Merge commit '517ffce4e1a03aea979fe3a18a3dd1761a24fafb' into arch-sparc

Backmerge from the point in mainline where a trivial conflict had been
introduced (arch/sparc/kernel/sys_sparc_64.c had grown sys_kern_features()
right after where kernel_execve() used to be)

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

18 files changed, 317 insertions, 90 deletions

diff --git a/arch/sparc/include/asm/compat.h b/arch/sparc/include/asm/compat.h
index cef99fbc0a2..830502fe62b 100644
--- a/arch/sparc/include/asm/compat.h
+++ b/arch/sparc/include/asm/compat.h
@@ -232,9 +232,10 @@ static inline void __user *arch_compat_alloc_user_space(long len)
 	struct pt_regs *regs = current_thread_info()->kregs;
 	unsigned long usp = regs->u_regs[UREG_I6];
 
-	if (!(test_thread_flag(TIF_32BIT)))
+	if (test_thread_64bit_stack(usp))
 		usp += STACK_BIAS;
-	else
+
+	if (test_thread_flag(TIF_32BIT))
 		usp &= 0xffffffffUL;
 
 	usp -= len;
diff --git a/arch/sparc/include/asm/ptrace.h b/arch/sparc/include/asm/ptrace.h
index 1e8b8180226..bdfafd7af46 100644
--- a/arch/sparc/include/asm/ptrace.h
+++ b/arch/sparc/include/asm/ptrace.h
@@ -45,7 +45,18 @@ struct global_reg_snapshot {
 	struct thread_info	*thread;
 	unsigned long		pad1;
 };
-extern struct global_reg_snapshot global_reg_snapshot[NR_CPUS];
+
+struct global_pmu_snapshot {
+	unsigned long		pcr[4];
+	unsigned long		pic[4];
+};
+
+union global_cpu_snapshot {
+	struct global_reg_snapshot	reg;
+	struct global_pmu_snapshot	pmu;
+};
+
+extern union global_cpu_snapshot global_cpu_snapshot[NR_CPUS];
 
 #define force_successful_syscall_return() set_thread_noerror(1)
 #define user_mode(regs) (!((regs)->tstate & TSTATE_PRIV))
diff --git a/arch/sparc/include/asm/smp_64.h b/arch/sparc/include/asm/smp_64.h
index 29862a9e906..dd3bef4b989 100644
--- a/arch/sparc/include/asm/smp_64.h
+++ b/arch/sparc/include/asm/smp_64.h
@@ -48,6 +48,7 @@ extern void smp_fill_in_sib_core_maps(void);
 extern void cpu_play_dead(void);
 
 extern void smp_fetch_global_regs(void);
+extern void smp_fetch_global_pmu(void);
 
 struct seq_file;
 void smp_bogo(struct seq_file *);
@@ -65,6 +66,7 @@ extern void __cpu_die(unsigned int cpu);
 #define hard_smp_processor_id()		0
 #define smp_fill_in_sib_core_maps() do { } while (0)
 #define smp_fetch_global_regs() do { } while (0)
+#define smp_fetch_global_pmu() do { } while (0)
 
 #endif /* !(CONFIG_SMP) */
 
diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index ea2ba22d595..269bd92313d 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -258,6 +258,11 @@ static inline bool test_and_clear_restore_sigmask(void)
 
 #define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
 
+#define thread32_stack_is_64bit(__SP) (((__SP) & 0x1) != 0)
+#define test_thread_64bit_stack(__SP) \
+	((test_thread_flag(TIF_32BIT) && !thread32_stack_is_64bit(__SP)) ? \
+	 false : true)
+
 #endif	/* !__ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
diff --git a/arch/sparc/include/asm/ttable.h b/arch/sparc/include/asm/ttable.h
index 48f2807d326..71b5a67522a 100644
--- a/arch/sparc/include/asm/ttable.h
+++ b/arch/sparc/include/asm/ttable.h
@@ -372,7 +372,9 @@ etrap_spill_fixup_64bit:				\
 
