9 files changed, 341 insertions, 197 deletions

diff --git a/arch/sparc64/kernel/ds.c b/arch/sparc64/kernel/ds.c
index 1c587107cef..fa1f04d756a 100644
--- a/arch/sparc64/kernel/ds.c
+++ b/arch/sparc64/kernel/ds.c
@@ -228,7 +228,7 @@ static struct ds_cap_state *find_cap_by_string(const char *name)
 	return NULL;
 }
 
-static int ds_send(struct ldc_channel *lp, void *data, int len)
+static int __ds_send(struct ldc_channel *lp, void *data, int len)
 {
 	int err, limit = 1000;
 
@@ -243,6 +243,18 @@ static int ds_send(struct ldc_channel *lp, void *data, int len)
 	return err;
 }
 
+static int ds_send(struct ldc_channel *lp, void *data, int len)
+{
+	unsigned long flags;
+	int err;
+
+	spin_lock_irqsave(&ds_lock, flags);
+	err = __ds_send(lp, data, len);
+	spin_unlock_irqrestore(&ds_lock, flags);
+
+	return err;
+}
+
 struct ds_md_update_req {
 	__u64				req_num;
 };
@@ -267,6 +279,8 @@ static void md_update_data(struct ldc_channel *lp,
 
 	printk(KERN_INFO PFX "Machine description update.\n");
 
+	mdesc_update();
+
 	memset(&pkt, 0, sizeof(pkt));
 	pkt.data.tag.type = DS_DATA;
 	pkt.data.tag.len = sizeof(pkt) - sizeof(struct ds_msg_tag);
@@ -275,8 +289,6 @@ static void md_update_data(struct ldc_channel *lp,
 	pkt.res.result = DS_OK;
 
 	ds_send(lp, &pkt, sizeof(pkt));
-
-	mdesc_update();
 }
 
 struct ds_shutdown_req {
@@ -391,18 +403,6 @@ struct dr_cpu_resp_entry {
 	__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);
@@ -425,7 +425,7 @@ static void __dr_cpu_send_error(struct ds_cap_state *cp, struct ds_data *data)
 
 	pkt.data.tag.len = msg_len - sizeof(struct ds_msg_tag);
 
-	ds_send(dp->lp, &pkt, msg_len);
+	__ds_send(dp->lp, &pkt, msg_len);
 }
 
 static void dr_cpu_send_error(struct ds_cap_state *cp, struct ds_data *data)
@@ -555,7 +555,7 @@ static int dr_cpu_configure(struct ds_cap_state *cp, u64 req_num,
 	}
 
 	spin_lock_irqsave(&ds_lock, flags);
-	ds_send(ds_info->lp, resp, resp_len);
+	__ds_send(ds_info->lp, resp, resp_len);
 	spin_unlock_irqrestore(&ds_lock, flags);
 
 	kfree(resp);
@@ -596,7 +596,7 @@ static int dr_cpu_unconfigure(struct ds_cap_state *cp, u64 req_num,
 	}
 
 	spin_lock_irqsave(&ds_lock, flags);
-	ds_send(ds_info->lp, resp, resp_len);
+	__ds_send(ds_info->lp, resp, resp_len);
 	spin_unlock_irqrestore(&ds_lock, flags);
 
 	kfree(resp);
@@ -604,107 +604,49 @@ static int dr_cpu_unconfigure(struct ds_cap_state *cp, u64 req_num,
 	return 0;
 }
 
-static void process_dr_cpu_list(struct ds_cap_state *cp)
+static void dr_cpu_data(struct ldc_channel *lp,
+			struct ds_cap_state *cp,
+			void *buf, int len)
 {
-	struct dr_cpu_queue_entry *qp, *tmp;
-	unsigned long flags;
-	LIST_HEAD(todo);
+	struct ds_data *data = buf;
+	struct dr_cpu_tag *tag = (struct dr_cpu_tag *) (data + 1);
+	u32 *cpu_list = (u32 *) (tag + 1);
+	u64 req_num = tag->req_num;
 	cpumask_t mask;
+	unsigned int i;
+	int err;
 
