25 files changed, 990 insertions, 478 deletions

diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile
index 706f9397479..d9da7112aaf 100644
--- a/arch/mips/kernel/Makefile
+++ b/arch/mips/kernel/Makefile
@@ -6,10 +6,11 @@ extra-y		:= head.o init_task.o vmlinux.lds
 
 obj-y		+= cpu-probe.o branch.o entry.o genex.o irq.o process.o \
 		   ptrace.o reset.o setup.o signal.o syscall.o \
-		   time.o topology.o traps.o unaligned.o
+		   time.o topology.o traps.o unaligned.o watch.o
 
 obj-$(CONFIG_CEVT_BCM1480)	+= cevt-bcm1480.o
 obj-$(CONFIG_CEVT_R4K)		+= cevt-r4k.o
+obj-$(CONFIG_MIPS_MT_SMTC)	+= cevt-smtc.o
 obj-$(CONFIG_CEVT_DS1287)	+= cevt-ds1287.o
 obj-$(CONFIG_CEVT_GT641XX)	+= cevt-gt641xx.o
 obj-$(CONFIG_CEVT_SB1250)	+= cevt-sb1250.o
diff --git a/arch/mips/kernel/cevt-r4k.c b/arch/mips/kernel/cevt-r4k.c
index 24a2d907aa0..4a4c59f2737 100644
--- a/arch/mips/kernel/cevt-r4k.c
+++ b/arch/mips/kernel/cevt-r4k.c
@@ -12,6 +12,14 @@
 
 #include <asm/smtc_ipi.h>
 #include <asm/time.h>
+#include <asm/cevt-r4k.h>
+
+/*
+ * The SMTC Kernel for the 34K, 1004K, et. al. replaces several
+ * of these routines with SMTC-specific variants.
+ */
+
+#ifndef CONFIG_MIPS_MT_SMTC
 
 static int mips_next_event(unsigned long delta,
                            struct clock_event_device *evt)
@@ -19,60 +27,27 @@ static int mips_next_event(unsigned long delta,
 	unsigned int cnt;
 	int res;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-	{
-	unsigned long flags, vpflags;
-	local_irq_save(flags);
-	vpflags = dvpe();
-#endif
 	cnt = read_c0_count();
 	cnt += delta;
 	write_c0_compare(cnt);
 	res = ((int)(read_c0_count() - cnt) > 0) ? -ETIME : 0;
-#ifdef CONFIG_MIPS_MT_SMTC
-	evpe(vpflags);
-	local_irq_restore(flags);
-	}
-#endif
 	return res;
 }
 
-static void mips_set_mode(enum clock_event_mode mode,
-                          struct clock_event_device *evt)
+#endif /* CONFIG_MIPS_MT_SMTC */
+
+void mips_set_clock_mode(enum clock_event_mode mode,
+				struct clock_event_device *evt)
 {
 	/* Nothing to do ...  */
 }
 
-static DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device);
-static int cp0_timer_irq_installed;
+DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device);
+int cp0_timer_irq_installed;
 
-/*
- * Timer ack for an R4k-compatible timer of a known frequency.
- */
-static void c0_timer_ack(void)
-{
-	write_c0_compare(read_c0_compare());
-}
+#ifndef CONFIG_MIPS_MT_SMTC
 
-/*
- * Possibly handle a performance counter interrupt.
- * Return true if the timer interrupt should not be checked
- */
-static inline int handle_perf_irq(int r2)
-{
-	/*
-	 * The performance counter overflow interrupt may be shared with the
-	 * timer interrupt (cp0_perfcount_irq < 0). If it is and a
-	 * performance counter has overflowed (perf_irq() == IRQ_HANDLED)
-	 * and we can't reliably determine if a counter interrupt has also
-	 * happened (!r2) then don't check for a timer interrupt.
-	 */
-	return (cp0_perfcount_irq < 0) &&
-		perf_irq() == IRQ_HANDLED &&
-		!r2;
-}
-
-static irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
+irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
 {
 	const int r2 = cpu_has_mips_r2;
 	struct clock_event_device *cd;
@@ -93,12 +68,8 @@ static irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
 	 * interrupt.  Being the paranoiacs we are we check anyway.
 	 */
 	if (!r2 || (read_c0_cause() & (1 << 30))) {
-		c0_timer_ack();
-#ifdef CONFIG_MIPS_MT_SMTC
-		if (cpu_data[cpu].vpe_id)
-			goto out;
-		cpu = 0;
-#endif
+		/* Clear Count/Compare Interrupt */
+		write_c0_compare(read_c0_compare());
 		cd = &per_cpu(mips_clockevent_device, cpu);
 		cd->event_handler(cd);
 	}
@@ -107,65 +78,16 @@ out:
 	return IRQ_HANDLED;
 }
 