 /* Normal 32bit spill */
 #define SPILL_2_GENERIC(ASI)				\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, (. - (128 + 4));			\
+	 srl	%sp, 0, %sp;				\
 	stwa	%l0, [%sp + %g0] ASI;			\
 	mov	0x04, %g3;				\
 	stwa	%l1, [%sp + %g3] ASI;			\
@@ -398,14 +400,16 @@ etrap_spill_fixup_64bit:				\
 	stwa	%i6, [%g1 + %g0] ASI;			\
 	stwa	%i7, [%g1 + %g3] ASI;			\
 	saved;						\
-        retry; nop; nop;				\
+        retry;						\
 	b,a,pt	%xcc, spill_fixup_dax;			\
 	b,a,pt	%xcc, spill_fixup_mna;			\
 	b,a,pt	%xcc, spill_fixup;
 
 #define SPILL_2_GENERIC_ETRAP		\
 etrap_user_spill_32bit:			\
-	srl	%sp, 0, %sp;		\
+	and	%sp, 1, %g3;		\
+	brnz,pn	%g3, etrap_user_spill_64bit;	\
+	 srl	%sp, 0, %sp;		\
 	stwa	%l0, [%sp + 0x00] %asi;	\
 	stwa	%l1, [%sp + 0x04] %asi;	\
 	stwa	%l2, [%sp + 0x08] %asi;	\
@@ -427,7 +431,7 @@ etrap_user_spill_32bit:			\
 	ba,pt	%xcc, etrap_save;	\
 	 wrpr	%g1, %cwp;		\
 	nop; nop; nop; nop;		\
-	nop; nop; nop; nop;		\
+	nop; nop;			\
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit;
@@ -592,7 +596,9 @@ user_rtt_fill_64bit:					\
 
 /* Normal 32bit fill */
 #define FILL_2_GENERIC(ASI)				\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, (. - (128 + 4));			\
+	 srl	%sp, 0, %sp;				\
 	lduwa	[%sp + %g0] ASI, %l0;			\
 	mov	0x04, %g2;				\
 	mov	0x08, %g3;				\
@@ -616,14 +622,16 @@ user_rtt_fill_64bit:					\
 	lduwa	[%g1 + %g3] ASI, %i6;			\
 	lduwa	[%g1 + %g5] ASI, %i7;			\
 	restored;					\
-	retry; nop; nop; nop; nop;			\
+	retry; nop; nop;				\
 	b,a,pt	%xcc, fill_fixup_dax;			\
 	b,a,pt	%xcc, fill_fixup_mna;			\
 	b,a,pt	%xcc, fill_fixup;
 
 #define FILL_2_GENERIC_RTRAP				\
 user_rtt_fill_32bit:					\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, user_rtt_fill_64bit;		\
+	 srl	%sp, 0, %sp;				\
 	lduwa	[%sp + 0x00] %asi, %l0;			\
 	lduwa	[%sp + 0x04] %asi, %l1;			\
 	lduwa	[%sp + 0x08] %asi, %l2;			\
@@ -643,7 +651,7 @@ user_rtt_fill_32bit:					\
 	ba,pt	%xcc, user_rtt_pre_restore;		\
 	 restored;					\
 	nop; nop; nop; nop; nop;			\
-	nop; nop; nop; nop; nop;			\
+	nop; nop; nop;					\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;
diff --git a/arch/sparc/include/uapi/asm/sigcontext.h b/arch/sparc/include/uapi/asm/sigcontext.h
index e69de29bb2d..ae5704fa77a 100644
--- a/arch/sparc/include/uapi/asm/sigcontext.h
+++ b/arch/sparc/include/uapi/asm/sigcontext.h
@@ -0,0 +1,4 @@
+/*
+ * There isn't anything here anymore, but the file must not be empty or patch
+ * will delete it.
+ */
diff --git a/arch/sparc/include/uapi/asm/unistd.h b/arch/sparc/include/uapi/asm/unistd.h
index 8974ef7ae92..bed86a820d0 100644
--- a/arch/sparc/include/uapi/asm/unistd.h
+++ b/arch/sparc/include/uapi/asm/unistd.h
@@ -405,8 +405,12 @@
 #define __NR_setns		337
 #define __NR_process_vm_readv	338
 #define __NR_process_vm_writev	339
+#define __NR_kern_features	340
 
-#define NR_syscalls		340
+#define NR_syscalls		341
+
+/* Bitmask values returned from kern_features system call.  */
+#define KERN_FEATURE_MIXED_MODE_STACK	0x00000001
 