-	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);
+	switch (tag->type) {
+	case DR_CPU_CONFIGURE:
+	case DR_CPU_UNCONFIGURE:
+	case DR_CPU_FORCE_UNCONFIGURE:
+		break;
 
-next:
-		list_del(&qp->list);
-		kfree(qp);
+	default:
+		dr_cpu_send_error(cp, data);
+		return;
 	}
-}
 
-static int dr_cpu_thread(void *__unused)
-{
-	struct ds_cap_state *cp;
-	DEFINE_WAIT(wait);
+	purge_dups(cpu_list, tag->num_records);
 
-	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;
+	cpus_clear(mask);
+	for (i = 0; i < tag->num_records; i++) {
+		if (cpu_list[i] == CPU_SENTINEL)
+			continue;
 
-		process_dr_cpu_list(cp);
+		if (cpu_list[i] < NR_CPUS)
+			cpu_set(cpu_list[i], mask);
 	}
 
-	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);
+	if (tag->type == DR_CPU_CONFIGURE)
+		err = dr_cpu_configure(cp, req_num, &mask);
+	else
+		err = dr_cpu_unconfigure(cp, req_num, &mask);
 
-	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);
-	}
+	if (err)
+		dr_cpu_send_error(cp, data);
 }
-#endif
+#endif /* CONFIG_HOTPLUG_CPU */
 
 struct ds_pri_msg {
 	__u64				req_num;
@@ -820,7 +762,7 @@ void ldom_set_var(const char *var, const char *value)
 		ds_var_doorbell = 0;
 		ds_var_response = -1;
 
-		ds_send(dp->lp, &pkt, msg_len);
+		__ds_send(dp->lp, &pkt, msg_len);
 		spin_unlock_irqrestore(&ds_lock, flags);
 
 		loops = 1000;
@@ -904,7 +846,7 @@ static int register_services(struct ds_info *dp)
 		pbuf.req.minor = 0;
 		strcpy(pbuf.req.svc_id, cp->service_id);
 
-		err = ds_send(lp, &pbuf, msg_len);
+		err = __ds_send(lp, &pbuf, msg_len);
 		if (err > 0)
 			cp->state = CAP_STATE_REG_SENT;
 	}
@@ -960,27 +902,97 @@ conn_reset:
 	return -ECONNRESET;
 }
 
