diff --git a/arch/sparc64/kernel/process.c b/arch/sparc64/kernel/process.c
index 0a0c05fc3a33..2084f81a76e1 100644
--- a/arch/sparc64/kernel/process.c
+++ b/arch/sparc64/kernel/process.c
@@ -657,20 +657,39 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 		struct task_struct *p, struct pt_regs *regs)
 {
 	struct thread_info *t = task_thread_info(p);
+	struct sparc_stackf *parent_sf;
+	unsigned long child_stack_sz;
 	char *child_trap_frame;
+	int kernel_thread;
+
+	kernel_thread = (regs->tstate & TSTATE_PRIV) ? 1 : 0;
+	parent_sf = ((struct sparc_stackf *) regs) - 1;
 
 	/* Calculate offset to stack_frame & pt_regs */
-	child_trap_frame = task_stack_page(p) + (THREAD_SIZE - (TRACEREG_SZ+STACKFRAME_SZ));
-	memcpy(child_trap_frame, (((struct sparc_stackf *)regs)-1), (TRACEREG_SZ+STACKFRAME_SZ));
+	child_stack_sz = ((STACKFRAME_SZ + TRACEREG_SZ) +
+			  (kernel_thread ? STACKFRAME_SZ : 0));
+	child_trap_frame = (task_stack_page(p) +
+			    (THREAD_SIZE - child_stack_sz));
+	memcpy(child_trap_frame, parent_sf, child_stack_sz);
 
-	t->flags = (t->flags & ~((0xffUL << TI_FLAG_CWP_SHIFT) | (0xffUL << TI_FLAG_CURRENT_DS_SHIFT))) |
+	t->flags = (t->flags & ~((0xffUL << TI_FLAG_CWP_SHIFT) |
+				 (0xffUL << TI_FLAG_CURRENT_DS_SHIFT))) |
 		(((regs->tstate + 1) & TSTATE_CWP) << TI_FLAG_CWP_SHIFT);
 	t->new_child = 1;
 	t->ksp = ((unsigned long) child_trap_frame) - STACK_BIAS;
-	t->kregs = (struct pt_regs *)(child_trap_frame+sizeof(struct sparc_stackf));
+	t->kregs = (struct pt_regs *) (child_trap_frame +
+				       sizeof(struct sparc_stackf));
 	t->fpsaved[0] = 0;
 
-	if (regs->tstate & TSTATE_PRIV) {
+	if (kernel_thread) {
+		struct sparc_stackf *child_sf = (struct sparc_stackf *)
+			(child_trap_frame + (STACKFRAME_SZ + TRACEREG_SZ));
+
+		/* Zero terminate the stack backtrace.  */
+		child_sf->fp = NULL;
+		t->kregs->u_regs[UREG_FP] =
+		  ((unsigned long) child_sf) - STACK_BIAS;
+
 		/* Special case, if we are spawning a kernel thread from
 		 * a userspace task (via KMOD, NFS, or similar) we must
 		 * disable performance counters in the child because the
@@ -681,12 +700,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 			t->pcr_reg = 0;
 			t->flags &= ~_TIF_PERFCTR;
 		}
-		t->kregs->u_regs[UREG_FP] = t->ksp;
 		t->flags |= ((long)ASI_P << TI_FLAG_CURRENT_DS_SHIFT);
-		flush_register_windows();
-		memcpy((void *)(t->ksp + STACK_BIAS),
-		       (void *)(regs->u_regs[UREG_FP] + STACK_BIAS),
-		       sizeof(struct sparc_stackf));
 		t->kregs->u_regs[UREG_G6] = (unsigned long) t;
 		t->kregs->u_regs[UREG_G4] = (unsigned long) t->task;
 	} else {
diff --git a/arch/sparc64/kernel/rtrap.S b/arch/sparc64/kernel/rtrap.S
index 3afacbb5781d..c6fc695fe1fe 100644
--- a/arch/sparc64/kernel/rtrap.S
+++ b/arch/sparc64/kernel/rtrap.S
@@ -363,6 +363,7 @@ kern_rtt:	rdpr			%canrestore, %g1
 		brz,pn			%g1, kern_rtt_fill
 		 nop
 kern_rtt_restore:
+		stw			%g0, [%sp + PTREGS_OFF + PT_V9_MAGIC]
 		restore
 		retry
 