 #ifdef __32bit_syscall_numbers__
 /* Sparc 32-bit only has the "setresuid32", "getresuid32" variants,
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c
index e48651dace1..b5c38faa4ea 100644
--- a/arch/sparc/kernel/perf_event.c
+++ b/arch/sparc/kernel/perf_event.c
@@ -817,15 +817,17 @@ static u64 nop_for_index(int idx)
 
 static inline void sparc_pmu_enable_event(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc, int idx)
 {
-	u64 val, mask = mask_for_index(idx);
+	u64 enc, val, mask = mask_for_index(idx);
 	int pcr_index = 0;
 
 	if (sparc_pmu->num_pcrs > 1)
 		pcr_index = idx;
 
+	enc = perf_event_get_enc(cpuc->events[idx]);
+
 	val = cpuc->pcr[pcr_index];
 	val &= ~mask;
-	val |= hwc->config;
+	val |= event_encoding(enc, idx);
 	cpuc->pcr[pcr_index] = val;
 
 	pcr_ops->write_pcr(pcr_index, cpuc->pcr[pcr_index]);
@@ -1738,8 +1740,6 @@ static void perf_callchain_user_64(struct perf_callchain_entry *entry,
 {
 	unsigned long ufp;
 
-	perf_callchain_store(entry, regs->tpc);
-
 	ufp = regs->u_regs[UREG_I6] + STACK_BIAS;
 	do {
 		struct sparc_stackf *usf, sf;
@@ -1760,19 +1760,27 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 {
 	unsigned long ufp;
 
-	perf_callchain_store(entry, regs->tpc);
-
 	ufp = regs->u_regs[UREG_I6] & 0xffffffffUL;
 	do {
-		struct sparc_stackf32 *usf, sf;
 		unsigned long pc;
 
-		usf = (struct sparc_stackf32 *) ufp;
-		if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
-			break;
+		if (thread32_stack_is_64bit(ufp)) {
+			struct sparc_stackf *usf, sf;
 
-		pc = sf.callers_pc;
-		ufp = (unsigned long)sf.fp;
+			ufp += STACK_BIAS;
+			usf = (struct sparc_stackf *) ufp;
+			if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+				break;
+			pc = sf.callers_pc & 0xffffffff;
+			ufp = ((unsigned long) sf.fp) & 0xffffffff;
+		} else {
+			struct sparc_stackf32 *usf, sf;
+			usf = (struct sparc_stackf32 *) ufp;
+			if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+				break;
+			pc = sf.callers_pc;
+			ufp = (unsigned long)sf.fp;
+		}
 		perf_callchain_store(entry, pc);
 	} while (entry->nr < PERF_MAX_STACK_DEPTH);
 }
@@ -1780,6 +1788,11 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 void
 perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 {
+	perf_callchain_store(entry, regs->tpc);
+
+	if (!current->mm)
+		return;
+
 	flushw_user();
 	if (test_thread_flag(TIF_32BIT))
 		perf_callchain_user_32(entry, regs);
diff --git a/arch/sparc/kernel/process_64.c b/arch/sparc/kernel/process_64.c
index 6b36e879b2a..dff54f46728 100644
--- a/arch/sparc/kernel/process_64.c
+++ b/arch/sparc/kernel/process_64.c
@@ -27,6 +27,7 @@
 #include <linux/tick.h>
 #include <linux/init.h>
 #include <linux/cpu.h>
+#include <linux/perf_event.h>
 #include <linux/elfcore.h>
 #include <linux/sysrq.h>
 #include <linux/nmi.h>
@@ -47,6 +48,7 @@
 #include <asm/syscalls.h>
 #include <asm/irq_regs.h>
 #include <asm/smp.h>
+#include <asm/pcr.h>
 
 #include "kstack.h"
 
@@ -204,18 +206,22 @@ void show_regs(struct pt_regs *regs)
 	show_stack(current, (unsigned long *) regs->u_regs[UREG_FP]);
 }
 
-struct global_reg_snapshot global_reg_snapshot[NR_CPUS];
-static DEFINE_SPINLOCK(global_reg_snapshot_lock);
+union global_cpu_snapshot global_cpu_snapshot[NR_CPUS];
+static DEFINE_SPINLOCK(global_cpu_snapshot_lock);
 
 static void __global_reg_self(struct thread_info *tp, struct pt_regs *regs,
 			      int this_cpu)
 {
+	struct global_reg_snapshot *rp;
+
 	flushw_all();
 