-static struct irqaction c0_compare_irqaction = {
+#endif /* Not CONFIG_MIPS_MT_SMTC */
+
+struct irqaction c0_compare_irqaction = {
 	.handler = c0_compare_interrupt,
-#ifdef CONFIG_MIPS_MT_SMTC
-	.flags = IRQF_DISABLED,
-#else
 	.flags = IRQF_DISABLED | IRQF_PERCPU,
-#endif
 	.name = "timer",
 };
 
-#ifdef CONFIG_MIPS_MT_SMTC
-DEFINE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device);
-
-static void smtc_set_mode(enum clock_event_mode mode,
-                          struct clock_event_device *evt)
-{
-}
-
-static void mips_broadcast(cpumask_t mask)
-{
-	unsigned int cpu;
-
-	for_each_cpu_mask(cpu, mask)
-		smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0);
-}
-
-static void setup_smtc_dummy_clockevent_device(void)
-{
-	//uint64_t mips_freq = mips_hpt_^frequency;
-	unsigned int cpu = smp_processor_id();
-	struct clock_event_device *cd;
 
-	cd = &per_cpu(smtc_dummy_clockevent_device, cpu);
-
-	cd->name		= "SMTC";
-	cd->features		= CLOCK_EVT_FEAT_DUMMY;
-
-	/* Calculate the min / max delta */
-	cd->mult	= 0; //div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32);
-	cd->shift		= 0; //32;
-	cd->max_delta_ns	= 0; //clockevent_delta2ns(0x7fffffff, cd);
-	cd->min_delta_ns	= 0; //clockevent_delta2ns(0x30, cd);
-
-	cd->rating		= 200;
-	cd->irq			= 17; //-1;
-//	if (cpu)
-//		cd->cpumask	= CPU_MASK_ALL; // cpumask_of_cpu(cpu);
-//	else
-		cd->cpumask	= cpumask_of_cpu(cpu);
-
-	cd->set_mode		= smtc_set_mode;
-
-	cd->broadcast		= mips_broadcast;
-
-	clockevents_register_device(cd);
-}
-#endif
-
-static void mips_event_handler(struct clock_event_device *dev)
+void mips_event_handler(struct clock_event_device *dev)
 {
 }
 
@@ -177,7 +99,23 @@ static int c0_compare_int_pending(void)
 	return (read_c0_cause() >> cp0_compare_irq) & 0x100;
 }
 
-static int c0_compare_int_usable(void)
+/*
+ * Compare interrupt can be routed and latched outside the core,
+ * so a single execution hazard barrier may not be enough to give
+ * it time to clear as seen in the Cause register.  4 time the
+ * pipeline depth seems reasonably conservative, and empirically
+ * works better in configurations with high CPU/bus clock ratios.
+ */
+
+#define compare_change_hazard() \
+	do { \
+		irq_disable_hazard(); \
+		irq_disable_hazard(); \
+		irq_disable_hazard(); \
+		irq_disable_hazard(); \
+	} while (0)
+
+int c0_compare_int_usable(void)
 {
 	unsigned int delta;
 	unsigned int cnt;
@@ -187,7 +125,7 @@ static int c0_compare_int_usable(void)
 	 */
 	if (c0_compare_int_pending()) {
 		write_c0_compare(read_c0_count());
-		irq_disable_hazard();
+		compare_change_hazard();
 		if (c0_compare_int_pending())
 			return 0;
 	}
@@ -196,7 +134,7 @@ static int c0_compare_int_usable(void)
 		cnt = read_c0_count();
 		cnt += delta;
 		write_c0_compare(cnt);
-		irq_disable_hazard();
+		compare_change_hazard();
 		if ((int)(read_c0_count() - cnt) < 0)
 		    break;
 		/* increase delta if the timer was already expired */
@@ -205,11 +143,12 @@ static int c0_compare_int_usable(void)
 	while ((int)(read_c0_count() - cnt) <= 0)
 		;	/* Wait for expiry  */
 
+	compare_change_hazard();
 	if (!c0_compare_int_pending())
 		return 0;
 
 	write_c0_compare(read_c0_count());
-	irq_disable_hazard();
+	compare_change_hazard();
 	if (c0_compare_int_pending())
 		return 0;
 
