[PATCH 00/17] xtensa SMP queue

[PATCH 00/17] xtensa SMP queue

[Max Filippov]

Hi Chris,

this series adds SMP support for xtensa.
There's a number of trivial cleanups:
      xtensa: remove NO_IRQ definitions
      xtensa: fix build warning in 64-bit build environment
      xtensa: fix build warning from gcc-4.7.2
      xtensa: call check_s32c1i after trap_init
      xtensa: move init_mmu declaration to mmu_context.h

a couple of fixes for spinlocks:
      xtensa: fix arch spinlock function names
      xtensa: fix __delay for small loop count

three fixes for the clocksources, with these we get high precision timers
usable in SMP and IRQ time accounting:
      xtensa: mark ccount as continuous clocksource
      xtensa: update clockevent setup for SMP
      xtensa: enable HAVE_IRQ_TIME_ACCOUNTING

fixes and rearrangement of IRQ handling code and a new driver for MX
interrupt distributor:
      xtensa: move built-in PIC to drivers/irqchip
      xtensa: clean up do_interrupt/do_IRQ
      xtensa: clear timer IRQ unconditionally in its handler
      xtensa: add MX irqchip

and finally SMP and hotplug CPU support and a sample SMP-capable CPU variant:
      xtensa: add SMP support
      xtensa: implement CPU hotplug
      xtensa: add Three Core HiFi-2 MX Variant.

Please review.

You can pull these changes from my git tree,
the following changes since commit 61e6cfa80de5760bbe406f4e815b7739205754d2:

  Linux 3.12-rc5 (2013-10-13 15:41:28 -0700)

are available in the git repository at:

  git://github.com/jcmvbkbc/linux-xtensa.git tags/xtensa-smp-20131017

for you to fetch changes up to 4862c5ae570b8a40275f413dfa2e3f184ecfa1d9:

  xtensa: implement CPU hotplug (2013-10-17 01:50:17 +0400)

----------------------------------------------------------------
Clean up minor warnings.
Fix spinlock function names, provide missing definitions.
Extract PIC into drivers/irqchip.
Clean up IRQ handling.
Add MX interrupt distributor.
Add SMP support.
Add hotplug CPU support.
Add SMP-capable core variant.

----------------------------------------------------------------
Baruch Siach (1):
      xtensa: mark ccount as continuous clocksource

Max Filippov (15):
      xtensa: remove NO_IRQ definitions
      xtensa: fix build warning in 64-bit build environment
      xtensa: fix build warning from gcc-4.7.2
      xtensa: fix arch spinlock function names
      xtensa: fix __delay for small loop count
      xtensa: enable HAVE_IRQ_TIME_ACCOUNTING
      xtensa: update clockevent setup for SMP
      xtensa: call check_s32c1i after trap_init
      xtensa: move init_mmu declaration to mmu_context.h
      xtensa: move built-in PIC to drivers/irqchip
      xtensa: clean up do_interrupt/do_IRQ
      xtensa: clear timer IRQ unconditionally in its handler
      xtensa: add MX irqchip
      xtensa: add SMP support
      xtensa: implement CPU hotplug

Piet Delaney (1):
      xtensa: add Three Core HiFi-2 MX Variant.

 arch/xtensa/Kconfig                                |  54 +-
 arch/xtensa/Makefile                               |   1 +
 arch/xtensa/include/asm/barrier.h                  |   4 +-
 arch/xtensa/include/asm/bitops.h                   |   8 +-
 arch/xtensa/include/asm/cacheflush.h               |  40 +-
 arch/xtensa/include/asm/delay.h                    |   9 +-
 arch/xtensa/include/asm/irq.h                      |   9 +
 arch/xtensa/include/asm/mmu.h                      |  10 +-
 arch/xtensa/include/asm/mmu_context.h              | 106 ++--
 arch/xtensa/include/asm/mxregs.h                   |  46 ++
 arch/xtensa/include/asm/processor.h                |  20 +
 arch/xtensa/include/asm/ptrace.h                   |   8 +
 arch/xtensa/include/asm/smp.h                      |  38 +-
 arch/xtensa/include/asm/spinlock.h                 |  31 +-
 arch/xtensa/include/asm/spinlock_types.h           |  20 +
 arch/xtensa/include/asm/timex.h                    |  14 +-
 arch/xtensa/include/asm/tlbflush.h                 |  42 +-
 arch/xtensa/include/asm/traps.h                    |   1 +
 arch/xtensa/include/asm/vectors.h                  |   4 +
 arch/xtensa/kernel/Makefile                        |   1 +
 arch/xtensa/kernel/head.S                          | 181 +++++--
 arch/xtensa/kernel/irq.c                           | 207 ++++----
 arch/xtensa/kernel/mxhead.S                        |  85 +++
 arch/xtensa/kernel/setup.c                         |  92 ++--
 arch/xtensa/kernel/smp.c                           | 591 +++++++++++++++++++++
 arch/xtensa/kernel/time.c                          |  56 +-
 arch/xtensa/kernel/traps.c                         |  56 +-
 arch/xtensa/kernel/vmlinux.lds.S                   |  26 +
 arch/xtensa/mm/cache.c                             |   7 +-
 arch/xtensa/mm/fault.c                             |   2 +-
 arch/xtensa/mm/misc.S                              |   4 +-
 arch/xtensa/mm/mmu.c                               |   4 +-
 arch/xtensa/mm/tlb.c                               |  37 +-
 .../platforms/xtfpga/include/platform/hardware.h   |   6 +-
 arch/xtensa/variants/s6000/include/variant/irq.h   |   1 -
 .../test_mmuhifi_c3/include/variant/core.h         | 384 +++++++++++++
 .../test_mmuhifi_c3/include/variant/tie-asm.h      | 183 +++++++
 .../variants/test_mmuhifi_c3/include/variant/tie.h | 141 +++++
 drivers/irqchip/Kconfig                            |   4 +
 drivers/irqchip/Makefile                           |   2 +
 drivers/irqchip/irq-xtensa-mx.c                    | 166 ++++++
 drivers/irqchip/irq-xtensa-pic.c                   | 108 ++++
 include/linux/irqchip/xtensa-mx.h                  |  17 +
 include/linux/irqchip/xtensa-pic.h                 |  18 +
 44 files changed, 2455 insertions(+), 389 deletions(-)
 create mode 100644 arch/xtensa/include/asm/mxregs.h
 create mode 100644 arch/xtensa/include/asm/spinlock_types.h
 create mode 100644 arch/xtensa/kernel/mxhead.S
 create mode 100644 arch/xtensa/kernel/smp.c
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h
 create mode 100644 drivers/irqchip/irq-xtensa-mx.c
 create mode 100644 drivers/irqchip/irq-xtensa-pic.c
 create mode 100644 include/linux/irqchip/xtensa-mx.h
 create mode 100644 include/linux/irqchip/xtensa-pic.h

-- 
1.8.1.4

[PATCH 01/17] xtensa: remove NO_IRQ definitions

[Max Filippov]

These definitions (-1) were correct when used as a hardware IRQ number,
but are incorrect as a mapped IRQ number.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/platforms/xtfpga/include/platform/hardware.h | 4 ----
 arch/xtensa/variants/s6000/include/variant/irq.h         | 1 -
 2 files changed, 5 deletions(-)

diff --git a/arch/xtensa/platforms/xtfpga/include/platform/hardware.h b/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
index 4416773..4b43ff1 100644
--- a/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
+++ b/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
@@ -15,10 +15,6 @@
 #ifndef __XTENSA_XTAVNET_HARDWARE_H
 #define __XTENSA_XTAVNET_HARDWARE_H
 
-/* By default NO_IRQ is defined to 0 in Linux, but we use the
-   interrupt 0 for UART... */
-#define NO_IRQ                 -1
-
 /* Memory configuration. */
 
 #define PLATFORM_DEFAULT_MEM_START 0x00000000
diff --git a/arch/xtensa/variants/s6000/include/variant/irq.h b/arch/xtensa/variants/s6000/include/variant/irq.h
index 97d6fc4..39ca751 100644
--- a/arch/xtensa/variants/s6000/include/variant/irq.h
+++ b/arch/xtensa/variants/s6000/include/variant/irq.h
@@ -1,7 +1,6 @@
 #ifndef _XTENSA_S6000_IRQ_H
 #define _XTENSA_S6000_IRQ_H
 
-#define NO_IRQ		(-1)
 #define VARIANT_NR_IRQS 8 /* GPIO interrupts */
 
 extern void variant_irq_enable(unsigned int irq);
-- 
1.8.1.4

[PATCH 02/17] xtensa: fix build warning in 64-bit build environment

[Max Filippov]

This fixes the following build warnings:

arch/xtensa/mm/misc.S: Assembler messages:
arch/xtensa/mm/misc.S:143: Warning: value 0xffffffff30000106 truncated to 0x30000106
arch/xtensa/mm/misc.S:197: Warning: value 0xffffffff30000106 truncated to 0x30000106

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/mm/misc.S | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/xtensa/mm/misc.S b/arch/xtensa/mm/misc.S
index d97ed1b..1f68558 100644
--- a/arch/xtensa/mm/misc.S
+++ b/arch/xtensa/mm/misc.S
@@ -140,7 +140,7 @@ ENTRY(clear_user_page)
 
 	/* Setup a temporary DTLB with the color of the VPN */
 
-	movi	a4, -PAGE_OFFSET + (PAGE_KERNEL | _PAGE_HW_WRITE)
+	movi	a4, ((PAGE_KERNEL | _PAGE_HW_WRITE) - PAGE_OFFSET) & 0xffffffff
 	movi	a5, TLBTEMP_BASE_1			# virt
 	add	a6, a2, a4				# ppn
 	add	a2, a5, a3				# add 'color'
@@ -194,7 +194,7 @@ ENTRY(copy_user_page)
 	or	a9, a9, a8
 	slli	a4, a4, PAGE_SHIFT
 	s32i	a9, a5, PAGE_FLAGS
-	movi	a5, -PAGE_OFFSET + (PAGE_KERNEL | _PAGE_HW_WRITE)
+	movi	a5, ((PAGE_KERNEL | _PAGE_HW_WRITE) - PAGE_OFFSET) & 0xffffffff
 
 	beqz	a6, 1f
 
-- 
1.8.1.4

[PATCH 03/17] xtensa: fix build warning from gcc-4.7.2

[Max Filippov]

This fixes the following build warning:

arch/xtensa/include/asm/vectors.h:70:5: warning: "XCHAL_HAVE_VECBASE" is not defined [-Wundef]

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/vectors.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/arch/xtensa/include/asm/vectors.h b/arch/xtensa/include/asm/vectors.h
index c52b656..6e0efa1 100644
--- a/arch/xtensa/include/asm/vectors.h
+++ b/arch/xtensa/include/asm/vectors.h
@@ -67,6 +67,10 @@
 						VECBASE_RESET_VADDR)
 #define RESET_VECTOR1_VADDR		XC_VADDR(RESET_VECTOR1_VECOFS)
 
+#ifndef XCHAL_HAVE_VECBASE
+#define XCHAL_HAVE_VECBASE 0
+#endif
+
 #if XCHAL_HAVE_VECBASE
 
 #define USER_VECTOR_VADDR		XC_VADDR(XCHAL_USER_VECOFS)
-- 
1.8.1.4

[PATCH 04/17] xtensa: fix arch spinlock function names

[Max Filippov]

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/spinlock.h       | 31 +++++++++++++++++--------------
 arch/xtensa/include/asm/spinlock_types.h | 20 ++++++++++++++++++++
 2 files changed, 37 insertions(+), 14 deletions(-)
 create mode 100644 arch/xtensa/include/asm/spinlock_types.h

diff --git a/arch/xtensa/include/asm/spinlock.h b/arch/xtensa/include/asm/spinlock.h
index 0397590..1d95fa5 100644
--- a/arch/xtensa/include/asm/spinlock.h
+++ b/arch/xtensa/include/asm/spinlock.h
@@ -28,13 +28,13 @@
  *    1         somebody owns the spinlock
  */
 
-#define __raw_spin_is_locked(x) ((x)->slock != 0)
-#define __raw_spin_unlock_wait(lock) \
-	do { while (__raw_spin_is_locked(lock)) cpu_relax(); } while (0)
+#define arch_spin_is_locked(x) ((x)->slock != 0)
+#define arch_spin_unlock_wait(lock) \
+	do { while (arch_spin_is_locked(lock)) cpu_relax(); } while (0)
 
-#define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
+#define arch_spin_lock_flags(lock, flags) arch_spin_lock(lock)
 
-static inline void __raw_spin_lock(raw_spinlock_t *lock)
+static inline void arch_spin_lock(arch_spinlock_t *lock)
 {
 	unsigned long tmp;
 
@@ -51,7 +51,7 @@ static inline void __raw_spin_lock(raw_spinlock_t *lock)
 
 /* Returns 1 if the lock is obtained, 0 otherwise. */
 
-static inline int __raw_spin_trylock(raw_spinlock_t *lock)
+static inline int arch_spin_trylock(arch_spinlock_t *lock)
 {
 	unsigned long tmp;
 
@@ -67,7 +67,7 @@ static inline int __raw_spin_trylock(raw_spinlock_t *lock)
 	return tmp == 0 ? 1 : 0;
 }
 
-static inline void __raw_spin_unlock(raw_spinlock_t *lock)
+static inline void arch_spin_unlock(arch_spinlock_t *lock)
 {
 	unsigned long tmp;
 
@@ -96,9 +96,9 @@ static inline void __raw_spin_unlock(raw_spinlock_t *lock)
  *  0x80000000  one writer owns the rwlock, no other writers, no readers
  */
 
-#define __raw_write_can_lock(x)  ((x)->lock == 0)
+#define arch_write_can_lock(x)  ((x)->lock == 0)
 
-static inline void __raw_write_lock(raw_rwlock_t *rw)
+static inline void arch_write_lock(arch_rwlock_t *rw)
 {
 	unsigned long tmp;
 
@@ -116,7 +116,7 @@ static inline void __raw_write_lock(raw_rwlock_t *rw)
 
 /* Returns 1 if the lock is obtained, 0 otherwise. */
 
-static inline int __raw_write_trylock(raw_rwlock_t *rw)
+static inline int arch_write_trylock(arch_rwlock_t *rw)
 {
 	unsigned long tmp;
 
@@ -133,7 +133,7 @@ static inline int __raw_write_trylock(raw_rwlock_t *rw)
 	return tmp == 0 ? 1 : 0;
 }
 
-static inline void __raw_write_unlock(raw_rwlock_t *rw)
+static inline void arch_write_unlock(arch_rwlock_t *rw)
 {
 	unsigned long tmp;
 
@@ -145,7 +145,7 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
 			: "memory");
 }
 
-static inline void __raw_read_lock(raw_rwlock_t *rw)
+static inline void arch_read_lock(arch_rwlock_t *rw)
 {
 	unsigned long tmp;
 	unsigned long result;
@@ -164,7 +164,7 @@ static inline void __raw_read_lock(raw_rwlock_t *rw)
 
 /* Returns 1 if the lock is obtained, 0 otherwise. */
 
-static inline int __raw_read_trylock(raw_rwlock_t *rw)
+static inline int arch_read_trylock(arch_rwlock_t *rw)
 {
 	unsigned long result;
 	unsigned long tmp;
@@ -184,7 +184,7 @@ static inline int __raw_read_trylock(raw_rwlock_t *rw)
 	return result == 0;
 }
 
-static inline void __raw_read_unlock(raw_rwlock_t *rw)
+static inline void arch_read_unlock(arch_rwlock_t *rw)
 {
 	unsigned long tmp1, tmp2;
 
@@ -199,4 +199,7 @@ static inline void __raw_read_unlock(raw_rwlock_t *rw)
 			: "memory");
 }
 
+#define arch_read_lock_flags(lock, flags)	arch_read_lock(lock)
+#define arch_write_lock_flags(lock, flags)	arch_write_lock(lock)
+
 #endif	/* _XTENSA_SPINLOCK_H */
diff --git a/arch/xtensa/include/asm/spinlock_types.h b/arch/xtensa/include/asm/spinlock_types.h
new file mode 100644
index 0000000..7ec5ce1
--- /dev/null
+++ b/arch/xtensa/include/asm/spinlock_types.h
@@ -0,0 +1,20 @@
+#ifndef __ASM_SPINLOCK_TYPES_H
+#define __ASM_SPINLOCK_TYPES_H
+
+#ifndef __LINUX_SPINLOCK_TYPES_H
+# error "please don't include this file directly"
+#endif
+
+typedef struct {
+	volatile unsigned int slock;
+} arch_spinlock_t;
+
+#define __ARCH_SPIN_LOCK_UNLOCKED	{ 0 }
+
+typedef struct {
+	volatile unsigned int lock;
+} arch_rwlock_t;
+
+#define __ARCH_RW_LOCK_UNLOCKED		{ 0 }
+
+#endif
-- 
1.8.1.4

[PATCH 05/17] xtensa: fix __delay for small loop count

[Max Filippov]

Avoid __delay counter underflow for loop counts < 2.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/delay.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/xtensa/include/asm/delay.h b/arch/xtensa/include/asm/delay.h
index 3899610..742b89f 100644
--- a/arch/xtensa/include/asm/delay.h
+++ b/arch/xtensa/include/asm/delay.h
@@ -19,9 +19,12 @@ extern unsigned long loops_per_jiffy;
 
 static inline void __delay(unsigned long loops)
 {
-	/* 2 cycles per loop. */
-	__asm__ __volatile__ ("1: addi %0, %0, -2; bgeui %0, 2, 1b"
-			      : "=r" (loops) : "0" (loops));
+	if (__builtin_constant_p(loops) && loops < 2)
+		__asm__ __volatile__ ("nop");
+	else if (loops >= 2)
+		/* 2 cycles per loop. */
+		__asm__ __volatile__ ("1: addi %0, %0, -2; bgeui %0, 2, 1b"
+				: "+r" (loops));
 }
 
 /* For SMP/NUMA systems, change boot_cpu_data to something like
-- 
1.8.1.4

[PATCH 06/17] xtensa: enable HAVE_IRQ_TIME_ACCOUNTING

[Max Filippov]

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index 8d24dcb..fb140ae 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -19,6 +19,7 @@ config XTENSA
 	select IRQ_DOMAIN
 	select HAVE_OPROFILE
 	select HAVE_FUNCTION_TRACER
+	select HAVE_IRQ_TIME_ACCOUNTING
 	help
 	  Xtensa processors are 32-bit RISC machines designed by Tensilica
 	  primarily for embedded systems.  These processors are both
-- 
1.8.1.4

[PATCH 07/17] xtensa: mark ccount as continuous clocksource

[Max Filippov]

From: Baruch Siach <baruch@tkos.co.il>

This allows ccount to be used as highres timer.

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/kernel/time.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c
index 9af3dd8..26eb6a9 100644
--- a/arch/xtensa/kernel/time.c
+++ b/arch/xtensa/kernel/time.c
@@ -46,6 +46,7 @@ static struct clocksource ccount_clocksource = {
 	.rating = 200,
 	.read = ccount_read,
 	.mask = CLOCKSOURCE_MASK(32),
+	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
 static int ccount_timer_set_next_event(unsigned long delta,
-- 
1.8.1.4

[PATCH 08/17] xtensa: update clockevent setup for SMP

[Max Filippov]

Provide per-cpu ccount_timer objects and use them appropriately.
Extract per-cpu clockevent setup function.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/kernel/time.c | 54 +++++++++++++++++++++++++----------------------
 1 file changed, 29 insertions(+), 25 deletions(-)

diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c
index 26eb6a9..4ce7aae 100644
--- a/arch/xtensa/kernel/time.c
+++ b/arch/xtensa/kernel/time.c
@@ -53,18 +53,12 @@ static int ccount_timer_set_next_event(unsigned long delta,
 		struct clock_event_device *dev);
 static void ccount_timer_set_mode(enum clock_event_mode mode,
 		struct clock_event_device *evt);
-static struct ccount_timer_t {
+struct ccount_timer {
 	struct clock_event_device evt;
 	int irq_enabled;
-} ccount_timer = {
-	.evt = {
-		.name		= "ccount_clockevent",
-		.features	= CLOCK_EVT_FEAT_ONESHOT,
-		.rating		= 300,
-		.set_next_event	= ccount_timer_set_next_event,
-		.set_mode	= ccount_timer_set_mode,
-	},
+	char name[24];
 };
+static DEFINE_PER_CPU(struct ccount_timer, ccount_timer);
 
 static int ccount_timer_set_next_event(unsigned long delta,
 		struct clock_event_device *dev)
@@ -85,8 +79,8 @@ static int ccount_timer_set_next_event(unsigned long delta,
 static void ccount_timer_set_mode(enum clock_event_mode mode,
 		struct clock_event_device *evt)
 {
-	struct ccount_timer_t *timer =
-		container_of(evt, struct ccount_timer_t, evt);
+	struct ccount_timer *timer =
+		container_of(evt, struct ccount_timer, evt);
 
 	/*
 	 * There is no way to disable the timer interrupt at the device level,
@@ -118,9 +112,28 @@ static struct irqaction timer_irqaction = {
 	.handler =	timer_interrupt,
 	.flags =	IRQF_TIMER,
 	.name =		"timer",
-	.dev_id =	&ccount_timer,
 };
 
+void local_timer_setup(unsigned cpu)
+{
+	struct ccount_timer *timer = &per_cpu(ccount_timer, cpu);
+	struct clock_event_device *clockevent = &timer->evt;
+
+	timer->irq_enabled = 1;
+	clockevent->name = timer->name;
+	snprintf(timer->name, sizeof(timer->name), "ccount_clockevent_%u", cpu);
+	clockevent->features = CLOCK_EVT_FEAT_ONESHOT;
+	clockevent->rating = 300;
+	clockevent->set_next_event = ccount_timer_set_next_event;
+	clockevent->set_mode = ccount_timer_set_mode;
+	clockevent->cpumask = cpumask_of(cpu);
+	clockevent->irq = irq_create_mapping(NULL, LINUX_TIMER_INT);
+	if (WARN(!clockevent->irq, "error: can't map timer irq"))
+		return;
+	clockevents_config_and_register(clockevent, ccount_freq,
+					0xf, 0xffffffff);
+}
+
 void __init time_init(void)
 {
 #ifdef CONFIG_XTENSA_CALIBRATE_CCOUNT
@@ -132,16 +145,8 @@ void __init time_init(void)
 	ccount_freq = CONFIG_XTENSA_CPU_CLOCK*1000000UL;
 #endif
 	clocksource_register_hz(&ccount_clocksource, ccount_freq);
-
-	ccount_timer.evt.cpumask = cpumask_of(0);
-	ccount_timer.evt.irq = irq_create_mapping(NULL, LINUX_TIMER_INT);
-	if (WARN(!ccount_timer.evt.irq, "error: can't map timer irq"))
-		return;
-	clockevents_config_and_register(&ccount_timer.evt, ccount_freq, 0xf,
-			0xffffffff);
-	setup_irq(ccount_timer.evt.irq, &timer_irqaction);
-	ccount_timer.irq_enabled = 1;
-
+	local_timer_setup(0);
+	setup_irq(this_cpu_ptr(&ccount_timer)->evt.irq, &timer_irqaction);
 	setup_sched_clock(ccount_sched_clock_read, 32, ccount_freq);
 }
 
@@ -149,10 +154,9 @@ void __init time_init(void)
  * The timer interrupt is called HZ times per second.
  */
 
-irqreturn_t timer_interrupt (int irq, void *dev_id)
+irqreturn_t timer_interrupt(int irq, void *dev_id)
 {
-	struct ccount_timer_t *timer = dev_id;
-	struct clock_event_device *evt = &timer->evt;
+	struct clock_event_device *evt = &this_cpu_ptr(&ccount_timer)->evt;
 
 	evt->event_handler(evt);
 
-- 
1.8.1.4

[PATCH 09/17] xtensa: call check_s32c1i after trap_init

[Max Filippov]

Otherwise exceptions may occur prior to exception handling mechanism
initialization, resulting in silently dead system.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/kernel/setup.c | 18 +++++++-----------
 1 file changed, 7 insertions(+), 11 deletions(-)

diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
index 946fb8d..00904b6 100644
--- a/arch/xtensa/kernel/setup.c
+++ b/arch/xtensa/kernel/setup.c
@@ -378,7 +378,8 @@ static inline int probed_compare_swap(int *v, int cmp, int set)
 
 /* Handle probed exception */
 
-void __init do_probed_exception(struct pt_regs *regs, unsigned long exccause)
+static void __init do_probed_exception(struct pt_regs *regs,
+		unsigned long exccause)
 {
 	if (regs->pc == rcw_probe_pc) {	/* exception on s32c1i ? */
 		regs->pc += 3;		/* skip the s32c1i instruction */
@@ -390,7 +391,7 @@ void __init do_probed_exception(struct pt_regs *regs, unsigned long exccause)
 
 /* Simple test of S32C1I (soc bringup assist) */
 
-void __init check_s32c1i(void)
+static int __init check_s32c1i(void)
 {
 	int n, cause1, cause2;
 	void *handbus, *handdata, *handaddr; /* temporarily saved handlers */
@@ -445,24 +446,21 @@ void __init check_s32c1i(void)
 	trap_set_handler(EXCCAUSE_LOAD_STORE_ERROR, handbus);
 	trap_set_handler(EXCCAUSE_LOAD_STORE_DATA_ERROR, handdata);
 	trap_set_handler(EXCCAUSE_LOAD_STORE_ADDR_ERROR, handaddr);
+	return 0;
 }
 
