Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6: (26 commits)
  [SPARC64]: Fix UP build.
  [SPARC64]: dr-cpu unconfigure support.
  [SERIAL]: Fix console write locking in sparc drivers.
  [SPARC64]: Give more accurate errors in dr_cpu_configure().
  [SPARC64]: Clear cpu_{core,sibling}_map[] in smp_fill_in_sib_core_maps()
  [SPARC64]: Fix leak when DR added cpu does not bootup.
  [SPARC64]: Add ->set_affinity IRQ handlers.
  [SPARC64]: Process dr-cpu events in a kthread instead of workqueue.
  [SPARC64]: More sensible udelay implementation.
  [SPARC64]: SMP build fixes.
  [SPARC64]: mdesc.c needs linux/mm.h
  [SPARC64]: Fix build regressions added by dr-cpu changes.
  [SPARC64]: Unconditionally register vio_bus_type.
  [SPARC64]: Initial LDOM cpu hotplug support.
  [SPARC64]: Fix setting of variables in LDOM guest.
  [SPARC64]: Fix MD property lifetime bugs.
  [SPARC64]: Abstract out mdesc accesses for better MD update handling.
  [SPARC64]: Use more mearningful names for IRQ registry.
  [SPARC64]: Initial domain-services driver.
  [SPARC64]: Export powerd facilities for external entities.
  ...

arch/sparc64/Kconfig

@@ -108,6 +108,15 @@ config SECCOMP
 
 source kernel/Kconfig.hz
 
+config HOTPLUG_CPU
+	bool "Support for hot-pluggable CPUs"
+	depends on SMP
+	select HOTPLUG
+	---help---
+	  Say Y here to experiment with turning CPUs off and on.  CPUs
+	  can be controlled through /sys/devices/system/cpu/cpu#.
+	  Say N if you want to disable CPU hotplug.
+
 source "init/Kconfig"
 
 config SYSVIPC_COMPAT
@@ -305,6 +314,12 @@ config SUN_IO
 	bool
 	default y
 
+config SUN_LDOMS
+	bool "Sun Logical Domains support"
+	help
+	  Say Y here is you want to support virtual devices via
+	  Logical Domains.
+
 config PCI
 	bool "PCI support"
 	select ARCH_SUPPORTS_MSI

arch/sparc64/kernel/Makefile

@@ -18,7 +18,7 @@ obj-$(CONFIG_STACKTRACE) += stacktrace.o
 obj-$(CONFIG_PCI)	 += ebus.o isa.o pci_common.o pci_iommu.o \
 			    pci_psycho.o pci_sabre.o pci_schizo.o \
 			    pci_sun4v.o pci_sun4v_asm.o pci_fire.o
-obj-$(CONFIG_SMP)	 += smp.o trampoline.o
+obj-$(CONFIG_SMP)	 += smp.o trampoline.o hvtramp.o
 obj-$(CONFIG_SPARC32_COMPAT) += sys32.o sys_sparc32.o signal32.o
 obj-$(CONFIG_BINFMT_ELF32) += binfmt_elf32.o
 obj-$(CONFIG_BINFMT_AOUT32) += binfmt_aout32.o
@@ -26,6 +26,7 @@ obj-$(CONFIG_MODULES) += module.o
 obj-$(CONFIG_US3_FREQ) += us3_cpufreq.o
 obj-$(CONFIG_US2E_FREQ) += us2e_cpufreq.o
 obj-$(CONFIG_KPROBES) += kprobes.o
+obj-$(CONFIG_SUN_LDOMS) += ldc.o vio.o viohs.o ds.o
 obj-$(CONFIG_AUDIT) += audit.o
 obj-$(CONFIG_AUDIT)$(CONFIG_SPARC32_COMPAT) += compat_audit.o
 obj-y += $(obj-yy)

arch/sparc64/kernel/ds.c

@@ -0,0 +1,1158 @@
+/* ds.c: Domain Services driver for Logical Domains
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/sched.h>
+#include <linux/delay.h>
+#include <linux/mutex.h>
+#include <linux/kthread.h>
+#include <linux/cpu.h>
+
+#include <asm/ldc.h>
+#include <asm/vio.h>
+#include <asm/power.h>
+#include <asm/mdesc.h>
+#include <asm/head.h>
+#include <asm/irq.h>
+
+#define DRV_MODULE_NAME		"ds"
+#define PFX DRV_MODULE_NAME	": "
+#define DRV_MODULE_VERSION	"1.0"
+#define DRV_MODULE_RELDATE	"Jul 11, 2007"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+MODULE_AUTHOR("David S. Miller (davem@davemloft.net)");
+MODULE_DESCRIPTION("Sun LDOM domain services driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+struct ds_msg_tag {
+	__u32			type;
+#define DS_INIT_REQ		0x00
+#define DS_INIT_ACK		0x01
+#define DS_INIT_NACK		0x02
+#define DS_REG_REQ		0x03
+#define DS_REG_ACK		0x04
+#define DS_REG_NACK		0x05
+#define DS_UNREG_REQ		0x06
+#define DS_UNREG_ACK		0x07
+#define DS_UNREG_NACK		0x08
+#define DS_DATA			0x09
+#define DS_NACK			0x0a
+
+	__u32			len;
+};
+
+/* Result codes */
+#define DS_OK			0x00
+#define DS_REG_VER_NACK		0x01
+#define DS_REG_DUP		0x02
+#define DS_INV_HDL		0x03
+#define DS_TYPE_UNKNOWN		0x04
+
+struct ds_version {
+	__u16			major;
+	__u16			minor;
+};
+
+struct ds_ver_req {
+	struct ds_msg_tag	tag;
+	struct ds_version	ver;
+};
+
+struct ds_ver_ack {
+	struct ds_msg_tag	tag;
+	__u16			minor;
+};
+
+struct ds_ver_nack {
+	struct ds_msg_tag	tag;
+	__u16			major;
+};
+
+struct ds_reg_req {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+	__u16			major;
+	__u16			minor;
+	char			svc_id[0];
+};
+
+struct ds_reg_ack {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+	__u16			minor;
+};
+
+struct ds_reg_nack {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+	__u16			major;
+};
+
+struct ds_unreg_req {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+};
+
+struct ds_unreg_ack {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+};
+
+struct ds_unreg_nack {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+};
+
+struct ds_data {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+};
+
+struct ds_data_nack {
+	struct ds_msg_tag	tag;
+	__u64			handle;
+	__u64			result;
+};
+
+struct ds_cap_state {
+	__u64			handle;
+
+	void			(*data)(struct ldc_channel *lp,
+					struct ds_cap_state *cp,
+					void *buf, int len);
+
+	const char		*service_id;
+
+	u8			state;
+#define CAP_STATE_UNKNOWN	0x00
+#define CAP_STATE_REG_SENT	0x01
+#define CAP_STATE_REGISTERED	0x02
+};
+
+static void md_update_data(struct ldc_channel *lp, struct ds_cap_state *cp,
+			   void *buf, int len);
+static void domain_shutdown_data(struct ldc_channel *lp,
+				 struct ds_cap_state *cp,
+				 void *buf, int len);
+static void domain_panic_data(struct ldc_channel *lp,
+			      struct ds_cap_state *cp,
+			      void *buf, int len);
+#ifdef CONFIG_HOTPLUG_CPU
+static void dr_cpu_data(struct ldc_channel *lp,
+			struct ds_cap_state *cp,
+			void *buf, int len);
+#endif
+static void ds_pri_data(struct ldc_channel *lp,
+			struct ds_cap_state *cp,
+			void *buf, int len);
+static void ds_var_data(struct ldc_channel *lp,
+			struct ds_cap_state *cp,
+			void *buf, int len);
+
+struct ds_cap_state ds_states[] = {
+	{
+		.service_id	= "md-update",
+		.data		= md_update_data,
+	},
+	{
+		.service_id	= "domain-shutdown",
+		.data		= domain_shutdown_data,
+	},
+	{
+		.service_id	= "domain-panic",
+		.data		= domain_panic_data,
+	},
+#ifdef CONFIG_HOTPLUG_CPU
+	{
+		.service_id	= "dr-cpu",
+		.data		= dr_cpu_data,
+	},
+#endif
+	{
+		.service_id	= "pri",
+		.data		= ds_pri_data,
+	},
+	{
+		.service_id	= "var-config",
+		.data		= ds_var_data,
+	},
+	{
+		.service_id	= "var-config-backup",
+		.data		= ds_var_data,
+	},
+};
+
+static DEFINE_SPINLOCK(ds_lock);
+
+struct ds_info {
+	struct ldc_channel	*lp;
+	u8			hs_state;
+#define DS_HS_START		0x01
+#define DS_HS_DONE		0x02
+
+	void			*rcv_buf;
+	int			rcv_buf_len;
+};
+
+static struct ds_info *ds_info;
+
+static struct ds_cap_state *find_cap(u64 handle)
+{
+	unsigned int index = handle >> 32;
+
+	if (index >= ARRAY_SIZE(ds_states))
+		return NULL;
+	return &ds_states[index];
+}
+
+static struct ds_cap_state *find_cap_by_string(const char *name)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ds_states); i++) {
+		if (strcmp(ds_states[i].service_id, name))
+			continue;
+
+		return &ds_states[i];
+	}
+	return NULL;
+}
+
+static int ds_send(struct ldc_channel *lp, void *data, int len)
+{
+	int err, limit = 1000;
+
+	err = -EINVAL;
+	while (limit-- > 0) {
+		err = ldc_write(lp, data, len);
+		if (!err || (err != -EAGAIN))
+			break;
+		udelay(1);
+	}
+
+	return err;
+}
+
+struct ds_md_update_req {
+	__u64				req_num;
+};
+
+struct ds_md_update_res {
+	__u64				req_num;
+	__u32				result;
+};
+
+static void md_update_data(struct ldc_channel *lp,
+			   struct ds_cap_state *dp,
+			   void *buf, int len)
+{
+	struct ds_data *dpkt = buf;
+	struct ds_md_update_req *rp;
+	struct {
+		struct ds_data		data;
+		struct ds_md_update_res	res;
+	} pkt;
+
+	rp = (struct ds_md_update_req *) (dpkt + 1);
+
+	printk(KERN_INFO PFX "Machine description update.\n");
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.data.tag.type = DS_DATA;
+	pkt.data.tag.len = sizeof(pkt) - sizeof(struct ds_msg_tag);
+	pkt.data.handle = dp->handle;
+	pkt.res.req_num = rp->req_num;
+	pkt.res.result = DS_OK;
+
+	ds_send(lp, &pkt, sizeof(pkt));
+
+	mdesc_update();
+}
+
+struct ds_shutdown_req {
+	__u64				req_num;
+	__u32				ms_delay;
+};
+
+struct ds_shutdown_res {
+	__u64				req_num;
+	__u32				result;
+	char				reason[1];
+};
+
+static void domain_shutdown_data(struct ldc_channel *lp,
+				 struct ds_cap_state *dp,
+				 void *buf, int len)
+{
+	struct ds_data *dpkt = buf;
+	struct ds_shutdown_req *rp;
+	struct {
+		struct ds_data		data;
+		struct ds_shutdown_res	res;
+	} pkt;
+
+	rp = (struct ds_shutdown_req *) (dpkt + 1);
+
+	printk(KERN_ALERT PFX "Shutdown request from "
+	       "LDOM manager received.\n");
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.data.tag.type = DS_DATA;
+	pkt.data.tag.len = sizeof(pkt) - sizeof(struct ds_msg_tag);
+	pkt.data.handle = dp->handle;
+	pkt.res.req_num = rp->req_num;
+	pkt.res.result = DS_OK;
+	pkt.res.reason[0] = 0;
+
+	ds_send(lp, &pkt, sizeof(pkt));
+
+	wake_up_powerd();
+}
+
+struct ds_panic_req {
+	__u64				req_num;
+};
+
+struct ds_panic_res {
+	__u64				req_num;
+	__u32				result;
+	char				reason[1];
+};
+
+static void domain_panic_data(struct ldc_channel *lp,
+			      struct ds_cap_state *dp,
+			      void *buf, int len)
+{
+	struct ds_data *dpkt = buf;
+	struct ds_panic_req *rp;
+	struct {
+		struct ds_data		data;
+		struct ds_panic_res	res;
+	} pkt;
+
+	rp = (struct ds_panic_req *) (dpkt + 1);
+
+	printk(KERN_ALERT PFX "Panic request from "
+	       "LDOM manager received.\n");
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.data.tag.type = DS_DATA;
+	pkt.data.tag.len = sizeof(pkt) - sizeof(struct ds_msg_tag);
+	pkt.data.handle = dp->handle;
+	pkt.res.req_num = rp->req_num;
+	pkt.res.result = DS_OK;
+	pkt.res.reason[0] = 0;
+
+	ds_send(lp, &pkt, sizeof(pkt));
+
+	panic("PANIC requested by LDOM manager.");
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+struct dr_cpu_tag {
+	__u64				req_num;
+	__u32				type;
+#define DR_CPU_CONFIGURE		0x43
+#define DR_CPU_UNCONFIGURE		0x55
+#define DR_CPU_FORCE_UNCONFIGURE	0x46
+#define DR_CPU_STATUS			0x53
+
+/* Responses */
+#define DR_CPU_OK			0x6f
+#define DR_CPU_ERROR			0x65
+
+	__u32				num_records;
+};
+
+struct dr_cpu_resp_entry {
+	__u32				cpu;
+	__u32				result;
+#define DR_CPU_RES_OK			0x00
+#define DR_CPU_RES_FAILURE		0x01
+#define DR_CPU_RES_BLOCKED		0x02
+#define DR_CPU_RES_CPU_NOT_RESPONDING	0x03
+#define DR_CPU_RES_NOT_IN_MD		0x04
+
+	__u32				stat;
+#define DR_CPU_STAT_NOT_PRESENT		0x00
+#define DR_CPU_STAT_UNCONFIGURED	0x01
+#define DR_CPU_STAT_CONFIGURED		0x02
+
+	__u32				str_off;
+};
+
+/* DR cpu requests get queued onto the work list by the
+ * dr_cpu_data() callback.  The list is protected by
+ * ds_lock, and processed by dr_cpu_process() in order.
+ */
+static LIST_HEAD(dr_cpu_work_list);
+static DECLARE_WAIT_QUEUE_HEAD(dr_cpu_wait);
+
+struct dr_cpu_queue_entry {
+	struct list_head		list;
+	char				req[0];
+};
+
+static void __dr_cpu_send_error(struct ds_cap_state *cp, struct ds_data *data)
+{
+	struct dr_cpu_tag *tag = (struct dr_cpu_tag *) (data + 1);
+	struct ds_info *dp = ds_info;
+	struct {
+		struct ds_data		data;
+		struct dr_cpu_tag	tag;
+	} pkt;
+	int msg_len;
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.data.tag.type = DS_DATA;
+	pkt.data.handle = cp->handle;
+	pkt.tag.req_num = tag->req_num;
+	pkt.tag.type = DR_CPU_ERROR;
+	pkt.tag.num_records = 0;
+
+	msg_len = (sizeof(struct ds_data) +
+		   sizeof(struct dr_cpu_tag));
+
+	pkt.data.tag.len = msg_len - sizeof(struct ds_msg_tag);
+
+	ds_send(dp->lp, &pkt, msg_len);
+}
+
+static void dr_cpu_send_error(struct ds_cap_state *cp, struct ds_data *data)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&ds_lock, flags);
+	__dr_cpu_send_error(cp, data);
+	spin_unlock_irqrestore(&ds_lock, flags);
+}
+
+#define CPU_SENTINEL	0xffffffff
+
+static void purge_dups(u32 *list, u32 num_ents)
+{
+	unsigned int i;
+
+	for (i = 0; i < num_ents; i++) {
+		u32 cpu = list[i];
+		unsigned int j;
+
+		if (cpu == CPU_SENTINEL)
+			continue;
+
+		for (j = i + 1; j < num_ents; j++) {
+			if (list[j] == cpu)
+				list[j] = CPU_SENTINEL;
+		}
+	}
+}
+
+static int dr_cpu_size_response(int ncpus)
+{
+	return (sizeof(struct ds_data) +
+		sizeof(struct dr_cpu_tag) +
+		(sizeof(struct dr_cpu_resp_entry) * ncpus));
+}
+
+static void dr_cpu_init_response(struct ds_data *resp, u64 req_num,
+				 u64 handle, int resp_len, int ncpus,
+				 cpumask_t *mask, u32 default_stat)
+{
+	struct dr_cpu_resp_entry *ent;
+	struct dr_cpu_tag *tag;
+	int i, cpu;
+
+	tag = (struct dr_cpu_tag *) (resp + 1);
+	ent = (struct dr_cpu_resp_entry *) (tag + 1);
+
+	resp->tag.type = DS_DATA;
+	resp->tag.len = resp_len - sizeof(struct ds_msg_tag);
+	resp->handle = handle;
+	tag->req_num = req_num;
+	tag->type = DR_CPU_OK;
+	tag->num_records = ncpus;
+
+	i = 0;
+	for_each_cpu_mask(cpu, *mask) {
+		ent[i].cpu = cpu;
+		ent[i].result = DR_CPU_RES_OK;
+		ent[i].stat = default_stat;
+		i++;
+	}
+	BUG_ON(i != ncpus);
+}
+
+static void dr_cpu_mark(struct ds_data *resp, int cpu, int ncpus,
+			u32 res, u32 stat)
+{
+	struct dr_cpu_resp_entry *ent;
+	struct dr_cpu_tag *tag;
+	int i;
+
+	tag = (struct dr_cpu_tag *) (resp + 1);
+	ent = (struct dr_cpu_resp_entry *) (tag + 1);
+
+	for (i = 0; i < ncpus; i++) {
+		if (ent[i].cpu != cpu)
+			continue;
+		ent[i].result = res;
+		ent[i].stat = stat;
+		break;
+	}
+}
+
+static int dr_cpu_configure(struct ds_cap_state *cp, u64 req_num,
+			    cpumask_t *mask)
+{
+	struct ds_data *resp;
+	int resp_len, ncpus, cpu;
+	unsigned long flags;
+
+	ncpus = cpus_weight(*mask);
+	resp_len = dr_cpu_size_response(ncpus);
+	resp = kzalloc(resp_len, GFP_KERNEL);
+	if (!resp)
+		return -ENOMEM;
+
+	dr_cpu_init_response(resp, req_num, cp->handle,
+			     resp_len, ncpus, mask,
+			     DR_CPU_STAT_CONFIGURED);
+
+	mdesc_fill_in_cpu_data(*mask);
+
+	for_each_cpu_mask(cpu, *mask) {
+		int err;
+
+		printk(KERN_INFO PFX "Starting cpu %d...\n", cpu);
+		err = cpu_up(cpu);
+		if (err) {
+			__u32 res = DR_CPU_RES_FAILURE;
+			__u32 stat = DR_CPU_STAT_UNCONFIGURED;
+
+			if (!cpu_present(cpu)) {
+				/* CPU not present in MD */
+				res = DR_CPU_RES_NOT_IN_MD;
+				stat = DR_CPU_STAT_NOT_PRESENT;
+			} else if (err == -ENODEV) {
+				/* CPU did not call in successfully */
+				res = DR_CPU_RES_CPU_NOT_RESPONDING;
+			}
+
+			printk(KERN_INFO PFX "CPU startup failed err=%d\n",
+			       err);
+			dr_cpu_mark(resp, cpu, ncpus, res, stat);
+		}
+	}
+
+	spin_lock_irqsave(&ds_lock, flags);
+	ds_send(ds_info->lp, resp, resp_len);
+	spin_unlock_irqrestore(&ds_lock, flags);
+
+	kfree(resp);
+
+	/* Redistribute IRQs, taking into account the new cpus.  */
+	fixup_irqs();
+
+	return 0;
+}
+
+static int dr_cpu_unconfigure(struct ds_cap_state *cp, u64 req_num,
+			      cpumask_t *mask)
+{
+	struct ds_data *resp;
+	int resp_len, ncpus, cpu;
+	unsigned long flags;
+
+	ncpus = cpus_weight(*mask);
+	resp_len = dr_cpu_size_response(ncpus);
+	resp = kzalloc(resp_len, GFP_KERNEL);
+	if (!resp)
+		return -ENOMEM;
+
+	dr_cpu_init_response(resp, req_num, cp->handle,
+			     resp_len, ncpus, mask,
+			     DR_CPU_STAT_UNCONFIGURED);
+
+	for_each_cpu_mask(cpu, *mask) {
+		int err;
+
+		printk(KERN_INFO PFX "CPU[%d]: Shutting down cpu %d...\n",
+		       smp_processor_id(), cpu);
+		err = cpu_down(cpu);
+		if (err)
+			dr_cpu_mark(resp, cpu, ncpus,
+				    DR_CPU_RES_FAILURE,
+				    DR_CPU_STAT_CONFIGURED);
+	}
+
+	spin_lock_irqsave(&ds_lock, flags);
+	ds_send(ds_info->lp, resp, resp_len);
+	spin_unlock_irqrestore(&ds_lock, flags);
+
+	kfree(resp);
+
+	return 0;
+}
+
+static void process_dr_cpu_list(struct ds_cap_state *cp)
+{
+	struct dr_cpu_queue_entry *qp, *tmp;
+	unsigned long flags;
+	LIST_HEAD(todo);
+	cpumask_t mask;
+
+	spin_lock_irqsave(&ds_lock, flags);
+	list_splice(&dr_cpu_work_list, &todo);
+	INIT_LIST_HEAD(&dr_cpu_work_list);
+	spin_unlock_irqrestore(&ds_lock, flags);
+
+	list_for_each_entry_safe(qp, tmp, &todo, list) {
+		struct ds_data *data = (struct ds_data *) qp->req;
+		struct dr_cpu_tag *tag = (struct dr_cpu_tag *) (data + 1);
+		u32 *cpu_list = (u32 *) (tag + 1);
+		u64 req_num = tag->req_num;
+		unsigned int i;
+		int err;
+
+		switch (tag->type) {
+		case DR_CPU_CONFIGURE:
+		case DR_CPU_UNCONFIGURE:
+		case DR_CPU_FORCE_UNCONFIGURE:
+			break;
+
+		default:
+			dr_cpu_send_error(cp, data);
+			goto next;
+		}
+
+		purge_dups(cpu_list, tag->num_records);
+
+		cpus_clear(mask);
+		for (i = 0; i < tag->num_records; i++) {
+			if (cpu_list[i] == CPU_SENTINEL)
+				continue;
+
+			if (cpu_list[i] < NR_CPUS)
+				cpu_set(cpu_list[i], mask);
+		}
+
+		if (tag->type == DR_CPU_CONFIGURE)
+			err = dr_cpu_configure(cp, req_num, &mask);
+		else
+			err = dr_cpu_unconfigure(cp, req_num, &mask);
+
+		if (err)
+			dr_cpu_send_error(cp, data);
+
+next:
+		list_del(&qp->list);
+		kfree(qp);
+	}
+}
+
+static int dr_cpu_thread(void *__unused)
+{
+	struct ds_cap_state *cp;
+	DEFINE_WAIT(wait);
+
+	cp = find_cap_by_string("dr-cpu");
+
+	while (1) {
+		prepare_to_wait(&dr_cpu_wait, &wait, TASK_INTERRUPTIBLE);
+		if (list_empty(&dr_cpu_work_list))
+			schedule();
+		finish_wait(&dr_cpu_wait, &wait);
+
+		if (kthread_should_stop())
+			break;
+
+		process_dr_cpu_list(cp);
+	}
+
+	return 0;
+}
+
+static void dr_cpu_data(struct ldc_channel *lp,
+			struct ds_cap_state *dp,
+			void *buf, int len)
+{
+	struct dr_cpu_queue_entry *qp;
+	struct ds_data *dpkt = buf;
+	struct dr_cpu_tag *rp;
+
+	rp = (struct dr_cpu_tag *) (dpkt + 1);
+
+	qp = kmalloc(sizeof(struct dr_cpu_queue_entry) + len, GFP_ATOMIC);
+	if (!qp) {
+		struct ds_cap_state *cp;
+
+		cp = find_cap_by_string("dr-cpu");
+		__dr_cpu_send_error(cp, dpkt);
+	} else {
+		memcpy(&qp->req, buf, len);
+		list_add_tail(&qp->list, &dr_cpu_work_list);
+		wake_up(&dr_cpu_wait);
+	}
+}
+#endif
+
+struct ds_pri_msg {
+	__u64				req_num;
+	__u64				type;
+#define DS_PRI_REQUEST			0x00
+#define DS_PRI_DATA			0x01
+#define DS_PRI_UPDATE			0x02
+};
+
+static void ds_pri_data(struct ldc_channel *lp,
+			struct ds_cap_state *dp,
+			void *buf, int len)
+{
+	struct ds_data *dpkt = buf;
+	struct ds_pri_msg *rp;
+
+	rp = (struct ds_pri_msg *) (dpkt + 1);
+
+	printk(KERN_INFO PFX "PRI REQ [%lx:%lx], len=%d\n",
+	       rp->req_num, rp->type, len);
+}
+
+struct ds_var_hdr {
+	__u32				type;
+#define DS_VAR_SET_REQ			0x00
+#define DS_VAR_DELETE_REQ		0x01
+#define DS_VAR_SET_RESP			0x02
+#define DS_VAR_DELETE_RESP		0x03
+};
+
+struct ds_var_set_msg {
+	struct ds_var_hdr		hdr;
+	char				name_and_value[0];
+};
+
+struct ds_var_delete_msg {
+	struct ds_var_hdr		hdr;
+	char				name[0];
+};
+
+struct ds_var_resp {
+	struct ds_var_hdr		hdr;
+	__u32				result;
+#define DS_VAR_SUCCESS			0x00
+#define DS_VAR_NO_SPACE			0x01
+#define DS_VAR_INVALID_VAR		0x02
+#define DS_VAR_INVALID_VAL		0x03
+#define DS_VAR_NOT_PRESENT		0x04
+};
+
+static DEFINE_MUTEX(ds_var_mutex);
+static int ds_var_doorbell;
+static int ds_var_response;
+
+static void ds_var_data(struct ldc_channel *lp,
+			struct ds_cap_state *dp,
+			void *buf, int len)
+{
+	struct ds_data *dpkt = buf;
+	struct ds_var_resp *rp;
+
+	rp = (struct ds_var_resp *) (dpkt + 1);
+
+	if (rp->hdr.type != DS_VAR_SET_RESP &&
+	    rp->hdr.type != DS_VAR_DELETE_RESP)
+		return;
+
+	ds_var_response = rp->result;
+	wmb();
+	ds_var_doorbell = 1;
+}
+
+void ldom_set_var(const char *var, const char *value)
+{
+	struct ds_info *dp = ds_info;
+	struct ds_cap_state *cp;
+
+	cp = find_cap_by_string("var-config");
+	if (cp->state != CAP_STATE_REGISTERED)
+		cp = find_cap_by_string("var-config-backup");
+
+	if (cp->state == CAP_STATE_REGISTERED) {
+		union {
+			struct {
+				struct ds_data		data;
+				struct ds_var_set_msg	msg;
+			} header;
+			char			all[512];
+		} pkt;
+		unsigned long flags;
+		char  *base, *p;
+		int msg_len, loops;
+
+		memset(&pkt, 0, sizeof(pkt));
+		pkt.header.data.tag.type = DS_DATA;
+		pkt.header.data.handle = cp->handle;
+		pkt.header.msg.hdr.type = DS_VAR_SET_REQ;
+		base = p = &pkt.header.msg.name_and_value[0];
+		strcpy(p, var);
+		p += strlen(var) + 1;
+		strcpy(p, value);
+		p += strlen(value) + 1;
+
+		msg_len = (sizeof(struct ds_data) +
+			   sizeof(struct ds_var_set_msg) +
+			   (p - base));
+		msg_len = (msg_len + 3) & ~3;
+		pkt.header.data.tag.len = msg_len - sizeof(struct ds_msg_tag);
+
+		mutex_lock(&ds_var_mutex);
+
+		spin_lock_irqsave(&ds_lock, flags);
+		ds_var_doorbell = 0;
+		ds_var_response = -1;
+
+		ds_send(dp->lp, &pkt, msg_len);
+		spin_unlock_irqrestore(&ds_lock, flags);
+
+		loops = 1000;
+		while (ds_var_doorbell == 0) {
+			if (loops-- < 0)
+				break;
+			barrier();
+			udelay(100);
+		}
+
+		mutex_unlock(&ds_var_mutex);
+
+		if (ds_var_doorbell == 0 ||
+		    ds_var_response != DS_VAR_SUCCESS)
+			printk(KERN_ERR PFX "var-config [%s:%s] "
+			       "failed, response(%d).\n",
+			       var, value,
+			       ds_var_response);
+	} else {
+		printk(KERN_ERR PFX "var-config not registered so "
+		       "could not set (%s) variable to (%s).\n",
+		       var, value);
+	}
+}
+
+void ldom_reboot(const char *boot_command)
+{
+	/* Don't bother with any of this if the boot_command
+	 * is empty.
+	 */
+	if (boot_command && strlen(boot_command)) {
+		char full_boot_str[256];
+
+		strcpy(full_boot_str, "boot ");
+		strcpy(full_boot_str + strlen("boot "), boot_command);
+
+		ldom_set_var("reboot-command", full_boot_str);
+	}
+	sun4v_mach_sir();
+}
+
+void ldom_power_off(void)
+{
+	sun4v_mach_exit(0);
+}
+
+static void ds_conn_reset(struct ds_info *dp)
+{
+	printk(KERN_ERR PFX "ds_conn_reset() from %p\n",
+	       __builtin_return_address(0));
+}
+
+static int register_services(struct ds_info *dp)
+{
+	struct ldc_channel *lp = dp->lp;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ds_states); i++) {
+		struct {
+			struct ds_reg_req req;
+			u8 id_buf[256];
+		} pbuf;
+		struct ds_cap_state *cp = &ds_states[i];
+		int err, msg_len;
+		u64 new_count;
+
+		if (cp->state == CAP_STATE_REGISTERED)
+			continue;
+
+		new_count = sched_clock() & 0xffffffff;
+		cp->handle = ((u64) i << 32) | new_count;
+
+		msg_len = (sizeof(struct ds_reg_req) +
+			   strlen(cp->service_id));
+
+		memset(&pbuf, 0, sizeof(pbuf));
+		pbuf.req.tag.type = DS_REG_REQ;
+		pbuf.req.tag.len = (msg_len - sizeof(struct ds_msg_tag));
+		pbuf.req.handle = cp->handle;
+		pbuf.req.major = 1;
+		pbuf.req.minor = 0;
+		strcpy(pbuf.req.svc_id, cp->service_id);
+
+		err = ds_send(lp, &pbuf, msg_len);
+		if (err > 0)
+			cp->state = CAP_STATE_REG_SENT;
+	}
+	return 0;
+}
+
+static int ds_handshake(struct ds_info *dp, struct ds_msg_tag *pkt)
+{
+
+	if (dp->hs_state == DS_HS_START) {
+		if (pkt->type != DS_INIT_ACK)
+			goto conn_reset;
+
+		dp->hs_state = DS_HS_DONE;
+
+		return register_services(dp);
+	}
+
+	if (dp->hs_state != DS_HS_DONE)
+		goto conn_reset;
+
+	if (pkt->type == DS_REG_ACK) {
+		struct ds_reg_ack *ap = (struct ds_reg_ack *) pkt;
+		struct ds_cap_state *cp = find_cap(ap->handle);
+
+		if (!cp) {
+			printk(KERN_ERR PFX "REG ACK for unknown handle %lx\n",
+			       ap->handle);
+			return 0;
+		}
+		printk(KERN_INFO PFX "Registered %s service.\n",
+		       cp->service_id);
+		cp->state = CAP_STATE_REGISTERED;
+	} else if (pkt->type == DS_REG_NACK) {
+		struct ds_reg_nack *np = (struct ds_reg_nack *) pkt;
+		struct ds_cap_state *cp = find_cap(np->handle);
+
+		if (!cp) {
+			printk(KERN_ERR PFX "REG NACK for "
+			       "unknown handle %lx\n",
+			       np->handle);
+			return 0;
+		}
+		printk(KERN_INFO PFX "Could not register %s service\n",
+		       cp->service_id);
+		cp->state = CAP_STATE_UNKNOWN;
+	}
+
+	return 0;
+
+conn_reset:
+	ds_conn_reset(dp);
+	return -ECONNRESET;
+}
+
+static int ds_data(struct ds_info *dp, struct ds_msg_tag *pkt, int len)
+{
+	struct ds_data *dpkt = (struct ds_data *) pkt;
+	struct ds_cap_state *cp = find_cap(dpkt->handle);
+
+	if (!cp) {
+		struct ds_data_nack nack = {
+			.tag = {
+				.type = DS_NACK,
+				.len = (sizeof(struct ds_data_nack) -
+					sizeof(struct ds_msg_tag)),
+			},
+			.handle = dpkt->handle,
+			.result = DS_INV_HDL,
+		};
+
+		printk(KERN_ERR PFX "Data for unknown handle %lu\n",
+		       dpkt->handle);
+		ds_send(dp->lp, &nack, sizeof(nack));
+	} else {
+		cp->data(dp->lp, cp, dpkt, len);
+	}
+	return 0;
+}
+
+static void ds_up(struct ds_info *dp)
+{
+	struct ldc_channel *lp = dp->lp;
+	struct ds_ver_req req;
+	int err;
+
+	req.tag.type = DS_INIT_REQ;
+	req.tag.len = sizeof(req) - sizeof(struct ds_msg_tag);
+	req.ver.major = 1;
+	req.ver.minor = 0;
+
+	err = ds_send(lp, &req, sizeof(req));
+	if (err > 0)
+		dp->hs_state = DS_HS_START;
+}
+
+static void ds_event(void *arg, int event)
+{
+	struct ds_info *dp = arg;
+	struct ldc_channel *lp = dp->lp;
+	unsigned long flags;
+	int err;
+
+	spin_lock_irqsave(&ds_lock, flags);
+
+	if (event == LDC_EVENT_UP) {
+		ds_up(dp);
+		spin_unlock_irqrestore(&ds_lock, flags);
+		return;
+	}
+
+	if (event != LDC_EVENT_DATA_READY) {
+		printk(KERN_WARNING PFX "Unexpected LDC event %d\n", event);
+		spin_unlock_irqrestore(&ds_lock, flags);
+		return;
+	}
+
+	err = 0;
+	while (1) {
+		struct ds_msg_tag *tag;
+
+		err = ldc_read(lp, dp->rcv_buf, sizeof(*tag));
+
+		if (unlikely(err < 0)) {
+			if (err == -ECONNRESET)
+				ds_conn_reset(dp);
+			break;
+		}
+		if (err == 0)
+			break;
+
+		tag = dp->rcv_buf;
+		err = ldc_read(lp, tag + 1, tag->len);
+
+		if (unlikely(err < 0)) {
+			if (err == -ECONNRESET)
+				ds_conn_reset(dp);
+			break;
+		}
+		if (err < tag->len)
+			break;
+
+		if (tag->type < DS_DATA)
+			err = ds_handshake(dp, dp->rcv_buf);
+		else
+			err = ds_data(dp, dp->rcv_buf,
+				      sizeof(*tag) + err);
+		if (err == -ECONNRESET)
+			break;
+	}
+
+	spin_unlock_irqrestore(&ds_lock, flags);
+}
+
+static int __devinit ds_probe(struct vio_dev *vdev,
+			      const struct vio_device_id *id)
+{
+	static int ds_version_printed;
+	struct ldc_channel_config ds_cfg = {
+		.event		= ds_event,
+		.mtu		= 4096,
+		.mode		= LDC_MODE_STREAM,
+	};
+	struct ldc_channel *lp;
+	struct ds_info *dp;
+	int err;
+
+	if (ds_version_printed++ == 0)
+		printk(KERN_INFO "%s", version);
+
+	dp = kzalloc(sizeof(*dp), GFP_KERNEL);
+	err = -ENOMEM;
+	if (!dp)
+		goto out_err;
+
+	dp->rcv_buf = kzalloc(4096, GFP_KERNEL);
+	if (!dp->rcv_buf)
+		goto out_free_dp;
+
+	dp->rcv_buf_len = 4096;
+
+	ds_cfg.tx_irq = vdev->tx_irq;
+	ds_cfg.rx_irq = vdev->rx_irq;
+
+	lp = ldc_alloc(vdev->channel_id, &ds_cfg, dp);
+	if (IS_ERR(lp)) {
+		err = PTR_ERR(lp);
+		goto out_free_rcv_buf;
+	}
+	dp->lp = lp;
+
+	err = ldc_bind(lp, "DS");
+	if (err)
+		goto out_free_ldc;
+
+	ds_info = dp;
+
+	start_powerd();
+
+	return err;
+
+out_free_ldc:
+	ldc_free(dp->lp);
+
+out_free_rcv_buf:
+	kfree(dp->rcv_buf);
+
+out_free_dp:
+	kfree(dp);
+
+out_err:
+	return err;
+}
+
+static int ds_remove(struct vio_dev *vdev)
+{
+	return 0;
+}
+
+static struct vio_device_id ds_match[] = {
+	{
+		.type = "domain-services-port",
+	},
+	{},
+};
+
+static struct vio_driver ds_driver = {
+	.id_table	= ds_match,
+	.probe		= ds_probe,
+	.remove		= ds_remove,
+	.driver		= {
+		.name	= "ds",
+		.owner	= THIS_MODULE,
+	}
+};
+
+static int __init ds_init(void)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ds_states); i++)
+		ds_states[i].handle = ((u64)i << 32);
+
+#ifdef CONFIG_HOTPLUG_CPU
+	kthread_run(dr_cpu_thread, NULL, "kdrcpud");
+#endif
+
+	return vio_register_driver(&ds_driver);
+}
+
+subsys_initcall(ds_init);

arch/sparc64/kernel/hvtramp.S

@@ -0,0 +1,139 @@
+/* hvtramp.S: Hypervisor start-cpu trampoline code.
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <asm/thread_info.h>
+#include <asm/hypervisor.h>
+#include <asm/scratchpad.h>
+#include <asm/spitfire.h>
+#include <asm/hvtramp.h>
+#include <asm/pstate.h>
+#include <asm/ptrace.h>
+#include <asm/asi.h>
+
+	.text
+	.align		8
+	.globl		hv_cpu_startup, hv_cpu_startup_end
+
+	/* This code executes directly out of the hypervisor
+	 * with physical addressing (va==pa).  %o0 contains
+	 * our client argument which for Linux points to
+	 * a descriptor data structure which defines the
+	 * MMU entries we need to load up.
+	 *
+	 * After we set things up we enable the MMU and call
+	 * into the kernel.
+	 *
+	 * First setup basic privileged cpu state.
+	 */
+hv_cpu_startup:
+	wrpr		%g0, 0, %gl
+	wrpr		%g0, 15, %pil
+	wrpr		%g0, 0, %canrestore
+	wrpr		%g0, 0, %otherwin
+	wrpr		%g0, 6, %cansave
+	wrpr		%g0, 6, %cleanwin
+	wrpr		%g0, 0, %cwp
+	wrpr		%g0, 0, %wstate
+	wrpr		%g0, 0, %tl
+
+	sethi		%hi(sparc64_ttable_tl0), %g1
+	wrpr		%g1, %tba
+
+	mov		%o0, %l0
+
+	lduw		[%l0 + HVTRAMP_DESCR_CPU], %g1
+	mov		SCRATCHPAD_CPUID, %g2
+	stxa		%g1, [%g2] ASI_SCRATCHPAD
+
+	ldx		[%l0 + HVTRAMP_DESCR_FAULT_INFO_VA], %g2
+	stxa		%g2, [%g0] ASI_SCRATCHPAD
+
+	mov		0, %l1
+	lduw		[%l0 + HVTRAMP_DESCR_NUM_MAPPINGS], %l2
+	add		%l0, HVTRAMP_DESCR_MAPS, %l3
+
+1:	ldx		[%l3 + HVTRAMP_MAPPING_VADDR], %o0
+	clr		%o1
+	ldx		[%l3 + HVTRAMP_MAPPING_TTE], %o2
+	mov		HV_MMU_IMMU | HV_MMU_DMMU, %o3
+	mov		HV_FAST_MMU_MAP_PERM_ADDR, %o5
+	ta		HV_FAST_TRAP
+
+	brnz,pn		%o0, 80f
+	 nop
+
+	add		%l1, 1, %l1
+	cmp		%l1, %l2
+	blt,a,pt	%xcc, 1b
+	 add		%l3, HVTRAMP_MAPPING_SIZE, %l3
+
+	ldx		[%l0 + HVTRAMP_DESCR_FAULT_INFO_PA], %o0
+	mov		HV_FAST_MMU_FAULT_AREA_CONF, %o5
+	ta		HV_FAST_TRAP
+
+	brnz,pn		%o0, 80f
+	 nop
+
+	wrpr		%g0, (PSTATE_PRIV | PSTATE_PEF), %pstate
+
+	ldx		[%l0 + HVTRAMP_DESCR_THREAD_REG], %l6
+
+	mov		1, %o0
+	set		1f, %o1
+	mov		HV_FAST_MMU_ENABLE, %o5
+	ta		HV_FAST_TRAP
+
+	ba,pt		%xcc, 80f
+	 nop
+
+1:
+	wr		%g0, 0, %fprs
+	wr		%g0, ASI_P, %asi
+
+	mov		PRIMARY_CONTEXT, %g7
+	stxa		%g0, [%g7] ASI_MMU
+	membar		#Sync
+
+	mov		SECONDARY_CONTEXT, %g7
+	stxa		%g0, [%g7] ASI_MMU
+	membar		#Sync
+
+	mov		%l6, %g6
+	ldx		[%g6 + TI_TASK], %g4
+
+	mov		1, %g5
+	sllx		%g5, THREAD_SHIFT, %g5
+	sub		%g5, (STACKFRAME_SZ + STACK_BIAS), %g5
+	add		%g6, %g5, %sp
+	mov		0, %fp
+
+	call		init_irqwork_curcpu
+	 nop
+	call		hard_smp_processor_id
+	 nop
+
+	mov		%o0, %o1
+	mov		0, %o0
+	mov		0, %o2
+	call		sun4v_init_mondo_queues
+	 mov		1, %o3
+
+	call		init_cur_cpu_trap
+	 mov		%g6, %o0
+
+	wrpr		%g0, (PSTATE_PRIV | PSTATE_PEF | PSTATE_IE), %pstate
+
+	call		smp_callin
+	 nop
+	call		cpu_idle
+	 mov		0, %o0
+	call		cpu_panic
+	 nop
+
+80:	ba,pt		%xcc, 80b
+	 nop
+
+	.align		8
+hv_cpu_startup_end:

arch/sparc64/kernel/irq.c

@@ -293,6 +293,11 @@ static void sun4u_irq_enable(unsigned int virt_irq)
 	}
 }
 
+static void sun4u_set_affinity(unsigned int virt_irq, cpumask_t mask)
+{
+	sun4u_irq_enable(virt_irq);
+}
+
 static void sun4u_irq_disable(unsigned int virt_irq)
 {
 	struct irq_handler_data *data = get_irq_chip_data(virt_irq);
@@ -309,6 +314,10 @@ static void sun4u_irq_disable(unsigned int virt_irq)
 static void sun4u_irq_end(unsigned int virt_irq)
 {
 	struct irq_handler_data *data = get_irq_chip_data(virt_irq);
+	struct irq_desc *desc = irq_desc + virt_irq;
+
+	if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+		return;
 
 	if (likely(data))
 		upa_writeq(ICLR_IDLE, data->iclr);
@@ -340,6 +349,24 @@ static void sun4v_irq_enable(unsigned int virt_irq)
 	}
 }
 