@@ -219,6 +158,8 @@ static int c0_compare_int_usable(void)
 	return 1;
 }
 
+#ifndef CONFIG_MIPS_MT_SMTC
+
 int __cpuinit mips_clockevent_init(void)
 {
 	uint64_t mips_freq = mips_hpt_frequency;
@@ -229,17 +170,6 @@ int __cpuinit mips_clockevent_init(void)
 	if (!cpu_has_counter || !mips_hpt_frequency)
 		return -ENXIO;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-	setup_smtc_dummy_clockevent_device();
-
-	/*
-	 * On SMTC we only register VPE0's compare interrupt as clockevent
-	 * device.
-	 */
-	if (cpu)
-		return 0;
-#endif
-
 	if (!c0_compare_int_usable())
 		return -ENXIO;
 
@@ -265,13 +195,9 @@ int __cpuinit mips_clockevent_init(void)
 
 	cd->rating		= 300;
 	cd->irq			= irq;
-#ifdef CONFIG_MIPS_MT_SMTC
-	cd->cpumask		= CPU_MASK_ALL;
-#else
 	cd->cpumask		= cpumask_of_cpu(cpu);
-#endif
 	cd->set_next_event	= mips_next_event;
-	cd->set_mode		= mips_set_mode;
+	cd->set_mode		= mips_set_clock_mode;
 	cd->event_handler	= mips_event_handler;
 
 	clockevents_register_device(cd);
@@ -281,12 +207,9 @@ int __cpuinit mips_clockevent_init(void)
 
 	cp0_timer_irq_installed = 1;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-#define CPUCTR_IMASKBIT (0x100 << cp0_compare_irq)
-	setup_irq_smtc(irq, &c0_compare_irqaction, CPUCTR_IMASKBIT);
-#else
 	setup_irq(irq, &c0_compare_irqaction);
-#endif
 
