_socket_pa_sn2(unsigned long gpa)
+{
+	return gpa;
+}
+
+/*
+ * Wrapper for bte_copy().
+ *
+ *	dst_pa - physical address of the destination of the transfer.
+ *	src_pa - physical address of the source of the transfer.
+ *	len - number of bytes to transfer from source to destination.
+ *
+ * Note: xp_remote_memcpy_sn2() should never be called while holding a spinlock.
+ */
+static enum xp_retval
+xp_remote_memcpy_sn2(unsigned long dst_pa, const unsigned long src_pa,
+		     size_t len)
+{
+	bte_result_t ret;
+
+	ret = bte_copy(src_pa, dst_pa, len, (BTE_NOTIFY | BTE_WACQUIRE), NULL);
+	if (ret == BTE_SUCCESS)
+		return xpSuccess;
+
+	if (is_shub2()) {
+		dev_err(xp, "bte_copy() on shub2 failed, error=0x%x dst_pa="
+			"0x%016lx src_pa=0x%016lx len=%ld\\n", ret, dst_pa,
+			src_pa, len);
+	} else {
+		dev_err(xp, "bte_copy() failed, error=%d dst_pa=0x%016lx "
+			"src_pa=0x%016lx len=%ld\\n", ret, dst_pa, src_pa, len);
+	}
+
+	return xpBteCopyError;
+}
+
+static int
+xp_cpu_to_nasid_sn2(int cpuid)
+{
+	return cpuid_to_nasid(cpuid);
+}
+
+static enum xp_retval
+xp_expand_memprotect_sn2(unsigned long phys_addr, unsigned long size)
+{
+	u64 nasid_array = 0;
+	int ret;
+
+	ret = sn_change_memprotect(phys_addr, size, SN_MEMPROT_ACCESS_CLASS_1,
+				   &nasid_array);
+	if (ret != 0) {
+		dev_err(xp, "sn_change_memprotect(,, "
+			"SN_MEMPROT_ACCESS_CLASS_1,) failed ret=%d\n", ret);
+		return xpSalError;
+	}
+	return xpSuccess;
+}
+
+static enum xp_retval
+xp_restrict_memprotect_sn2(unsigned long phys_addr, unsigned long size)
+{
+	u64 nasid_array = 0;
+	int ret;
+
+	ret = sn_change_memprotect(phys_addr, size, SN_MEMPROT_ACCESS_CLASS_0,
+				   &nasid_array);
+	if (ret != 0) {
+		dev_err(xp, "sn_change_memprotect(,, "
+			"SN_MEMPROT_ACCESS_CLASS_0,) failed ret=%d\n", ret);
+		return xpSalError;
+	}
+	return xpSuccess;
+}
+
+enum xp_retval
+xp_init_sn2(void)
+{
+	BUG_ON(!is_shub());
+
+	xp_max_npartitions = XP_MAX_NPARTITIONS_SN2;
+	xp_partition_id = sn_partition_id;
+	xp_region_size = sn_region_size;
+
+	xp_pa = xp_pa_sn2;
+	xp_socket_pa = xp_socket_pa_sn2;
+	xp_remote_memcpy = xp_remote_memcpy_sn2;
+	xp_cpu_to_nasid = xp_cpu_to_nasid_sn2;
+	xp_expand_memprotect = xp_expand_memprotect_sn2;
+	xp_restrict_memprotect = xp_restrict_memprotect_sn2;
+
+	return xp_register_nofault_code_sn2();
+}
+
+void
+xp_exit_sn2(void)
+{
+	BUG_ON(!is_shub());
+
+	xp_unregister_nofault_code_sn2();
+}
+

kernel/drivers/misc/sgi-xp/xp_uv.c

@@ -0,0 +1,171 @@
+/*
+ * 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) 2008 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+/*
+ * Cross Partition (XP) uv-based functions.
+ *
+ *      Architecture specific implementation of common functions.
+ *
+ */
+
+#include <linux/device.h>
+#include <asm/uv/uv_hub.h>
+#if defined CONFIG_X86_64
+#include <asm/uv/bios.h>
+#elif defined CONFIG_IA64_GENERIC || defined CONFIG_IA64_SGI_UV
+#include <asm/sn/sn_sal.h>
+#endif
+#include "../sgi-gru/grukservices.h"
+#include "xp.h"
+
+/*
+ * Convert a virtual memory address to a physical memory address.
+ */
+static unsigned long
+xp_pa_uv(void *addr)
+{
+	return uv_gpa(addr);
+}
+
+/*
+ * Convert a global physical to socket physical address.
+ */
+static unsigned long
+xp_socket_pa_uv(unsigned long gpa)
+{
+	return uv_gpa_to_soc_phys_ram(gpa);
+}
+
+static enum xp_retval
+xp_remote_mmr_read(unsigned long dst_gpa, const unsigned long src_gpa,
+		   size_t len)
+{
+	int ret;
+	unsigned long *dst_va = __va(uv_gpa_to_soc_phys_ram(dst_gpa));
+
+	BUG_ON(!uv_gpa_in_mmr_space(src_gpa));
+	BUG_ON(len != 8);
+
+	ret = gru_read_gpa(dst_va, src_gpa);
+	if (ret == 0)
+		return xpSuccess;
+
+	dev_err(xp, "gru_read_gpa() failed, dst_gpa=0x%016lx src_gpa=0x%016lx "
+		"len=%ld\n", dst_gpa, src_gpa, len);
+	return xpGruCopyError;
+}
+
+
+static enum xp_retval
+xp_remote_memcpy_uv(unsigned long dst_gpa, const unsigned long src_gpa,
+		    size_t len)
+{
+	int ret;
+
+	if (uv_gpa_in_mmr_space(src_gpa))
+		return xp_remote_mmr_read(dst_gpa, src_gpa, len);
+
+	ret = gru_copy_gpa(dst_gpa, src_gpa, len);
+	if (ret == 0)
+		return xpSuccess;
+
+	dev_err(xp, "gru_copy_gpa() failed, dst_gpa=0x%016lx src_gpa=0x%016lx "
+		"len=%ld\n", dst_gpa, src_gpa, len);
+	return xpGruCopyError;
+}
+
+static int
+xp_cpu_to_nasid_uv(int cpuid)
+{
+	/* ??? Is this same as sn2 nasid in mach/part bitmaps set up by SAL? */