+static void __send_ds_nack(struct ds_info *dp, u64 handle)
+{
+	struct ds_data_nack nack = {
+		.tag = {
+			.type = DS_NACK,
+			.len = (sizeof(struct ds_data_nack) -
+				sizeof(struct ds_msg_tag)),
+		},
+		.handle = handle,
+		.result = DS_INV_HDL,
+	};
+
+	__ds_send(dp->lp, &nack, sizeof(nack));
+}
+
+static LIST_HEAD(ds_work_list);
+static DECLARE_WAIT_QUEUE_HEAD(ds_wait);
+
+struct ds_queue_entry {
+	struct list_head		list;
+	int				req_len;
+	int				__pad;
+	u64				req[0];
+};
+
+static void process_ds_work(void)
+{
+	struct ds_queue_entry *qp, *tmp;
+	static struct ds_info *dp;
+	unsigned long flags;
+	LIST_HEAD(todo);
+
+	spin_lock_irqsave(&ds_lock, flags);
+	list_splice(&ds_work_list, &todo);
+	INIT_LIST_HEAD(&ds_work_list);
+	spin_unlock_irqrestore(&ds_lock, flags);
+
+	dp = ds_info;
+
+	list_for_each_entry_safe(qp, tmp, &todo, list) {
+		struct ds_data *dpkt = (struct ds_data *) qp->req;
+		struct ds_cap_state *cp = find_cap(dpkt->handle);
+		int req_len = qp->req_len;
+
+		if (!cp) {
+			printk(KERN_ERR PFX "Data for unknown handle %lu\n",
+			       dpkt->handle);
+
+			spin_lock_irqsave(&ds_lock, flags);
+			__send_ds_nack(dp, dpkt->handle);
+			spin_unlock_irqrestore(&ds_lock, flags);
+		} else {
+			cp->data(dp->lp, cp, dpkt, req_len);
+		}
+
+		list_del(&qp->list);
+		kfree(qp);
+	}
+}
+
+static int ds_thread(void *__unused)
+{
+	DEFINE_WAIT(wait);
+
+	while (1) {
+		prepare_to_wait(&ds_wait, &wait, TASK_INTERRUPTIBLE);
+		if (list_empty(&ds_work_list))
+			schedule();
+		finish_wait(&ds_wait, &wait);
+
+		if (kthread_should_stop())
+			break;
+
+		process_ds_work();
+	}
+
+	return 0;
+}
+
 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));
+	struct ds_queue_entry *qp;
+
+	qp = kmalloc(sizeof(struct ds_queue_entry) + len, GFP_ATOMIC);
+	if (!qp) {
+		__send_ds_nack(dp, dpkt->handle);
 	} else {
-		cp->data(dp->lp, cp, dpkt, len);
+		memcpy(&qp->req, pkt, len);
+		list_add_tail(&qp->list, &ds_work_list);
+		wake_up(&ds_wait);
 	}
 	return 0;
 }
@@ -996,11 +1008,24 @@ static void ds_up(struct ds_info *dp)
 	req.ver.major = 1;
 	req.ver.minor = 0;
 
-	err = ds_send(lp, &req, sizeof(req));
+	err = __ds_send(lp, &req, sizeof(req));
 	if (err > 0)
 		dp->hs_state = DS_HS_START;
 }
 
+static void ds_reset(struct ds_info *dp)
+{
+	int i;
+
+	dp->hs_state = 0;
+
+	for (i = 0; i < ARRAY_SIZE(ds_states); i++) {
+		struct ds_cap_state *cp = &ds_states[i];
+
+		cp->state = CAP_STATE_UNKNOWN;
+	}
+}
+
 static void ds_event(void *arg, int event)
 {
 	struct ds_info *dp = arg;
@@ -1016,6 +1041,12 @@ static void ds_event(void *arg, int event)
 		return;
 	}
 
+	if (event == LDC_EVENT_RESET) {
+		ds_reset(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);
@@ -1148,9 +1179,7 @@ static int __init ds_init(void)
 	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
+	kthread_run(ds_thread, NULL, "kldomd");
 
 	return vio_register_driver(&ds_driver);
 }
diff --git a/arch/sparc64/kernel/hvtramp.S b/arch/sparc64/kernel/hvtramp.S
index 76a090e2c2a..a55c252e18c 100644
--- a/arch/sparc64/kernel/hvtramp.S
+++ b/arch/sparc64/kernel/hvtramp.S
@@ -10,6 +10,7 @@
 #include <asm/hvtramp.h>
 #include <asm/pstate.h>
 #include <asm/ptrace.h>
+#include <asm/head.h>
 #include <asm/asi.h>
 
 	.text
@@ -28,7 +29,7 @@
 	 * First setup basic privileged cpu state.
 	 */
 hv_cpu_startup:
-	wrpr		%g0, 0, %gl
+	SET_GL(0)
 	wrpr		%g0, 15, %pil
 	wrpr		%g0, 0, %canrestore
 	wrpr		%g0, 0, %otherwin
diff --git a/arch/sparc64/kernel/mdesc.c b/arch/sparc64/kernel/mdesc.c
index 62a38979394..302ba5e5a0b 100644
--- a/arch/sparc64/kernel/mdesc.c
+++ b/arch/sparc64/kernel/mdesc.c
@@ -137,7 +137,7 @@ static struct mdesc_handle *mdesc_kmalloc(unsigned int mdesc_size)
 		       sizeof(struct mdesc_hdr) +
 		       mdesc_size);
 