 	return 0;
 }
+
+#endif /* Not CONFIG_MIPS_MT_SMTC */
diff --git a/arch/mips/kernel/cevt-smtc.c b/arch/mips/kernel/cevt-smtc.c
new file mode 100644
index 00000000000..5162fe4b595
--- /dev/null
+++ b/arch/mips/kernel/cevt-smtc.c
@@ -0,0 +1,321 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2007 MIPS Technologies, Inc.
+ * Copyright (C) 2007 Ralf Baechle <ralf@linux-mips.org>
+ * Copyright (C) 2008 Kevin D. Kissell, Paralogos sarl
+ */
+#include <linux/clockchips.h>
+#include <linux/interrupt.h>
+#include <linux/percpu.h>
+
+#include <asm/smtc_ipi.h>
+#include <asm/time.h>
+#include <asm/cevt-r4k.h>
+
+/*
+ * Variant clock event timer support for SMTC on MIPS 34K, 1004K
+ * or other MIPS MT cores.
+ *
+ * Notes on SMTC Support:
+ *
+ * SMTC has multiple microthread TCs pretending to be Linux CPUs.
+ * But there's only one Count/Compare pair per VPE, and Compare
+ * interrupts are taken opportunisitically by available TCs
+ * bound to the VPE with the Count register.  The new timer
+ * framework provides for global broadcasts, but we really
+ * want VPE-level multicasts for best behavior. So instead
+ * of invoking the high-level clock-event broadcast code,
+ * this version of SMTC support uses the historical SMTC
+ * multicast mechanisms "under the hood", appearing to the
+ * generic clock layer as if the interrupts are per-CPU.
+ *
+ * The approach taken here is to maintain a set of NR_CPUS
+ * virtual timers, and track which "CPU" needs to be alerted
+ * at each event.
+ *
+ * It's unlikely that we'll see a MIPS MT core with more than
+ * 2 VPEs, but we *know* that we won't need to handle more
+ * VPEs than we have "CPUs".  So NCPUs arrays of NCPUs elements
+ * is always going to be overkill, but always going to be enough.
+ */
+
+unsigned long smtc_nexttime[NR_CPUS][NR_CPUS];
+static int smtc_nextinvpe[NR_CPUS];
+
+/*
+ * Timestamps stored are absolute values to be programmed
+ * into Count register.  Valid timestamps will never be zero.
+ * If a Zero Count value is actually calculated, it is converted
+ * to be a 1, which will introduce 1 or two CPU cycles of error
+ * roughly once every four billion events, which at 1000 HZ means
+ * about once every 50 days.  If that's actually a problem, one
+ * could alternate squashing 0 to 1 and to -1.
+ */
+
+#define MAKEVALID(x) (((x) == 0L) ? 1L : (x))
+#define ISVALID(x) ((x) != 0L)
+
+/*
+ * Time comparison is subtle, as it's really truncated
+ * modular arithmetic.
+ */
+
+#define IS_SOONER(a, b, reference) \
+    (((a) - (unsigned long)(reference)) < ((b) - (unsigned long)(reference)))
+
+/*
+ * CATCHUP_INCREMENT, used when the function falls behind the counter.
+ * Could be an increasing function instead of a constant;
+ */
+
+#define CATCHUP_INCREMENT 64
+
+static int mips_next_event(unsigned long delta,
+				struct clock_event_device *evt)
+{
+	unsigned long flags;
+	unsigned int mtflags;
+	unsigned long timestamp, reference, previous;
+	unsigned long nextcomp = 0L;
+	int vpe = current_cpu_data.vpe_id;
+	int cpu = smp_processor_id();
+	local_irq_save(flags);
+	mtflags = dmt();
+
+	/*
+	 * Maintain the per-TC virtual timer
+	 * and program the per-VPE shared Count register
+	 * as appropriate here...
+	 */
+	reference = (unsigned long)read_c0_count();
+	timestamp = MAKEVALID(reference + delta);
+	/*
+	 * To really model the clock, we have to catch the case
+	 * where the current next-in-VPE timestamp is the old
+	 * timestamp for the calling CPE, but the new value is
+	 * in fact later.  In that case, we have to do a full
+	 * scan and discover the new next-in-VPE CPU id and
+	 * timestamp.
+	 */
+	previous = smtc_nexttime[vpe][cpu];
+	if (cpu == smtc_nextinvpe[vpe] && ISVALID(previous)
+	    && IS_SOONER(previous, timestamp, reference)) {
+		int i;
+		int soonest = cpu;
+
+		/*
+		 * Update timestamp array here, so that new
+		 * value gets considered along with those of
+		 * other virtual CPUs on the VPE.
+		 */
+		smtc_nexttime[vpe][cpu] = timestamp;
+		for_each_online_cpu(i) {
+			if (ISVALID(smtc_nexttime[vpe][i])
+			    && IS_SOONER(smtc_nexttime[vpe][i],
+				smtc_nexttime[vpe][soonest], reference)) {
+				    soonest = i;
+			}
+		}
+		smtc_nextinvpe[vpe] = soonest;
+		nextcomp = smtc_nexttime[vpe][soonest];
+	/*
+	 * Otherwise, we don't have to process the whole array rank,
+	 * we just have to see if the event horizon has gotten closer.
+	 */
+	} else {
+		if (!ISVALID(smtc_nexttime[vpe][smtc_nextinvpe[vpe]]) ||
+		    IS_SOONER(timestamp,
+			smtc_nexttime[vpe][smtc_nextinvpe[vpe]], reference)) {
+			    smtc_nextinvpe[vpe] = cpu;
+			    nextcomp = timestamp;
+		}
+		/*
+		 * Since next-in-VPE may me the same as the executing
+		 * virtual CPU, we update the array *after* checking
+		 * its value.
+		 */
+		smtc_nexttime[vpe][cpu] = timestamp;
+	}
+
+	/*