+	return UV_PNODE_TO_NASID(uv_cpu_to_pnode(cpuid));
+}
+
+static enum xp_retval
+xp_expand_memprotect_uv(unsigned long phys_addr, unsigned long size)
+{
+	int ret;
+
+#if defined CONFIG_X86_64
+	ret = uv_bios_change_memprotect(phys_addr, size, UV_MEMPROT_ALLOW_RW);
+	if (ret != BIOS_STATUS_SUCCESS) {
+		dev_err(xp, "uv_bios_change_memprotect(,, "
+			"UV_MEMPROT_ALLOW_RW) failed, ret=%d\n", ret);
+		return xpBiosError;
+	}
+
+#elif defined CONFIG_IA64_GENERIC || defined CONFIG_IA64_SGI_UV
+	u64 nasid_array;
+
+	ret = sn_change_memprotect(phys_addr, size, SN_MEMPROT_ACCESS_CLASS_1,
+				   &nasid_array);
+	if (ret != 0) {
+		dev_err(xp, "sn_change_memprotect(,, "
+			"SN_MEMPROT_ACCESS_CLASS_1,) failed ret=%d\n", ret);
+		return xpSalError;
+	}
+#else
+	#error not a supported configuration
+#endif
+	return xpSuccess;
+}
+
+static enum xp_retval
+xp_restrict_memprotect_uv(unsigned long phys_addr, unsigned long size)
+{
+	int ret;
+
+#if defined CONFIG_X86_64
+	ret = uv_bios_change_memprotect(phys_addr, size,
+					UV_MEMPROT_RESTRICT_ACCESS);
+	if (ret != BIOS_STATUS_SUCCESS) {
+		dev_err(xp, "uv_bios_change_memprotect(,, "
+			"UV_MEMPROT_RESTRICT_ACCESS) failed, ret=%d\n", ret);
+		return xpBiosError;
+	}
+
+#elif defined CONFIG_IA64_GENERIC || defined CONFIG_IA64_SGI_UV
+	u64 nasid_array;
+
+	ret = sn_change_memprotect(phys_addr, size, SN_MEMPROT_ACCESS_CLASS_0,
+				   &nasid_array);
+	if (ret != 0) {
+		dev_err(xp, "sn_change_memprotect(,, "
+			"SN_MEMPROT_ACCESS_CLASS_0,) failed ret=%d\n", ret);
+		return xpSalError;
+	}
+#else
+	#error not a supported configuration
+#endif
+	return xpSuccess;
+}
+
+enum xp_retval
+xp_init_uv(void)
+{
+	BUG_ON(!is_uv());
+
+	xp_max_npartitions = XP_MAX_NPARTITIONS_UV;
+	xp_partition_id = sn_partition_id;
+	xp_region_size = sn_region_size;
+
+	xp_pa = xp_pa_uv;
+	xp_socket_pa = xp_socket_pa_uv;
+	xp_remote_memcpy = xp_remote_memcpy_uv;
+	xp_cpu_to_nasid = xp_cpu_to_nasid_uv;
+	xp_expand_memprotect = xp_expand_memprotect_uv;
+	xp_restrict_memprotect = xp_restrict_memprotect_uv;
+
+	return xpSuccess;
+}
+
+void
+xp_exit_uv(void)
+{
+	BUG_ON(!is_uv());
+}

kernel/drivers/misc/sgi-xp/xpc_partition.c

@@ -0,0 +1,541 @@
+/*
+ * 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) 2004-2008 Silicon Graphics, Inc.  All Rights Reserved.
+ */
+
+/*
+ * Cross Partition Communication (XPC) partition support.
+ *
+ *	This is the part of XPC that detects the presence/absence of
+ *	other partitions. It provides a heartbeat and monitors the
+ *	heartbeats of other partitions.
+ *
+ */
+
+#include <linux/device.h>
+#include <linux/hardirq.h>
+#include <linux/slab.h>
+#include "xpc.h"
+#include <asm/uv/uv_hub.h>
+
+/* XPC is exiting flag */
+int xpc_exiting;
+
+/* this partition's reserved page pointers */
+struct xpc_rsvd_page *xpc_rsvd_page;
+static unsigned long *xpc_part_nasids;
+unsigned long *xpc_mach_nasids;
+
+static int xpc_nasid_mask_nbytes;	/* #of bytes in nasid mask */
+int xpc_nasid_mask_nlongs;	/* #of longs in nasid mask */
+
+struct xpc_partition *xpc_partitions;
+
+/*
+ * Guarantee that the kmalloc'd memory is cacheline aligned.
+ */
+void *
+xpc_kmalloc_cacheline_aligned(size_t size, gfp_t flags, void **base)
+{
+	/* see if kmalloc will give us cachline aligned memory by default */
+	*base = kmalloc(size, flags);
+	if (*base == NULL)
+		return NULL;
+
+	if ((u64)*base == L1_CACHE_ALIGN((u64)*base))
+		return *base;
+
+	kfree(*base);
+
+	/* nope, we'll have to do it ourselves */
+	*base = kmalloc(size + L1_CACHE_BYTES, flags);
+	if (*base == NULL)
+		return NULL;
+
+	return (void *)L1_CACHE_ALIGN((u64)*base);
+}
+
+/*
+ * Given a nasid, get the physical address of the  partition's reserved page
+ * for that nasid. This function returns 0 on any error.
+ */
+static unsigned long
+xpc_get_rsvd_page_pa(int nasid)
+{
+	enum xp_retval ret;
+	u64 cookie = 0;
+	unsigned long rp_pa = nasid;	/* seed with nasid */
+	size_t len = 0;
+	size_t buf_len = 0;
+	void *buf = buf;
+	void *buf_base = NULL;
+	enum xp_retval (*get_partition_rsvd_page_pa)
+		(void *, u64 *, unsigned long *, size_t *) =
+		xpc_arch_ops.get_partition_rsvd_page_pa;
+
+	while (1) {
+
+		/* !!! rp_pa will need to be _gpa on UV.
+		 * ??? So do we save it into the architecture specific parts
+		 * ??? of the xpc_partition structure? Do we rename this
+		 * ??? function or have two versions? Rename rp_pa for UV to
+		 * ??? rp_gpa?