 #else /* XCHAL_HAVE_S32C1I */
 
 /* This condition should not occur with a commercially deployed processor.
    Display reminder for early engr test or demo chips / FPGA bitstreams */
-void __init check_s32c1i(void)
+static int __init check_s32c1i(void)
 {
 	pr_warn("Processor configuration lacks atomic compare-and-swap support!\n");
+	return 0;
 }
 
 #endif /* XCHAL_HAVE_S32C1I */
-#else /* CONFIG_S32C1I_SELFTEST */
-
-void __init check_s32c1i(void)
-{
-}
-
+early_initcall(check_s32c1i);
 #endif /* CONFIG_S32C1I_SELFTEST */
 
 
@@ -471,8 +469,6 @@ void __init setup_arch(char **cmdline_p)
 	strlcpy(boot_command_line, command_line, COMMAND_LINE_SIZE);
 	*cmdline_p = command_line;
 
-	check_s32c1i();
-
 	/* Reserve some memory regions */
 
 #ifdef CONFIG_BLK_DEV_INITRD
-- 
1.8.1.4

[PATCH 10/17] xtensa: move init_mmu declaration to mmu_context.h

[Max Filippov]

Secondary CPUs need this declaration to initialize their MMUs.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/mmu_context.h | 6 ++++++
 arch/xtensa/kernel/setup.c            | 7 +------
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/arch/xtensa/include/asm/mmu_context.h b/arch/xtensa/include/asm/mmu_context.h
index d43525a..86292c2 100644
--- a/arch/xtensa/include/asm/mmu_context.h
+++ b/arch/xtensa/include/asm/mmu_context.h
@@ -49,6 +49,12 @@ extern unsigned long asid_cache;
 #define ASID_MASK	((1 << XCHAL_MMU_ASID_BITS) - 1)
 #define ASID_INSERT(x)	(0x03020001 | (((x) & ASID_MASK) << 8))
 
+#ifdef CONFIG_MMU
+void init_mmu(void);
+#else
+static inline void init_mmu(void) { }
+#endif
+
 static inline void set_rasid_register (unsigned long val)
 {
 	__asm__ __volatile__ (" wsr %0, rasid\n\t"
diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
index 00904b6..84b26ee 100644
--- a/arch/xtensa/kernel/setup.c
+++ b/arch/xtensa/kernel/setup.c
@@ -40,6 +40,7 @@
 #endif
 
 #include <asm/bootparam.h>
+#include <asm/mmu_context.h>
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/timex.h>
@@ -88,12 +89,6 @@ static char default_command_line[COMMAND_LINE_SIZE] __initdata = CONFIG_CMDLINE;
 
 sysmem_info_t __initdata sysmem;
 
-#ifdef CONFIG_MMU
-extern void init_mmu(void);
-#else
-static inline void init_mmu(void) { }
-#endif
-
 extern int mem_reserve(unsigned long, unsigned long, int);
 extern void bootmem_init(void);
 extern void zones_init(void);
-- 
1.8.1.4

[PATCH 11/17] xtensa: move built-in PIC to drivers/irqchip

[Max Filippov]

Extract xtensa built-in interrupt controller implementation from
xtensa/kernel/irq.c and move it to other irqchips, providing way to
instantiate it from the device tree.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/irq.h      |   7 ++
 arch/xtensa/kernel/irq.c           | 137 ++++++++-----------------------------
 drivers/irqchip/Makefile           |   1 +
 drivers/irqchip/irq-xtensa-pic.c   | 108 +++++++++++++++++++++++++++++
 include/linux/irqchip/xtensa-pic.h |  18 +++++
 5 files changed, 164 insertions(+), 107 deletions(-)
 create mode 100644 drivers/irqchip/irq-xtensa-pic.c
 create mode 100644 include/linux/irqchip/xtensa-pic.h

diff --git a/arch/xtensa/include/asm/irq.h b/arch/xtensa/include/asm/irq.h
index 4c0ccc9..16464f2 100644
--- a/arch/xtensa/include/asm/irq.h
+++ b/arch/xtensa/include/asm/irq.h
@@ -43,5 +43,12 @@ static __inline__ int irq_canonicalize(int irq)
 }
 
 struct irqaction;
+struct irq_domain;
+
+int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize,
+		unsigned long int_irq, unsigned long ext_irq,
+		unsigned long *out_hwirq, unsigned int *out_type);
+int xtensa_irq_map(struct irq_domain *d, unsigned int irq, irq_hw_number_t hw);
+unsigned xtensa_map_ext_irq(unsigned ext_irq);
 
 #endif	/* _XTENSA_IRQ_H */
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index 6f4f974..ada1e48 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -18,28 +18,20 @@
 #include <linux/interrupt.h>
 #include <linux/irq.h>
 #include <linux/kernel_stat.h>
+#include <linux/irqchip.h>
+#include <linux/irqchip/xtensa-pic.h>
 #include <linux/irqdomain.h>
 #include <linux/of.h>
 
 #include <asm/uaccess.h>
 #include <asm/platform.h>
 
-static unsigned int cached_irq_mask;
-
 atomic_t irq_err_count;
 
-static struct irq_domain *root_domain;
-
-/*
- * do_IRQ handles all normal device IRQ's (the special
- * SMP cross-CPU interrupts have their own specific
- * handlers).
- */
-
 asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs)
 {
 	struct pt_regs *old_regs = set_irq_regs(regs);
-	int irq = irq_find_mapping(root_domain, hwirq);
+	int irq = irq_find_mapping(NULL, hwirq);
 
 	if (hwirq >= NR_IRQS) {
 		printk(KERN_EMERG "%s: cannot handle IRQ %d\n",
@@ -74,83 +66,57 @@ int arch_show_interrupts(struct seq_file *p, int prec)
 	return 0;
 }
 
-static void xtensa_irq_mask(struct irq_data *d)
-{
-	cached_irq_mask &= ~(1 << d->hwirq);
-	set_sr (cached_irq_mask, intenable);
-}
-
-static void xtensa_irq_unmask(struct irq_data *d)
-{
-	cached_irq_mask |= 1 << d->hwirq;
-	set_sr (cached_irq_mask, intenable);
-}
-
-static void xtensa_irq_enable(struct irq_data *d)
-{
-	variant_irq_enable(d->hwirq);
-	xtensa_irq_unmask(d);
-}
-
-static void xtensa_irq_disable(struct irq_data *d)
-{
-	xtensa_irq_mask(d);
-	variant_irq_disable(d->hwirq);
-}
-
-static void xtensa_irq_ack(struct irq_data *d)
-{
-	set_sr(1 << d->hwirq, intclear);
-}
-
-static int xtensa_irq_retrigger(struct irq_data *d)
+int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize,
+		unsigned long int_irq, unsigned long ext_irq,
+		unsigned long *out_hwirq, unsigned int *out_type)
 {
-	set_sr(1 << d->hwirq, intset);
-	return 1;
+	if (WARN_ON(intsize < 1 || intsize > 2))
+		return -EINVAL;
+	if (intsize == 2 && intspec[1] == 1) {
+		int_irq = xtensa_map_ext_irq(ext_irq);
+		if (int_irq < XCHAL_NUM_INTERRUPTS)
+			*out_hwirq = int_irq;
+		else
+			return -EINVAL;
+	} else {
+		*out_hwirq = int_irq;
+	}
+	*out_type = IRQ_TYPE_NONE;
+	return 0;
 }
 
-static struct irq_chip xtensa_irq_chip = {
-	.name		= "xtensa",
-	.irq_enable	= xtensa_irq_enable,
-	.irq_disable	= xtensa_irq_disable,
-	.irq_mask	= xtensa_irq_mask,
-	.irq_unmask	= xtensa_irq_unmask,
-	.irq_ack	= xtensa_irq_ack,
-	.irq_retrigger	= xtensa_irq_retrigger,
-};
-
-static int xtensa_irq_map(struct irq_domain *d, unsigned int irq,
+int xtensa_irq_map(struct irq_domain *d, unsigned int irq,
 		irq_hw_number_t hw)
 {
+	struct irq_chip *irq_chip = d->host_data;
 	u32 mask = 1 << hw;
 
 	if (mask & XCHAL_INTTYPE_MASK_SOFTWARE) {
-		irq_set_chip_and_handler_name(irq, &xtensa_irq_chip,
+		irq_set_chip_and_handler_name(irq, irq_chip,
 				handle_simple_irq, "level");
 		irq_set_status_flags(irq, IRQ_LEVEL);
 	} else if (mask & XCHAL_INTTYPE_MASK_EXTERN_EDGE) {
-		irq_set_chip_and_handler_name(irq, &xtensa_irq_chip,
+		irq_set_chip_and_handler_name(irq, irq_chip,
 				handle_edge_irq, "edge");
 		irq_clear_status_flags(irq, IRQ_LEVEL);
 	} else if (mask & XCHAL_INTTYPE_MASK_EXTERN_LEVEL) {
-		irq_set_chip_and_handler_name(irq, &xtensa_irq_chip,
+		irq_set_chip_and_handler_name(irq, irq_chip,
 				handle_level_irq, "level");
 		irq_set_status_flags(irq, IRQ_LEVEL);
 	} else if (mask & XCHAL_INTTYPE_MASK_TIMER) {
-		irq_set_chip_and_handler_name(irq, &xtensa_irq_chip,
-				handle_edge_irq, "edge");
+		irq_set_chip_and_handler_name(irq, irq_chip,
+				handle_percpu_irq, "timer");
 		irq_clear_status_flags(irq, IRQ_LEVEL);
 	} else {/* XCHAL_INTTYPE_MASK_WRITE_ERROR */
 		/* XCHAL_INTTYPE_MASK_NMI */
-
-		irq_set_chip_and_handler_name(irq, &xtensa_irq_chip,
+		irq_set_chip_and_handler_name(irq, irq_chip,
 				handle_level_irq, "level");
 		irq_set_status_flags(irq, IRQ_LEVEL);
 	}
 	return 0;
 }
 
-static unsigned map_ext_irq(unsigned ext_irq)
+unsigned xtensa_map_ext_irq(unsigned ext_irq)
 {
 	unsigned mask = XCHAL_INTTYPE_MASK_EXTERN_EDGE |
 		XCHAL_INTTYPE_MASK_EXTERN_LEVEL;
@@ -163,55 +129,12 @@ static unsigned map_ext_irq(unsigned ext_irq)
 	return XCHAL_NUM_INTERRUPTS;
 }
 
-/*
- * Device Tree IRQ specifier translation function which works with one or
- * two cell bindings. First cell value maps directly to the hwirq number.
- * Second cell if present specifies whether hwirq number is external (1) or
- * internal (0).
- */
-int xtensa_irq_domain_xlate(struct irq_domain *d, struct device_node *ctrlr,
-		const u32 *intspec, unsigned int intsize,
-		unsigned long *out_hwirq, unsigned int *out_type)
-{
-	if (WARN_ON(intsize < 1 || intsize > 2))
-		return -EINVAL;
-	if (intsize == 2 && intspec[1] == 1) {
-		unsigned int_irq = map_ext_irq(intspec[0]);
-		if (int_irq < XCHAL_NUM_INTERRUPTS)
-			*out_hwirq = int_irq;
-		else
-			return -EINVAL;
-	} else {
-		*out_hwirq = intspec[0];
-	}
-	*out_type = IRQ_TYPE_NONE;
-	return 0;
-}
-
-static const struct irq_domain_ops xtensa_irq_domain_ops = {
-	.xlate = xtensa_irq_domain_xlate,
-	.map = xtensa_irq_map,
-};
-
 void __init init_IRQ(void)
 {
-	struct device_node *intc = NULL;
-
-	cached_irq_mask = 0;
-	set_sr(~0, intclear);
-
 #ifdef CONFIG_OF
-	/* The interrupt controller device node is mandatory */
-	intc = of_find_compatible_node(NULL, NULL, "xtensa,pic");
-	BUG_ON(!intc);
-
-	root_domain = irq_domain_add_linear(intc, NR_IRQS,
-			&xtensa_irq_domain_ops, NULL);
+	irqchip_init();
 #else
-	root_domain = irq_domain_add_legacy(intc, NR_IRQS, 0, 0,
-			&xtensa_irq_domain_ops, NULL);
+	xtensa_pic_init_legacy(NULL);
 #endif
-	irq_set_default_host(root_domain);
-
 	variant_init_irq();
 }
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index c60b901..c81a7f3 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -22,3 +22,4 @@ obj-$(CONFIG_RENESAS_IRQC)		+= irq-renesas-irqc.o
 obj-$(CONFIG_VERSATILE_FPGA_IRQ)	+= irq-versatile-fpga.o
 obj-$(CONFIG_ARCH_VT8500)		+= irq-vt8500.o
 obj-$(CONFIG_TB10X_IRQC)		+= irq-tb10x.o
+obj-$(CONFIG_XTENSA)			+= irq-xtensa-pic.o
diff --git a/drivers/irqchip/irq-xtensa-pic.c b/drivers/irqchip/irq-xtensa-pic.c
new file mode 100644
index 0000000..846d181
--- /dev/null
+++ b/drivers/irqchip/irq-xtensa-pic.c
@@ -0,0 +1,108 @@
+/*
+ * Xtensa built-in interrupt controller
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
+ *
+ * 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.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Kevin Chea
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irqdomain.h>
+#include <linux/irq.h>
+#include <linux/of.h>
+
+#include "irqchip.h"
+
+unsigned int cached_irq_mask;
+
+/*
+ * Device Tree IRQ specifier translation function which works with one or
+ * two cell bindings. First cell value maps directly to the hwirq number.
+ * Second cell if present specifies whether hwirq number is external (1) or
+ * internal (0).
+ */
+static int xtensa_pic_irq_domain_xlate(struct irq_domain *d,
+		struct device_node *ctrlr,
+		const u32 *intspec, unsigned int intsize,
+		unsigned long *out_hwirq, unsigned int *out_type)
+{
+	return xtensa_irq_domain_xlate(intspec, intsize,
+			intspec[0], intspec[0],
+			out_hwirq, out_type);
+}
+
+static const struct irq_domain_ops xtensa_irq_domain_ops = {
+	.xlate = xtensa_pic_irq_domain_xlate,
+	.map = xtensa_irq_map,
+};
+
+static void xtensa_irq_mask(struct irq_data *d)
+{
+	cached_irq_mask &= ~(1 << d->hwirq);
+	set_sr(cached_irq_mask, intenable);
+}
+
+static void xtensa_irq_unmask(struct irq_data *d)
+{
+	cached_irq_mask |= 1 << d->hwirq;
+	set_sr(cached_irq_mask, intenable);
+}
+
+static void xtensa_irq_enable(struct irq_data *d)
+{
+	variant_irq_enable(d->hwirq);
+	xtensa_irq_unmask(d);
+}
+
+static void xtensa_irq_disable(struct irq_data *d)
+{
+	xtensa_irq_mask(d);
+	variant_irq_disable(d->hwirq);
+}
+
+static void xtensa_irq_ack(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intclear);
+}
+
+static int xtensa_irq_retrigger(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intset);
+	return 1;
+}
+
+static struct irq_chip xtensa_irq_chip = {
+	.name		= "xtensa",
+	.irq_enable	= xtensa_irq_enable,
+	.irq_disable	= xtensa_irq_disable,
+	.irq_mask	= xtensa_irq_mask,
+	.irq_unmask	= xtensa_irq_unmask,
+	.irq_ack	= xtensa_irq_ack,
+	.irq_retrigger	= xtensa_irq_retrigger,
+};
+
+int __init xtensa_pic_init_legacy(struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_legacy(NULL, NR_IRQS, 0, 0,
+				&xtensa_irq_domain_ops, &xtensa_irq_chip);
+	irq_set_default_host(root_domain);
+	return 0;
+}
+
+static int __init xtensa_pic_init(struct device_node *np,
+		struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_linear(np, NR_IRQS, &xtensa_irq_domain_ops,
+				&xtensa_irq_chip);
+	irq_set_default_host(root_domain);
+	return 0;
+}
+IRQCHIP_DECLARE(xtensa_irq_chip, "xtensa,pic", xtensa_pic_init);
diff --git a/include/linux/irqchip/xtensa-pic.h b/include/linux/irqchip/xtensa-pic.h
new file mode 100644
index 0000000..48718ae
--- /dev/null
+++ b/include/linux/irqchip/xtensa-pic.h
@@ -0,0 +1,18 @@
+/*
+ * Xtensa built-in interrupt controller
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
+ *
+ * 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.
+ */
+
+#ifndef __LINUX_IRQCHIP_XTENSA_PIC_H
+#define __LINUX_IRQCHIP_XTENSA_PIC_H
+
+struct device_node;
+int xtensa_pic_init_legacy(struct device_node *interrupt_parent);
+
+#endif /* __LINUX_IRQCHIP_XTENSA_PIC_H */
-- 
1.8.1.4

[PATCH 12/17] xtensa: clean up do_interrupt/do_IRQ

[Max Filippov]

- set up irq registers and call irq_enter/irq_exit once for each kernel
  entry due to interrupt;
- don't attempt to clear current IRQ in the do_interrupt, IRQ handler
  will take care of it;
- find pending interrupt with highest priority before every ISR
  invocation.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/kernel/irq.c   |  6 ------
 arch/xtensa/kernel/traps.c | 23 +++++++++--------------
 2 files changed, 9 insertions(+), 20 deletions(-)

diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index ada1e48..3cef58e 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -30,7 +30,6 @@ atomic_t irq_err_count;
 
 asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs)
 {
-	struct pt_regs *old_regs = set_irq_regs(regs);
 	int irq = irq_find_mapping(NULL, hwirq);
 
 	if (hwirq >= NR_IRQS) {
@@ -38,8 +37,6 @@ asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs)
 				__func__, hwirq);
 	}
 
-	irq_enter();
-
 #ifdef CONFIG_DEBUG_STACKOVERFLOW
 	/* Debugging check for stack overflow: is there less than 1KB free? */
 	{
@@ -54,9 +51,6 @@ asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs)
 	}
 #endif
 	generic_handle_irq(irq);
-
-	irq_exit();
-	set_irq_regs(old_regs);
 }
 
 int arch_show_interrupts(struct seq_file *p, int prec)
diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c
index 3e8a05c..3dbe864 100644
--- a/arch/xtensa/kernel/traps.c
+++ b/arch/xtensa/kernel/traps.c
@@ -212,6 +212,9 @@ void do_interrupt(struct pt_regs *regs)
 		XCHAL_INTLEVEL6_MASK,
 		XCHAL_INTLEVEL7_MASK,
 	};
+	struct pt_regs *old_regs = set_irq_regs(regs);
+
+	irq_enter();
 
 	for (;;) {
 		unsigned intread = get_sr(interrupt);
@@ -227,21 +230,13 @@ void do_interrupt(struct pt_regs *regs)
 		}
 
 		if (level == 0)
-			return;
-
-		/*
-		 * Clear the interrupt before processing, in case it's
-		 *  edge-triggered or software-generated
-		 */
-		while (int_at_level) {
-			unsigned i = __ffs(int_at_level);
-			unsigned mask = 1 << i;
-
-			int_at_level ^= mask;
-			set_sr(mask, intclear);
-			do_IRQ(i, regs);
-		}
+			break;
+
+		do_IRQ(__ffs(int_at_level), regs);
 	}
+
+	irq_exit();
+	set_irq_regs(old_regs);
 }
 
 /*
-- 
1.8.1.4

[PATCH 13/17] xtensa: clear timer IRQ unconditionally in its handler

[Max Filippov]

PIC irq_ack doesn't clear timer IRQ, because timer interrupt handler
usually set up new timer by writing to ccompare register and thus
clearing timer IRQ. However timer may not be set up in the IRQ handler,
e.g. with tickless idle on SMP, or when CPU is going offline, leaving
timer IRQ raised and making do_interrupt attempting to handle it
forever.

To fix this always write current value of ccompare SR chosen to be linux
timer back to that SR on entry to timer interrupt handler.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/kernel/time.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c
index 4ce7aae..60dcb3f 100644
--- a/arch/xtensa/kernel/time.c
+++ b/arch/xtensa/kernel/time.c
@@ -158,6 +158,7 @@ irqreturn_t timer_interrupt(int irq, void *dev_id)
 {
 	struct clock_event_device *evt = &this_cpu_ptr(&ccount_timer)->evt;
 
+	set_linux_timer(get_linux_timer());
 	evt->event_handler(evt);
 
 	/* Allow platform to do something useful (Wdog). */
-- 
1.8.1.4

[PATCH 14/17] xtensa: add MX irqchip

[Max Filippov]

MX is an interrupt distributor used in some SMP-capable xtensa
configurations.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/include/asm/irq.h       |   1 +
 arch/xtensa/include/asm/mxregs.h    |  46 ++++++++++
 arch/xtensa/include/asm/processor.h |  20 +++++
 arch/xtensa/kernel/irq.c            |   8 ++
 drivers/irqchip/Kconfig             |   4 +
 drivers/irqchip/Makefile            |   1 +
 drivers/irqchip/irq-xtensa-mx.c     | 166 ++++++++++++++++++++++++++++++++++++
 include/linux/irqchip/xtensa-mx.h   |  17 ++++
 8 files changed, 263 insertions(+)
 create mode 100644 arch/xtensa/include/asm/mxregs.h
 create mode 100644 drivers/irqchip/irq-xtensa-mx.c
 create mode 100644 include/linux/irqchip/xtensa-mx.h

diff --git a/arch/xtensa/include/asm/irq.h b/arch/xtensa/include/asm/irq.h
index 16464f2..7d194d4 100644
--- a/arch/xtensa/include/asm/irq.h
+++ b/arch/xtensa/include/asm/irq.h
@@ -50,5 +50,6 @@ int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize,
 		unsigned long *out_hwirq, unsigned int *out_type);
 int xtensa_irq_map(struct irq_domain *d, unsigned int irq, irq_hw_number_t hw);
 unsigned xtensa_map_ext_irq(unsigned ext_irq);
+unsigned xtensa_get_ext_irq_no(unsigned irq);
 
 #endif	/* _XTENSA_IRQ_H */
diff --git a/arch/xtensa/include/asm/mxregs.h b/arch/xtensa/include/asm/mxregs.h
new file mode 100644
index 0000000..73dcc54
--- /dev/null
+++ b/arch/xtensa/include/asm/mxregs.h
@@ -0,0 +1,46 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * 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) 2008 - 2013 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_MXREGS_H
+#define _XTENSA_MXREGS_H
+
+/*
+ * RER/WER at, as	Read/write external register
+ *	at: value
+ *	as: address
+ *
+ * Address	Value
+ * 00nn		0...0p..p	Interrupt Routing, route IRQ n to processor p
+ * 01pp		0...0d..d	16 bits (d) 'ored' as single IPI to processor p
+ * 0180		0...0m..m	Clear enable specified by mask (m)
+ * 0184		0...0m..m	Set enable specified by mask (m)
+ * 0190		0...0x..x	8-bit IPI partition register
+ *				VVVVVVVVPPPPUUUUUUUUUUUUUUUUU
+ *				V (10-bit) Release/Version
+ *				P ( 4-bit) Number of cores - 1
+ *				U (18-bit) ID
+ * 01a0		i.......i	32-bit ConfigID
+ * 0200		0...0m..m	RunStall core 'n'
+ * 0220		c		Cache coherency enabled
+ */
+
+#define MIROUT(irq)	(0x000 + (irq))
+#define MIPICAUSE(cpu)	(0x100 + (cpu))
+#define MIPISET(cause)	(0x140 + (cause))
+#define MIENG		0x180
+#define MIENGSET	0x184
+#define MIASG		0x188	/* Read Global Assert Register */
+#define MIASGSET	0x18c	/* Set Global Addert Regiter */
+#define MIPIPART	0x190
+#define SYSCFGID	0x1a0
+#define MPSCORE		0x200
+#define CCON		0x220
+
+#endif /* _XTENSA_MXREGS_H */
diff --git a/arch/xtensa/include/asm/processor.h b/arch/xtensa/include/asm/processor.h
index 7e409a5..abb5970 100644
--- a/arch/xtensa/include/asm/processor.h
+++ b/arch/xtensa/include/asm/processor.h
@@ -191,5 +191,25 @@ extern unsigned long get_wchan(struct task_struct *p);
 #define set_sr(x,sr) ({unsigned int v=(unsigned int)x; WSR(v,sr);})
 #define get_sr(sr) ({unsigned int v; RSR(v,sr); v; })
 
+#ifndef XCHAL_HAVE_EXTERN_REGS
+#define XCHAL_HAVE_EXTERN_REGS 0
+#endif
+
+#if XCHAL_HAVE_EXTERN_REGS
+
+static inline void set_er(unsigned long value, unsigned long addr)
+{
+	asm volatile ("wer %0, %1" : : "a" (value), "a" (addr) : "memory");
+}
+
+static inline unsigned long get_er(unsigned long addr)
+{
+	register unsigned long value;
+	asm volatile ("rer %0, %1" : "=a" (value) : "a" (addr) : "memory");
+	return value;
+}
+
+#endif /* XCHAL_HAVE_EXTERN_REGS */
+
 #endif	/* __ASSEMBLY__ */
 #endif	/* _XTENSA_PROCESSOR_H */
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index 3cef58e..7d49730 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -123,6 +123,14 @@ unsigned xtensa_map_ext_irq(unsigned ext_irq)
 	return XCHAL_NUM_INTERRUPTS;
 }
 