+static void sun4v_set_affinity(unsigned int virt_irq, cpumask_t mask)
+{
+	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
+	unsigned int ino = bucket - &ivector_table[0];
+
+	if (likely(bucket)) {
+		unsigned long cpuid;
+		int err;
+
+		cpuid = irq_choose_cpu(virt_irq);
+
+		err = sun4v_intr_settarget(ino, cpuid);
+		if (err != HV_EOK)
+			printk("sun4v_intr_settarget(%x,%lu): err(%d)\n",
+			       ino, cpuid, err);
+	}
+}
+
 static void sun4v_irq_disable(unsigned int virt_irq)
 {
 	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
@@ -373,6 +400,10 @@ static void sun4v_irq_end(unsigned int virt_irq)
 {
 	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
 	unsigned int ino = bucket - &ivector_table[0];
+	struct irq_desc *desc = irq_desc + virt_irq;
+
+	if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+		return;
 
 	if (likely(bucket)) {
 		int err;
@@ -418,6 +449,28 @@ static void sun4v_virq_enable(unsigned int virt_irq)
 	}
 }
 
+static void sun4v_virt_set_affinity(unsigned int virt_irq, cpumask_t mask)
+{
+	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
+	unsigned int ino = bucket - &ivector_table[0];
+
+	if (likely(bucket)) {
+		unsigned long cpuid, dev_handle, dev_ino;
+		int err;
+
+		cpuid = irq_choose_cpu(virt_irq);
+
+		dev_handle = ino & IMAP_IGN;
+		dev_ino = ino & IMAP_INO;
+
+		err = sun4v_vintr_set_target(dev_handle, dev_ino, cpuid);
+		if (err != HV_EOK)
+			printk("sun4v_vintr_set_target(%lx,%lx,%lu): "
+			       "err(%d)\n",
+			       dev_handle, dev_ino, cpuid, err);
+	}
+}
+
 static void sun4v_virq_disable(unsigned int virt_irq)
 {
 	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
@@ -443,6 +496,10 @@ static void sun4v_virq_end(unsigned int virt_irq)
 {
 	struct ino_bucket *bucket = virt_irq_to_bucket(virt_irq);
 	unsigned int ino = bucket - &ivector_table[0];
+	struct irq_desc *desc = irq_desc + virt_irq;
+
+	if (unlikely(desc->status & (IRQ_DISABLED|IRQ_INPROGRESS)))
+		return;
 
 	if (likely(bucket)) {
 		unsigned long dev_handle, dev_ino;
@@ -477,6 +534,7 @@ static struct irq_chip sun4u_irq = {
 	.enable		= sun4u_irq_enable,
 	.disable	= sun4u_irq_disable,
 	.end		= sun4u_irq_end,
+	.set_affinity	= sun4u_set_affinity,
 };
 
 static struct irq_chip sun4u_irq_ack = {
@@ -485,6 +543,7 @@ static struct irq_chip sun4u_irq_ack = {
 	.disable	= sun4u_irq_disable,
 	.ack		= run_pre_handler,
 	.end		= sun4u_irq_end,
+	.set_affinity	= sun4u_set_affinity,
 };
 
 static struct irq_chip sun4v_irq = {
@@ -492,6 +551,7 @@ static struct irq_chip sun4v_irq = {
 	.enable		= sun4v_irq_enable,
 	.disable	= sun4v_irq_disable,
 	.end		= sun4v_irq_end,
+	.set_affinity	= sun4v_set_affinity,
 };
 
 static struct irq_chip sun4v_irq_ack = {
@@ -500,6 +560,7 @@ static struct irq_chip sun4v_irq_ack = {
 	.disable	= sun4v_irq_disable,
 	.ack		= run_pre_handler,
 	.end		= sun4v_irq_end,
+	.set_affinity	= sun4v_set_affinity,
 };
 
 #ifdef CONFIG_PCI_MSI
@@ -511,6 +572,7 @@ static struct irq_chip sun4v_msi = {
 	.disable	= sun4v_msi_disable,
 	.ack		= run_pre_handler,
 	.end		= sun4v_irq_end,
+	.set_affinity	= sun4v_set_affinity,
 };
 #endif
 
@@ -519,6 +581,7 @@ static struct irq_chip sun4v_virq = {
 	.enable		= sun4v_virq_enable,
 	.disable	= sun4v_virq_disable,
 	.end		= sun4v_virq_end,
+	.set_affinity	= sun4v_virt_set_affinity,
 };
 
 static struct irq_chip sun4v_virq_ack = {
@@ -527,6 +590,7 @@ static struct irq_chip sun4v_virq_ack = {
 	.disable	= sun4v_virq_disable,
 	.ack		= run_pre_handler,
 	.end		= sun4v_virq_end,
+	.set_affinity	= sun4v_virt_set_affinity,
 };
 
 void irq_install_pre_handler(int virt_irq,
@@ -739,6 +803,26 @@ void handler_irq(int irq, struct pt_regs *regs)
 	set_irq_regs(old_regs);
 }
 
+#ifdef CONFIG_HOTPLUG_CPU
+void fixup_irqs(void)
+{
+	unsigned int irq;
+
+	for (irq = 0; irq < NR_IRQS; irq++) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&irq_desc[irq].lock, flags);
+		if (irq_desc[irq].action &&
+		    !(irq_desc[irq].status & IRQ_PER_CPU)) {
+			if (irq_desc[irq].chip->set_affinity)
+				irq_desc[irq].chip->set_affinity(irq,
+					irq_desc[irq].affinity);
+		}
+		spin_unlock_irqrestore(&irq_desc[irq].lock, flags);
+	}
+}
+#endif
+
 struct sun5_timer {
 	u64	count0;
 	u64	limit0;