+		 */
+		ret = get_partition_rsvd_page_pa(buf, &cookie, &rp_pa, &len);
+
+		dev_dbg(xpc_part, "SAL returned with ret=%d, cookie=0x%016lx, "
+			"address=0x%016lx, len=0x%016lx\n", ret,
+			(unsigned long)cookie, rp_pa, len);
+
+		if (ret != xpNeedMoreInfo)
+			break;
+
+		/* !!! L1_CACHE_ALIGN() is only a sn2-bte_copy requirement */
+		if (is_shub())
+			len = L1_CACHE_ALIGN(len);
+
+		if (len > buf_len) {
+			if (buf_base != NULL)
+				kfree(buf_base);
+			buf_len = L1_CACHE_ALIGN(len);
+			buf = xpc_kmalloc_cacheline_aligned(buf_len, GFP_KERNEL,
+							    &buf_base);
+			if (buf_base == NULL) {
+				dev_err(xpc_part, "unable to kmalloc "
+					"len=0x%016lx\n", buf_len);
+				ret = xpNoMemory;
+				break;
+			}
+		}
+
+		ret = xp_remote_memcpy(xp_pa(buf), rp_pa, len);
+		if (ret != xpSuccess) {
+			dev_dbg(xpc_part, "xp_remote_memcpy failed %d\n", ret);
+			break;
+		}
+	}
+
+	kfree(buf_base);
+
+	if (ret != xpSuccess)
+		rp_pa = 0;
+
+	dev_dbg(xpc_part, "reserved page at phys address 0x%016lx\n", rp_pa);
+	return rp_pa;
+}
+
+/*
+ * Fill the partition reserved page with the information needed by
+ * other partitions to discover we are alive and establish initial
+ * communications.
+ */
+int
+xpc_setup_rsvd_page(void)
+{
+	int ret;
+	struct xpc_rsvd_page *rp;
+	unsigned long rp_pa;
+	unsigned long new_ts_jiffies;
+
+	/* get the local reserved page's address */
+
+	preempt_disable();
+	rp_pa = xpc_get_rsvd_page_pa(xp_cpu_to_nasid(smp_processor_id()));
+	preempt_enable();
+	if (rp_pa == 0) {
+		dev_err(xpc_part, "SAL failed to locate the reserved page\n");
+		return -ESRCH;
+	}
+	rp = (struct xpc_rsvd_page *)__va(xp_socket_pa(rp_pa));
+
+	if (rp->SAL_version < 3) {
+		/* SAL_versions < 3 had a SAL_partid defined as a u8 */
+		rp->SAL_partid &= 0xff;
+	}
+	BUG_ON(rp->SAL_partid != xp_partition_id);
+
+	if (rp->SAL_partid < 0 || rp->SAL_partid >= xp_max_npartitions) {
+		dev_err(xpc_part, "the reserved page's partid of %d is outside "
+			"supported range (< 0 || >= %d)\n", rp->SAL_partid,
+			xp_max_npartitions);
+		return -EINVAL;
+	}
+
+	rp->version = XPC_RP_VERSION;
+	rp->max_npartitions = xp_max_npartitions;
+
+	/* establish the actual sizes of the nasid masks */
+	if (rp->SAL_version == 1) {
+		/* SAL_version 1 didn't set the nasids_size field */
+		rp->SAL_nasids_size = 128;
+	}
+	xpc_nasid_mask_nbytes = rp->SAL_nasids_size;
+	xpc_nasid_mask_nlongs = BITS_TO_LONGS(rp->SAL_nasids_size *
+					      BITS_PER_BYTE);
+
+	/* setup the pointers to the various items in the reserved page */
+	xpc_part_nasids = XPC_RP_PART_NASIDS(rp);
+	xpc_mach_nasids = XPC_RP_MACH_NASIDS(rp);
+
+	ret = xpc_arch_ops.setup_rsvd_page(rp);
+	if (ret != 0)
+		return ret;
+
+	/*
+	 * Set timestamp of when reserved page was setup by XPC.
+	 * This signifies to the remote partition that our reserved
+	 * page is initialized.
+	 */
+	new_ts_jiffies = jiffies;
+	if (new_ts_jiffies == 0 || new_ts_jiffies == rp->ts_jiffies)
+		new_ts_jiffies++;
+	rp->ts_jiffies = new_ts_jiffies;
+
+	xpc_rsvd_page = rp;
+	return 0;
+}
+
+void
+xpc_teardown_rsvd_page(void)
+{
+	/* a zero timestamp indicates our rsvd page is not initialized */
+	xpc_rsvd_page->ts_jiffies = 0;
+}
+
+/*
+ * Get a copy of a portion of the remote partition's rsvd page.
+ *
+ * remote_rp points to a buffer that is cacheline aligned for BTE copies and
+ * is large enough to contain a copy of their reserved page header and
+ * part_nasids mask.
+ */
+enum xp_retval
+xpc_get_remote_rp(int nasid, unsigned long *discovered_nasids,
+		  struct xpc_rsvd_page *remote_rp, unsigned long *remote_rp_pa)
+{
+	int l;
+	enum xp_retval ret;
+
+	/* get the reserved page's physical address */
+
+	*remote_rp_pa = xpc_get_rsvd_page_pa(nasid);
+	if (*remote_rp_pa == 0)
+		return xpNoRsvdPageAddr;
+
+	/* pull over the reserved page header and part_nasids mask */
+	ret = xp_remote_memcpy(xp_pa(remote_rp), *remote_rp_pa,
+			       XPC_RP_HEADER_SIZE + xpc_nasid_mask_nbytes);
+	if (ret != xpSuccess)
+		return ret;
+
+	if (discovered_nasids != NULL) {
+		unsigned long *remote_part_nasids =
+		    XPC_RP_PART_NASIDS(remote_rp);
+
+		for (l = 0; l < xpc_nasid_mask_nlongs; l++)
+			discovered_nasids[l] |= remote_part_nasids[l];
+	}
+
+	/* zero timestamp indicates the reserved page has not been setup */
+	if (remote_rp->ts_jiffies == 0)
+		return xpRsvdPageNotSet;
+
+	if (XPC_VERSION_MAJOR(remote_rp->version) !=
+	    XPC_VERSION_MAJOR(XPC_RP_VERSION)) {
+		return xpBadVersion;
+	}
+
+	/* check that both remote and local partids are valid for each side */
+	if (remote_rp->SAL_partid < 0 ||
+	    remote_rp->SAL_partid >= xp_max_npartitions ||
+	    remote_rp->max_npartitions <= xp_partition_id) {
+		return xpInvalidPartid;
+	}
+
+	if (remote_rp->SAL_partid == xp_partition_id)
+		return xpLocalPartid;
+
+	return xpSuccess;
+}
+
+/*
+ * See if the other side has responded to a partition deactivate request
+ * from us. Though we requested the remote partition to deactivate with regard
+ * to us, we really only need to wait for the other side to disengage from us.