+unsigned xtensa_get_ext_irq_no(unsigned irq)
+{
+	unsigned mask = (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+		XCHAL_INTTYPE_MASK_EXTERN_LEVEL) &
+		((1u << irq) - 1);
+	return hweight32(mask);
+}
+
 void __init init_IRQ(void)
 {
 #ifdef CONFIG_OF
diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
index 3792a1a..07bc79c 100644
--- a/drivers/irqchip/Kconfig
+++ b/drivers/irqchip/Kconfig
@@ -61,3 +61,7 @@ config VERSATILE_FPGA_IRQ_NR
        int
        default 4
        depends on VERSATILE_FPGA_IRQ
+
+config XTENSA_MX
+	bool
+	select IRQ_DOMAIN
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index c81a7f3..66913f9 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -23,3 +23,4 @@ obj-$(CONFIG_VERSATILE_FPGA_IRQ)	+= irq-versatile-fpga.o
 obj-$(CONFIG_ARCH_VT8500)		+= irq-vt8500.o
 obj-$(CONFIG_TB10X_IRQC)		+= irq-tb10x.o
 obj-$(CONFIG_XTENSA)			+= irq-xtensa-pic.o
+obj-$(CONFIG_XTENSA_MX)			+= irq-xtensa-mx.o
diff --git a/drivers/irqchip/irq-xtensa-mx.c b/drivers/irqchip/irq-xtensa-mx.c
new file mode 100644
index 0000000..86f91f0
--- /dev/null
+++ b/drivers/irqchip/irq-xtensa-mx.c
@@ -0,0 +1,166 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ *
+ * 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.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irqdomain.h>
+#include <linux/irq.h>
+#include <linux/of.h>
+
+#include <asm/mxregs.h>
+
+#include "irqchip.h"
+
+#define HW_IRQ_IPI_COUNT 2
+#define HW_IRQ_MX_BASE 2
+#define HW_IRQ_EXTERN_BASE 3
+
+static DEFINE_PER_CPU(unsigned int, cached_irq_mask);
+
+static int xtensa_mx_irq_map(struct irq_domain *d, unsigned int irq,
+		irq_hw_number_t hw)
+{
+	if (hw < HW_IRQ_IPI_COUNT) {
+		struct irq_chip *irq_chip = d->host_data;
+		irq_set_chip_and_handler_name(irq, irq_chip,
+				handle_percpu_irq, "ipi");
+		irq_set_status_flags(irq, IRQ_LEVEL);
+		return 0;
+	}
+	return xtensa_irq_map(d, irq, hw);
+}
+
+/*
+ * Device Tree IRQ specifier translation function which works with one or
+ * two cell bindings. First cell value maps directly to the hwirq number.
+ * Second cell if present specifies whether hwirq number is external (1) or
+ * internal (0).
+ */
+static int xtensa_mx_irq_domain_xlate(struct irq_domain *d,
+		struct device_node *ctrlr,
+		const u32 *intspec, unsigned int intsize,
+		unsigned long *out_hwirq, unsigned int *out_type)
+{
+	return xtensa_irq_domain_xlate(intspec, intsize,
+			intspec[0], intspec[0] + HW_IRQ_EXTERN_BASE,
+			out_hwirq, out_type);
+}
+
+static const struct irq_domain_ops xtensa_mx_irq_domain_ops = {
+	.xlate = xtensa_mx_irq_domain_xlate,
+	.map = xtensa_mx_irq_map,
+};
+
+void secondary_init_irq(void)
+{
+	this_cpu_write(cached_irq_mask,
+			XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+			XCHAL_INTTYPE_MASK_EXTERN_LEVEL);
+	set_sr(XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+			XCHAL_INTTYPE_MASK_EXTERN_LEVEL, intenable);
+}
+
+static void xtensa_mx_irq_mask(struct irq_data *d)
+{
+	unsigned long mask = 1u << d->hwirq;
+
+	if (mask & (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+				XCHAL_INTTYPE_MASK_EXTERN_LEVEL)) {
+		set_er(1u << (xtensa_get_ext_irq_no(d->hwirq) -
+					HW_IRQ_MX_BASE), MIENG);
+	} else {
+		int cpu = get_cpu();
+		per_cpu(cached_irq_mask, cpu) &= ~mask;
+		set_sr(per_cpu(cached_irq_mask, cpu), intenable);
+		put_cpu();
+	}
+}
+
+static void xtensa_mx_irq_unmask(struct irq_data *d)
+{
+	unsigned long mask = 1u << d->hwirq;
+
+	if (mask & (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+				XCHAL_INTTYPE_MASK_EXTERN_LEVEL)) {
+		set_er(1u << (xtensa_get_ext_irq_no(d->hwirq) -
+					HW_IRQ_MX_BASE), MIENGSET);
+	} else {
+		int cpu = get_cpu();
+		per_cpu(cached_irq_mask, cpu) |= mask;
+		set_sr(per_cpu(cached_irq_mask, cpu), intenable);
+		put_cpu();
+	}
+}
+
+static void xtensa_mx_irq_enable(struct irq_data *d)
+{
+	variant_irq_enable(d->hwirq);
+	xtensa_mx_irq_unmask(d);
+}
+
+static void xtensa_mx_irq_disable(struct irq_data *d)
+{
+	xtensa_mx_irq_mask(d);
+	variant_irq_disable(d->hwirq);
+}
+
+static void xtensa_mx_irq_ack(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intclear);
+}
+
+static int xtensa_mx_irq_retrigger(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intset);
+	return 1;
+}
+
+static int xtensa_mx_irq_set_affinity(struct irq_data *d,
+		const struct cpumask *dest, bool force)
+{
+	unsigned mask = 1u << cpumask_any(dest);
+
+	set_er(mask, MIROUT(d->hwirq - HW_IRQ_MX_BASE));
+	return 0;
+
+}
+
+static struct irq_chip xtensa_mx_irq_chip = {
+	.name		= "xtensa-mx",
+	.irq_enable	= xtensa_mx_irq_enable,
+	.irq_disable	= xtensa_mx_irq_disable,
+	.irq_mask	= xtensa_mx_irq_mask,
+	.irq_unmask	= xtensa_mx_irq_unmask,
+	.irq_ack	= xtensa_mx_irq_ack,
+	.irq_retrigger	= xtensa_mx_irq_retrigger,
+	.irq_set_affinity = xtensa_mx_irq_set_affinity,
+};
+
+int __init xtensa_mx_init_legacy(struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_legacy(NULL, NR_IRQS, 0, 0,
+				&xtensa_mx_irq_domain_ops,
+				&xtensa_mx_irq_chip);
+	irq_set_default_host(root_domain);
+	secondary_init_irq();
+	return 0;
+}
+
+static int __init xtensa_mx_init(struct device_node *np,
+		struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_linear(np, NR_IRQS, &xtensa_mx_irq_domain_ops,
+				&xtensa_mx_irq_chip);
+	irq_set_default_host(root_domain);
+	secondary_init_irq();
+	return 0;
+}
+IRQCHIP_DECLARE(xtensa_mx_irq_chip, "xtensa,mx", xtensa_mx_init);
diff --git a/include/linux/irqchip/xtensa-mx.h b/include/linux/irqchip/xtensa-mx.h
new file mode 100644
index 0000000..9c3b6ec
--- /dev/null
+++ b/include/linux/irqchip/xtensa-mx.h
@@ -0,0 +1,17 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ *
+ * 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.
+ */
+
+#ifndef __LINUX_IRQCHIP_XTENSA_MX_H
+#define __LINUX_IRQCHIP_XTENSA_MX_H
+
+struct device_node;
+int xtensa_mx_init_legacy(struct device_node *interrupt_parent);
+
+#endif /* __LINUX_IRQCHIP_XTENSA_MX_H */
-- 
1.8.1.4

[PATCH 15/17] xtensa: add SMP support

[Max Filippov]

This is largely based on SMP code from the xtensa-2.6.29-smp tree by
Piet Delaney, Marc Gauthier, Joe Taylor, Christian Zankel (and possibly
other Tensilica folks).

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/Kconfig                                |  37 +-
 arch/xtensa/include/asm/barrier.h                  |   4 +-
 arch/xtensa/include/asm/bitops.h                   |   8 +-
 arch/xtensa/include/asm/cacheflush.h               |  40 +-
 arch/xtensa/include/asm/mmu.h                      |  10 +-
 arch/xtensa/include/asm/mmu_context.h              | 100 +++--
 arch/xtensa/include/asm/ptrace.h                   |   8 +
 arch/xtensa/include/asm/smp.h                      |  29 +-
 arch/xtensa/include/asm/timex.h                    |  14 +-
 arch/xtensa/include/asm/tlbflush.h                 |  42 +-
 arch/xtensa/include/asm/traps.h                    |   1 +
 arch/xtensa/kernel/Makefile                        |   1 +
 arch/xtensa/kernel/head.S                          | 130 ++++--
 arch/xtensa/kernel/irq.c                           |  15 +-
 arch/xtensa/kernel/mxhead.S                        |  85 ++++
 arch/xtensa/kernel/setup.c                         |  66 ++-
 arch/xtensa/kernel/smp.c                           | 465 +++++++++++++++++++++
 arch/xtensa/kernel/traps.c                         |  33 +-
 arch/xtensa/kernel/vmlinux.lds.S                   |  26 ++
 arch/xtensa/mm/cache.c                             |   7 +-
 arch/xtensa/mm/fault.c                             |   2 +-
 arch/xtensa/mm/mmu.c                               |   4 +-
 arch/xtensa/mm/tlb.c                               |  37 +-
 .../platforms/xtfpga/include/platform/hardware.h   |   2 +-
 24 files changed, 967 insertions(+), 199 deletions(-)
 create mode 100644 arch/xtensa/kernel/mxhead.S
 create mode 100644 arch/xtensa/kernel/smp.c

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index fb140ae..4b09c60 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -9,7 +9,6 @@ config XTENSA
 	select GENERIC_CLOCKEVENTS
 	select VIRT_TO_BUS
 	select GENERIC_IRQ_SHOW
-	select GENERIC_CPU_DEVICES
 	select GENERIC_SCHED_CLOCK
 	select MODULES_USE_ELF_RELA
 	select GENERIC_PCI_IOMAP
@@ -65,6 +64,9 @@ config MMU
 config VARIANT_IRQ_SWITCH
 	def_bool n
 
+config MAY_HAVE_SMP
+	def_bool n
+
 menu "Processor type and features"
 
 choice
@@ -105,6 +107,39 @@ config XTENSA_UNALIGNED_USER
 
 source "kernel/Kconfig.preempt"
 
+config HAVE_SMP
+	bool "System Supports SMP (MX)"
+	depends on MAY_HAVE_SMP
+	select XTENSA_MX
+	help
+	  This option is use to indicate that the system-on-a-chip (SOC)
+	  supports Multiprocessing. Multiprocessor support implemented above
+	  the CPU core definition and currently needs to be selected manually.
+
+	  Multiprocessor support in implemented with external cache and
+	  interrupt controlers.
+
+	  The MX interrupt distributer adds Interprocessor Interrupts
+	  and causes the IRQ numbers to be increased by 4 for devices
+	  like the open cores ethernet driver and the serial interface.
+
+	  You still have to select "Enable SMP" to enable SMP on this SOC.
+
+config SMP
+	bool "Enable Symmetric multi-processing support"
+	depends on HAVE_SMP
+	select USE_GENERIC_SMP_HELPERS
+	select GENERIC_SMP_IDLE_THREAD
+	help
+	  Enabled SMP Software; allows more than one CPU/CORE
+	  to be activated during startup.
+
+config NR_CPUS
+	depends on SMP
+	int "Maximum number of CPUs (2-32)"
+	range 2 32
+	default "4"
+
 config MATH_EMULATION
 	bool "Math emulation"
 	help
diff --git a/arch/xtensa/include/asm/barrier.h b/arch/xtensa/include/asm/barrier.h
index ef02167..8e5e5c9 100644
--- a/arch/xtensa/include/asm/barrier.h
+++ b/arch/xtensa/include/asm/barrier.h
@@ -17,7 +17,9 @@
 #define wmb() mb()
 
 #ifdef CONFIG_SMP
-#error smp_* not defined
+#define smp_mb()	mb()
+#define smp_rmb()	rmb()
+#define smp_wmb()	wmb()
 #else
 #define smp_mb()	barrier()
 #define smp_rmb()	barrier()
diff --git a/arch/xtensa/include/asm/bitops.h b/arch/xtensa/include/asm/bitops.h
index 84afe58..7b6873a 100644
--- a/arch/xtensa/include/asm/bitops.h
+++ b/arch/xtensa/include/asm/bitops.h
@@ -22,12 +22,8 @@
 #include <asm/processor.h>
 #include <asm/byteorder.h>
 
-#ifdef CONFIG_SMP
-# error SMP not supported on this architecture
-#endif
-
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
+#define smp_mb__before_clear_bit()	smp_mb()
+#define smp_mb__after_clear_bit()	smp_mb()
 
 #include <asm-generic/bitops/non-atomic.h>
 
diff --git a/arch/xtensa/include/asm/cacheflush.h b/arch/xtensa/include/asm/cacheflush.h
index 127cd48..555a98a 100644
--- a/arch/xtensa/include/asm/cacheflush.h
+++ b/arch/xtensa/include/asm/cacheflush.h
@@ -1,18 +1,14 @@
 /*
- * include/asm-xtensa/cacheflush.h
- *
  * 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.
  *
- * (C) 2001 - 2007 Tensilica Inc.
+ * (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_CACHEFLUSH_H
 #define _XTENSA_CACHEFLUSH_H
 
-#ifdef __KERNEL__
-
 #include <linux/mm.h>
 #include <asm/processor.h>
 #include <asm/page.h>
@@ -51,7 +47,6 @@ extern void __invalidate_icache_page(unsigned long);
 extern void __invalidate_icache_range(unsigned long, unsigned long);
 extern void __invalidate_dcache_range(unsigned long, unsigned long);
 
-
 #if XCHAL_DCACHE_IS_WRITEBACK
 extern void __flush_invalidate_dcache_all(void);
 extern void __flush_dcache_page(unsigned long);
@@ -87,9 +82,22 @@ static inline void __invalidate_icache_page_alias(unsigned long virt,
  * (see also Documentation/cachetlb.txt)
  */
 
-#if (DCACHE_WAY_SIZE > PAGE_SIZE)
+#if (DCACHE_WAY_SIZE > PAGE_SIZE) || defined(CONFIG_SMP)
+
+#ifdef CONFIG_SMP
+void flush_cache_all(void);
+void flush_cache_range(struct vm_area_struct*, ulong, ulong);
+void flush_icache_range(unsigned long start, unsigned long end);
+void flush_cache_page(struct vm_area_struct*,
+			     unsigned long, unsigned long);
+#else
+#define flush_cache_all local_flush_cache_all
+#define flush_cache_range local_flush_cache_range
+#define flush_icache_range local_flush_icache_range
+#define flush_cache_page  local_flush_cache_page
+#endif
 
-#define flush_cache_all()						\
+#define local_flush_cache_all()						\
 	do {								\
 		__flush_invalidate_dcache_all();			\
 		__invalidate_icache_all();				\
@@ -103,9 +111,11 @@ static inline void __invalidate_icache_page_alias(unsigned long virt,
 
 #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
 extern void flush_dcache_page(struct page*);
-extern void flush_cache_range(struct vm_area_struct*, ulong, ulong);
-extern void flush_cache_page(struct vm_area_struct*,
-			     unsigned long, unsigned long);
+
+void local_flush_cache_range(struct vm_area_struct *vma,
+		unsigned long start, unsigned long end);
+void local_flush_cache_page(struct vm_area_struct *vma,
+		unsigned long address, unsigned long pfn);
 
 #else
 
@@ -119,13 +129,14 @@ extern void flush_cache_page(struct vm_area_struct*,
 #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 0
 #define flush_dcache_page(page)				do { } while (0)
 
-#define flush_cache_page(vma,addr,pfn)			do { } while (0)
-#define flush_cache_range(vma,start,end)		do { } while (0)
+#define flush_icache_range local_flush_icache_range
+#define flush_cache_page(vma, addr, pfn)		do { } while (0)
+#define flush_cache_range(vma, start, end)		do { } while (0)
 
 #endif
 
 /* Ensure consistency between data and instruction cache. */
-#define flush_icache_range(start,end) 					\
+#define local_flush_icache_range(start, end)				\
 	do {								\
 		__flush_dcache_range(start, (end) - (start));		\
 		__invalidate_icache_range(start,(end) - (start));	\
@@ -253,5 +264,4 @@ static inline void flush_invalidate_dcache_unaligned(u32 addr, u32 size)
 	}
 }
 
-#endif /* __KERNEL__ */
 #endif /* _XTENSA_CACHEFLUSH_H */
diff --git a/arch/xtensa/include/asm/mmu.h b/arch/xtensa/include/asm/mmu.h
index 8554b2c..71afe41 100644
--- a/arch/xtensa/include/asm/mmu.h
+++ b/arch/xtensa/include/asm/mmu.h
@@ -1,11 +1,9 @@
 /*
- * include/asm-xtensa/mmu.h
- *
  * 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) 2001 - 2005 Tensilica Inc.
+ * Copyright (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_MMU_H
@@ -15,8 +13,10 @@
 #include <asm-generic/mmu.h>
 #else
 
-/* Default "unsigned long" context */
-typedef unsigned long mm_context_t;
+typedef struct {
+	unsigned long asid[NR_CPUS];
+	unsigned int cpu;
+} mm_context_t;
 
 #endif /* CONFIG_MMU */
 #endif	/* _XTENSA_MMU_H */
diff --git a/arch/xtensa/include/asm/mmu_context.h b/arch/xtensa/include/asm/mmu_context.h
index 86292c2..d33c71a 100644
--- a/arch/xtensa/include/asm/mmu_context.h
+++ b/arch/xtensa/include/asm/mmu_context.h
@@ -1,13 +1,11 @@
 /*
- * include/asm-xtensa/mmu_context.h
- *
  * Switch an MMU context.
  *
  * 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) 2001 - 2005 Tensilica Inc.
+ * Copyright (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_MMU_CONTEXT_H
@@ -20,22 +18,25 @@
 #include <linux/stringify.h>
 #include <linux/sched.h>
 
-#include <variant/core.h>
+#include <asm/vectors.h>
 
 #include <asm/pgtable.h>
 #include <asm/cacheflush.h>
 #include <asm/tlbflush.h>
 #include <asm-generic/mm_hooks.h>
+#include <asm-generic/percpu.h>
 
 #if (XCHAL_HAVE_TLBS != 1)
 # error "Linux must have an MMU!"
 #endif
 
-extern unsigned long asid_cache;
+DECLARE_PER_CPU(unsigned long, asid_cache);
+#define cpu_asid_cache(cpu) per_cpu(asid_cache, cpu)
 
 /*
  * NO_CONTEXT is the invalid ASID value that we don't ever assign to
- * any user or kernel context.
+ * any user or kernel context.  We use the reserved values in the
+ * ASID_INSERT macro below.
  *
  * 0 invalid
  * 1 kernel
@@ -68,64 +69,77 @@ static inline unsigned long get_rasid_register (void)
 	return tmp;
 }
 
-static inline void
-__get_new_mmu_context(struct mm_struct *mm)
+static inline void get_new_mmu_context(struct mm_struct *mm, unsigned int cpu)
+{
+	unsigned long asid = cpu_asid_cache(cpu);
+	if ((++asid & ASID_MASK) == 0) {
+		/*
+		 * Start new asid cycle; continue counting with next
+		 * incarnation bits; skipping over 0, 1, 2, 3.
+		 */
+		local_flush_tlb_all();
+		asid += ASID_USER_FIRST;
+	}
+	cpu_asid_cache(cpu) = asid;
+	mm->context.asid[cpu] = asid;
+	mm->context.cpu = cpu;
+}
+
+static inline void get_mmu_context(struct mm_struct *mm, unsigned int cpu)
 {
-	extern void flush_tlb_all(void);
-	if (! (++asid_cache & ASID_MASK) ) {
-		flush_tlb_all(); /* start new asid cycle */
-		asid_cache += ASID_USER_FIRST;
+	/*
+	 * Check if our ASID is of an older version and thus invalid.
+	 */
+
+	if (mm) {
+		unsigned long asid = mm->context.asid[cpu];
+
+		if (asid == NO_CONTEXT ||
+				((asid ^ cpu_asid_cache(cpu)) & ~ASID_MASK))
+			get_new_mmu_context(mm, cpu);
 	}
-	mm->context = asid_cache;
 }
 
-static inline void
-__load_mmu_context(struct mm_struct *mm)
+static inline void activate_context(struct mm_struct *mm, unsigned int cpu)
 {
-	set_rasid_register(ASID_INSERT(mm->context));
+	get_mmu_context(mm, cpu);
+	set_rasid_register(ASID_INSERT(mm->context.asid[cpu]));
 	invalidate_page_directory();
 }
 
 /*
  * Initialize the context related info for a new mm_struct
- * instance.
+ * instance.  Valid cpu values are 0..(NR_CPUS-1), so initializing
+ * to -1 says the process has never run on any core.
  */
 
-static inline int
-init_new_context(struct task_struct *tsk, struct mm_struct *mm)
+static inline int init_new_context(struct task_struct *tsk,
+		struct mm_struct *mm)
 {
-	mm->context = NO_CONTEXT;
+	int cpu;
+	for_each_possible_cpu(cpu) {
+		mm->context.asid[cpu] = NO_CONTEXT;
+	}
+	mm->context.cpu = -1;
 	return 0;
 }
 
-/*
- * After we have set current->mm to a new value, this activates
- * the context for the new mm so we see the new mappings.
- */
-static inline void
-activate_mm(struct mm_struct *prev, struct mm_struct *next)
-{
-	/* Unconditionally get a new ASID.  */
-
-	__get_new_mmu_context(next);
-	__load_mmu_context(next);
-}
-
-
 static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			     struct task_struct *tsk)
 {
-	unsigned long asid = asid_cache;
-
-	/* Check if our ASID is of an older version and thus invalid */
-
-	if (next->context == NO_CONTEXT || ((next->context^asid) & ~ASID_MASK))
-		__get_new_mmu_context(next);
-
-	__load_mmu_context(next);
+	unsigned int cpu = smp_processor_id();
+	int migrated = next->context.cpu != cpu;
+	/* Flush the icache if we migrated to a new core. */
+	if (migrated) {
+		__invalidate_icache_all();
+		next->context.cpu = cpu;
+	}
+	if (migrated || prev != next)
+		activate_context(next, cpu);
 }
 
-#define deactivate_mm(tsk, mm)	do { } while(0)
+#define activate_mm(prev, next)	switch_mm((prev), (next), NULL)
+#define deactivate_mm(tsk, mm)	do { } while (0)
 
 /*
  * Destroy context related info for an mm_struct that is about
diff --git a/arch/xtensa/include/asm/ptrace.h b/arch/xtensa/include/asm/ptrace.h
index 81f31bc..598e752 100644
--- a/arch/xtensa/include/asm/ptrace.h
+++ b/arch/xtensa/include/asm/ptrace.h
@@ -59,9 +59,17 @@ struct pt_regs {
 	(task_stack_page(tsk) + KERNEL_STACK_SIZE - (XCHAL_NUM_AREGS-16)*4) - 1)
 # define user_mode(regs) (((regs)->ps & 0x00000020)!=0)
 # define instruction_pointer(regs) ((regs)->pc)
+# define return_pointer(regs) (MAKE_PC_FROM_RA((regs)->areg[0], \
+					       (regs)->areg[1]))
 
 # ifndef CONFIG_SMP
 #  define profile_pc(regs) instruction_pointer(regs)
+# else
+#  define profile_pc(regs)						\
+	({								\
+		in_lock_functions(instruction_pointer(regs)) ?		\
+		return_pointer(regs) : instruction_pointer(regs);	\
+	})
 # endif
 
 #define user_stack_pointer(regs) ((regs)->areg[1])
diff --git a/arch/xtensa/include/asm/smp.h b/arch/xtensa/include/asm/smp.h
index 83c569e..30ac58c 100644
--- a/arch/xtensa/include/asm/smp.h
+++ b/arch/xtensa/include/asm/smp.h
@@ -1,27 +1,34 @@
 /*
- * include/asm-xtensa/smp.h
- *
  * 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) 2001 - 2005 Tensilica Inc.
+ * Copyright (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_SMP_H
 #define _XTENSA_SMP_H
 
-extern struct xtensa_cpuinfo boot_cpu_data;
+#ifdef CONFIG_SMP
 
-#define cpu_data (&boot_cpu_data)
-#define current_cpu_data boot_cpu_data
+#define raw_smp_processor_id()	(current_thread_info()->cpu)
+#define cpu_logical_map(cpu)	(cpu)
 
-struct xtensa_cpuinfo {
-	unsigned long	*pgd_cache;
-	unsigned long	*pte_cache;
-	unsigned long	pgtable_cache_sz;
+struct start_info {
+	unsigned long stack;
 };
+extern struct start_info start_info;
 