arch/sparc64/kernel/ldc.c

@@ -0,0 +1,2373 @@
+/* ldc.c: Logical Domain Channel link-layer protocol driver.
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/scatterlist.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/init.h>
+
+#include <asm/hypervisor.h>
+#include <asm/iommu.h>
+#include <asm/page.h>
+#include <asm/ldc.h>
+#include <asm/mdesc.h>
+
+#define DRV_MODULE_NAME		"ldc"
+#define PFX DRV_MODULE_NAME	": "
+#define DRV_MODULE_VERSION	"1.0"
+#define DRV_MODULE_RELDATE	"June 25, 2007"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+#define LDC_PACKET_SIZE		64
+
+/* Packet header layout for unreliable and reliable mode frames.
+ * When in RAW mode, packets are simply straight 64-byte payloads
+ * with no headers.
+ */
+struct ldc_packet {
+	u8			type;
+#define LDC_CTRL		0x01
+#define LDC_DATA		0x02
+#define LDC_ERR			0x10
+
+	u8			stype;
+#define LDC_INFO		0x01
+#define LDC_ACK			0x02
+#define LDC_NACK		0x04
+
+	u8			ctrl;
+#define LDC_VERS		0x01 /* Link Version		*/
+#define LDC_RTS			0x02 /* Request To Send		*/
+#define LDC_RTR			0x03 /* Ready To Receive	*/
+#define LDC_RDX			0x04 /* Ready for Data eXchange	*/
+#define LDC_CTRL_MSK		0x0f
+
+	u8			env;
+#define LDC_LEN			0x3f
+#define LDC_FRAG_MASK		0xc0
+#define LDC_START		0x40
+#define LDC_STOP		0x80
+
+	u32			seqid;
+
+	union {
+		u8		u_data[LDC_PACKET_SIZE - 8];
+		struct {
+			u32	pad;
+			u32	ackid;
+			u8	r_data[LDC_PACKET_SIZE - 8 - 8];
+		} r;
+	} u;
+};
+
+struct ldc_version {
+	u16 major;
+	u16 minor;
+};
+
+/* Ordered from largest major to lowest.  */
+static struct ldc_version ver_arr[] = {
+	{ .major = 1, .minor = 0 },
+};
+
+#define LDC_DEFAULT_MTU			(4 * LDC_PACKET_SIZE)
+#define LDC_DEFAULT_NUM_ENTRIES		(PAGE_SIZE / LDC_PACKET_SIZE)
+
+struct ldc_channel;
+
+struct ldc_mode_ops {
+	int (*write)(struct ldc_channel *, const void *, unsigned int);
+	int (*read)(struct ldc_channel *, void *, unsigned int);
+};
+
+static const struct ldc_mode_ops raw_ops;
+static const struct ldc_mode_ops nonraw_ops;
+static const struct ldc_mode_ops stream_ops;
+
+int ldom_domaining_enabled;
+
+struct ldc_iommu {
+	/* Protects arena alloc/free.  */
+	spinlock_t			lock;
+	struct iommu_arena		arena;
+	struct ldc_mtable_entry		*page_table;
+};
+
+struct ldc_channel {
+	/* Protects all operations that depend upon channel state.  */
+	spinlock_t			lock;
+
+	unsigned long			id;
+
+	u8				*mssbuf;
+	u32				mssbuf_len;
+	u32				mssbuf_off;
+
+	struct ldc_packet		*tx_base;
+	unsigned long			tx_head;
+	unsigned long			tx_tail;
+	unsigned long			tx_num_entries;
+	unsigned long			tx_ra;
+
+	unsigned long			tx_acked;
+
+	struct ldc_packet		*rx_base;
+	unsigned long			rx_head;
+	unsigned long			rx_tail;
+	unsigned long			rx_num_entries;
+	unsigned long			rx_ra;
+
+	u32				rcv_nxt;
+	u32				snd_nxt;
+
+	unsigned long			chan_state;
+
+	struct ldc_channel_config	cfg;
+	void				*event_arg;
+
+	const struct ldc_mode_ops	*mops;
+
+	struct ldc_iommu		iommu;
+
+	struct ldc_version		ver;
+
+	u8				hs_state;
+#define LDC_HS_CLOSED			0x00
+#define LDC_HS_OPEN			0x01
+#define LDC_HS_GOTVERS			0x02
+#define LDC_HS_SENTRTR			0x03
+#define LDC_HS_GOTRTR			0x04
+#define LDC_HS_COMPLETE			0x10
+
+	u8				flags;
+#define LDC_FLAG_ALLOCED_QUEUES		0x01
+#define LDC_FLAG_REGISTERED_QUEUES	0x02
+#define LDC_FLAG_REGISTERED_IRQS	0x04
+#define LDC_FLAG_RESET			0x10
+
+	u8				mss;
+	u8				state;
+
+#define LDC_IRQ_NAME_MAX		32
+	char				rx_irq_name[LDC_IRQ_NAME_MAX];
+	char				tx_irq_name[LDC_IRQ_NAME_MAX];
+
+	struct hlist_head		mh_list;
+
+	struct hlist_node		list;
+};
+
+#define ldcdbg(TYPE, f, a...) \
+do {	if (lp->cfg.debug & LDC_DEBUG_##TYPE) \
+		printk(KERN_INFO PFX "ID[%lu] " f, lp->id, ## a); \
+} while (0)
+
+static const char *state_to_str(u8 state)
+{
+	switch (state) {
+	case LDC_STATE_INVALID:
+		return "INVALID";
+	case LDC_STATE_INIT:
+		return "INIT";
+	case LDC_STATE_BOUND:
+		return "BOUND";
+	case LDC_STATE_READY:
+		return "READY";
+	case LDC_STATE_CONNECTED:
+		return "CONNECTED";
+	default:
+		return "<UNKNOWN>";
+	}
+}
+
+static void ldc_set_state(struct ldc_channel *lp, u8 state)
+{
+	ldcdbg(STATE, "STATE (%s) --> (%s)\n",
+	       state_to_str(lp->state),
+	       state_to_str(state));
+
+	lp->state = state;
+}
+
+static unsigned long __advance(unsigned long off, unsigned long num_entries)
+{
+	off += LDC_PACKET_SIZE;
+	if (off == (num_entries * LDC_PACKET_SIZE))
+		off = 0;
+
+	return off;
+}
+
+static unsigned long rx_advance(struct ldc_channel *lp, unsigned long off)
+{
+	return __advance(off, lp->rx_num_entries);
+}
+
+static unsigned long tx_advance(struct ldc_channel *lp, unsigned long off)
+{
+	return __advance(off, lp->tx_num_entries);
+}
+
+static struct ldc_packet *handshake_get_tx_packet(struct ldc_channel *lp,
+						  unsigned long *new_tail)
+{
+	struct ldc_packet *p;
+	unsigned long t;
+
+	t = tx_advance(lp, lp->tx_tail);
+	if (t == lp->tx_head)
+		return NULL;
+
+	*new_tail = t;
+
+	p = lp->tx_base;
+	return p + (lp->tx_tail / LDC_PACKET_SIZE);
+}
+
+/* When we are in reliable or stream mode, have to track the next packet
+ * we haven't gotten an ACK for in the TX queue using tx_acked.  We have
+ * to be careful not to stomp over the queue past that point.  During
+ * the handshake, we don't have TX data packets pending in the queue
+ * and that's why handshake_get_tx_packet() need not be mindful of
+ * lp->tx_acked.
+ */
+static unsigned long head_for_data(struct ldc_channel *lp)
+{
+	if (lp->cfg.mode == LDC_MODE_STREAM)
+		return lp->tx_acked;
+	return lp->tx_head;
+}
+
+static int tx_has_space_for(struct ldc_channel *lp, unsigned int size)
+{
+	unsigned long limit, tail, new_tail, diff;
+	unsigned int mss;
+
+	limit = head_for_data(lp);
+	tail = lp->tx_tail;
+	new_tail = tx_advance(lp, tail);
+	if (new_tail == limit)
+		return 0;
+
+	if (limit > new_tail)
+		diff = limit - new_tail;
+	else
+		diff = (limit +
+			((lp->tx_num_entries * LDC_PACKET_SIZE) - new_tail));
+	diff /= LDC_PACKET_SIZE;
+	mss = lp->mss;
+
+	if (diff * mss < size)
+		return 0;
+
+	return 1;
+}
+
+static struct ldc_packet *data_get_tx_packet(struct ldc_channel *lp,
+					     unsigned long *new_tail)
+{
+	struct ldc_packet *p;
+	unsigned long h, t;
+
+	h = head_for_data(lp);
+	t = tx_advance(lp, lp->tx_tail);
+	if (t == h)
+		return NULL;
+
+	*new_tail = t;
+
+	p = lp->tx_base;
+	return p + (lp->tx_tail / LDC_PACKET_SIZE);
+}
+
+static int set_tx_tail(struct ldc_channel *lp, unsigned long tail)
+{
+	unsigned long orig_tail = lp->tx_tail;
+	int limit = 1000;
+
+	lp->tx_tail = tail;
+	while (limit-- > 0) {
+		unsigned long err;
+
+		err = sun4v_ldc_tx_set_qtail(lp->id, tail);
+		if (!err)
+			return 0;
+
+		if (err != HV_EWOULDBLOCK) {
+			lp->tx_tail = orig_tail;
+			return -EINVAL;
+		}
+		udelay(1);
+	}
+
+	lp->tx_tail = orig_tail;
+	return -EBUSY;
+}
+
+/* This just updates the head value in the hypervisor using
+ * a polling loop with a timeout.  The caller takes care of
+ * upating software state representing the head change, if any.
+ */
+static int __set_rx_head(struct ldc_channel *lp, unsigned long head)
+{
+	int limit = 1000;
+
+	while (limit-- > 0) {
+		unsigned long err;
+
+		err = sun4v_ldc_rx_set_qhead(lp->id, head);
+		if (!err)
+			return 0;
+
+		if (err != HV_EWOULDBLOCK)
+			return -EINVAL;
+
+		udelay(1);
+	}
+
+	return -EBUSY;
+}
+
+static int send_tx_packet(struct ldc_channel *lp,
+			  struct ldc_packet *p,
+			  unsigned long new_tail)
+{
+	BUG_ON(p != (lp->tx_base + (lp->tx_tail / LDC_PACKET_SIZE)));
+
+	return set_tx_tail(lp, new_tail);
+}
+
+static struct ldc_packet *handshake_compose_ctrl(struct ldc_channel *lp,
+						 u8 stype, u8 ctrl,
+						 void *data, int dlen,
+						 unsigned long *new_tail)
+{
+	struct ldc_packet *p = handshake_get_tx_packet(lp, new_tail);
+
+	if (p) {
+		memset(p, 0, sizeof(*p));
+		p->type = LDC_CTRL;
+		p->stype = stype;
+		p->ctrl = ctrl;
+		if (data)
+			memcpy(p->u.u_data, data, dlen);
+	}
+	return p;
+}
+
+static int start_handshake(struct ldc_channel *lp)
+{
+	struct ldc_packet *p;
+	struct ldc_version *ver;
+	unsigned long new_tail;
+
+	ver = &ver_arr[0];
+
+	ldcdbg(HS, "SEND VER INFO maj[%u] min[%u]\n",
+	       ver->major, ver->minor);
+
+	p = handshake_compose_ctrl(lp, LDC_INFO, LDC_VERS,
+				   ver, sizeof(*ver), &new_tail);
+	if (p) {
+		int err = send_tx_packet(lp, p, new_tail);
+		if (!err)
+			lp->flags &= ~LDC_FLAG_RESET;
+		return err;
+	}
+	return -EBUSY;
+}
+
+static int send_version_nack(struct ldc_channel *lp,
+			     u16 major, u16 minor)
+{
+	struct ldc_packet *p;
+	struct ldc_version ver;
+	unsigned long new_tail;
+
+	ver.major = major;
+	ver.minor = minor;
+
+	p = handshake_compose_ctrl(lp, LDC_NACK, LDC_VERS,
+				   &ver, sizeof(ver), &new_tail);
+	if (p) {
+		ldcdbg(HS, "SEND VER NACK maj[%u] min[%u]\n",
+		       ver.major, ver.minor);
+
+		return send_tx_packet(lp, p, new_tail);
+	}
+	return -EBUSY;
+}
+
+static int send_version_ack(struct ldc_channel *lp,
+			    struct ldc_version *vp)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+
+	p = handshake_compose_ctrl(lp, LDC_ACK, LDC_VERS,
+				   vp, sizeof(*vp), &new_tail);
+	if (p) {
+		ldcdbg(HS, "SEND VER ACK maj[%u] min[%u]\n",
+		       vp->major, vp->minor);
+
+		return send_tx_packet(lp, p, new_tail);
+	}
+	return -EBUSY;
+}
+
+static int send_rts(struct ldc_channel *lp)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+
+	p = handshake_compose_ctrl(lp, LDC_INFO, LDC_RTS, NULL, 0,
+				   &new_tail);
+	if (p) {
+		p->env = lp->cfg.mode;
+		p->seqid = 0;
+		lp->rcv_nxt = 0;
+
+		ldcdbg(HS, "SEND RTS env[0x%x] seqid[0x%x]\n",
+		       p->env, p->seqid);
+
+		return send_tx_packet(lp, p, new_tail);
+	}
+	return -EBUSY;
+}
+
+static int send_rtr(struct ldc_channel *lp)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+
+	p = handshake_compose_ctrl(lp, LDC_INFO, LDC_RTR, NULL, 0,
+				   &new_tail);
+	if (p) {
+		p->env = lp->cfg.mode;
+		p->seqid = 0;
+
+		ldcdbg(HS, "SEND RTR env[0x%x] seqid[0x%x]\n",
+		       p->env, p->seqid);
+
+		return send_tx_packet(lp, p, new_tail);
+	}
+	return -EBUSY;
+}
+
+static int send_rdx(struct ldc_channel *lp)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+
+	p = handshake_compose_ctrl(lp, LDC_INFO, LDC_RDX, NULL, 0,
+				   &new_tail);
+	if (p) {
+		p->env = 0;
+		p->seqid = ++lp->snd_nxt;
+		p->u.r.ackid = lp->rcv_nxt;
+
+		ldcdbg(HS, "SEND RDX env[0x%x] seqid[0x%x] ackid[0x%x]\n",
+		       p->env, p->seqid, p->u.r.ackid);
+
+		return send_tx_packet(lp, p, new_tail);
+	}
+	return -EBUSY;
+}
+
+static int send_data_nack(struct ldc_channel *lp, struct ldc_packet *data_pkt)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+	int err;
+
+	p = data_get_tx_packet(lp, &new_tail);
+	if (!p)
+		return -EBUSY;
+	memset(p, 0, sizeof(*p));
+	p->type = data_pkt->type;
+	p->stype = LDC_NACK;
+	p->ctrl = data_pkt->ctrl & LDC_CTRL_MSK;
+	p->seqid = lp->snd_nxt + 1;
+	p->u.r.ackid = lp->rcv_nxt;
+
+	ldcdbg(HS, "SEND DATA NACK type[0x%x] ctl[0x%x] seq[0x%x] ack[0x%x]\n",
+	       p->type, p->ctrl, p->seqid, p->u.r.ackid);
+
+	err = send_tx_packet(lp, p, new_tail);
+	if (!err)
+		lp->snd_nxt++;
+
+	return err;
+}
+
+static int ldc_abort(struct ldc_channel *lp)
+{
+	unsigned long hv_err;
+
+	ldcdbg(STATE, "ABORT\n");
+
+	/* We report but do not act upon the hypervisor errors because
+	 * there really isn't much we can do if they fail at this point.
+	 */
+	hv_err = sun4v_ldc_tx_qconf(lp->id, lp->tx_ra, lp->tx_num_entries);
+	if (hv_err)
+		printk(KERN_ERR PFX "ldc_abort: "
+		       "sun4v_ldc_tx_qconf(%lx,%lx,%lx) failed, err=%lu\n",
+		       lp->id, lp->tx_ra, lp->tx_num_entries, hv_err);
+
+	hv_err = sun4v_ldc_tx_get_state(lp->id,
+					&lp->tx_head,
+					&lp->tx_tail,
+					&lp->chan_state);
+	if (hv_err)
+		printk(KERN_ERR PFX "ldc_abort: "
+		       "sun4v_ldc_tx_get_state(%lx,...) failed, err=%lu\n",
+		       lp->id, hv_err);
+
+	hv_err = sun4v_ldc_rx_qconf(lp->id, lp->rx_ra, lp->rx_num_entries);
+	if (hv_err)
+		printk(KERN_ERR PFX "ldc_abort: "
+		       "sun4v_ldc_rx_qconf(%lx,%lx,%lx) failed, err=%lu\n",
+		       lp->id, lp->rx_ra, lp->rx_num_entries, hv_err);
+
+	/* Refetch the RX queue state as well, because we could be invoked
+	 * here in the queue processing context.
+	 */
+	hv_err = sun4v_ldc_rx_get_state(lp->id,
+					&lp->rx_head,
+					&lp->rx_tail,
+					&lp->chan_state);
+	if (hv_err)
+		printk(KERN_ERR PFX "ldc_abort: "
+		       "sun4v_ldc_rx_get_state(%lx,...) failed, err=%lu\n",
+		       lp->id, hv_err);
+
+	return -ECONNRESET;
+}
+
+static struct ldc_version *find_by_major(u16 major)
+{
+	struct ldc_version *ret = NULL;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ver_arr); i++) {
+		struct ldc_version *v = &ver_arr[i];
+		if (v->major <= major) {
+			ret = v;
+			break;
+		}
+	}
+	return ret;
+}
+
+static int process_ver_info(struct ldc_channel *lp, struct ldc_version *vp)
+{
+	struct ldc_version *vap;
+	int err;
+
+	ldcdbg(HS, "GOT VERSION INFO major[%x] minor[%x]\n",
+	       vp->major, vp->minor);
+
+	if (lp->hs_state == LDC_HS_GOTVERS) {
+		lp->hs_state = LDC_HS_OPEN;
+		memset(&lp->ver, 0, sizeof(lp->ver));
+	}
+
+	vap = find_by_major(vp->major);
+	if (!vap) {
+		err = send_version_nack(lp, 0, 0);
+	} else if (vap->major != vp->major) {
+		err = send_version_nack(lp, vap->major, vap->minor);
+	} else {
+		struct ldc_version ver = *vp;
+		if (ver.minor > vap->minor)
+			ver.minor = vap->minor;
+		err = send_version_ack(lp, &ver);
+		if (!err) {
+			lp->ver = ver;
+			lp->hs_state = LDC_HS_GOTVERS;
+		}
+	}
+	if (err)
+		return ldc_abort(lp);
+
+	return 0;
+}
+
+static int process_ver_ack(struct ldc_channel *lp, struct ldc_version *vp)
+{
+	ldcdbg(HS, "GOT VERSION ACK major[%x] minor[%x]\n",
+	       vp->major, vp->minor);
+
+	if (lp->hs_state == LDC_HS_GOTVERS) {
+		if (lp->ver.major != vp->major ||
+		    lp->ver.minor != vp->minor)
+			return ldc_abort(lp);
+	} else {
+		lp->ver = *vp;
+		lp->hs_state = LDC_HS_GOTVERS;
+	}
+	if (send_rts(lp))
+		return ldc_abort(lp);
+	return 0;
+}
+
+static int process_ver_nack(struct ldc_channel *lp, struct ldc_version *vp)
+{
+	struct ldc_version *vap;
+
+	if ((vp->major == 0 && vp->minor == 0) ||
+	    !(vap = find_by_major(vp->major))) {
+		return ldc_abort(lp);
+	} else {
+		struct ldc_packet *p;
+		unsigned long new_tail;
+
+		p = handshake_compose_ctrl(lp, LDC_INFO, LDC_VERS,
+					   vap, sizeof(*vap),
+					   &new_tail);
+		if (p)
+			return send_tx_packet(lp, p, new_tail);
+		else
+			return ldc_abort(lp);
+	}
+}
+
+static int process_version(struct ldc_channel *lp,
+			   struct ldc_packet *p)
+{
+	struct ldc_version *vp;
+
+	vp = (struct ldc_version *) p->u.u_data;
+
+	switch (p->stype) {
+	case LDC_INFO:
+		return process_ver_info(lp, vp);
+
+	case LDC_ACK:
+		return process_ver_ack(lp, vp);
+
+	case LDC_NACK:
+		return process_ver_nack(lp, vp);
+
+	default:
+		return ldc_abort(lp);
+	}
+}
+
+static int process_rts(struct ldc_channel *lp,
+		       struct ldc_packet *p)
+{
+	ldcdbg(HS, "GOT RTS stype[%x] seqid[%x] env[%x]\n",
+	       p->stype, p->seqid, p->env);
+
+	if (p->stype     != LDC_INFO	   ||
+	    lp->hs_state != LDC_HS_GOTVERS ||
+	    p->env       != lp->cfg.mode)
+		return ldc_abort(lp);
+
+	lp->snd_nxt = p->seqid;
+	lp->rcv_nxt = p->seqid;
+	lp->hs_state = LDC_HS_SENTRTR;
+	if (send_rtr(lp))
+		return ldc_abort(lp);
+
+	return 0;
+}
+
+static int process_rtr(struct ldc_channel *lp,
+		       struct ldc_packet *p)
+{
+	ldcdbg(HS, "GOT RTR stype[%x] seqid[%x] env[%x]\n",
+	       p->stype, p->seqid, p->env);
+
+	if (p->stype     != LDC_INFO ||
+	    p->env       != lp->cfg.mode)
+		return ldc_abort(lp);
+
+	lp->snd_nxt = p->seqid;
+	lp->hs_state = LDC_HS_COMPLETE;
+	ldc_set_state(lp, LDC_STATE_CONNECTED);
+	send_rdx(lp);
+
+	return LDC_EVENT_UP;
+}
+
+static int rx_seq_ok(struct ldc_channel *lp, u32 seqid)
+{
+	return lp->rcv_nxt + 1 == seqid;
+}
+
+static int process_rdx(struct ldc_channel *lp,
+		       struct ldc_packet *p)
+{
+	ldcdbg(HS, "GOT RDX stype[%x] seqid[%x] env[%x] ackid[%x]\n",
+	       p->stype, p->seqid, p->env, p->u.r.ackid);
+
+	if (p->stype != LDC_INFO ||
+	    !(rx_seq_ok(lp, p->seqid)))
+		return ldc_abort(lp);
+
+	lp->rcv_nxt = p->seqid;
+
+	lp->hs_state = LDC_HS_COMPLETE;
+	ldc_set_state(lp, LDC_STATE_CONNECTED);
+
+	return LDC_EVENT_UP;
+}
+
+static int process_control_frame(struct ldc_channel *lp,
+				 struct ldc_packet *p)
+{
+	switch (p->ctrl) {
+	case LDC_VERS:
+		return process_version(lp, p);
+
+	case LDC_RTS:
+		return process_rts(lp, p);
+
+	case LDC_RTR:
+		return process_rtr(lp, p);
+
+	case LDC_RDX:
+		return process_rdx(lp, p);
+
+	default:
+		return ldc_abort(lp);
+	}
+}
+
+static int process_error_frame(struct ldc_channel *lp,
+			       struct ldc_packet *p)
+{
+	return ldc_abort(lp);
+}
+
+static int process_data_ack(struct ldc_channel *lp,
+			    struct ldc_packet *ack)
+{
+	unsigned long head = lp->tx_acked;
+	u32 ackid = ack->u.r.ackid;
+
+	while (1) {
+		struct ldc_packet *p = lp->tx_base + (head / LDC_PACKET_SIZE);
+
+		head = tx_advance(lp, head);
+
+		if (p->seqid == ackid) {
+			lp->tx_acked = head;
+			return 0;
+		}
+		if (head == lp->tx_tail)
+			return ldc_abort(lp);
+	}
+
+	return 0;
+}
+
+static void send_events(struct ldc_channel *lp, unsigned int event_mask)
+{
+	if (event_mask & LDC_EVENT_RESET)
+		lp->cfg.event(lp->event_arg, LDC_EVENT_RESET);
+	if (event_mask & LDC_EVENT_UP)
+		lp->cfg.event(lp->event_arg, LDC_EVENT_UP);
+	if (event_mask & LDC_EVENT_DATA_READY)
+		lp->cfg.event(lp->event_arg, LDC_EVENT_DATA_READY);
+}
+
+static irqreturn_t ldc_rx(int irq, void *dev_id)
+{
+	struct ldc_channel *lp = dev_id;
+	unsigned long orig_state, hv_err, flags;
+	unsigned int event_mask;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	orig_state = lp->chan_state;
+	hv_err = sun4v_ldc_rx_get_state(lp->id,
+					&lp->rx_head,
+					&lp->rx_tail,
+					&lp->chan_state);
+
+	ldcdbg(RX, "RX state[0x%02lx:0x%02lx] head[0x%04lx] tail[0x%04lx]\n",
+	       orig_state, lp->chan_state, lp->rx_head, lp->rx_tail);
+
+	event_mask = 0;
+
+	if (lp->cfg.mode == LDC_MODE_RAW &&
+	    lp->chan_state == LDC_CHANNEL_UP) {
+		lp->hs_state = LDC_HS_COMPLETE;
+		ldc_set_state(lp, LDC_STATE_CONNECTED);
+
+		event_mask |= LDC_EVENT_UP;
+
+		orig_state = lp->chan_state;
+	}
+
+	/* If we are in reset state, flush the RX queue and ignore
+	 * everything.
+	 */
+	if (lp->flags & LDC_FLAG_RESET) {
+		(void) __set_rx_head(lp, lp->rx_tail);
+		goto out;
+	}
+
+	/* Once we finish the handshake, we let the ldc_read()
+	 * paths do all of the control frame and state management.
+	 * Just trigger the callback.
+	 */
+	if (lp->hs_state == LDC_HS_COMPLETE) {
+handshake_complete:
+		if (lp->chan_state != orig_state) {
+			unsigned int event = LDC_EVENT_RESET;
+
+			if (lp->chan_state == LDC_CHANNEL_UP)
+				event = LDC_EVENT_UP;
+
+			event_mask |= event;
+		}
+		if (lp->rx_head != lp->rx_tail)
+			event_mask |= LDC_EVENT_DATA_READY;
+
+		goto out;
+	}
+
+	if (lp->chan_state != orig_state)
+		goto out;
+
+	while (lp->rx_head != lp->rx_tail) {
+		struct ldc_packet *p;
+		unsigned long new;
+		int err;
+
+		p = lp->rx_base + (lp->rx_head / LDC_PACKET_SIZE);
+
+		switch (p->type) {
+		case LDC_CTRL:
+			err = process_control_frame(lp, p);
+			if (err > 0)
+				event_mask |= err;
+			break;
+
+		case LDC_DATA:
+			event_mask |= LDC_EVENT_DATA_READY;
+			err = 0;
+			break;
+
+		case LDC_ERR:
+			err = process_error_frame(lp, p);
+			break;
+
+		default:
+			err = ldc_abort(lp);
+			break;
+		}
+
+		if (err < 0)
+			break;
+
+		new = lp->rx_head;
+		new += LDC_PACKET_SIZE;
+		if (new == (lp->rx_num_entries * LDC_PACKET_SIZE))
+			new = 0;
+		lp->rx_head = new;
+
+		err = __set_rx_head(lp, new);
+		if (err < 0) {
+			(void) ldc_abort(lp);
+			break;
+		}
+		if (lp->hs_state == LDC_HS_COMPLETE)
+			goto handshake_complete;
+	}
+
+out:
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	send_events(lp, event_mask);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t ldc_tx(int irq, void *dev_id)
+{
+	struct ldc_channel *lp = dev_id;
+	unsigned long flags, hv_err, orig_state;
+	unsigned int event_mask = 0;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	orig_state = lp->chan_state;
+	hv_err = sun4v_ldc_tx_get_state(lp->id,
+					&lp->tx_head,
+					&lp->tx_tail,
+					&lp->chan_state);
+
+	ldcdbg(TX, " TX state[0x%02lx:0x%02lx] head[0x%04lx] tail[0x%04lx]\n",
+	       orig_state, lp->chan_state, lp->tx_head, lp->tx_tail);
+
+	if (lp->cfg.mode == LDC_MODE_RAW &&
+	    lp->chan_state == LDC_CHANNEL_UP) {
+		lp->hs_state = LDC_HS_COMPLETE;
+		ldc_set_state(lp, LDC_STATE_CONNECTED);
+
+		event_mask |= LDC_EVENT_UP;
+	}
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	send_events(lp, event_mask);
+
+	return IRQ_HANDLED;
+}
+
+/* XXX ldc_alloc() and ldc_free() needs to run under a mutex so
+ * XXX that addition and removal from the ldc_channel_list has
+ * XXX atomicity, otherwise the __ldc_channel_exists() check is
+ * XXX totally pointless as another thread can slip into ldc_alloc()
+ * XXX and add a channel with the same ID.  There also needs to be
+ * XXX a spinlock for ldc_channel_list.
+ */
+static HLIST_HEAD(ldc_channel_list);
+
+static int __ldc_channel_exists(unsigned long id)
+{
+	struct ldc_channel *lp;
+	struct hlist_node *n;
+
+	hlist_for_each_entry(lp, n, &ldc_channel_list, list) {
+		if (lp->id == id)
+			return 1;
+	}
+	return 0;
+}
+
+static int alloc_queue(const char *name, unsigned long num_entries,
+		       struct ldc_packet **base, unsigned long *ra)
+{
+	unsigned long size, order;
+	void *q;
+
+	size = num_entries * LDC_PACKET_SIZE;
+	order = get_order(size);
+
+	q = (void *) __get_free_pages(GFP_KERNEL, order);
+	if (!q) {
+		printk(KERN_ERR PFX "Alloc of %s queue failed with "
+		       "size=%lu order=%lu\n", name, size, order);
+		return -ENOMEM;
+	}
+
+	memset(q, 0, PAGE_SIZE << order);
+
+	*base = q;
+	*ra = __pa(q);
+
+	return 0;
+}
+
+static void free_queue(unsigned long num_entries, struct ldc_packet *q)
+{
+	unsigned long size, order;
+
+	if (!q)
+		return;
+
+	size = num_entries * LDC_PACKET_SIZE;
+	order = get_order(size);
+
+	free_pages((unsigned long)q, order);
+}
+
+/* XXX Make this configurable... XXX */
+#define LDC_IOTABLE_SIZE	(8 * 1024)
+
+static int ldc_iommu_init(struct ldc_channel *lp)
+{
+	unsigned long sz, num_tsb_entries, tsbsize, order;
+	struct ldc_iommu *iommu = &lp->iommu;
+	struct ldc_mtable_entry *table;
+	unsigned long hv_err;
+	int err;
+
+	num_tsb_entries = LDC_IOTABLE_SIZE;
+	tsbsize = num_tsb_entries * sizeof(struct ldc_mtable_entry);
+
+	spin_lock_init(&iommu->lock);
+
+	sz = num_tsb_entries / 8;
+	sz = (sz + 7UL) & ~7UL;
+	iommu->arena.map = kzalloc(sz, GFP_KERNEL);
+	if (!iommu->arena.map) {
+		printk(KERN_ERR PFX "Alloc of arena map failed, sz=%lu\n", sz);
+		return -ENOMEM;
+	}
+
+	iommu->arena.limit = num_tsb_entries;
+
+	order = get_order(tsbsize);
+
+	table = (struct ldc_mtable_entry *)
+		__get_free_pages(GFP_KERNEL, order);
+	err = -ENOMEM;
+	if (!table) {
+		printk(KERN_ERR PFX "Alloc of MTE table failed, "
+		       "size=%lu order=%lu\n", tsbsize, order);
+		goto out_free_map;
+	}
+
+	memset(table, 0, PAGE_SIZE << order);
+
+	iommu->page_table = table;
+
+	hv_err = sun4v_ldc_set_map_table(lp->id, __pa(table),
+					 num_tsb_entries);
+	err = -EINVAL;
+	if (hv_err)
+		goto out_free_table;
+
+	return 0;
+
+out_free_table:
+	free_pages((unsigned long) table, order);
+	iommu->page_table = NULL;
+
+out_free_map:
+	kfree(iommu->arena.map);
+	iommu->arena.map = NULL;
+
+	return err;
+}
+
+static void ldc_iommu_release(struct ldc_channel *lp)
+{
+	struct ldc_iommu *iommu = &lp->iommu;
+	unsigned long num_tsb_entries, tsbsize, order;
+
+	(void) sun4v_ldc_set_map_table(lp->id, 0, 0);
+
+	num_tsb_entries = iommu->arena.limit;
+	tsbsize = num_tsb_entries * sizeof(struct ldc_mtable_entry);
+	order = get_order(tsbsize);
+
+	free_pages((unsigned long) iommu->page_table, order);
+	iommu->page_table = NULL;
+
+	kfree(iommu->arena.map);
+	iommu->arena.map = NULL;
+}
+
+struct ldc_channel *ldc_alloc(unsigned long id,
+			      const struct ldc_channel_config *cfgp,
+			      void *event_arg)
+{
+	struct ldc_channel *lp;
+	const struct ldc_mode_ops *mops;
+	unsigned long dummy1, dummy2, hv_err;
+	u8 mss, *mssbuf;
+	int err;
+
+	err = -ENODEV;
+	if (!ldom_domaining_enabled)
+		goto out_err;
+
+	err = -EINVAL;
+	if (!cfgp)
+		goto out_err;
+
+	switch (cfgp->mode) {
+	case LDC_MODE_RAW:
+		mops = &raw_ops;
+		mss = LDC_PACKET_SIZE;
+		break;
+
+	case LDC_MODE_UNRELIABLE:
+		mops = &nonraw_ops;
+		mss = LDC_PACKET_SIZE - 8;
+		break;
+
+	case LDC_MODE_STREAM:
+		mops = &stream_ops;
+		mss = LDC_PACKET_SIZE - 8 - 8;
+		break;
+
+	default:
+		goto out_err;
+	}
+
+	if (!cfgp->event || !event_arg || !cfgp->rx_irq || !cfgp->tx_irq)
+		goto out_err;
+
+	hv_err = sun4v_ldc_tx_qinfo(id, &dummy1, &dummy2);
+	err = -ENODEV;
+	if (hv_err == HV_ECHANNEL)
+		goto out_err;
+
+	err = -EEXIST;
+	if (__ldc_channel_exists(id))
+		goto out_err;
+
+	mssbuf = NULL;
+
+	lp = kzalloc(sizeof(*lp), GFP_KERNEL);
+	err = -ENOMEM;
+	if (!lp)
+		goto out_err;
+
+	spin_lock_init(&lp->lock);
+
+	lp->id = id;
+
+	err = ldc_iommu_init(lp);
+	if (err)
+		goto out_free_ldc;
+
+	lp->mops = mops;
+	lp->mss = mss;
+
+	lp->cfg = *cfgp;
+	if (!lp->cfg.mtu)
+		lp->cfg.mtu = LDC_DEFAULT_MTU;
+
+	if (lp->cfg.mode == LDC_MODE_STREAM) {
+		mssbuf = kzalloc(lp->cfg.mtu, GFP_KERNEL);
+		if (!mssbuf) {
+			err = -ENOMEM;
+			goto out_free_iommu;
+		}
+		lp->mssbuf = mssbuf;
+	}
+
+	lp->event_arg = event_arg;
+
+	/* XXX allow setting via ldc_channel_config to override defaults
+	 * XXX or use some formula based upon mtu
+	 */
+	lp->tx_num_entries = LDC_DEFAULT_NUM_ENTRIES;
+	lp->rx_num_entries = LDC_DEFAULT_NUM_ENTRIES;
+
+	err = alloc_queue("TX", lp->tx_num_entries,
+			  &lp->tx_base, &lp->tx_ra);
+	if (err)
+		goto out_free_mssbuf;
+
+	err = alloc_queue("RX", lp->rx_num_entries,
+			  &lp->rx_base, &lp->rx_ra);
+	if (err)
+		goto out_free_txq;
+
+	lp->flags |= LDC_FLAG_ALLOCED_QUEUES;
+
+	lp->hs_state = LDC_HS_CLOSED;
+	ldc_set_state(lp, LDC_STATE_INIT);
+
+	INIT_HLIST_NODE(&lp->list);
+	hlist_add_head(&lp->list, &ldc_channel_list);
+
+	INIT_HLIST_HEAD(&lp->mh_list);
+
+	return lp;
+
+out_free_txq:
+	free_queue(lp->tx_num_entries, lp->tx_base);
+
+out_free_mssbuf:
+	if (mssbuf)
+		kfree(mssbuf);
+
+out_free_iommu:
+	ldc_iommu_release(lp);
+
+out_free_ldc:
+	kfree(lp);
+
+out_err:
+	return ERR_PTR(err);
+}
+EXPORT_SYMBOL(ldc_alloc);
+
+void ldc_free(struct ldc_channel *lp)
+{
+	if (lp->flags & LDC_FLAG_REGISTERED_IRQS) {
+		free_irq(lp->cfg.rx_irq, lp);
+		free_irq(lp->cfg.tx_irq, lp);
+	}
+
+	if (lp->flags & LDC_FLAG_REGISTERED_QUEUES) {
+		sun4v_ldc_tx_qconf(lp->id, 0, 0);
+		sun4v_ldc_rx_qconf(lp->id, 0, 0);
+		lp->flags &= ~LDC_FLAG_REGISTERED_QUEUES;
+	}
+	if (lp->flags & LDC_FLAG_ALLOCED_QUEUES) {
+		free_queue(lp->tx_num_entries, lp->tx_base);
+		free_queue(lp->rx_num_entries, lp->rx_base);
+		lp->flags &= ~LDC_FLAG_ALLOCED_QUEUES;
+	}
+
+	hlist_del(&lp->list);
+
+	if (lp->mssbuf)
+		kfree(lp->mssbuf);
+
+	ldc_iommu_release(lp);
+
+	kfree(lp);
+}
+EXPORT_SYMBOL(ldc_free);
+
+/* Bind the channel.  This registers the LDC queues with
+ * the hypervisor and puts the channel into a pseudo-listening
+ * state.  This does not initiate a handshake, ldc_connect() does
+ * that.
+ */
+int ldc_bind(struct ldc_channel *lp, const char *name)
+{
+	unsigned long hv_err, flags;
+	int err = -EINVAL;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	if (!name)
+		goto out_err;
+
+	if (lp->state != LDC_STATE_INIT)
+		goto out_err;
+
+	snprintf(lp->rx_irq_name, LDC_IRQ_NAME_MAX, "%s RX", name);
+	snprintf(lp->tx_irq_name, LDC_IRQ_NAME_MAX, "%s TX", name);
+
+	err = request_irq(lp->cfg.rx_irq, ldc_rx,
+			  IRQF_SAMPLE_RANDOM | IRQF_SHARED,
+			  lp->rx_irq_name, lp);
+	if (err)
+		goto out_err;
+
+	err = request_irq(lp->cfg.tx_irq, ldc_tx,
+			  IRQF_SAMPLE_RANDOM | IRQF_SHARED,
+			  lp->tx_irq_name, lp);
+	if (err)
+		goto out_free_rx_irq;
+
+
+	lp->flags |= LDC_FLAG_REGISTERED_IRQS;
+
+	err = -ENODEV;
+	hv_err = sun4v_ldc_tx_qconf(lp->id, 0, 0);
+	if (hv_err)
+		goto out_free_tx_irq;
+
+	hv_err = sun4v_ldc_tx_qconf(lp->id, lp->tx_ra, lp->tx_num_entries);
+	if (hv_err)
+		goto out_free_tx_irq;
+
+	hv_err = sun4v_ldc_rx_qconf(lp->id, 0, 0);
+	if (hv_err)
+		goto out_unmap_tx;
+
+	hv_err = sun4v_ldc_rx_qconf(lp->id, lp->rx_ra, lp->rx_num_entries);
+	if (hv_err)
+		goto out_unmap_tx;
+
+	lp->flags |= LDC_FLAG_REGISTERED_QUEUES;
+
+	hv_err = sun4v_ldc_tx_get_state(lp->id,
+					&lp->tx_head,
+					&lp->tx_tail,
+					&lp->chan_state);
+	err = -EBUSY;
+	if (hv_err)
+		goto out_unmap_rx;
+
+	lp->tx_acked = lp->tx_head;
+
+	lp->hs_state = LDC_HS_OPEN;
+	ldc_set_state(lp, LDC_STATE_BOUND);
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return 0;
+
+out_unmap_rx:
+	lp->flags &= ~LDC_FLAG_REGISTERED_QUEUES;
+	sun4v_ldc_rx_qconf(lp->id, 0, 0);
+
+out_unmap_tx:
+	sun4v_ldc_tx_qconf(lp->id, 0, 0);
+
+out_free_tx_irq:
+	lp->flags &= ~LDC_FLAG_REGISTERED_IRQS;
+	free_irq(lp->cfg.tx_irq, lp);
+
+out_free_rx_irq:
+	free_irq(lp->cfg.rx_irq, lp);
+
+out_err:
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return err;
+}
+EXPORT_SYMBOL(ldc_bind);
+
+int ldc_connect(struct ldc_channel *lp)
+{
+	unsigned long flags;
+	int err;
+
+	if (lp->cfg.mode == LDC_MODE_RAW)
+		return -EINVAL;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	if (!(lp->flags & LDC_FLAG_ALLOCED_QUEUES) ||
+	    !(lp->flags & LDC_FLAG_REGISTERED_QUEUES) ||
+	    lp->hs_state != LDC_HS_OPEN)
+		err = -EINVAL;
+	else
+		err = start_handshake(lp);
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return err;
+}
+EXPORT_SYMBOL(ldc_connect);
+
+int ldc_disconnect(struct ldc_channel *lp)
+{
+	unsigned long hv_err, flags;
+	int err;
+
+	if (lp->cfg.mode == LDC_MODE_RAW)
+		return -EINVAL;
+
+	if (!(lp->flags & LDC_FLAG_ALLOCED_QUEUES) ||
+	    !(lp->flags & LDC_FLAG_REGISTERED_QUEUES))
+		return -EINVAL;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	err = -ENODEV;
+	hv_err = sun4v_ldc_tx_qconf(lp->id, 0, 0);
+	if (hv_err)
+		goto out_err;
+
+	hv_err = sun4v_ldc_tx_qconf(lp->id, lp->tx_ra, lp->tx_num_entries);
+	if (hv_err)
+		goto out_err;
+
+	hv_err = sun4v_ldc_rx_qconf(lp->id, 0, 0);
+	if (hv_err)
+		goto out_err;
+
+	hv_err = sun4v_ldc_rx_qconf(lp->id, lp->rx_ra, lp->rx_num_entries);
+	if (hv_err)
+		goto out_err;
+
+	ldc_set_state(lp, LDC_STATE_BOUND);
+	lp->hs_state = LDC_HS_OPEN;
+	lp->flags |= LDC_FLAG_RESET;
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return 0;
+
+out_err:
+	sun4v_ldc_tx_qconf(lp->id, 0, 0);
+	sun4v_ldc_rx_qconf(lp->id, 0, 0);
+	free_irq(lp->cfg.tx_irq, lp);
+	free_irq(lp->cfg.rx_irq, lp);
+	lp->flags &= ~(LDC_FLAG_REGISTERED_IRQS |
+		       LDC_FLAG_REGISTERED_QUEUES);
+	ldc_set_state(lp, LDC_STATE_INIT);
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return err;
+}
+EXPORT_SYMBOL(ldc_disconnect);
+
+int ldc_state(struct ldc_channel *lp)
+{
+	return lp->state;
+}
+EXPORT_SYMBOL(ldc_state);
+
+static int write_raw(struct ldc_channel *lp, const void *buf, unsigned int size)
+{
+	struct ldc_packet *p;
+	unsigned long new_tail;
+	int err;
+
+	if (size > LDC_PACKET_SIZE)
+		return -EMSGSIZE;
+
+	p = data_get_tx_packet(lp, &new_tail);
+	if (!p)
+		return -EAGAIN;
+
+	memcpy(p, buf, size);
+
+	err = send_tx_packet(lp, p, new_tail);
+	if (!err)
+		err = size;
+
+	return err;
+}
+
+static int read_raw(struct ldc_channel *lp, void *buf, unsigned int size)
+{
+	struct ldc_packet *p;
+	unsigned long hv_err, new;
+	int err;
+
+	if (size < LDC_PACKET_SIZE)
+		return -EINVAL;
+
+	hv_err = sun4v_ldc_rx_get_state(lp->id,
+					&lp->rx_head,
+					&lp->rx_tail,
+					&lp->chan_state);
+	if (hv_err)
+		return ldc_abort(lp);
+
+	if (lp->chan_state == LDC_CHANNEL_DOWN ||
+	    lp->chan_state == LDC_CHANNEL_RESETTING)
+		return -ECONNRESET;
+
+	if (lp->rx_head == lp->rx_tail)
+		return 0;
+
+	p = lp->rx_base + (lp->rx_head / LDC_PACKET_SIZE);
+	memcpy(buf, p, LDC_PACKET_SIZE);
+
+	new = rx_advance(lp, lp->rx_head);
+	lp->rx_head = new;
+
+	err = __set_rx_head(lp, new);
+	if (err < 0)
+		err = -ECONNRESET;
+	else
+		err = LDC_PACKET_SIZE;
+
+	return err;
+}
+
+static const struct ldc_mode_ops raw_ops = {
+	.write		=	write_raw,
+	.read		=	read_raw,
+};
+
+static int write_nonraw(struct ldc_channel *lp, const void *buf,
+			unsigned int size)
+{
+	unsigned long hv_err, tail;
+	unsigned int copied;
+	u32 seq;
+	int err;
+
+	hv_err = sun4v_ldc_tx_get_state(lp->id, &lp->tx_head, &lp->tx_tail,
+					&lp->chan_state);
+	if (unlikely(hv_err))
+		return -EBUSY;
+
+	if (unlikely(lp->chan_state != LDC_CHANNEL_UP))
+		return ldc_abort(lp);
+
+	if (!tx_has_space_for(lp, size))
+		return -EAGAIN;
+
+	seq = lp->snd_nxt;
+	copied = 0;
+	tail = lp->tx_tail;
+	while (copied < size) {
+		struct ldc_packet *p = lp->tx_base + (tail / LDC_PACKET_SIZE);
+		u8 *data = ((lp->cfg.mode == LDC_MODE_UNRELIABLE) ?
+			    p->u.u_data :
+			    p->u.r.r_data);
+		int data_len;
+
+		p->type = LDC_DATA;
+		p->stype = LDC_INFO;
+		p->ctrl = 0;
+
+		data_len = size - copied;
+		if (data_len > lp->mss)
+			data_len = lp->mss;
+
+		BUG_ON(data_len > LDC_LEN);
+
+		p->env = (data_len |
+			  (copied == 0 ? LDC_START : 0) |
+			  (data_len == size - copied ? LDC_STOP : 0));
+
+		p->seqid = ++seq;
+
+		ldcdbg(DATA, "SENT DATA [%02x:%02x:%02x:%02x:%08x]\n",
+		       p->type,
+		       p->stype,
+		       p->ctrl,
+		       p->env,
+		       p->seqid);
+
+		memcpy(data, buf, data_len);
+		buf += data_len;
+		copied += data_len;
+
+		tail = tx_advance(lp, tail);
+	}
+
+	err = set_tx_tail(lp, tail);
+	if (!err) {
+		lp->snd_nxt = seq;
+		err = size;
+	}
+
+	return err;
+}
+
+static int rx_bad_seq(struct ldc_channel *lp, struct ldc_packet *p,
+		      struct ldc_packet *first_frag)
+{
+	int err;
+
+	if (first_frag)
+		lp->rcv_nxt = first_frag->seqid - 1;
+
+	err = send_data_nack(lp, p);
+	if (err)
+		return err;
+
+	err = __set_rx_head(lp, lp->rx_tail);
+	if (err < 0)
+		return ldc_abort(lp);
+
+	return 0;
+}
+
+static int data_ack_nack(struct ldc_channel *lp, struct ldc_packet *p)
+{
+	if (p->stype & LDC_ACK) {
+		int err = process_data_ack(lp, p);
+		if (err)
+			return err;
+	}
+	if (p->stype & LDC_NACK)
+		return ldc_abort(lp);
+
+	return 0;
+}
+
+static int rx_data_wait(struct ldc_channel *lp, unsigned long cur_head)
+{
+	unsigned long dummy;
+	int limit = 1000;
+
+	ldcdbg(DATA, "DATA WAIT cur_head[%lx] rx_head[%lx] rx_tail[%lx]\n",
+	       cur_head, lp->rx_head, lp->rx_tail);
+	while (limit-- > 0) {
+		unsigned long hv_err;
+
+		hv_err = sun4v_ldc_rx_get_state(lp->id,
+						&dummy,
+						&lp->rx_tail,
+						&lp->chan_state);
+		if (hv_err)
+			return ldc_abort(lp);
+
+		if (lp->chan_state == LDC_CHANNEL_DOWN ||
+		    lp->chan_state == LDC_CHANNEL_RESETTING)
+			return -ECONNRESET;
+
+		if (cur_head != lp->rx_tail) {
+			ldcdbg(DATA, "DATA WAIT DONE "
+			       "head[%lx] tail[%lx] chan_state[%lx]\n",
+			       dummy, lp->rx_tail, lp->chan_state);
+			return 0;
+		}
+
+		udelay(1);
+	}
+	return -EAGAIN;
+}
+
+static int rx_set_head(struct ldc_channel *lp, unsigned long head)
+{
+	int err = __set_rx_head(lp, head);
+
+	if (err < 0)
+		return ldc_abort(lp);
+
+	lp->rx_head = head;
+	return 0;
+}
+
+static void send_data_ack(struct ldc_channel *lp)
+{
+	unsigned long new_tail;
+	struct ldc_packet *p;
+
+	p = data_get_tx_packet(lp, &new_tail);
+	if (likely(p)) {
+		int err;
+
+		memset(p, 0, sizeof(*p));
+		p->type = LDC_DATA;
+		p->stype = LDC_ACK;
+		p->ctrl = 0;
+		p->seqid = lp->snd_nxt + 1;
+		p->u.r.ackid = lp->rcv_nxt;
+
+		err = send_tx_packet(lp, p, new_tail);
+		if (!err)
+			lp->snd_nxt++;
+	}
+}
+
+static int read_nonraw(struct ldc_channel *lp, void *buf, unsigned int size)
+{
+	struct ldc_packet *first_frag;
+	unsigned long hv_err, new;
+	int err, copied;
+
+	hv_err = sun4v_ldc_rx_get_state(lp->id,
+					&lp->rx_head,
+					&lp->rx_tail,
+					&lp->chan_state);
+	if (hv_err)
+		return ldc_abort(lp);
+
+	if (lp->chan_state == LDC_CHANNEL_DOWN ||
+	    lp->chan_state == LDC_CHANNEL_RESETTING)
+		return -ECONNRESET;
+
+	if (lp->rx_head == lp->rx_tail)
+		return 0;
+
+	first_frag = NULL;
+	copied = err = 0;
+	new = lp->rx_head;
+	while (1) {
+		struct ldc_packet *p;
+		int pkt_len;
+
+		BUG_ON(new == lp->rx_tail);
+		p = lp->rx_base + (new / LDC_PACKET_SIZE);
+
+		ldcdbg(RX, "RX read pkt[%02x:%02x:%02x:%02x:%08x:%08x] "
+		       "rcv_nxt[%08x]\n",
+		       p->type,
+		       p->stype,
+		       p->ctrl,
+		       p->env,
+		       p->seqid,
+		       p->u.r.ackid,
+		       lp->rcv_nxt);
+
+		if (unlikely(!rx_seq_ok(lp, p->seqid))) {
+			err = rx_bad_seq(lp, p, first_frag);
+			copied = 0;
+			break;
+		}
+
+		if (p->type & LDC_CTRL) {
+			err = process_control_frame(lp, p);
+			if (err < 0)
+				break;
+			err = 0;
+		}
+
+		lp->rcv_nxt = p->seqid;
+
+		if (!(p->type & LDC_DATA)) {
+			new = rx_advance(lp, new);
+			goto no_data;
+		}
+		if (p->stype & (LDC_ACK | LDC_NACK)) {
+			err = data_ack_nack(lp, p);
+			if (err)
+				break;
+		}
+		if (!(p->stype & LDC_INFO)) {
+			new = rx_advance(lp, new);
+			err = rx_set_head(lp, new);
+			if (err)
+				break;
+			goto no_data;
+		}
+
+		pkt_len = p->env & LDC_LEN;
+
+		/* Every initial packet starts with the START bit set.
+		 *
+		 * Singleton packets will have both START+STOP set.
+		 *
+		 * Fragments will have START set in the first frame, STOP
+		 * set in the last frame, and neither bit set in middle
+		 * frames of the packet.
+		 *
+		 * Therefore if we are at the beginning of a packet and
+		 * we don't see START, or we are in the middle of a fragmented
+		 * packet and do see START, we are unsynchronized and should
+		 * flush the RX queue.
+		 */
+		if ((first_frag == NULL && !(p->env & LDC_START)) ||
+		    (first_frag != NULL &&  (p->env & LDC_START))) {
+			if (!first_frag)
+				new = rx_advance(lp, new);
+
+			err = rx_set_head(lp, new);
+			if (err)
+				break;
+
+			if (!first_frag)
+				goto no_data;
+		}
+		if (!first_frag)
+			first_frag = p;
+
+		if (pkt_len > size - copied) {
+			/* User didn't give us a big enough buffer,
+			 * what to do?  This is a pretty serious error.
+			 *
+			 * Since we haven't updated the RX ring head to
+			 * consume any of the packets, signal the error
+			 * to the user and just leave the RX ring alone.
+			 *
+			 * This seems the best behavior because this allows
+			 * a user of the LDC layer to start with a small
+			 * RX buffer for ldc_read() calls and use -EMSGSIZE
+			 * as a cue to enlarge it's read buffer.
+			 */
+			err = -EMSGSIZE;
+			break;
+		}
+
+		/* Ok, we are gonna eat this one.  */
+		new = rx_advance(lp, new);
+
+		memcpy(buf,
+		       (lp->cfg.mode == LDC_MODE_UNRELIABLE ?
+			p->u.u_data : p->u.r.r_data), pkt_len);
+		buf += pkt_len;
+		copied += pkt_len;
+
+		if (p->env & LDC_STOP)
+			break;
+
+no_data:
+		if (new == lp->rx_tail) {
+			err = rx_data_wait(lp, new);
+			if (err)
+				break;
+		}
+	}
+
+	if (!err)
+		err = rx_set_head(lp, new);
+
+	if (err && first_frag)
+		lp->rcv_nxt = first_frag->seqid - 1;
+
+	if (!err) {
+		err = copied;
+		if (err > 0 && lp->cfg.mode != LDC_MODE_UNRELIABLE)
+			send_data_ack(lp);
+	}
+
+	return err;
+}
+
+static const struct ldc_mode_ops nonraw_ops = {
+	.write		=	write_nonraw,
+	.read		=	read_nonraw,
+};
+
+static int write_stream(struct ldc_channel *lp, const void *buf,
+			unsigned int size)
+{
+	if (size > lp->cfg.mtu)
+		size = lp->cfg.mtu;
+	return write_nonraw(lp, buf, size);
+}
+
+static int read_stream(struct ldc_channel *lp, void *buf, unsigned int size)
+{
+	if (!lp->mssbuf_len) {
+		int err = read_nonraw(lp, lp->mssbuf, lp->cfg.mtu);
+		if (err < 0)
+			return err;
+
+		lp->mssbuf_len = err;
+		lp->mssbuf_off = 0;
+	}
+
+	if (size > lp->mssbuf_len)
+		size = lp->mssbuf_len;
+	memcpy(buf, lp->mssbuf + lp->mssbuf_off, size);
+
+	lp->mssbuf_off += size;
+	lp->mssbuf_len -= size;
+
+	return size;
+}
+
+static const struct ldc_mode_ops stream_ops = {
+	.write		=	write_stream,
+	.read		=	read_stream,
+};
+
+int ldc_write(struct ldc_channel *lp, const void *buf, unsigned int size)
+{
+	unsigned long flags;
+	int err;
+
+	if (!buf)
+		return -EINVAL;
+
+	if (!size)
+		return 0;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	if (lp->hs_state != LDC_HS_COMPLETE)
+		err = -ENOTCONN;
+	else
+		err = lp->mops->write(lp, buf, size);
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return err;
+}
+EXPORT_SYMBOL(ldc_write);
+
+int ldc_read(struct ldc_channel *lp, void *buf, unsigned int size)
+{
+	unsigned long flags;
+	int err;
+
+	if (!buf)
+		return -EINVAL;
+
+	if (!size)
+		return 0;
+
+	spin_lock_irqsave(&lp->lock, flags);
+
+	if (lp->hs_state != LDC_HS_COMPLETE)
+		err = -ENOTCONN;
+	else
+		err = lp->mops->read(lp, buf, size);
+
+	spin_unlock_irqrestore(&lp->lock, flags);
+
+	return err;
+}
+EXPORT_SYMBOL(ldc_read);
+
+static long arena_alloc(struct ldc_iommu *iommu, unsigned long npages)
+{
+	struct iommu_arena *arena = &iommu->arena;
+	unsigned long n, i, start, end, limit;
+	int pass;
+
+	limit = arena->limit;
+	start = arena->hint;
+	pass = 0;
+
+again:
+	n = find_next_zero_bit(arena->map, limit, start);
+	end = n + npages;
+	if (unlikely(end >= limit)) {
+		if (likely(pass < 1)) {
+			limit = start;
+			start = 0;
+			pass++;
+			goto again;
+		} else {
+			/* Scanned the whole thing, give up. */
+			return -1;
+		}
+	}
+
+	for (i = n; i < end; i++) {
+		if (test_bit(i, arena->map)) {
+			start = i + 1;
+			goto again;
+		}
+	}
+
+	for (i = n; i < end; i++)
+		__set_bit(i, arena->map);
+
+	arena->hint = end;
+
+	return n;
+}
+
+#define COOKIE_PGSZ_CODE	0xf000000000000000ULL
+#define COOKIE_PGSZ_CODE_SHIFT	60ULL
+
+static u64 pagesize_code(void)
+{
+	switch (PAGE_SIZE) {
+	default:
+	case (8ULL * 1024ULL):
+		return 0;
+	case (64ULL * 1024ULL):
+		return 1;
+	case (512ULL * 1024ULL):
+		return 2;
+	case (4ULL * 1024ULL * 1024ULL):
+		return 3;
+	case (32ULL * 1024ULL * 1024ULL):
+		return 4;
+	case (256ULL * 1024ULL * 1024ULL):
+		return 5;
+	}
+}
+
+static u64 make_cookie(u64 index, u64 pgsz_code, u64 page_offset)
+{
+	return ((pgsz_code << COOKIE_PGSZ_CODE_SHIFT) |
+		(index << PAGE_SHIFT) |
+		page_offset);
+}
+
+static u64 cookie_to_index(u64 cookie, unsigned long *shift)
+{
+	u64 szcode = cookie >> COOKIE_PGSZ_CODE_SHIFT;
+
+	cookie &= ~COOKIE_PGSZ_CODE;
+
+	*shift = szcode * 3;
+
+	return (cookie >> (13ULL + (szcode * 3ULL)));
+}
+
+static struct ldc_mtable_entry *alloc_npages(struct ldc_iommu *iommu,
+					     unsigned long npages)
+{
+	long entry;
+
+	entry = arena_alloc(iommu, npages);
+	if (unlikely(entry < 0))
+		return NULL;
+
+	return iommu->page_table + entry;
+}
+
+static u64 perm_to_mte(unsigned int map_perm)
+{
+	u64 mte_base;
+
+	mte_base = pagesize_code();
+
+	if (map_perm & LDC_MAP_SHADOW) {
+		if (map_perm & LDC_MAP_R)
+			mte_base |= LDC_MTE_COPY_R;
+		if (map_perm & LDC_MAP_W)
+			mte_base |= LDC_MTE_COPY_W;
+	}
+	if (map_perm & LDC_MAP_DIRECT) {
+		if (map_perm & LDC_MAP_R)
+			mte_base |= LDC_MTE_READ;
+		if (map_perm & LDC_MAP_W)
+			mte_base |= LDC_MTE_WRITE;
+		if (map_perm & LDC_MAP_X)
+			mte_base |= LDC_MTE_EXEC;
+	}
+	if (map_perm & LDC_MAP_IO) {
+		if (map_perm & LDC_MAP_R)
+			mte_base |= LDC_MTE_IOMMU_R;
+		if (map_perm & LDC_MAP_W)
+			mte_base |= LDC_MTE_IOMMU_W;
+	}
+
+	return mte_base;
+}
+
+static int pages_in_region(unsigned long base, long len)
+{
+	int count = 0;
+
+	do {
+		unsigned long new = (base + PAGE_SIZE) & PAGE_MASK;
+
+		len -= (new - base);
+		base = new;
+		count++;
+	} while (len > 0);
+
+	return count;
+}
+
+struct cookie_state {
+	struct ldc_mtable_entry		*page_table;
+	struct ldc_trans_cookie		*cookies;
+	u64				mte_base;
+	u64				prev_cookie;
+	u32				pte_idx;
+	u32				nc;
+};
+
+static void fill_cookies(struct cookie_state *sp, unsigned long pa,
+			 unsigned long off, unsigned long len)
+{
+	do {
+		unsigned long tlen, new = pa + PAGE_SIZE;
+		u64 this_cookie;
+
+		sp->page_table[sp->pte_idx].mte = sp->mte_base | pa;
+
+		tlen = PAGE_SIZE;
+		if (off)
+			tlen = PAGE_SIZE - off;
+		if (tlen > len)
+			tlen = len;
+
+		this_cookie = make_cookie(sp->pte_idx,
+					  pagesize_code(), off);
+
+		off = 0;
+
+		if (this_cookie == sp->prev_cookie) {
+			sp->cookies[sp->nc - 1].cookie_size += tlen;
+		} else {
+			sp->cookies[sp->nc].cookie_addr = this_cookie;
+			sp->cookies[sp->nc].cookie_size = tlen;
+			sp->nc++;
+		}
+		sp->prev_cookie = this_cookie + tlen;
+
+		sp->pte_idx++;
+
+		len -= tlen;
+		pa = new;
+	} while (len > 0);
+}
+
+static int sg_count_one(struct scatterlist *sg)
+{
+	unsigned long base = page_to_pfn(sg->page) << PAGE_SHIFT;
+	long len = sg->length;
+
+	if ((sg->offset | len) & (8UL - 1))
+		return -EFAULT;
+
+	return pages_in_region(base + sg->offset, len);
+}
+
+static int sg_count_pages(struct scatterlist *sg, int num_sg)
+{
+	int count;
+	int i;
+
+	count = 0;
+	for (i = 0; i < num_sg; i++) {
+		int err = sg_count_one(sg + i);
+		if (err < 0)
+			return err;
+		count += err;
+	}
+
+	return count;
+}
+
+int ldc_map_sg(struct ldc_channel *lp,
+	       struct scatterlist *sg, int num_sg,
+	       struct ldc_trans_cookie *cookies, int ncookies,
+	       unsigned int map_perm)
+{
+	unsigned long i, npages, flags;
+	struct ldc_mtable_entry *base;
+	struct cookie_state state;
+	struct ldc_iommu *iommu;
+	int err;
+
+	if (map_perm & ~LDC_MAP_ALL)
+		return -EINVAL;
+
+	err = sg_count_pages(sg, num_sg);
+	if (err < 0)
+		return err;
+
+	npages = err;
+	if (err > ncookies)
+		return -EMSGSIZE;
+
+	iommu = &lp->iommu;
+
+	spin_lock_irqsave(&iommu->lock, flags);
+	base = alloc_npages(iommu, npages);
+	spin_unlock_irqrestore(&iommu->lock, flags);
+
+	if (!base)
+		return -ENOMEM;
+
+	state.page_table = iommu->page_table;
+	state.cookies = cookies;
+	state.mte_base = perm_to_mte(map_perm);
+	state.prev_cookie = ~(u64)0;
+	state.pte_idx = (base - iommu->page_table);
+	state.nc = 0;
+
+	for (i = 0; i < num_sg; i++)
+		fill_cookies(&state, page_to_pfn(sg[i].page) << PAGE_SHIFT,
+			     sg[i].offset, sg[i].length);
+
+	return state.nc;
+}
+EXPORT_SYMBOL(ldc_map_sg);
+
+int ldc_map_single(struct ldc_channel *lp,
+		   void *buf, unsigned int len,
+		   struct ldc_trans_cookie *cookies, int ncookies,
+		   unsigned int map_perm)
+{
+	unsigned long npages, pa, flags;
+	struct ldc_mtable_entry *base;
+	struct cookie_state state;
+	struct ldc_iommu *iommu;
+
+	if ((map_perm & ~LDC_MAP_ALL) || (ncookies < 1))
+		return -EINVAL;
+
+	pa = __pa(buf);
+	if ((pa | len) & (8UL - 1))
+		return -EFAULT;
+
+	npages = pages_in_region(pa, len);
+
+	iommu = &lp->iommu;
+
+	spin_lock_irqsave(&iommu->lock, flags);
+	base = alloc_npages(iommu, npages);
+	spin_unlock_irqrestore(&iommu->lock, flags);
+
+	if (!base)
+		return -ENOMEM;
+
+	state.page_table = iommu->page_table;
+	state.cookies = cookies;
+	state.mte_base = perm_to_mte(map_perm);
+	state.prev_cookie = ~(u64)0;
+	state.pte_idx = (base - iommu->page_table);
+	state.nc = 0;
+	fill_cookies(&state, (pa & PAGE_MASK), (pa & ~PAGE_MASK), len);
+	BUG_ON(state.nc != 1);
+
+	return state.nc;
+}
+EXPORT_SYMBOL(ldc_map_single);
+
+static void free_npages(unsigned long id, struct ldc_iommu *iommu,
+			u64 cookie, u64 size)
+{
+	struct iommu_arena *arena = &iommu->arena;
+	unsigned long i, shift, index, npages;
+	struct ldc_mtable_entry *base;
+
+	npages = PAGE_ALIGN(((cookie & ~PAGE_MASK) + size)) >> PAGE_SHIFT;
+	index = cookie_to_index(cookie, &shift);
+	base = iommu->page_table + index;
+
+	BUG_ON(index > arena->limit ||
+	       (index + npages) > arena->limit);
+
+	for (i = 0; i < npages; i++) {
+		if (base->cookie)
+			sun4v_ldc_revoke(id, cookie + (i << shift),
+					 base->cookie);
+		base->mte = 0;
+		__clear_bit(index + i, arena->map);
+	}
+}
+
+void ldc_unmap(struct ldc_channel *lp, struct ldc_trans_cookie *cookies,
+	       int ncookies)
+{
+	struct ldc_iommu *iommu = &lp->iommu;
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&iommu->lock, flags);
+	for (i = 0; i < ncookies; i++) {
+		u64 addr = cookies[i].cookie_addr;
+		u64 size = cookies[i].cookie_size;
+
+		free_npages(lp->id, iommu, addr, size);
+	}
+	spin_unlock_irqrestore(&iommu->lock, flags);
+}
+EXPORT_SYMBOL(ldc_unmap);
+
+int ldc_copy(struct ldc_channel *lp, int copy_dir,
+	     void *buf, unsigned int len, unsigned long offset,
+	     struct ldc_trans_cookie *cookies, int ncookies)
+{
+	unsigned int orig_len;
+	unsigned long ra;
+	int i;
+
+	if (copy_dir != LDC_COPY_IN && copy_dir != LDC_COPY_OUT) {
+		printk(KERN_ERR PFX "ldc_copy: ID[%lu] Bad copy_dir[%d]\n",
+		       lp->id, copy_dir);
+		return -EINVAL;
+	}
+
+	ra = __pa(buf);
+	if ((ra | len | offset) & (8UL - 1)) {
+		printk(KERN_ERR PFX "ldc_copy: ID[%lu] Unaligned buffer "
+		       "ra[%lx] len[%x] offset[%lx]\n",
+		       lp->id, ra, len, offset);
+		return -EFAULT;
+	}
+
+	if (lp->hs_state != LDC_HS_COMPLETE ||
+	    (lp->flags & LDC_FLAG_RESET)) {
+		printk(KERN_ERR PFX "ldc_copy: ID[%lu] Link down hs_state[%x] "
+		       "flags[%x]\n", lp->id, lp->hs_state, lp->flags);
+		return -ECONNRESET;
+	}
+
+	orig_len = len;
+	for (i = 0; i < ncookies; i++) {
+		unsigned long cookie_raddr = cookies[i].cookie_addr;
+		unsigned long this_len = cookies[i].cookie_size;
+		unsigned long actual_len;
+
+		if (unlikely(offset)) {
+			unsigned long this_off = offset;
+
+			if (this_off > this_len)
+				this_off = this_len;
+
+			offset -= this_off;
+			this_len -= this_off;
+			if (!this_len)
+				continue;
+			cookie_raddr += this_off;
+		}
+
+		if (this_len > len)
+			this_len = len;
+
+		while (1) {
+			unsigned long hv_err;
+
+			hv_err = sun4v_ldc_copy(lp->id, copy_dir,
+						cookie_raddr, ra,
+						this_len, &actual_len);
+			if (unlikely(hv_err)) {
+				printk(KERN_ERR PFX "ldc_copy: ID[%lu] "
+				       "HV error %lu\n",
+				       lp->id, hv_err);
+				if (lp->hs_state != LDC_HS_COMPLETE ||
+				    (lp->flags & LDC_FLAG_RESET))
+					return -ECONNRESET;
+				else
+					return -EFAULT;
+			}
+
+			cookie_raddr += actual_len;
+			ra += actual_len;
+			len -= actual_len;
+			if (actual_len == this_len)
+				break;
+
+			this_len -= actual_len;
+		}
+
+		if (!len)
+			break;
+	}
+
+	/* It is caller policy what to do about short copies.
+	 * For example, a networking driver can declare the
+	 * packet a runt and drop it.
+	 */
+
+	return orig_len - len;
+}
+EXPORT_SYMBOL(ldc_copy);
+
+void *ldc_alloc_exp_dring(struct ldc_channel *lp, unsigned int len,
+			  struct ldc_trans_cookie *cookies, int *ncookies,
+			  unsigned int map_perm)
+{
+	void *buf;
+	int err;
+
+	if (len & (8UL - 1))
+		return ERR_PTR(-EINVAL);
+
+	buf = kzalloc(len, GFP_KERNEL);
+	if (!buf)
+		return ERR_PTR(-ENOMEM);
+
+	err = ldc_map_single(lp, buf, len, cookies, *ncookies, map_perm);
+	if (err < 0) {
+		kfree(buf);
+		return ERR_PTR(err);
+	}
+	*ncookies = err;
+
+	return buf;
+}
+EXPORT_SYMBOL(ldc_alloc_exp_dring);
+
+void ldc_free_exp_dring(struct ldc_channel *lp, void *buf, unsigned int len,
+			struct ldc_trans_cookie *cookies, int ncookies)
+{
+	ldc_unmap(lp, cookies, ncookies);
+	kfree(buf);
+}
+EXPORT_SYMBOL(ldc_free_exp_dring);
+
+static int __init ldc_init(void)
+{
+	unsigned long major, minor;
+	struct mdesc_handle *hp;
+	const u64 *v;
+	u64 mp;
+
+	hp = mdesc_grab();
+	if (!hp)
+		return -ENODEV;
+
+	mp = mdesc_node_by_name(hp, MDESC_NODE_NULL, "platform");
+	if (mp == MDESC_NODE_NULL)
+		return -ENODEV;
+
+	v = mdesc_get_property(hp, mp, "domaining-enabled", NULL);
+	if (!v)
+		return -ENODEV;
+
+	major = 1;
+	minor = 0;
+	if (sun4v_hvapi_register(HV_GRP_LDOM, major, &minor)) {
+		printk(KERN_INFO PFX "Could not register LDOM hvapi.\n");
+		return -ENODEV;
+	}
+
+	printk(KERN_INFO "%s", version);
+
+	if (!*v) {
+		printk(KERN_INFO PFX "Domaining disabled.\n");
+		return -ENODEV;
+	}
+	ldom_domaining_enabled = 1;
+
+	return 0;
+}
+
+core_initcall(ldc_init);

arch/sparc64/kernel/mdesc.c

@@ -6,6 +6,9 @@
 #include <linux/types.h>
 #include <linux/bootmem.h>
 #include <linux/log2.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
 
 #include <asm/hypervisor.h>
 #include <asm/mdesc.h>
@@ -29,7 +32,7 @@ struct mdesc_hdr {
 	u32	node_sz; /* node block size */
 	u32	name_sz; /* name block size */
 	u32	data_sz; /* data block size */
-};
+} __attribute__((aligned(16)));
 
 struct mdesc_elem {
 	u8	tag;
@@ -53,306 +56,402 @@ struct mdesc_elem {
 	} d;
 };
 
-static struct mdesc_hdr *main_mdesc;
-static struct mdesc_node *allnodes;
-
-static struct mdesc_node *allnodes_tail;
-static unsigned int unique_id;
+struct mdesc_mem_ops {
+	struct mdesc_handle *(*alloc)(unsigned int mdesc_size);
+	void (*free)(struct mdesc_handle *handle);
+};
 
-static struct mdesc_node **mdesc_hash;
-static unsigned int mdesc_hash_size;
+struct mdesc_handle {
+	struct list_head	list;
+	struct mdesc_mem_ops	*mops;
+	void			*self_base;
+	atomic_t		refcnt;
+	unsigned int		handle_size;
+	struct mdesc_hdr	mdesc;
+};
 
-static inline unsigned int node_hashfn(u64 node)
+static void mdesc_handle_init(struct mdesc_handle *hp,
+			      unsigned int handle_size,
+			      void *base)
 {
-	return ((unsigned int) (node ^ (node >> 8) ^ (node >> 16)))
-		& (mdesc_hash_size - 1);
+	BUG_ON(((unsigned long)&hp->mdesc) & (16UL - 1));
+
+	memset(hp, 0, handle_size);
+	INIT_LIST_HEAD(&hp->list);
+	hp->self_base = base;
+	atomic_set(&hp->refcnt, 1);
+	hp->handle_size = handle_size;
 }
 
-static inline void hash_node(struct mdesc_node *mp)
+static struct mdesc_handle *mdesc_bootmem_alloc(unsigned int mdesc_size)
 {
-	struct mdesc_node **head = &mdesc_hash[node_hashfn(mp->node)];
+	struct mdesc_handle *hp;
+	unsigned int handle_size, alloc_size;
 
-	mp->hash_next = *head;
-	*head = mp;
+	handle_size = (sizeof(struct mdesc_handle) -
+		       sizeof(struct mdesc_hdr) +
+		       mdesc_size);
+	alloc_size = PAGE_ALIGN(handle_size);
 
-	if (allnodes_tail) {
-		allnodes_tail->allnodes_next = mp;
-		allnodes_tail = mp;
-	} else {
-		allnodes = allnodes_tail = mp;
-	}
+	hp = __alloc_bootmem(alloc_size, PAGE_SIZE, 0UL);
+	if (hp)
+		mdesc_handle_init(hp, handle_size, hp);
+
+	return hp;
 }
 