+ */
+int
+xpc_partition_disengaged(struct xpc_partition *part)
+{
+	short partid = XPC_PARTID(part);
+	int disengaged;
+
+	disengaged = !xpc_arch_ops.partition_engaged(partid);
+	if (part->disengage_timeout) {
+		if (!disengaged) {
+			if (time_is_after_jiffies(part->disengage_timeout)) {
+				/* timelimit hasn't been reached yet */
+				return 0;
+			}
+
+			/*
+			 * Other side hasn't responded to our deactivate
+			 * request in a timely fashion, so assume it's dead.
+			 */
+
+			dev_info(xpc_part, "deactivate request to remote "
+				 "partition %d timed out\n", partid);
+			xpc_disengage_timedout = 1;
+			xpc_arch_ops.assume_partition_disengaged(partid);
+			disengaged = 1;
+		}
+		part->disengage_timeout = 0;
+
+		/* cancel the timer function, provided it's not us */
+		if (!in_interrupt())
+			del_singleshot_timer_sync(&part->disengage_timer);
+
+		DBUG_ON(part->act_state != XPC_P_AS_DEACTIVATING &&
+			part->act_state != XPC_P_AS_INACTIVE);
+		if (part->act_state != XPC_P_AS_INACTIVE)
+			xpc_wakeup_channel_mgr(part);
+
+		xpc_arch_ops.cancel_partition_deactivation_request(part);
+	}
+	return disengaged;
+}
+
+/*
+ * Mark specified partition as active.
+ */
+enum xp_retval
+xpc_mark_partition_active(struct xpc_partition *part)
+{
+	unsigned long irq_flags;
+	enum xp_retval ret;
+
+	dev_dbg(xpc_part, "setting partition %d to ACTIVE\n", XPC_PARTID(part));
+
+	spin_lock_irqsave(&part->act_lock, irq_flags);
+	if (part->act_state == XPC_P_AS_ACTIVATING) {
+		part->act_state = XPC_P_AS_ACTIVE;
+		ret = xpSuccess;
+	} else {
+		DBUG_ON(part->reason == xpSuccess);
+		ret = part->reason;
+	}
+	spin_unlock_irqrestore(&part->act_lock, irq_flags);
+
+	return ret;
+}
+
+/*
+ * Start the process of deactivating the specified partition.
+ */
+void
+xpc_deactivate_partition(const int line, struct xpc_partition *part,
+			 enum xp_retval reason)
+{
+	unsigned long irq_flags;
+
+	spin_lock_irqsave(&part->act_lock, irq_flags);
+
+	if (part->act_state == XPC_P_AS_INACTIVE) {
+		XPC_SET_REASON(part, reason, line);
+		spin_unlock_irqrestore(&part->act_lock, irq_flags);
+		if (reason == xpReactivating) {
+			/* we interrupt ourselves to reactivate partition */
+			xpc_arch_ops.request_partition_reactivation(part);
+		}
+		return;
+	}
+	if (part->act_state == XPC_P_AS_DEACTIVATING) {
+		if ((part->reason == xpUnloading && reason != xpUnloading) ||
+		    reason == xpReactivating) {
+			XPC_SET_REASON(part, reason, line);
+		}
+		spin_unlock_irqrestore(&part->act_lock, irq_flags);
+		return;
+	}
+
+	part->act_state = XPC_P_AS_DEACTIVATING;
+	XPC_SET_REASON(part, reason, line);
+
+	spin_unlock_irqrestore(&part->act_lock, irq_flags);
+
+	/* ask remote partition to deactivate with regard to us */
+	xpc_arch_ops.request_partition_deactivation(part);
+
+	/* set a timelimit on the disengage phase of the deactivation request */
+	part->disengage_timeout = jiffies + (xpc_disengage_timelimit * HZ);
+	part->disengage_timer.expires = part->disengage_timeout;
+	add_timer(&part->disengage_timer);
+
+	dev_dbg(xpc_part, "bringing partition %d down, reason = %d\n",
+		XPC_PARTID(part), reason);
+
+	xpc_partition_going_down(part, reason);
+}
+
+/*
+ * Mark specified partition as inactive.
+ */
+void
+xpc_mark_partition_inactive(struct xpc_partition *part)
+{
+	unsigned long irq_flags;
+
+	dev_dbg(xpc_part, "setting partition %d to INACTIVE\n",
+		XPC_PARTID(part));
+
+	spin_lock_irqsave(&part->act_lock, irq_flags);
+	part->act_state = XPC_P_AS_INACTIVE;
+	spin_unlock_irqrestore(&part->act_lock, irq_flags);
+	part->remote_rp_pa = 0;
+}
+
+/*
+ * SAL has provided a partition and machine mask.  The partition mask
+ * contains a bit for each even nasid in our partition.  The machine
+ * mask contains a bit for each even nasid in the entire machine.
+ *
+ * Using those two bit arrays, we can determine which nasids are
+ * known in the machine.  Each should also have a reserved page
+ * initialized if they are available for partitioning.
+ */
+void
+xpc_discovery(void)
+{
+	void *remote_rp_base;
+	struct xpc_rsvd_page *remote_rp;
+	unsigned long remote_rp_pa;
+	int region;
+	int region_size;
+	int max_regions;
+	int nasid;
+	struct xpc_rsvd_page *rp;
+	unsigned long *discovered_nasids;
+	enum xp_retval ret;
+
+	remote_rp = xpc_kmalloc_cacheline_aligned(XPC_RP_HEADER_SIZE +
+						  xpc_nasid_mask_nbytes,
+						  GFP_KERNEL, &remote_rp_base);
+	if (remote_rp == NULL)
+		return;
+
+	discovered_nasids = kzalloc(sizeof(long) * xpc_nasid_mask_nlongs,
+				    GFP_KERNEL);
+	if (discovered_nasids == NULL) {
+		kfree(remote_rp_base);
+		return;
+	}
+
+	rp = (struct xpc_rsvd_page *)xpc_rsvd_page;
+
+	/*
+	 * The term 'region' in this context refers to the minimum number of
+	 * nodes that can comprise an access protection grouping. The access
+	 * protection is in regards to memory, IOI and IPI.