-#define cpu_logical_map(cpu)	(cpu)
+struct cpumask;
+void arch_send_call_function_ipi_mask(const struct cpumask *mask);
+void arch_send_call_function_single_ipi(int cpu);
+
+void smp_init_cpus(void);
+void secondary_init_irq(void);
+void ipi_init(void);
+struct seq_file;
+void show_ipi_list(struct seq_file *p, int prec);
+
+#endif /* CONFIG_SMP */
 
 #endif	/* _XTENSA_SMP_H */
diff --git a/arch/xtensa/include/asm/timex.h b/arch/xtensa/include/asm/timex.h
index 27fa3c1..ca929e6 100644
--- a/arch/xtensa/include/asm/timex.h
+++ b/arch/xtensa/include/asm/timex.h
@@ -1,18 +1,14 @@
 /*
- * include/asm-xtensa/timex.h
- *
  * 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) 2001 - 2008 Tensilica Inc.
+ * Copyright (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_TIMEX_H
 #define _XTENSA_TIMEX_H
 
-#ifdef __KERNEL__
-
 #include <asm/processor.h>
 #include <linux/stringify.h>
 
@@ -39,14 +35,9 @@ extern unsigned long ccount_freq;
 
 typedef unsigned long long cycles_t;
 
-/*
- * Only used for SMP.
- */
-
-extern cycles_t cacheflush_time;
-
 #define get_cycles()	(0)
 
+void local_timer_setup(unsigned cpu);
 
 /*
  * Register access.
@@ -81,5 +72,4 @@ static inline void set_linux_timer (unsigned long ccompare)
 	WSR_CCOMPARE(LINUX_TIMER, ccompare);
 }
 
-#endif	/* __KERNEL__ */
 #endif	/* _XTENSA_TIMEX_H */
diff --git a/arch/xtensa/include/asm/tlbflush.h b/arch/xtensa/include/asm/tlbflush.h
index 43dd348..fc34274 100644
--- a/arch/xtensa/include/asm/tlbflush.h
+++ b/arch/xtensa/include/asm/tlbflush.h
@@ -1,18 +1,14 @@
 /*
- * include/asm-xtensa/tlbflush.h
- *
  * 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) 2001 - 2005 Tensilica Inc.
+ * Copyright (C) 2001 - 2013 Tensilica Inc.
  */
 
 #ifndef _XTENSA_TLBFLUSH_H
 #define _XTENSA_TLBFLUSH_H
 
-#ifdef __KERNEL__
-
 #include <linux/stringify.h>
 #include <asm/processor.h>
 
@@ -34,12 +30,37 @@
  *  - flush_tlb_range(mm, start, end) flushes a range of pages
  */
 
-extern void flush_tlb_all(void);
-extern void flush_tlb_mm(struct mm_struct*);
-extern void flush_tlb_page(struct vm_area_struct*,unsigned long);
-extern void flush_tlb_range(struct vm_area_struct*,unsigned long,unsigned long);
+void local_flush_tlb_all(void);
+void local_flush_tlb_mm(struct mm_struct *mm);
+void local_flush_tlb_page(struct vm_area_struct *vma,
+		unsigned long page);
+void local_flush_tlb_range(struct vm_area_struct *vma,
+		unsigned long start, unsigned long end);
+
+#ifdef CONFIG_SMP
+
+void flush_tlb_all(void);
+void flush_tlb_mm(struct mm_struct *);
+void flush_tlb_page(struct vm_area_struct *, unsigned long);
+void flush_tlb_range(struct vm_area_struct *, unsigned long,
+		unsigned long);
+
+static inline void flush_tlb_kernel_range(unsigned long start,
+		unsigned long end)
+{
+	flush_tlb_all();
+}
+
+#else /* !CONFIG_SMP */
+
+#define flush_tlb_all()			   local_flush_tlb_all()
+#define flush_tlb_mm(mm)		   local_flush_tlb_mm(mm)
+#define flush_tlb_page(vma, page)	   local_flush_tlb_page(vma, page)
+#define flush_tlb_range(vma, vmaddr, end)  local_flush_tlb_range(vma, vmaddr, \
+								 end)
+#define flush_tlb_kernel_range(start, end) local_flush_tlb_all()
 
-#define flush_tlb_kernel_range(start,end) flush_tlb_all()
+#endif /* CONFIG_SMP */
 
 /* TLB operations. */
 
@@ -187,5 +208,4 @@ static inline unsigned long read_itlb_translation (int way)
 }
 
 #endif	/* __ASSEMBLY__ */
-#endif	/* __KERNEL__ */
 #endif	/* _XTENSA_TLBFLUSH_H */
diff --git a/arch/xtensa/include/asm/traps.h b/arch/xtensa/include/asm/traps.h
index 917488a..8c194f6 100644
--- a/arch/xtensa/include/asm/traps.h
+++ b/arch/xtensa/include/asm/traps.h
@@ -19,6 +19,7 @@
  */
 extern void * __init trap_set_handler(int cause, void *handler);
 extern void do_unhandled(struct pt_regs *regs, unsigned long exccause);
+void secondary_trap_init(void);
 
 static inline void spill_registers(void)
 {
diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile
index f90265e..18d962a 100644
--- a/arch/xtensa/kernel/Makefile
+++ b/arch/xtensa/kernel/Makefile
@@ -12,6 +12,7 @@ obj-$(CONFIG_KGDB) += xtensa-stub.o
 obj-$(CONFIG_PCI) += pci.o
 obj-$(CONFIG_MODULES) += xtensa_ksyms.o module.o
 obj-$(CONFIG_FUNCTION_TRACER) += mcount.o
+obj-$(CONFIG_SMP) += smp.o mxhead.o
 
 AFLAGS_head.o += -mtext-section-literals
 
diff --git a/arch/xtensa/kernel/head.S b/arch/xtensa/kernel/head.S
index 7d740eb..74ec62c 100644
--- a/arch/xtensa/kernel/head.S
+++ b/arch/xtensa/kernel/head.S
@@ -19,6 +19,7 @@
 #include <asm/page.h>
 #include <asm/cacheasm.h>
 #include <asm/initialize_mmu.h>
+#include <asm/mxregs.h>
 
 #include <linux/init.h>
 #include <linux/linkage.h>
@@ -54,7 +55,7 @@ ENTRY(_start)
 
 	/* Preserve the pointer to the boot parameter list in EXCSAVE_1 */
 	wsr     a2, excsave1
-	_j	_SetupMMU
+	_j	_SetupOCD
 
 	.align	4
 	.literal_position
@@ -62,6 +63,23 @@ ENTRY(_start)
 	.word	_startup
 
 	.align	4
+_SetupOCD:
+	/*
+	 * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions).
+	 * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow
+	 * xt-gdb to single step via DEBUG exceptions received directly
+	 * by ocd.
+	 */
+	movi	a1, 1
+	movi	a0, 0
+	wsr	a1, windowstart
+	wsr	a0, windowbase
+	rsync
+
+	movi	a1, LOCKLEVEL
+	wsr	a1, ps
+	rsync
+
 	.global _SetupMMU
 _SetupMMU:
 	Offset = _SetupMMU - _start
@@ -90,19 +108,6 @@ ENDPROC(_start)
 
 ENTRY(_startup)
 
-	/* Disable interrupts and exceptions. */
-
-	movi	a0, LOCKLEVEL
-	wsr	a0, ps
-
-	/* Start with a fresh windowbase and windowstart.  */
-
-	movi	a1, 1
-	movi	a0, 0
-	wsr	a1, windowstart
-	wsr	a0, windowbase
-	rsync
-
 	/* Set a0 to 0 for the remaining initialization. */
 
 	movi	a0, 0
@@ -154,17 +159,6 @@ ENTRY(_startup)
 	wsr	a0, cpenable
 #endif
 
-	/* Set PS.INTLEVEL=LOCKLEVEL, PS.WOE=0, kernel stack, PS.EXCM=0
-	 *
-	 * Note: PS.EXCM must be cleared before using any loop
-	 *	 instructions; otherwise, they are silently disabled, and
-	 * 	 at most one iteration of the loop is executed.
-	 */
-
-	movi	a1, LOCKLEVEL
-	wsr	a1, ps
-	rsync
-
 	/*  Initialize the caches.
 	 *  a2, a3 are just working registers (clobbered).
 	 */
@@ -182,6 +176,37 @@ ENTRY(_startup)
 
 	isync
 
+#ifdef CONFIG_HAVE_SMP
+	movi	a2, CCON	# MX External Register to Configure Cache
+	movi	a3, 1
+	wer	a3, a2
+#endif
+
+	/* Setup stack and enable window exceptions (keep irqs disabled) */
+
+	movi	a1, start_info
+	l32i	a1, a1, 0
+
+	movi	a2, (1 << PS_WOE_BIT) | LOCKLEVEL
+					# WOE=1, INTLEVEL=LOCKLEVEL, UM=0
+	wsr	a2, ps			# (enable reg-windows; progmode stack)
+	rsync
+
+	/* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/
+
+	movi	a2, debug_exception
+	wsr	a2, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
+
+#ifdef CONFIG_SMP
+	/*
+	 * Notice that we assume with SMP that cores have PRID
+	 * supported by the cores.
+	 */
+	rsr	a2, prid
+	bnez	a2, .Lboot_secondary
+
+#endif  /* CONFIG_SMP */
+
 	/* Unpack data sections
 	 *
 	 * The linker script used to build the Linux kernel image
@@ -234,24 +259,7 @@ ENTRY(_startup)
 	___invalidate_icache_all a2 a3
 	isync
 
-	/* Setup stack and enable window exceptions (keep irqs disabled) */
-
-	movi	a1, init_thread_union
-	addi	a1, a1, KERNEL_STACK_SIZE
-
-	movi	a2, (1 << PS_WOE_BIT) | LOCKLEVEL
-					# WOE=1, INTLEVEL=LOCKLEVEL, UM=0
-	wsr	a2, ps			# (enable reg-windows; progmode stack)
-	rsync
-
-	/* Set up EXCSAVE[DEBUGLEVEL] to point to the Debug Exception Handler.*/
-
-	movi	a2, debug_exception
-	wsr	a2, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
-
-	/* Set up EXCSAVE[1] to point to the exc_table. */
-
-	movi	a6, exc_table
+	movi	a6, 0
 	xsr	a6, excsave1
 
 	/* init_arch kick-starts the linux kernel */
@@ -265,9 +273,45 @@ ENTRY(_startup)
 should_never_return:
 	j	should_never_return
 
+#ifdef CONFIG_SMP
+.Lboot_secondary:
+
+	movi	a2, cpu_start_ccount
+1:
+	l32i	a3, a2, 0
+	beqi	a3, 0, 1b
+	movi	a3, 0
+	s32i	a3, a2, 0
+	memw
+1:
+	l32i	a3, a2, 0
+	beqi	a3, 0, 1b
+	wsr	a3, ccount
+	movi	a3, 0
+	s32i	a3, a2, 0
+	memw
+
+	movi	a6, 0
+	wsr	a6, excsave1
+
+	movi	a4, secondary_start_kernel
+	callx4	a4
+	j	should_never_return
+
+#endif  /* CONFIG_SMP */
+
 ENDPROC(_startup)
 
 /*
+ * DATA section
+ */
+
+        .section ".data.init.refok"
+        .align  4
+ENTRY(start_info)
+        .long   init_thread_union + KERNEL_STACK_SIZE
+
+/*
  * BSS section
  */
 	
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index 7d49730..fad9e00 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -4,7 +4,7 @@
  * Xtensa built-in interrupt controller and some generic functions copied
  * from i386.
  *
- * Copyright (C) 2002 - 2006 Tensilica, Inc.
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
  * Copyright (C) 1992, 1998 Linus Torvalds, Ingo Molnar
  *
  *
@@ -19,10 +19,12 @@
 #include <linux/irq.h>
 #include <linux/kernel_stat.h>
 #include <linux/irqchip.h>
+#include <linux/irqchip/xtensa-mx.h>
 #include <linux/irqchip/xtensa-pic.h>
 #include <linux/irqdomain.h>
 #include <linux/of.h>
 
+#include <asm/mxregs.h>
 #include <asm/uaccess.h>
 #include <asm/platform.h>
 
@@ -55,6 +57,9 @@ asmlinkage void do_IRQ(int hwirq, struct pt_regs *regs)
 
 int arch_show_interrupts(struct seq_file *p, int prec)
 {
+#ifdef CONFIG_SMP
+	show_ipi_list(p, prec);
+#endif
 	seq_printf(p, "%*s: ", prec, "ERR");
 	seq_printf(p, "%10u\n", atomic_read(&irq_err_count));
 	return 0;
@@ -136,7 +141,15 @@ void __init init_IRQ(void)
 #ifdef CONFIG_OF
 	irqchip_init();
 #else
+#ifdef CONFIG_HAVE_SMP
+	xtensa_mx_init_legacy(NULL);
+#else
 	xtensa_pic_init_legacy(NULL);
 #endif
+#endif
+
+#ifdef CONFIG_SMP
+	ipi_init();
+#endif
 	variant_init_irq();
 }
diff --git a/arch/xtensa/kernel/mxhead.S b/arch/xtensa/kernel/mxhead.S
new file mode 100644
index 0000000..77a161a
--- /dev/null
+++ b/arch/xtensa/kernel/mxhead.S
@@ -0,0 +1,85 @@
+/*
+ * Xtensa Secondary Processors startup code.
+ *
+ * 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) 2001 - 2013 Tensilica Inc.
+ *
+ * Joe Taylor <joe@tensilica.com>
+ * Chris Zankel <chris@zankel.net>
+ * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca>
+ * Pete Delaney <piet@tensilica.com>
+ */
+
+#include <linux/linkage.h>
+
+#include <asm/cacheasm.h>
+#include <asm/initialize_mmu.h>
+#include <asm/mxregs.h>
+#include <asm/regs.h>
+
+
+	.section .SecondaryResetVector.text, "ax"
+
+
+ENTRY(_SecondaryResetVector)
+	_j _SetupOCD
+
+	.begin  no-absolute-literals
+	.literal_position
+
+_SetupOCD:
+	/*
+	 * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions).
+	 * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow
+	 * xt-gdb to single step via DEBUG exceptions received directly
+	 * by ocd.
+	 */
+	movi	a1, 1
+	movi	a0, 0
+	wsr	a1, windowstart
+	wsr	a0, windowbase
+	rsync
+
+	movi	a1, LOCKLEVEL
+	wsr	a1, ps
+	rsync
+
+_SetupMMU:
+	Offset = _SetupMMU - _SecondaryResetVector
+
+#ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX
+	initialize_mmu
+#endif
+
+	/*
+	 * Start Secondary Processors with NULL pointer to boot params.
+	 */
+	movi	a2, 0				#  a2 == NULL
+	movi	a3, _startup
+	jx	a3
+
+	.end    no-absolute-literals
+
+
+	.section 	.SecondaryResetVector.remapped_text, "ax"
+	.global         _RemappedSecondaryResetVector
+
+	.org 0                                  # Need to do org before literals
+
+_RemappedSecondaryResetVector:
+	.begin  no-absolute-literals
+	.literal_position
+
+	_j      _RemappedSetupMMU
+	. = _RemappedSecondaryResetVector + Offset
+
+_RemappedSetupMMU:
+
+#ifdef CONFIG_INITIALIZE_XTENSA_MMU_INSIDE_VMLINUX
+	initialize_mmu
+#endif
+
+	.end    no-absolute-literals
diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
index 84b26ee..1f6e377 100644
--- a/arch/xtensa/kernel/setup.c
+++ b/arch/xtensa/kernel/setup.c
@@ -21,6 +21,8 @@
 #include <linux/screen_info.h>
 #include <linux/bootmem.h>
 #include <linux/kernel.h>
+#include <linux/percpu.h>
+#include <linux/cpu.h>
 
 #ifdef CONFIG_OF
 #include <linux/of_fdt.h>
@@ -49,6 +51,7 @@
 #include <asm/setup.h>
 #include <asm/param.h>
 #include <asm/traps.h>
+#include <asm/smp.h>
 
 #include <platform/hardware.h>
 
@@ -523,6 +526,10 @@ void __init setup_arch(char **cmdline_p)
 
 	platform_setup(cmdline_p);
 
+#ifdef CONFIG_SMP
+	smp_init_cpus();
+#endif
+
 	paging_init();
 	zones_init();
 
@@ -539,6 +546,21 @@ void __init setup_arch(char **cmdline_p)
 #endif
 }
 
+static DEFINE_PER_CPU(struct cpu, cpu_data);
+
+static int __init topology_init(void)
+{
+	int i;
+
+	for_each_possible_cpu(i) {
+		struct cpu *cpu = &per_cpu(cpu_data, i);
+		register_cpu(cpu, i);
+	}
+
+	return 0;
+}
+subsys_initcall(topology_init);
+
 void machine_restart(char * cmd)
 {
 	platform_restart();
@@ -564,21 +586,27 @@ void machine_power_off(void)
 static int
 c_show(struct seq_file *f, void *slot)
 {
+	char buf[NR_CPUS * 5];
+
+	cpulist_scnprintf(buf, sizeof(buf), cpu_online_mask);
 	/* high-level stuff */
-	seq_printf(f,"processor\t: 0\n"
-		     "vendor_id\t: Tensilica\n"
-		     "model\t\t: Xtensa " XCHAL_HW_VERSION_NAME "\n"
-		     "core ID\t\t: " XCHAL_CORE_ID "\n"
-		     "build ID\t: 0x%x\n"
-		     "byte order\t: %s\n"
-		     "cpu MHz\t\t: %lu.%02lu\n"
-		     "bogomips\t: %lu.%02lu\n",
-		     XCHAL_BUILD_UNIQUE_ID,
-		     XCHAL_HAVE_BE ?  "big" : "little",
-		     ccount_freq/1000000,
-		     (ccount_freq/10000) % 100,
-		     loops_per_jiffy/(500000/HZ),
-		     (loops_per_jiffy/(5000/HZ)) % 100);
+	seq_printf(f, "CPU count\t: %u\n"
+		      "CPU list\t: %s\n"
+		      "vendor_id\t: Tensilica\n"
+		      "model\t\t: Xtensa " XCHAL_HW_VERSION_NAME "\n"
+		      "core ID\t\t: " XCHAL_CORE_ID "\n"
+		      "build ID\t: 0x%x\n"
+		      "byte order\t: %s\n"
+		      "cpu MHz\t\t: %lu.%02lu\n"
+		      "bogomips\t: %lu.%02lu\n",
+		      num_online_cpus(),
+		      buf,
+		      XCHAL_BUILD_UNIQUE_ID,
+		      XCHAL_HAVE_BE ?  "big" : "little",
+		      ccount_freq/1000000,
+		      (ccount_freq/10000) % 100,
+		      loops_per_jiffy/(500000/HZ),
+		      (loops_per_jiffy/(5000/HZ)) % 100);
 
 	seq_printf(f,"flags\t\t: "
 #if XCHAL_HAVE_NMI
@@ -690,7 +718,7 @@ c_show(struct seq_file *f, void *slot)
 static void *
 c_start(struct seq_file *f, loff_t *pos)
 {
-	return (void *) ((*pos == 0) ? (void *)1 : NULL);
+	return (*pos == 0) ? (void *)1 : NULL;
 }
 
 static void *
@@ -706,10 +734,10 @@ c_stop(struct seq_file *f, void *v)
 
 const struct seq_operations cpuinfo_op =
 {
-	start:  c_start,
-	next:   c_next,
-	stop:   c_stop,
-	show:   c_show
+	.start	= c_start,
+	.next	= c_next,
+	.stop	= c_stop,
+	.show	= c_show,
 };
 