+	 * It may be that, in fact, we don't need to update Compare,
+	 * but if we do, we want to make sure we didn't fall into
+	 * a crack just behind Count.
+	 */
+	if (ISVALID(nextcomp)) {
+		write_c0_compare(nextcomp);
+		ehb();
+		/*
+		 * We never return an error, we just make sure
+		 * that we trigger the handlers as quickly as
+		 * we can if we fell behind.
+		 */
+		while ((nextcomp - (unsigned long)read_c0_count())
+			> (unsigned long)LONG_MAX) {
+			nextcomp += CATCHUP_INCREMENT;
+			write_c0_compare(nextcomp);
+			ehb();
+		}
+	}
+	emt(mtflags);
+	local_irq_restore(flags);
+	return 0;
+}
+
+
+void smtc_distribute_timer(int vpe)
+{
+	unsigned long flags;
+	unsigned int mtflags;
+	int cpu;
+	struct clock_event_device *cd;
+	unsigned long nextstamp = 0L;
+	unsigned long reference;
+
+
+repeat:
+	for_each_online_cpu(cpu) {
+	    /*
+	     * Find virtual CPUs within the current VPE who have
+	     * unserviced timer requests whose time is now past.
+	     */
+	    local_irq_save(flags);
+	    mtflags = dmt();
+	    if (cpu_data[cpu].vpe_id == vpe &&
+		ISVALID(smtc_nexttime[vpe][cpu])) {
+		reference = (unsigned long)read_c0_count();
+		if ((smtc_nexttime[vpe][cpu] - reference)
+			 > (unsigned long)LONG_MAX) {
+			    smtc_nexttime[vpe][cpu] = 0L;
+			    emt(mtflags);
+			    local_irq_restore(flags);
+			    /*
+			     * We don't send IPIs to ourself.
+			     */
+			    if (cpu != smp_processor_id()) {
+				smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0);
+			    } else {
+				cd = &per_cpu(mips_clockevent_device, cpu);
+				cd->event_handler(cd);
+			    }
+		} else {
+			/* Local to VPE but Valid Time not yet reached. */
+			if (!ISVALID(nextstamp) ||
+			    IS_SOONER(smtc_nexttime[vpe][cpu], nextstamp,
+			    reference)) {
+				smtc_nextinvpe[vpe] = cpu;
+				nextstamp = smtc_nexttime[vpe][cpu];
+			}
+			emt(mtflags);
+			local_irq_restore(flags);
+		}
+	    } else {
+		emt(mtflags);
+		local_irq_restore(flags);
+
+	    }
+	}
+	/* Reprogram for interrupt at next soonest timestamp for VPE */
+	if (ISVALID(nextstamp)) {
+		write_c0_compare(nextstamp);
+		ehb();
+		if ((nextstamp - (unsigned long)read_c0_count())
+			> (unsigned long)LONG_MAX)
+				goto repeat;
+	}
+}
+
+
+irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
+{
+	int cpu = smp_processor_id();
+
+	/* If we're running SMTC, we've got MIPS MT and therefore MIPS32R2 */
+	handle_perf_irq(1);
+
+	if (read_c0_cause() & (1 << 30)) {
+		/* Clear Count/Compare Interrupt */
+		write_c0_compare(read_c0_compare());
+		smtc_distribute_timer(cpu_data[cpu].vpe_id);
+	}
+	return IRQ_HANDLED;
+}
+
+
+int __cpuinit mips_clockevent_init(void)
+{
+	uint64_t mips_freq = mips_hpt_frequency;
+	unsigned int cpu = smp_processor_id();
+	struct clock_event_device *cd;
+	unsigned int irq;
+	int i;
+	int j;
+
+	if (!cpu_has_counter || !mips_hpt_frequency)
+		return -ENXIO;
+	if (cpu == 0) {
+		for (i = 0; i < num_possible_cpus(); i++) {
+			smtc_nextinvpe[i] = 0;
+			for (j = 0; j < num_possible_cpus(); j++)
+				smtc_nexttime[i][j] = 0L;
+		}
+		/*
+		 * SMTC also can't have the usablility test
+		 * run by secondary TCs once Compare is in use.
+		 */
+		if (!c0_compare_int_usable())
+			return -ENXIO;
+	}
+
+	/*
+	 * With vectored interrupts things are getting platform specific.
+	 * get_c0_compare_int is a hook to allow a platform to return the
+	 * interrupt number of it's liking.
+	 */
+	irq = MIPS_CPU_IRQ_BASE + cp0_compare_irq;
+	if (get_c0_compare_int)
+		irq = get_c0_compare_int();
+
+	cd = &per_cpu(mips_clockevent_device, cpu);
+
+	cd->name		= "MIPS";
+	cd->features		= CLOCK_EVT_FEAT_ONESHOT;
+
+	/* Calculate the min / max delta */
+	cd->mult	= div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32);
+	cd->shift		= 32;
+	cd->max_delta_ns	= clockevent_delta2ns(0x7fffffff, cd);
+	cd->min_delta_ns	= clockevent_delta2ns(0x300, cd);
+
+	cd->rating		= 300;
+	cd->irq			= irq;
+	cd->cpumask		= cpumask_of_cpu(cpu);
+	cd->set_next_event	= mips_next_event;
+	cd->set_mode		= mips_set_clock_mode;
+	cd->event_handler	= mips_event_handler;
+
+	clockevents_register_device(cd);
+
+	/*
+	 * On SMTC we only want to do the data structure
+	 * initialization and IRQ setup once.
+	 */
+	if (cpu)
+		return 0;
+	/*
+	 * And we need the hwmask associated with the c0_compare
+	 * vector to be initialized.
+	 */
+	irq_hwmask[irq] = (0x100 << cp0_compare_irq);
+	if (cp0_timer_irq_installed)
+		return 0;
+
+	cp0_timer_irq_installed = 1;
+
+	setup_irq(irq, &c0_compare_irqaction);
+
+	return 0;
+}
diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c
index 335a6ae3d59..0cf15457eca 100644
--- a/arch/mips/kernel/cpu-probe.c
+++ b/arch/mips/kernel/cpu-probe.c
@@ -21,6 +21,7 @@
 #include <asm/fpu.h>
 #include <asm/mipsregs.h>
 #include <asm/system.h>
+#include <asm/watch.h>
 