+	 */
+	region_size = xp_region_size;
+
+	if (is_uv())
+		max_regions = 256;
+	else {
+		max_regions = 64;
+
+		switch (region_size) {
+		case 128:
+			max_regions *= 2;
+		case 64:
+			max_regions *= 2;
+		case 32:
+			max_regions *= 2;
+			region_size = 16;
+			DBUG_ON(!is_shub2());
+		}
+	}
+
+	for (region = 0; region < max_regions; region++) {
+
+		if (xpc_exiting)
+			break;
+
+		dev_dbg(xpc_part, "searching region %d\n", region);
+
+		for (nasid = (region * region_size * 2);
+		     nasid < ((region + 1) * region_size * 2); nasid += 2) {
+
+			if (xpc_exiting)
+				break;
+
+			dev_dbg(xpc_part, "checking nasid %d\n", nasid);
+
+			if (test_bit(nasid / 2, xpc_part_nasids)) {
+				dev_dbg(xpc_part, "PROM indicates Nasid %d is "
+					"part of the local partition; skipping "
+					"region\n", nasid);
+				break;
+			}
+
+			if (!(test_bit(nasid / 2, xpc_mach_nasids))) {
+				dev_dbg(xpc_part, "PROM indicates Nasid %d was "
+					"not on Numa-Link network at reset\n",
+					nasid);
+				continue;
+			}
+
+			if (test_bit(nasid / 2, discovered_nasids)) {
+				dev_dbg(xpc_part, "Nasid %d is part of a "
+					"partition which was previously "
+					"discovered\n", nasid);
+				continue;
+			}
+
+			/* pull over the rsvd page header & part_nasids mask */
+
+			ret = xpc_get_remote_rp(nasid, discovered_nasids,
+						remote_rp, &remote_rp_pa);
+			if (ret != xpSuccess) {
+				dev_dbg(xpc_part, "unable to get reserved page "
+					"from nasid %d, reason=%d\n", nasid,
+					ret);
+
+				if (ret == xpLocalPartid)
+					break;
+
+				continue;
+			}
+
+			xpc_arch_ops.request_partition_activation(remote_rp,
+							 remote_rp_pa, nasid);
+		}
+	}
+
+	kfree(discovered_nasids);
+	kfree(remote_rp_base);
+}
+
+/*
+ * Given a partid, get the nasids owned by that partition from the
+ * remote partition's reserved page.
+ */
+enum xp_retval
+xpc_initiate_partid_to_nasids(short partid, void *nasid_mask)
+{
+	struct xpc_partition *part;
+	unsigned long part_nasid_pa;
+
+	part = &xpc_partitions[partid];
+	if (part->remote_rp_pa == 0)
+		return xpPartitionDown;
+
+	memset(nasid_mask, 0, xpc_nasid_mask_nbytes);
+
+	part_nasid_pa = (unsigned long)XPC_RP_PART_NASIDS(part->remote_rp_pa);
+
+	return xp_remote_memcpy(xp_pa(nasid_mask), part_nasid_pa,
+				xpc_nasid_mask_nbytes);
+}

kernel/drivers/misc/sgi-xp/xpnet.c

@@ -0,0 +1,607 @@
+/*
+ * 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) 1999-2009 Silicon Graphics, Inc. All rights reserved.
+ */
+
+/*
+ * Cross Partition Network Interface (XPNET) support
+ *
+ *	XPNET provides a virtual network layered on top of the Cross
+ *	Partition communication layer.
+ *
+ *	XPNET provides direct point-to-point and broadcast-like support
+ *	for an ethernet-like device.  The ethernet broadcast medium is
+ *	replaced with a point-to-point message structure which passes
+ *	pointers to a DMA-capable block that a remote partition should
+ *	retrieve and pass to the upper level networking layer.
+ *
+ */
+
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include "xp.h"
+
+/*
+ * The message payload transferred by XPC.
+ *
+ * buf_pa is the physical address where the DMA should pull from.
+ *
+ * NOTE: for performance reasons, buf_pa should _ALWAYS_ begin on a
+ * cacheline boundary.  To accomplish this, we record the number of
+ * bytes from the beginning of the first cacheline to the first useful
+ * byte of the skb (leadin_ignore) and the number of bytes from the
+ * last useful byte of the skb to the end of the last cacheline
+ * (tailout_ignore).
+ *
+ * size is the number of bytes to transfer which includes the skb->len
+ * (useful bytes of the senders skb) plus the leadin and tailout
+ */
+struct xpnet_message {
+	u16 version;		/* Version for this message */
+	u16 embedded_bytes;	/* #of bytes embedded in XPC message */
+	u32 magic;		/* Special number indicating this is xpnet */
+	unsigned long buf_pa;	/* phys address of buffer to retrieve */
+	u32 size;		/* #of bytes in buffer */
+	u8 leadin_ignore;	/* #of bytes to ignore at the beginning */
+	u8 tailout_ignore;	/* #of bytes to ignore at the end */
+	unsigned char data;	/* body of small packets */
+};
+
+/*
+ * Determine the size of our message, the cacheline aligned size,
+ * and then the number of message will request from XPC.
+ *
+ * XPC expects each message to exist in an individual cacheline.
+ */
+#define XPNET_MSG_SIZE		XPC_MSG_PAYLOAD_MAX_SIZE
+#define XPNET_MSG_DATA_MAX	\
+		(XPNET_MSG_SIZE - offsetof(struct xpnet_message, data))
+#define XPNET_MSG_NENTRIES	(PAGE_SIZE / XPC_MSG_MAX_SIZE)
+
+#define XPNET_MAX_KTHREADS	(XPNET_MSG_NENTRIES + 1)
+#define XPNET_MAX_IDLE_KTHREADS	(XPNET_MSG_NENTRIES + 1)
+
+/*
+ * Version number of XPNET implementation. XPNET can always talk to versions
+ * with same major #, and never talk to versions with a different version.
+ */
+#define _XPNET_VERSION(_major, _minor)	(((_major) << 4) | (_minor))
+#define XPNET_VERSION_MAJOR(_v)		((_v) >> 4)
+#define XPNET_VERSION_MINOR(_v)		((_v) & 0xf)
+
+#define	XPNET_VERSION _XPNET_VERSION(1, 0)	/* version 1.0 */
+#define	XPNET_VERSION_EMBED _XPNET_VERSION(1, 1)	/* version 1.1 */
+#define XPNET_MAGIC	0x88786984	/* "XNET" */
+
+#define XPNET_VALID_MSG(_m)						     \
+   ((XPNET_VERSION_MAJOR(_m->version) == XPNET_VERSION_MAJOR(XPNET_VERSION)) \
+    && (msg->magic == XPNET_MAGIC))
+
+#define XPNET_DEVICE_NAME		"xp0"
+
+/*
+ * When messages are queued with xpc_send_notify, a kmalloc'd buffer
+ * of the following type is passed as a notification cookie.  When the
+ * notification function is called, we use the cookie to decide
+ * whether all outstanding message sends have completed.  The skb can
+ * then be released.