 #endif /* CONFIG_PROC_FS */
diff --git a/arch/xtensa/kernel/smp.c b/arch/xtensa/kernel/smp.c
new file mode 100644
index 0000000..46bdd14
--- /dev/null
+++ b/arch/xtensa/kernel/smp.c
@@ -0,0 +1,465 @@
+/*
+ * Xtensa SMP support functions.
+ *
+ * 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) 2008 - 2013 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Joe Taylor <joe@tensilica.com>
+ * Pete Delaney <piet@tensilica.com
+ */
+
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/irqdomain.h>
+#include <linux/irq.h>
+#include <linux/kdebug.h>
+#include <linux/module.h>
+#include <linux/reboot.h>
+#include <linux/seq_file.h>
+#include <linux/smp.h>
+#include <linux/thread_info.h>
+
+#include <asm/cacheflush.h>
+#include <asm/kdebug.h>
+#include <asm/mmu_context.h>
+#include <asm/mxregs.h>
+#include <asm/platform.h>
+#include <asm/tlbflush.h>
+#include <asm/traps.h>
+
+#ifdef CONFIG_SMP
+# if XCHAL_HAVE_S32C1I == 0
+#  error "The S32C1I option is required for SMP."
+# endif
+#endif
+
+/* IPI (Inter Process Interrupt) */
+
+#define IPI_IRQ	0
+
+static irqreturn_t ipi_interrupt(int irq, void *dev_id);
+static struct irqaction ipi_irqaction = {
+	.handler =	ipi_interrupt,
+	.flags =	IRQF_PERCPU,
+	.name =		"ipi",
+};
+
+void ipi_init(void)
+{
+	unsigned irq = irq_create_mapping(NULL, IPI_IRQ);
+	setup_irq(irq, &ipi_irqaction);
+}
+
+static inline unsigned int get_core_count(void)
+{
+	/* Bits 18..21 of SYSCFGID contain the core count minus 1. */
+	unsigned int syscfgid = get_er(SYSCFGID);
+	return ((syscfgid >> 18) & 0xf) + 1;
+}
+
+static inline int get_core_id(void)
+{
+	/* Bits 0...18 of SYSCFGID contain the core id  */
+	unsigned int core_id = get_er(SYSCFGID);
+	return core_id & 0x3fff;
+}
+
+void __init smp_prepare_cpus(unsigned int max_cpus)
+{
+	unsigned i;
+
+	for (i = 0; i < max_cpus; ++i)
+		set_cpu_present(i, true);
+}
+
+void __init smp_init_cpus(void)
+{
+	unsigned i;
+	unsigned int ncpus = get_core_count();
+	unsigned int core_id = get_core_id();
+
+	pr_info("%s: Core Count = %d\n", __func__, ncpus);
+	pr_info("%s: Core Id = %d\n", __func__, core_id);
+
+	for (i = 0; i < ncpus; ++i)
+		set_cpu_possible(i, true);
+}
+
+void __init smp_prepare_boot_cpu(void)
+{
+	unsigned int cpu = smp_processor_id();
+	BUG_ON(cpu != 0);
+	cpu_asid_cache(cpu) = ASID_USER_FIRST;
+}
+
+void __init smp_cpus_done(unsigned int max_cpus)
+{
+}
+
+static int boot_secondary_processors = 1; /* Set with xt-gdb via .xt-gdb */
+static DECLARE_COMPLETION(cpu_running);
+
+void __init secondary_start_kernel(void)
+{
+	struct mm_struct *mm = &init_mm;
+	unsigned int cpu = smp_processor_id();
+
+	init_mmu();
+
+#ifdef CONFIG_DEBUG_KERNEL
+	if (boot_secondary_processors == 0) {
+		pr_debug("%s: boot_secondary_processors:%d; Hanging cpu:%d\n",
+			__func__, boot_secondary_processors, cpu);
+		for (;;)
+			__asm__ __volatile__ ("waiti " __stringify(LOCKLEVEL));
+	}
+
+	pr_debug("%s: boot_secondary_processors:%d; Booting cpu:%d\n",
+		__func__, boot_secondary_processors, cpu);
+#endif
+	/* Init EXCSAVE1 */
+
+	secondary_trap_init();
+
+	/* All kernel threads share the same mm context. */
+
+	atomic_inc(&mm->mm_users);
+	atomic_inc(&mm->mm_count);
+	current->active_mm = mm;
+	cpumask_set_cpu(cpu, mm_cpumask(mm));
+	enter_lazy_tlb(mm, current);
+
+	preempt_disable();
+	trace_hardirqs_off();
+
+	calibrate_delay();
+
+	notify_cpu_starting(cpu);
+
+	secondary_init_irq();
+	local_timer_setup(cpu);
+
+	local_irq_enable();
+
+	set_cpu_online(cpu, true);
+	complete(&cpu_running);
+
+	cpu_startup_entry(CPUHP_ONLINE);
+}
+
+static void mx_cpu_start(void *p)
+{
+	unsigned cpu = (unsigned)p;
+	unsigned long run_stall_mask = get_er(MPSCORE);
+
+	set_er(run_stall_mask & ~(1u << cpu), MPSCORE);
+	pr_debug("%s: cpu: %d, run_stall_mask: %lx ---> %lx\n",
+			__func__, cpu, run_stall_mask, get_er(MPSCORE));
+}
+
+static void mx_cpu_stop(void *p)
+{
+	unsigned cpu = (unsigned)p;
+	unsigned long run_stall_mask = get_er(MPSCORE);
+
+	set_er(run_stall_mask | (1u << cpu), MPSCORE);
+	pr_debug("%s: cpu: %d, run_stall_mask: %lx ---> %lx\n",
+			__func__, cpu, run_stall_mask, get_er(MPSCORE));
+}
+
+unsigned long cpu_start_ccount;
+
+static int boot_secondary(unsigned int cpu, struct task_struct *ts)
+{
+	unsigned long timeout = jiffies + msecs_to_jiffies(1000);
+	unsigned long ccount;
+	int i;
+
+	smp_call_function_single(0, mx_cpu_start, (void *)cpu, 1);
+
+	for (i = 0; i < 2; ++i) {
+		do
+			ccount = get_ccount();
+		while (!ccount);
+
+		cpu_start_ccount = ccount;
+
+		while (time_before(jiffies, timeout)) {
+			mb();
+			if (!cpu_start_ccount)
+				break;
+		}
+
+		if (cpu_start_ccount) {
+			smp_call_function_single(0, mx_cpu_stop,
+					(void *)cpu, 1);
+			cpu_start_ccount = 0;
+			return -EIO;
+		}
+	}
+	return 0;
+}
+
+int __cpu_up(unsigned int cpu, struct task_struct *idle)
+{
+	int ret = 0;
+
+	if (cpu_asid_cache(cpu) == 0)
+		cpu_asid_cache(cpu) = ASID_USER_FIRST;
+
+	start_info.stack = (unsigned long)task_pt_regs(idle);
+	wmb();
+
+	pr_debug("%s: Calling wakeup_secondary(cpu:%d, idle:%p, sp: %08lx)\n",
+			__func__, cpu, idle, start_info.stack);
+
+	ret = boot_secondary(cpu, idle);
+	if (ret == 0) {
+		wait_for_completion_timeout(&cpu_running,
+				msecs_to_jiffies(1000));
+		if (!cpu_online(cpu))
+			ret = -EIO;
+	}
+
+	if (ret)
+		pr_err("CPU %u failed to boot\n", cpu);
+
+	return ret;
+}
+
+enum ipi_msg_type {
+	IPI_RESCHEDULE = 0,
+	IPI_CALL_FUNC,
+	IPI_CPU_STOP,
+	IPI_MAX
+};
+
+static const struct {
+	const char *short_text;
+	const char *long_text;
+} ipi_text[] = {
+	{ .short_text = "RES", .long_text = "Rescheduling interrupts" },
+	{ .short_text = "CAL", .long_text = "Function call interrupts" },
+	{ .short_text = "DIE", .long_text = "CPU shutdown interrupts" },
+};
+
+struct ipi_data {
+	unsigned long ipi_count[IPI_MAX];
+};
+
+static DEFINE_PER_CPU(struct ipi_data, ipi_data);
+
+static void send_ipi_message(const struct cpumask *callmask,
+		enum ipi_msg_type msg_id)
+{
+	int index;
+	unsigned long mask = 0;
+
+	for_each_cpu(index, callmask)
+		if (index != smp_processor_id())
+			mask |= 1 << index;
+
+	set_er(mask, MIPISET(msg_id));
+}
+
+void arch_send_call_function_ipi_mask(const struct cpumask *mask)
+{
+	send_ipi_message(mask, IPI_CALL_FUNC);
+}
+
+void arch_send_call_function_single_ipi(int cpu)
+{
+	send_ipi_message(cpumask_of(cpu), IPI_CALL_FUNC);
+}
+
+void smp_send_reschedule(int cpu)
+{
+	send_ipi_message(cpumask_of(cpu), IPI_RESCHEDULE);
+}
+
+void smp_send_stop(void)
+{
+	struct cpumask targets;
+
+	cpumask_copy(&targets, cpu_online_mask);
+	cpumask_clear_cpu(smp_processor_id(), &targets);
+	send_ipi_message(&targets, IPI_CPU_STOP);
+}
+
+static void ipi_cpu_stop(unsigned int cpu)
+{
+	set_cpu_online(cpu, false);
+	machine_halt();
+}
+
+irqreturn_t ipi_interrupt(int irq, void *dev_id)
+{
+	unsigned int cpu = smp_processor_id();
+	struct ipi_data *ipi = &per_cpu(ipi_data, cpu);
+	unsigned int msg;
+	unsigned i;
+
+	msg = get_er(MIPICAUSE(cpu));
+	for (i = 0; i < IPI_MAX; i++)
+		if (msg & (1 << i)) {
+			set_er(1 << i, MIPICAUSE(cpu));
+			++ipi->ipi_count[i];
+		}
+
+	if (msg & (1 << IPI_RESCHEDULE))
+		scheduler_ipi();
+	if (msg & (1 << IPI_CALL_FUNC))
+		generic_smp_call_function_interrupt();
+	if (msg & (1 << IPI_CPU_STOP))
+		ipi_cpu_stop(cpu);
+
+	return IRQ_HANDLED;
+}
+
+void show_ipi_list(struct seq_file *p, int prec)
+{
+	unsigned int cpu;
+	unsigned i;
+
+	for (i = 0; i < IPI_MAX; ++i) {
+		seq_printf(p, "%*s:", prec, ipi_text[i].short_text);
+		for_each_online_cpu(cpu)
+			seq_printf(p, " %10lu",
+					per_cpu(ipi_data, cpu).ipi_count[i]);
+		seq_printf(p, "   %s\n", ipi_text[i].long_text);
+	}
+}
+
+int setup_profiling_timer(unsigned int multiplier)
+{
+	pr_debug("setup_profiling_timer %d\n", multiplier);
+	return 0;
+}
+
+/* TLB flush functions */
+
+struct flush_data {
+	struct vm_area_struct *vma;
+	unsigned long addr1;
+	unsigned long addr2;
+};
+
+static void ipi_flush_tlb_all(void *arg)
+{
+	local_flush_tlb_all();
+}
+
+void flush_tlb_all(void)
+{
+	on_each_cpu(ipi_flush_tlb_all, NULL, 1);
+}
+
+static void ipi_flush_tlb_mm(void *arg)
+{
+	local_flush_tlb_mm(arg);
+}
+
+void flush_tlb_mm(struct mm_struct *mm)
+{
+	on_each_cpu(ipi_flush_tlb_mm, mm, 1);
+}
+
+static void ipi_flush_tlb_page(void *arg)
+{
+	struct flush_data *fd = arg;
+	local_flush_tlb_page(fd->vma, fd->addr1);
+}
+
+void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
+{
+	struct flush_data fd = {
+		.vma = vma,
+		.addr1 = addr,
+	};
+	on_each_cpu(ipi_flush_tlb_page, &fd, 1);
+}
+
+static void ipi_flush_tlb_range(void *arg)
+{
+	struct flush_data *fd = arg;
+	local_flush_tlb_range(fd->vma, fd->addr1, fd->addr2);
+}
+
+void flush_tlb_range(struct vm_area_struct *vma,
+		     unsigned long start, unsigned long end)
+{
+	struct flush_data fd = {
+		.vma = vma,
+		.addr1 = start,
+		.addr2 = end,
+	};
+	on_each_cpu(ipi_flush_tlb_range, &fd, 1);
+}
+
+/* Cache flush functions */
+
+static void ipi_flush_cache_all(void *arg)
+{
+	local_flush_cache_all();
+}
+
+void flush_cache_all(void)
+{
+	on_each_cpu(ipi_flush_cache_all, NULL, 1);
+}
+
+static void ipi_flush_cache_page(void *arg)
+{
+	struct flush_data *fd = arg;
+	local_flush_cache_page(fd->vma, fd->addr1, fd->addr2);
+}
+
+void flush_cache_page(struct vm_area_struct *vma,
+		     unsigned long address, unsigned long pfn)
+{
+	struct flush_data fd = {
+		.vma = vma,
+		.addr1 = address,
+		.addr2 = pfn,
+	};
+	on_each_cpu(ipi_flush_cache_page, &fd, 1);
+}
+
+static void ipi_flush_cache_range(void *arg)
+{
+	struct flush_data *fd = arg;
+	local_flush_cache_range(fd->vma, fd->addr1, fd->addr2);
+}
+
+void flush_cache_range(struct vm_area_struct *vma,
+		     unsigned long start, unsigned long end)
+{
+	struct flush_data fd = {
+		.vma = vma,
+		.addr1 = start,
+		.addr2 = end,
+	};
+	on_each_cpu(ipi_flush_cache_range, &fd, 1);
+}
+
+static void ipi_flush_icache_range(void *arg)
+{
+	struct flush_data *fd = arg;
+	local_flush_icache_range(fd->addr1, fd->addr2);
+}
+
+void flush_icache_range(unsigned long start, unsigned long end)
+{
+	struct flush_data fd = {
+		.addr1 = start,
+		.addr2 = end,
+	};
+	on_each_cpu(ipi_flush_icache_range, &fd, 1);
+}
diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c
index 3dbe864..3c0ff57 100644
--- a/arch/xtensa/kernel/traps.c
+++ b/arch/xtensa/kernel/traps.c
@@ -157,7 +157,7 @@ COPROCESSOR(7),
  * 2. it is a temporary memory buffer for the exception handlers.
  */
 
-unsigned long exc_table[EXC_TABLE_SIZE/4];
+DEFINE_PER_CPU(unsigned long, exc_table[EXC_TABLE_SIZE/4]);
 
 void die(const char*, struct pt_regs*, long);
 
@@ -313,17 +313,31 @@ do_debug(struct pt_regs *regs)
 }
 
 
+static void set_handler(int idx, void *handler)
+{
+	unsigned int cpu;
+
+	for_each_possible_cpu(cpu)
+		per_cpu(exc_table, cpu)[idx] = (unsigned long)handler;
+}
+
 /* Set exception C handler - for temporary use when probing exceptions */
 
 void * __init trap_set_handler(int cause, void *handler)
 {
-	unsigned long *entry = &exc_table[EXC_TABLE_DEFAULT / 4 + cause];
-	void *previous = (void *)*entry;
-	*entry = (unsigned long)handler;
+	void *previous = (void *)per_cpu(exc_table, 0)[
+		EXC_TABLE_DEFAULT / 4 + cause];
+	set_handler(EXC_TABLE_DEFAULT / 4 + cause, handler);
 	return previous;
 }
 
 
+static void __init trap_init_excsave(void)
+{
+	unsigned long excsave1 = (unsigned long)this_cpu_ptr(exc_table);
+	__asm__ __volatile__("wsr  %0, excsave1\n" : : "a" (excsave1));
+}
+
 /*
  * Initialize dispatch tables.
  *
@@ -337,8 +351,6 @@ void * __init trap_set_handler(int cause, void *handler)
  * See vectors.S for more details.
  */
 
-#define set_handler(idx,handler) (exc_table[idx] = (unsigned long) (handler))
-
 void __init trap_init(void)
 {
 	int i;
@@ -368,10 +380,15 @@ void __init trap_init(void)
 	}
 
 	/* Initialize EXCSAVE_1 to hold the address of the exception table. */
+	trap_init_excsave();
+}
 
-	i = (unsigned long)exc_table;
-	__asm__ __volatile__("wsr  %0, excsave1\n" : : "a" (i));
+#ifdef CONFIG_SMP
+void __init secondary_trap_init(void)
+{
+	trap_init_excsave();
 }
+#endif
 
 /*
  * This function dumps the current valid window frame and other base registers.
diff --git a/arch/xtensa/kernel/vmlinux.lds.S b/arch/xtensa/kernel/vmlinux.lds.S
index 21acd11..ee32c00 100644
--- a/arch/xtensa/kernel/vmlinux.lds.S
+++ b/arch/xtensa/kernel/vmlinux.lds.S
@@ -165,6 +165,13 @@ SECTIONS
 		   .DoubleExceptionVector.text);
     RELOCATE_ENTRY(_DebugInterruptVector_text,
 		   .DebugInterruptVector.text);
+#if defined(CONFIG_SMP)
+    RELOCATE_ENTRY(_SecondaryResetVector_literal,
+		   .SecondaryResetVector.literal);
+    RELOCATE_ENTRY(_SecondaryResetVector_text,
+		   .SecondaryResetVector.text);
+#endif
+
   
     __boot_reloc_table_end = ABSOLUTE(.) ;
 
@@ -272,6 +279,25 @@ SECTIONS
 		  .DoubleExceptionVector.literal)
 
   . = (LOADADDR( .DoubleExceptionVector.text ) + SIZEOF( .DoubleExceptionVector.text ) + 3) & ~ 3;
+
+#if defined(CONFIG_SMP)
+
+  SECTION_VECTOR (_SecondaryResetVector_literal,
+		  .SecondaryResetVector.literal,
+		  RESET_VECTOR1_VADDR - 4,
+		  SIZEOF(.DoubleExceptionVector.text),
+		  .DoubleExceptionVector.text)
+
+  SECTION_VECTOR (_SecondaryResetVector_text,
+		  .SecondaryResetVector.text,
+		  RESET_VECTOR1_VADDR,
+		  4,
+		  .SecondaryResetVector.literal)
+
+  . = LOADADDR(.SecondaryResetVector.text)+SIZEOF(.SecondaryResetVector.text);
+
+#endif
+
   . = ALIGN(PAGE_SIZE);
 
   __init_end = .;
diff --git a/arch/xtensa/mm/cache.c b/arch/xtensa/mm/cache.c
index 81edeab..ba4c47f 100644
--- a/arch/xtensa/mm/cache.c
+++ b/arch/xtensa/mm/cache.c
@@ -118,7 +118,7 @@ void flush_dcache_page(struct page *page)
  * For now, flush the whole cache. FIXME??
  */
 
-void flush_cache_range(struct vm_area_struct* vma,
+void local_flush_cache_range(struct vm_area_struct *vma,
 		       unsigned long start, unsigned long end)
 {
 	__flush_invalidate_dcache_all();
@@ -132,7 +132,7 @@ void flush_cache_range(struct vm_area_struct* vma,
  * alias versions of the cache flush functions.
  */
 
-void flush_cache_page(struct vm_area_struct* vma, unsigned long address,
+void local_flush_cache_page(struct vm_area_struct *vma, unsigned long address,
 		      unsigned long pfn)
 {
 	/* Note that we have to use the 'alias' address to avoid multi-hit */
@@ -159,8 +159,7 @@ update_mmu_cache(struct vm_area_struct * vma, unsigned long addr, pte_t *ptep)
 
 	/* Invalidate old entry in TLBs */
 
-	invalidate_itlb_mapping(addr);
-	invalidate_dtlb_mapping(addr);
+	flush_tlb_page(vma, addr);
 
 #if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK
 
diff --git a/arch/xtensa/mm/fault.c b/arch/xtensa/mm/fault.c
index 70fa7bc..b57c4f9 100644
--- a/arch/xtensa/mm/fault.c
+++ b/arch/xtensa/mm/fault.c
@@ -21,7 +21,7 @@
 #include <asm/uaccess.h>
 #include <asm/pgalloc.h>
 
-unsigned long asid_cache = ASID_USER_FIRST;
+DEFINE_PER_CPU(unsigned long, asid_cache) = ASID_USER_FIRST;
 void bad_page_fault(struct pt_regs*, unsigned long, int);
 
 #undef DEBUG_PAGE_FAULT
diff --git a/arch/xtensa/mm/mmu.c b/arch/xtensa/mm/mmu.c
index a107757..3fdd69a 100644
--- a/arch/xtensa/mm/mmu.c
+++ b/arch/xtensa/mm/mmu.c
@@ -22,7 +22,7 @@ void __init paging_init(void)
 /*
  * Flush the mmu and reset associated register to default values.
  */
-void __init init_mmu(void)
+void init_mmu(void)
 {
 #if !(XCHAL_HAVE_PTP_MMU && XCHAL_HAVE_SPANNING_WAY)
 	/*
@@ -37,7 +37,7 @@ void __init init_mmu(void)
 	set_itlbcfg_register(0);
 	set_dtlbcfg_register(0);
 #endif
-	flush_tlb_all();
+	local_flush_tlb_all();
 
 	/* Set rasid register to a known value. */
 
diff --git a/arch/xtensa/mm/tlb.c b/arch/xtensa/mm/tlb.c
index ca9d236..ade62382 100644
--- a/arch/xtensa/mm/tlb.c
+++ b/arch/xtensa/mm/tlb.c
@@ -48,7 +48,7 @@ static inline void __flush_dtlb_all (void)
 }
 
 
-void flush_tlb_all (void)
+void local_flush_tlb_all(void)
 {
 	__flush_itlb_all();
 	__flush_dtlb_all();
@@ -60,19 +60,23 @@ void flush_tlb_all (void)
  * a new context will be assigned to it.
  */
 
-void flush_tlb_mm(struct mm_struct *mm)
+void local_flush_tlb_mm(struct mm_struct *mm)
 {
+	int cpu = smp_processor_id();
+
 	if (mm == current->active_mm) {
 		unsigned long flags;
 		local_irq_save(flags);
-		__get_new_mmu_context(mm);
-		__load_mmu_context(mm);
+		mm->context.asid[cpu] = NO_CONTEXT;
+		activate_context(mm, cpu);
 		local_irq_restore(flags);
+	} else {
+		mm->context.asid[cpu] = NO_CONTEXT;
+		mm->context.cpu = -1;
 	}
-	else
-		mm->context = 0;
 }
 
+
 #define _ITLB_ENTRIES (ITLB_ARF_WAYS << XCHAL_ITLB_ARF_ENTRIES_LOG2)
 #define _DTLB_ENTRIES (DTLB_ARF_WAYS << XCHAL_DTLB_ARF_ENTRIES_LOG2)
 #if _ITLB_ENTRIES > _DTLB_ENTRIES
@@ -81,24 +85,26 @@ void flush_tlb_mm(struct mm_struct *mm)
 # define _TLB_ENTRIES _DTLB_ENTRIES
 #endif
 
-void flush_tlb_range (struct vm_area_struct *vma,
-		      unsigned long start, unsigned long end)
+void local_flush_tlb_range(struct vm_area_struct *vma,
+		unsigned long start, unsigned long end)
 {
+	int cpu = smp_processor_id();
 	struct mm_struct *mm = vma->vm_mm;
 	unsigned long flags;
 
-	if (mm->context == NO_CONTEXT)
+	if (mm->context.asid[cpu] == NO_CONTEXT)
 		return;
 
 #if 0
 	printk("[tlbrange<%02lx,%08lx,%08lx>]\n",
-			(unsigned long)mm->context, start, end);
+			(unsigned long)mm->context.asid[cpu], start, end);
 #endif
 	local_irq_save(flags);
 
 	if (end-start + (PAGE_SIZE-1) <= _TLB_ENTRIES << PAGE_SHIFT) {
 		int oldpid = get_rasid_register();
-		set_rasid_register (ASID_INSERT(mm->context));
+
+		set_rasid_register(ASID_INSERT(mm->context.asid[cpu]));
 		start &= PAGE_MASK;
 		if (vma->vm_flags & VM_EXEC)
 			while(start < end) {
@@ -114,24 +120,25 @@ void flush_tlb_range (struct vm_area_struct *vma,
 
 		set_rasid_register(oldpid);
 	} else {
-		flush_tlb_mm(mm);
+		local_flush_tlb_mm(mm);
 	}
 	local_irq_restore(flags);
 }
 
-void flush_tlb_page (struct vm_area_struct *vma, unsigned long page)
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
 {
+	int cpu = smp_processor_id();
 	struct mm_struct* mm = vma->vm_mm;
 	unsigned long flags;
 	int oldpid;
 
-	if(mm->context == NO_CONTEXT)
+	if (mm->context.asid[cpu] == NO_CONTEXT)
 		return;
 
 	local_irq_save(flags);
 
 	oldpid = get_rasid_register();
-	set_rasid_register(ASID_INSERT(mm->context));
+	set_rasid_register(ASID_INSERT(mm->context.asid[cpu]));
 
 	if (vma->vm_flags & VM_EXEC)
 		invalidate_itlb_mapping(page);
diff --git a/arch/xtensa/platforms/xtfpga/include/platform/hardware.h b/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
index 4b43ff1..aeb316b 100644
--- a/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
+++ b/arch/xtensa/platforms/xtfpga/include/platform/hardware.h
@@ -26,7 +26,7 @@
 
 /* Default assignment of LX60 devices to external interrupts. */
 
-#ifdef CONFIG_ARCH_HAS_SMP
+#ifdef CONFIG_XTENSA_MX
 #define DUART16552_INTNUM	XCHAL_EXTINT3_NUM
 #define OETH_IRQ		XCHAL_EXTINT4_NUM
 #else
-- 
1.8.1.4

[PATCH 16/17] xtensa: add Three Core HiFi-2 MX Variant.

[Max Filippov]

From: Piet Delaney <piet.delaney@gmail.com>

This variant is equipped with MX and is capable of running SMP linux.

Signed-off-by: Piet Delaney <piet.delaney@gmail.com>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/Kconfig                                |   7 +
 arch/xtensa/Makefile                               |   1 +
 .../test_mmuhifi_c3/include/variant/core.h         | 384 +++++++++++++++++++++
 .../test_mmuhifi_c3/include/variant/tie-asm.h      | 183 ++++++++++
 .../variants/test_mmuhifi_c3/include/variant/tie.h | 141 ++++++++
 5 files changed, 716 insertions(+)
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
 create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index 4b09c60..9cfb514 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -89,6 +89,13 @@ config XTENSA_VARIANT_DC233C
 	help
 	  This variant refers to Tensilica's Diamond 233L Standard core Rev.C (LE).
 
+config XTENSA_VARIANT_TEST_MMUHIFI_C3
+	bool "test_mmuhifi_c3 - Experimental 3 Core SMP HiFi-2 with a V2 MMU; Extension of a Diamond DC_D_330HiFi"
+	select MMU
+	select MAY_HAVE_SMP
+	help
+	  This variant refers to Tensilica's Diamond 330HiFi 3 core SMP cluster.
+
 config XTENSA_VARIANT_S6000
 	bool "s6000 - Stretch software configurable processor"
 	select VARIANT_IRQ_SWITCH
diff --git a/arch/xtensa/Makefile b/arch/xtensa/Makefile
index 81250ec..8edd7c0 100644
--- a/arch/xtensa/Makefile
+++ b/arch/xtensa/Makefile
@@ -16,6 +16,7 @@
 variant-$(CONFIG_XTENSA_VARIANT_FSF)		:= fsf
 variant-$(CONFIG_XTENSA_VARIANT_DC232B)		:= dc232b
 variant-$(CONFIG_XTENSA_VARIANT_DC233C)		:= dc233c
+variant-$(CONFIG_XTENSA_VARIANT_TEST_MMUHIFI_C3) := test_mmuhifi_c3
 variant-$(CONFIG_XTENSA_VARIANT_S6000)		:= s6000
 variant-$(CONFIG_XTENSA_VARIANT_LINUX_CUSTOM)	:= custom
 