-static struct mdesc_node *find_node(u64 node)
+static void mdesc_bootmem_free(struct mdesc_handle *hp)
 {
-	struct mdesc_node *mp = mdesc_hash[node_hashfn(node)];
+	unsigned int alloc_size, handle_size = hp->handle_size;
+	unsigned long start, end;
+
+	BUG_ON(atomic_read(&hp->refcnt) != 0);
+	BUG_ON(!list_empty(&hp->list));
 
-	while (mp) {
-		if (mp->node == node)
-			return mp;
+	alloc_size = PAGE_ALIGN(handle_size);
 
-		mp = mp->hash_next;
+	start = (unsigned long) hp;
+	end = start + alloc_size;
+
+	while (start < end) {
+		struct page *p;
+
+		p = virt_to_page(start);
+		ClearPageReserved(p);
+		__free_page(p);
+		start += PAGE_SIZE;
 	}
-	return NULL;
 }
 
-struct property *md_find_property(const struct mdesc_node *mp,
-				  const char *name,
-				  int *lenp)
+static struct mdesc_mem_ops bootmem_mdesc_memops = {
+	.alloc = mdesc_bootmem_alloc,
+	.free  = mdesc_bootmem_free,
+};
+
+static struct mdesc_handle *mdesc_kmalloc(unsigned int mdesc_size)
 {
-	struct property *pp;
+	unsigned int handle_size;
+	void *base;
 
-	for (pp = mp->properties; pp != 0; pp = pp->next) {
-		if (strcasecmp(pp->name, name) == 0) {
-			if (lenp)
-				*lenp = pp->length;
-			break;
-		}
+	handle_size = (sizeof(struct mdesc_handle) -
+		       sizeof(struct mdesc_hdr) +
+		       mdesc_size);
+
+	base = kmalloc(handle_size + 15, GFP_KERNEL);
+	if (base) {
+		struct mdesc_handle *hp;
+		unsigned long addr;
+
+		addr = (unsigned long)base;
+		addr = (addr + 15UL) & ~15UL;
+		hp = (struct mdesc_handle *) addr;
+
+		mdesc_handle_init(hp, handle_size, base);
+		return hp;
 	}
-	return pp;
+
+	return NULL;
 }
-EXPORT_SYMBOL(md_find_property);
 
-/*
- * Find a property with a given name for a given node
- * and return the value.
- */
-const void *md_get_property(const struct mdesc_node *mp, const char *name,
-			    int *lenp)
+static void mdesc_kfree(struct mdesc_handle *hp)
 {
-	struct property *pp = md_find_property(mp, name, lenp);
-	return pp ? pp->value : NULL;
+	BUG_ON(atomic_read(&hp->refcnt) != 0);
+	BUG_ON(!list_empty(&hp->list));
+
+	kfree(hp->self_base);
 }
-EXPORT_SYMBOL(md_get_property);
 
-struct mdesc_node *md_find_node_by_name(struct mdesc_node *from,
-					const char *name)
+static struct mdesc_mem_ops kmalloc_mdesc_memops = {
+	.alloc = mdesc_kmalloc,
+	.free  = mdesc_kfree,
+};
+
+static struct mdesc_handle *mdesc_alloc(unsigned int mdesc_size,
+					struct mdesc_mem_ops *mops)
 {
-	struct mdesc_node *mp;
+	struct mdesc_handle *hp = mops->alloc(mdesc_size);
 
-	mp = from ? from->allnodes_next : allnodes;
-	for (; mp != NULL; mp = mp->allnodes_next) {
-		if (strcmp(mp->name, name) == 0)
-			break;
-	}
-	return mp;
-}
-EXPORT_SYMBOL(md_find_node_by_name);
+	if (hp)
+		hp->mops = mops;
 
-static unsigned int mdesc_early_allocated;
+	return hp;
+}
 
-static void * __init mdesc_early_alloc(unsigned long size)
+static void mdesc_free(struct mdesc_handle *hp)
 {
-	void *ret;
+	hp->mops->free(hp);
+}
 
-	ret = __alloc_bootmem(size, SMP_CACHE_BYTES, 0UL);
-	if (ret == NULL) {
-		prom_printf("MDESC: alloc of %lu bytes failed.\n", size);
-		prom_halt();
-	}
+static struct mdesc_handle *cur_mdesc;
+static LIST_HEAD(mdesc_zombie_list);
+static DEFINE_SPINLOCK(mdesc_lock);
 
-	memset(ret, 0, size);
+struct mdesc_handle *mdesc_grab(void)
+{
+	struct mdesc_handle *hp;
+	unsigned long flags;
 
-	mdesc_early_allocated += size;
+	spin_lock_irqsave(&mdesc_lock, flags);
+	hp = cur_mdesc;
+	if (hp)
+		atomic_inc(&hp->refcnt);
+	spin_unlock_irqrestore(&mdesc_lock, flags);
 
-	return ret;
+	return hp;
 }
+EXPORT_SYMBOL(mdesc_grab);
 
-static unsigned int __init count_arcs(struct mdesc_elem *ep)
+void mdesc_release(struct mdesc_handle *hp)
 {
-	unsigned int ret = 0;
+	unsigned long flags;
 
-	ep++;
-	while (ep->tag != MD_NODE_END) {
-		if (ep->tag == MD_PROP_ARC)
-			ret++;
-		ep++;
+	spin_lock_irqsave(&mdesc_lock, flags);
+	if (atomic_dec_and_test(&hp->refcnt)) {
+		list_del_init(&hp->list);
+		hp->mops->free(hp);
 	}
-	return ret;
+	spin_unlock_irqrestore(&mdesc_lock, flags);
 }
+EXPORT_SYMBOL(mdesc_release);
 
-static void __init mdesc_node_alloc(u64 node, struct mdesc_elem *ep, const char *names)
+static void do_mdesc_update(struct work_struct *work)
 {
-	unsigned int num_arcs = count_arcs(ep);
-	struct mdesc_node *mp;
+	unsigned long len, real_len, status;
+	struct mdesc_handle *hp, *orig_hp;
+	unsigned long flags;
+
+	(void) sun4v_mach_desc(0UL, 0UL, &len);
+
+	hp = mdesc_alloc(len, &kmalloc_mdesc_memops);
+	if (!hp) {
+		printk(KERN_ERR "MD: mdesc alloc fails\n");
+		return;
+	}
+
+	status = sun4v_mach_desc(__pa(&hp->mdesc), len, &real_len);
+	if (status != HV_EOK || real_len > len) {
+		printk(KERN_ERR "MD: mdesc reread fails with %lu\n",
+		       status);
+		atomic_dec(&hp->refcnt);
+		mdesc_free(hp);
+		return;
+	}
 
-	mp = mdesc_early_alloc(sizeof(*mp) +
-			       (num_arcs * sizeof(struct mdesc_arc)));
-	mp->name = names + ep->name_offset;
-	mp->node = node;
-	mp->unique_id = unique_id++;
-	mp->num_arcs = num_arcs;
+	spin_lock_irqsave(&mdesc_lock, flags);
+	orig_hp = cur_mdesc;
+	cur_mdesc = hp;
 
-	hash_node(mp);
+	if (atomic_dec_and_test(&orig_hp->refcnt))
+		mdesc_free(orig_hp);
+	else
+		list_add(&orig_hp->list, &mdesc_zombie_list);
+	spin_unlock_irqrestore(&mdesc_lock, flags);
 }
 
-static inline struct mdesc_elem *node_block(struct mdesc_hdr *mdesc)
+static DECLARE_WORK(mdesc_update_work, do_mdesc_update);
+
+void mdesc_update(void)
+{
+	schedule_work(&mdesc_update_work);
+}
+
+static struct mdesc_elem *node_block(struct mdesc_hdr *mdesc)
 {
 	return (struct mdesc_elem *) (mdesc + 1);
 }
 
-static inline void *name_block(struct mdesc_hdr *mdesc)
+static void *name_block(struct mdesc_hdr *mdesc)
 {
 	return ((void *) node_block(mdesc)) + mdesc->node_sz;
 }
 
-static inline void *data_block(struct mdesc_hdr *mdesc)
+static void *data_block(struct mdesc_hdr *mdesc)
 {
 	return ((void *) name_block(mdesc)) + mdesc->name_sz;
 }
 
-/* In order to avoid recursion (the graph can be very deep) we use a
- * two pass algorithm.  First we allocate all the nodes and hash them.
- * Then we iterate over each node, filling in the arcs and properties.
- */
-static void __init build_all_nodes(struct mdesc_hdr *mdesc)
+u64 mdesc_node_by_name(struct mdesc_handle *hp,
+		       u64 from_node, const char *name)
 {
-	struct mdesc_elem *start, *ep;
-	struct mdesc_node *mp;
-	const char *names;
-	void *data;
-	u64 last_node;
-
-	start = ep = node_block(mdesc);
-	last_node = mdesc->node_sz / 16;
+	struct mdesc_elem *ep = node_block(&hp->mdesc);
+	const char *names = name_block(&hp->mdesc);
+	u64 last_node = hp->mdesc.node_sz / 16;
+	u64 ret;
+
+	if (from_node == MDESC_NODE_NULL)
+		from_node = 0;
+
+	if (from_node >= last_node)
+		return MDESC_NODE_NULL;
+
+	ret = ep[from_node].d.val;
+	while (ret < last_node) {
+		if (ep[ret].tag != MD_NODE)
+			return MDESC_NODE_NULL;
+		if (!strcmp(names + ep[ret].name_offset, name))
+			break;
+		ret = ep[ret].d.val;
+	}
+	if (ret >= last_node)
+		ret = MDESC_NODE_NULL;
+	return ret;
+}
+EXPORT_SYMBOL(mdesc_node_by_name);
 
-	names = name_block(mdesc);
+const void *mdesc_get_property(struct mdesc_handle *hp, u64 node,
+			       const char *name, int *lenp)
+{
+	const char *names = name_block(&hp->mdesc);
+	u64 last_node = hp->mdesc.node_sz / 16;
+	void *data = data_block(&hp->mdesc);
+	struct mdesc_elem *ep;
 
-	while (1) {
-		u64 node = ep - start;
+	if (node == MDESC_NODE_NULL || node >= last_node)
+		return NULL;
 
-		if (ep->tag == MD_LIST_END)
+	ep = node_block(&hp->mdesc) + node;
+	ep++;
+	for (; ep->tag != MD_NODE_END; ep++) {
+		void *val = NULL;
+		int len = 0;
+
+		switch (ep->tag) {
+		case MD_PROP_VAL:
+			val = &ep->d.val;
+			len = 8;
 			break;
 
-		if (ep->tag != MD_NODE) {
-			prom_printf("MDESC: Inconsistent element list.\n");
-			prom_halt();
-		}
-
-		mdesc_node_alloc(node, ep, names);
+		case MD_PROP_STR:
+		case MD_PROP_DATA:
+			val = data + ep->d.data.data_offset;
+			len = ep->d.data.data_len;
+			break;
 
-		if (ep->d.val >= last_node) {
-			printk("MDESC: Warning, early break out of node scan.\n");
-			printk("MDESC: Next node [%lu] last_node [%lu].\n",
-			       node, last_node);
+		default:
 			break;
 		}
+		if (!val)
+			continue;
 
-		ep = start + ep->d.val;
+		if (!strcmp(names + ep->name_offset, name)) {
+			if (lenp)
+				*lenp = len;
+			return val;
+		}
 	}
 
-	data = data_block(mdesc);
-	for (mp = allnodes; mp; mp = mp->allnodes_next) {
-		struct mdesc_elem *ep = start + mp->node;
-		struct property **link = &mp->properties;
-		unsigned int this_arc = 0;
-
-		ep++;
-		while (ep->tag != MD_NODE_END) {
-			switch (ep->tag) {
-			case MD_PROP_ARC: {
-				struct mdesc_node *target;
-
-				if (this_arc >= mp->num_arcs) {
-					prom_printf("MDESC: ARC overrun [%u:%u]\n",
-						    this_arc, mp->num_arcs);
-					prom_halt();
-				}
-				target = find_node(ep->d.val);
-				if (!target) {
-					printk("MDESC: Warning, arc points to "
-					       "missing node, ignoring.\n");
-					break;
-				}
-				mp->arcs[this_arc].name =
-					(names + ep->name_offset);
-				mp->arcs[this_arc].arc = target;
-				this_arc++;
-				break;
-			}
+	return NULL;
+}
+EXPORT_SYMBOL(mdesc_get_property);
 
-			case MD_PROP_VAL:
-			case MD_PROP_STR:
-			case MD_PROP_DATA: {
-				struct property *p = mdesc_early_alloc(sizeof(*p));
-
-				p->unique_id = unique_id++;
-				p->name = (char *) names + ep->name_offset;
-				if (ep->tag == MD_PROP_VAL) {
-					p->value = &ep->d.val;
-					p->length = 8;
-				} else {
-					p->value = data + ep->d.data.data_offset;
-					p->length = ep->d.data.data_len;
-				}
-				*link = p;
-				link = &p->next;
-				break;
-			}
+u64 mdesc_next_arc(struct mdesc_handle *hp, u64 from, const char *arc_type)
+{
+	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
+	const char *names = name_block(&hp->mdesc);
+	u64 last_node = hp->mdesc.node_sz / 16;
 
-			case MD_NOOP:
-				break;
+	if (from == MDESC_NODE_NULL || from >= last_node)
+		return MDESC_NODE_NULL;
 
-			default:
-				printk("MDESC: Warning, ignoring unknown tag type %02x\n",
-				       ep->tag);
-			}
-			ep++;
-		}
+	ep = base + from;
+
+	ep++;
+	for (; ep->tag != MD_NODE_END; ep++) {
+		if (ep->tag != MD_PROP_ARC)
+			continue;
+
+		if (strcmp(names + ep->name_offset, arc_type))
+			continue;
+
+		return ep - base;
 	}
+
+	return MDESC_NODE_NULL;
 }
+EXPORT_SYMBOL(mdesc_next_arc);
 
-static unsigned int __init count_nodes(struct mdesc_hdr *mdesc)
+u64 mdesc_arc_target(struct mdesc_handle *hp, u64 arc)
 {
-	struct mdesc_elem *ep = node_block(mdesc);
-	struct mdesc_elem *end;
-	unsigned int cnt = 0;
-
-	end = ((void *)ep) + mdesc->node_sz;
-	while (ep < end) {
-		if (ep->tag == MD_NODE)
-			cnt++;
-		ep++;
-	}
-	return cnt;
+	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
+
+	ep = base + arc;
+
+	return ep->d.val;
+}
+EXPORT_SYMBOL(mdesc_arc_target);
+
+const char *mdesc_node_name(struct mdesc_handle *hp, u64 node)
+{
+	struct mdesc_elem *ep, *base = node_block(&hp->mdesc);
+	const char *names = name_block(&hp->mdesc);
+	u64 last_node = hp->mdesc.node_sz / 16;
+
+	if (node == MDESC_NODE_NULL || node >= last_node)
+		return NULL;
+
+	ep = base + node;
+	if (ep->tag != MD_NODE)
+		return NULL;
+
+	return names + ep->name_offset;
 }
+EXPORT_SYMBOL(mdesc_node_name);
 
 static void __init report_platform_properties(void)
 {
-	struct mdesc_node *pn = md_find_node_by_name(NULL, "platform");
+	struct mdesc_handle *hp = mdesc_grab();
+	u64 pn = mdesc_node_by_name(hp, MDESC_NODE_NULL, "platform");
 	const char *s;
 	const u64 *v;
 
-	if (!pn) {
+	if (pn == MDESC_NODE_NULL) {
 		prom_printf("No platform node in machine-description.\n");
 		prom_halt();
 	}
 
-	s = md_get_property(pn, "banner-name", NULL);
+	s = mdesc_get_property(hp, pn, "banner-name", NULL);
 	printk("PLATFORM: banner-name [%s]\n", s);
-	s = md_get_property(pn, "name", NULL);
+	s = mdesc_get_property(hp, pn, "name", NULL);
 	printk("PLATFORM: name [%s]\n", s);
 
-	v = md_get_property(pn, "hostid", NULL);
+	v = mdesc_get_property(hp, pn, "hostid", NULL);
 	if (v)
 		printk("PLATFORM: hostid [%08lx]\n", *v);
-	v = md_get_property(pn, "serial#", NULL);
+	v = mdesc_get_property(hp, pn, "serial#", NULL);
 	if (v)
 		printk("PLATFORM: serial# [%08lx]\n", *v);
-	v = md_get_property(pn, "stick-frequency", NULL);
+	v = mdesc_get_property(hp, pn, "stick-frequency", NULL);
 	printk("PLATFORM: stick-frequency [%08lx]\n", *v);
-	v = md_get_property(pn, "mac-address", NULL);
+	v = mdesc_get_property(hp, pn, "mac-address", NULL);
 	if (v)
 		printk("PLATFORM: mac-address [%lx]\n", *v);
-	v = md_get_property(pn, "watchdog-resolution", NULL);
+	v = mdesc_get_property(hp, pn, "watchdog-resolution", NULL);
 	if (v)
 		printk("PLATFORM: watchdog-resolution [%lu ms]\n", *v);
-	v = md_get_property(pn, "watchdog-max-timeout", NULL);
+	v = mdesc_get_property(hp, pn, "watchdog-max-timeout", NULL);
 	if (v)
 		printk("PLATFORM: watchdog-max-timeout [%lu ms]\n", *v);
-	v = md_get_property(pn, "max-cpus", NULL);
+	v = mdesc_get_property(hp, pn, "max-cpus", NULL);
 	if (v)
 		printk("PLATFORM: max-cpus [%lu]\n", *v);
+
+#ifdef CONFIG_SMP
+	{
+		int max_cpu, i;
+
+		if (v) {
+			max_cpu = *v;
+			if (max_cpu > NR_CPUS)
+				max_cpu = NR_CPUS;
+		} else {
+			max_cpu = NR_CPUS;
+		}
+		for (i = 0; i < max_cpu; i++)
+			cpu_set(i, cpu_possible_map);
+	}
+#endif
+
+	mdesc_release(hp);
 }
 
 static int inline find_in_proplist(const char *list, const char *match, int len)
@@ -369,15 +468,17 @@ static int inline find_in_proplist(const char *list, const char *match, int len)
 	return 0;
 }
 
-static void __init fill_in_one_cache(cpuinfo_sparc *c, struct mdesc_node *mp)
+static void __devinit fill_in_one_cache(cpuinfo_sparc *c,
+					struct mdesc_handle *hp,
+					u64 mp)
 {
-	const u64 *level = md_get_property(mp, "level", NULL);
-	const u64 *size = md_get_property(mp, "size", NULL);
-	const u64 *line_size = md_get_property(mp, "line-size", NULL);
+	const u64 *level = mdesc_get_property(hp, mp, "level", NULL);
+	const u64 *size = mdesc_get_property(hp, mp, "size", NULL);
+	const u64 *line_size = mdesc_get_property(hp, mp, "line-size", NULL);
 	const char *type;
 	int type_len;
 
-	type = md_get_property(mp, "type", &type_len);
+	type = mdesc_get_property(hp, mp, "type", &type_len);
 
 	switch (*level) {
 	case 1:
@@ -400,48 +501,45 @@ static void __init fill_in_one_cache(cpuinfo_sparc *c, struct mdesc_node *mp)
 	}
 
 	if (*level == 1) {
-		unsigned int i;
-
-		for (i = 0; i < mp->num_arcs; i++) {
-			struct mdesc_node *t = mp->arcs[i].arc;
+		u64 a;
 
-			if (strcmp(mp->arcs[i].name, "fwd"))
-				continue;
+		mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
+			u64 target = mdesc_arc_target(hp, a);
+			const char *name = mdesc_node_name(hp, target);
 
-			if (!strcmp(t->name, "cache"))
-				fill_in_one_cache(c, t);
+			if (!strcmp(name, "cache"))
+				fill_in_one_cache(c, hp, target);
 		}
 	}
 }
 
-static void __init mark_core_ids(struct mdesc_node *mp, int core_id)
+static void __devinit mark_core_ids(struct mdesc_handle *hp, u64 mp,
+				    int core_id)
 {
-	unsigned int i;
+	u64 a;
 
-	for (i = 0; i < mp->num_arcs; i++) {
-		struct mdesc_node *t = mp->arcs[i].arc;
+	mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_BACK) {
+		u64 t = mdesc_arc_target(hp, a);
+		const char *name;
 		const u64 *id;
 
-		if (strcmp(mp->arcs[i].name, "back"))
-			continue;
-
-		if (!strcmp(t->name, "cpu")) {
-			id = md_get_property(t, "id", NULL);
+		name = mdesc_node_name(hp, t);
+		if (!strcmp(name, "cpu")) {
+			id = mdesc_get_property(hp, t, "id", NULL);
 			if (*id < NR_CPUS)
 				cpu_data(*id).core_id = core_id;
 		} else {
-			unsigned int j;
+			u64 j;
 
-			for (j = 0; j < t->num_arcs; j++) {
-				struct mdesc_node *n = t->arcs[j].arc;
+			mdesc_for_each_arc(j, hp, t, MDESC_ARC_TYPE_BACK) {
+				u64 n = mdesc_arc_target(hp, j);
+				const char *n_name;
 
-				if (strcmp(t->arcs[j].name, "back"))
+				n_name = mdesc_node_name(hp, n);
+				if (strcmp(n_name, "cpu"))
 					continue;
 
-				if (strcmp(n->name, "cpu"))
-					continue;
-
-				id = md_get_property(n, "id", NULL);
+				id = mdesc_get_property(hp, n, "id", NULL);
 				if (*id < NR_CPUS)
 					cpu_data(*id).core_id = core_id;
 			}
@@ -449,78 +547,81 @@ static void __init mark_core_ids(struct mdesc_node *mp, int core_id)
 	}
 }
 
-static void __init set_core_ids(void)
+static void __devinit set_core_ids(struct mdesc_handle *hp)
 {
-	struct mdesc_node *mp;
 	int idx;
+	u64 mp;
 
 	idx = 1;
-	md_for_each_node_by_name(mp, "cache") {
-		const u64 *level = md_get_property(mp, "level", NULL);
+	mdesc_for_each_node_by_name(hp, mp, "cache") {
+		const u64 *level;
 		const char *type;
 		int len;
 
+		level = mdesc_get_property(hp, mp, "level", NULL);
 		if (*level != 1)
 			continue;
 
-		type = md_get_property(mp, "type", &len);
+		type = mdesc_get_property(hp, mp, "type", &len);
 		if (!find_in_proplist(type, "instn", len))
 			continue;
 
-		mark_core_ids(mp, idx);
+		mark_core_ids(hp, mp, idx);
 
 		idx++;
 	}
 }
 
-static void __init mark_proc_ids(struct mdesc_node *mp, int proc_id)
+static void __devinit mark_proc_ids(struct mdesc_handle *hp, u64 mp,
+				    int proc_id)
 {
-	int i;
+	u64 a;
 
-	for (i = 0; i < mp->num_arcs; i++) {
-		struct mdesc_node *t = mp->arcs[i].arc;
+	mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_BACK) {
+		u64 t = mdesc_arc_target(hp, a);
+		const char *name;
 		const u64 *id;
 
-		if (strcmp(mp->arcs[i].name, "back"))
-			continue;
-
-		if (strcmp(t->name, "cpu"))
+		name = mdesc_node_name(hp, t);
+		if (strcmp(name, "cpu"))
 			continue;
 
-		id = md_get_property(t, "id", NULL);
+		id = mdesc_get_property(hp, t, "id", NULL);
 		if (*id < NR_CPUS)
 			cpu_data(*id).proc_id = proc_id;
 	}
 }
 
-static void __init __set_proc_ids(const char *exec_unit_name)
+static void __devinit __set_proc_ids(struct mdesc_handle *hp,
+				     const char *exec_unit_name)
 {
-	struct mdesc_node *mp;
 	int idx;
+	u64 mp;
 
 	idx = 0;
-	md_for_each_node_by_name(mp, exec_unit_name) {
+	mdesc_for_each_node_by_name(hp, mp, exec_unit_name) {
 		const char *type;
 		int len;
 
-		type = md_get_property(mp, "type", &len);
+		type = mdesc_get_property(hp, mp, "type", &len);
 		if (!find_in_proplist(type, "int", len) &&
 		    !find_in_proplist(type, "integer", len))
 			continue;
 
-		mark_proc_ids(mp, idx);
+		mark_proc_ids(hp, mp, idx);
 
 		idx++;
 	}
 }
 
-static void __init set_proc_ids(void)
+static void __devinit set_proc_ids(struct mdesc_handle *hp)
 {
-	__set_proc_ids("exec_unit");
-	__set_proc_ids("exec-unit");
+	__set_proc_ids(hp, "exec_unit");
+	__set_proc_ids(hp, "exec-unit");
 }
 
-static void __init get_one_mondo_bits(const u64 *p, unsigned int *mask, unsigned char def)
+static void __devinit get_one_mondo_bits(const u64 *p, unsigned int *mask,
+					 unsigned char def)
 {
 	u64 val;
 
@@ -538,35 +639,37 @@ use_default:
 	*mask = ((1U << def) * 64U) - 1U;
 }
 
-static void __init get_mondo_data(struct mdesc_node *mp, struct trap_per_cpu *tb)
+static void __devinit get_mondo_data(struct mdesc_handle *hp, u64 mp,
+				     struct trap_per_cpu *tb)
 {
 	const u64 *val;
 
-	val = md_get_property(mp, "q-cpu-mondo-#bits", NULL);
+	val = mdesc_get_property(hp, mp, "q-cpu-mondo-#bits", NULL);
 	get_one_mondo_bits(val, &tb->cpu_mondo_qmask, 7);
 
-	val = md_get_property(mp, "q-dev-mondo-#bits", NULL);
+	val = mdesc_get_property(hp, mp, "q-dev-mondo-#bits", NULL);
 	get_one_mondo_bits(val, &tb->dev_mondo_qmask, 7);
 
-	val = md_get_property(mp, "q-resumable-#bits", NULL);
+	val = mdesc_get_property(hp, mp, "q-resumable-#bits", NULL);
 	get_one_mondo_bits(val, &tb->resum_qmask, 6);
 
-	val = md_get_property(mp, "q-nonresumable-#bits", NULL);
+	val = mdesc_get_property(hp, mp, "q-nonresumable-#bits", NULL);
 	get_one_mondo_bits(val, &tb->nonresum_qmask, 2);
 }
 
-static void __init mdesc_fill_in_cpu_data(void)
+void __devinit mdesc_fill_in_cpu_data(cpumask_t mask)
 {
-	struct mdesc_node *mp;
+	struct mdesc_handle *hp = mdesc_grab();
+	u64 mp;
 
 	ncpus_probed = 0;
-	md_for_each_node_by_name(mp, "cpu") {
-		const u64 *id = md_get_property(mp, "id", NULL);
-		const u64 *cfreq = md_get_property(mp, "clock-frequency", NULL);
+	mdesc_for_each_node_by_name(hp, mp, "cpu") {
+		const u64 *id = mdesc_get_property(hp, mp, "id", NULL);
+		const u64 *cfreq = mdesc_get_property(hp, mp, "clock-frequency", NULL);
 		struct trap_per_cpu *tb;
 		cpuinfo_sparc *c;
-		unsigned int i;
 		int cpuid;
+		u64 a;
 
 		ncpus_probed++;
 
@@ -575,6 +678,8 @@ static void __init mdesc_fill_in_cpu_data(void)
 #ifdef CONFIG_SMP
 		if (cpuid >= NR_CPUS)
 			continue;
+		if (!cpu_isset(cpuid, mask))
+			continue;
 #else
 		/* On uniprocessor we only want the values for the
 		 * real physical cpu the kernel booted onto, however
@@ -589,35 +694,30 @@ static void __init mdesc_fill_in_cpu_data(void)
 		c->clock_tick = *cfreq;
 
 		tb = &trap_block[cpuid];
-		get_mondo_data(mp, tb);
-
-		for (i = 0; i < mp->num_arcs; i++) {
-			struct mdesc_node *t = mp->arcs[i].arc;
-			unsigned int j;
+		get_mondo_data(hp, mp, tb);
 
-			if (strcmp(mp->arcs[i].name, "fwd"))
-				continue;
+		mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
+			u64 j, t = mdesc_arc_target(hp, a);
+			const char *t_name;
 
-			if (!strcmp(t->name, "cache")) {
-				fill_in_one_cache(c, t);
+			t_name = mdesc_node_name(hp, t);
+			if (!strcmp(t_name, "cache")) {
+				fill_in_one_cache(c, hp, t);
 				continue;
 			}
 
-			for (j = 0; j < t->num_arcs; j++) {
-				struct mdesc_node *n;
+			mdesc_for_each_arc(j, hp, t, MDESC_ARC_TYPE_FWD) {
+				u64 n = mdesc_arc_target(hp, j);
+				const char *n_name;
 
-				n = t->arcs[j].arc;
-				if (strcmp(t->arcs[j].name, "fwd"))
-					continue;
-
-				if (!strcmp(n->name, "cache"))
-					fill_in_one_cache(c, n);
+				n_name = mdesc_node_name(hp, n);
+				if (!strcmp(n_name, "cache"))
+					fill_in_one_cache(c, hp, n);
 			}
 		}
 
 #ifdef CONFIG_SMP
 		cpu_set(cpuid, cpu_present_map);
-		cpu_set(cpuid, phys_cpu_present_map);
 #endif
 
 		c->core_id = 0;
@@ -628,45 +728,43 @@ static void __init mdesc_fill_in_cpu_data(void)
 	sparc64_multi_core = 1;
 #endif
 
-	set_core_ids();
-	set_proc_ids();
+	set_core_ids(hp);
+	set_proc_ids(hp);
 
 	smp_fill_in_sib_core_maps();
+
+	mdesc_release(hp);
 }
 
 void __init sun4v_mdesc_init(void)
 {
+	struct mdesc_handle *hp;
 	unsigned long len, real_len, status;
+	cpumask_t mask;
 
 	(void) sun4v_mach_desc(0UL, 0UL, &len);
 
 	printk("MDESC: Size is %lu bytes.\n", len);
 
-	main_mdesc = mdesc_early_alloc(len);
+	hp = mdesc_alloc(len, &bootmem_mdesc_memops);
+	if (hp == NULL) {
+		prom_printf("MDESC: alloc of %lu bytes failed.\n", len);
+		prom_halt();
+	}
 
-	status = sun4v_mach_desc(__pa(main_mdesc), len, &real_len);
+	status = sun4v_mach_desc(__pa(&hp->mdesc), len, &real_len);
 	if (status != HV_EOK || real_len > len) {
 		prom_printf("sun4v_mach_desc fails, err(%lu), "
 			    "len(%lu), real_len(%lu)\n",
 			    status, len, real_len);
+		mdesc_free(hp);
 		prom_halt();
 	}
 
-	len = count_nodes(main_mdesc);
-	printk("MDESC: %lu nodes.\n", len);
-
-	len = roundup_pow_of_two(len);
-
-	mdesc_hash = mdesc_early_alloc(len * sizeof(struct mdesc_node *));
-	mdesc_hash_size = len;
-
-	printk("MDESC: Hash size %lu entries.\n", len);
-
-	build_all_nodes(main_mdesc);
-
-	printk("MDESC: Built graph with %u bytes of memory.\n",
-	       mdesc_early_allocated);
+	cur_mdesc = hp;
 
 	report_platform_properties();
-	mdesc_fill_in_cpu_data();
+
+	cpus_setall(mask);
+	mdesc_fill_in_cpu_data(mask);
 }

arch/sparc64/kernel/power.c

@@ -1,7 +1,6 @@
-/* $Id: power.c,v 1.10 2001/12/11 01:57:16 davem Exp $
- * power.c: Power management driver.
+/* power.c: Power management driver.
  *
- * Copyright (C) 1999 David S. Miller (davem@redhat.com)
+ * Copyright (C) 1999, 2007 David S. Miller (davem@davemloft.net)
  */
 
 #include <linux/kernel.h>
@@ -19,6 +18,7 @@
 #include <asm/prom.h>
 #include <asm/of_device.h>
 #include <asm/io.h>
+#include <asm/power.h>
 #include <asm/sstate.h>
 
 #include <linux/unistd.h>
@@ -29,24 +29,26 @@
  */
 int scons_pwroff = 1; 
 
-#ifdef CONFIG_PCI
-#include <linux/pci.h>
 static void __iomem *power_reg;
 
 static DECLARE_WAIT_QUEUE_HEAD(powerd_wait);
 static int button_pressed;
 
-static irqreturn_t power_handler(int irq, void *dev_id)
+void wake_up_powerd(void)
 {
 	if (button_pressed == 0) {
 		button_pressed = 1;
 		wake_up(&powerd_wait);
 	}
+}
+
+static irqreturn_t power_handler(int irq, void *dev_id)
+{
+	wake_up_powerd();
 
 	/* FIXME: Check registers for status... */
 	return IRQ_HANDLED;
 }
-#endif /* CONFIG_PCI */
 
 extern void machine_halt(void);
 extern void machine_alt_power_off(void);
@@ -56,19 +58,18 @@ void machine_power_off(void)
 {
 	sstate_poweroff();
 	if (!serial_console || scons_pwroff) {
-#ifdef CONFIG_PCI
 		if (power_reg) {
 			/* Both register bits seem to have the
 			 * same effect, so until I figure out
 			 * what the difference is...
 			 */
 			writel(AUXIO_PCIO_CPWR_OFF | AUXIO_PCIO_SPWR_OFF, power_reg);
-		} else
-#endif /* CONFIG_PCI */
+		} else {
 			if (poweroff_method != NULL) {
 				poweroff_method();
 				/* not reached */
 			}
+		}
 	}
 	machine_halt();
 }
@@ -76,7 +77,6 @@ void machine_power_off(void)
 void (*pm_power_off)(void) = machine_power_off;
 EXPORT_SYMBOL(pm_power_off);
 
-#ifdef CONFIG_PCI
 static int powerd(void *__unused)
 {
 	static char *envp[] = { "HOME=/", "TERM=linux", "PATH=/sbin:/usr/sbin:/bin:/usr/bin", NULL };
@@ -86,7 +86,7 @@ static int powerd(void *__unused)
 	daemonize("powerd");
 
 	add_wait_queue(&powerd_wait, &wait);
-again:
+
 	for (;;) {
 		set_task_state(current, TASK_INTERRUPTIBLE);
 		if (button_pressed)
@@ -100,16 +100,28 @@ again:
 	/* Ok, down we go... */
 	button_pressed = 0;
 	if (kernel_execve("/sbin/shutdown", argv, envp) < 0) {
-		printk("powerd: shutdown execution failed\n");
-		add_wait_queue(&powerd_wait, &wait);
-		goto again;
+		printk(KERN_ERR "powerd: shutdown execution failed\n");
+		machine_power_off();
 	}
 	return 0;
 }
 
+int start_powerd(void)
+{
+	int err;
+
+	err = kernel_thread(powerd, NULL, CLONE_FS);
+	if (err < 0)
+		printk(KERN_ERR "power: Failed to start power daemon.\n");
+	else
+		printk(KERN_INFO "power: powerd running.\n");
+
+	return err;
+}
+
 static int __init has_button_interrupt(unsigned int irq, struct device_node *dp)
 {
-	if (irq == PCI_IRQ_NONE)
+	if (irq == 0xffffffff)
 		return 0;
 	if (!of_find_property(dp, "button", NULL))
 		return 0;
@@ -130,17 +142,14 @@ static int __devinit power_probe(struct of_device *op, const struct of_device_id
 	poweroff_method = machine_halt;  /* able to use the standard halt */
 
 	if (has_button_interrupt(irq, op->node)) {
-		if (kernel_thread(powerd, NULL, CLONE_FS) < 0) {
-			printk("Failed to start power daemon.\n");
+		if (start_powerd() < 0)
 			return 0;
-		}
-		printk("powerd running.\n");
 
 		if (request_irq(irq,
 				power_handler, 0, "power", NULL) < 0)
-			printk("power: Error, cannot register IRQ handler.\n");
+			printk(KERN_ERR "power: Cannot setup IRQ handler.\n");
 	} else {
-		printk("not using powerd.\n");
+		printk(KERN_INFO "power: Not using powerd.\n");
 	}
 
 	return 0;
@@ -164,4 +173,3 @@ void __init power_init(void)
 	of_register_driver(&power_driver, &of_bus_type);
 	return;
 }
-#endif /* CONFIG_PCI */

arch/sparc64/kernel/process.c

@@ -29,6 +29,7 @@
 #include <linux/compat.h>
 #include <linux/tick.h>
 #include <linux/init.h>
+#include <linux/cpu.h>
 
 #include <asm/oplib.h>
 #include <asm/uaccess.h>
@@ -49,7 +50,7 @@
 
 /* #define VERBOSE_SHOWREGS */
 
-static void sparc64_yield(void)
+static void sparc64_yield(int cpu)
 {
 	if (tlb_type != hypervisor)
 		return;
@@ -57,7 +58,7 @@ static void sparc64_yield(void)
 	clear_thread_flag(TIF_POLLING_NRFLAG);
 	smp_mb__after_clear_bit();
 
-	while (!need_resched()) {
+	while (!need_resched() && !cpu_is_offline(cpu)) {
 		unsigned long pstate;
 
 		/* Disable interrupts. */
@@ -68,7 +69,7 @@ static void sparc64_yield(void)
 			: "=&r" (pstate)
 			: "i" (PSTATE_IE));
 
-		if (!need_resched())
+		if (!need_resched() && !cpu_is_offline(cpu))
 			sun4v_cpu_yield();
 
 		/* Re-enable interrupts. */
@@ -86,15 +87,25 @@ static void sparc64_yield(void)
 /* The idle loop on sparc64. */
 void cpu_idle(void)
 {
+	int cpu = smp_processor_id();
+
 	set_thread_flag(TIF_POLLING_NRFLAG);
 
 	while(1) {
 		tick_nohz_stop_sched_tick();
-		while (!need_resched())
-			sparc64_yield();
+
+		while (!need_resched() && !cpu_is_offline(cpu))
+			sparc64_yield(cpu);
+
 		tick_nohz_restart_sched_tick();
 
 		preempt_enable_no_resched();
+
+#ifdef CONFIG_HOTPLUG_CPU
+		if (cpu_is_offline(cpu))
+			cpu_play_dead();
+#endif
+
 		schedule();
 		preempt_disable();
 	}

arch/sparc64/kernel/prom.c

@@ -1808,7 +1808,7 @@ static void __init of_fill_in_cpu_data(void)
 
 #ifdef CONFIG_SMP
 		cpu_set(cpuid, cpu_present_map);
-		cpu_set(cpuid, phys_cpu_present_map);
+		cpu_set(cpuid, cpu_possible_map);
 #endif
 	}
 

arch/sparc64/kernel/setup.c

@@ -442,7 +442,6 @@ static int show_cpuinfo(struct seq_file *m, void *__unused)
 		   "D$ parity tl1\t: %u\n"
 		   "I$ parity tl1\t: %u\n"
 #ifndef CONFIG_SMP
-		   "Cpu0Bogo\t: %lu.%02lu\n"
 		   "Cpu0ClkTck\t: %016lx\n"
 #endif
 		   ,
@@ -455,10 +454,8 @@ static int show_cpuinfo(struct seq_file *m, void *__unused)
 		   ncpus_probed,
 		   num_online_cpus(),
 		   dcache_parity_tl1_occurred,
-		   icache_parity_tl1_occurred
+		   icache_parity_tl1_occurred,
 #ifndef CONFIG_SMP
-		   , cpu_data(0).udelay_val/(500000/HZ),
-		   (cpu_data(0).udelay_val/(5000/HZ)) % 100,
 		   cpu_data(0).clock_tick
 #endif
 		);

arch/sparc64/kernel/smp.c

@@ -1,6 +1,6 @@
 /* smp.c: Sparc64 SMP support.
  *
- * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ * Copyright (C) 1997, 2007 David S. Miller (davem@davemloft.net)
  */
 
 #include <linux/module.h>
@@ -28,6 +28,8 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 #include <asm/cpudata.h>
+#include <asm/hvtramp.h>
+#include <asm/io.h>
 
 #include <asm/irq.h>
 #include <asm/irq_regs.h>
@@ -41,22 +43,26 @@
 #include <asm/sections.h>
 #include <asm/prom.h>
 #include <asm/mdesc.h>
+#include <asm/ldc.h>
+#include <asm/hypervisor.h>
 
 extern void calibrate_delay(void);
 
 int sparc64_multi_core __read_mostly;
 
-/* Please don't make this stuff initdata!!!  --DaveM */
-unsigned char boot_cpu_id;
-
+cpumask_t cpu_possible_map __read_mostly = CPU_MASK_NONE;
 cpumask_t cpu_online_map __read_mostly = CPU_MASK_NONE;
-cpumask_t phys_cpu_present_map __read_mostly = CPU_MASK_NONE;
 cpumask_t cpu_sibling_map[NR_CPUS] __read_mostly =
 	{ [0 ... NR_CPUS-1] = CPU_MASK_NONE };
 cpumask_t cpu_core_map[NR_CPUS] __read_mostly =
 	{ [0 ... NR_CPUS-1] = CPU_MASK_NONE };
+
+EXPORT_SYMBOL(cpu_possible_map);
+EXPORT_SYMBOL(cpu_online_map);
+EXPORT_SYMBOL(cpu_sibling_map);
+EXPORT_SYMBOL(cpu_core_map);
+
 static cpumask_t smp_commenced_mask;
-static cpumask_t cpu_callout_map;
 
 void smp_info(struct seq_file *m)
 {
@@ -73,18 +79,17 @@ void smp_bogo(struct seq_file *m)
 	
 	for_each_online_cpu(i)
 		seq_printf(m,
-			   "Cpu%dBogo\t: %lu.%02lu\n"
 			   "Cpu%dClkTck\t: %016lx\n",
-			   i, cpu_data(i).udelay_val / (500000/HZ),
-			   (cpu_data(i).udelay_val / (5000/HZ)) % 100,
 			   i, cpu_data(i).clock_tick);
 }
 