-	global_reg_snapshot[this_cpu].tstate = regs->tstate;
-	global_reg_snapshot[this_cpu].tpc = regs->tpc;
-	global_reg_snapshot[this_cpu].tnpc = regs->tnpc;
-	global_reg_snapshot[this_cpu].o7 = regs->u_regs[UREG_I7];
+	rp = &global_cpu_snapshot[this_cpu].reg;
+
+	rp->tstate = regs->tstate;
+	rp->tpc = regs->tpc;
+	rp->tnpc = regs->tnpc;
+	rp->o7 = regs->u_regs[UREG_I7];
 
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *rw;
@@ -223,17 +229,17 @@ static void __global_reg_self(struct thread_info *tp, struct pt_regs *regs,
 		rw = (struct reg_window *)
 			(regs->u_regs[UREG_FP] + STACK_BIAS);
 		if (kstack_valid(tp, (unsigned long) rw)) {
-			global_reg_snapshot[this_cpu].i7 = rw->ins[7];
+			rp->i7 = rw->ins[7];
 			rw = (struct reg_window *)
 				(rw->ins[6] + STACK_BIAS);
 			if (kstack_valid(tp, (unsigned long) rw))
-				global_reg_snapshot[this_cpu].rpc = rw->ins[7];
+				rp->rpc = rw->ins[7];
 		}
 	} else {
-		global_reg_snapshot[this_cpu].i7 = 0;
-		global_reg_snapshot[this_cpu].rpc = 0;
+		rp->i7 = 0;
+		rp->rpc = 0;
 	}
-	global_reg_snapshot[this_cpu].thread = tp;
+	rp->thread = tp;
 }
 
 /* In order to avoid hangs we do not try to synchronize with the
@@ -261,9 +267,9 @@ void arch_trigger_all_cpu_backtrace(void)
 	if (!regs)
 		regs = tp->kregs;
 
-	spin_lock_irqsave(&global_reg_snapshot_lock, flags);
+	spin_lock_irqsave(&global_cpu_snapshot_lock, flags);
 
-	memset(global_reg_snapshot, 0, sizeof(global_reg_snapshot));
+	memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot));
 
 	this_cpu = raw_smp_processor_id();
 
@@ -272,7 +278,7 @@ void arch_trigger_all_cpu_backtrace(void)
 	smp_fetch_global_regs();
 
 	for_each_online_cpu(cpu) {
-		struct global_reg_snapshot *gp = &global_reg_snapshot[cpu];
+		struct global_reg_snapshot *gp = &global_cpu_snapshot[cpu].reg;
 
 		__global_reg_poll(gp);
 
@@ -295,9 +301,9 @@ void arch_trigger_all_cpu_backtrace(void)
 		}
 	}
 
-	memset(global_reg_snapshot, 0, sizeof(global_reg_snapshot));
+	memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot));
 
-	spin_unlock_irqrestore(&global_reg_snapshot_lock, flags);
+	spin_unlock_irqrestore(&global_cpu_snapshot_lock, flags);
 }
 
 #ifdef CONFIG_MAGIC_SYSRQ
@@ -309,16 +315,90 @@ static void sysrq_handle_globreg(int key)
 
 static struct sysrq_key_op sparc_globalreg_op = {
 	.handler	= sysrq_handle_globreg,
-	.help_msg	= "Globalregs",
+	.help_msg	= "global-regs(Y)",
 	.action_msg	= "Show Global CPU Regs",
 };
 
-static int __init sparc_globreg_init(void)
+static void __global_pmu_self(int this_cpu)
 {
-	return register_sysrq_key('y', &sparc_globalreg_op);
+	struct global_pmu_snapshot *pp;
+	int i, num;
+
+	pp = &global_cpu_snapshot[this_cpu].pmu;
+
+	num = 1;
+	if (tlb_type == hypervisor &&
+	    sun4v_chip_type >= SUN4V_CHIP_NIAGARA4)
+		num = 4;
+
+	for (i = 0; i < num; i++) {
+		pp->pcr[i] = pcr_ops->read_pcr(i);
+		pp->pic[i] = pcr_ops->read_pic(i);
+	}
 }
 