diff --git a/arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
new file mode 100644
index 0000000..309caa1
--- /dev/null
+++ b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
@@ -0,0 +1,384 @@
+/*
+ * Xtensa processor core configuration information.
+ *
+ * This file is subject to the terms and conditions of version 2.1 of the GNU
+ * Lesser General Public License as published by the Free Software Foundation.
+ *
+ * Copyright (c) 1999-2009 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_CORE_CONFIGURATION_H
+#define _XTENSA_CORE_CONFIGURATION_H
+
+
+/****************************************************************************
+	    Parameters Useful for Any Code, USER or PRIVILEGED
+ ****************************************************************************/
+
+/*
+ *  Note:  Macros of the form XCHAL_HAVE_*** have a value of 1 if the option is
+ *  configured, and a value of 0 otherwise.  These macros are always defined.
+ */
+
+
+/*----------------------------------------------------------------------
+				ISA
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_HAVE_BE			0	/* big-endian byte ordering */
+#define XCHAL_HAVE_WINDOWED		1	/* windowed registers option */
+#define XCHAL_NUM_AREGS			32	/* num of physical addr regs */
+#define XCHAL_NUM_AREGS_LOG2		5	/* log2(XCHAL_NUM_AREGS) */
+#define XCHAL_MAX_INSTRUCTION_SIZE	8	/* max instr bytes (3..8) */
+#define XCHAL_HAVE_DEBUG		1	/* debug option */
+#define XCHAL_HAVE_DENSITY		1	/* 16-bit instructions */
+#define XCHAL_HAVE_LOOPS		1	/* zero-overhead loops */
+#define XCHAL_HAVE_NSA			1	/* NSA/NSAU instructions */
+#define XCHAL_HAVE_MINMAX		1	/* MIN/MAX instructions */
+#define XCHAL_HAVE_SEXT			1	/* SEXT instruction */
+#define XCHAL_HAVE_CLAMPS		1	/* CLAMPS instruction */
+#define XCHAL_HAVE_MUL16		1	/* MUL16S/MUL16U instructions */
+#define XCHAL_HAVE_MUL32		1	/* MULL instruction */
+#define XCHAL_HAVE_MUL32_HIGH		0	/* MULUH/MULSH instructions */
+#define XCHAL_HAVE_DIV32		0	/* QUOS/QUOU/REMS/REMU instructions */
+#define XCHAL_HAVE_L32R			1	/* L32R instruction */
+#define XCHAL_HAVE_ABSOLUTE_LITERALS	1	/* non-PC-rel (extended) L32R */
+#define XCHAL_HAVE_CONST16		0	/* CONST16 instruction */
+#define XCHAL_HAVE_ADDX			1	/* ADDX#/SUBX# instructions */
+#define XCHAL_HAVE_WIDE_BRANCHES	0	/* B*.W18 or B*.W15 instr's */
+#define XCHAL_HAVE_PREDICTED_BRANCHES	0	/* B[EQ/EQZ/NE/NEZ]T instr's */
+#define XCHAL_HAVE_CALL4AND12		1	/* (obsolete option) */
+#define XCHAL_HAVE_ABS			1	/* ABS instruction */
+/*#define XCHAL_HAVE_POPC		0*/	/* POPC instruction */
+/*#define XCHAL_HAVE_CRC		0*/	/* CRC instruction */
+#define XCHAL_HAVE_RELEASE_SYNC		1	/* L32AI/S32RI instructions */
+#define XCHAL_HAVE_S32C1I		1	/* S32C1I instruction */
+#define XCHAL_HAVE_SPECULATION		0	/* speculation */
+#define XCHAL_HAVE_FULL_RESET		1	/* all regs/state reset */
+#define XCHAL_NUM_CONTEXTS		1	/* */
+#define XCHAL_NUM_MISC_REGS		2	/* num of scratch regs (0..4) */
+#define XCHAL_HAVE_TAP_MASTER		0	/* JTAG TAP control instr's */
+#define XCHAL_HAVE_PRID			1	/* processor ID register */
+#define XCHAL_HAVE_EXTERN_REGS		1	/* WER/RER instructions */
+#define XCHAL_HAVE_MP_INTERRUPTS	1	/* interrupt distributor port */
+#define XCHAL_HAVE_MP_RUNSTALL		1	/* core RunStall control port */
+#define XCHAL_HAVE_THREADPTR		1	/* THREADPTR register */
+#define XCHAL_HAVE_BOOLEANS		1	/* boolean registers */
+#define XCHAL_HAVE_CP			1	/* CPENABLE reg (coprocessor) */
+#define XCHAL_CP_MAXCFG			2	/* max allowed cp id plus one */
+#define XCHAL_HAVE_MAC16		0	/* MAC16 package */
+#define XCHAL_HAVE_VECTORFPU2005	0	/* vector floating-point pkg */
+#define XCHAL_HAVE_FP			0	/* floating point pkg */
+#define XCHAL_HAVE_DFP			0	/* double precision FP pkg */
+#define XCHAL_HAVE_DFP_accel		0	/* double precision FP acceleration pkg */
+#define XCHAL_HAVE_VECTRA1		0	/* Vectra I  pkg */
+#define XCHAL_HAVE_VECTRALX		0	/* Vectra LX pkg */
+#define XCHAL_HAVE_HIFIPRO		0	/* HiFiPro Audio Engine pkg */
+#define XCHAL_HAVE_HIFI2		1	/* HiFi2 Audio Engine pkg */
+#define XCHAL_HAVE_CONNXD2		0	/* ConnX D2 pkg */
+
+
+/*----------------------------------------------------------------------
+				MISC
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_NUM_WRITEBUFFER_ENTRIES	8	/* size of write buffer */
+#define XCHAL_INST_FETCH_WIDTH		8	/* instr-fetch width in bytes */
+#define XCHAL_DATA_WIDTH		8	/* data width in bytes */
+/*  In T1050, applies to selected core load and store instructions (see ISA): */
+#define XCHAL_UNALIGNED_LOAD_EXCEPTION	1	/* unaligned loads cause exc. */
+#define XCHAL_UNALIGNED_STORE_EXCEPTION	1	/* unaligned stores cause exc.*/
+#define XCHAL_UNALIGNED_LOAD_HW		0	/* unaligned loads work in hw */
+#define XCHAL_UNALIGNED_STORE_HW	0	/* unaligned stores work in hw*/
+
+#define XCHAL_SW_VERSION		800000	/* sw version of this header */
+
+#define XCHAL_CORE_ID			"test_mmuhifi_c3"	/* alphanum core name
+						   (CoreID) set in the Xtensa
+						   Processor Generator */
+
+#define XCHAL_CORE_DESCRIPTION		"test_mmuhifi_c3"
+#define XCHAL_BUILD_UNIQUE_ID		0x00005A6A	/* 22-bit sw build ID */
+
+/*
+ *  These definitions describe the hardware targeted by this software.
+ */
+#define XCHAL_HW_CONFIGID0		0xC1B3CBFE	/* ConfigID hi 32 bits*/
+#define XCHAL_HW_CONFIGID1		0x10405A6A	/* ConfigID lo 32 bits*/
+#define XCHAL_HW_VERSION_NAME		"LX3.0.0"	/* full version name */
+#define XCHAL_HW_VERSION_MAJOR		2300	/* major ver# of targeted hw */
+#define XCHAL_HW_VERSION_MINOR		0	/* minor ver# of targeted hw */
+#define XCHAL_HW_VERSION		230000	/* major*100+minor */
+#define XCHAL_HW_REL_LX3		1
+#define XCHAL_HW_REL_LX3_0		1
+#define XCHAL_HW_REL_LX3_0_0		1
+#define XCHAL_HW_CONFIGID_RELIABLE	1
+/*  If software targets a *range* of hardware versions, these are the bounds: */
+#define XCHAL_HW_MIN_VERSION_MAJOR	2300	/* major v of earliest tgt hw */
+#define XCHAL_HW_MIN_VERSION_MINOR	0	/* minor v of earliest tgt hw */
+#define XCHAL_HW_MIN_VERSION		230000	/* earliest targeted hw */
+#define XCHAL_HW_MAX_VERSION_MAJOR	2300	/* major v of latest tgt hw */
+#define XCHAL_HW_MAX_VERSION_MINOR	0	/* minor v of latest tgt hw */
+#define XCHAL_HW_MAX_VERSION		230000	/* latest targeted hw */
+
+
+/*----------------------------------------------------------------------
+				CACHE
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_ICACHE_LINESIZE		32	/* I-cache line size in bytes */
+#define XCHAL_DCACHE_LINESIZE		32	/* D-cache line size in bytes */
+#define XCHAL_ICACHE_LINEWIDTH		5	/* log2(I line size in bytes) */
+#define XCHAL_DCACHE_LINEWIDTH		5	/* log2(D line size in bytes) */
+
+#define XCHAL_ICACHE_SIZE		16384	/* I-cache size in bytes or 0 */
+#define XCHAL_DCACHE_SIZE		16384	/* D-cache size in bytes or 0 */
+
+#define XCHAL_DCACHE_IS_WRITEBACK	1	/* writeback feature */
+#define XCHAL_DCACHE_IS_COHERENT	1	/* MP coherence feature */
+
+
+
+
+/****************************************************************************
+    Parameters Useful for PRIVILEGED (Supervisory or Non-Virtualized) Code
+ ****************************************************************************/
+
+
+#ifndef XTENSA_HAL_NON_PRIVILEGED_ONLY
+
+/*----------------------------------------------------------------------
+				CACHE
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_HAVE_PIF			1	/* any outbound PIF present */
+
+/*  If present, cache size in bytes == (ways * 2^(linewidth + setwidth)).  */
+
+/*  Number of cache sets in log2(lines per way):  */
+#define XCHAL_ICACHE_SETWIDTH		8
+#define XCHAL_DCACHE_SETWIDTH		8
+
+/*  Cache set associativity (number of ways):  */
+#define XCHAL_ICACHE_WAYS		2
+#define XCHAL_DCACHE_WAYS		2
+
+/*  Cache features:  */
+#define XCHAL_ICACHE_LINE_LOCKABLE	0
+#define XCHAL_DCACHE_LINE_LOCKABLE	0
+#define XCHAL_ICACHE_ECC_PARITY		0
+#define XCHAL_DCACHE_ECC_PARITY		0
+
+/*  Cache access size in bytes (affects operation of SICW instruction):  */
+#define XCHAL_ICACHE_ACCESS_SIZE	8
+#define XCHAL_DCACHE_ACCESS_SIZE	8
+
+/*  Number of encoded cache attr bits (see <xtensa/hal.h> for decoded bits):  */
+#define XCHAL_CA_BITS			4
+
+
+/*----------------------------------------------------------------------
+			INTERNAL I/D RAM/ROMs and XLMI
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_NUM_INSTROM		0	/* number of core instr. ROMs */
+#define XCHAL_NUM_INSTRAM		0	/* number of core instr. RAMs */
+#define XCHAL_NUM_DATAROM		0	/* number of core data ROMs */
+#define XCHAL_NUM_DATARAM		0	/* number of core data RAMs */
+#define XCHAL_NUM_URAM			0	/* number of core unified RAMs*/
+#define XCHAL_NUM_XLMI			0	/* number of core XLMI ports */
+
+
+/*----------------------------------------------------------------------
+			INTERRUPTS and TIMERS
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_HAVE_INTERRUPTS		1	/* interrupt option */
+#define XCHAL_HAVE_HIGHPRI_INTERRUPTS	1	/* med/high-pri. interrupts */
+#define XCHAL_HAVE_NMI			0	/* non-maskable interrupt */
+#define XCHAL_HAVE_CCOUNT		1	/* CCOUNT reg. (timer option) */
+#define XCHAL_NUM_TIMERS		2	/* number of CCOMPAREn regs */
+#define XCHAL_NUM_INTERRUPTS		12	/* number of interrupts */
+#define XCHAL_NUM_INTERRUPTS_LOG2	4	/* ceil(log2(NUM_INTERRUPTS)) */
+#define XCHAL_NUM_EXTINTERRUPTS		9	/* num of external interrupts */
+#define XCHAL_NUM_INTLEVELS		2	/* number of interrupt levels
+						   (not including level zero) */
+#define XCHAL_EXCM_LEVEL		1	/* level masked by PS.EXCM */
+	/* (always 1 in XEA1; levels 2 .. EXCM_LEVEL are "medium priority") */
+
+/*  Masks of interrupts at each interrupt level:  */
+#define XCHAL_INTLEVEL1_MASK		0x00000FFF
+#define XCHAL_INTLEVEL2_MASK		0x00000000
+#define XCHAL_INTLEVEL3_MASK		0x00000000
+#define XCHAL_INTLEVEL4_MASK		0x00000000
+#define XCHAL_INTLEVEL5_MASK		0x00000000
+#define XCHAL_INTLEVEL6_MASK		0x00000000
+#define XCHAL_INTLEVEL7_MASK		0x00000000
+
+/*  Masks of interrupts at each range 1..n of interrupt levels:  */
+#define XCHAL_INTLEVEL1_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL2_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL3_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL4_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL5_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL6_ANDBELOW_MASK	0x00000FFF
+#define XCHAL_INTLEVEL7_ANDBELOW_MASK	0x00000FFF
+
+/*  Level of each interrupt:  */
+#define XCHAL_INT0_LEVEL		1
+#define XCHAL_INT1_LEVEL		1
+#define XCHAL_INT2_LEVEL		1
+#define XCHAL_INT3_LEVEL		1
+#define XCHAL_INT4_LEVEL		1
+#define XCHAL_INT5_LEVEL		1
+#define XCHAL_INT6_LEVEL		1
+#define XCHAL_INT7_LEVEL		1
+#define XCHAL_INT8_LEVEL		1
+#define XCHAL_INT9_LEVEL		1
+#define XCHAL_INT10_LEVEL		1
+#define XCHAL_INT11_LEVEL		1
+#define XCHAL_DEBUGLEVEL		2	/* debug interrupt level */
+#define XCHAL_HAVE_DEBUG_EXTERN_INT	1	/* OCD external db interrupt */
+
+/*  Type of each interrupt:  */
+#define XCHAL_INT0_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT1_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT2_TYPE 	XTHAL_INTTYPE_EXTERN_EDGE
+#define XCHAL_INT3_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT4_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT5_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT6_TYPE 	XTHAL_INTTYPE_TIMER
+#define XCHAL_INT7_TYPE 	XTHAL_INTTYPE_SOFTWARE
+#define XCHAL_INT8_TYPE 	XTHAL_INTTYPE_TIMER
+#define XCHAL_INT9_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT10_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+#define XCHAL_INT11_TYPE 	XTHAL_INTTYPE_EXTERN_LEVEL
+
+/*  Masks of interrupts for each type of interrupt:  */
+#define XCHAL_INTTYPE_MASK_UNCONFIGURED	0xFFFFF000
+#define XCHAL_INTTYPE_MASK_SOFTWARE	0x00000080
+#define XCHAL_INTTYPE_MASK_EXTERN_EDGE	0x00000004
+#define XCHAL_INTTYPE_MASK_EXTERN_LEVEL	0x00000E3B
+#define XCHAL_INTTYPE_MASK_TIMER	0x00000140
+#define XCHAL_INTTYPE_MASK_NMI		0x00000000
+#define XCHAL_INTTYPE_MASK_WRITE_ERROR	0x00000000
+
+/*  Interrupt numbers assigned to specific interrupt sources:  */
+#define XCHAL_TIMER0_INTERRUPT		6	/* CCOMPARE0 */
+#define XCHAL_TIMER1_INTERRUPT		8	/* CCOMPARE1 */
+#define XCHAL_TIMER2_INTERRUPT		XTHAL_TIMER_UNCONFIGURED
+#define XCHAL_TIMER3_INTERRUPT		XTHAL_TIMER_UNCONFIGURED
+
+/*  Interrupt numbers for levels at which only one interrupt is configured:  */
+/*  (There are many interrupts each at level(s) 1.)  */
+
+
+/*
+ *  External interrupt vectors/levels.
+ *  These macros describe how Xtensa processor interrupt numbers
+ *  (as numbered internally, eg. in INTERRUPT and INTENABLE registers)
+ *  map to external BInterrupt<n> pins, for those interrupts
+ *  configured as external (level-triggered, edge-triggered, or NMI).
+ *  See the Xtensa processor databook for more details.
+ */
+
+/*  Core interrupt numbers mapped to each EXTERNAL interrupt number:  */
+#define XCHAL_EXTINT0_NUM		0	/* (intlevel 1) */
+#define XCHAL_EXTINT1_NUM		1	/* (intlevel 1) */
+#define XCHAL_EXTINT2_NUM		2	/* (intlevel 1) */
+#define XCHAL_EXTINT3_NUM		3	/* (intlevel 1) */
+#define XCHAL_EXTINT4_NUM		4	/* (intlevel 1) */
+#define XCHAL_EXTINT5_NUM		5	/* (intlevel 1) */
+#define XCHAL_EXTINT6_NUM		9	/* (intlevel 1) */
+#define XCHAL_EXTINT7_NUM		10	/* (intlevel 1) */
+#define XCHAL_EXTINT8_NUM		11	/* (intlevel 1) */
+
+
+/*----------------------------------------------------------------------
+			EXCEPTIONS and VECTORS
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_XEA_VERSION		2	/* Xtensa Exception Architecture
+						   number: 1 == XEA1 (old)
+							   2 == XEA2 (new)
+							   0 == XEAX (extern) */
+#define XCHAL_HAVE_XEA1			0	/* Exception Architecture 1 */
+#define XCHAL_HAVE_XEA2			1	/* Exception Architecture 2 */
+#define XCHAL_HAVE_XEAX			0	/* External Exception Arch. */
+#define XCHAL_HAVE_EXCEPTIONS		1	/* exception option */
+#define XCHAL_HAVE_MEM_ECC_PARITY	0	/* local memory ECC/parity */
+#define XCHAL_HAVE_VECTOR_SELECT	1	/* relocatable vectors */
+#define XCHAL_HAVE_VECBASE		1	/* relocatable vectors */
+#define XCHAL_VECBASE_RESET_VADDR	0xD0000000  /* VECBASE reset value */
+#define XCHAL_VECBASE_RESET_PADDR	0x00000000
+#define XCHAL_RESET_VECBASE_OVERLAP	0
+
+#define XCHAL_RESET_VECTOR0_VADDR	0xFE000000
+#define XCHAL_RESET_VECTOR0_PADDR	0xFE000000
+#define XCHAL_RESET_VECTOR1_VADDR	0xD8000500
+#define XCHAL_RESET_VECTOR1_PADDR	0x00000500
+#define XCHAL_RESET_VECTOR_VADDR	0xFE000000
+#define XCHAL_RESET_VECTOR_PADDR	0xFE000000
+#define XCHAL_USER_VECOFS		0x00000340
+#define XCHAL_USER_VECTOR_VADDR		0xD0000340
+#define XCHAL_USER_VECTOR_PADDR		0x00000340
+#define XCHAL_KERNEL_VECOFS		0x00000300
+#define XCHAL_KERNEL_VECTOR_VADDR	0xD0000300
+#define XCHAL_KERNEL_VECTOR_PADDR	0x00000300
+#define XCHAL_DOUBLEEXC_VECOFS		0x000003C0
+#define XCHAL_DOUBLEEXC_VECTOR_VADDR	0xD00003C0
+#define XCHAL_DOUBLEEXC_VECTOR_PADDR	0x000003C0
+#define XCHAL_WINDOW_OF4_VECOFS		0x00000000
+#define XCHAL_WINDOW_UF4_VECOFS		0x00000040
+#define XCHAL_WINDOW_OF8_VECOFS		0x00000080
+#define XCHAL_WINDOW_UF8_VECOFS		0x000000C0
+#define XCHAL_WINDOW_OF12_VECOFS	0x00000100
+#define XCHAL_WINDOW_UF12_VECOFS	0x00000140
+#define XCHAL_WINDOW_VECTORS_VADDR	0xD0000000
+#define XCHAL_WINDOW_VECTORS_PADDR	0x00000000
+#define XCHAL_INTLEVEL2_VECOFS		0x00000280
+#define XCHAL_INTLEVEL2_VECTOR_VADDR	0xD0000280
+#define XCHAL_INTLEVEL2_VECTOR_PADDR	0x00000280
+#define XCHAL_DEBUG_VECOFS		XCHAL_INTLEVEL2_VECOFS
+#define XCHAL_DEBUG_VECTOR_VADDR	XCHAL_INTLEVEL2_VECTOR_VADDR
+#define XCHAL_DEBUG_VECTOR_PADDR	XCHAL_INTLEVEL2_VECTOR_PADDR
+
+
+/*----------------------------------------------------------------------
+				DEBUG
+  ----------------------------------------------------------------------*/
+
+#define XCHAL_HAVE_OCD			1	/* OnChipDebug option */
+#define XCHAL_NUM_IBREAK		0	/* number of IBREAKn regs */
+#define XCHAL_NUM_DBREAK		0	/* number of DBREAKn regs */
+#define XCHAL_HAVE_OCD_DIR_ARRAY	0	/* faster OCD option */
+
+
+/*----------------------------------------------------------------------
+				MMU
+  ----------------------------------------------------------------------*/
+
+/*  See core-matmap.h header file for more details.  */
+
+#define XCHAL_HAVE_TLBS			1	/* inverse of HAVE_CACHEATTR */
+#define XCHAL_HAVE_SPANNING_WAY		0	/* one way maps I+D 4GB vaddr */
+#define XCHAL_HAVE_IDENTITY_MAP		0	/* vaddr == paddr always */
+#define XCHAL_HAVE_CACHEATTR		0	/* CACHEATTR register present */
+#define XCHAL_HAVE_MIMIC_CACHEATTR	0	/* region protection */
+#define XCHAL_HAVE_XLT_CACHEATTR	0	/* region prot. w/translation */
+#define XCHAL_HAVE_PTP_MMU		1	/* full MMU (with page table
+						   [autorefill] and protection)
+						   usable for an MMU-based OS */
+/*  If none of the above last 4 are set, it's a custom TLB configuration.  */
+#define XCHAL_ITLB_ARF_ENTRIES_LOG2	2	/* log2(autorefill way size) */
+#define XCHAL_DTLB_ARF_ENTRIES_LOG2	2	/* log2(autorefill way size) */
+
+#define XCHAL_MMU_ASID_BITS		8	/* number of bits in ASIDs */
+#define XCHAL_MMU_RINGS			4	/* number of rings (1..4) */
+#define XCHAL_MMU_RING_BITS		2	/* num of bits in RING field */
+
+#endif /* !XTENSA_HAL_NON_PRIVILEGED_ONLY */
+
+
+#endif /* _XTENSA_CORE_CONFIGURATION_H */
+
diff --git a/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
new file mode 100644
index 0000000..ffade04
--- /dev/null
+++ b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
@@ -0,0 +1,183 @@
+/*
+ * This header file contains assembly-language definitions (assembly
+ * macros, etc.) for this specific Xtensa processor's TIE extensions
+ * and options.  It is customized to this Xtensa processor configuration.
+ *
+ * This file is subject to the terms and conditions of version 2.1 of the GNU
+ * Lesser General Public License as published by the Free Software Foundation.
+ *
+ * Copyright (C) 1999-2009 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_CORE_TIE_ASM_H
+#define _XTENSA_CORE_TIE_ASM_H
+
+/*  Selection parameter values for save-area save/restore macros:  */
+/*  Option vs. TIE:  */
+#define XTHAL_SAS_TIE	0x0001	/* custom extension or coprocessor */
+#define XTHAL_SAS_OPT	0x0002	/* optional (and not a coprocessor) */
+/*  Whether used automatically by compiler:  */
+#define XTHAL_SAS_NOCC	0x0004	/* not used by compiler w/o special opts/code */
+#define XTHAL_SAS_CC	0x0008	/* used by compiler without special opts/code */
+/*  ABI handling across function calls:  */
+#define XTHAL_SAS_CALR	0x0010	/* caller-saved */
+#define XTHAL_SAS_CALE	0x0020	/* callee-saved */
+#define XTHAL_SAS_GLOB	0x0040	/* global across function calls (in thread) */
+/*  Misc  */
+#define XTHAL_SAS_ALL	0xFFFF	/* include all default NCP contents */
+
+
+
+/* Macro to save all non-coprocessor (extra) custom TIE and optional state
+ * (not including zero-overhead loop registers).
+ * Save area ptr (clobbered):  ptr  (8 byte aligned)
+ * Scratch regs  (clobbered):  at1..at4  (only first XCHAL_NCP_NUM_ATMPS needed)
+ */
+	.macro xchal_ncp_store  ptr at1 at2 at3 at4  continue=0 ofs=-1 select=XTHAL_SAS_ALL
+	xchal_sa_start	\continue, \ofs
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	rsr	\at1, BR		// boolean option
+	s32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	rsr	\at1, SCOMPARE1		// conditional store option
+	s32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_CC | XTHAL_SAS_GLOB) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	rur	\at1, THREADPTR		// threadptr option
+	s32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.endm	// xchal_ncp_store
+
+/* Macro to save all non-coprocessor (extra) custom TIE and optional state
+ * (not including zero-overhead loop registers).
+ * Save area ptr (clobbered):  ptr  (8 byte aligned)
+ * Scratch regs  (clobbered):  at1..at4  (only first XCHAL_NCP_NUM_ATMPS needed)
+ */
+	.macro xchal_ncp_load  ptr at1 at2 at3 at4  continue=0 ofs=-1 select=XTHAL_SAS_ALL
+	xchal_sa_start	\continue, \ofs
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	l32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	wsr	\at1, BR		// boolean option
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	l32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	wsr	\at1, SCOMPARE1		// conditional store option
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.ifeq (XTHAL_SAS_OPT | XTHAL_SAS_CC | XTHAL_SAS_GLOB) & ~\select
+	xchal_sa_align	\ptr, 0, 1024-4, 4, 4
+	l32i	\at1, \ptr, .Lxchal_ofs_ + 0
+	wur	\at1, THREADPTR		// threadptr option
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 4
+	.endif
+	.endm	// xchal_ncp_load
+
+
+
+#define XCHAL_NCP_NUM_ATMPS	1
+
+
+
+/* Macro to save the state of TIE coprocessor AudioEngineLX.
+ * Save area ptr (clobbered):  ptr  (8 byte aligned)
+ * Scratch regs  (clobbered):  at1..at4  (only first XCHAL_CP1_NUM_ATMPS needed)
+ */
+#define xchal_cp_AudioEngineLX_store	xchal_cp1_store
+/* #define xchal_cp_AudioEngineLX_store_a2	xchal_cp1_store a2 a3 a4 a5 a6 */
+	.macro	xchal_cp1_store  ptr at1 at2 at3 at4  continue=0 ofs=-1 select=XTHAL_SAS_ALL
+	xchal_sa_start \continue, \ofs
+	.ifeq (XTHAL_SAS_TIE | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 0, 1, 8
+	rur240	\at1		// AE_OVF_SAR
+	s32i	\at1, \ptr, 0
+	rur241	\at1		// AE_BITHEAD
+	s32i	\at1, \ptr, 4
+	rur242	\at1		// AE_TS_FTS_BU_BP
+	s32i	\at1, \ptr, 8
+	rur243	\at1		// AE_SD_NO
+	s32i	\at1, \ptr, 12
+	AE_SP24X2S.I aep0, \ptr,  16
+	AE_SP24X2S.I aep1, \ptr,  24
+	AE_SP24X2S.I aep2, \ptr,  32
+	AE_SP24X2S.I aep3, \ptr,  40
+	AE_SP24X2S.I aep4, \ptr,  48
+	AE_SP24X2S.I aep5, \ptr,  56
+	addi	\ptr, \ptr, 64
+	AE_SP24X2S.I aep6, \ptr,  0
+	AE_SP24X2S.I aep7, \ptr,  8
+	AE_SQ56S.I aeq0, \ptr,  16
+	AE_SQ56S.I aeq1, \ptr,  24
+	AE_SQ56S.I aeq2, \ptr,  32
+	AE_SQ56S.I aeq3, \ptr,  40
+	.set	.Lxchal_pofs_, .Lxchal_pofs_ + 64
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 112
+	.endif
+	.endm	// xchal_cp1_store
+
+/* Macro to restore the state of TIE coprocessor AudioEngineLX.
+ * Save area ptr (clobbered):  ptr  (8 byte aligned)
+ * Scratch regs  (clobbered):  at1..at4  (only first XCHAL_CP1_NUM_ATMPS needed)
+ */
+#define xchal_cp_AudioEngineLX_load	xchal_cp1_load
+/* #define xchal_cp_AudioEngineLX_load_a2	xchal_cp1_load a2 a3 a4 a5 a6 */
+	.macro	xchal_cp1_load  ptr at1 at2 at3 at4  continue=0 ofs=-1 select=XTHAL_SAS_ALL
+	xchal_sa_start \continue, \ofs
+	.ifeq (XTHAL_SAS_TIE | XTHAL_SAS_NOCC | XTHAL_SAS_CALR) & ~\select
+	xchal_sa_align	\ptr, 0, 0, 1, 8
+	l32i	\at1, \ptr, 0
+	wur240	\at1		// AE_OVF_SAR
+	l32i	\at1, \ptr, 4
+	wur241	\at1		// AE_BITHEAD
+	l32i	\at1, \ptr, 8
+	wur242	\at1		// AE_TS_FTS_BU_BP
+	l32i	\at1, \ptr, 12
+	wur243	\at1		// AE_SD_NO
+	addi	\ptr, \ptr, 80
+	AE_LQ56.I aeq0, \ptr,  0
+	AE_LQ56.I aeq1, \ptr,  8
+	AE_LQ56.I aeq2, \ptr,  16
+	AE_LQ56.I aeq3, \ptr,  24
+	AE_LP24X2.I aep0, \ptr,  -64
+	AE_LP24X2.I aep1, \ptr,  -56
+	AE_LP24X2.I aep2, \ptr,  -48
+	AE_LP24X2.I aep3, \ptr,  -40
+	AE_LP24X2.I aep4, \ptr,  -32
+	AE_LP24X2.I aep5, \ptr,  -24
+	AE_LP24X2.I aep6, \ptr,  -16
+	AE_LP24X2.I aep7, \ptr,  -8
+	.set	.Lxchal_pofs_, .Lxchal_pofs_ + 80
+	.set	.Lxchal_ofs_, .Lxchal_ofs_ + 112
+	.endif
+	.endm	// xchal_cp1_load
+
+#define XCHAL_CP1_NUM_ATMPS	1
+#define XCHAL_SA_NUM_ATMPS	1
+
+	/*  Empty macros for unconfigured coprocessors:  */
+	.macro xchal_cp0_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp0_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp2_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp2_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp3_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp3_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp4_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp4_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp5_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp5_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp6_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp6_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp7_store	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+	.macro xchal_cp7_load	p a b c d continue=0 ofs=-1 select=-1 ; .endm
+
+#endif /*_XTENSA_CORE_TIE_ASM_H*/
+
diff --git a/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h
new file mode 100644
index 0000000..6ec45c0
--- /dev/null
+++ b/arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h
@@ -0,0 +1,141 @@
+/*
+ * This header file describes this specific Xtensa processor's TIE extensions
+ * that extend basic Xtensa core functionality.  It is customized to this
+ * Xtensa processor configuration.
+ *
+ * This file is subject to the terms and conditions of version 2.1 of the GNU
+ * Lesser General Public License as published by the Free Software Foundation.
+ *
+ * Copyright (C) 1999-2009 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_CORE_TIE_H
+#define _XTENSA_CORE_TIE_H
+
+#define XCHAL_CP_NUM			1	/* number of coprocessors */
+#define XCHAL_CP_MAX			2	/* max CP ID + 1 (0 if none) */
+#define XCHAL_CP_MASK			0x02	/* bitmask of all CPs by ID */
+#define XCHAL_CP_PORT_MASK		0x00	/* bitmask of only port CPs */
+
+/*  Basic parameters of each coprocessor:  */
+#define XCHAL_CP1_NAME			"AudioEngineLX"
+#define XCHAL_CP1_IDENT			AudioEngineLX
+#define XCHAL_CP1_SA_SIZE		112	/* size of state save area */
+#define XCHAL_CP1_SA_ALIGN		8	/* min alignment of save area */
+#define XCHAL_CP_ID_AUDIOENGINELX	1	/* coprocessor ID (0..7) */
+
+/*  Filler info for unassigned coprocessors, to simplify arrays etc:  */
+#define XCHAL_CP0_SA_SIZE		0
+#define XCHAL_CP0_SA_ALIGN		1
+#define XCHAL_CP2_SA_SIZE		0
+#define XCHAL_CP2_SA_ALIGN		1
+#define XCHAL_CP3_SA_SIZE		0
+#define XCHAL_CP3_SA_ALIGN		1
+#define XCHAL_CP4_SA_SIZE		0
+#define XCHAL_CP4_SA_ALIGN		1
+#define XCHAL_CP5_SA_SIZE		0
+#define XCHAL_CP5_SA_ALIGN		1
+#define XCHAL_CP6_SA_SIZE		0
+#define XCHAL_CP6_SA_ALIGN		1
+#define XCHAL_CP7_SA_SIZE		0
+#define XCHAL_CP7_SA_ALIGN		1
+
+/*  Save area for non-coprocessor optional and custom (TIE) state:  */
+#define XCHAL_NCP_SA_SIZE		12
+#define XCHAL_NCP_SA_ALIGN		4
+
+/*  Total save area for optional and custom state (NCP + CPn):  */
+#define XCHAL_TOTAL_SA_SIZE		128	/* with 16-byte align padding */
+#define XCHAL_TOTAL_SA_ALIGN		8	/* actual minimum alignment */
+
+/*
+ * Detailed contents of save areas.
+ * NOTE:  caller must define the XCHAL_SA_REG macro (not defined here)
+ * before expanding the XCHAL_xxx_SA_LIST() macros.
+ *
+ * XCHAL_SA_REG(s,ccused,abikind,kind,opt,name,galign,align,asize,
+ *		dbnum,base,regnum,bitsz,gapsz,reset,x...)
+ *
+ *	s = passed from XCHAL_*_LIST(s), eg. to select how to expand
+ *	ccused = set if used by compiler without special options or code
+ *	abikind = 0 (caller-saved), 1 (callee-saved), or 2 (thread-global)
+ *	kind = 0 (special reg), 1 (TIE user reg), or 2 (TIE regfile reg)
+ *	opt = 0 (custom TIE extension or coprocessor), or 1 (optional reg)
+ *	name = lowercase reg name (no quotes)
+ *	galign = group byte alignment (power of 2) (galign >= align)
+ *	align = register byte alignment (power of 2)
+ *	asize = allocated size in bytes (asize*8 == bitsz + gapsz + padsz)
+ *	  (not including any pad bytes required to galign this or next reg)
+ *	dbnum = unique target number f/debug (see <xtensa-libdb-macros.h>)
+ *	base = reg shortname w/o index (or sr=special, ur=TIE user reg)
+ *	regnum = reg index in regfile, or special/TIE-user reg number
+ *	bitsz = number of significant bits (regfile width, or ur/sr mask bits)
+ *	gapsz = intervening bits, if bitsz bits not stored contiguously
+ *	(padsz = pad bits at end [TIE regfile] or at msbits [ur,sr] of asize)
+ *	reset = register reset value (or 0 if undefined at reset)
+ *	x = reserved for future use (0 until then)
+ *
+ *  To filter out certain registers, e.g. to expand only the non-global
+ *  registers used by the compiler, you can do something like this:
+ *
+ *  #define XCHAL_SA_REG(s,ccused,p...)	SELCC##ccused(p)
+ *  #define SELCC0(p...)
+ *  #define SELCC1(abikind,p...)	SELAK##abikind(p)
+ *  #define SELAK0(p...)		REG(p)
+ *  #define SELAK1(p...)		REG(p)
+ *  #define SELAK2(p...)
+ *  #define REG(kind,tie,name,galn,aln,asz,csz,dbnum,base,rnum,bsz,rst,x...) \
+ *		...what you want to expand...
+ */
+
+#define XCHAL_NCP_SA_NUM	3
+#define XCHAL_NCP_SA_LIST(s)	\
+ XCHAL_SA_REG(s,0,0,0,1,             br, 4, 4, 4,0x0204,  sr,4  , 16,0,0,0) \
+ XCHAL_SA_REG(s,0,0,0,1,      scompare1, 4, 4, 4,0x020C,  sr,12 , 32,0,0,0) \
+ XCHAL_SA_REG(s,1,2,1,1,      threadptr, 4, 4, 4,0x03E7,  ur,231, 32,0,0,0)
+
+#define XCHAL_CP0_SA_NUM	0
+#define XCHAL_CP0_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP1_SA_NUM	16
+#define XCHAL_CP1_SA_LIST(s)	\
+ XCHAL_SA_REG(s,0,0,1,0,     ae_ovf_sar, 8, 4, 4,0x03F0,  ur,240,  7,0,0,0) \
+ XCHAL_SA_REG(s,0,0,1,0,     ae_bithead, 4, 4, 4,0x03F1,  ur,241, 32,0,0,0) \
+ XCHAL_SA_REG(s,0,0,1,0,ae_ts_fts_bu_bp, 4, 4, 4,0x03F2,  ur,242, 16,0,0,0) \
+ XCHAL_SA_REG(s,0,0,1,0,       ae_sd_no, 4, 4, 4,0x03F3,  ur,243, 28,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep0, 8, 8, 8,0x0060, aep,0  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep1, 8, 8, 8,0x0061, aep,1  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep2, 8, 8, 8,0x0062, aep,2  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep3, 8, 8, 8,0x0063, aep,3  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep4, 8, 8, 8,0x0064, aep,4  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep5, 8, 8, 8,0x0065, aep,5  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep6, 8, 8, 8,0x0066, aep,6  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aep7, 8, 8, 8,0x0067, aep,7  , 48,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aeq0, 8, 8, 8,0x0068, aeq,0  , 56,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aeq1, 8, 8, 8,0x0069, aeq,1  , 56,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aeq2, 8, 8, 8,0x006A, aeq,2  , 56,0,0,0) \
+ XCHAL_SA_REG(s,0,0,2,0,           aeq3, 8, 8, 8,0x006B, aeq,3  , 56,0,0,0)
+
+#define XCHAL_CP2_SA_NUM	0
+#define XCHAL_CP2_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP3_SA_NUM	0
+#define XCHAL_CP3_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP4_SA_NUM	0
+#define XCHAL_CP4_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP5_SA_NUM	0
+#define XCHAL_CP5_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP6_SA_NUM	0
+#define XCHAL_CP6_SA_LIST(s)	/* empty */
+
+#define XCHAL_CP7_SA_NUM	0
+#define XCHAL_CP7_SA_LIST(s)	/* empty */
+
+/* Byte length of instruction from its first nibble (op0 field), per FLIX.  */
+#define XCHAL_OP0_FORMAT_LENGTHS	3,3,3,3,3,3,3,3,2,2,2,2,2,2,3,8
+
+#endif /*_XTENSA_CORE_TIE_H*/
+
-- 
1.8.1.4