+static __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
+
 extern void setup_sparc64_timer(void);
 
 static volatile unsigned long callin_flag = 0;
 
-void __init smp_callin(void)
+void __devinit smp_callin(void)
 {
 	int cpuid = hard_smp_processor_id();
 
@@ -102,8 +107,6 @@ void __init smp_callin(void)
 
 	local_irq_enable();
 
-	calibrate_delay();
-	cpu_data(cpuid).udelay_val = loops_per_jiffy;
 	callin_flag = 1;
 	__asm__ __volatile__("membar #Sync\n\t"
 			     "flush  %%g6" : : : "memory");
@@ -120,7 +123,9 @@ void __init smp_callin(void)
 	while (!cpu_isset(cpuid, smp_commenced_mask))
 		rmb();
 
+	spin_lock(&call_lock);
 	cpu_set(cpuid, cpu_online_map);
+	spin_unlock(&call_lock);
 
 	/* idle thread is expected to have preempt disabled */
 	preempt_disable();
@@ -268,6 +273,67 @@ static void smp_synchronize_one_tick(int cpu)
 	spin_unlock_irqrestore(&itc_sync_lock, flags);
 }
 
+#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
+/* XXX Put this in some common place. XXX */
+static unsigned long kimage_addr_to_ra(void *p)
+{
+	unsigned long val = (unsigned long) p;
+
+	return kern_base + (val - KERNBASE);
+}
+
+static void ldom_startcpu_cpuid(unsigned int cpu, unsigned long thread_reg)
+{
+	extern unsigned long sparc64_ttable_tl0;
+	extern unsigned long kern_locked_tte_data;
+	extern int bigkernel;
+	struct hvtramp_descr *hdesc;
+	unsigned long trampoline_ra;
+	struct trap_per_cpu *tb;
+	u64 tte_vaddr, tte_data;
+	unsigned long hv_err;
+
+	hdesc = kzalloc(sizeof(*hdesc), GFP_KERNEL);
+	if (!hdesc) {
+		printk(KERN_ERR "ldom_startcpu_cpuid: Cannot allocate "
+		       "hvtramp_descr.\n");
+		return;
+	}
+
+	hdesc->cpu = cpu;
+	hdesc->num_mappings = (bigkernel ? 2 : 1);
+
+	tb = &trap_block[cpu];
+	tb->hdesc = hdesc;
+
+	hdesc->fault_info_va = (unsigned long) &tb->fault_info;
+	hdesc->fault_info_pa = kimage_addr_to_ra(&tb->fault_info);
+
+	hdesc->thread_reg = thread_reg;
+
+	tte_vaddr = (unsigned long) KERNBASE;
+	tte_data = kern_locked_tte_data;
+
+	hdesc->maps[0].vaddr = tte_vaddr;
+	hdesc->maps[0].tte   = tte_data;
+	if (bigkernel) {
+		tte_vaddr += 0x400000;
+		tte_data  += 0x400000;
+		hdesc->maps[1].vaddr = tte_vaddr;
+		hdesc->maps[1].tte   = tte_data;
+	}
+
+	trampoline_ra = kimage_addr_to_ra(hv_cpu_startup);
+
+	hv_err = sun4v_cpu_start(cpu, trampoline_ra,
+				 kimage_addr_to_ra(&sparc64_ttable_tl0),
+				 __pa(hdesc));
+	if (hv_err)
+		printk(KERN_ERR "ldom_startcpu_cpuid: sun4v_cpu_start() "
+		       "gives error %lu\n", hv_err);
+}
+#endif
+
 extern void sun4v_init_mondo_queues(int use_bootmem, int cpu, int alloc, int load);
 
 extern unsigned long sparc64_cpu_startup;
@@ -280,6 +346,7 @@ static struct thread_info *cpu_new_thread = NULL;
 
 static int __devinit smp_boot_one_cpu(unsigned int cpu)
 {
+	struct trap_per_cpu *tb = &trap_block[cpu];
 	unsigned long entry =
 		(unsigned long)(&sparc64_cpu_startup);
 	unsigned long cookie =
@@ -290,20 +357,25 @@ static int __devinit smp_boot_one_cpu(unsigned int cpu)
 	p = fork_idle(cpu);
 	callin_flag = 0;
 	cpu_new_thread = task_thread_info(p);
-	cpu_set(cpu, cpu_callout_map);
 
 	if (tlb_type == hypervisor) {
 		/* Alloc the mondo queues, cpu will load them.  */
 		sun4v_init_mondo_queues(0, cpu, 1, 0);
 
-		prom_startcpu_cpuid(cpu, entry, cookie);
+#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
+		if (ldom_domaining_enabled)
+			ldom_startcpu_cpuid(cpu,
+					    (unsigned long) cpu_new_thread);
+		else
+#endif
+			prom_startcpu_cpuid(cpu, entry, cookie);
 	} else {
 		struct device_node *dp = of_find_node_by_cpuid(cpu);
 
 		prom_startcpu(dp->node, entry, cookie);
 	}
 
-	for (timeout = 0; timeout < 5000000; timeout++) {
+	for (timeout = 0; timeout < 50000; timeout++) {
 		if (callin_flag)
 			break;
 		udelay(100);
@@ -313,11 +385,15 @@ static int __devinit smp_boot_one_cpu(unsigned int cpu)
 		ret = 0;
 	} else {
 		printk("Processor %d is stuck.\n", cpu);
-		cpu_clear(cpu, cpu_callout_map);
 		ret = -ENODEV;
 	}
 	cpu_new_thread = NULL;
 
+	if (tb->hdesc) {
+		kfree(tb->hdesc);
+		tb->hdesc = NULL;
+	}
+
 	return ret;
 }
 
@@ -720,7 +796,6 @@ struct call_data_struct {
 	int wait;
 };
 
-static __cacheline_aligned_in_smp DEFINE_SPINLOCK(call_lock);
 static struct call_data_struct *call_data;
 
 extern unsigned long xcall_call_function;
@@ -1152,34 +1227,14 @@ void smp_penguin_jailcell(int irq, struct pt_regs *regs)
 	preempt_enable();
 }
 
-void __init smp_tick_init(void)
-{
-	boot_cpu_id = hard_smp_processor_id();
-}
-
 /* /proc/profile writes can call this, don't __init it please. */
 int setup_profiling_timer(unsigned int multiplier)
 {
 	return -EINVAL;
 }
 
-/* Constrain the number of cpus to max_cpus.  */
 void __init smp_prepare_cpus(unsigned int max_cpus)
 {
-	int i;
-
-	if (num_possible_cpus() > max_cpus) {
-		for_each_possible_cpu(i) {
-			if (i != boot_cpu_id) {
-				cpu_clear(i, phys_cpu_present_map);
-				cpu_clear(i, cpu_present_map);
-				if (num_possible_cpus() <= max_cpus)
-					break;
-			}
-		}
-	}
-
-	cpu_data(boot_cpu_id).udelay_val = loops_per_jiffy;
 }
 
 void __devinit smp_prepare_boot_cpu(void)
@@ -1190,30 +1245,32 @@ void __devinit smp_fill_in_sib_core_maps(void)
 {
 	unsigned int i;
 
-	for_each_possible_cpu(i) {
+	for_each_present_cpu(i) {
 		unsigned int j;
 
+		cpus_clear(cpu_core_map[i]);
 		if (cpu_data(i).core_id == 0) {
 			cpu_set(i, cpu_core_map[i]);
 			continue;
 		}
 
-		for_each_possible_cpu(j) {
+		for_each_present_cpu(j) {
 			if (cpu_data(i).core_id ==
 			    cpu_data(j).core_id)
 				cpu_set(j, cpu_core_map[i]);
 		}
 	}
 
-	for_each_possible_cpu(i) {
+	for_each_present_cpu(i) {
 		unsigned int j;
 
+		cpus_clear(cpu_sibling_map[i]);
 		if (cpu_data(i).proc_id == -1) {
 			cpu_set(i, cpu_sibling_map[i]);
 			continue;
 		}
 
-		for_each_possible_cpu(j) {
+		for_each_present_cpu(j) {
 			if (cpu_data(i).proc_id ==
 			    cpu_data(j).proc_id)
 				cpu_set(j, cpu_sibling_map[i]);
@@ -1242,18 +1299,112 @@ int __cpuinit __cpu_up(unsigned int cpu)
 	return ret;
 }
 
-void __init smp_cpus_done(unsigned int max_cpus)
+#ifdef CONFIG_HOTPLUG_CPU
+void cpu_play_dead(void)
+{
+	int cpu = smp_processor_id();
+	unsigned long pstate;
+
+	idle_task_exit();
+
+	if (tlb_type == hypervisor) {
+		struct trap_per_cpu *tb = &trap_block[cpu];
+
+		sun4v_cpu_qconf(HV_CPU_QUEUE_CPU_MONDO,
+				tb->cpu_mondo_pa, 0);
+		sun4v_cpu_qconf(HV_CPU_QUEUE_DEVICE_MONDO,
+				tb->dev_mondo_pa, 0);
+		sun4v_cpu_qconf(HV_CPU_QUEUE_RES_ERROR,
+				tb->resum_mondo_pa, 0);
+		sun4v_cpu_qconf(HV_CPU_QUEUE_NONRES_ERROR,
+				tb->nonresum_mondo_pa, 0);
+	}
+
+	cpu_clear(cpu, smp_commenced_mask);
+	membar_safe("#Sync");
+
+	local_irq_disable();
+
+	__asm__ __volatile__(
+		"rdpr	%%pstate, %0\n\t"
+		"wrpr	%0, %1, %%pstate"
+		: "=r" (pstate)
+		: "i" (PSTATE_IE));
+
+	while (1)
+		barrier();
+}
+
+int __cpu_disable(void)
 {
-	unsigned long bogosum = 0;
+	int cpu = smp_processor_id();
+	cpuinfo_sparc *c;
 	int i;
 
-	for_each_online_cpu(i)
-		bogosum += cpu_data(i).udelay_val;
-	printk("Total of %ld processors activated "
-	       "(%lu.%02lu BogoMIPS).\n",
-	       (long) num_online_cpus(),
-	       bogosum/(500000/HZ),
-	       (bogosum/(5000/HZ))%100);
+	for_each_cpu_mask(i, cpu_core_map[cpu])
+		cpu_clear(cpu, cpu_core_map[i]);
+	cpus_clear(cpu_core_map[cpu]);
+
+	for_each_cpu_mask(i, cpu_sibling_map[cpu])
+		cpu_clear(cpu, cpu_sibling_map[i]);
+	cpus_clear(cpu_sibling_map[cpu]);
+
+	c = &cpu_data(cpu);
+
+	c->core_id = 0;
+	c->proc_id = -1;
+
+	spin_lock(&call_lock);
+	cpu_clear(cpu, cpu_online_map);
+	spin_unlock(&call_lock);
+
+	smp_wmb();
+
+	/* Make sure no interrupts point to this cpu.  */
+	fixup_irqs();
+
+	local_irq_enable();
+	mdelay(1);
+	local_irq_disable();
+
+	return 0;
+}
+
+void __cpu_die(unsigned int cpu)
+{
+	int i;
+
+	for (i = 0; i < 100; i++) {
+		smp_rmb();
+		if (!cpu_isset(cpu, smp_commenced_mask))
+			break;
+		msleep(100);
+	}
+	if (cpu_isset(cpu, smp_commenced_mask)) {
+		printk(KERN_ERR "CPU %u didn't die...\n", cpu);
+	} else {
+#if defined(CONFIG_SUN_LDOMS)
+		unsigned long hv_err;
+		int limit = 100;
+
+		do {
+			hv_err = sun4v_cpu_stop(cpu);
+			if (hv_err == HV_EOK) {
+				cpu_clear(cpu, cpu_present_map);
+				break;
+			}
+		} while (--limit > 0);
+		if (limit <= 0) {
+			printk(KERN_ERR "sun4v_cpu_stop() fails err=%lu\n",
+			       hv_err);
+		}
+#endif
+	}
+}
+#endif
+
+void __init smp_cpus_done(unsigned int max_cpus)
+{
 }
 
 void smp_send_reschedule(int cpu)

arch/sparc64/kernel/sparc64_ksyms.c

@@ -1,7 +1,6 @@
-/* $Id: sparc64_ksyms.c,v 1.121 2002/02/09 19:49:31 davem Exp $
- * arch/sparc64/kernel/sparc64_ksyms.c: Sparc64 specific ksyms support.
+/* arch/sparc64/kernel/sparc64_ksyms.c: Sparc64 specific ksyms support.
  *
- * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
+ * Copyright (C) 1996, 2007 David S. Miller (davem@davemloft.net)
  * Copyright (C) 1996 Eddie C. Dost (ecd@skynet.be)
  * Copyright (C) 1999 Jakub Jelinek (jj@ultra.linux.cz)
  */
@@ -28,7 +27,6 @@
 #include <net/compat.h>
 
 #include <asm/oplib.h>
-#include <asm/delay.h>
 #include <asm/system.h>
 #include <asm/auxio.h>
 #include <asm/pgtable.h>
@@ -124,10 +122,6 @@ EXPORT_SYMBOL(__write_lock);
 EXPORT_SYMBOL(__write_unlock);
 EXPORT_SYMBOL(__write_trylock);
 
-/* CPU online map and active count.  */
-EXPORT_SYMBOL(cpu_online_map);
-EXPORT_SYMBOL(phys_cpu_present_map);
-
 EXPORT_SYMBOL(smp_call_function);
 #endif /* CONFIG_SMP */
 
@@ -330,12 +324,6 @@ EXPORT_SYMBOL(memset);
 EXPORT_SYMBOL(memmove);
 EXPORT_SYMBOL(strncmp);
 
-/* Delay routines. */
-EXPORT_SYMBOL(__udelay);
-EXPORT_SYMBOL(__ndelay);
-EXPORT_SYMBOL(__const_udelay);
-EXPORT_SYMBOL(__delay);
-
 void VISenter(void);
 /* RAID code needs this */
 EXPORT_SYMBOL(VISenter);

arch/sparc64/kernel/sysfs.c

@@ -193,7 +193,6 @@ static ssize_t show_##NAME(struct sys_device *dev, char *buf) \
 }
 
 SHOW_CPUDATA_ULONG_NAME(clock_tick, clock_tick);
-SHOW_CPUDATA_ULONG_NAME(udelay_val, udelay_val);
 SHOW_CPUDATA_UINT_NAME(l1_dcache_size, dcache_size);
 SHOW_CPUDATA_UINT_NAME(l1_dcache_line_size, dcache_line_size);
 SHOW_CPUDATA_UINT_NAME(l1_icache_size, icache_size);
@@ -203,7 +202,6 @@ SHOW_CPUDATA_UINT_NAME(l2_cache_line_size, ecache_line_size);
 
 static struct sysdev_attribute cpu_core_attrs[] = {
 	_SYSDEV_ATTR(clock_tick,          0444, show_clock_tick, NULL),
-	_SYSDEV_ATTR(udelay_val,          0444, show_udelay_val, NULL),
 	_SYSDEV_ATTR(l1_dcache_size,      0444, show_l1_dcache_size, NULL),
 	_SYSDEV_ATTR(l1_dcache_line_size, 0444, show_l1_dcache_line_size, NULL),
 	_SYSDEV_ATTR(l1_icache_size,      0444, show_l1_icache_size, NULL),

arch/sparc64/kernel/time.c

@@ -849,9 +849,6 @@ static unsigned long sparc64_init_timers(void)
 {
 	struct device_node *dp;
 	unsigned long clock;
-#ifdef CONFIG_SMP
-	extern void smp_tick_init(void);
-#endif
 
 	dp = of_find_node_by_path("/");
 	if (tlb_type == spitfire) {
@@ -874,10 +871,6 @@ static unsigned long sparc64_init_timers(void)
 		clock = of_getintprop_default(dp, "stick-frequency", 0);
 	}
 
-#ifdef CONFIG_SMP
-	smp_tick_init();
-#endif
-
 	return clock;
 }
 
@@ -1038,10 +1031,31 @@ static void __init setup_clockevent_multiplier(unsigned long hz)
 	sparc64_clockevent.mult = mult;
 }
 
+static unsigned long tb_ticks_per_usec __read_mostly;
+
+void __delay(unsigned long loops)
+{
+	unsigned long bclock, now;
+
+	bclock = tick_ops->get_tick();
+	do {
+		now = tick_ops->get_tick();
+	} while ((now-bclock) < loops);
+}
+EXPORT_SYMBOL(__delay);
+
+void udelay(unsigned long usecs)
+{
+	__delay(tb_ticks_per_usec * usecs);
+}
+EXPORT_SYMBOL(udelay);
+
 void __init time_init(void)
 {
 	unsigned long clock = sparc64_init_timers();
 
+	tb_ticks_per_usec = clock / USEC_PER_SEC;
+
 	timer_ticks_per_nsec_quotient =
 		clocksource_hz2mult(clock, SPARC64_NSEC_PER_CYC_SHIFT);
 

arch/sparc64/kernel/vio.c

@@ -0,0 +1,395 @@
+/* vio.c: Virtual I/O channel devices probing infrastructure.
+ *
+ *    Copyright (c) 2003-2005 IBM Corp.
+ *     Dave Engebretsen engebret@us.ibm.com
+ *     Santiago Leon santil@us.ibm.com
+ *     Hollis Blanchard <hollisb@us.ibm.com>
+ *     Stephen Rothwell
+ *
+ * Adapted to sparc64 by David S. Miller davem@davemloft.net
+ */
+
+#include <linux/kernel.h>
+#include <linux/irq.h>
+#include <linux/init.h>
+
+#include <asm/mdesc.h>
+#include <asm/vio.h>
+
+static inline int find_in_proplist(const char *list, const char *match,
+				   int len)
+{
+	while (len > 0) {
+		int l;
+
+		if (!strcmp(list, match))
+			return 1;
+		l = strlen(list) + 1;
+		list += l;
+		len -= l;
+	}
+	return 0;
+}
+
+static const struct vio_device_id *vio_match_device(
+	const struct vio_device_id *matches,
+	const struct vio_dev *dev)
+{
+	const char *type, *compat;
+	int len;
+
+	type = dev->type;
+	compat = dev->compat;
+	len = dev->compat_len;
+
+	while (matches->type[0] || matches->compat[0]) {
+		int match = 1;
+		if (matches->type[0])
+			match &= !strcmp(matches->type, type);
+
+		if (matches->compat[0]) {
+			match &= len &&
+				find_in_proplist(compat, matches->compat, len);
+		}
+		if (match)
+			return matches;
+		matches++;
+	}
+	return NULL;
+}
+
+static int vio_bus_match(struct device *dev, struct device_driver *drv)
+{
+	struct vio_dev *vio_dev = to_vio_dev(dev);
+	struct vio_driver *vio_drv = to_vio_driver(drv);
+	const struct vio_device_id *matches = vio_drv->id_table;
+
+	if (!matches)
+		return 0;
+
+	return vio_match_device(matches, vio_dev) != NULL;
+}
+
+static int vio_device_probe(struct device *dev)
+{
+	struct vio_dev *vdev = to_vio_dev(dev);
+	struct vio_driver *drv = to_vio_driver(dev->driver);
+	const struct vio_device_id *id;
+	int error = -ENODEV;
+
+	if (drv->probe) {
+		id = vio_match_device(drv->id_table, vdev);
+		if (id)
+			error = drv->probe(vdev, id);
+	}
+
+	return error;
+}
+
+static int vio_device_remove(struct device *dev)
+{
+	struct vio_dev *vdev = to_vio_dev(dev);
+	struct vio_driver *drv = to_vio_driver(dev->driver);
+
+	if (drv->remove)
+		return drv->remove(vdev);
+
+	return 1;
+}
+
+static ssize_t devspec_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct vio_dev *vdev = to_vio_dev(dev);
+	const char *str = "none";
+
+	if (!strcmp(vdev->type, "network"))
+		str = "vnet";
+	else if (!strcmp(vdev->type, "block"))
+		str = "vdisk";
+
+	return sprintf(buf, "%s\n", str);
+}
+
+static ssize_t type_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct vio_dev *vdev = to_vio_dev(dev);
+	return sprintf(buf, "%s\n", vdev->type);
+}
+
+static struct device_attribute vio_dev_attrs[] = {
+	__ATTR_RO(devspec),
+	__ATTR_RO(type),
+	__ATTR_NULL
+};
+
+static struct bus_type vio_bus_type = {
+	.name		= "vio",
+	.dev_attrs	= vio_dev_attrs,
+	.match		= vio_bus_match,
+	.probe		= vio_device_probe,
+	.remove		= vio_device_remove,
+};
+
+int vio_register_driver(struct vio_driver *viodrv)
+{
+	viodrv->driver.bus = &vio_bus_type;
+
+	return driver_register(&viodrv->driver);
+}
+EXPORT_SYMBOL(vio_register_driver);
+
+void vio_unregister_driver(struct vio_driver *viodrv)
+{
+	driver_unregister(&viodrv->driver);
+}
+EXPORT_SYMBOL(vio_unregister_driver);
+
+static void __devinit vio_dev_release(struct device *dev)
+{
+	kfree(to_vio_dev(dev));
+}
+
+static ssize_t
+show_pciobppath_attr(struct device *dev, struct device_attribute *attr,
+		     char *buf)
+{
+	struct vio_dev *vdev;
+	struct device_node *dp;
+
+	vdev = to_vio_dev(dev);
+	dp = vdev->dp;
+
+	return snprintf (buf, PAGE_SIZE, "%s\n", dp->full_name);
+}
+
+static DEVICE_ATTR(obppath, S_IRUSR | S_IRGRP | S_IROTH,
+		   show_pciobppath_attr, NULL);
+
+struct device_node *cdev_node;
+
+static struct vio_dev *root_vdev;
+static u64 cdev_cfg_handle;
+
+static void vio_fill_channel_info(struct mdesc_handle *hp, u64 mp,
+				  struct vio_dev *vdev)
+{
+	u64 a;
+
+	mdesc_for_each_arc(a, hp, mp, MDESC_ARC_TYPE_FWD) {
+		const u64 *chan_id;
+		const u64 *irq;
+		u64 target;
+
+		target = mdesc_arc_target(hp, a);
+
+		irq = mdesc_get_property(hp, target, "tx-ino", NULL);
+		if (irq)
+			vdev->tx_irq = sun4v_build_virq(cdev_cfg_handle, *irq);
+
+		irq = mdesc_get_property(hp, target, "rx-ino", NULL);
+		if (irq)
+			vdev->rx_irq = sun4v_build_virq(cdev_cfg_handle, *irq);
+
+		chan_id = mdesc_get_property(hp, target, "id", NULL);
+		if (chan_id)
+			vdev->channel_id = *chan_id;
+	}
+}
+
+static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
+				      struct device *parent)
+{
+	const char *type, *compat;
+	struct device_node *dp;
+	struct vio_dev *vdev;
+	int err, tlen, clen;
+
+	type = mdesc_get_property(hp, mp, "device-type", &tlen);
+	if (!type) {
+		type = mdesc_get_property(hp, mp, "name", &tlen);
+		if (!type) {
+			type = mdesc_node_name(hp, mp);
+			tlen = strlen(type) + 1;
+		}
+	}
+	if (tlen > VIO_MAX_TYPE_LEN) {
+		printk(KERN_ERR "VIO: Type string [%s] is too long.\n",
+		       type);
+		return NULL;
+	}
+
+	compat = mdesc_get_property(hp, mp, "device-type", &clen);
+	if (!compat) {
+		clen = 0;
+	} else if (clen > VIO_MAX_COMPAT_LEN) {
+		printk(KERN_ERR "VIO: Compat len %d for [%s] is too long.\n",
+		       clen, type);
+		return NULL;
+	}
+
+	vdev = kzalloc(sizeof(*vdev), GFP_KERNEL);
+	if (!vdev) {
+		printk(KERN_ERR "VIO: Could not allocate vio_dev\n");
+		return NULL;
+	}
+
+	vdev->mp = mp;
+	memcpy(vdev->type, type, tlen);
+	if (compat)
+		memcpy(vdev->compat, compat, clen);
+	else
+		memset(vdev->compat, 0, sizeof(vdev->compat));
+	vdev->compat_len = clen;
+
+	vdev->channel_id = ~0UL;
+	vdev->tx_irq = ~0;
+	vdev->rx_irq = ~0;
+
+	vio_fill_channel_info(hp, mp, vdev);
+
+	snprintf(vdev->dev.bus_id, BUS_ID_SIZE, "%lx", mp);
+	vdev->dev.parent = parent;
+	vdev->dev.bus = &vio_bus_type;
+	vdev->dev.release = vio_dev_release;
+
+	if (parent == NULL) {
+		dp = cdev_node;
+	} else if (to_vio_dev(parent) == root_vdev) {
+		dp = of_get_next_child(cdev_node, NULL);
+		while (dp) {
+			if (!strcmp(dp->type, type))
+				break;
+
+			dp = of_get_next_child(cdev_node, dp);
+		}
+	} else {
+		dp = to_vio_dev(parent)->dp;
+	}
+	vdev->dp = dp;
+
+	err = device_register(&vdev->dev);
+	if (err) {
+		printk(KERN_ERR "VIO: Could not register device %s, err=%d\n",
+		       vdev->dev.bus_id, err);
+		kfree(vdev);
+		return NULL;
+	}
+	if (vdev->dp)
+		err = sysfs_create_file(&vdev->dev.kobj,
+					&dev_attr_obppath.attr);
+
+	return vdev;
+}
+
+static void walk_tree(struct mdesc_handle *hp, u64 n, struct vio_dev *parent)
+{
+	u64 a;
+
+	mdesc_for_each_arc(a, hp, n, MDESC_ARC_TYPE_FWD) {
+		struct vio_dev *vdev;
+		u64 target;
+
+		target = mdesc_arc_target(hp, a);
+		vdev = vio_create_one(hp, target, &parent->dev);
+		if (vdev)
+			walk_tree(hp, target, vdev);
+	}
+}
+
+static void create_devices(struct mdesc_handle *hp, u64 root)
+{
+	u64 mp;
+
+	root_vdev = vio_create_one(hp, root, NULL);
+	if (!root_vdev) {
+		printk(KERN_ERR "VIO: Coult not create root device.\n");
+		return;
+	}
+
+	walk_tree(hp, root, root_vdev);
+
+	/* Domain services is odd as it doesn't sit underneath the
+	 * channel-devices node, so we plug it in manually.
+	 */
+	mp = mdesc_node_by_name(hp, MDESC_NODE_NULL, "domain-services");
+	if (mp != MDESC_NODE_NULL) {
+		struct vio_dev *parent = vio_create_one(hp, mp,
+							&root_vdev->dev);
+
+		if (parent)
+			walk_tree(hp, mp, parent);
+	}
+}
+
+const char *channel_devices_node = "channel-devices";
+const char *channel_devices_compat = "SUNW,sun4v-channel-devices";
+const char *cfg_handle_prop = "cfg-handle";
+
+static int __init vio_init(void)
+{
+	struct mdesc_handle *hp;
+	const char *compat;
+	const u64 *cfg_handle;
+	int err, len;
+	u64 root;
+
+	err = bus_register(&vio_bus_type);
+	if (err) {
+		printk(KERN_ERR "VIO: Could not register bus type err=%d\n",
+		       err);
+		return err;
+	}
+
+	hp = mdesc_grab();
+	if (!hp)
+		return 0;
+
+	root = mdesc_node_by_name(hp, MDESC_NODE_NULL, channel_devices_node);
+	if (root == MDESC_NODE_NULL) {
+		printk(KERN_INFO "VIO: No channel-devices MDESC node.\n");
+		mdesc_release(hp);
+		return 0;
+	}
+
+	cdev_node = of_find_node_by_name(NULL, "channel-devices");
+	err = -ENODEV;
+	if (!cdev_node) {
+		printk(KERN_INFO "VIO: No channel-devices OBP node.\n");
+		goto out_release;
+	}
+
+	compat = mdesc_get_property(hp, root, "compatible", &len);
+	if (!compat) {
+		printk(KERN_ERR "VIO: Channel devices lacks compatible "
+		       "property\n");
+		goto out_release;
+	}
+	if (!find_in_proplist(compat, channel_devices_compat, len)) {
+		printk(KERN_ERR "VIO: Channel devices node lacks (%s) "
+		       "compat entry.\n", channel_devices_compat);
+		goto out_release;
+	}
+
+	cfg_handle = mdesc_get_property(hp, root, cfg_handle_prop, NULL);
+	if (!cfg_handle) {
+		printk(KERN_ERR "VIO: Channel devices lacks %s property\n",
+		       cfg_handle_prop);
+		goto out_release;
+	}
+
+	cdev_cfg_handle = *cfg_handle;
+
+	create_devices(hp, root);
+
+	mdesc_release(hp);
+
+	return 0;
+
+out_release:
+	mdesc_release(hp);
+	return err;
+}
+
+postcore_initcall(vio_init);