+ */
+struct xpnet_pending_msg {
+	struct sk_buff *skb;
+	atomic_t use_count;
+};
+
+struct net_device *xpnet_device;
+
+/*
+ * When we are notified of other partitions activating, we add them to
+ * our bitmask of partitions to which we broadcast.
+ */
+static unsigned long *xpnet_broadcast_partitions;
+/* protect above */
+static DEFINE_SPINLOCK(xpnet_broadcast_lock);
+
+/*
+ * Since the Block Transfer Engine (BTE) is being used for the transfer
+ * and it relies upon cache-line size transfers, we need to reserve at
+ * least one cache-line for head and tail alignment.  The BTE is
+ * limited to 8MB transfers.
+ *
+ * Testing has shown that changing MTU to greater than 64KB has no effect
+ * on TCP as the two sides negotiate a Max Segment Size that is limited
+ * to 64K.  Other protocols May use packets greater than this, but for
+ * now, the default is 64KB.
+ */
+#define XPNET_MAX_MTU (0x800000UL - L1_CACHE_BYTES)
+/* 32KB has been determined to be the ideal */
+#define XPNET_DEF_MTU (0x8000UL)
+
+/*
+ * The partid is encapsulated in the MAC address beginning in the following
+ * octet and it consists of two octets.
+ */
+#define XPNET_PARTID_OCTET	2
+
+/* Define the XPNET debug device structures to be used with dev_dbg() et al */
+
+struct device_driver xpnet_dbg_name = {
+	.name = "xpnet"
+};
+
+struct device xpnet_dbg_subname = {
+	.init_name = "",	/* set to "" */
+	.driver = &xpnet_dbg_name
+};
+
+struct device *xpnet = &xpnet_dbg_subname;
+
+/*
+ * Packet was recevied by XPC and forwarded to us.
+ */
+static void
+xpnet_receive(short partid, int channel, struct xpnet_message *msg)
+{
+	struct sk_buff *skb;
+	void *dst;
+	enum xp_retval ret;
+
+	if (!XPNET_VALID_MSG(msg)) {
+		/*
+		 * Packet with a different XPC version.  Ignore.
+		 */
+		xpc_received(partid, channel, (void *)msg);
+
+		xpnet_device->stats.rx_errors++;
+
+		return;
+	}
+	dev_dbg(xpnet, "received 0x%lx, %d, %d, %d\n", msg->buf_pa, msg->size,
+		msg->leadin_ignore, msg->tailout_ignore);
+
+	/* reserve an extra cache line */
+	skb = dev_alloc_skb(msg->size + L1_CACHE_BYTES);
+	if (!skb) {
+		dev_err(xpnet, "failed on dev_alloc_skb(%d)\n",
+			msg->size + L1_CACHE_BYTES);
+
+		xpc_received(partid, channel, (void *)msg);
+
+		xpnet_device->stats.rx_errors++;
+
+		return;
+	}
+
+	/*
+	 * The allocated skb has some reserved space.
+	 * In order to use xp_remote_memcpy(), we need to get the
+	 * skb->data pointer moved forward.
+	 */
+	skb_reserve(skb, (L1_CACHE_BYTES - ((u64)skb->data &
+					    (L1_CACHE_BYTES - 1)) +
+			  msg->leadin_ignore));
+
+	/*
+	 * Update the tail pointer to indicate data actually
+	 * transferred.
+	 */
+	skb_put(skb, (msg->size - msg->leadin_ignore - msg->tailout_ignore));
+
+	/*
+	 * Move the data over from the other side.
+	 */
+	if ((XPNET_VERSION_MINOR(msg->version) == 1) &&
+	    (msg->embedded_bytes != 0)) {
+		dev_dbg(xpnet, "copying embedded message. memcpy(0x%p, 0x%p, "
+			"%lu)\n", skb->data, &msg->data,
+			(size_t)msg->embedded_bytes);
+
+		skb_copy_to_linear_data(skb, &msg->data,
+					(size_t)msg->embedded_bytes);
+	} else {
+		dst = (void *)((u64)skb->data & ~(L1_CACHE_BYTES - 1));
+		dev_dbg(xpnet, "transferring buffer to the skb->data area;\n\t"
+			"xp_remote_memcpy(0x%p, 0x%p, %hu)\n", dst,
+					  (void *)msg->buf_pa, msg->size);
+
+		ret = xp_remote_memcpy(xp_pa(dst), msg->buf_pa, msg->size);
+		if (ret != xpSuccess) {
+			/*
+			 * !!! Need better way of cleaning skb.  Currently skb
+			 * !!! appears in_use and we can't just call
+			 * !!! dev_kfree_skb.
+			 */
+			dev_err(xpnet, "xp_remote_memcpy(0x%p, 0x%p, 0x%hx) "
+				"returned error=0x%x\n", dst,
+				(void *)msg->buf_pa, msg->size, ret);
+
+			xpc_received(partid, channel, (void *)msg);
+
+			xpnet_device->stats.rx_errors++;
+
+			return;
+		}
+	}
+
+	dev_dbg(xpnet, "<skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n", (void *)skb->head,
+		(void *)skb->data, skb_tail_pointer(skb), skb_end_pointer(skb),
+		skb->len);
+
+	skb->protocol = eth_type_trans(skb, xpnet_device);
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+	dev_dbg(xpnet, "passing skb to network layer\n"
+		"\tskb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n",
+		(void *)skb->head, (void *)skb->data, skb_tail_pointer(skb),
+		skb_end_pointer(skb), skb->len);
+
+	xpnet_device->stats.rx_packets++;
+	xpnet_device->stats.rx_bytes += skb->len + ETH_HLEN;
+
+	netif_rx_ni(skb);
+	xpc_received(partid, channel, (void *)msg);
+}
+
+/*
+ * This is the handler which XPC calls during any sort of change in
+ * state or message reception on a connection.