-core_initcall(sparc_globreg_init);
+static void __global_pmu_poll(struct global_pmu_snapshot *pp)
+{
+	int limit = 0;
+
+	while (!pp->pcr[0] && ++limit < 100) {
+		barrier();
+		udelay(1);
+	}
+}
+
+static void pmu_snapshot_all_cpus(void)
+{
+	unsigned long flags;
+	int this_cpu, cpu;
+
+	spin_lock_irqsave(&global_cpu_snapshot_lock, flags);
+
+	memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot));
+
+	this_cpu = raw_smp_processor_id();
+
+	__global_pmu_self(this_cpu);
+
+	smp_fetch_global_pmu();
+
+	for_each_online_cpu(cpu) {
+		struct global_pmu_snapshot *pp = &global_cpu_snapshot[cpu].pmu;
+
+		__global_pmu_poll(pp);
+
+		printk("%c CPU[%3d]: PCR[%08lx:%08lx:%08lx:%08lx] PIC[%08lx:%08lx:%08lx:%08lx]\n",
+		       (cpu == this_cpu ? '*' : ' '), cpu,
+		       pp->pcr[0], pp->pcr[1], pp->pcr[2], pp->pcr[3],
+		       pp->pic[0], pp->pic[1], pp->pic[2], pp->pic[3]);
+	}
+
+	memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot));
+
+	spin_unlock_irqrestore(&global_cpu_snapshot_lock, flags);
+}
+
+static void sysrq_handle_globpmu(int key)
+{
+	pmu_snapshot_all_cpus();
+}
+
+static struct sysrq_key_op sparc_globalpmu_op = {
+	.handler	= sysrq_handle_globpmu,
+	.help_msg	= "global-pmu(X)",
+	.action_msg	= "Show Global PMU Regs",
+};
+
+static int __init sparc_sysrq_init(void)
+{
+	int ret = register_sysrq_key('y', &sparc_globalreg_op);
+
+	if (!ret)
+		ret = register_sysrq_key('x', &sparc_globalpmu_op);
+	return ret;
+}
+
+core_initcall(sparc_sysrq_init);
 
 #endif
 
@@ -372,13 +452,16 @@ void flush_thread(void)
 /* It's a bit more tricky when 64-bit tasks are involved... */
 static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
 {
+	bool stack_64bit = test_thread_64bit_stack(psp);
 	unsigned long fp, distance, rval;
 
-	if (!(test_thread_flag(TIF_32BIT))) {
+	if (stack_64bit) {
 		csp += STACK_BIAS;
 		psp += STACK_BIAS;
 		__get_user(fp, &(((struct reg_window __user *)psp)->ins[6]));
 		fp += STACK_BIAS;
+		if (test_thread_flag(TIF_32BIT))
+			fp &= 0xffffffff;
 	} else
 		__get_user(fp, &(((struct reg_window32 __user *)psp)->ins[6]));
 
@@ -392,7 +475,7 @@ static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
 	rval = (csp - distance);
 	if (copy_in_user((void __user *) rval, (void __user *) psp, distance))
 		rval = 0;
-	else if (test_thread_flag(TIF_32BIT)) {
+	else if (!stack_64bit) {
 		if (put_user(((u32)csp),
 			     &(((struct reg_window32 __user *)rval)->ins[6])))
 			rval = 0;
@@ -427,18 +510,18 @@ void synchronize_user_stack(void)
 
 	flush_user_windows();
 	if ((window = get_thread_wsaved()) != 0) {
-		int winsize = sizeof(struct reg_window);
-		int bias = 0;
-
-		if (test_thread_flag(TIF_32BIT))
-			winsize = sizeof(struct reg_window32);
-		else
-			bias = STACK_BIAS;
-
 		window -= 1;
 		do {
-			unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
 			struct reg_window *rwin = &t->reg_window[window];
+			int winsize = sizeof(struct reg_window);
+			unsigned long sp;
+
+			sp = t->rwbuf_stkptrs[window];
+
+			if (test_thread_64bit_stack(sp))
+				sp += STACK_BIAS;
+			else
+				winsize = sizeof(struct reg_window32);
 
 			if (!copy_to_user((char __user *)sp, rwin, winsize)) {
 				shift_window_buffer(window, get_thread_wsaved() - 1, t);
@@ -464,13 +547,6 @@ void fault_in_user_windows(void)
 {
 	struct thread_info *t = current_thread_info();
 	unsigned long window;
-	int winsize = sizeof(struct reg_window);
-	int bias = 0;
-
-	if (test_thread_flag(TIF_32BIT))
-		winsize = sizeof(struct reg_window32);
-	else
-		bias = STACK_BIAS;
 