arch/sparc64/kernel/viohs.c

@@ -0,0 +1,792 @@
+/* viohs.c: LDOM Virtual I/O handshake helper layer.
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+
+#include <asm/ldc.h>
+#include <asm/vio.h>
+
+int vio_ldc_send(struct vio_driver_state *vio, void *data, int len)
+{
+	int err, limit = 1000;
+
+	err = -EINVAL;
+	while (limit-- > 0) {
+		err = ldc_write(vio->lp, data, len);
+		if (!err || (err != -EAGAIN))
+			break;
+		udelay(1);
+	}
+
+	return err;
+}
+EXPORT_SYMBOL(vio_ldc_send);
+
+static int send_ctrl(struct vio_driver_state *vio,
+		     struct vio_msg_tag *tag, int len)
+{
+	tag->sid = vio_send_sid(vio);
+	return vio_ldc_send(vio, tag, len);
+}
+
+static void init_tag(struct vio_msg_tag *tag, u8 type, u8 stype, u16 stype_env)
+{
+	tag->type = type;
+	tag->stype = stype;
+	tag->stype_env = stype_env;
+}
+
+static int send_version(struct vio_driver_state *vio, u16 major, u16 minor)
+{
+	struct vio_ver_info pkt;
+
+	vio->_local_sid = (u32) sched_clock();
+
+	memset(&pkt, 0, sizeof(pkt));
+	init_tag(&pkt.tag, VIO_TYPE_CTRL, VIO_SUBTYPE_INFO, VIO_VER_INFO);
+	pkt.major = major;
+	pkt.minor = minor;
+	pkt.dev_class = vio->dev_class;
+
+	viodbg(HS, "SEND VERSION INFO maj[%u] min[%u] devclass[%u]\n",
+	       major, minor, vio->dev_class);
+
+	return send_ctrl(vio, &pkt.tag, sizeof(pkt));
+}
+
+static int start_handshake(struct vio_driver_state *vio)
+{
+	int err;
+
+	viodbg(HS, "START HANDSHAKE\n");
+
+	vio->hs_state = VIO_HS_INVALID;
+
+	err = send_version(vio,
+			   vio->ver_table[0].major,
+			   vio->ver_table[0].minor);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+void vio_link_state_change(struct vio_driver_state *vio, int event)
+{
+	if (event == LDC_EVENT_UP) {
+		vio->hs_state = VIO_HS_INVALID;
+
+		switch (vio->dev_class) {
+		case VDEV_NETWORK:
+		case VDEV_NETWORK_SWITCH:
+			vio->dr_state = (VIO_DR_STATE_TXREQ |
+					 VIO_DR_STATE_RXREQ);
+			break;
+
+		case VDEV_DISK:
+			vio->dr_state = VIO_DR_STATE_TXREQ;
+			break;
+		case VDEV_DISK_SERVER:
+			vio->dr_state = VIO_DR_STATE_RXREQ;
+			break;
+		}
+		start_handshake(vio);
+	}
+}
+EXPORT_SYMBOL(vio_link_state_change);
+
+static int handshake_failure(struct vio_driver_state *vio)
+{
+	struct vio_dring_state *dr;
+
+	/* XXX Put policy here...  Perhaps start a timer to fire
+	 * XXX in 100 ms, which will bring the link up and retry
+	 * XXX the handshake.
+	 */
+
+	viodbg(HS, "HANDSHAKE FAILURE\n");
+
+	vio->dr_state &= ~(VIO_DR_STATE_TXREG |
+			   VIO_DR_STATE_RXREG);
+
+	dr = &vio->drings[VIO_DRIVER_RX_RING];
+	memset(dr, 0, sizeof(*dr));
+
+	kfree(vio->desc_buf);
+	vio->desc_buf = NULL;
+	vio->desc_buf_len = 0;
+
+	vio->hs_state = VIO_HS_INVALID;
+
+	return -ECONNRESET;
+}
+
+static int process_unknown(struct vio_driver_state *vio, void *arg)
+{
+	struct vio_msg_tag *pkt = arg;
+
+	viodbg(HS, "UNKNOWN CONTROL [%02x:%02x:%04x:%08x]\n",
+	       pkt->type, pkt->stype, pkt->stype_env, pkt->sid);
+
+	printk(KERN_ERR "vio: ID[%lu] Resetting connection.\n",
+	       vio->vdev->channel_id);
+
+	ldc_disconnect(vio->lp);
+
+	return -ECONNRESET;
+}
+
+static int send_dreg(struct vio_driver_state *vio)
+{
+	struct vio_dring_state *dr = &vio->drings[VIO_DRIVER_TX_RING];
+	union {
+		struct vio_dring_register pkt;
+		char all[sizeof(struct vio_dring_register) +
+			 (sizeof(struct ldc_trans_cookie) *
+			  dr->ncookies)];
+	} u;
+	int i;
+
+	memset(&u, 0, sizeof(u));
+	init_tag(&u.pkt.tag, VIO_TYPE_CTRL, VIO_SUBTYPE_INFO, VIO_DRING_REG);
+	u.pkt.dring_ident = 0;
+	u.pkt.num_descr = dr->num_entries;
+	u.pkt.descr_size = dr->entry_size;
+	u.pkt.options = VIO_TX_DRING;
+	u.pkt.num_cookies = dr->ncookies;
+
+	viodbg(HS, "SEND DRING_REG INFO ndesc[%u] dsz[%u] opt[0x%x] "
+	       "ncookies[%u]\n",
+	       u.pkt.num_descr, u.pkt.descr_size, u.pkt.options,
+	       u.pkt.num_cookies);
+
+	for (i = 0; i < dr->ncookies; i++) {
+		u.pkt.cookies[i] = dr->cookies[i];
+
+		viodbg(HS, "DRING COOKIE(%d) [%016llx:%016llx]\n",
+		       i,
+		       (unsigned long long) u.pkt.cookies[i].cookie_addr,
+		       (unsigned long long) u.pkt.cookies[i].cookie_size);
+	}
+
+	return send_ctrl(vio, &u.pkt.tag, sizeof(u));
+}
+
+static int send_rdx(struct vio_driver_state *vio)
+{
+	struct vio_rdx pkt;
+
+	memset(&pkt, 0, sizeof(pkt));
+
+	init_tag(&pkt.tag, VIO_TYPE_CTRL, VIO_SUBTYPE_INFO, VIO_RDX);
+
+	viodbg(HS, "SEND RDX INFO\n");
+
+	return send_ctrl(vio, &pkt.tag, sizeof(pkt));
+}
+
+static int send_attr(struct vio_driver_state *vio)
+{
+	return vio->ops->send_attr(vio);
+}
+
+static struct vio_version *find_by_major(struct vio_driver_state *vio,
+					 u16 major)
+{
+	struct vio_version *ret = NULL;
+	int i;
+
+	for (i = 0; i < vio->ver_table_entries; i++) {
+		struct vio_version *v = &vio->ver_table[i];
+		if (v->major <= major) {
+			ret = v;
+			break;
+		}
+	}
+	return ret;
+}
+
+static int process_ver_info(struct vio_driver_state *vio,
+			    struct vio_ver_info *pkt)
+{
+	struct vio_version *vap;
+	int err;
+
+	viodbg(HS, "GOT VERSION INFO maj[%u] min[%u] devclass[%u]\n",
+	       pkt->major, pkt->minor, pkt->dev_class);
+
+	if (vio->hs_state != VIO_HS_INVALID) {
+		/* XXX Perhaps invoke start_handshake? XXX */
+		memset(&vio->ver, 0, sizeof(vio->ver));
+		vio->hs_state = VIO_HS_INVALID;
+	}
+
+	vap = find_by_major(vio, pkt->major);
+
+	vio->_peer_sid = pkt->tag.sid;
+
+	if (!vap) {
+		pkt->tag.stype = VIO_SUBTYPE_NACK;
+		pkt->major = 0;
+		pkt->minor = 0;
+		viodbg(HS, "SEND VERSION NACK maj[0] min[0]\n");
+		err = send_ctrl(vio, &pkt->tag, sizeof(*pkt));
+	} else if (vap->major != pkt->major) {
+		pkt->tag.stype = VIO_SUBTYPE_NACK;
+		pkt->major = vap->major;
+		pkt->minor = vap->minor;
+		viodbg(HS, "SEND VERSION NACK maj[%u] min[%u]\n",
+		       pkt->major, pkt->minor);
+		err = send_ctrl(vio, &pkt->tag, sizeof(*pkt));
+	} else {
+		struct vio_version ver = {
+			.major = pkt->major,
+			.minor = pkt->minor,
+		};
+		if (ver.minor > vap->minor)
+			ver.minor = vap->minor;
+		pkt->minor = ver.minor;
+		pkt->tag.stype = VIO_SUBTYPE_ACK;
+		viodbg(HS, "SEND VERSION ACK maj[%u] min[%u]\n",
+		       pkt->major, pkt->minor);
+		err = send_ctrl(vio, &pkt->tag, sizeof(*pkt));
+		if (err > 0) {
+			vio->ver = ver;
+			vio->hs_state = VIO_HS_GOTVERS;
+		}
+	}
+	if (err < 0)
+		return handshake_failure(vio);
+
+	return 0;
+}
+
+static int process_ver_ack(struct vio_driver_state *vio,
+			   struct vio_ver_info *pkt)
+{
+	viodbg(HS, "GOT VERSION ACK maj[%u] min[%u] devclass[%u]\n",
+	       pkt->major, pkt->minor, pkt->dev_class);
+
+	if (vio->hs_state & VIO_HS_GOTVERS) {
+		if (vio->ver.major != pkt->major ||
+		    vio->ver.minor != pkt->minor) {
+			pkt->tag.stype = VIO_SUBTYPE_NACK;
+			(void) send_ctrl(vio, &pkt->tag, sizeof(*pkt));
+			return handshake_failure(vio);
+		}
+	} else {
+		vio->ver.major = pkt->major;
+		vio->ver.minor = pkt->minor;
+		vio->hs_state = VIO_HS_GOTVERS;
+	}
+
+	switch (vio->dev_class) {
+	case VDEV_NETWORK:
+	case VDEV_DISK:
+		if (send_attr(vio) < 0)
+			return handshake_failure(vio);
+		break;
+
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int process_ver_nack(struct vio_driver_state *vio,
+			    struct vio_ver_info *pkt)
+{
+	struct vio_version *nver;
+
+	viodbg(HS, "GOT VERSION NACK maj[%u] min[%u] devclass[%u]\n",
+	       pkt->major, pkt->minor, pkt->dev_class);
+
+	if ((pkt->major == 0 && pkt->minor == 0) ||
+	    !(nver = find_by_major(vio, pkt->major)))
+		return handshake_failure(vio);
+
+	if (send_version(vio, nver->major, nver->minor) < 0)
+		return handshake_failure(vio);
+
+	return 0;
+}
+
+static int process_ver(struct vio_driver_state *vio, struct vio_ver_info *pkt)
+{
+	switch (pkt->tag.stype) {
+	case VIO_SUBTYPE_INFO:
+		return process_ver_info(vio, pkt);
+
+	case VIO_SUBTYPE_ACK:
+		return process_ver_ack(vio, pkt);
+
+	case VIO_SUBTYPE_NACK:
+		return process_ver_nack(vio, pkt);
+
+	default:
+		return handshake_failure(vio);
+	};
+}
+
+static int process_attr(struct vio_driver_state *vio, void *pkt)
+{
+	int err;
+
+	if (!(vio->hs_state & VIO_HS_GOTVERS))
+		return handshake_failure(vio);
+
+	err = vio->ops->handle_attr(vio, pkt);
+	if (err < 0) {
+		return handshake_failure(vio);
+	} else {
+		vio->hs_state |= VIO_HS_GOT_ATTR;
+
+		if ((vio->dr_state & VIO_DR_STATE_TXREQ) &&
+		    !(vio->hs_state & VIO_HS_SENT_DREG)) {
+			if (send_dreg(vio) < 0)
+				return handshake_failure(vio);
+
+			vio->hs_state |= VIO_HS_SENT_DREG;
+		}
+	}
+	return 0;
+}
+
+static int all_drings_registered(struct vio_driver_state *vio)
+{
+	int need_rx, need_tx;
+
+	need_rx = (vio->dr_state & VIO_DR_STATE_RXREQ);
+	need_tx = (vio->dr_state & VIO_DR_STATE_TXREQ);
+
+	if (need_rx &&
+	    !(vio->dr_state & VIO_DR_STATE_RXREG))
+		return 0;
+
+	if (need_tx &&
+	    !(vio->dr_state & VIO_DR_STATE_TXREG))
+		return 0;
+
+	return 1;
+}
+
+static int process_dreg_info(struct vio_driver_state *vio,
+			     struct vio_dring_register *pkt)
+{
+	struct vio_dring_state *dr;
+	int i, len;
+
+	viodbg(HS, "GOT DRING_REG INFO ident[%llx] "
+	       "ndesc[%u] dsz[%u] opt[0x%x] ncookies[%u]\n",
+	       (unsigned long long) pkt->dring_ident,
+	       pkt->num_descr, pkt->descr_size, pkt->options,
+	       pkt->num_cookies);
+
+	if (!(vio->dr_state & VIO_DR_STATE_RXREQ))
+		goto send_nack;
+
+	if (vio->dr_state & VIO_DR_STATE_RXREG)
+		goto send_nack;
+
+	vio->desc_buf = kzalloc(pkt->descr_size, GFP_ATOMIC);
+	if (!vio->desc_buf)
+		goto send_nack;
+
+	vio->desc_buf_len = pkt->descr_size;
+
+	dr = &vio->drings[VIO_DRIVER_RX_RING];
+
+	dr->num_entries = pkt->num_descr;
+	dr->entry_size = pkt->descr_size;
+	dr->ncookies = pkt->num_cookies;
+	for (i = 0; i < dr->ncookies; i++) {
+		dr->cookies[i] = pkt->cookies[i];
+
+		viodbg(HS, "DRING COOKIE(%d) [%016llx:%016llx]\n",
+		       i,
+		       (unsigned long long)
+		       pkt->cookies[i].cookie_addr,
+		       (unsigned long long)
+		       pkt->cookies[i].cookie_size);
+	}
+
+	pkt->tag.stype = VIO_SUBTYPE_ACK;
+	pkt->dring_ident = ++dr->ident;
+
+	viodbg(HS, "SEND DRING_REG ACK ident[%llx]\n",
+	       (unsigned long long) pkt->dring_ident);
+
+	len = (sizeof(*pkt) +
+	       (dr->ncookies * sizeof(struct ldc_trans_cookie)));
+	if (send_ctrl(vio, &pkt->tag, len) < 0)
+		goto send_nack;
+
+	vio->dr_state |= VIO_DR_STATE_RXREG;
+
+	return 0;
+
+send_nack:
+	pkt->tag.stype = VIO_SUBTYPE_NACK;
+	viodbg(HS, "SEND DRING_REG NACK\n");
+	(void) send_ctrl(vio, &pkt->tag, sizeof(*pkt));
+
+	return handshake_failure(vio);
+}
+
+static int process_dreg_ack(struct vio_driver_state *vio,
+			    struct vio_dring_register *pkt)
+{
+	struct vio_dring_state *dr;
+
+	viodbg(HS, "GOT DRING_REG ACK ident[%llx] "
+	       "ndesc[%u] dsz[%u] opt[0x%x] ncookies[%u]\n",
+	       (unsigned long long) pkt->dring_ident,
+	       pkt->num_descr, pkt->descr_size, pkt->options,
+	       pkt->num_cookies);
+
+	dr = &vio->drings[VIO_DRIVER_TX_RING];
+
+	if (!(vio->dr_state & VIO_DR_STATE_TXREQ))
+		return handshake_failure(vio);
+
+	dr->ident = pkt->dring_ident;
+	vio->dr_state |= VIO_DR_STATE_TXREG;
+
+	if (all_drings_registered(vio)) {
+		if (send_rdx(vio) < 0)
+			return handshake_failure(vio);
+		vio->hs_state = VIO_HS_SENT_RDX;
+	}
+	return 0;
+}
+
+static int process_dreg_nack(struct vio_driver_state *vio,
+			     struct vio_dring_register *pkt)
+{
+	viodbg(HS, "GOT DRING_REG NACK ident[%llx] "
+	       "ndesc[%u] dsz[%u] opt[0x%x] ncookies[%u]\n",
+	       (unsigned long long) pkt->dring_ident,
+	       pkt->num_descr, pkt->descr_size, pkt->options,
+	       pkt->num_cookies);
+
+	return handshake_failure(vio);
+}
+
+static int process_dreg(struct vio_driver_state *vio,
+			struct vio_dring_register *pkt)
+{
+	if (!(vio->hs_state & VIO_HS_GOTVERS))
+		return handshake_failure(vio);
+
+	switch (pkt->tag.stype) {
+	case VIO_SUBTYPE_INFO:
+		return process_dreg_info(vio, pkt);
+
+	case VIO_SUBTYPE_ACK:
+		return process_dreg_ack(vio, pkt);
+
+	case VIO_SUBTYPE_NACK:
+		return process_dreg_nack(vio, pkt);
+
+	default:
+		return handshake_failure(vio);
+	}
+}
+
+static int process_dunreg(struct vio_driver_state *vio,
+			  struct vio_dring_unregister *pkt)
+{
+	struct vio_dring_state *dr = &vio->drings[VIO_DRIVER_RX_RING];
+
+	viodbg(HS, "GOT DRING_UNREG\n");
+
+	if (pkt->dring_ident != dr->ident)
+		return 0;
+
+	vio->dr_state &= ~VIO_DR_STATE_RXREG;
+
+	memset(dr, 0, sizeof(*dr));
+
+	kfree(vio->desc_buf);
+	vio->desc_buf = NULL;
+	vio->desc_buf_len = 0;
+
+	return 0;
+}
+
+static int process_rdx_info(struct vio_driver_state *vio, struct vio_rdx *pkt)
+{
+	viodbg(HS, "GOT RDX INFO\n");
+
+	pkt->tag.stype = VIO_SUBTYPE_ACK;
+	viodbg(HS, "SEND RDX ACK\n");
+	if (send_ctrl(vio, &pkt->tag, sizeof(*pkt)) < 0)
+		return handshake_failure(vio);
+
+	vio->hs_state |= VIO_HS_SENT_RDX_ACK;
+	return 0;
+}
+
+static int process_rdx_ack(struct vio_driver_state *vio, struct vio_rdx *pkt)
+{
+	viodbg(HS, "GOT RDX ACK\n");
+
+	if (!(vio->hs_state & VIO_HS_SENT_RDX))
+		return handshake_failure(vio);
+
+	vio->hs_state |= VIO_HS_GOT_RDX_ACK;
+	return 0;
+}
+
+static int process_rdx_nack(struct vio_driver_state *vio, struct vio_rdx *pkt)
+{
+	viodbg(HS, "GOT RDX NACK\n");
+
+	return handshake_failure(vio);
+}
+
+static int process_rdx(struct vio_driver_state *vio, struct vio_rdx *pkt)
+{
+	if (!all_drings_registered(vio))
+		handshake_failure(vio);
+
+	switch (pkt->tag.stype) {
+	case VIO_SUBTYPE_INFO:
+		return process_rdx_info(vio, pkt);
+
+	case VIO_SUBTYPE_ACK:
+		return process_rdx_ack(vio, pkt);
+
+	case VIO_SUBTYPE_NACK:
+		return process_rdx_nack(vio, pkt);
+
+	default:
+		return handshake_failure(vio);
+	}
+}
+
+int vio_control_pkt_engine(struct vio_driver_state *vio, void *pkt)
+{
+	struct vio_msg_tag *tag = pkt;
+	u8 prev_state = vio->hs_state;
+	int err;
+
+	switch (tag->stype_env) {
+	case VIO_VER_INFO:
+		err = process_ver(vio, pkt);
+		break;
+
+	case VIO_ATTR_INFO:
+		err = process_attr(vio, pkt);
+		break;
+
+	case VIO_DRING_REG:
+		err = process_dreg(vio, pkt);
+		break;
+
+	case VIO_DRING_UNREG:
+		err = process_dunreg(vio, pkt);
+		break;
+
+	case VIO_RDX:
+		err = process_rdx(vio, pkt);
+		break;
+
+	default:
+		err = process_unknown(vio, pkt);
+		break;
+	}
+	if (!err &&
+	    vio->hs_state != prev_state &&
+	    (vio->hs_state & VIO_HS_COMPLETE))
+		vio->ops->handshake_complete(vio);
+
+	return err;
+}
+EXPORT_SYMBOL(vio_control_pkt_engine);
+
+void vio_conn_reset(struct vio_driver_state *vio)
+{
+}
+EXPORT_SYMBOL(vio_conn_reset);
+
+/* The issue is that the Solaris virtual disk server just mirrors the
+ * SID values it gets from the client peer.  So we work around that
+ * here in vio_{validate,send}_sid() so that the drivers don't need
+ * to be aware of this crap.
+ */
+int vio_validate_sid(struct vio_driver_state *vio, struct vio_msg_tag *tp)
+{
+	u32 sid;
+
+	/* Always let VERSION+INFO packets through unchecked, they
+	 * define the new SID.
+	 */
+	if (tp->type == VIO_TYPE_CTRL &&
+	    tp->stype == VIO_SUBTYPE_INFO &&
+	    tp->stype_env == VIO_VER_INFO)
+		return 0;
+
+	/* Ok, now figure out which SID to use.  */
+	switch (vio->dev_class) {
+	case VDEV_NETWORK:
+	case VDEV_NETWORK_SWITCH:
+	case VDEV_DISK_SERVER:
+	default:
+		sid = vio->_peer_sid;
+		break;
+
+	case VDEV_DISK:
+		sid = vio->_local_sid;
+		break;
+	}
+
+	if (sid == tp->sid)
+		return 0;
+	viodbg(DATA, "BAD SID tag->sid[%08x] peer_sid[%08x] local_sid[%08x]\n",
+	       tp->sid, vio->_peer_sid, vio->_local_sid);
+	return -EINVAL;
+}
+EXPORT_SYMBOL(vio_validate_sid);
+
+u32 vio_send_sid(struct vio_driver_state *vio)
+{
+	switch (vio->dev_class) {
+	case VDEV_NETWORK:
+	case VDEV_NETWORK_SWITCH:
+	case VDEV_DISK:
+	default:
+		return vio->_local_sid;
+
+	case VDEV_DISK_SERVER:
+		return vio->_peer_sid;
+	}
+}
+EXPORT_SYMBOL(vio_send_sid);
+
+extern int vio_ldc_alloc(struct vio_driver_state *vio,
+			 struct ldc_channel_config *base_cfg,
+			 void *event_arg)
+{
+	struct ldc_channel_config cfg = *base_cfg;
+	struct ldc_channel *lp;
+
+	cfg.tx_irq = vio->vdev->tx_irq;
+	cfg.rx_irq = vio->vdev->rx_irq;
+
+	lp = ldc_alloc(vio->vdev->channel_id, &cfg, event_arg);
+	if (IS_ERR(lp))
+		return PTR_ERR(lp);
+
+	vio->lp = lp;
+
+	return 0;
+}
+EXPORT_SYMBOL(vio_ldc_alloc);
+
+void vio_ldc_free(struct vio_driver_state *vio)
+{
+	ldc_free(vio->lp);
+	vio->lp = NULL;
+
+	kfree(vio->desc_buf);
+	vio->desc_buf = NULL;
+	vio->desc_buf_len = 0;
+}
+EXPORT_SYMBOL(vio_ldc_free);
+
+void vio_port_up(struct vio_driver_state *vio)
+{
+	unsigned long flags;
+	int err, state;
+
+	spin_lock_irqsave(&vio->lock, flags);
+
+	state = ldc_state(vio->lp);
+
+	err = 0;
+	if (state == LDC_STATE_INIT) {
+		err = ldc_bind(vio->lp, vio->name);
+		if (err)
+			printk(KERN_WARNING "%s: Port %lu bind failed, "
+			       "err=%d\n",
+			       vio->name, vio->vdev->channel_id, err);
+	}
+
+	if (!err) {
+		err = ldc_connect(vio->lp);
+		if (err)
+			printk(KERN_WARNING "%s: Port %lu connect failed, "
+			       "err=%d\n",
+			       vio->name, vio->vdev->channel_id, err);
+	}
+	if (err) {
+		unsigned long expires = jiffies + HZ;
+
+		expires = round_jiffies(expires);
+		mod_timer(&vio->timer, expires);
+	}
+
+	spin_unlock_irqrestore(&vio->lock, flags);
+}
+EXPORT_SYMBOL(vio_port_up);
+
+static void vio_port_timer(unsigned long _arg)
+{
+	struct vio_driver_state *vio = (struct vio_driver_state *) _arg;
+
+	vio_port_up(vio);
+}
+
+int vio_driver_init(struct vio_driver_state *vio, struct vio_dev *vdev,
+		    u8 dev_class, struct vio_version *ver_table,
+		    int ver_table_size, struct vio_driver_ops *ops,
+		    char *name)
+{
+	switch (dev_class) {
+	case VDEV_NETWORK:
+	case VDEV_NETWORK_SWITCH:
+	case VDEV_DISK:
+	case VDEV_DISK_SERVER:
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if (!ops->send_attr ||
+	    !ops->handle_attr ||
+	    !ops->handshake_complete)
+		return -EINVAL;
+
+	if (!ver_table || ver_table_size < 0)
+		return -EINVAL;
+
+	if (!name)
+		return -EINVAL;
+
+	spin_lock_init(&vio->lock);
+
+	vio->name = name;
+
+	vio->dev_class = dev_class;
+	vio->vdev = vdev;
+
+	vio->ver_table = ver_table;
+	vio->ver_table_entries = ver_table_size;
+
+	vio->ops = ops;
+
+	setup_timer(&vio->timer, vio_port_timer, (unsigned long) vio);
+
+	return 0;
+}
+EXPORT_SYMBOL(vio_driver_init);

arch/sparc64/lib/Makefile

@@ -14,6 +14,6 @@ lib-y := PeeCeeI.o copy_page.o clear_page.o strlen.o strncmp.o \
 	 NGmemcpy.o NGcopy_from_user.o NGcopy_to_user.o NGpatch.o \
 	 NGpage.o NGbzero.o \
 	 copy_in_user.o user_fixup.o memmove.o \
-	 mcount.o ipcsum.o rwsem.o xor.o delay.o
+	 mcount.o ipcsum.o rwsem.o xor.o
 
 obj-y += iomap.o

arch/sparc64/lib/delay.c

@@ -1,46 +0,0 @@
-/* delay.c: Delay loops for sparc64
- *
- * Copyright (C) 2004, 2006 David S. Miller <davem@davemloft.net>
- *
- * Based heavily upon x86 variant which is:
- *	Copyright (C) 1993 Linus Torvalds
- *	Copyright (C) 1997 Martin Mares <mj@atrey.karlin.mff.cuni.cz>
- */
-
-#include <linux/delay.h>
-#include <asm/timer.h>
-
-void __delay(unsigned long loops)
-{
-	unsigned long bclock, now;
-	
-	bclock = tick_ops->get_tick();
-	do {
-		now = tick_ops->get_tick();
-	} while ((now-bclock) < loops);
-}
-
-/* We used to multiply by HZ after shifting down by 32 bits
- * but that runs into problems for higher values of HZ and
- * slow cpus.
- */
-void __const_udelay(unsigned long n)
-{
-	n *= 4;
-
-	n *= (cpu_data(raw_smp_processor_id()).udelay_val * (HZ/4));
-	n >>= 32;
-
-	__delay(n + 1);
-}
-
-void __udelay(unsigned long n)
-{
-	__const_udelay(n * 0x10c7UL);
-}
-
-
-void __ndelay(unsigned long n)
-{
-	__const_udelay(n * 0x5UL);
-}

arch/sparc64/prom/misc.c

@@ -14,6 +14,7 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 #include <asm/system.h>
+#include <asm/ldc.h>
 
 int prom_service_exists(const char *service_name)
 {
@@ -37,6 +38,10 @@ void prom_sun4v_guest_soft_state(void)
 /* Reset and reboot the machine with the command 'bcommand'. */
 void prom_reboot(const char *bcommand)
 {
+#ifdef CONFIG_SUN_LDOMS
+	if (ldom_domaining_enabled)
+		ldom_reboot(bcommand);
+#endif
 	p1275_cmd("boot", P1275_ARG(0, P1275_ARG_IN_STRING) |
 		  P1275_INOUT(1, 0), bcommand);
 }
@@ -91,6 +96,10 @@ void prom_cmdline(void)
  */
 void prom_halt(void)
 {
+#ifdef CONFIG_SUN_LDOMS
+	if (ldom_domaining_enabled)
+		ldom_power_off();
+#endif
 again:
 	p1275_cmd("exit", P1275_INOUT(0, 0));
 	goto again; /* PROM is out to get me -DaveM */
@@ -98,6 +107,10 @@ again:
 
 void prom_halt_power_off(void)
 {
+#ifdef CONFIG_SUN_LDOMS
+	if (ldom_domaining_enabled)
+		ldom_power_off();
+#endif
 	p1275_cmd("SUNW,power-off", P1275_INOUT(0, 0));
 
 	/* if nothing else helps, we just halt */

arch/sparc64/prom/p1275.c

@@ -16,6 +16,7 @@
 #include <asm/system.h>
 #include <asm/spitfire.h>
 #include <asm/pstate.h>
+#include <asm/ldc.h>
 
 struct {
 	long prom_callback;			/* 0x00 */

arch/sparc64/prom/tree.c

@@ -13,6 +13,7 @@
 
 #include <asm/openprom.h>
 #include <asm/oplib.h>
+#include <asm/ldc.h>
 
 /* Return the child of node 'node' or zero if no this node has no
  * direct descendent.
@@ -261,9 +262,17 @@ int prom_node_has_property(int node, const char *prop)
 int
 prom_setprop(int node, const char *pname, char *value, int size)
 {
-	if(size == 0) return 0;
-	if((pname == 0) || (value == 0)) return 0;
+	if (size == 0)
+		return 0;
+	if ((pname == 0) || (value == 0))
+		return 0;
 	
+#ifdef CONFIG_SUN_LDOMS
+	if (ldom_domaining_enabled) {
+		ldom_set_var(pname, value);
+		return 0;
+	}
+#endif
 	return p1275_cmd ("setprop", P1275_ARG(1,P1275_ARG_IN_STRING)|
 					  P1275_ARG(2,P1275_ARG_IN_BUF)|
 					  P1275_INOUT(4, 1), 

drivers/block/Kconfig

@@ -423,6 +423,13 @@ config ATA_OVER_ETH
 	This driver provides Support for ATA over Ethernet block
 	devices like the Coraid EtherDrive (R) Storage Blade.
 
+config SUNVDC
+	tristate "Sun Virtual Disk Client support"
+	depends on SUN_LDOMS
+	help
+	  Support for virtual disk devices as a client under Sun
+	  Logical Domains.
+
 source "drivers/s390/block/Kconfig"
 
 endif # BLK_DEV

drivers/block/Makefile

@@ -19,6 +19,7 @@ obj-$(CONFIG_BLK_CPQ_DA)	+= cpqarray.o
 obj-$(CONFIG_BLK_CPQ_CISS_DA)  += cciss.o
 obj-$(CONFIG_BLK_DEV_DAC960)	+= DAC960.o
 obj-$(CONFIG_CDROM_PKTCDVD)	+= pktcdvd.o
+obj-$(CONFIG_SUNVDC)		+= sunvdc.o
 
 obj-$(CONFIG_BLK_DEV_UMEM)	+= umem.o
 obj-$(CONFIG_BLK_DEV_NBD)	+= nbd.o

drivers/block/sunvdc.c

@@ -0,0 +1,972 @@
+/* sunvdc.c: Sun LDOM Virtual Disk Client.
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/blkdev.h>
+#include <linux/hdreg.h>
+#include <linux/genhd.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/list.h>
+
+#include <asm/vio.h>
+#include <asm/ldc.h>
+
+#define DRV_MODULE_NAME		"sunvdc"
+#define PFX DRV_MODULE_NAME	": "
+#define DRV_MODULE_VERSION	"1.0"
+#define DRV_MODULE_RELDATE	"June 25, 2007"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+MODULE_AUTHOR("David S. Miller (davem@davemloft.net)");
+MODULE_DESCRIPTION("Sun LDOM virtual disk client driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+#define VDC_TX_RING_SIZE	256
+
+#define WAITING_FOR_LINK_UP	0x01
+#define WAITING_FOR_TX_SPACE	0x02
+#define WAITING_FOR_GEN_CMD	0x04
+#define WAITING_FOR_ANY		-1
+
+struct vdc_req_entry {
+	struct request		*req;
+};
+
+struct vdc_port {
+	struct vio_driver_state	vio;
+
+	struct vdc		*vp;
+
+	struct gendisk		*disk;
+
+	struct vdc_completion	*cmp;
+
+	u64			req_id;
+	u64			seq;
+	struct vdc_req_entry	rq_arr[VDC_TX_RING_SIZE];
+
+	unsigned long		ring_cookies;
+
+	u64			max_xfer_size;
+	u32			vdisk_block_size;
+
+	/* The server fills these in for us in the disk attribute
+	 * ACK packet.
+	 */
+	u64			operations;
+	u32			vdisk_size;
+	u8			vdisk_type;
+	u8			dev_no;
+
+	char			disk_name[32];
+
+	struct vio_disk_geom	geom;
+	struct vio_disk_vtoc	label;
+
+	struct list_head	list;
+};
+
+static inline struct vdc_port *to_vdc_port(struct vio_driver_state *vio)
+{
+	return container_of(vio, struct vdc_port, vio);
+}
+
+struct vdc {
+	/* Protects prot_list.  */
+	spinlock_t		lock;
+
+	struct vio_dev		*dev;
+
+	struct list_head	port_list;
+};
+
+/* Ordered from largest major to lowest */
+static struct vio_version vdc_versions[] = {
+	{ .major = 1, .minor = 0 },
+};
+
+#define VDCBLK_NAME	"vdisk"
+static int vdc_major;
+#define PARTITION_SHIFT	3
+
+static inline u32 vdc_tx_dring_avail(struct vio_dring_state *dr)
+{
+	return vio_dring_avail(dr, VDC_TX_RING_SIZE);
+}
+
+static int vdc_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+{
+	struct gendisk *disk = bdev->bd_disk;
+	struct vdc_port *port = disk->private_data;
+
+	geo->heads = (u8) port->geom.num_hd;
+	geo->sectors = (u8) port->geom.num_sec;
+	geo->cylinders = port->geom.num_cyl;
+
+	return 0;
+}
+
+static struct block_device_operations vdc_fops = {
+	.owner		= THIS_MODULE,
+	.getgeo		= vdc_getgeo,
+};
+
+static void vdc_finish(struct vio_driver_state *vio, int err, int waiting_for)
+{
+	if (vio->cmp &&
+	    (waiting_for == -1 ||
+	     vio->cmp->waiting_for == waiting_for)) {
+		vio->cmp->err = err;
+		complete(&vio->cmp->com);
+		vio->cmp = NULL;
+	}
+}
+
+static void vdc_handshake_complete(struct vio_driver_state *vio)
+{
+	vdc_finish(vio, 0, WAITING_FOR_LINK_UP);
+}
+
+static int vdc_handle_unknown(struct vdc_port *port, void *arg)
+{
+	struct vio_msg_tag *pkt = arg;
+
+	printk(KERN_ERR PFX "Received unknown msg [%02x:%02x:%04x:%08x]\n",
+	       pkt->type, pkt->stype, pkt->stype_env, pkt->sid);
+	printk(KERN_ERR PFX "Resetting connection.\n");
+
+	ldc_disconnect(port->vio.lp);
+
+	return -ECONNRESET;
+}
+
+static int vdc_send_attr(struct vio_driver_state *vio)
+{
+	struct vdc_port *port = to_vdc_port(vio);
+	struct vio_disk_attr_info pkt;
+
+	memset(&pkt, 0, sizeof(pkt));
+
+	pkt.tag.type = VIO_TYPE_CTRL;
+	pkt.tag.stype = VIO_SUBTYPE_INFO;
+	pkt.tag.stype_env = VIO_ATTR_INFO;
+	pkt.tag.sid = vio_send_sid(vio);
+
+	pkt.xfer_mode = VIO_DRING_MODE;
+	pkt.vdisk_block_size = port->vdisk_block_size;
+	pkt.max_xfer_size = port->max_xfer_size;
+
+	viodbg(HS, "SEND ATTR xfer_mode[0x%x] blksz[%u] max_xfer[%lu]\n",
+	       pkt.xfer_mode, pkt.vdisk_block_size, pkt.max_xfer_size);
+
+	return vio_ldc_send(&port->vio, &pkt, sizeof(pkt));
+}
+
+static int vdc_handle_attr(struct vio_driver_state *vio, void *arg)
+{
+	struct vdc_port *port = to_vdc_port(vio);
+	struct vio_disk_attr_info *pkt = arg;
+
+	viodbg(HS, "GOT ATTR stype[0x%x] ops[%lx] disk_size[%lu] disk_type[%x] "
+	       "xfer_mode[0x%x] blksz[%u] max_xfer[%lu]\n",
+	       pkt->tag.stype, pkt->operations,
+	       pkt->vdisk_size, pkt->vdisk_type,
+	       pkt->xfer_mode, pkt->vdisk_block_size,
+	       pkt->max_xfer_size);
+
+	if (pkt->tag.stype == VIO_SUBTYPE_ACK) {
+		switch (pkt->vdisk_type) {
+		case VD_DISK_TYPE_DISK:
+		case VD_DISK_TYPE_SLICE:
+			break;
+
+		default:
+			printk(KERN_ERR PFX "%s: Bogus vdisk_type 0x%x\n",
+			       vio->name, pkt->vdisk_type);
+			return -ECONNRESET;
+		}
+
+		if (pkt->vdisk_block_size > port->vdisk_block_size) {
+			printk(KERN_ERR PFX "%s: BLOCK size increased "
+			       "%u --> %u\n",
+			       vio->name,
+			       port->vdisk_block_size, pkt->vdisk_block_size);
+			return -ECONNRESET;
+		}
+
+		port->operations = pkt->operations;
+		port->vdisk_size = pkt->vdisk_size;
+		port->vdisk_type = pkt->vdisk_type;
+		if (pkt->max_xfer_size < port->max_xfer_size)
+			port->max_xfer_size = pkt->max_xfer_size;
+		port->vdisk_block_size = pkt->vdisk_block_size;
+		return 0;
+	} else {
+		printk(KERN_ERR PFX "%s: Attribute NACK\n", vio->name);
+
+		return -ECONNRESET;
+	}
+}
+
+static void vdc_end_special(struct vdc_port *port, struct vio_disk_desc *desc)
+{
+	int err = desc->status;
+
+	vdc_finish(&port->vio, -err, WAITING_FOR_GEN_CMD);
+}
+
+static void vdc_end_request(struct request *req, int uptodate, int num_sectors)
+{
+	if (end_that_request_first(req, uptodate, num_sectors))
+		return;
+	add_disk_randomness(req->rq_disk);
+	end_that_request_last(req, uptodate);
+}
+
+static void vdc_end_one(struct vdc_port *port, struct vio_dring_state *dr,
+			unsigned int index)
+{
+	struct vio_disk_desc *desc = vio_dring_entry(dr, index);
+	struct vdc_req_entry *rqe = &port->rq_arr[index];
+	struct request *req;
+
+	if (unlikely(desc->hdr.state != VIO_DESC_DONE))
+		return;
+
+	ldc_unmap(port->vio.lp, desc->cookies, desc->ncookies);
+	desc->hdr.state = VIO_DESC_FREE;
+	dr->cons = (index + 1) & (VDC_TX_RING_SIZE - 1);
+
+	req = rqe->req;
+	if (req == NULL) {
+		vdc_end_special(port, desc);
+		return;
+	}
+
+	rqe->req = NULL;
+
+	vdc_end_request(req, !desc->status, desc->size >> 9);
+
+	if (blk_queue_stopped(port->disk->queue))
+		blk_start_queue(port->disk->queue);
+}
+
+static int vdc_ack(struct vdc_port *port, void *msgbuf)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data *pkt = msgbuf;
+
+	if (unlikely(pkt->dring_ident != dr->ident ||
+		     pkt->start_idx != pkt->end_idx ||
+		     pkt->start_idx >= VDC_TX_RING_SIZE))
+		return 0;
+
+	vdc_end_one(port, dr, pkt->start_idx);
+
+	return 0;
+}
+
+static int vdc_nack(struct vdc_port *port, void *msgbuf)
+{
+	/* XXX Implement me XXX */
+	return 0;
+}
+
+static void vdc_event(void *arg, int event)
+{
+	struct vdc_port *port = arg;
+	struct vio_driver_state *vio = &port->vio;
+	unsigned long flags;
+	int err;
+
+	spin_lock_irqsave(&vio->lock, flags);
+
+	if (unlikely(event == LDC_EVENT_RESET ||
+		     event == LDC_EVENT_UP)) {
+		vio_link_state_change(vio, event);
+		spin_unlock_irqrestore(&vio->lock, flags);
+		return;
+	}
+
+	if (unlikely(event != LDC_EVENT_DATA_READY)) {
+		printk(KERN_WARNING PFX "Unexpected LDC event %d\n", event);
+		spin_unlock_irqrestore(&vio->lock, flags);
+		return;
+	}
+
+	err = 0;
+	while (1) {
+		union {
+			struct vio_msg_tag tag;
+			u64 raw[8];
+		} msgbuf;
+
+		err = ldc_read(vio->lp, &msgbuf, sizeof(msgbuf));
+		if (unlikely(err < 0)) {
+			if (err == -ECONNRESET)
+				vio_conn_reset(vio);
+			break;
+		}
+		if (err == 0)
+			break;
+		viodbg(DATA, "TAG [%02x:%02x:%04x:%08x]\n",
+		       msgbuf.tag.type,
+		       msgbuf.tag.stype,
+		       msgbuf.tag.stype_env,
+		       msgbuf.tag.sid);
+		err = vio_validate_sid(vio, &msgbuf.tag);
+		if (err < 0)
+			break;
+
+		if (likely(msgbuf.tag.type == VIO_TYPE_DATA)) {
+			if (msgbuf.tag.stype == VIO_SUBTYPE_ACK)
+				err = vdc_ack(port, &msgbuf);
+			else if (msgbuf.tag.stype == VIO_SUBTYPE_NACK)
+				err = vdc_nack(port, &msgbuf);
+			else
+				err = vdc_handle_unknown(port, &msgbuf);
+		} else if (msgbuf.tag.type == VIO_TYPE_CTRL) {
+			err = vio_control_pkt_engine(vio, &msgbuf);
+		} else {
+			err = vdc_handle_unknown(port, &msgbuf);
+		}
+		if (err < 0)
+			break;
+	}
+	if (err < 0)
+		vdc_finish(&port->vio, err, WAITING_FOR_ANY);
+	spin_unlock_irqrestore(&vio->lock, flags);
+}
+
+static int __vdc_tx_trigger(struct vdc_port *port)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data hdr = {
+		.tag = {
+			.type		= VIO_TYPE_DATA,
+			.stype		= VIO_SUBTYPE_INFO,
+			.stype_env	= VIO_DRING_DATA,
+			.sid		= vio_send_sid(&port->vio),
+		},
+		.dring_ident		= dr->ident,
+		.start_idx		= dr->prod,
+		.end_idx		= dr->prod,
+	};
+	int err, delay;
+
+	hdr.seq = dr->snd_nxt;
+	delay = 1;
+	do {
+		err = vio_ldc_send(&port->vio, &hdr, sizeof(hdr));
+		if (err > 0) {
+			dr->snd_nxt++;
+			break;
+		}
+		udelay(delay);
+		if ((delay <<= 1) > 128)
+			delay = 128;
+	} while (err == -EAGAIN);
+
+	return err;
+}
+
+static int __send_request(struct request *req)
+{
+	struct vdc_port *port = req->rq_disk->private_data;
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct scatterlist sg[port->ring_cookies];
+	struct vdc_req_entry *rqe;
+	struct vio_disk_desc *desc;
+	unsigned int map_perm;
+	int nsg, err, i;
+	u64 len;
+	u8 op;
+
+	map_perm = LDC_MAP_SHADOW | LDC_MAP_DIRECT | LDC_MAP_IO;
+
+	if (rq_data_dir(req) == READ) {
+		map_perm |= LDC_MAP_W;
+		op = VD_OP_BREAD;
+	} else {
+		map_perm |= LDC_MAP_R;
+		op = VD_OP_BWRITE;
+	}
+
+	nsg = blk_rq_map_sg(req->q, req, sg);
+
+	len = 0;
+	for (i = 0; i < nsg; i++)
+		len += sg[i].length;
+
+	if (unlikely(vdc_tx_dring_avail(dr) < 1)) {
+		blk_stop_queue(port->disk->queue);
+		err = -ENOMEM;
+		goto out;
+	}
+
+	desc = vio_dring_cur(dr);
+
+	err = ldc_map_sg(port->vio.lp, sg, nsg,
+			 desc->cookies, port->ring_cookies,
+			 map_perm);
+	if (err < 0) {
+		printk(KERN_ERR PFX "ldc_map_sg() failure, err=%d.\n", err);
+		return err;
+	}
+
+	rqe = &port->rq_arr[dr->prod];
+	rqe->req = req;
+
+	desc->hdr.ack = VIO_ACK_ENABLE;
+	desc->req_id = port->req_id;
+	desc->operation = op;
+	if (port->vdisk_type == VD_DISK_TYPE_DISK) {
+		desc->slice = 2;
+	} else {
+		desc->slice = 0;
+	}
+	desc->status = ~0;
+	desc->offset = (req->sector << 9) / port->vdisk_block_size;
+	desc->size = len;
+	desc->ncookies = err;
+
+	/* This has to be a non-SMP write barrier because we are writing
+	 * to memory which is shared with the peer LDOM.
+	 */
+	wmb();
+	desc->hdr.state = VIO_DESC_READY;
+
+	err = __vdc_tx_trigger(port);
+	if (err < 0) {
+		printk(KERN_ERR PFX "vdc_tx_trigger() failure, err=%d\n", err);
+	} else {
+		port->req_id++;
+		dr->prod = (dr->prod + 1) & (VDC_TX_RING_SIZE - 1);
+	}
+out:
+
+	return err;
+}
+
+static void do_vdc_request(request_queue_t *q)
+{
+	while (1) {
+		struct request *req = elv_next_request(q);
+
+		if (!req)
+			break;
+
+		blkdev_dequeue_request(req);
+		if (__send_request(req) < 0)
+			vdc_end_request(req, 0, req->hard_nr_sectors);
+	}
+}
+
+static int generic_request(struct vdc_port *port, u8 op, void *buf, int len)
+{
+	struct vio_dring_state *dr;
+	struct vio_completion comp;
+	struct vio_disk_desc *desc;
+	unsigned int map_perm;
+	unsigned long flags;
+	int op_len, err;
+	void *req_buf;
+
+	if (!(((u64)1 << ((u64)op - 1)) & port->operations))
+		return -EOPNOTSUPP;
+
+	switch (op) {
+	case VD_OP_BREAD:
+	case VD_OP_BWRITE:
+	default:
+		return -EINVAL;
+
+	case VD_OP_FLUSH:
+		op_len = 0;
+		map_perm = 0;
+		break;
+
+	case VD_OP_GET_WCE:
+		op_len = sizeof(u32);
+		map_perm = LDC_MAP_W;
+		break;
+
+	case VD_OP_SET_WCE:
+		op_len = sizeof(u32);
+		map_perm = LDC_MAP_R;
+		break;
+
+	case VD_OP_GET_VTOC:
+		op_len = sizeof(struct vio_disk_vtoc);
+		map_perm = LDC_MAP_W;
+		break;
+
+	case VD_OP_SET_VTOC:
+		op_len = sizeof(struct vio_disk_vtoc);
+		map_perm = LDC_MAP_R;
+		break;
+
+	case VD_OP_GET_DISKGEOM:
+		op_len = sizeof(struct vio_disk_geom);
+		map_perm = LDC_MAP_W;
+		break;
+
+	case VD_OP_SET_DISKGEOM:
+		op_len = sizeof(struct vio_disk_geom);
+		map_perm = LDC_MAP_R;
+		break;
+
+	case VD_OP_SCSICMD:
+		op_len = 16;
+		map_perm = LDC_MAP_RW;
+		break;
+
+	case VD_OP_GET_DEVID:
+		op_len = sizeof(struct vio_disk_devid);
+		map_perm = LDC_MAP_W;
+		break;
+
+	case VD_OP_GET_EFI:
+	case VD_OP_SET_EFI:
+		return -EOPNOTSUPP;
+		break;
+	};
+
+	map_perm |= LDC_MAP_SHADOW | LDC_MAP_DIRECT | LDC_MAP_IO;
+
+	op_len = (op_len + 7) & ~7;
+	req_buf = kzalloc(op_len, GFP_KERNEL);
+	if (!req_buf)
+		return -ENOMEM;
+
+	if (len > op_len)
+		len = op_len;
+
+	if (map_perm & LDC_MAP_R)
+		memcpy(req_buf, buf, len);
+
+	spin_lock_irqsave(&port->vio.lock, flags);
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+
+	/* XXX If we want to use this code generically we have to
+	 * XXX handle TX ring exhaustion etc.
+	 */
+	desc = vio_dring_cur(dr);
+
+	err = ldc_map_single(port->vio.lp, req_buf, op_len,
+			     desc->cookies, port->ring_cookies,
+			     map_perm);
+	if (err < 0) {
+		spin_unlock_irqrestore(&port->vio.lock, flags);
+		kfree(req_buf);
+		return err;
+	}
+
+	init_completion(&comp.com);
+	comp.waiting_for = WAITING_FOR_GEN_CMD;
+	port->vio.cmp = &comp;
+
+	desc->hdr.ack = VIO_ACK_ENABLE;
+	desc->req_id = port->req_id;
+	desc->operation = op;
+	desc->slice = 0;
+	desc->status = ~0;
+	desc->offset = 0;
+	desc->size = op_len;
+	desc->ncookies = err;
+
+	/* This has to be a non-SMP write barrier because we are writing
+	 * to memory which is shared with the peer LDOM.
+	 */
+	wmb();
+	desc->hdr.state = VIO_DESC_READY;
+
+	err = __vdc_tx_trigger(port);
+	if (err >= 0) {
+		port->req_id++;
+		dr->prod = (dr->prod + 1) & (VDC_TX_RING_SIZE - 1);
+		spin_unlock_irqrestore(&port->vio.lock, flags);
+
+		wait_for_completion(&comp.com);
+		err = comp.err;
+	} else {
+		port->vio.cmp = NULL;
+		spin_unlock_irqrestore(&port->vio.lock, flags);
+	}
+
+	if (map_perm & LDC_MAP_W)
+		memcpy(buf, req_buf, len);
+
+	kfree(req_buf);
+
+	return err;
+}
+
+static int __devinit vdc_alloc_tx_ring(struct vdc_port *port)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	unsigned long len, entry_size;
+	int ncookies;
+	void *dring;
+
+	entry_size = sizeof(struct vio_disk_desc) +
+		(sizeof(struct ldc_trans_cookie) * port->ring_cookies);
+	len = (VDC_TX_RING_SIZE * entry_size);
+
+	ncookies = VIO_MAX_RING_COOKIES;
+	dring = ldc_alloc_exp_dring(port->vio.lp, len,
+				    dr->cookies, &ncookies,
+				    (LDC_MAP_SHADOW |
+				     LDC_MAP_DIRECT |
+				     LDC_MAP_RW));
+	if (IS_ERR(dring))
+		return PTR_ERR(dring);
+
+	dr->base = dring;
+	dr->entry_size = entry_size;
+	dr->num_entries = VDC_TX_RING_SIZE;
+	dr->prod = dr->cons = 0;
+	dr->pending = VDC_TX_RING_SIZE;
+	dr->ncookies = ncookies;
+
+	return 0;
+}
+
+static void vdc_free_tx_ring(struct vdc_port *port)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+
+	if (dr->base) {
+		ldc_free_exp_dring(port->vio.lp, dr->base,
+				   (dr->entry_size * dr->num_entries),
+				   dr->cookies, dr->ncookies);
+		dr->base = NULL;
+		dr->entry_size = 0;
+		dr->num_entries = 0;
+		dr->pending = 0;
+		dr->ncookies = 0;
+	}
+}
+
+static int probe_disk(struct vdc_port *port)
+{
+	struct vio_completion comp;
+	struct request_queue *q;
+	struct gendisk *g;
+	int err;
+
+	init_completion(&comp.com);
+	comp.err = 0;
+	comp.waiting_for = WAITING_FOR_LINK_UP;
+	port->vio.cmp = &comp;
+
+	vio_port_up(&port->vio);
+
+	wait_for_completion(&comp.com);
+	if (comp.err)
+		return comp.err;
+
+	err = generic_request(port, VD_OP_GET_VTOC,
+			      &port->label, sizeof(port->label));
+	if (err < 0) {
+		printk(KERN_ERR PFX "VD_OP_GET_VTOC returns error %d\n", err);
+		return err;
+	}
+
+	err = generic_request(port, VD_OP_GET_DISKGEOM,
+			      &port->geom, sizeof(port->geom));
+	if (err < 0) {
+		printk(KERN_ERR PFX "VD_OP_GET_DISKGEOM returns "
+		       "error %d\n", err);
+		return err;
+	}
+
+	port->vdisk_size = ((u64)port->geom.num_cyl *
+			    (u64)port->geom.num_hd *
+			    (u64)port->geom.num_sec);
+
+	q = blk_init_queue(do_vdc_request, &port->vio.lock);
+	if (!q) {
+		printk(KERN_ERR PFX "%s: Could not allocate queue.\n",
+		       port->vio.name);
+		return -ENOMEM;
+	}
+	g = alloc_disk(1 << PARTITION_SHIFT);
+	if (!g) {
+		printk(KERN_ERR PFX "%s: Could not allocate gendisk.\n",
+		       port->vio.name);
+		blk_cleanup_queue(q);
+		return -ENOMEM;
+	}
+
+	port->disk = g;
+
+	blk_queue_max_hw_segments(q, port->ring_cookies);
+	blk_queue_max_phys_segments(q, port->ring_cookies);
+	blk_queue_max_sectors(q, port->max_xfer_size);
+	g->major = vdc_major;
+	g->first_minor = port->dev_no << PARTITION_SHIFT;
+	strcpy(g->disk_name, port->disk_name);
+
+	g->fops = &vdc_fops;
+	g->queue = q;
+	g->private_data = port;
+	g->driverfs_dev = &port->vio.vdev->dev;
+
+	set_capacity(g, port->vdisk_size);
+
+	printk(KERN_INFO PFX "%s: %u sectors (%u MB)\n",
+	       g->disk_name,
+	       port->vdisk_size, (port->vdisk_size >> (20 - 9)));
+
+	add_disk(g);
+
+	return 0;
+}
+
+static struct ldc_channel_config vdc_ldc_cfg = {
+	.event		= vdc_event,
+	.mtu		= 64,
+	.mode		= LDC_MODE_UNRELIABLE,
+};
+
+static struct vio_driver_ops vdc_vio_ops = {
+	.send_attr		= vdc_send_attr,
+	.handle_attr		= vdc_handle_attr,
+	.handshake_complete	= vdc_handshake_complete,
+};
+
+static int __devinit vdc_port_probe(struct vio_dev *vdev,
+				    const struct vio_device_id *id)
+{
+	struct mdesc_handle *hp;
+	struct vdc_port *port;
+	unsigned long flags;
+	struct vdc *vp;
+	const u64 *port_id;
+	int err;
+
+	vp = dev_get_drvdata(vdev->dev.parent);
+	if (!vp) {
+		printk(KERN_ERR PFX "Cannot find port parent vdc.\n");
+		return -ENODEV;
+	}
+
+	hp = mdesc_grab();
+
+	port_id = mdesc_get_property(hp, vdev->mp, "id", NULL);
+	err = -ENODEV;
+	if (!port_id) {
+		printk(KERN_ERR PFX "Port lacks id property.\n");
+		goto err_out_release_mdesc;
+	}
+	if ((*port_id << PARTITION_SHIFT) & ~(u64)MINORMASK) {
+		printk(KERN_ERR PFX "Port id [%lu] too large.\n", *port_id);
+		goto err_out_release_mdesc;
+	}
+
+	port = kzalloc(sizeof(*port), GFP_KERNEL);
+	err = -ENOMEM;
+	if (!port) {
+		printk(KERN_ERR PFX "Cannot allocate vdc_port.\n");
+		goto err_out_release_mdesc;
+	}
+
+	port->vp = vp;
+	port->dev_no = *port_id;
+
+	if (port->dev_no >= 26)
+		snprintf(port->disk_name, sizeof(port->disk_name),
+			 VDCBLK_NAME "%c%c",
+			 'a' + (port->dev_no / 26) - 1,
+			 'a' + (port->dev_no % 26));
+	else
+		snprintf(port->disk_name, sizeof(port->disk_name),
+			 VDCBLK_NAME "%c", 'a' + (port->dev_no % 26));
+
+	err = vio_driver_init(&port->vio, vdev, VDEV_DISK,
+			      vdc_versions, ARRAY_SIZE(vdc_versions),
+			      &vdc_vio_ops, port->disk_name);
+	if (err)
+		goto err_out_free_port;
+
+	port->vdisk_block_size = 512;
+	port->max_xfer_size = ((128 * 1024) / port->vdisk_block_size);
+	port->ring_cookies = ((port->max_xfer_size *
+			       port->vdisk_block_size) / PAGE_SIZE) + 2;
+
+	err = vio_ldc_alloc(&port->vio, &vdc_ldc_cfg, port);
+	if (err)
+		goto err_out_free_port;
+
+	err = vdc_alloc_tx_ring(port);
+	if (err)
+		goto err_out_free_ldc;
+
+	err = probe_disk(port);
+	if (err)
+		goto err_out_free_tx_ring;
+
+	INIT_LIST_HEAD(&port->list);
+
+	spin_lock_irqsave(&vp->lock, flags);
+	list_add(&port->list, &vp->port_list);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	dev_set_drvdata(&vdev->dev, port);
+
+	mdesc_release(hp);
+
+	return 0;
+
+err_out_free_tx_ring:
+	vdc_free_tx_ring(port);
+
+err_out_free_ldc:
+	vio_ldc_free(&port->vio);
+
+err_out_free_port:
+	kfree(port);
+
+err_out_release_mdesc:
+	mdesc_release(hp);
+	return err;
+}
+
+static int vdc_port_remove(struct vio_dev *vdev)
+{
+	struct vdc_port *port = dev_get_drvdata(&vdev->dev);
+
+	if (port) {
+		del_timer_sync(&port->vio.timer);
+
+		vdc_free_tx_ring(port);
+		vio_ldc_free(&port->vio);
+
+		dev_set_drvdata(&vdev->dev, NULL);
+
+		kfree(port);
+	}
+	return 0;
+}
+
+static struct vio_device_id vdc_port_match[] = {
+	{
+		.type = "vdc-port",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vdc_match);
+
+static struct vio_driver vdc_port_driver = {
+	.id_table	= vdc_port_match,
+	.probe		= vdc_port_probe,
+	.remove		= vdc_port_remove,
+	.driver		= {
+		.name	= "vdc_port",
+		.owner	= THIS_MODULE,
+	}
+};
+
+static int __devinit vdc_probe(struct vio_dev *vdev,
+			       const struct vio_device_id *id)
+{
+	static int vdc_version_printed;
+	struct vdc *vp;
+
+	if (vdc_version_printed++ == 0)
+		printk(KERN_INFO "%s", version);
+
+	vp = kzalloc(sizeof(struct vdc), GFP_KERNEL);
+	if (!vp)
+		return -ENOMEM;
+
+	spin_lock_init(&vp->lock);
+	vp->dev = vdev;
+	INIT_LIST_HEAD(&vp->port_list);
+
+	dev_set_drvdata(&vdev->dev, vp);
+
+	return 0;
+}
+
+static int vdc_remove(struct vio_dev *vdev)
+{
+
+	struct vdc *vp = dev_get_drvdata(&vdev->dev);
+
+	if (vp) {
+		kfree(vp);
+		dev_set_drvdata(&vdev->dev, NULL);
+	}
+	return 0;
+}
+
+static struct vio_device_id vdc_match[] = {
+	{
+		.type = "block",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vdc_match);
+
+static struct vio_driver vdc_driver = {
+	.id_table	= vdc_match,
+	.probe		= vdc_probe,
+	.remove		= vdc_remove,
+	.driver		= {
+		.name	= "vdc",
+		.owner	= THIS_MODULE,
+	}
+};
+
+static int __init vdc_init(void)
+{
+	int err;
+
+	err = register_blkdev(0, VDCBLK_NAME);
+	if (err < 0)
+		goto out_err;
+
+	vdc_major = err;
+	err = vio_register_driver(&vdc_driver);
+	if (err)
+		goto out_unregister_blkdev;
+
+	err = vio_register_driver(&vdc_port_driver);
+	if (err)
+		goto out_unregister_vdc;
+
+	return 0;
+
+out_unregister_vdc:
+	vio_unregister_driver(&vdc_driver);
+
+out_unregister_blkdev:
+	unregister_blkdev(vdc_major, VDCBLK_NAME);
+	vdc_major = 0;
+
+out_err:
+	return err;
+}
+
+static void __exit vdc_exit(void)
+{
+	vio_unregister_driver(&vdc_port_driver);
+	vio_unregister_driver(&vdc_driver);
+	unregister_blkdev(vdc_major, VDCBLK_NAME);
+}
+
+module_init(vdc_init);
+module_exit(vdc_exit);