[PATCH 17/17] xtensa: implement CPU hotplug

[Max Filippov]

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
 arch/xtensa/Kconfig           |   9 +++
 arch/xtensa/include/asm/irq.h |   1 +
 arch/xtensa/include/asm/smp.h |   9 +++
 arch/xtensa/kernel/head.S     |  51 ++++++++++++++++-
 arch/xtensa/kernel/irq.c      |  49 ++++++++++++++++
 arch/xtensa/kernel/setup.c    |   1 +
 arch/xtensa/kernel/smp.c      | 128 +++++++++++++++++++++++++++++++++++++++++-
 arch/xtensa/kernel/traps.c    |   4 +-
 8 files changed, 248 insertions(+), 4 deletions(-)

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index 9cfb514..e3af36b 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -147,6 +147,15 @@ config NR_CPUS
 	range 2 32
 	default "4"
 
+config HOTPLUG_CPU
+	bool "Enable CPU hotplug support"
+	depends on SMP
+	help
+	  Say Y here to allow turning CPUs off and on. CPUs can be
+	  controlled through /sys/devices/system/cpu.
+
+	  Say N if you want to disable CPU hotplug.
+
 config MATH_EMULATION
 	bool "Math emulation"
 	help
diff --git a/arch/xtensa/include/asm/irq.h b/arch/xtensa/include/asm/irq.h
index 7d194d4..f71f88e 100644
--- a/arch/xtensa/include/asm/irq.h
+++ b/arch/xtensa/include/asm/irq.h
@@ -45,6 +45,7 @@ static __inline__ int irq_canonicalize(int irq)
 struct irqaction;
 struct irq_domain;
 
+void migrate_irqs(void);
 int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize,
 		unsigned long int_irq, unsigned long ext_irq,
 		unsigned long *out_hwirq, unsigned int *out_type);
diff --git a/arch/xtensa/include/asm/smp.h b/arch/xtensa/include/asm/smp.h
index 30ac58c..4e43f56 100644
--- a/arch/xtensa/include/asm/smp.h
+++ b/arch/xtensa/include/asm/smp.h
@@ -29,6 +29,15 @@ void ipi_init(void);
 struct seq_file;
 void show_ipi_list(struct seq_file *p, int prec);
 
+#ifdef CONFIG_HOTPLUG_CPU
+
+void __cpu_die(unsigned int cpu);
+int __cpu_disable(void);
+void cpu_die(void);
+void cpu_restart(void);
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
 #endif /* CONFIG_SMP */
 
 #endif	/* _XTENSA_SMP_H */
diff --git a/arch/xtensa/kernel/head.S b/arch/xtensa/kernel/head.S
index 74ec62c..aeeb3cc 100644
--- a/arch/xtensa/kernel/head.S
+++ b/arch/xtensa/kernel/head.S
@@ -103,7 +103,7 @@ _SetupMMU:
 
 ENDPROC(_start)
 
-	__INIT
+	__REF
 	.literal_position
 
 ENTRY(_startup)
@@ -302,6 +302,55 @@ should_never_return:
 
 ENDPROC(_startup)
 
+#ifdef CONFIG_HOTPLUG_CPU
+
+ENTRY(cpu_restart)
+
+#if XCHAL_DCACHE_IS_WRITEBACK
+	___flush_invalidate_dcache_all a2 a3
+#else
+	___invalidate_dcache_all a2 a3
+#endif
+	memw
+	movi	a2, CCON	# MX External Register to Configure Cache
+	movi	a3, 0
+	wer	a3, a2
+	extw
+
+	rsr	a0, prid
+	neg	a2, a0
+	movi	a3, cpu_start_id
+	s32i	a2, a3, 0
+#if XCHAL_DCACHE_IS_WRITEBACK
+	dhwbi	a3, 0
+#endif
+1:
+	l32i	a2, a3, 0
+	dhi	a3, 0
+	bne	a2, a0, 1b
+
+	/*
+	 * Initialize WB, WS, and clear PS.EXCM (to allow loop instructions).
+	 * Set Interrupt Level just below XCHAL_DEBUGLEVEL to allow
+	 * xt-gdb to single step via DEBUG exceptions received directly
+	 * by ocd.
+	 */
+	movi	a1, 1
+	movi	a0, 0
+	wsr	a1, windowstart
+	wsr	a0, windowbase
+	rsync
+
+	movi	a1, LOCKLEVEL
+	wsr	a1, ps
+	rsync
+
+	j	_startup
+
+ENDPROC(cpu_restart)
+
+#endif  /* CONFIG_HOTPLUG_CPU */
+
 /*
  * DATA section
  */
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index fad9e00..482868a 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -153,3 +153,52 @@ void __init init_IRQ(void)
 #endif
 	variant_init_irq();
 }
+
+#ifdef CONFIG_HOTPLUG_CPU
+static void route_irq(struct irq_data *data, unsigned int irq, unsigned int cpu)
+{
+	struct irq_desc *desc = irq_to_desc(irq);
+	struct irq_chip *chip = irq_data_get_irq_chip(data);
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&desc->lock, flags);
+	if (chip->irq_set_affinity)
+		chip->irq_set_affinity(data, cpumask_of(cpu), false);
+	raw_spin_unlock_irqrestore(&desc->lock, flags);
+}
+
+/*
+ * The CPU has been marked offline.  Migrate IRQs off this CPU.  If
+ * the affinity settings do not allow other CPUs, force them onto any
+ * available CPU.
+ */
+void migrate_irqs(void)
+{
+	unsigned int i, cpu = smp_processor_id();
+	struct irq_desc *desc;
+
+	for_each_irq_desc(i, desc) {
+		struct irq_data *data = irq_desc_get_irq_data(desc);
+		unsigned int newcpu;
+
+		if (irqd_is_per_cpu(data))
+			continue;
+
+		if (!cpumask_test_cpu(cpu, data->affinity))
+			continue;
+
+		newcpu = cpumask_any_and(data->affinity, cpu_online_mask);
+
+		if (newcpu >= nr_cpu_ids) {
+			pr_info_ratelimited("IRQ%u no longer affine to CPU%u\n",
+					    i, cpu);
+
+			cpumask_setall(data->affinity);
+			newcpu = cpumask_any_and(data->affinity,
+						 cpu_online_mask);
+		}
+
+		route_irq(data, i, newcpu);
+	}
+}
+#endif /* CONFIG_HOTPLUG_CPU */
diff --git a/arch/xtensa/kernel/setup.c b/arch/xtensa/kernel/setup.c
index 1f6e377..963a4b2 100644
--- a/arch/xtensa/kernel/setup.c
+++ b/arch/xtensa/kernel/setup.c
@@ -554,6 +554,7 @@ static int __init topology_init(void)
 
 	for_each_possible_cpu(i) {
 		struct cpu *cpu = &per_cpu(cpu_data, i);
+		cpu->hotpluggable = !!i;
 		register_cpu(cpu, i);
 	}
 
diff --git a/arch/xtensa/kernel/smp.c b/arch/xtensa/kernel/smp.c
index 46bdd14..1c7a209 100644
--- a/arch/xtensa/kernel/smp.c
+++ b/arch/xtensa/kernel/smp.c
@@ -40,6 +40,11 @@
 # endif
 #endif
 
+static void system_invalidate_dcache_range(unsigned long start,
+		unsigned long size);
+static void system_flush_invalidate_dcache_range(unsigned long start,
+		unsigned long size);
+
 /* IPI (Inter Process Interrupt) */
 
 #define IPI_IRQ	0
@@ -106,7 +111,7 @@ void __init smp_cpus_done(unsigned int max_cpus)
 static int boot_secondary_processors = 1; /* Set with xt-gdb via .xt-gdb */
 static DECLARE_COMPLETION(cpu_running);
 
-void __init secondary_start_kernel(void)
+void secondary_start_kernel(void)
 {
 	struct mm_struct *mm = &init_mm;
 	unsigned int cpu = smp_processor_id();
@@ -174,6 +179,9 @@ static void mx_cpu_stop(void *p)
 			__func__, cpu, run_stall_mask, get_er(MPSCORE));
 }
 
+#ifdef CONFIG_HOTPLUG_CPU
+unsigned long cpu_start_id __cacheline_aligned;
+#endif
 unsigned long cpu_start_ccount;
 
 static int boot_secondary(unsigned int cpu, struct task_struct *ts)
@@ -182,6 +190,11 @@ static int boot_secondary(unsigned int cpu, struct task_struct *ts)
 	unsigned long ccount;
 	int i;
 
+#ifdef CONFIG_HOTPLUG_CPU
+	cpu_start_id = cpu;
+	system_flush_invalidate_dcache_range(
+			(unsigned long)&cpu_start_id, sizeof(cpu_start_id));
+#endif
 	smp_call_function_single(0, mx_cpu_start, (void *)cpu, 1);
 
 	for (i = 0; i < 2; ++i) {
@@ -234,6 +247,85 @@ int __cpu_up(unsigned int cpu, struct task_struct *idle)
 	return ret;
 }
 
+#ifdef CONFIG_HOTPLUG_CPU
+
+/*
+ * __cpu_disable runs on the processor to be shutdown.
+ */
+int __cpu_disable(void)
+{
+	unsigned int cpu = smp_processor_id();
+
+	/*
+	 * Take this CPU offline.  Once we clear this, we can't return,
+	 * and we must not schedule until we're ready to give up the cpu.
+	 */
+	set_cpu_online(cpu, false);
+
+	/*
+	 * OK - migrate IRQs away from this CPU
+	 */
+	migrate_irqs();
+
+	/*
+	 * Flush user cache and TLB mappings, and then remove this CPU
+	 * from the vm mask set of all processes.
+	 */
+	local_flush_cache_all();
+	local_flush_tlb_all();
+	invalidate_page_directory();
+
+	clear_tasks_mm_cpumask(cpu);
+
+	return 0;
+}
+
+static void platform_cpu_kill(unsigned int cpu)
+{
+	smp_call_function_single(0, mx_cpu_stop, (void *)cpu, true);
+}
+
+/*
+ * called on the thread which is asking for a CPU to be shutdown -
+ * waits until shutdown has completed, or it is timed out.
+ */
+void __cpu_die(unsigned int cpu)
+{
+	unsigned long timeout = jiffies + msecs_to_jiffies(1000);
+	while (time_before(jiffies, timeout)) {
+		system_invalidate_dcache_range((unsigned long)&cpu_start_id,
+				sizeof(cpu_start_id));
+		if (cpu_start_id == -cpu) {
+			platform_cpu_kill(cpu);
+			return;
+		}
+	}
+	pr_err("CPU%u: unable to kill\n", cpu);
+}
+
+void arch_cpu_idle_dead(void)
+{
+	cpu_die();
+}
+/*
+ * Called from the idle thread for the CPU which has been shutdown.
+ *
+ * Note that we disable IRQs here, but do not re-enable them
+ * before returning to the caller. This is also the behaviour
+ * of the other hotplug-cpu capable cores, so presumably coming
+ * out of idle fixes this.
+ */
+void __ref cpu_die(void)
+{
+	idle_task_exit();
+	local_irq_disable();
+	__asm__ __volatile__(
+			"	movi	a2, cpu_restart\n"
+			"	jx	a2\n");
+}
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
 enum ipi_msg_type {
 	IPI_RESCHEDULE = 0,
 	IPI_CALL_FUNC,
@@ -463,3 +555,37 @@ void flush_icache_range(unsigned long start, unsigned long end)
 	};
 	on_each_cpu(ipi_flush_icache_range, &fd, 1);
 }
+
+/* ------------------------------------------------------------------------- */
+
+static void ipi_invalidate_dcache_range(void *arg)
+{
+	struct flush_data *fd = arg;
+	__invalidate_dcache_range(fd->addr1, fd->addr2);
+}
+
+static void system_invalidate_dcache_range(unsigned long start,
+		unsigned long size)
+{
+	struct flush_data fd = {
+		.addr1 = start,
+		.addr2 = size,
+	};
+	on_each_cpu(ipi_invalidate_dcache_range, &fd, 1);
+}
+
+static void ipi_flush_invalidate_dcache_range(void *arg)
+{
+	struct flush_data *fd = arg;
+	__flush_invalidate_dcache_range(fd->addr1, fd->addr2);
+}
+
+static void system_flush_invalidate_dcache_range(unsigned long start,
+		unsigned long size)
+{
+	struct flush_data fd = {
+		.addr1 = start,
+		.addr2 = size,
+	};
+	on_each_cpu(ipi_flush_invalidate_dcache_range, &fd, 1);
+}
diff --git a/arch/xtensa/kernel/traps.c b/arch/xtensa/kernel/traps.c
index 3c0ff57..eebbfd8 100644
--- a/arch/xtensa/kernel/traps.c
+++ b/arch/xtensa/kernel/traps.c
@@ -332,7 +332,7 @@ void * __init trap_set_handler(int cause, void *handler)
 }
 
 
-static void __init trap_init_excsave(void)
+static void trap_init_excsave(void)
 {
 	unsigned long excsave1 = (unsigned long)this_cpu_ptr(exc_table);
 	__asm__ __volatile__("wsr  %0, excsave1\n" : : "a" (excsave1));
@@ -384,7 +384,7 @@ void __init trap_init(void)
 }
 
 #ifdef CONFIG_SMP
-void __init secondary_trap_init(void)
+void secondary_trap_init(void)
 {
 	trap_init_excsave();
 }
-- 
1.8.1.4

Re: [PATCH 00/17] xtensa SMP queue

[Chris Zankel]

Hi Max,

I have added the SMP patch-set to the for_next tree (for the time 
being). Let me know if you wanted it in for this release or if you think 
it's not ready for prime time. Note that I skipped the variant 
configuration as it is a lot of code and it is more of a 'test' 
configuration.