diff --git a/arch/sparc64/kernel/stacktrace.c b/arch/sparc64/kernel/stacktrace.c
index 01b52f561af4..c73ce3f4197e 100644
--- a/arch/sparc64/kernel/stacktrace.c
+++ b/arch/sparc64/kernel/stacktrace.c
@@ -19,7 +19,7 @@ void save_stack_trace(struct stack_trace *trace)
 	fp = ksp + STACK_BIAS;
 	thread_base = (unsigned long) tp;
 	do {
-		struct reg_window *rw;
+		struct sparc_stackf *sf;
 		struct pt_regs *regs;
 		unsigned long pc;
 
@@ -28,15 +28,17 @@ void save_stack_trace(struct stack_trace *trace)
 		    fp >= (thread_base + THREAD_SIZE))
 			break;
 
-		rw = (struct reg_window *) fp;
-		regs = (struct pt_regs *) (rw + 1);
+		sf = (struct sparc_stackf *) fp;
+		regs = (struct pt_regs *) (sf + 1);
 
 		if ((regs->magic & ~0x1ff) == PT_REGS_MAGIC) {
+			if (!(regs->tstate & TSTATE_PRIV))
+				break;
 			pc = regs->tpc;
 			fp = regs->u_regs[UREG_I6] + STACK_BIAS;
 		} else {
-			pc = rw->ins[7];
-			fp = rw->ins[6] + STACK_BIAS;
+			pc = sf->callers_pc;
+			fp = (unsigned long)sf->fp + STACK_BIAS;
 		}
 
 		if (trace->skip > 0)
diff --git a/arch/sparc64/kernel/traps.c b/arch/sparc64/kernel/traps.c
index d9b8d46707d1..369749262653 100644
--- a/arch/sparc64/kernel/traps.c
+++ b/arch/sparc64/kernel/traps.c
@@ -2116,7 +2116,7 @@ void show_stack(struct task_struct *tsk, unsigned long *_ksp)
 	printk("\n");
 #endif
 	do {
-		struct reg_window *rw;
+		struct sparc_stackf *sf;
 		struct pt_regs *regs;
 		unsigned long pc;
 
@@ -2124,15 +2124,17 @@ void show_stack(struct task_struct *tsk, unsigned long *_ksp)
 		if (fp < (thread_base + sizeof(struct thread_info)) ||
 		    fp >= (thread_base + THREAD_SIZE))
 			break;
-		rw = (struct reg_window *)fp;
-		regs = (struct pt_regs *) (rw + 1);
+		sf = (struct sparc_stackf *) fp;
+		regs = (struct pt_regs *) (sf + 1);
 
 		if ((regs->magic & ~0x1ff) == PT_REGS_MAGIC) {
+			if (!(regs->tstate & TSTATE_PRIV))
+				break;
 			pc = regs->tpc;
 			fp = regs->u_regs[UREG_I6] + STACK_BIAS;
 		} else {
-			pc = rw->ins[7];
-			fp = rw->ins[6] + STACK_BIAS;
+			pc = sf->callers_pc;
+			fp = (unsigned long)sf->fp + STACK_BIAS;
 		}
 
 		printk(" [%016lx] ", pc);
diff --git a/drivers/serial/sunhv.c b/drivers/serial/sunhv.c
index 145c0281495d..2847336742d7 100644
--- a/drivers/serial/sunhv.c
+++ b/drivers/serial/sunhv.c
@@ -499,7 +499,6 @@ static void sunhv_console_write_bychar(struct console *con, const char *s, unsig
 	} else
 		spin_lock(&port->lock);
 
-	spin_lock_irqsave(&port->lock, flags);
 	for (i = 0; i < n; i++) {
 		if (*s == '\n')
 			sunhv_console_putchar(port, '\r');