-	base = kmalloc(handle_size + 15, GFP_KERNEL);
+	base = kmalloc(handle_size + 15, GFP_KERNEL | __GFP_NOFAIL);
 	if (base) {
 		struct mdesc_handle *hp;
 		unsigned long addr;
@@ -214,18 +214,83 @@ void mdesc_release(struct mdesc_handle *hp)
 }
 EXPORT_SYMBOL(mdesc_release);
 
-static void do_mdesc_update(struct work_struct *work)
+static DEFINE_MUTEX(mdesc_mutex);
+static struct mdesc_notifier_client *client_list;
+
+void mdesc_register_notifier(struct mdesc_notifier_client *client)
+{
+	u64 node;
+
+	mutex_lock(&mdesc_mutex);
+	client->next = client_list;
+	client_list = client;
+
+	mdesc_for_each_node_by_name(cur_mdesc, node, client->node_name)
+		client->add(cur_mdesc, node);
+
+	mutex_unlock(&mdesc_mutex);
+}
+
+/* Run 'func' on nodes which are in A but not in B.  */
+static void invoke_on_missing(const char *name,
+			      struct mdesc_handle *a,
+			      struct mdesc_handle *b,
+			      void (*func)(struct mdesc_handle *, u64))
+{
+	u64 node;
+
+	mdesc_for_each_node_by_name(a, node, name) {
+		const u64 *id = mdesc_get_property(a, node, "id", NULL);
+		int found = 0;
+		u64 fnode;
+
+		mdesc_for_each_node_by_name(b, fnode, name) {
+			const u64 *fid = mdesc_get_property(b, fnode,
+							    "id", NULL);
+
+			if (*id == *fid) {
+				found = 1;
+				break;
+			}
+		}
+		if (!found)
+			func(a, node);
+	}
+}
+
+static void notify_one(struct mdesc_notifier_client *p,
+		       struct mdesc_handle *old_hp,
+		       struct mdesc_handle *new_hp)
+{
+	invoke_on_missing(p->node_name, old_hp, new_hp, p->remove);
+	invoke_on_missing(p->node_name, new_hp, old_hp, p->add);
+}
+
+static void mdesc_notify_clients(struct mdesc_handle *old_hp,
+				 struct mdesc_handle *new_hp)
+{
+	struct mdesc_notifier_client *p = client_list;
+
+	while (p) {
+		notify_one(p, old_hp, new_hp);
+		p = p->next;
+	}
+}
+
+void mdesc_update(void)
 {
 	unsigned long len, real_len, status;
 	struct mdesc_handle *hp, *orig_hp;
 	unsigned long flags;
 
+	mutex_lock(&mdesc_mutex);
+
 	(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;
+		goto out;
 	}
 
 	status = sun4v_mach_desc(__pa(&hp->mdesc), len, &real_len);
@@ -234,25 +299,25 @@ static void do_mdesc_update(struct work_struct *work)
 		       status);
 		atomic_dec(&hp->refcnt);
 		mdesc_free(hp);
-		return;
+		goto out;
 	}
 
 	spin_lock_irqsave(&mdesc_lock, flags);
 	orig_hp = cur_mdesc;
 	cur_mdesc = hp;
+	spin_unlock_irqrestore(&mdesc_lock, flags);
 
+	mdesc_notify_clients(orig_hp, hp);
+
+	spin_lock_irqsave(&mdesc_lock, flags);
 	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 DECLARE_WORK(mdesc_update_work, do_mdesc_update);
 