 /*
  * Not all of the MIPS CPUs have the "wait" instruction available. Moreover,
@@ -45,18 +46,7 @@ static void r39xx_wait(void)
 	local_irq_enable();
 }
 
-/*
- * There is a race when WAIT instruction executed with interrupt
- * enabled.
- * But it is implementation-dependent wheter the pipelie restarts when
- * a non-enabled interrupt is requested.
- */
-static void r4k_wait(void)
-{
-	__asm__("	.set	mips3			\n"
-		"	wait				\n"
-		"	.set	mips0			\n");
-}
+extern void r4k_wait(void);
 
 /*
  * This variant is preferable as it allows testing need_resched and going to
@@ -65,14 +55,18 @@ static void r4k_wait(void)
  * interrupt is requested" restriction in the MIPS32/MIPS64 architecture makes
  * using this version a gamble.
  */
-static void r4k_wait_irqoff(void)
+void r4k_wait_irqoff(void)
 {
 	local_irq_disable();
 	if (!need_resched())
-		__asm__("	.set	mips3		\n"
+		__asm__("	.set	push		\n"
+			"	.set	mips3		\n"
 			"	wait			\n"
-			"	.set	mips0		\n");
+			"	.set	pop		\n");
 	local_irq_enable();
+	__asm__(" 	.globl __pastwait	\n"
+		"__pastwait:			\n");
+	return;
 }
 
 /*
@@ -128,7 +122,7 @@ static int __init wait_disable(char *s)
 
 __setup("nowait", wait_disable);
 
-static inline void check_wait(void)
+void __init check_wait(void)
 {
 	struct cpuinfo_mips *c = &current_cpu_data;
 
@@ -242,7 +236,6 @@ static inline void check_errata(void)
 
 void __init check_bugs32(void)
 {
-	check_wait();
 	check_errata();
 }
 
@@ -685,6 +678,7 @@ static inline void spram_config(void) {}
 static inline void cpu_probe_mips(struct cpuinfo_mips *c)
 {
 	decode_configs(c);
+	mips_probe_watch_registers(c);
 	switch (c->processor_id & 0xff00) {
 	case PRID_IMP_4KC:
 		c->cputype = CPU_4KC;
diff --git a/arch/mips/kernel/entry.S b/arch/mips/kernel/entry.S
index e29598ae939..ffa331029e0 100644
--- a/arch/mips/kernel/entry.S
+++ b/arch/mips/kernel/entry.S
@@ -79,11 +79,6 @@ FEXPORT(syscall_exit)
 
 FEXPORT(restore_all)			# restore full frame
 #ifdef CONFIG_MIPS_MT_SMTC
-/* Detect and execute deferred IPI "interrupts" */
-	LONG_L	s0, TI_REGS($28)
-	LONG_S	sp, TI_REGS($28)
-	jal	deferred_smtc_ipi
-	LONG_S	s0, TI_REGS($28)
 #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP
 /* Re-arm any temporarily masked interrupts not explicitly "acked" */
 	mfc0	v0, CP0_TCSTATUS
@@ -112,6 +107,11 @@ FEXPORT(restore_all)			# restore full frame
 	xor	t0, t0, t3
 	mtc0	t0, CP0_TCCONTEXT
 #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */
+/* Detect and execute deferred IPI "interrupts" */
+	LONG_L	s0, TI_REGS($28)
+	LONG_S	sp, TI_REGS($28)
+	jal	deferred_smtc_ipi
+	LONG_S	s0, TI_REGS($28)
 #endif /* CONFIG_MIPS_MT_SMTC */
 	.set	noat
 	RESTORE_TEMP
diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S
index c6ada98ee04..757d48f0d80 100644
--- a/arch/mips/kernel/genex.S
+++ b/arch/mips/kernel/genex.S
@@ -20,6 +20,7 @@
 #include <asm/stackframe.h>
 #include <asm/war.h>
 #include <asm/page.h>
+#include <asm/thread_info.h>
 