 	flush_user_windows();
 	window = get_thread_wsaved();
@@ -478,8 +554,16 @@ void fault_in_user_windows(void)
 	if (likely(window != 0)) {
 		window -= 1;
 		do {
-			unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
 			struct reg_window *rwin = &t->reg_window[window];
+			int winsize = sizeof(struct reg_window);
+			unsigned long sp;
+
+			sp = t->rwbuf_stkptrs[window];
+
+			if (test_thread_64bit_stack(sp))
+				sp += STACK_BIAS;
+			else
+				winsize = sizeof(struct reg_window32);
 
 			if (unlikely(sp & 0x7UL))
 				stack_unaligned(sp);
diff --git a/arch/sparc/kernel/ptrace_64.c b/arch/sparc/kernel/ptrace_64.c
index 484dabac704..7ff45e4ba68 100644
--- a/arch/sparc/kernel/ptrace_64.c
+++ b/arch/sparc/kernel/ptrace_64.c
@@ -151,7 +151,7 @@ static int regwindow64_get(struct task_struct *target,
 {
 	unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-	if (test_tsk_thread_flag(current, TIF_32BIT)) {
+	if (!test_thread_64bit_stack(rw_addr)) {
 		struct reg_window32 win32;
 		int i;
 
@@ -176,7 +176,7 @@ static int regwindow64_set(struct task_struct *target,
 {
 	unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-	if (test_tsk_thread_flag(current, TIF_32BIT)) {
+	if (!test_thread_64bit_stack(rw_addr)) {
 		struct reg_window32 win32;
 		int i;
 
diff --git a/arch/sparc/kernel/smp_64.c b/arch/sparc/kernel/smp_64.c
index 781bcb10b8b..d94b878577b 100644
--- a/arch/sparc/kernel/smp_64.c
+++ b/arch/sparc/kernel/smp_64.c
@@ -852,6 +852,8 @@ extern unsigned long xcall_flush_tlb_mm;
 extern unsigned long xcall_flush_tlb_pending;
 extern unsigned long xcall_flush_tlb_kernel_range;
 extern unsigned long xcall_fetch_glob_regs;
+extern unsigned long xcall_fetch_glob_pmu;
+extern unsigned long xcall_fetch_glob_pmu_n4;
 extern unsigned long xcall_receive_signal;
 extern unsigned long xcall_new_mmu_context_version;
 #ifdef CONFIG_KGDB
@@ -1000,6 +1002,15 @@ void smp_fetch_global_regs(void)
 	smp_cross_call(&xcall_fetch_glob_regs, 0, 0, 0);
 }
 
+void smp_fetch_global_pmu(void)
+{
+	if (tlb_type == hypervisor &&
+	    sun4v_chip_type >= SUN4V_CHIP_NIAGARA4)
+		smp_cross_call(&xcall_fetch_glob_pmu_n4, 0, 0, 0);
+	else
+		smp_cross_call(&xcall_fetch_glob_pmu, 0, 0, 0);
+}
+
 /* We know that the window frames of the user have been flushed
  * to the stack before we get here because all callers of us
  * are flush_tlb_*() routines, and these run after flush_cache_*()
diff --git a/arch/sparc/kernel/sys_sparc_64.c b/arch/sparc/kernel/sys_sparc_64.c
index adfe60ece05..51b85feb8b9 100644
--- a/arch/sparc/kernel/sys_sparc_64.c
+++ b/arch/sparc/kernel/sys_sparc_64.c
@@ -729,3 +729,8 @@ SYSCALL_DEFINE5(rt_sigaction, int, sig, const struct sigaction __user *, act,
 
 	return ret;
 }
+
+asmlinkage long sys_kern_features(void)
+{
+	return KERN_FEATURE_MIXED_MODE_STACK;
+}
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index acad343667d..d8b22b3266d 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -86,6 +86,7 @@ sys_call_table32:
 	.word compat_sys_pwritev, compat_sys_rt_tgsigqueueinfo, sys_perf_event_open, compat_sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys32_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, compat_sys_open_by_handle_at, compat_sys_clock_adjtime
 	.word sys_syncfs, compat_sys_sendmmsg, sys_setns, compat_sys_process_vm_readv, compat_sys_process_vm_writev
+/*340*/	.word sys_kern_features
 