-void mdesc_update(void)
-{
-	schedule_work(&mdesc_update_work);
+out:
+	mutex_unlock(&mdesc_mutex);
 }
 
 static struct mdesc_elem *node_block(struct mdesc_hdr *mdesc)
@@ -278,13 +343,14 @@ u64 mdesc_node_by_name(struct mdesc_handle *hp,
 	u64 last_node = hp->mdesc.node_sz / 16;
 	u64 ret;
 
-	if (from_node == MDESC_NODE_NULL)
-		from_node = 0;
-
-	if (from_node >= last_node)
+	if (from_node == MDESC_NODE_NULL) {
+		ret = from_node = 0;
+	} else if (from_node >= last_node) {
 		return MDESC_NODE_NULL;
+	} else {
+		ret = ep[from_node].d.val;
+	}
 
-	ret = ep[from_node].d.val;
 	while (ret < last_node) {
 		if (ep[ret].tag != MD_NODE)
 			return MDESC_NODE_NULL;
diff --git a/arch/sparc64/kernel/setup.c b/arch/sparc64/kernel/setup.c
index dc928e49e34..aafde3dd9fd 100644
--- a/arch/sparc64/kernel/setup.c
+++ b/arch/sparc64/kernel/setup.c
@@ -454,9 +454,9 @@ 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).clock_tick
+		   , cpu_data(0).clock_tick
 #endif
 		);
 #ifdef CONFIG_SMP
diff --git a/arch/sparc64/kernel/signal.c b/arch/sparc64/kernel/signal.c
index 203e8730100..fb13775b368 100644
--- a/arch/sparc64/kernel/signal.c
+++ b/arch/sparc64/kernel/signal.c
@@ -289,9 +289,7 @@ void do_rt_sigreturn(struct pt_regs *regs)
 	struct rt_signal_frame __user *sf;
 	unsigned long tpc, tnpc, tstate;
 	__siginfo_fpu_t __user *fpu_save;
-	mm_segment_t old_fs;
 	sigset_t set;
-	stack_t st;
 	int err;
 
 	/* Always make any pending restarted system calls return -EINTR */
@@ -327,20 +325,13 @@ void do_rt_sigreturn(struct pt_regs *regs)
 		err |= restore_fpu_state(regs, &sf->fpu_state);
 
 	err |= __copy_from_user(&set, &sf->mask, sizeof(sigset_t));
-	err |= __copy_from_user(&st, &sf->stack, sizeof(stack_t));
-	
+	err |= do_sigaltstack(&sf->stack, NULL, (unsigned long)sf);
+
 	if (err)
 		goto segv;
-		
+
 	regs->tpc = tpc;
 	regs->tnpc = tnpc;
-	
-	/* It is more difficult to avoid calling this function than to
-	   call it and ignore errors.  */
-	old_fs = get_fs();
-	set_fs(KERNEL_DS);
-	do_sigaltstack((const stack_t __user *) &st, NULL, (unsigned long)sf);
-	set_fs(old_fs);
 
 	sigdelsetmask(&set, ~_BLOCKABLE);
 	spin_lock_irq(&current->sighand->siglock);
diff --git a/arch/sparc64/kernel/traps.c b/arch/sparc64/kernel/traps.c
index 00a9e3286c8..6ef2d299fb1 100644
--- a/arch/sparc64/kernel/traps.c
+++ b/arch/sparc64/kernel/traps.c
@@ -2225,6 +2225,7 @@ void die_if_kernel(char *str, struct pt_regs *regs)
 	notify_die(DIE_OOPS, str, regs, 0, 255, SIGSEGV);
 	__asm__ __volatile__("flushw");
 	__show_regs(regs);
+	add_taint(TAINT_DIE);
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *rw = (struct reg_window *)
 			(regs->u_regs[UREG_FP] + STACK_BIAS);
diff --git a/arch/sparc64/kernel/vio.c b/arch/sparc64/kernel/vio.c
index 49569b44ea1..8d3cc4fdb55 100644
--- a/arch/sparc64/kernel/vio.c
+++ b/arch/sparc64/kernel/vio.c
@@ -201,10 +201,11 @@ static void vio_fill_channel_info(struct mdesc_handle *hp, u64 mp,
 static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
 				      struct device *parent)
 {
-	const char *type, *compat;
+	const char *type, *compat, *bus_id_name;
 	struct device_node *dp;
 	struct vio_dev *vdev;
 	int err, tlen, clen;
+	const u64 *id;
 