 #define PANIC_PIC(msg)					\
 		.set push;				\
@@ -126,7 +127,42 @@ handle_vcei:
 
 	__FINIT
 
+	.align	5	/* 32 byte rollback region */
+LEAF(r4k_wait)
+	.set	push
+	.set	noreorder
+	/* start of rollback region */
+	LONG_L	t0, TI_FLAGS($28)
+	nop
+	andi	t0, _TIF_NEED_RESCHED
+	bnez	t0, 1f
+	 nop
+	nop
+	nop
+	.set	mips3
+	wait
+	/* end of rollback region (the region size must be power of two) */
+	.set	pop
+1:
+	jr	ra
+	END(r4k_wait)
+
+	.macro	BUILD_ROLLBACK_PROLOGUE handler
+	FEXPORT(rollback_\handler)
+	.set	push
+	.set	noat
+	MFC0	k0, CP0_EPC
+	PTR_LA	k1, r4k_wait
+	ori	k0, 0x1f	/* 32 byte rollback region */
+	xori	k0, 0x1f
+	bne	k0, k1, 9f
+	MTC0	k0, CP0_EPC
+9:
+	.set pop
+	.endm
+
 	.align  5
+BUILD_ROLLBACK_PROLOGUE handle_int
 NESTED(handle_int, PT_SIZE, sp)
 #ifdef CONFIG_TRACE_IRQFLAGS
 	/*
@@ -201,6 +237,7 @@ NESTED(except_vec_ejtag_debug, 0, sp)
  * This prototype is copied to ebase + n*IntCtl.VS and patched
  * to invoke the handler
  */
+BUILD_ROLLBACK_PROLOGUE except_vec_vi
 NESTED(except_vec_vi, 0, sp)
 	SAVE_SOME
 	SAVE_AT
@@ -245,8 +282,8 @@ NESTED(except_vec_vi_handler, 0, sp)
 	and	t0, a0, t1
 #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP
 	mfc0	t2, CP0_TCCONTEXT
-	or	t0, t0, t2
-	mtc0	t0, CP0_TCCONTEXT
+	or	t2, t0, t2
+	mtc0	t2, CP0_TCCONTEXT
 #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */
 	xor	t1, t1, t0
 	mtc0	t1, CP0_STATUS
@@ -416,7 +453,11 @@ NESTED(nmi_handler, PT_SIZE, sp)
 	BUILD_HANDLER tr tr sti silent			/* #13 */
 	BUILD_HANDLER fpe fpe fpe silent		/* #15 */
 	BUILD_HANDLER mdmx mdmx sti silent		/* #22 */
+#ifdef 	CONFIG_HARDWARE_WATCHPOINTS
+	BUILD_HANDLER watch watch sti silent		/* #23 */
+#else
 	BUILD_HANDLER watch watch sti verbose		/* #23 */
+#endif
 	BUILD_HANDLER mcheck mcheck cli verbose		/* #24 */
 	BUILD_HANDLER mt mt sti silent			/* #25 */
 	BUILD_HANDLER dsp dsp sti silent		/* #26 */
diff --git a/arch/mips/kernel/head.S b/arch/mips/kernel/head.S
index 361364501d3..492a0a8d70f 100644
--- a/arch/mips/kernel/head.S
+++ b/arch/mips/kernel/head.S
@@ -22,6 +22,7 @@
 #include <asm/irqflags.h>
 #include <asm/regdef.h>
 #include <asm/page.h>
+#include <asm/pgtable-bits.h>
 #include <asm/mipsregs.h>
 #include <asm/stackframe.h>
 
diff --git a/arch/mips/kernel/kgdb.c b/arch/mips/kernel/kgdb.c
index 8f6d58ede33..6e152c80cd4 100644
--- a/arch/mips/kernel/kgdb.c
+++ b/arch/mips/kernel/kgdb.c
@@ -236,8 +236,7 @@ int kgdb_arch_handle_exception(int vector, int signo, int err_code,
 
 		atomic_set(&kgdb_cpu_doing_single_step, -1);
 		if (remcom_in_buffer[0] == 's')
-			if (kgdb_contthread)
-				atomic_set(&kgdb_cpu_doing_single_step, cpu);
+			atomic_set(&kgdb_cpu_doing_single_step, cpu);
 
 		return 0;
 	}
diff --git a/arch/mips/kernel/linux32.c b/arch/mips/kernel/linux32.c
index 2fefb14414b..aa2c55e3b55 100644
--- a/arch/mips/kernel/linux32.c
+++ b/arch/mips/kernel/linux32.c
@@ -63,41 +63,6 @@
 #define merge_64(r1, r2) ((((r2) & 0xffffffffUL) << 32) + ((r1) & 0xffffffffUL))
 #endif
 