+ */
+static void
+xpnet_connection_activity(enum xp_retval reason, short partid, int channel,
+			  void *data, void *key)
+{
+	DBUG_ON(partid < 0 || partid >= xp_max_npartitions);
+	DBUG_ON(channel != XPC_NET_CHANNEL);
+
+	switch (reason) {
+	case xpMsgReceived:	/* message received */
+		DBUG_ON(data == NULL);
+
+		xpnet_receive(partid, channel, (struct xpnet_message *)data);
+		break;
+
+	case xpConnected:	/* connection completed to a partition */
+		spin_lock_bh(&xpnet_broadcast_lock);
+		__set_bit(partid, xpnet_broadcast_partitions);
+		spin_unlock_bh(&xpnet_broadcast_lock);
+
+		netif_carrier_on(xpnet_device);
+
+		dev_dbg(xpnet, "%s connected to partition %d\n",
+			xpnet_device->name, partid);
+		break;
+
+	default:
+		spin_lock_bh(&xpnet_broadcast_lock);
+		__clear_bit(partid, xpnet_broadcast_partitions);
+		spin_unlock_bh(&xpnet_broadcast_lock);
+
+		if (bitmap_empty((unsigned long *)xpnet_broadcast_partitions,
+				 xp_max_npartitions)) {
+			netif_carrier_off(xpnet_device);
+		}
+
+		dev_dbg(xpnet, "%s disconnected from partition %d\n",
+			xpnet_device->name, partid);
+		break;
+	}
+}
+
+static int
+xpnet_dev_open(struct net_device *dev)
+{
+	enum xp_retval ret;
+
+	dev_dbg(xpnet, "calling xpc_connect(%d, 0x%p, NULL, %ld, %ld, %ld, "
+		"%ld)\n", XPC_NET_CHANNEL, xpnet_connection_activity,
+		(unsigned long)XPNET_MSG_SIZE,
+		(unsigned long)XPNET_MSG_NENTRIES,
+		(unsigned long)XPNET_MAX_KTHREADS,
+		(unsigned long)XPNET_MAX_IDLE_KTHREADS);
+
+	ret = xpc_connect(XPC_NET_CHANNEL, xpnet_connection_activity, NULL,
+			  XPNET_MSG_SIZE, XPNET_MSG_NENTRIES,
+			  XPNET_MAX_KTHREADS, XPNET_MAX_IDLE_KTHREADS);
+	if (ret != xpSuccess) {
+		dev_err(xpnet, "ifconfig up of %s failed on XPC connect, "
+			"ret=%d\n", dev->name, ret);
+
+		return -ENOMEM;
+	}
+
+	dev_dbg(xpnet, "ifconfig up of %s; XPC connected\n", dev->name);
+
+	return 0;
+}
+
+static int
+xpnet_dev_stop(struct net_device *dev)
+{
+	xpc_disconnect(XPC_NET_CHANNEL);
+
+	dev_dbg(xpnet, "ifconfig down of %s; XPC disconnected\n", dev->name);
+
+	return 0;
+}
+
+static int
+xpnet_dev_change_mtu(struct net_device *dev, int new_mtu)
+{
+	/* 68 comes from min TCP+IP+MAC header */
+	if ((new_mtu < 68) || (new_mtu > XPNET_MAX_MTU)) {
+		dev_err(xpnet, "ifconfig %s mtu %d failed; value must be "
+			"between 68 and %ld\n", dev->name, new_mtu,
+			XPNET_MAX_MTU);
+		return -EINVAL;
+	}
+
+	dev->mtu = new_mtu;
+	dev_dbg(xpnet, "ifconfig %s mtu set to %d\n", dev->name, new_mtu);
+	return 0;
+}
+
+/*
+ * Notification that the other end has received the message and
+ * DMA'd the skb information.  At this point, they are done with
+ * our side.  When all recipients are done processing, we
+ * release the skb and then release our pending message structure.
+ */
+static void
+xpnet_send_completed(enum xp_retval reason, short partid, int channel,
+		     void *__qm)
+{
+	struct xpnet_pending_msg *queued_msg = (struct xpnet_pending_msg *)__qm;
+
+	DBUG_ON(queued_msg == NULL);
+
+	dev_dbg(xpnet, "message to %d notified with reason %d\n",
+		partid, reason);
+
+	if (atomic_dec_return(&queued_msg->use_count) == 0) {
+		dev_dbg(xpnet, "all acks for skb->head=-x%p\n",
+			(void *)queued_msg->skb->head);
+
+		dev_kfree_skb_any(queued_msg->skb);
+		kfree(queued_msg);
+	}
+}
+
+static void
+xpnet_send(struct sk_buff *skb, struct xpnet_pending_msg *queued_msg,
+	   u64 start_addr, u64 end_addr, u16 embedded_bytes, int dest_partid)
+{
+	u8 msg_buffer[XPNET_MSG_SIZE];
+	struct xpnet_message *msg = (struct xpnet_message *)&msg_buffer;
+	u16 msg_size = sizeof(struct xpnet_message);
+	enum xp_retval ret;
+
+	msg->embedded_bytes = embedded_bytes;
+	if (unlikely(embedded_bytes != 0)) {
+		msg->version = XPNET_VERSION_EMBED;
+		dev_dbg(xpnet, "calling memcpy(0x%p, 0x%p, 0x%lx)\n",
+			&msg->data, skb->data, (size_t)embedded_bytes);
+		skb_copy_from_linear_data(skb, &msg->data,
+					  (size_t)embedded_bytes);
+		msg_size += embedded_bytes - 1;
+	} else {
+		msg->version = XPNET_VERSION;
+	}
+	msg->magic = XPNET_MAGIC;
+	msg->size = end_addr - start_addr;
+	msg->leadin_ignore = (u64)skb->data - start_addr;
+	msg->tailout_ignore = end_addr - (u64)skb_tail_pointer(skb);
+	msg->buf_pa = xp_pa((void *)start_addr);
+
+	dev_dbg(xpnet, "sending XPC message to %d:%d\n"
+		"msg->buf_pa=0x%lx, msg->size=%u, "
+		"msg->leadin_ignore=%u, msg->tailout_ignore=%u\n",
+		dest_partid, XPC_NET_CHANNEL, msg->buf_pa, msg->size,
+		msg->leadin_ignore, msg->tailout_ignore);
+
+	atomic_inc(&queued_msg->use_count);
+
+	ret = xpc_send_notify(dest_partid, XPC_NET_CHANNEL, XPC_NOWAIT, msg,
+			      msg_size, xpnet_send_completed, queued_msg);
+	if (unlikely(ret != xpSuccess))
+		atomic_dec(&queued_msg->use_count);
+}
+
+/*
+ * Network layer has formatted a packet (skb) and is ready to place it
+ * "on the wire".  Prepare and send an xpnet_message to all partitions
+ * which have connected with us and are targets of this packet.
+ *
+ * MAC-NOTE:  For the XPNET driver, the MAC address contains the
+ * destination partid.  If the destination partid octets are 0xffff,
+ * this packet is to be broadcast to all connected partitions.