 	type = mdesc_get_property(hp, mp, "device-type", &tlen);
 	if (!type) {
@@ -220,6 +221,16 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
 		return NULL;
 	}
 
+	bus_id_name = type;
+	if (!strcmp(type, "domain-services-port"))
+		bus_id_name = "ds";
+
+	if (strlen(bus_id_name) >= KOBJ_NAME_LEN - 4) {
+		printk(KERN_ERR "VIO: bus_id_name [%s] is too long.\n",
+		       bus_id_name);
+		return NULL;
+	}
+
 	compat = mdesc_get_property(hp, mp, "device-type", &clen);
 	if (!compat) {
 		clen = 0;
@@ -249,7 +260,14 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
 
 	vio_fill_channel_info(hp, mp, vdev);
 
-	snprintf(vdev->dev.bus_id, BUS_ID_SIZE, "%lx", mp);
+	id = mdesc_get_property(hp, mp, "id", NULL);
+	if (!id)
+		snprintf(vdev->dev.bus_id, BUS_ID_SIZE, "%s",
+			 bus_id_name);
+	else
+		snprintf(vdev->dev.bus_id, BUS_ID_SIZE, "%s-%lu",
+			 bus_id_name, *id);
+
 	vdev->dev.parent = parent;
 	vdev->dev.bus = &vio_bus_type;
 	vdev->dev.release = vio_dev_release;
@@ -269,6 +287,8 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
 	}
 	vdev->dp = dp;
 
+	printk(KERN_ERR "VIO: Adding device %s\n", vdev->dev.bus_id);
+
 	err = device_register(&vdev->dev);
 	if (err) {
 		printk(KERN_ERR "VIO: Could not register device %s, err=%d\n",
@@ -283,46 +303,46 @@ static struct vio_dev *vio_create_one(struct mdesc_handle *hp, u64 mp,
 	return vdev;
 }
 
-static void walk_tree(struct mdesc_handle *hp, u64 n, struct vio_dev *parent)
+static void vio_add(struct mdesc_handle *hp, u64 node)
 {
-	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);
-	}
+	(void) vio_create_one(hp, node, &root_vdev->dev);
 }
 
-static void create_devices(struct mdesc_handle *hp, u64 root)
+static int vio_md_node_match(struct device *dev, void *arg)
 {
-	u64 mp;
+	struct vio_dev *vdev = to_vio_dev(dev);
 
-	root_vdev = vio_create_one(hp, root, NULL);
-	if (!root_vdev) {
-		printk(KERN_ERR "VIO: Coult not create root device.\n");
-		return;
-	}
+	if (vdev->mp == (u64) arg)
+		return 1;
 
-	walk_tree(hp, root, root_vdev);
+	return 0;
+}
+
+static void vio_remove(struct mdesc_handle *hp, u64 node)
+{
+	struct device *dev;
 
-	/* 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);
+	dev = device_find_child(&root_vdev->dev, (void *) node,
+				vio_md_node_match);
+	if (dev) {
+		printk(KERN_INFO "VIO: Removing device %s\n", dev->bus_id);
 