 #endif /* CONFIG_COMPAT */
 
@@ -163,3 +164,4 @@ sys_call_table:
 	.word sys_pwritev, sys_rt_tgsigqueueinfo, sys_perf_event_open, sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, sys_open_by_handle_at, sys_clock_adjtime
 	.word sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
+/*340*/	.word sys_kern_features
diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index f81d038f734..8201c25e766 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -113,21 +113,24 @@ static inline long sign_extend_imm13(long imm)
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-	unsigned long value;
+	unsigned long value, fp;
 	
 	if (reg < 16)
 		return (!reg ? 0 : regs->u_regs[reg]);
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		value = win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		get_user(value, &win32->locals[reg - 16]);
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		get_user(value, &win->locals[reg - 16]);
 	}
 	return value;
@@ -135,19 +138,24 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 
 static unsigned long *fetch_reg_addr(unsigned int reg, struct pt_regs *regs)
 {
+	unsigned long fp;
+
 	if (reg < 16)
 		return &regs->u_regs[reg];
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 *win32;
-		win32 = (struct reg_window32 *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 *)((unsigned long)((u32)fp));
 		return (unsigned long *)&win32->locals[reg - 16];
 	} else {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
 	}
 }
@@ -392,13 +400,15 @@ int handle_popc(u32 insn, struct pt_regs *regs)
 		if (rd)
 			regs->u_regs[rd] = ret;
 	} else {
-		if (test_thread_flag(TIF_32BIT)) {
+		unsigned long fp = regs->u_regs[UREG_FP];
+
+		if (!test_thread_64bit_stack(fp)) {
 			struct reg_window32 __user *win32;
-			win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+			win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 			put_user(ret, &win32->locals[rd - 16]);
 		} else {
 			struct reg_window __user *win;
-			win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+			win = (struct reg_window __user *)(fp + STACK_BIAS);
 			put_user(ret, &win->locals[rd - 16]);
 		}
 	}
@@ -554,7 +564,7 @@ void handle_ld_nf(u32 insn, struct pt_regs *regs)
 		reg[0] = 0;
 		if ((insn & 0x780000) == 0x180000)
 			reg[1] = 0;
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
 		put_user(0, (int __user *) reg);
 		if ((insn & 0x780000) == 0x180000)
 			put_user(0, ((int __user *) reg) + 1);
diff --git a/arch/sparc/kernel/visemul.c b/arch/sparc/kernel/visemul.c
index 08e074b7eb6..c096c624ac4 100644
--- a/arch/sparc/kernel/visemul.c
+++ b/arch/sparc/kernel/visemul.c
@@ -149,21 +149,24 @@ static inline void maybe_flush_windows(unsigned int rs1, unsigned int rs2,
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-	unsigned long value;
+	unsigned long value, fp;
 	
 	if (reg < 16)
 		return (!reg ? 0 : regs->u_regs[reg]);
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		value = win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		get_user(value, &win32->locals[reg - 16]);
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		get_user(value, &win->locals[reg - 16]);
 	}
 	return value;
@@ -172,16 +175,18 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 static inline unsigned long __user *__fetch_reg_addr_user(unsigned int reg,
 							  struct pt_regs *regs)
 {
+	unsigned long fp = regs->u_regs[UREG_FP];
+
 	BUG_ON(reg < 16);
 	BUG_ON(regs->tstate & TSTATE_PRIV);
 
-	if (test_thread_flag(TIF_32BIT)) {
+	if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		return (unsigned long __user *)&win32->locals[reg - 16];
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
 	}
 }
@@ -204,7 +209,7 @@ static void store_reg(struct pt_regs *regs, unsigned long val, unsigned long rd)
 	} else {
 		unsigned long __user *rd_user = __fetch_reg_addr_user(rd, regs);
 