+ */
+static int
+xpnet_dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct xpnet_pending_msg *queued_msg;
+	u64 start_addr, end_addr;
+	short dest_partid;
+	u16 embedded_bytes = 0;
+
+	dev_dbg(xpnet, ">skb->head=0x%p skb->data=0x%p skb->tail=0x%p "
+		"skb->end=0x%p skb->len=%d\n", (void *)skb->head,
+		(void *)skb->data, skb_tail_pointer(skb), skb_end_pointer(skb),
+		skb->len);
+
+	if (skb->data[0] == 0x33) {
+		dev_kfree_skb(skb);
+		return NETDEV_TX_OK;	/* nothing needed to be done */
+	}
+
+	/*
+	 * The xpnet_pending_msg tracks how many outstanding
+	 * xpc_send_notifies are relying on this skb.  When none
+	 * remain, release the skb.
+	 */
+	queued_msg = kmalloc(sizeof(struct xpnet_pending_msg), GFP_ATOMIC);
+	if (queued_msg == NULL) {
+		dev_warn(xpnet, "failed to kmalloc %ld bytes; dropping "
+			 "packet\n", sizeof(struct xpnet_pending_msg));
+
+		dev->stats.tx_errors++;
+		dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	/* get the beginning of the first cacheline and end of last */
+	start_addr = ((u64)skb->data & ~(L1_CACHE_BYTES - 1));
+	end_addr = L1_CACHE_ALIGN((u64)skb_tail_pointer(skb));
+
+	/* calculate how many bytes to embed in the XPC message */
+	if (unlikely(skb->len <= XPNET_MSG_DATA_MAX)) {
+		/* skb->data does fit so embed */
+		embedded_bytes = skb->len;
+	}
+
+	/*
+	 * Since the send occurs asynchronously, we set the count to one
+	 * and begin sending.  Any sends that happen to complete before
+	 * we are done sending will not free the skb.  We will be left
+	 * with that task during exit.  This also handles the case of
+	 * a packet destined for a partition which is no longer up.
+	 */
+	atomic_set(&queued_msg->use_count, 1);
+	queued_msg->skb = skb;
+
+	if (skb->data[0] == 0xff) {
+		/* we are being asked to broadcast to all partitions */
+		for_each_set_bit(dest_partid, xpnet_broadcast_partitions,
+			     xp_max_npartitions) {
+
+			xpnet_send(skb, queued_msg, start_addr, end_addr,
+				   embedded_bytes, dest_partid);
+		}
+	} else {
+		dest_partid = (short)skb->data[XPNET_PARTID_OCTET + 1];
+		dest_partid |= (short)skb->data[XPNET_PARTID_OCTET + 0] << 8;
+
+		if (dest_partid >= 0 &&
+		    dest_partid < xp_max_npartitions &&
+		    test_bit(dest_partid, xpnet_broadcast_partitions) != 0) {
+
+			xpnet_send(skb, queued_msg, start_addr, end_addr,
+				   embedded_bytes, dest_partid);
+		}
+	}
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += skb->len;
+
+	if (atomic_dec_return(&queued_msg->use_count) == 0) {
+		dev_kfree_skb(skb);
+		kfree(queued_msg);
+	}
+
+	return NETDEV_TX_OK;
+}
+
+/*
+ * Deal with transmit timeouts coming from the network layer.
+ */
+static void
+xpnet_dev_tx_timeout(struct net_device *dev)
+{
+	dev->stats.tx_errors++;
+}
+
+static const struct net_device_ops xpnet_netdev_ops = {
+	.ndo_open		= xpnet_dev_open,
+	.ndo_stop		= xpnet_dev_stop,
+	.ndo_start_xmit		= xpnet_dev_hard_start_xmit,
+	.ndo_change_mtu		= xpnet_dev_change_mtu,
+	.ndo_tx_timeout		= xpnet_dev_tx_timeout,
+	.ndo_set_mac_address 	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+};
+
+static int __init
+xpnet_init(void)
+{
+	int result;
+
+	if (!is_shub() && !is_uv())
+		return -ENODEV;
+
+	dev_info(xpnet, "registering network device %s\n", XPNET_DEVICE_NAME);
+
+	xpnet_broadcast_partitions = kzalloc(BITS_TO_LONGS(xp_max_npartitions) *
+					     sizeof(long), GFP_KERNEL);
+	if (xpnet_broadcast_partitions == NULL)
+		return -ENOMEM;
+
+	/*
+	 * use ether_setup() to init the majority of our device
+	 * structure and then override the necessary pieces.
+	 */
+	xpnet_device = alloc_netdev(0, XPNET_DEVICE_NAME, ether_setup);
+	if (xpnet_device == NULL) {
+		kfree(xpnet_broadcast_partitions);
+		return -ENOMEM;
+	}
+
+	netif_carrier_off(xpnet_device);
+
+	xpnet_device->netdev_ops = &xpnet_netdev_ops;
+	xpnet_device->mtu = XPNET_DEF_MTU;
+
+	/*
+	 * Multicast assumes the LSB of the first octet is set for multicast
+	 * MAC addresses.  We chose the first octet of the MAC to be unlikely
+	 * to collide with any vendor's officially issued MAC.
+	 */
+	xpnet_device->dev_addr[0] = 0x02;     /* locally administered, no OUI */
+
+	xpnet_device->dev_addr[XPNET_PARTID_OCTET + 1] = xp_partition_id;
+	xpnet_device->dev_addr[XPNET_PARTID_OCTET + 0] = (xp_partition_id >> 8);
+
+	/*
+	 * ether_setup() sets this to a multicast device.  We are
+	 * really not supporting multicast at this time.
+	 */
+	xpnet_device->flags &= ~IFF_MULTICAST;
+
+	/*
+	 * No need to checksum as it is a DMA transfer.  The BTE will
+	 * report an error if the data is not retrievable and the
+	 * packet will be dropped.
+	 */
+	xpnet_device->features = NETIF_F_HW_CSUM;
+
+	result = register_netdev(xpnet_device);
+	if (result != 0) {
+		free_netdev(xpnet_device);
+		kfree(xpnet_broadcast_partitions);
+	}
+
+	return result;
+}
+
+module_init(xpnet_init);
+
+static void __exit
+xpnet_exit(void)
+{
+	dev_info(xpnet, "unregistering network device %s\n",
+		 xpnet_device[0].name);
+
+	unregister_netdev(xpnet_device);
+	free_netdev(xpnet_device);
+	kfree(xpnet_broadcast_partitions);
+}
+
+module_exit(xpnet_exit);
+
+MODULE_AUTHOR("Silicon Graphics, Inc.");
+MODULE_DESCRIPTION("Cross Partition Network adapter (XPNET)");
+MODULE_LICENSE("GPL");