-		if (parent)
-			walk_tree(hp, mp, parent);
+		device_unregister(dev);
 	}
 }
 
+static struct mdesc_notifier_client vio_device_notifier = {
+	.add		= vio_add,
+	.remove		= vio_remove,
+	.node_name	= "virtual-device-port",
+};
+
+static struct mdesc_notifier_client vio_ds_notifier = {
+	.add		= vio_add,
+	.remove		= vio_remove,
+	.node_name	= "domain-services-port",
+};
+
 const char *channel_devices_node = "channel-devices";
 const char *channel_devices_compat = "SUNW,sun4v-channel-devices";
 const char *cfg_handle_prop = "cfg-handle";
@@ -381,11 +401,19 @@ static int __init vio_init(void)
 
 	cdev_cfg_handle = *cfg_handle;
 
-	create_devices(hp, root);
+	root_vdev = vio_create_one(hp, root, NULL);
+	err = -ENODEV;
+	if (!root_vdev) {
+		printk(KERN_ERR "VIO: Coult not create root device.\n");
+		goto out_release;
+	}
+
+	mdesc_register_notifier(&vio_device_notifier);
+	mdesc_register_notifier(&vio_ds_notifier);
 
 	mdesc_release(hp);
 
-	return 0;
+	return err;
 
 out_release:
 	mdesc_release(hp);
diff --git a/arch/sparc64/kernel/viohs.c b/arch/sparc64/kernel/viohs.c
index 15613add45d..09126fc338b 100644
--- a/arch/sparc64/kernel/viohs.c
+++ b/arch/sparc64/kernel/viohs.c
@@ -78,6 +78,24 @@ static int start_handshake(struct vio_driver_state *vio)
 	return 0;
 }
 
+static void flush_rx_dring(struct vio_driver_state *vio)
+{
+	struct vio_dring_state *dr;
+	u64 ident;
+
+	BUG_ON(!(vio->dr_state & VIO_DR_STATE_RXREG));
+
+	dr = &vio->drings[VIO_DRIVER_RX_RING];
+	ident = dr->ident;
+
+	BUG_ON(!vio->desc_buf);
+	kfree(vio->desc_buf);
+	vio->desc_buf = NULL;
+
+	memset(dr, 0, sizeof(*dr));
+	dr->ident = ident;
+}
+
 void vio_link_state_change(struct vio_driver_state *vio, int event)
 {
 	if (event == LDC_EVENT_UP) {
@@ -98,6 +116,16 @@ void vio_link_state_change(struct vio_driver_state *vio, int event)
 			break;
 		}
 		start_handshake(vio);
+	} else if (event == LDC_EVENT_RESET) {
+		vio->hs_state = VIO_HS_INVALID;
+
+		if (vio->dr_state & VIO_DR_STATE_RXREG)
+			flush_rx_dring(vio);
+
+		vio->dr_state = 0x00;
+		memset(&vio->ver, 0, sizeof(vio->ver));
+
+		ldc_disconnect(vio->lp);
 	}
 }
 EXPORT_SYMBOL(vio_link_state_change);
@@ -396,6 +424,8 @@ static int process_dreg_info(struct vio_driver_state *vio,
 	if (vio->dr_state & VIO_DR_STATE_RXREG)
 		goto send_nack;
 
+	BUG_ON(vio->desc_buf);
+
 	vio->desc_buf = kzalloc(pkt->descr_size, GFP_ATOMIC);
 	if (!vio->desc_buf)
 		goto send_nack;
diff --git a/arch/sparc64/kernel/vmlinux.lds.S b/arch/sparc64/kernel/vmlinux.lds.S
index 3ad10f3027e..481861764de 100644
--- a/arch/sparc64/kernel/vmlinux.lds.S
+++ b/arch/sparc64/kernel/vmlinux.lds.S
@@ -90,10 +90,8 @@ SECTIONS
   __initramfs_end = .;
 #endif
 
-  . = ALIGN(PAGE_SIZE);
-  __per_cpu_start = .;
-  .data.percpu  : { *(.data.percpu) }
-  __per_cpu_end = .;
+  PERCPU(PAGE_SIZE)
+
   . = ALIGN(PAGE_SIZE);
   __init_end = .;
   __bss_start = .;