-/*
- * Revalidate the inode. This is required for proper NFS attribute caching.
- */
-
-int cp_compat_stat(struct kstat *stat, struct compat_stat __user *statbuf)
-{
-	struct compat_stat tmp;
-
-	if (!new_valid_dev(stat->dev) || !new_valid_dev(stat->rdev))
-		return -EOVERFLOW;
-
-	memset(&tmp, 0, sizeof(tmp));
-	tmp.st_dev = new_encode_dev(stat->dev);
-	tmp.st_ino = stat->ino;
-	if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
-		return -EOVERFLOW;
-	tmp.st_mode = stat->mode;
-	tmp.st_nlink = stat->nlink;
-	SET_UID(tmp.st_uid, stat->uid);
-	SET_GID(tmp.st_gid, stat->gid);
-	tmp.st_rdev = new_encode_dev(stat->rdev);
-	tmp.st_size = stat->size;
-	tmp.st_atime = stat->atime.tv_sec;
-	tmp.st_mtime = stat->mtime.tv_sec;
-	tmp.st_ctime = stat->ctime.tv_sec;
-#ifdef STAT_HAVE_NSEC
-	tmp.st_atime_nsec = stat->atime.tv_nsec;
-	tmp.st_mtime_nsec = stat->mtime.tv_nsec;
-	tmp.st_ctime_nsec = stat->ctime.tv_nsec;
-#endif
-	tmp.st_blocks = stat->blocks;
-	tmp.st_blksize = stat->blksize;
-	return copy_to_user(statbuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
-}
-
 asmlinkage unsigned long
 sys32_mmap2(unsigned long addr, unsigned long len, unsigned long prot,
          unsigned long flags, unsigned long fd, unsigned long pgoff)
@@ -168,72 +133,6 @@ asmlinkage long sys32_ftruncate64(unsigned int fd, unsigned long __dummy,
 	return sys_ftruncate(fd, merge_64(a2, a3));
 }
 
-static inline long
-get_tv32(struct timeval *o, struct compat_timeval __user *i)
-{
-	return (!access_ok(VERIFY_READ, i, sizeof(*i)) ||
-		(__get_user(o->tv_sec, &i->tv_sec) |
-		 __get_user(o->tv_usec, &i->tv_usec)));
-}
-
-static inline long
-put_tv32(struct compat_timeval __user *o, struct timeval *i)
-{
-	return (!access_ok(VERIFY_WRITE, o, sizeof(*o)) ||
-		(__put_user(i->tv_sec, &o->tv_sec) |
-		 __put_user(i->tv_usec, &o->tv_usec)));
-}
-
-extern struct timezone sys_tz;
-
-asmlinkage int
-sys32_gettimeofday(struct compat_timeval __user *tv, struct timezone __user *tz)
-{
-	if (tv) {
-		struct timeval ktv;
-		do_gettimeofday(&ktv);
-		if (put_tv32(tv, &ktv))
-			return -EFAULT;
-	}
-	if (tz) {
-		if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
-			return -EFAULT;
-	}
-	return 0;
-}
-
-static inline long get_ts32(struct timespec *o, struct compat_timeval __user *i)
-{
-	long usec;
-
-	if (!access_ok(VERIFY_READ, i, sizeof(*i)))
-		return -EFAULT;
-	if (__get_user(o->tv_sec, &i->tv_sec))
-		return -EFAULT;
-	if (__get_user(usec, &i->tv_usec))
-		return -EFAULT;
-	o->tv_nsec = usec * 1000;
-		return 0;
-}
-
-asmlinkage int
-sys32_settimeofday(struct compat_timeval __user *tv, struct timezone __user *tz)
-{
-	struct timespec kts;
-	struct timezone ktz;
-
- 	if (tv) {
-		if (get_ts32(&kts, tv))
-			return -EFAULT;
-	}
-	if (tz) {
-		if (copy_from_user(&ktz, tz, sizeof(ktz)))
-			return -EFAULT;
-	}
-
-	return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL);
-}
-
 asmlinkage int sys32_llseek(unsigned int fd, unsigned int offset_hi