drivers/net/Kconfig

@@ -604,6 +604,12 @@ config CASSINI
 	  Support for the Sun Cassini chip, aka Sun GigaSwift Ethernet. See also
 	  <http://www.sun.com/products-n-solutions/hardware/docs/pdf/817-4341-10.pdf>
 
+config SUNVNET
+	tristate "Sun Virtual Network support"
+	depends on SUN_LDOMS
+	help
+	  Support for virtual network devices under Sun Logical Domains.
+
 config NET_VENDOR_3COM
 	bool "3COM cards"
 	depends on ISA || EISA || MCA || PCI

drivers/net/Makefile

@@ -34,6 +34,7 @@ obj-$(CONFIG_SUNBMAC) += sunbmac.o
 obj-$(CONFIG_MYRI_SBUS) += myri_sbus.o
 obj-$(CONFIG_SUNGEM) += sungem.o sungem_phy.o
 obj-$(CONFIG_CASSINI) += cassini.o
+obj-$(CONFIG_SUNVNET) += sunvnet.o
 
 obj-$(CONFIG_MACE) += mace.o
 obj-$(CONFIG_BMAC) += bmac.o

drivers/net/sunvnet.c

@@ -0,0 +1,1164 @@
+/* sunvnet.c: Sun LDOM Virtual Network Driver.
+ *
+ * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/etherdevice.h>
+
+#include <asm/vio.h>
+#include <asm/ldc.h>
+
+#include "sunvnet.h"
+
+#define DRV_MODULE_NAME		"sunvnet"
+#define PFX DRV_MODULE_NAME	": "
+#define DRV_MODULE_VERSION	"1.0"
+#define DRV_MODULE_RELDATE	"June 25, 2007"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+MODULE_AUTHOR("David S. Miller (davem@davemloft.net)");
+MODULE_DESCRIPTION("Sun LDOM virtual network driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+/* Ordered from largest major to lowest */
+static struct vio_version vnet_versions[] = {
+	{ .major = 1, .minor = 0 },
+};
+
+static inline u32 vnet_tx_dring_avail(struct vio_dring_state *dr)
+{
+	return vio_dring_avail(dr, VNET_TX_RING_SIZE);
+}
+
+static int vnet_handle_unknown(struct vnet_port *port, void *arg)
+{
+	struct vio_msg_tag *pkt = arg;
+
+	printk(KERN_ERR PFX "Received unknown msg [%02x:%02x:%04x:%08x]\n",
+	       pkt->type, pkt->stype, pkt->stype_env, pkt->sid);
+	printk(KERN_ERR PFX "Resetting connection.\n");
+
+	ldc_disconnect(port->vio.lp);
+
+	return -ECONNRESET;
+}
+
+static int vnet_send_attr(struct vio_driver_state *vio)
+{
+	struct vnet_port *port = to_vnet_port(vio);
+	struct net_device *dev = port->vp->dev;
+	struct vio_net_attr_info pkt;
+	int i;
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.tag.type = VIO_TYPE_CTRL;
+	pkt.tag.stype = VIO_SUBTYPE_INFO;
+	pkt.tag.stype_env = VIO_ATTR_INFO;
+	pkt.tag.sid = vio_send_sid(vio);
+	pkt.xfer_mode = VIO_DRING_MODE;
+	pkt.addr_type = VNET_ADDR_ETHERMAC;
+	pkt.ack_freq = 0;
+	for (i = 0; i < 6; i++)
+		pkt.addr |= (u64)dev->dev_addr[i] << ((5 - i) * 8);
+	pkt.mtu = ETH_FRAME_LEN;
+
+	viodbg(HS, "SEND NET ATTR xmode[0x%x] atype[0x%x] addr[%llx] "
+	       "ackfreq[%u] mtu[%llu]\n",
+	       pkt.xfer_mode, pkt.addr_type,
+	       (unsigned long long) pkt.addr,
+	       pkt.ack_freq,
+	       (unsigned long long) pkt.mtu);
+
+	return vio_ldc_send(vio, &pkt, sizeof(pkt));
+}
+
+static int handle_attr_info(struct vio_driver_state *vio,
+			    struct vio_net_attr_info *pkt)
+{
+	viodbg(HS, "GOT NET ATTR INFO xmode[0x%x] atype[0x%x] addr[%llx] "
+	       "ackfreq[%u] mtu[%llu]\n",
+	       pkt->xfer_mode, pkt->addr_type,
+	       (unsigned long long) pkt->addr,
+	       pkt->ack_freq,
+	       (unsigned long long) pkt->mtu);
+
+	pkt->tag.sid = vio_send_sid(vio);
+
+	if (pkt->xfer_mode != VIO_DRING_MODE ||
+	    pkt->addr_type != VNET_ADDR_ETHERMAC ||
+	    pkt->mtu != ETH_FRAME_LEN) {
+		viodbg(HS, "SEND NET ATTR NACK\n");
+
+		pkt->tag.stype = VIO_SUBTYPE_NACK;
+
+		(void) vio_ldc_send(vio, pkt, sizeof(*pkt));
+
+		return -ECONNRESET;
+	} else {
+		viodbg(HS, "SEND NET ATTR ACK\n");
+
+		pkt->tag.stype = VIO_SUBTYPE_ACK;
+
+		return vio_ldc_send(vio, pkt, sizeof(*pkt));
+	}
+
+}
+
+static int handle_attr_ack(struct vio_driver_state *vio,
+			   struct vio_net_attr_info *pkt)
+{
+	viodbg(HS, "GOT NET ATTR ACK\n");
+
+	return 0;
+}
+
+static int handle_attr_nack(struct vio_driver_state *vio,
+			    struct vio_net_attr_info *pkt)
+{
+	viodbg(HS, "GOT NET ATTR NACK\n");
+
+	return -ECONNRESET;
+}
+
+static int vnet_handle_attr(struct vio_driver_state *vio, void *arg)
+{
+	struct vio_net_attr_info *pkt = arg;
+
+	switch (pkt->tag.stype) {
+	case VIO_SUBTYPE_INFO:
+		return handle_attr_info(vio, pkt);
+
+	case VIO_SUBTYPE_ACK:
+		return handle_attr_ack(vio, pkt);
+
+	case VIO_SUBTYPE_NACK:
+		return handle_attr_nack(vio, pkt);
+
+	default:
+		return -ECONNRESET;
+	}
+}
+
+static void vnet_handshake_complete(struct vio_driver_state *vio)
+{
+	struct vio_dring_state *dr;
+
+	dr = &vio->drings[VIO_DRIVER_RX_RING];
+	dr->snd_nxt = dr->rcv_nxt = 1;
+
+	dr = &vio->drings[VIO_DRIVER_TX_RING];
+	dr->snd_nxt = dr->rcv_nxt = 1;
+}
+
+/* The hypervisor interface that implements copying to/from imported
+ * memory from another domain requires that copies are done to 8-byte
+ * aligned buffers, and that the lengths of such copies are also 8-byte
+ * multiples.
+ *
+ * So we align skb->data to an 8-byte multiple and pad-out the data
+ * area so we can round the copy length up to the next multiple of
+ * 8 for the copy.
+ *
+ * The transmitter puts the actual start of the packet 6 bytes into
+ * the buffer it sends over, so that the IP headers after the ethernet
+ * header are aligned properly.  These 6 bytes are not in the descriptor
+ * length, they are simply implied.  This offset is represented using
+ * the VNET_PACKET_SKIP macro.
+ */
+static struct sk_buff *alloc_and_align_skb(struct net_device *dev,
+					   unsigned int len)
+{
+	struct sk_buff *skb = netdev_alloc_skb(dev, len+VNET_PACKET_SKIP+8+8);
+	unsigned long addr, off;
+
+	if (unlikely(!skb))
+		return NULL;
+
+	addr = (unsigned long) skb->data;
+	off = ((addr + 7UL) & ~7UL) - addr;
+	if (off)
+		skb_reserve(skb, off);
+
+	return skb;
+}
+
+static int vnet_rx_one(struct vnet_port *port, unsigned int len,
+		       struct ldc_trans_cookie *cookies, int ncookies)
+{
+	struct net_device *dev = port->vp->dev;
+	unsigned int copy_len;
+	struct sk_buff *skb;
+	int err;
+
+	err = -EMSGSIZE;
+	if (unlikely(len < ETH_ZLEN || len > ETH_FRAME_LEN)) {
+		dev->stats.rx_length_errors++;
+		goto out_dropped;
+	}
+
+	skb = alloc_and_align_skb(dev, len);
+	err = -ENOMEM;
+	if (unlikely(!skb)) {
+		dev->stats.rx_missed_errors++;
+		goto out_dropped;
+	}
+
+	copy_len = (len + VNET_PACKET_SKIP + 7U) & ~7U;
+	skb_put(skb, copy_len);
+	err = ldc_copy(port->vio.lp, LDC_COPY_IN,
+		       skb->data, copy_len, 0,
+		       cookies, ncookies);
+	if (unlikely(err < 0)) {
+		dev->stats.rx_frame_errors++;
+		goto out_free_skb;
+	}
+
+	skb_pull(skb, VNET_PACKET_SKIP);
+	skb_trim(skb, len);
+	skb->protocol = eth_type_trans(skb, dev);
+
+	dev->stats.rx_packets++;
+	dev->stats.rx_bytes += len;
+
+	netif_rx(skb);
+
+	return 0;
+
+out_free_skb:
+	kfree_skb(skb);
+
+out_dropped:
+	dev->stats.rx_dropped++;
+	return err;
+}
+
+static int vnet_send_ack(struct vnet_port *port, struct vio_dring_state *dr,
+			 u32 start, u32 end, u8 vio_dring_state)
+{
+	struct vio_dring_data hdr = {
+		.tag = {
+			.type		= VIO_TYPE_DATA,
+			.stype		= VIO_SUBTYPE_ACK,
+			.stype_env	= VIO_DRING_DATA,
+			.sid		= vio_send_sid(&port->vio),
+		},
+		.dring_ident		= dr->ident,
+		.start_idx		= start,
+		.end_idx		= end,
+		.state			= vio_dring_state,
+	};
+	int err, delay;
+
+	hdr.seq = dr->snd_nxt;
+	delay = 1;
+	do {
+		err = vio_ldc_send(&port->vio, &hdr, sizeof(hdr));
+		if (err > 0) {
+			dr->snd_nxt++;
+			break;
+		}
+		udelay(delay);
+		if ((delay <<= 1) > 128)
+			delay = 128;
+	} while (err == -EAGAIN);
+
+	return err;
+}
+
+static u32 next_idx(u32 idx, struct vio_dring_state *dr)
+{
+	if (++idx == dr->num_entries)
+		idx = 0;
+	return idx;
+}
+
+static u32 prev_idx(u32 idx, struct vio_dring_state *dr)
+{
+	if (idx == 0)
+		idx = dr->num_entries - 1;
+	else
+		idx--;
+
+	return idx;
+}
+
+static struct vio_net_desc *get_rx_desc(struct vnet_port *port,
+					struct vio_dring_state *dr,
+					u32 index)
+{
+	struct vio_net_desc *desc = port->vio.desc_buf;
+	int err;
+
+	err = ldc_get_dring_entry(port->vio.lp, desc, dr->entry_size,
+				  (index * dr->entry_size),
+				  dr->cookies, dr->ncookies);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	return desc;
+}
+
+static int put_rx_desc(struct vnet_port *port,
+		       struct vio_dring_state *dr,
+		       struct vio_net_desc *desc,
+		       u32 index)
+{
+	int err;
+
+	err = ldc_put_dring_entry(port->vio.lp, desc, dr->entry_size,
+				  (index * dr->entry_size),
+				  dr->cookies, dr->ncookies);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int vnet_walk_rx_one(struct vnet_port *port,
+			    struct vio_dring_state *dr,
+			    u32 index, int *needs_ack)
+{
+	struct vio_net_desc *desc = get_rx_desc(port, dr, index);
+	struct vio_driver_state *vio = &port->vio;
+	int err;
+
+	if (IS_ERR(desc))
+		return PTR_ERR(desc);
+
+	viodbg(DATA, "vio_walk_rx_one desc[%02x:%02x:%08x:%08x:%lx:%lx]\n",
+	       desc->hdr.state, desc->hdr.ack,
+	       desc->size, desc->ncookies,
+	       desc->cookies[0].cookie_addr,
+	       desc->cookies[0].cookie_size);
+
+	if (desc->hdr.state != VIO_DESC_READY)
+		return 1;
+	err = vnet_rx_one(port, desc->size, desc->cookies, desc->ncookies);
+	if (err == -ECONNRESET)
+		return err;
+	desc->hdr.state = VIO_DESC_DONE;
+	err = put_rx_desc(port, dr, desc, index);
+	if (err < 0)
+		return err;
+	*needs_ack = desc->hdr.ack;
+	return 0;
+}
+
+static int vnet_walk_rx(struct vnet_port *port, struct vio_dring_state *dr,
+			u32 start, u32 end)
+{
+	struct vio_driver_state *vio = &port->vio;
+	int ack_start = -1, ack_end = -1;
+
+	end = (end == (u32) -1) ? prev_idx(start, dr) : next_idx(end, dr);
+
+	viodbg(DATA, "vnet_walk_rx start[%08x] end[%08x]\n", start, end);
+
+	while (start != end) {
+		int ack = 0, err = vnet_walk_rx_one(port, dr, start, &ack);
+		if (err == -ECONNRESET)
+			return err;
+		if (err != 0)
+			break;
+		if (ack_start == -1)
+			ack_start = start;
+		ack_end = start;
+		start = next_idx(start, dr);
+		if (ack && start != end) {
+			err = vnet_send_ack(port, dr, ack_start, ack_end,
+					    VIO_DRING_ACTIVE);
+			if (err == -ECONNRESET)
+				return err;
+			ack_start = -1;
+		}
+	}
+	if (unlikely(ack_start == -1))
+		ack_start = ack_end = prev_idx(start, dr);
+	return vnet_send_ack(port, dr, ack_start, ack_end, VIO_DRING_STOPPED);
+}
+
+static int vnet_rx(struct vnet_port *port, void *msgbuf)
+{
+	struct vio_dring_data *pkt = msgbuf;
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_RX_RING];
+	struct vio_driver_state *vio = &port->vio;
+
+	viodbg(DATA, "vnet_rx stype_env[%04x] seq[%016lx] rcv_nxt[%016lx]\n",
+	       pkt->tag.stype_env, pkt->seq, dr->rcv_nxt);
+
+	if (unlikely(pkt->tag.stype_env != VIO_DRING_DATA))
+		return 0;
+	if (unlikely(pkt->seq != dr->rcv_nxt)) {
+		printk(KERN_ERR PFX "RX out of sequence seq[0x%lx] "
+		       "rcv_nxt[0x%lx]\n", pkt->seq, dr->rcv_nxt);
+		return 0;
+	}
+
+	dr->rcv_nxt++;
+
+	/* XXX Validate pkt->start_idx and pkt->end_idx XXX */
+
+	return vnet_walk_rx(port, dr, pkt->start_idx, pkt->end_idx);
+}
+
+static int idx_is_pending(struct vio_dring_state *dr, u32 end)
+{
+	u32 idx = dr->cons;
+	int found = 0;
+
+	while (idx != dr->prod) {
+		if (idx == end) {
+			found = 1;
+			break;
+		}
+		idx = next_idx(idx, dr);
+	}
+	return found;
+}
+
+static int vnet_ack(struct vnet_port *port, void *msgbuf)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data *pkt = msgbuf;
+	struct net_device *dev;
+	struct vnet *vp;
+	u32 end;
+
+	if (unlikely(pkt->tag.stype_env != VIO_DRING_DATA))
+		return 0;
+
+	end = pkt->end_idx;
+	if (unlikely(!idx_is_pending(dr, end)))
+		return 0;
+
+	dr->cons = next_idx(end, dr);
+
+	vp = port->vp;
+	dev = vp->dev;
+	if (unlikely(netif_queue_stopped(dev) &&
+		     vnet_tx_dring_avail(dr) >= VNET_TX_WAKEUP_THRESH(dr)))
+		return 1;
+
+	return 0;
+}
+
+static int vnet_nack(struct vnet_port *port, void *msgbuf)
+{
+	/* XXX just reset or similar XXX */
+	return 0;
+}
+
+static void maybe_tx_wakeup(struct vnet *vp)
+{
+	struct net_device *dev = vp->dev;
+
+	netif_tx_lock(dev);
+	if (likely(netif_queue_stopped(dev))) {
+		struct vnet_port *port;
+		int wake = 1;
+
+		list_for_each_entry(port, &vp->port_list, list) {
+			struct vio_dring_state *dr;
+
+			dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+			if (vnet_tx_dring_avail(dr) <
+			    VNET_TX_WAKEUP_THRESH(dr)) {
+				wake = 0;
+				break;
+			}
+		}
+		if (wake)
+			netif_wake_queue(dev);
+	}
+	netif_tx_unlock(dev);
+}
+
+static void vnet_event(void *arg, int event)
+{
+	struct vnet_port *port = arg;
+	struct vio_driver_state *vio = &port->vio;
+	unsigned long flags;
+	int tx_wakeup, err;
+
+	spin_lock_irqsave(&vio->lock, flags);
+
+	if (unlikely(event == LDC_EVENT_RESET ||
+		     event == LDC_EVENT_UP)) {
+		vio_link_state_change(vio, event);
+		spin_unlock_irqrestore(&vio->lock, flags);
+
+		return;
+	}
+
+	if (unlikely(event != LDC_EVENT_DATA_READY)) {
+		printk(KERN_WARNING PFX "Unexpected LDC event %d\n", event);
+		spin_unlock_irqrestore(&vio->lock, flags);
+		return;
+	}
+
+	tx_wakeup = err = 0;
+	while (1) {
+		union {
+			struct vio_msg_tag tag;
+			u64 raw[8];
+		} msgbuf;
+
+		err = ldc_read(vio->lp, &msgbuf, sizeof(msgbuf));
+		if (unlikely(err < 0)) {
+			if (err == -ECONNRESET)
+				vio_conn_reset(vio);
+			break;
+		}
+		if (err == 0)
+			break;
+		viodbg(DATA, "TAG [%02x:%02x:%04x:%08x]\n",
+		       msgbuf.tag.type,
+		       msgbuf.tag.stype,
+		       msgbuf.tag.stype_env,
+		       msgbuf.tag.sid);
+		err = vio_validate_sid(vio, &msgbuf.tag);
+		if (err < 0)
+			break;
+
+		if (likely(msgbuf.tag.type == VIO_TYPE_DATA)) {
+			if (msgbuf.tag.stype == VIO_SUBTYPE_INFO) {
+				err = vnet_rx(port, &msgbuf);
+			} else if (msgbuf.tag.stype == VIO_SUBTYPE_ACK) {
+				err = vnet_ack(port, &msgbuf);
+				if (err > 0)
+					tx_wakeup |= err;
+			} else if (msgbuf.tag.stype == VIO_SUBTYPE_NACK) {
+				err = vnet_nack(port, &msgbuf);
+			}
+		} else if (msgbuf.tag.type == VIO_TYPE_CTRL) {
+			err = vio_control_pkt_engine(vio, &msgbuf);
+			if (err)
+				break;
+		} else {
+			err = vnet_handle_unknown(port, &msgbuf);
+		}
+		if (err == -ECONNRESET)
+			break;
+	}
+	spin_unlock(&vio->lock);
+	if (unlikely(tx_wakeup && err != -ECONNRESET))
+		maybe_tx_wakeup(port->vp);
+	local_irq_restore(flags);
+}
+
+static int __vnet_tx_trigger(struct vnet_port *port)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data hdr = {
+		.tag = {
+			.type		= VIO_TYPE_DATA,
+			.stype		= VIO_SUBTYPE_INFO,
+			.stype_env	= VIO_DRING_DATA,
+			.sid		= vio_send_sid(&port->vio),
+		},
+		.dring_ident		= dr->ident,
+		.start_idx		= dr->prod,
+		.end_idx		= (u32) -1,
+	};
+	int err, delay;
+
+	hdr.seq = dr->snd_nxt;
+	delay = 1;
+	do {
+		err = vio_ldc_send(&port->vio, &hdr, sizeof(hdr));
+		if (err > 0) {
+			dr->snd_nxt++;
+			break;
+		}
+		udelay(delay);
+		if ((delay <<= 1) > 128)
+			delay = 128;
+	} while (err == -EAGAIN);
+
+	return err;
+}
+
+struct vnet_port *__tx_port_find(struct vnet *vp, struct sk_buff *skb)
+{
+	unsigned int hash = vnet_hashfn(skb->data);
+	struct hlist_head *hp = &vp->port_hash[hash];
+	struct hlist_node *n;
+	struct vnet_port *port;
+
+	hlist_for_each_entry(port, n, hp, hash) {
+		if (!compare_ether_addr(port->raddr, skb->data))
+			return port;
+	}
+	port = NULL;
+	if (!list_empty(&vp->port_list))
+		port = list_entry(vp->port_list.next, struct vnet_port, list);
+
+	return port;
+}
+
+struct vnet_port *tx_port_find(struct vnet *vp, struct sk_buff *skb)
+{
+	struct vnet_port *ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&vp->lock, flags);
+	ret = __tx_port_find(vp, skb);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	return ret;
+}
+
+static int vnet_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+	struct vnet_port *port = tx_port_find(vp, skb);
+	struct vio_dring_state *dr;
+	struct vio_net_desc *d;
+	unsigned long flags;
+	unsigned int len;
+	void *tx_buf;
+	int i, err;
+
+	if (unlikely(!port))
+		goto out_dropped;
+
+	spin_lock_irqsave(&port->vio.lock, flags);
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	if (unlikely(vnet_tx_dring_avail(dr) < 2)) {
+		if (!netif_queue_stopped(dev)) {
+			netif_stop_queue(dev);
+
+			/* This is a hard error, log it. */
+			printk(KERN_ERR PFX "%s: BUG! Tx Ring full when "
+			       "queue awake!\n", dev->name);
+			dev->stats.tx_errors++;
+		}
+		spin_unlock_irqrestore(&port->vio.lock, flags);
+		return NETDEV_TX_BUSY;
+	}
+
+	d = vio_dring_cur(dr);
+
+	tx_buf = port->tx_bufs[dr->prod].buf;
+	skb_copy_from_linear_data(skb, tx_buf + VNET_PACKET_SKIP, skb->len);
+
+	len = skb->len;
+	if (len < ETH_ZLEN) {
+		len = ETH_ZLEN;
+		memset(tx_buf+VNET_PACKET_SKIP+skb->len, 0, len - skb->len);
+	}
+
+	d->hdr.ack = VIO_ACK_ENABLE;
+	d->size = len;
+	d->ncookies = port->tx_bufs[dr->prod].ncookies;
+	for (i = 0; i < d->ncookies; i++)
+		d->cookies[i] = port->tx_bufs[dr->prod].cookies[i];
+
+	/* This has to be a non-SMP write barrier because we are writing
+	 * to memory which is shared with the peer LDOM.
+	 */
+	wmb();
+
+	d->hdr.state = VIO_DESC_READY;
+
+	err = __vnet_tx_trigger(port);
+	if (unlikely(err < 0)) {
+		printk(KERN_INFO PFX "%s: TX trigger error %d\n",
+		       dev->name, err);
+		d->hdr.state = VIO_DESC_FREE;
+		dev->stats.tx_carrier_errors++;
+		goto out_dropped_unlock;
+	}
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += skb->len;
+
+	dr->prod = (dr->prod + 1) & (VNET_TX_RING_SIZE - 1);
+	if (unlikely(vnet_tx_dring_avail(dr) < 2)) {
+		netif_stop_queue(dev);
+		if (vnet_tx_dring_avail(dr) > VNET_TX_WAKEUP_THRESH(dr))
+			netif_wake_queue(dev);
+	}
+
+	spin_unlock_irqrestore(&port->vio.lock, flags);
+
+	dev_kfree_skb(skb);
+
+	dev->trans_start = jiffies;
+	return NETDEV_TX_OK;
+
+out_dropped_unlock:
+	spin_unlock_irqrestore(&port->vio.lock, flags);
+
+out_dropped:
+	dev_kfree_skb(skb);
+	dev->stats.tx_dropped++;
+	return NETDEV_TX_OK;
+}
+
+static void vnet_tx_timeout(struct net_device *dev)
+{
+	/* XXX Implement me XXX */
+}
+
+static int vnet_open(struct net_device *dev)
+{
+	netif_carrier_on(dev);
+	netif_start_queue(dev);
+
+	return 0;
+}
+
+static int vnet_close(struct net_device *dev)
+{
+	netif_stop_queue(dev);
+	netif_carrier_off(dev);
+
+	return 0;
+}
+
+static void vnet_set_rx_mode(struct net_device *dev)
+{
+	/* XXX Implement multicast support XXX */
+}
+
+static int vnet_change_mtu(struct net_device *dev, int new_mtu)
+{
+	if (new_mtu != ETH_DATA_LEN)
+		return -EINVAL;
+
+	dev->mtu = new_mtu;
+	return 0;
+}
+
+static int vnet_set_mac_addr(struct net_device *dev, void *p)
+{
+	return -EINVAL;
+}
+
+static void vnet_get_drvinfo(struct net_device *dev,
+			     struct ethtool_drvinfo *info)
+{
+	strcpy(info->driver, DRV_MODULE_NAME);
+	strcpy(info->version, DRV_MODULE_VERSION);
+}
+
+static u32 vnet_get_msglevel(struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+	return vp->msg_enable;
+}
+
+static void vnet_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct vnet *vp = netdev_priv(dev);
+	vp->msg_enable = value;
+}
+
+static const struct ethtool_ops vnet_ethtool_ops = {
+	.get_drvinfo		= vnet_get_drvinfo,
+	.get_msglevel		= vnet_get_msglevel,
+	.set_msglevel		= vnet_set_msglevel,
+	.get_link		= ethtool_op_get_link,
+	.get_perm_addr		= ethtool_op_get_perm_addr,
+};
+
+static void vnet_port_free_tx_bufs(struct vnet_port *port)
+{
+	struct vio_dring_state *dr;
+	int i;
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	if (dr->base) {
+		ldc_free_exp_dring(port->vio.lp, dr->base,
+				   (dr->entry_size * dr->num_entries),
+				   dr->cookies, dr->ncookies);
+		dr->base = NULL;
+		dr->entry_size = 0;
+		dr->num_entries = 0;
+		dr->pending = 0;
+		dr->ncookies = 0;
+	}
+
+	for (i = 0; i < VNET_TX_RING_SIZE; i++) {
+		void *buf = port->tx_bufs[i].buf;
+
+		if (!buf)
+			continue;
+
+		ldc_unmap(port->vio.lp,
+			  port->tx_bufs[i].cookies,
+			  port->tx_bufs[i].ncookies);
+
+		kfree(buf);
+		port->tx_bufs[i].buf = NULL;
+	}
+}
+
+static int __devinit vnet_port_alloc_tx_bufs(struct vnet_port *port)
+{
+	struct vio_dring_state *dr;
+	unsigned long len;
+	int i, err, ncookies;
+	void *dring;
+
+	for (i = 0; i < VNET_TX_RING_SIZE; i++) {
+		void *buf = kzalloc(ETH_FRAME_LEN + 8, GFP_KERNEL);
+		int map_len = (ETH_FRAME_LEN + 7) & ~7;
+
+		err = -ENOMEM;
+		if (!buf) {
+			printk(KERN_ERR "TX buffer allocation failure\n");
+			goto err_out;
+		}
+		err = -EFAULT;
+		if ((unsigned long)buf & (8UL - 1)) {
+			printk(KERN_ERR "TX buffer misaligned\n");
+			kfree(buf);
+			goto err_out;
+		}
+
+		err = ldc_map_single(port->vio.lp, buf, map_len,
+				     port->tx_bufs[i].cookies, 2,
+				     (LDC_MAP_SHADOW |
+				      LDC_MAP_DIRECT |
+				      LDC_MAP_RW));
+		if (err < 0) {
+			kfree(buf);
+			goto err_out;
+		}
+		port->tx_bufs[i].buf = buf;
+		port->tx_bufs[i].ncookies = err;
+	}
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+
+	len = (VNET_TX_RING_SIZE *
+	       (sizeof(struct vio_net_desc) +
+		(sizeof(struct ldc_trans_cookie) * 2)));
+
+	ncookies = VIO_MAX_RING_COOKIES;
+	dring = ldc_alloc_exp_dring(port->vio.lp, len,
+				    dr->cookies, &ncookies,
+				    (LDC_MAP_SHADOW |
+				     LDC_MAP_DIRECT |
+				     LDC_MAP_RW));
+	if (IS_ERR(dring)) {
+		err = PTR_ERR(dring);
+		goto err_out;
+	}
+
+	dr->base = dring;
+	dr->entry_size = (sizeof(struct vio_net_desc) +
+			  (sizeof(struct ldc_trans_cookie) * 2));
+	dr->num_entries = VNET_TX_RING_SIZE;
+	dr->prod = dr->cons = 0;
+	dr->pending = VNET_TX_RING_SIZE;
+	dr->ncookies = ncookies;
+
+	return 0;
+
+err_out:
+	vnet_port_free_tx_bufs(port);
+
+	return err;
+}
+
+static struct ldc_channel_config vnet_ldc_cfg = {
+	.event		= vnet_event,
+	.mtu		= 64,
+	.mode		= LDC_MODE_UNRELIABLE,
+};
+
+static struct vio_driver_ops vnet_vio_ops = {
+	.send_attr		= vnet_send_attr,
+	.handle_attr		= vnet_handle_attr,
+	.handshake_complete	= vnet_handshake_complete,
+};
+
+const char *remote_macaddr_prop = "remote-mac-address";
+
+static int __devinit vnet_port_probe(struct vio_dev *vdev,
+				     const struct vio_device_id *id)
+{
+	struct mdesc_handle *hp;
+	struct vnet_port *port;
+	unsigned long flags;
+	struct vnet *vp;
+	const u64 *rmac;
+	int len, i, err, switch_port;
+
+	vp = dev_get_drvdata(vdev->dev.parent);
+	if (!vp) {
+		printk(KERN_ERR PFX "Cannot find port parent vnet.\n");
+		return -ENODEV;
+	}
+
+	hp = mdesc_grab();
+
+	rmac = mdesc_get_property(hp, vdev->mp, remote_macaddr_prop, &len);
+	err = -ENODEV;
+	if (!rmac) {
+		printk(KERN_ERR PFX "Port lacks %s property.\n",
+		       remote_macaddr_prop);
+		goto err_out_put_mdesc;
+	}
+
+	port = kzalloc(sizeof(*port), GFP_KERNEL);
+	err = -ENOMEM;
+	if (!port) {
+		printk(KERN_ERR PFX "Cannot allocate vnet_port.\n");
+		goto err_out_put_mdesc;
+	}
+
+	for (i = 0; i < ETH_ALEN; i++)
+		port->raddr[i] = (*rmac >> (5 - i) * 8) & 0xff;
+
+	port->vp = vp;
+
+	err = vio_driver_init(&port->vio, vdev, VDEV_NETWORK,
+			      vnet_versions, ARRAY_SIZE(vnet_versions),
+			      &vnet_vio_ops, vp->dev->name);
+	if (err)
+		goto err_out_free_port;
+
+	err = vio_ldc_alloc(&port->vio, &vnet_ldc_cfg, port);
+	if (err)
+		goto err_out_free_port;
+
+	err = vnet_port_alloc_tx_bufs(port);
+	if (err)
+		goto err_out_free_ldc;
+
+	INIT_HLIST_NODE(&port->hash);
+	INIT_LIST_HEAD(&port->list);
+
+	switch_port = 0;
+	if (mdesc_get_property(hp, vdev->mp, "switch-port", NULL) != NULL)
+		switch_port = 1;
+
+	spin_lock_irqsave(&vp->lock, flags);
+	if (switch_port)
+		list_add(&port->list, &vp->port_list);
+	else
+		list_add_tail(&port->list, &vp->port_list);
+	hlist_add_head(&port->hash, &vp->port_hash[vnet_hashfn(port->raddr)]);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	dev_set_drvdata(&vdev->dev, port);
+
+	printk(KERN_INFO "%s: PORT ( remote-mac ", vp->dev->name);
+	for (i = 0; i < 6; i++)
+		printk("%2.2x%c", port->raddr[i], i == 5 ? ' ' : ':');
+	if (switch_port)
+		printk("switch-port ");
+	printk(")\n");
+
+	vio_port_up(&port->vio);
+
+	mdesc_release(hp);
+
+	return 0;
+
+err_out_free_ldc:
+	vio_ldc_free(&port->vio);
+
+err_out_free_port:
+	kfree(port);
+
+err_out_put_mdesc:
+	mdesc_release(hp);
+	return err;
+}
+
+static int vnet_port_remove(struct vio_dev *vdev)
+{
+	struct vnet_port *port = dev_get_drvdata(&vdev->dev);
+
+	if (port) {
+		struct vnet *vp = port->vp;
+		unsigned long flags;
+
+		del_timer_sync(&port->vio.timer);
+
+		spin_lock_irqsave(&vp->lock, flags);
+		list_del(&port->list);
+		hlist_del(&port->hash);
+		spin_unlock_irqrestore(&vp->lock, flags);
+
+		vnet_port_free_tx_bufs(port);
+		vio_ldc_free(&port->vio);
+
+		dev_set_drvdata(&vdev->dev, NULL);
+
+		kfree(port);
+	}
+	return 0;
+}
+
+static struct vio_device_id vnet_port_match[] = {
+	{
+		.type = "vnet-port",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vnet_match);
+
+static struct vio_driver vnet_port_driver = {
+	.id_table	= vnet_port_match,
+	.probe		= vnet_port_probe,
+	.remove		= vnet_port_remove,
+	.driver		= {
+		.name	= "vnet_port",
+		.owner	= THIS_MODULE,
+	}
+};
+
+const char *local_mac_prop = "local-mac-address";
+
+static int __devinit vnet_probe(struct vio_dev *vdev,
+				const struct vio_device_id *id)
+{
+	static int vnet_version_printed;
+	struct mdesc_handle *hp;
+	struct net_device *dev;
+	struct vnet *vp;
+	const u64 *mac;
+	int err, i, len;
+
+	if (vnet_version_printed++ == 0)
+		printk(KERN_INFO "%s", version);
+
+	hp = mdesc_grab();
+
+	mac = mdesc_get_property(hp, vdev->mp, local_mac_prop, &len);
+	if (!mac) {
+		printk(KERN_ERR PFX "vnet lacks %s property.\n",
+		       local_mac_prop);
+		err = -ENODEV;
+		goto err_out;
+	}
+
+	dev = alloc_etherdev(sizeof(*vp));
+	if (!dev) {
+		printk(KERN_ERR PFX "Etherdev alloc failed, aborting.\n");
+		err = -ENOMEM;
+		goto err_out;
+	}
+
+	for (i = 0; i < ETH_ALEN; i++)
+		dev->dev_addr[i] = (*mac >> (5 - i) * 8) & 0xff;
+
+	memcpy(dev->perm_addr, dev->dev_addr, dev->addr_len);
+
+	SET_NETDEV_DEV(dev, &vdev->dev);
+
+	vp = netdev_priv(dev);
+
+	spin_lock_init(&vp->lock);
+	vp->dev = dev;
+	vp->vdev = vdev;
+
+	INIT_LIST_HEAD(&vp->port_list);
+	for (i = 0; i < VNET_PORT_HASH_SIZE; i++)
+		INIT_HLIST_HEAD(&vp->port_hash[i]);
+
+	dev->open = vnet_open;
+	dev->stop = vnet_close;
+	dev->set_multicast_list = vnet_set_rx_mode;
+	dev->set_mac_address = vnet_set_mac_addr;
+	dev->tx_timeout = vnet_tx_timeout;
+	dev->ethtool_ops = &vnet_ethtool_ops;
+	dev->watchdog_timeo = VNET_TX_TIMEOUT;
+	dev->change_mtu = vnet_change_mtu;
+	dev->hard_start_xmit = vnet_start_xmit;
+
+	err = register_netdev(dev);
+	if (err) {
+		printk(KERN_ERR PFX "Cannot register net device, "
+		       "aborting.\n");
+		goto err_out_free_dev;
+	}
+
+	printk(KERN_INFO "%s: Sun LDOM vnet ", dev->name);
+
+	for (i = 0; i < 6; i++)
+		printk("%2.2x%c", dev->dev_addr[i], i == 5 ? '\n' : ':');
+
+	dev_set_drvdata(&vdev->dev, vp);
+
+	mdesc_release(hp);
+
+	return 0;
+
+err_out_free_dev:
+	free_netdev(dev);
+
+err_out:
+	mdesc_release(hp);
+	return err;
+}
+
+static int vnet_remove(struct vio_dev *vdev)
+{
+
+	struct vnet *vp = dev_get_drvdata(&vdev->dev);
+
+	if (vp) {
+		/* XXX unregister port, or at least check XXX */
+		unregister_netdevice(vp->dev);
+		dev_set_drvdata(&vdev->dev, NULL);
+	}
+	return 0;
+}
+
+static struct vio_device_id vnet_match[] = {
+	{
+		.type = "network",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vnet_match);
+
+static struct vio_driver vnet_driver = {
+	.id_table	= vnet_match,
+	.probe		= vnet_probe,
+	.remove		= vnet_remove,
+	.driver		= {
+		.name	= "vnet",
+		.owner	= THIS_MODULE,
+	}
+};
+
+static int __init vnet_init(void)
+{
+	int err = vio_register_driver(&vnet_driver);
+
+	if (!err) {
+		err = vio_register_driver(&vnet_port_driver);
+		if (err)
+			vio_unregister_driver(&vnet_driver);
+	}
+
+	return err;
+}
+
+static void __exit vnet_exit(void)
+{
+	vio_unregister_driver(&vnet_port_driver);
+	vio_unregister_driver(&vnet_driver);
+}
+
+module_init(vnet_init);
+module_exit(vnet_exit);