-		if (test_thread_flag(TIF_32BIT))
+		if (!test_thread_64bit_stack(regs->u_regs[UREG_FP]))
 			__put_user((u32)val, (u32 __user *)rd_user);
 		else
 			__put_user(val, rd_user);
diff --git a/arch/sparc/kernel/winfixup.S b/arch/sparc/kernel/winfixup.S
index a6b0863c27d..1e67ce95836 100644
--- a/arch/sparc/kernel/winfixup.S
+++ b/arch/sparc/kernel/winfixup.S
@@ -43,6 +43,8 @@ spill_fixup_mna:
 spill_fixup_dax:
 	TRAP_LOAD_THREAD_REG(%g6, %g1)
 	ldx	[%g6 + TI_FLAGS], %g1
+	andcc	%sp, 0x1, %g0
+	movne	%icc, 0, %g1
 	andcc	%g1, _TIF_32BIT, %g0
 	ldub	[%g6 + TI_WSAVED], %g1
 	sll	%g1, 3, %g3
diff --git a/arch/sparc/math-emu/math_64.c b/arch/sparc/math-emu/math_64.c
index 1704068da92..034aadbff03 100644
--- a/arch/sparc/math-emu/math_64.c
+++ b/arch/sparc/math-emu/math_64.c
@@ -320,7 +320,7 @@ int do_mathemu(struct pt_regs *regs, struct fpustate *f, bool illegal_insn_trap)
 					XR = 0;
 				else if (freg < 16)
 					XR = regs->u_regs[freg];
-				else if (test_thread_flag(TIF_32BIT)) {
+				else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
 					struct reg_window32 __user *win32;
 					flushw_user ();
 					win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
diff --git a/arch/sparc/mm/ultra.S b/arch/sparc/mm/ultra.S
index 874162a11ce..f8e13d421fc 100644
--- a/arch/sparc/mm/ultra.S
+++ b/arch/sparc/mm/ultra.S
@@ -481,8 +481,8 @@ xcall_sync_tick:
 
 	.globl		xcall_fetch_glob_regs
 xcall_fetch_glob_regs:
-	sethi		%hi(global_reg_snapshot), %g1
-	or		%g1, %lo(global_reg_snapshot), %g1
+	sethi		%hi(global_cpu_snapshot), %g1
+	or		%g1, %lo(global_cpu_snapshot), %g1
 	__GET_CPUID(%g2)
 	sllx		%g2, 6, %g3
 	add		%g1, %g3, %g1
@@ -509,6 +509,66 @@ xcall_fetch_glob_regs:
 	stx		%g3, [%g1 + GR_SNAP_THREAD]
 	retry
 
+	.globl		xcall_fetch_glob_pmu
+xcall_fetch_glob_pmu:
+	sethi		%hi(global_cpu_snapshot), %g1
+	or		%g1, %lo(global_cpu_snapshot), %g1
+	__GET_CPUID(%g2)
+	sllx		%g2, 6, %g3
+	add		%g1, %g3, %g1
+	rd		%pic, %g7
+	stx		%g7, [%g1 + (4 * 8)]
+	rd		%pcr, %g7
+	stx		%g7, [%g1 + (0 * 8)]
+	retry
+
+	.globl		xcall_fetch_glob_pmu_n4
+xcall_fetch_glob_pmu_n4:
+	sethi		%hi(global_cpu_snapshot), %g1
+	or		%g1, %lo(global_cpu_snapshot), %g1
+	__GET_CPUID(%g2)
+	sllx		%g2, 6, %g3
+	add		%g1, %g3, %g1
+
+	ldxa		[%g0] ASI_PIC, %g7
+	stx		%g7, [%g1 + (4 * 8)]
+	mov		0x08, %g3
+	ldxa		[%g3] ASI_PIC, %g7
+	stx		%g7, [%g1 + (5 * 8)]
+	mov		0x10, %g3
+	ldxa		[%g3] ASI_PIC, %g7
+	stx		%g7, [%g1 + (6 * 8)]
+	mov		0x18, %g3
+	ldxa		[%g3] ASI_PIC, %g7
+	stx		%g7, [%g1 + (7 * 8)]
+
+	mov		%o0, %g2
+	mov		%o1, %g3
+	mov		%o5, %g7
+
+	mov		HV_FAST_VT_GET_PERFREG, %o5
+	mov		3, %o0
+	ta		HV_FAST_TRAP
+	stx		%o1, [%g1 + (3 * 8)]
+	mov		HV_FAST_VT_GET_PERFREG, %o5
+	mov		2, %o0
+	ta		HV_FAST_TRAP
+	stx		%o1, [%g1 + (2 * 8)]
+	mov		HV_FAST_VT_GET_PERFREG, %o5
+	mov		1, %o0
+	ta		HV_FAST_TRAP
+	stx		%o1, [%g1 + (1 * 8)]
+	mov		HV_FAST_VT_GET_PERFREG, %o5
+	mov		0, %o0
+	ta		HV_FAST_TRAP
+	stx		%o1, [%g1 + (0 * 8)]
+
+	mov		%g2, %o0
+	mov		%g3, %o1
+	mov		%g7, %o5
+
+	retry
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 	.align		32
 	.globl		xcall_flush_dcache_page_cheetah