drivers/net/sunvnet.h

@@ -0,0 +1,70 @@
+#ifndef _SUNVNET_H
+#define _SUNVNET_H
+
+#define DESC_NCOOKIES(entry_size)	\
+	((entry_size) - sizeof(struct vio_net_desc))
+
+/* length of time before we decide the hardware is borked,
+ * and dev->tx_timeout() should be called to fix the problem
+ */
+#define VNET_TX_TIMEOUT			(5 * HZ)
+
+#define VNET_TX_RING_SIZE		512
+#define VNET_TX_WAKEUP_THRESH(dr)	((dr)->pending / 4)
+
+/* VNET packets are sent in buffers with the first 6 bytes skipped
+ * so that after the ethernet header the IPv4/IPv6 headers are aligned
+ * properly.
+ */
+#define VNET_PACKET_SKIP		6
+
+struct vnet_tx_entry {
+	void			*buf;
+	unsigned int		ncookies;
+	struct ldc_trans_cookie	cookies[2];
+};
+
+struct vnet;
+struct vnet_port {
+	struct vio_driver_state	vio;
+
+	struct hlist_node	hash;
+	u8			raddr[ETH_ALEN];
+
+	struct vnet		*vp;
+
+	struct vnet_tx_entry	tx_bufs[VNET_TX_RING_SIZE];
+
+	struct list_head	list;
+};
+
+static inline struct vnet_port *to_vnet_port(struct vio_driver_state *vio)
+{
+	return container_of(vio, struct vnet_port, vio);
+}
+
+#define VNET_PORT_HASH_SIZE	16
+#define VNET_PORT_HASH_MASK	(VNET_PORT_HASH_SIZE - 1)
+
+static inline unsigned int vnet_hashfn(u8 *mac)
+{
+	unsigned int val = mac[4] ^ mac[5];
+
+	return val & (VNET_PORT_HASH_MASK);
+}
+
+struct vnet {
+	/* Protects port_list and port_hash.  */
+	spinlock_t		lock;
+
+	struct net_device	*dev;
+
+	u32			msg_enable;
+	struct vio_dev		*vdev;
+
+	struct list_head	port_list;
+
+	struct hlist_head	port_hash[VNET_PORT_HASH_SIZE];
+};
+
+#endif /* _SUNVNET_H */

drivers/serial/sunhv.c

@@ -440,8 +440,16 @@ static void sunhv_console_write_paged(struct console *con, const char *s, unsign
 {
 	struct uart_port *port = sunhv_port;
 	unsigned long flags;
+	int locked = 1;
+
+	local_irq_save(flags);
+	if (port->sysrq) {
+		locked = 0;
+	} else if (oops_in_progress) {
+		locked = spin_trylock(&port->lock);
+	} else
+		spin_lock(&port->lock);
 
-	spin_lock_irqsave(&port->lock, flags);
 	while (n > 0) {
 		unsigned long ra = __pa(con_write_page);
 		unsigned long page_bytes;
@@ -469,7 +477,10 @@ static void sunhv_console_write_paged(struct console *con, const char *s, unsign
 			ra += written;
 		}
 	}
-	spin_unlock_irqrestore(&port->lock, flags);
+
+	if (locked)
+		spin_unlock(&port->lock);
+	local_irq_restore(flags);
 }
 
 static inline void sunhv_console_putchar(struct uart_port *port, char c)
@@ -488,7 +499,15 @@ static void sunhv_console_write_bychar(struct console *con, const char *s, unsig
 {
 	struct uart_port *port = sunhv_port;
 	unsigned long flags;
-	int i;
+	int i, locked = 1;
+
+	local_irq_save(flags);
+	if (port->sysrq) {
+		locked = 0;
+	} else if (oops_in_progress) {
+		locked = spin_trylock(&port->lock);
+	} else
+		spin_lock(&port->lock);
 
 	spin_lock_irqsave(&port->lock, flags);
 	for (i = 0; i < n; i++) {
@@ -496,7 +515,10 @@ static void sunhv_console_write_bychar(struct console *con, const char *s, unsig
 			sunhv_console_putchar(port, '\r');
 		sunhv_console_putchar(port, *s++);
 	}
-	spin_unlock_irqrestore(&port->lock, flags);
+
+	if (locked)
+		spin_unlock(&port->lock);
+	local_irq_restore(flags);
 }
 
 static struct console sunhv_console = {

drivers/serial/sunsab.c

@@ -860,22 +860,31 @@ static int num_channels;
 static void sunsab_console_putchar(struct uart_port *port, int c)
 {
 	struct uart_sunsab_port *up = (struct uart_sunsab_port *)port;
-	unsigned long flags;
-
-	spin_lock_irqsave(&up->port.lock, flags);
 
 	sunsab_tec_wait(up);
 	writeb(c, &up->regs->w.tic);
-
-	spin_unlock_irqrestore(&up->port.lock, flags);
 }
 
 static void sunsab_console_write(struct console *con, const char *s, unsigned n)
 {
 	struct uart_sunsab_port *up = &sunsab_ports[con->index];
+	unsigned long flags;
+	int locked = 1;
+
+	local_irq_save(flags);
+	if (up->port.sysrq) {
+		locked = 0;
+	} else if (oops_in_progress) {
+		locked = spin_trylock(&up->port.lock);
+	} else
+		spin_lock(&up->port.lock);
 
 	uart_console_write(&up->port, s, n, sunsab_console_putchar);
 	sunsab_tec_wait(up);
+
+	if (locked)
+		spin_unlock(&up->port.lock);
+	local_irq_restore(flags);
 }
 
 static int sunsab_console_setup(struct console *con, char *options)

drivers/serial/sunsu.c

@@ -1288,7 +1288,17 @@ static void sunsu_console_write(struct console *co, const char *s,
 				unsigned int count)
 {
 	struct uart_sunsu_port *up = &sunsu_ports[co->index];
+	unsigned long flags;
 	unsigned int ier;
+	int locked = 1;
+
+	local_irq_save(flags);
+	if (up->port.sysrq) {
+		locked = 0;
+	} else if (oops_in_progress) {
+		locked = spin_trylock(&up->port.lock);
+	} else
+		spin_lock(&up->port.lock);
 
 	/*
 	 *	First save the UER then disable the interrupts
@@ -1304,6 +1314,10 @@ static void sunsu_console_write(struct console *co, const char *s,
 	 */
 	wait_for_xmitr(up);
 	serial_out(up, UART_IER, ier);
+
+	if (locked)
+		spin_unlock(&up->port.lock);
+	local_irq_restore(flags);
 }
 
 /*

drivers/serial/sunzilog.c

@@ -9,7 +9,7 @@
  * C. Dost, Pete Zaitcev, Ted Ts'o and Alex Buell for their
  * work there.
  *
- *  Copyright (C) 2002, 2006 David S. Miller (davem@davemloft.net)
+ * Copyright (C) 2002, 2006, 2007 David S. Miller (davem@davemloft.net)
  */
 
 #include <linux/module.h>
@@ -1151,11 +1151,22 @@ sunzilog_console_write(struct console *con, const char *s, unsigned int count)
 {
 	struct uart_sunzilog_port *up = &sunzilog_port_table[con->index];
 	unsigned long flags;
+	int locked = 1;
+
+	local_irq_save(flags);
+	if (up->port.sysrq) {
+		locked = 0;
+	} else if (oops_in_progress) {
+		locked = spin_trylock(&up->port.lock);
+	} else
+		spin_lock(&up->port.lock);
 
-	spin_lock_irqsave(&up->port.lock, flags);
 	uart_console_write(&up->port, s, count, sunzilog_putchar);
 	udelay(2);
-	spin_unlock_irqrestore(&up->port.lock, flags);
+
+	if (locked)
+		spin_unlock(&up->port.lock);
+	local_irq_restore(flags);
 }
 
 static int __init sunzilog_console_setup(struct console *con, char *options)

include/asm-sparc64/bugs.h

@@ -4,12 +4,7 @@
  */
 #include <asm/sstate.h>
 
-extern unsigned long loops_per_jiffy;
-
 static void __init check_bugs(void)
 {
-#ifndef CONFIG_SMP
-	cpu_data(0).udelay_val = loops_per_jiffy;
-#endif
 	sstate_running();
 }

include/asm-sparc64/cpudata.h

@@ -19,7 +19,7 @@ typedef struct {
 	unsigned int	__softirq_pending; /* must be 1st, see rtrap.S */
 	unsigned int	__pad0;
 	unsigned long	clock_tick;	/* %tick's per second */
-	unsigned long	udelay_val;
+	unsigned long	__pad;
 	unsigned int	__pad1;
 	unsigned int	__pad2;
 
@@ -80,7 +80,8 @@ struct trap_per_cpu {
 	unsigned int		dev_mondo_qmask;
 	unsigned int		resum_qmask;
 	unsigned int		nonresum_qmask;
-	unsigned int		__pad2[3];
+	unsigned int		__pad2[1];
+	void			*hdesc;
 } __attribute__((aligned(64)));
 extern struct trap_per_cpu trap_block[NR_CPUS];
 extern void init_cur_cpu_trap(struct thread_info *);

include/asm-sparc64/delay.h

@@ -1,37 +1,17 @@
 /* delay.h: Linux delay routines on sparc64.
  *
- * Copyright (C) 1996, 2004 David S. Miller (davem@davemloft.net).
- *
- * Based heavily upon x86 variant which is:
- * Copyright (C) 1993 Linus Torvalds
- *
- * Delay routines calling functions in arch/sparc64/lib/delay.c
+ * Copyright (C) 1996, 2004, 2007 David S. Miller (davem@davemloft.net).
  */
 
-#ifndef __SPARC64_DELAY_H
-#define __SPARC64_DELAY_H
-
-#include <linux/param.h>
-#include <asm/cpudata.h>
+#ifndef _SPARC64_DELAY_H
+#define _SPARC64_DELAY_H
 
 #ifndef __ASSEMBLY__
 
-extern void __bad_udelay(void);
-extern void __bad_ndelay(void);
-
-extern void __udelay(unsigned long usecs);
-extern void __ndelay(unsigned long nsecs);
-extern void __const_udelay(unsigned long usecs);
 extern void __delay(unsigned long loops);
-
-#define udelay(n) (__builtin_constant_p(n) ? \
-	((n) > 20000 ? __bad_udelay() : __const_udelay((n) * 0x10c7ul)) : \
-	__udelay(n))
-	
-#define ndelay(n) (__builtin_constant_p(n) ? \
-	((n) > 20000 ? __bad_ndelay() : __const_udelay((n) * 5ul)) : \
-	__ndelay(n))
+extern void udelay(unsigned long usecs);
+#define mdelay(n)	udelay((n) * 1000)
 
 #endif /* !__ASSEMBLY__ */
 
-#endif /* defined(__SPARC64_DELAY_H) */
+#endif /* _SPARC64_DELAY_H */

include/asm-sparc64/hvtramp.h

@@ -0,0 +1,37 @@
+#ifndef _SPARC64_HVTRAP_H
+#define _SPARC64_HVTRAP_H
+
+#ifndef __ASSEMBLY__
+
+#include <linux/types.h>
+
+struct hvtramp_mapping {
+	__u64		vaddr;
+	__u64		tte;
+};
+
+struct hvtramp_descr {
+	__u32			cpu;
+	__u32			num_mappings;
+	__u64			fault_info_va;
+	__u64			fault_info_pa;
+	__u64			thread_reg;
+	struct hvtramp_mapping	maps[2];
+};
+
+extern void hv_cpu_startup(unsigned long hvdescr_pa);
+
+#endif
+
+#define HVTRAMP_DESCR_CPU		0x00
+#define HVTRAMP_DESCR_NUM_MAPPINGS	0x04
+#define HVTRAMP_DESCR_FAULT_INFO_VA	0x08
+#define HVTRAMP_DESCR_FAULT_INFO_PA	0x10
+#define HVTRAMP_DESCR_THREAD_REG	0x18
+#define HVTRAMP_DESCR_MAPS		0x20
+
+#define HVTRAMP_MAPPING_VADDR		0x00
+#define HVTRAMP_MAPPING_TTE		0x08
+#define HVTRAMP_MAPPING_SIZE		0x10
+
+#endif /* _SPARC64_HVTRAP_H */

include/asm-sparc64/hypervisor.h

@@ -98,7 +98,7 @@
 #define HV_FAST_MACH_EXIT		0x00
 
 #ifndef __ASSEMBLY__
-extern void sun4v_mach_exit(unsigned long exit_core);
+extern void sun4v_mach_exit(unsigned long exit_code);
 #endif
 
 /* Domain services.  */

include/asm-sparc64/irq.h

@@ -53,6 +53,8 @@ extern unsigned int sun4v_build_msi(u32 devhandle, unsigned int *virt_irq_p,
 extern void sun4v_destroy_msi(unsigned int virt_irq);
 extern unsigned int sbus_build_irq(void *sbus, unsigned int ino);
 
+extern void fixup_irqs(void);
+
 static __inline__ void set_softint(unsigned long bits)
 {
 	__asm__ __volatile__("wr	%0, 0x0, %%set_softint"

include/asm-sparc64/ldc.h

@@ -0,0 +1,138 @@
+#ifndef _SPARC64_LDC_H
+#define _SPARC64_LDC_H
+
+#include <asm/hypervisor.h>
+
+extern int ldom_domaining_enabled;
+extern void ldom_set_var(const char *var, const char *value);
+extern void ldom_reboot(const char *boot_command);
+extern void ldom_power_off(void);
+
+/* The event handler will be evoked when link state changes
+ * or data becomes available on the receive side.
+ *
+ * For non-RAW links, if the LDC_EVENT_RESET event arrives the
+ * driver should reset all of it's internal state and reinvoke
+ * ldc_connect() to try and bring the link up again.
+ *
+ * For RAW links, ldc_connect() is not used.  Instead the driver
+ * just waits for the LDC_EVENT_UP event.
+ */
+struct ldc_channel_config {
+	void (*event)(void *arg, int event);
+
+	u32			mtu;
+	unsigned int		rx_irq;
+	unsigned int		tx_irq;
+	u8			mode;
+#define LDC_MODE_RAW		0x00
+#define LDC_MODE_UNRELIABLE	0x01
+#define LDC_MODE_RESERVED	0x02
+#define LDC_MODE_STREAM		0x03
+
+	u8			debug;
+#define LDC_DEBUG_HS		0x01
+#define LDC_DEBUG_STATE		0x02
+#define LDC_DEBUG_RX		0x04
+#define LDC_DEBUG_TX		0x08
+#define LDC_DEBUG_DATA		0x10
+};
+
+#define LDC_EVENT_RESET		0x01
+#define LDC_EVENT_UP		0x02
+#define LDC_EVENT_DATA_READY	0x04
+
+#define LDC_STATE_INVALID	0x00
+#define LDC_STATE_INIT		0x01
+#define LDC_STATE_BOUND		0x02
+#define LDC_STATE_READY		0x03
+#define LDC_STATE_CONNECTED	0x04
+
+struct ldc_channel;
+
+/* Allocate state for a channel.  */
+extern struct ldc_channel *ldc_alloc(unsigned long id,
+				     const struct ldc_channel_config *cfgp,
+				     void *event_arg);
+
+/* Shut down and free state for a channel.  */
+extern void ldc_free(struct ldc_channel *lp);
+
+/* Register TX and RX queues of the link with the hypervisor.  */
+extern int ldc_bind(struct ldc_channel *lp, const char *name);
+
+/* For non-RAW protocols we need to complete a handshake before
+ * communication can proceed.  ldc_connect() does that, if the
+ * handshake completes successfully, an LDC_EVENT_UP event will
+ * be sent up to the driver.
+ */
+extern int ldc_connect(struct ldc_channel *lp);
+extern int ldc_disconnect(struct ldc_channel *lp);
+
+extern int ldc_state(struct ldc_channel *lp);
+
+/* Read and write operations.  Only valid when the link is up.  */
+extern int ldc_write(struct ldc_channel *lp, const void *buf,
+		     unsigned int size);
+extern int ldc_read(struct ldc_channel *lp, void *buf, unsigned int size);
+
+#define LDC_MAP_SHADOW	0x01
+#define LDC_MAP_DIRECT	0x02
+#define LDC_MAP_IO	0x04
+#define LDC_MAP_R	0x08
+#define LDC_MAP_W	0x10
+#define LDC_MAP_X	0x20
+#define LDC_MAP_RW	(LDC_MAP_R | LDC_MAP_W)
+#define LDC_MAP_RWX	(LDC_MAP_R | LDC_MAP_W | LDC_MAP_X)
+#define LDC_MAP_ALL	0x03f
+
+struct ldc_trans_cookie {
+	u64			cookie_addr;
+	u64			cookie_size;
+};
+
+struct scatterlist;
+extern int ldc_map_sg(struct ldc_channel *lp,
+		      struct scatterlist *sg, int num_sg,
+		      struct ldc_trans_cookie *cookies, int ncookies,
+		      unsigned int map_perm);
+
+extern int ldc_map_single(struct ldc_channel *lp,
+			  void *buf, unsigned int len,
+			  struct ldc_trans_cookie *cookies, int ncookies,
+			  unsigned int map_perm);
+
+extern void ldc_unmap(struct ldc_channel *lp, struct ldc_trans_cookie *cookies,
+		      int ncookies);
+
+extern int ldc_copy(struct ldc_channel *lp, int copy_dir,
+		    void *buf, unsigned int len, unsigned long offset,
+		    struct ldc_trans_cookie *cookies, int ncookies);
+
+static inline int ldc_get_dring_entry(struct ldc_channel *lp,
+				      void *buf, unsigned int len,
+				      unsigned long offset,
+				      struct ldc_trans_cookie *cookies,
+				      int ncookies)
+{
+	return ldc_copy(lp, LDC_COPY_IN, buf, len, offset, cookies, ncookies);
+}
+
+static inline int ldc_put_dring_entry(struct ldc_channel *lp,
+				      void *buf, unsigned int len,
+				      unsigned long offset,
+				      struct ldc_trans_cookie *cookies,
+				      int ncookies)
+{
+	return ldc_copy(lp, LDC_COPY_OUT, buf, len, offset, cookies, ncookies);
+}
+
+extern void *ldc_alloc_exp_dring(struct ldc_channel *lp, unsigned int len,
+				 struct ldc_trans_cookie *cookies,
+				 int *ncookies, unsigned int map_perm);
+
+extern void ldc_free_exp_dring(struct ldc_channel *lp, void *buf,
+			       unsigned int len,
+			       struct ldc_trans_cookie *cookies, int ncookies);
+
+#endif /* _SPARC64_LDC_H */

include/asm-sparc64/mdesc.h

@@ -2,38 +2,66 @@
 #define _SPARC64_MDESC_H
 
 #include <linux/types.h>
+#include <linux/cpumask.h>
 #include <asm/prom.h>
 
-struct mdesc_node;
-struct mdesc_arc {
-	const char		*name;
-	struct mdesc_node	*arc;
-};
-
-struct mdesc_node {
-	const char		*name;
-	u64			node;
-	unsigned int		unique_id;
-	unsigned int		num_arcs;
-	unsigned int		irqs[2];
-	struct property		*properties;
-	struct mdesc_node	*hash_next;
-	struct mdesc_node	*allnodes_next;
-	struct mdesc_arc	arcs[0];
-};
-
-extern struct mdesc_node *md_find_node_by_name(struct mdesc_node *from,
-					       const char *name);
-#define md_for_each_node_by_name(__mn, __name) \
-	for (__mn = md_find_node_by_name(NULL, __name); __mn; \
-	     __mn = md_find_node_by_name(__mn, __name))
-
-extern struct property *md_find_property(const struct mdesc_node *mp,
-					 const char *name,
-					 int *lenp);
-extern const void *md_get_property(const struct mdesc_node *mp,
-				   const char *name,
-				   int *lenp);
+struct mdesc_handle;
+
+/* Machine description operations are to be surrounded by grab and
+ * release calls.  The mdesc_handle returned from the grab is
+ * the first argument to all of the operational calls that work
+ * on mdescs.
+ */
+extern struct mdesc_handle *mdesc_grab(void);
+extern void mdesc_release(struct mdesc_handle *);
+
+#define MDESC_NODE_NULL		(~(u64)0)
+
+extern u64 mdesc_node_by_name(struct mdesc_handle *handle,
+			      u64 from_node, const char *name);
+#define mdesc_for_each_node_by_name(__hdl, __node, __name) \
+	for (__node = mdesc_node_by_name(__hdl, MDESC_NODE_NULL, __name); \
+	     (__node) != MDESC_NODE_NULL; \
+	     __node = mdesc_node_by_name(__hdl, __node, __name))
+
+/* Access to property values returned from mdesc_get_property() are
+ * only valid inside of a mdesc_grab()/mdesc_release() sequence.
+ * Once mdesc_release() is called, the memory backed up by these
+ * pointers may reference freed up memory.
+ *
+ * Therefore callers must make copies of any property values
+ * they need.
+ *
+ * These same rules apply to mdesc_node_name().
+ */
+extern const void *mdesc_get_property(struct mdesc_handle *handle,
+				      u64 node, const char *name, int *lenp);
+extern const char *mdesc_node_name(struct mdesc_handle *hp, u64 node);
+
+/* MD arc iteration, the standard sequence is:
+ *
+ *	unsigned long arc;
+ *	mdesc_for_each_arc(arc, handle, node, MDESC_ARC_TYPE_{FWD,BACK}) {
+ *		unsigned long target = mdesc_arc_target(handle, arc);
+ *		...
+ *	}
+ */
+
+#define MDESC_ARC_TYPE_FWD	"fwd"
+#define MDESC_ARC_TYPE_BACK	"back"
+
+extern u64 mdesc_next_arc(struct mdesc_handle *handle, u64 from,
+			  const char *arc_type);
+#define mdesc_for_each_arc(__arc, __hdl, __node, __type) \
+	for (__arc = mdesc_next_arc(__hdl, __node, __type); \
+	     (__arc) != MDESC_NODE_NULL; \
+	     __arc = mdesc_next_arc(__hdl, __arc, __type))
+
+extern u64 mdesc_arc_target(struct mdesc_handle *hp, u64 arc);
+
+extern void mdesc_update(void);
+
+extern void mdesc_fill_in_cpu_data(cpumask_t mask);
 
 extern void sun4v_mdesc_init(void);
 

include/asm-sparc64/mmu_context.h

@@ -76,6 +76,9 @@ static inline void switch_mm(struct mm_struct *old_mm, struct mm_struct *mm, str
 	unsigned long ctx_valid, flags;
 	int cpu;
 
+	if (unlikely(mm == &init_mm))
+		return;
+
 	spin_lock_irqsave(&mm->context.lock, flags);
 	ctx_valid = CTX_VALID(mm->context);
 	if (!ctx_valid)

include/asm-sparc64/power.h

@@ -0,0 +1,7 @@
+#ifndef _SPARC64_POWER_H
+#define _SPARC64_POWER_H
+
+extern void wake_up_powerd(void);
+extern int start_powerd(void);
+
+#endif /* !(_SPARC64_POWER_H) */

include/asm-sparc64/smp.h

@@ -29,9 +29,6 @@
 #include <asm/bitops.h>
 #include <asm/atomic.h>
 
-extern cpumask_t phys_cpu_present_map;
-#define cpu_possible_map phys_cpu_present_map
-
 extern cpumask_t cpu_sibling_map[NR_CPUS];
 extern cpumask_t cpu_core_map[NR_CPUS];
 extern int sparc64_multi_core;
@@ -44,7 +41,12 @@ extern int hard_smp_processor_id(void);
 #define raw_smp_processor_id() (current_thread_info()->cpu)
 
 extern void smp_fill_in_sib_core_maps(void);
-extern unsigned char boot_cpu_id;
+extern void cpu_play_dead(void);
+
+#ifdef CONFIG_HOTPLUG_CPU
+extern int __cpu_disable(void);
+extern void __cpu_die(unsigned int cpu);
+#endif
 
 #endif /* !(__ASSEMBLY__) */
 
@@ -52,7 +54,6 @@ extern unsigned char boot_cpu_id;
 
 #define hard_smp_processor_id()		0
 #define smp_fill_in_sib_core_maps() do { } while (0)
-#define boot_cpu_id	(0)
 
 #endif /* !(CONFIG_SMP) */
 

include/asm-sparc64/vio.h

@@ -0,0 +1,404 @@
+#ifndef _SPARC64_VIO_H
+#define _SPARC64_VIO_H
+
+#include <linux/kernel.h>
+#include <linux/device.h>
+#include <linux/mod_devicetable.h>
+#include <linux/timer.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/list.h>
+
+#include <asm/ldc.h>
+#include <asm/mdesc.h>
+
+struct vio_msg_tag {
+	u8			type;
+#define VIO_TYPE_CTRL		0x01
+#define VIO_TYPE_DATA		0x02
+#define VIO_TYPE_ERR		0x04
+
+	u8			stype;
+#define VIO_SUBTYPE_INFO	0x01
+#define VIO_SUBTYPE_ACK		0x02
+#define VIO_SUBTYPE_NACK	0x04
+
+	u16			stype_env;
+#define VIO_VER_INFO		0x0001
+#define VIO_ATTR_INFO		0x0002
+#define VIO_DRING_REG		0x0003
+#define VIO_DRING_UNREG		0x0004
+#define VIO_RDX			0x0005
+#define VIO_PKT_DATA		0x0040
+#define VIO_DESC_DATA		0x0041
+#define VIO_DRING_DATA		0x0042
+#define VNET_MCAST_INFO		0x0101
+
+	u32		sid;
+};
+
+struct vio_rdx {
+	struct vio_msg_tag	tag;
+	u64			resv[6];
+};
+
+struct vio_ver_info {
+	struct vio_msg_tag	tag;
+	u16			major;
+	u16			minor;
+	u8			dev_class;
+#define VDEV_NETWORK		0x01
+#define VDEV_NETWORK_SWITCH	0x02
+#define VDEV_DISK		0x03
+#define VDEV_DISK_SERVER	0x04
+
+	u8			resv1[3];
+	u64			resv2[5];
+};
+
+struct vio_dring_register {
+	struct vio_msg_tag	tag;
+	u64			dring_ident;
+	u32			num_descr;
+	u32			descr_size;
+	u16			options;
+#define VIO_TX_DRING		0x0001
+#define VIO_RX_DRING		0x0002
+	u16			resv;
+	u32			num_cookies;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+struct vio_dring_unregister {
+	struct vio_msg_tag	tag;
+	u64			dring_ident;
+	u64			resv[5];
+};
+
+/* Data transfer modes */
+#define VIO_PKT_MODE		0x01 /* Packet based transfer	*/
+#define VIO_DESC_MODE		0x02 /* In-band descriptors	*/
+#define VIO_DRING_MODE		0x03 /* Descriptor rings	*/
+
+struct vio_dring_data {
+	struct vio_msg_tag	tag;
+	u64			seq;
+	u64			dring_ident;
+	u32			start_idx;
+	u32			end_idx;
+	u8			state;
+#define VIO_DRING_ACTIVE	0x01
+#define VIO_DRING_STOPPED	0x02
+
+	u8			__pad1;
+	u16			__pad2;
+	u32			__pad3;
+	u64			__par4[2];
+};
+
+struct vio_dring_hdr {
+	u8			state;
+#define VIO_DESC_FREE		0x01
+#define VIO_DESC_READY		0x02
+#define VIO_DESC_ACCEPTED	0x03
+#define VIO_DESC_DONE		0x04
+	u8			ack;
+#define VIO_ACK_ENABLE		0x01
+#define VIO_ACK_DISABLE		0x00
+
+	u16			__pad1;
+	u32			__pad2;
+};
+
+/* VIO disk specific structures and defines */
+struct vio_disk_attr_info {
+	struct vio_msg_tag	tag;
+	u8			xfer_mode;
+	u8			vdisk_type;
+#define VD_DISK_TYPE_SLICE	0x01 /* Slice in block device	*/
+#define VD_DISK_TYPE_DISK	0x02 /* Entire block device	*/
+	u16			resv1;
+	u32			vdisk_block_size;
+	u64			operations;
+	u64			vdisk_size;
+	u64			max_xfer_size;
+	u64			resv2[2];
+};
+
+struct vio_disk_desc {
+	struct vio_dring_hdr	hdr;
+	u64			req_id;
+	u8			operation;
+#define VD_OP_BREAD		0x01 /* Block read			*/
+#define VD_OP_BWRITE		0x02 /* Block write			*/
+#define VD_OP_FLUSH		0x03 /* Flush disk contents		*/
+#define VD_OP_GET_WCE		0x04 /* Get write-cache status		*/
+#define VD_OP_SET_WCE		0x05 /* Enable/disable write-cache	*/
+#define VD_OP_GET_VTOC		0x06 /* Get VTOC			*/
+#define VD_OP_SET_VTOC		0x07 /* Set VTOC			*/
+#define VD_OP_GET_DISKGEOM	0x08 /* Get disk geometry		*/
+#define VD_OP_SET_DISKGEOM	0x09 /* Set disk geometry		*/
+#define VD_OP_SCSICMD		0x0a /* SCSI control command		*/
+#define VD_OP_GET_DEVID		0x0b /* Get device ID			*/
+#define VD_OP_GET_EFI		0x0c /* Get EFI				*/
+#define VD_OP_SET_EFI		0x0d /* Set EFI				*/
+	u8			slice;
+	u16			resv1;
+	u32			status;
+	u64			offset;
+	u64			size;
+	u32			ncookies;
+	u32			resv2;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+#define VIO_DISK_VNAME_LEN	8
+#define VIO_DISK_ALABEL_LEN	128
+#define VIO_DISK_NUM_PART	8
+
+struct vio_disk_vtoc {
+	u8			volume_name[VIO_DISK_VNAME_LEN];
+	u16			sector_size;
+	u16			num_partitions;
+	u8			ascii_label[VIO_DISK_ALABEL_LEN];
+	struct {
+		u16		id;
+		u16		perm_flags;
+		u32		resv;
+		u64		start_block;
+		u64		num_blocks;
+	} partitions[VIO_DISK_NUM_PART];
+};
+
+struct vio_disk_geom {
+	u16			num_cyl; /* Num data cylinders		*/
+	u16			alt_cyl; /* Num alternate cylinders	*/
+	u16			beg_cyl; /* Cyl off of fixed head area	*/
+	u16			num_hd;  /* Num heads			*/
+	u16			num_sec; /* Num sectors			*/
+	u16			ifact;   /* Interleave factor		*/
+	u16			apc;     /* Alts per cylinder (SCSI)	*/
+	u16			rpm;	 /* Revolutions per minute	*/
+	u16			phy_cyl; /* Num physical cylinders	*/
+	u16			wr_skip; /* Num sects to skip, writes	*/
+	u16			rd_skip; /* Num sects to skip, writes	*/
+};
+
+struct vio_disk_devid {
+	u16			resv;
+	u16			type;
+	u32			len;
+	char			id[0];
+};
+
+struct vio_disk_efi {
+	u64			lba;
+	u64			len;
+	char			data[0];
+};
+
+/* VIO net specific structures and defines */
+struct vio_net_attr_info {
+	struct vio_msg_tag	tag;
+	u8			xfer_mode;
+	u8			addr_type;
+#define VNET_ADDR_ETHERMAC	0x01
+	u16			ack_freq;
+	u32			resv1;
+	u64			addr;
+	u64			mtu;
+	u64			resv2[3];
+};
+
+#define VNET_NUM_MCAST		7
+
+struct vio_net_mcast_info {
+	struct vio_msg_tag	tag;
+	u8			set;
+	u8			count;
+	u8			mcast_addr[VNET_NUM_MCAST * 6];
+	u32			resv;
+};
+
+struct vio_net_desc {
+	struct vio_dring_hdr	hdr;
+	u32			size;
+	u32			ncookies;
+	struct ldc_trans_cookie	cookies[0];
+};
+
+#define VIO_MAX_RING_COOKIES	24
+
+struct vio_dring_state {
+	u64			ident;
+	void			*base;
+	u64			snd_nxt;
+	u64			rcv_nxt;
+	u32			entry_size;
+	u32			num_entries;
+	u32			prod;
+	u32			cons;
+	u32			pending;
+	int			ncookies;
+	struct ldc_trans_cookie	cookies[VIO_MAX_RING_COOKIES];
+};
+
+static inline void *vio_dring_cur(struct vio_dring_state *dr)
+{
+	return dr->base + (dr->entry_size * dr->prod);
+}
+
+static inline void *vio_dring_entry(struct vio_dring_state *dr,
+				    unsigned int index)
+{
+	return dr->base + (dr->entry_size * index);
+}
+
+static inline u32 vio_dring_avail(struct vio_dring_state *dr,
+				  unsigned int ring_size)
+{
+	/* Ensure build-time power-of-2.  */
+	BUILD_BUG_ON(ring_size & (ring_size - 1));
+
+	return (dr->pending -
+		((dr->prod - dr->cons) & (ring_size - 1)));
+}
+
+#define VIO_MAX_TYPE_LEN	64
+#define VIO_MAX_COMPAT_LEN	64
+
+struct vio_dev {
+	u64			mp;
+	struct device_node	*dp;
+
+	char			type[VIO_MAX_TYPE_LEN];
+	char			compat[VIO_MAX_COMPAT_LEN];
+	int			compat_len;
+
+	unsigned long		channel_id;
+
+	unsigned int		tx_irq;
+	unsigned int		rx_irq;
+
+	struct device		dev;
+};
+
+struct vio_driver {
+	struct list_head		node;
+	const struct vio_device_id	*id_table;
+	int (*probe)(struct vio_dev *dev, const struct vio_device_id *id);
+	int (*remove)(struct vio_dev *dev);
+	void (*shutdown)(struct vio_dev *dev);
+	unsigned long			driver_data;
+	struct device_driver		driver;
+};
+
+struct vio_version {
+	u16		major;
+	u16		minor;
+};
+
+struct vio_driver_state;
+struct vio_driver_ops {
+	int	(*send_attr)(struct vio_driver_state *vio);
+	int	(*handle_attr)(struct vio_driver_state *vio, void *pkt);
+	void	(*handshake_complete)(struct vio_driver_state *vio);
+};
+
+struct vio_completion {
+	struct completion	com;
+	int			err;
+	int			waiting_for;
+};
+
+struct vio_driver_state {
+	/* Protects VIO handshake and, optionally, driver private state.  */
+	spinlock_t		lock;
+
+	struct ldc_channel	*lp;
+
+	u32			_peer_sid;
+	u32			_local_sid;
+	struct vio_dring_state	drings[2];
+#define VIO_DRIVER_TX_RING	0
+#define VIO_DRIVER_RX_RING	1
+
+	u8			hs_state;
+#define VIO_HS_INVALID		0x00
+#define VIO_HS_GOTVERS		0x01
+#define VIO_HS_GOT_ATTR		0x04
+#define VIO_HS_SENT_DREG	0x08
+#define VIO_HS_SENT_RDX		0x10
+#define VIO_HS_GOT_RDX_ACK	0x20
+#define VIO_HS_GOT_RDX		0x40
+#define VIO_HS_SENT_RDX_ACK	0x80
+#define VIO_HS_COMPLETE		(VIO_HS_GOT_RDX_ACK | VIO_HS_SENT_RDX_ACK)
+
+	u8			dev_class;
+
+	u8			dr_state;
+#define VIO_DR_STATE_TXREG	0x01
+#define VIO_DR_STATE_RXREG	0x02
+#define VIO_DR_STATE_TXREQ	0x10
+#define VIO_DR_STATE_RXREQ	0x20
+
+	u8			debug;
+#define VIO_DEBUG_HS		0x01
+#define VIO_DEBUG_DATA		0x02
+
+	void			*desc_buf;
+	unsigned int		desc_buf_len;
+
+	struct vio_completion	*cmp;
+
+	struct vio_dev		*vdev;
+
+	struct timer_list	timer;
+
+	struct vio_version	ver;
+
+	struct vio_version	*ver_table;
+	int			ver_table_entries;
+
+	char			*name;
+
+	struct vio_driver_ops	*ops;
+};
+
+#define viodbg(TYPE, f, a...) \
+do {	if (vio->debug & VIO_DEBUG_##TYPE) \
+		printk(KERN_INFO "vio: ID[%lu] " f, \
+		       vio->vdev->channel_id, ## a); \
+} while (0)
+
+extern int vio_register_driver(struct vio_driver *drv);
+extern void vio_unregister_driver(struct vio_driver *drv);
+
+static inline struct vio_driver *to_vio_driver(struct device_driver *drv)
+{
+	return container_of(drv, struct vio_driver, driver);
+}
+
+static inline struct vio_dev *to_vio_dev(struct device *dev)
+{
+	return container_of(dev, struct vio_dev, dev);
+}
+
+extern int vio_ldc_send(struct vio_driver_state *vio, void *data, int len);
+extern void vio_link_state_change(struct vio_driver_state *vio, int event);
+extern void vio_conn_reset(struct vio_driver_state *vio);
+extern int vio_control_pkt_engine(struct vio_driver_state *vio, void *pkt);
+extern int vio_validate_sid(struct vio_driver_state *vio,
+			    struct vio_msg_tag *tp);
+extern u32 vio_send_sid(struct vio_driver_state *vio);
+extern int vio_ldc_alloc(struct vio_driver_state *vio,
+			 struct ldc_channel_config *base_cfg, void *event_arg);
+extern void vio_ldc_free(struct vio_driver_state *vio);
+extern int vio_driver_init(struct vio_driver_state *vio, struct vio_dev *vdev,
+			   u8 dev_class, struct vio_version *ver_table,
+			   int ver_table_size, struct vio_driver_ops *ops,
+			   char *name);
+
+extern void vio_port_up(struct vio_driver_state *vio);
+
+#endif /* _SPARC64_VIO_H */