Thanks,
-Chris



On 10/16/13, 3:42 PM, Max Filippov wrote:
> Hi Chris,
>
> this series adds SMP support for xtensa.
> There's a number of trivial cleanups:
>        xtensa: remove NO_IRQ definitions
>        xtensa: fix build warning in 64-bit build environment
>        xtensa: fix build warning from gcc-4.7.2
>        xtensa: call check_s32c1i after trap_init
>        xtensa: move init_mmu declaration to mmu_context.h
>
> a couple of fixes for spinlocks:
>        xtensa: fix arch spinlock function names
>        xtensa: fix __delay for small loop count
>
> three fixes for the clocksources, with these we get high precision timers
> usable in SMP and IRQ time accounting:
>        xtensa: mark ccount as continuous clocksource
>        xtensa: update clockevent setup for SMP
>        xtensa: enable HAVE_IRQ_TIME_ACCOUNTING
>
> fixes and rearrangement of IRQ handling code and a new driver for MX
> interrupt distributor:
>        xtensa: move built-in PIC to drivers/irqchip
>        xtensa: clean up do_interrupt/do_IRQ
>        xtensa: clear timer IRQ unconditionally in its handler
>        xtensa: add MX irqchip
>
> and finally SMP and hotplug CPU support and a sample SMP-capable CPU variant:
>        xtensa: add SMP support
>        xtensa: implement CPU hotplug
>        xtensa: add Three Core HiFi-2 MX Variant.
>
> Please review.
>
> You can pull these changes from my git tree,
> the following changes since commit 61e6cfa80de5760bbe406f4e815b7739205754d2:
>
>    Linux 3.12-rc5 (2013-10-13 15:41:28 -0700)
>
> are available in the git repository at:
>
>    git://github.com/jcmvbkbc/linux-xtensa.git tags/xtensa-smp-20131017
>
> for you to fetch changes up to 4862c5ae570b8a40275f413dfa2e3f184ecfa1d9:
>
>    xtensa: implement CPU hotplug (2013-10-17 01:50:17 +0400)
>
> ----------------------------------------------------------------
> Clean up minor warnings.
> Fix spinlock function names, provide missing definitions.
> Extract PIC into drivers/irqchip.
> Clean up IRQ handling.
> Add MX interrupt distributor.
> Add SMP support.
> Add hotplug CPU support.
> Add SMP-capable core variant.
>
> ----------------------------------------------------------------
> Baruch Siach (1):
>        xtensa: mark ccount as continuous clocksource
>
> Max Filippov (15):
>        xtensa: remove NO_IRQ definitions
>        xtensa: fix build warning in 64-bit build environment
>        xtensa: fix build warning from gcc-4.7.2
>        xtensa: fix arch spinlock function names
>        xtensa: fix __delay for small loop count
>        xtensa: enable HAVE_IRQ_TIME_ACCOUNTING
>        xtensa: update clockevent setup for SMP
>        xtensa: call check_s32c1i after trap_init
>        xtensa: move init_mmu declaration to mmu_context.h
>        xtensa: move built-in PIC to drivers/irqchip
>        xtensa: clean up do_interrupt/do_IRQ
>        xtensa: clear timer IRQ unconditionally in its handler
>        xtensa: add MX irqchip
>        xtensa: add SMP support
>        xtensa: implement CPU hotplug
>
> Piet Delaney (1):
>        xtensa: add Three Core HiFi-2 MX Variant.
>
>   arch/xtensa/Kconfig                                |  54 +-
>   arch/xtensa/Makefile                               |   1 +
>   arch/xtensa/include/asm/barrier.h                  |   4 +-
>   arch/xtensa/include/asm/bitops.h                   |   8 +-
>   arch/xtensa/include/asm/cacheflush.h               |  40 +-
>   arch/xtensa/include/asm/delay.h                    |   9 +-
>   arch/xtensa/include/asm/irq.h                      |   9 +
>   arch/xtensa/include/asm/mmu.h                      |  10 +-
>   arch/xtensa/include/asm/mmu_context.h              | 106 ++--
>   arch/xtensa/include/asm/mxregs.h                   |  46 ++
>   arch/xtensa/include/asm/processor.h                |  20 +
>   arch/xtensa/include/asm/ptrace.h                   |   8 +
>   arch/xtensa/include/asm/smp.h                      |  38 +-
>   arch/xtensa/include/asm/spinlock.h                 |  31 +-
>   arch/xtensa/include/asm/spinlock_types.h           |  20 +
>   arch/xtensa/include/asm/timex.h                    |  14 +-
>   arch/xtensa/include/asm/tlbflush.h                 |  42 +-
>   arch/xtensa/include/asm/traps.h                    |   1 +
>   arch/xtensa/include/asm/vectors.h                  |   4 +
>   arch/xtensa/kernel/Makefile                        |   1 +
>   arch/xtensa/kernel/head.S                          | 181 +++++--
>   arch/xtensa/kernel/irq.c                           | 207 ++++----
>   arch/xtensa/kernel/mxhead.S                        |  85 +++
>   arch/xtensa/kernel/setup.c                         |  92 ++--
>   arch/xtensa/kernel/smp.c                           | 591 +++++++++++++++++++++
>   arch/xtensa/kernel/time.c                          |  56 +-
>   arch/xtensa/kernel/traps.c                         |  56 +-
>   arch/xtensa/kernel/vmlinux.lds.S                   |  26 +
>   arch/xtensa/mm/cache.c                             |   7 +-
>   arch/xtensa/mm/fault.c                             |   2 +-
>   arch/xtensa/mm/misc.S                              |   4 +-
>   arch/xtensa/mm/mmu.c                               |   4 +-
>   arch/xtensa/mm/tlb.c                               |  37 +-
>   .../platforms/xtfpga/include/platform/hardware.h   |   6 +-
>   arch/xtensa/variants/s6000/include/variant/irq.h   |   1 -
>   .../test_mmuhifi_c3/include/variant/core.h         | 384 +++++++++++++
>   .../test_mmuhifi_c3/include/variant/tie-asm.h      | 183 +++++++
>   .../variants/test_mmuhifi_c3/include/variant/tie.h | 141 +++++
>   drivers/irqchip/Kconfig                            |   4 +
>   drivers/irqchip/Makefile                           |   2 +
>   drivers/irqchip/irq-xtensa-mx.c                    | 166 ++++++
>   drivers/irqchip/irq-xtensa-pic.c                   | 108 ++++
>   include/linux/irqchip/xtensa-mx.h                  |  17 +
>   include/linux/irqchip/xtensa-pic.h                 |  18 +
>   44 files changed, 2455 insertions(+), 389 deletions(-)
>   create mode 100644 arch/xtensa/include/asm/mxregs.h
>   create mode 100644 arch/xtensa/include/asm/spinlock_types.h
>   create mode 100644 arch/xtensa/kernel/mxhead.S
>   create mode 100644 arch/xtensa/kernel/smp.c
>   create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
>   create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
>   create mode 100644 arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h
>   create mode 100644 drivers/irqchip/irq-xtensa-mx.c
>   create mode 100644 drivers/irqchip/irq-xtensa-pic.c
>   create mode 100644 include/linux/irqchip/xtensa-mx.h
>   create mode 100644 include/linux/irqchip/xtensa-pic.h
>

Re: [PATCH 00/17] xtensa SMP queue

[Max Filippov]

Hi Chris,

On Tue, Nov 19, 2013 at 8:05 AM, Chris Zankel <chris@zankel.net> wrote:
> Hi Max,
>
> I have added the SMP patch-set to the for_next tree (for the time being).
> Let me know if you wanted it in for this release or if you think it's not
> ready for prime time. Note that I skipped the variant configuration as it is
> a lot of code and it is more of a 'test' configuration.

Cool (: I currently only have a small optimization fix for the MX interrupt
controller, otherwise I think it's good. Should I send a follow-up patch
or respin the whole series?

> On 10/16/13, 3:42 PM, Max Filippov wrote:
>>
>> Hi Chris,
>>
>> this series adds SMP support for xtensa.
>> There's a number of trivial cleanups:
>>        xtensa: remove NO_IRQ definitions
>>        xtensa: fix build warning in 64-bit build environment
>>        xtensa: fix build warning from gcc-4.7.2
>>        xtensa: call check_s32c1i after trap_init
>>        xtensa: move init_mmu declaration to mmu_context.h
>>
>> a couple of fixes for spinlocks:
>>        xtensa: fix arch spinlock function names
>>        xtensa: fix __delay for small loop count
>>
>> three fixes for the clocksources, with these we get high precision timers
>> usable in SMP and IRQ time accounting:
>>        xtensa: mark ccount as continuous clocksource
>>        xtensa: update clockevent setup for SMP
>>        xtensa: enable HAVE_IRQ_TIME_ACCOUNTING
>>
>> fixes and rearrangement of IRQ handling code and a new driver for MX
>> interrupt distributor:
>>        xtensa: move built-in PIC to drivers/irqchip
>>        xtensa: clean up do_interrupt/do_IRQ
>>        xtensa: clear timer IRQ unconditionally in its handler
>>        xtensa: add MX irqchip
>>
>> and finally SMP and hotplug CPU support and a sample SMP-capable CPU
>> variant:
>>        xtensa: add SMP support
>>        xtensa: implement CPU hotplug
>>        xtensa: add Three Core HiFi-2 MX Variant.
>>
>> Please review.
>>
>> You can pull these changes from my git tree,
>> the following changes since commit
>> 61e6cfa80de5760bbe406f4e815b7739205754d2:
>>
>>    Linux 3.12-rc5 (2013-10-13 15:41:28 -0700)
>>
>> are available in the git repository at:
>>
>>    git://github.com/jcmvbkbc/linux-xtensa.git tags/xtensa-smp-20131017
>>
>> for you to fetch changes up to 4862c5ae570b8a40275f413dfa2e3f184ecfa1d9:
>>
>>    xtensa: implement CPU hotplug (2013-10-17 01:50:17 +0400)
>>
>> ----------------------------------------------------------------
>> Clean up minor warnings.
>> Fix spinlock function names, provide missing definitions.
>> Extract PIC into drivers/irqchip.
>> Clean up IRQ handling.
>> Add MX interrupt distributor.
>> Add SMP support.
>> Add hotplug CPU support.
>> Add SMP-capable core variant.
>>
>> ----------------------------------------------------------------
>> Baruch Siach (1):
>>        xtensa: mark ccount as continuous clocksource
>>
>> Max Filippov (15):
>>        xtensa: remove NO_IRQ definitions
>>        xtensa: fix build warning in 64-bit build environment
>>        xtensa: fix build warning from gcc-4.7.2
>>        xtensa: fix arch spinlock function names
>>        xtensa: fix __delay for small loop count
>>        xtensa: enable HAVE_IRQ_TIME_ACCOUNTING
>>        xtensa: update clockevent setup for SMP
>>        xtensa: call check_s32c1i after trap_init
>>        xtensa: move init_mmu declaration to mmu_context.h
>>        xtensa: move built-in PIC to drivers/irqchip
>>        xtensa: clean up do_interrupt/do_IRQ
>>        xtensa: clear timer IRQ unconditionally in its handler
>>        xtensa: add MX irqchip
>>        xtensa: add SMP support
>>        xtensa: implement CPU hotplug
>>
>> Piet Delaney (1):
>>        xtensa: add Three Core HiFi-2 MX Variant.
>>
>>   arch/xtensa/Kconfig                                |  54 +-
>>   arch/xtensa/Makefile                               |   1 +
>>   arch/xtensa/include/asm/barrier.h                  |   4 +-
>>   arch/xtensa/include/asm/bitops.h                   |   8 +-
>>   arch/xtensa/include/asm/cacheflush.h               |  40 +-
>>   arch/xtensa/include/asm/delay.h                    |   9 +-
>>   arch/xtensa/include/asm/irq.h                      |   9 +
>>   arch/xtensa/include/asm/mmu.h                      |  10 +-
>>   arch/xtensa/include/asm/mmu_context.h              | 106 ++--
>>   arch/xtensa/include/asm/mxregs.h                   |  46 ++
>>   arch/xtensa/include/asm/processor.h                |  20 +
>>   arch/xtensa/include/asm/ptrace.h                   |   8 +
>>   arch/xtensa/include/asm/smp.h                      |  38 +-
>>   arch/xtensa/include/asm/spinlock.h                 |  31 +-
>>   arch/xtensa/include/asm/spinlock_types.h           |  20 +
>>   arch/xtensa/include/asm/timex.h                    |  14 +-
>>   arch/xtensa/include/asm/tlbflush.h                 |  42 +-
>>   arch/xtensa/include/asm/traps.h                    |   1 +
>>   arch/xtensa/include/asm/vectors.h                  |   4 +
>>   arch/xtensa/kernel/Makefile                        |   1 +
>>   arch/xtensa/kernel/head.S                          | 181 +++++--
>>   arch/xtensa/kernel/irq.c                           | 207 ++++----
>>   arch/xtensa/kernel/mxhead.S                        |  85 +++
>>   arch/xtensa/kernel/setup.c                         |  92 ++--
>>   arch/xtensa/kernel/smp.c                           | 591
>> +++++++++++++++++++++
>>   arch/xtensa/kernel/time.c                          |  56 +-
>>   arch/xtensa/kernel/traps.c                         |  56 +-
>>   arch/xtensa/kernel/vmlinux.lds.S                   |  26 +
>>   arch/xtensa/mm/cache.c                             |   7 +-
>>   arch/xtensa/mm/fault.c                             |   2 +-
>>   arch/xtensa/mm/misc.S                              |   4 +-
>>   arch/xtensa/mm/mmu.c                               |   4 +-
>>   arch/xtensa/mm/tlb.c                               |  37 +-
>>   .../platforms/xtfpga/include/platform/hardware.h   |   6 +-
>>   arch/xtensa/variants/s6000/include/variant/irq.h   |   1 -
>>   .../test_mmuhifi_c3/include/variant/core.h         | 384 +++++++++++++
>>   .../test_mmuhifi_c3/include/variant/tie-asm.h      | 183 +++++++
>>   .../variants/test_mmuhifi_c3/include/variant/tie.h | 141 +++++
>>   drivers/irqchip/Kconfig                            |   4 +
>>   drivers/irqchip/Makefile                           |   2 +
>>   drivers/irqchip/irq-xtensa-mx.c                    | 166 ++++++
>>   drivers/irqchip/irq-xtensa-pic.c                   | 108 ++++
>>   include/linux/irqchip/xtensa-mx.h                  |  17 +
>>   include/linux/irqchip/xtensa-pic.h                 |  18 +
>>   44 files changed, 2455 insertions(+), 389 deletions(-)
>>   create mode 100644 arch/xtensa/include/asm/mxregs.h
>>   create mode 100644 arch/xtensa/include/asm/spinlock_types.h
>>   create mode 100644 arch/xtensa/kernel/mxhead.S
>>   create mode 100644 arch/xtensa/kernel/smp.c
>>   create mode 100644
>> arch/xtensa/variants/test_mmuhifi_c3/include/variant/core.h
>>   create mode 100644
>> arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie-asm.h
>>   create mode 100644
>> arch/xtensa/variants/test_mmuhifi_c3/include/variant/tie.h
>>   create mode 100644 drivers/irqchip/irq-xtensa-mx.c
>>   create mode 100644 drivers/irqchip/irq-xtensa-pic.c
>>   create mode 100644 include/linux/irqchip/xtensa-mx.h
>>   create mode 100644 include/linux/irqchip/xtensa-pic.h
>>
>



-- 
Thanks.
-- Max

Re: [PATCH 00/17] xtensa SMP queue

[czankel]

On 11/18/13 10:36 PM, Max Filippov wrote:
> Hi Chris,
>
> On Tue, Nov 19, 2013 at 8:05 AM, Chris Zankel <chris@zankel.net> wrote:
>> Hi Max,
>>
>> I have added the SMP patch-set to the for_next tree (for the time being).
>> Let me know if you wanted it in for this release or if you think it's not
>> ready for prime time. Note that I skipped the variant configuration as it is
>> a lot of code and it is more of a 'test' configuration.
> Cool (: I currently only have a small optimization fix for the MX interrupt
> controller, otherwise I think it's good. Should I send a follow-up patch
> or respin the whole series?
Feel free to just send me the patch, patch to the patch, or whatever is 
easiest for you.

Thanks,
-Chris

[PATCHv2 14/17] xtensa: add MX irqchip

[Max Filippov]

MX is an interrupt distributor used in some SMP-capable xtensa
configurations.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---

Hi Chris,

this is the updated version of the MX irqchip patch, you will also find it
in the SMP series tagged xtensa-smp-20131201 in my xtensa tree.

Changes v1 -> v2:
- don't disable preemption in the else clause of xtensa_mx_irq_[un]mask
  as these functions should be always called with preemption disabled.

 arch/xtensa/include/asm/irq.h       |   1 +
 arch/xtensa/include/asm/mxregs.h    |  46 ++++++++++
 arch/xtensa/include/asm/processor.h |  20 +++++
 arch/xtensa/kernel/irq.c            |   8 ++
 drivers/irqchip/Kconfig             |   4 +
 drivers/irqchip/Makefile            |   1 +
 drivers/irqchip/irq-xtensa-mx.c     | 164 ++++++++++++++++++++++++++++++++++++
 include/linux/irqchip/xtensa-mx.h   |  17 ++++
 8 files changed, 261 insertions(+)
 create mode 100644 arch/xtensa/include/asm/mxregs.h
 create mode 100644 drivers/irqchip/irq-xtensa-mx.c
 create mode 100644 include/linux/irqchip/xtensa-mx.h

diff --git a/arch/xtensa/include/asm/irq.h b/arch/xtensa/include/asm/irq.h
index 16464f2..7d194d4 100644
--- a/arch/xtensa/include/asm/irq.h
+++ b/arch/xtensa/include/asm/irq.h
@@ -50,5 +50,6 @@ int xtensa_irq_domain_xlate(const u32 *intspec, unsigned int intsize,
 		unsigned long *out_hwirq, unsigned int *out_type);
 int xtensa_irq_map(struct irq_domain *d, unsigned int irq, irq_hw_number_t hw);
 unsigned xtensa_map_ext_irq(unsigned ext_irq);
+unsigned xtensa_get_ext_irq_no(unsigned irq);
 
 #endif	/* _XTENSA_IRQ_H */
diff --git a/arch/xtensa/include/asm/mxregs.h b/arch/xtensa/include/asm/mxregs.h
new file mode 100644
index 0000000..73dcc54
--- /dev/null
+++ b/arch/xtensa/include/asm/mxregs.h
@@ -0,0 +1,46 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * 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) 2008 - 2013 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_MXREGS_H
+#define _XTENSA_MXREGS_H
+
+/*
+ * RER/WER at, as	Read/write external register
+ *	at: value
+ *	as: address
+ *
+ * Address	Value
+ * 00nn		0...0p..p	Interrupt Routing, route IRQ n to processor p
+ * 01pp		0...0d..d	16 bits (d) 'ored' as single IPI to processor p
+ * 0180		0...0m..m	Clear enable specified by mask (m)
+ * 0184		0...0m..m	Set enable specified by mask (m)
+ * 0190		0...0x..x	8-bit IPI partition register
+ *				VVVVVVVVPPPPUUUUUUUUUUUUUUUUU
+ *				V (10-bit) Release/Version
+ *				P ( 4-bit) Number of cores - 1
+ *				U (18-bit) ID
+ * 01a0		i.......i	32-bit ConfigID
+ * 0200		0...0m..m	RunStall core 'n'
+ * 0220		c		Cache coherency enabled
+ */
+
+#define MIROUT(irq)	(0x000 + (irq))
+#define MIPICAUSE(cpu)	(0x100 + (cpu))
+#define MIPISET(cause)	(0x140 + (cause))
+#define MIENG		0x180
+#define MIENGSET	0x184
+#define MIASG		0x188	/* Read Global Assert Register */
+#define MIASGSET	0x18c	/* Set Global Addert Regiter */
+#define MIPIPART	0x190
+#define SYSCFGID	0x1a0
+#define MPSCORE		0x200
+#define CCON		0x220
+
+#endif /* _XTENSA_MXREGS_H */
diff --git a/arch/xtensa/include/asm/processor.h b/arch/xtensa/include/asm/processor.h
index 7e409a5..abb5970 100644
--- a/arch/xtensa/include/asm/processor.h
+++ b/arch/xtensa/include/asm/processor.h
@@ -191,5 +191,25 @@ extern unsigned long get_wchan(struct task_struct *p);
 #define set_sr(x,sr) ({unsigned int v=(unsigned int)x; WSR(v,sr);})
 #define get_sr(sr) ({unsigned int v; RSR(v,sr); v; })
 
+#ifndef XCHAL_HAVE_EXTERN_REGS
+#define XCHAL_HAVE_EXTERN_REGS 0
+#endif
+
+#if XCHAL_HAVE_EXTERN_REGS
+
+static inline void set_er(unsigned long value, unsigned long addr)
+{
+	asm volatile ("wer %0, %1" : : "a" (value), "a" (addr) : "memory");
+}
+
+static inline unsigned long get_er(unsigned long addr)
+{
+	register unsigned long value;
+	asm volatile ("rer %0, %1" : "=a" (value) : "a" (addr) : "memory");
+	return value;
+}
+
+#endif /* XCHAL_HAVE_EXTERN_REGS */
+
 #endif	/* __ASSEMBLY__ */
 #endif	/* _XTENSA_PROCESSOR_H */
diff --git a/arch/xtensa/kernel/irq.c b/arch/xtensa/kernel/irq.c
index 3cef58e..7d49730 100644
--- a/arch/xtensa/kernel/irq.c
+++ b/arch/xtensa/kernel/irq.c
@@ -123,6 +123,14 @@ unsigned xtensa_map_ext_irq(unsigned ext_irq)
 	return XCHAL_NUM_INTERRUPTS;
 }
 
+unsigned xtensa_get_ext_irq_no(unsigned irq)
+{
+	unsigned mask = (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+		XCHAL_INTTYPE_MASK_EXTERN_LEVEL) &
+		((1u << irq) - 1);
+	return hweight32(mask);
+}
+
 void __init init_IRQ(void)
 {
 #ifdef CONFIG_OF
diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
index 3792a1a..07bc79c 100644
--- a/drivers/irqchip/Kconfig
+++ b/drivers/irqchip/Kconfig
@@ -61,3 +61,7 @@ config VERSATILE_FPGA_IRQ_NR
        int
        default 4
        depends on VERSATILE_FPGA_IRQ
+
+config XTENSA_MX
+	bool
+	select IRQ_DOMAIN
diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
index c81a7f3..66913f9 100644
--- a/drivers/irqchip/Makefile
+++ b/drivers/irqchip/Makefile
@@ -23,3 +23,4 @@ obj-$(CONFIG_VERSATILE_FPGA_IRQ)	+= irq-versatile-fpga.o
 obj-$(CONFIG_ARCH_VT8500)		+= irq-vt8500.o
 obj-$(CONFIG_TB10X_IRQC)		+= irq-tb10x.o
 obj-$(CONFIG_XTENSA)			+= irq-xtensa-pic.o
+obj-$(CONFIG_XTENSA_MX)			+= irq-xtensa-mx.o
diff --git a/drivers/irqchip/irq-xtensa-mx.c b/drivers/irqchip/irq-xtensa-mx.c
new file mode 100644
index 0000000..76647ca
--- /dev/null
+++ b/drivers/irqchip/irq-xtensa-mx.c
@@ -0,0 +1,164 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ *
+ * 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.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irqdomain.h>
+#include <linux/irq.h>
+#include <linux/of.h>
+
+#include <asm/mxregs.h>
+
+#include "irqchip.h"
+
+#define HW_IRQ_IPI_COUNT 2
+#define HW_IRQ_MX_BASE 2
+#define HW_IRQ_EXTERN_BASE 3
+
+static DEFINE_PER_CPU(unsigned int, cached_irq_mask);
+
+static int xtensa_mx_irq_map(struct irq_domain *d, unsigned int irq,
+		irq_hw_number_t hw)
+{
+	if (hw < HW_IRQ_IPI_COUNT) {
+		struct irq_chip *irq_chip = d->host_data;
+		irq_set_chip_and_handler_name(irq, irq_chip,
+				handle_percpu_irq, "ipi");
+		irq_set_status_flags(irq, IRQ_LEVEL);
+		return 0;
+	}
+	return xtensa_irq_map(d, irq, hw);
+}
+
+/*
+ * Device Tree IRQ specifier translation function which works with one or
+ * two cell bindings. First cell value maps directly to the hwirq number.
+ * Second cell if present specifies whether hwirq number is external (1) or
+ * internal (0).
+ */
+static int xtensa_mx_irq_domain_xlate(struct irq_domain *d,
+		struct device_node *ctrlr,
+		const u32 *intspec, unsigned int intsize,
+		unsigned long *out_hwirq, unsigned int *out_type)
+{
+	return xtensa_irq_domain_xlate(intspec, intsize,
+			intspec[0], intspec[0] + HW_IRQ_EXTERN_BASE,
+			out_hwirq, out_type);
+}
+
+static const struct irq_domain_ops xtensa_mx_irq_domain_ops = {
+	.xlate = xtensa_mx_irq_domain_xlate,
+	.map = xtensa_mx_irq_map,
+};
+
+void secondary_init_irq(void)
+{
+	__this_cpu_write(cached_irq_mask,
+			XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+			XCHAL_INTTYPE_MASK_EXTERN_LEVEL);
+	set_sr(XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+			XCHAL_INTTYPE_MASK_EXTERN_LEVEL, intenable);
+}
+
+static void xtensa_mx_irq_mask(struct irq_data *d)
+{
+	unsigned int mask = 1u << d->hwirq;
+
+	if (mask & (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+				XCHAL_INTTYPE_MASK_EXTERN_LEVEL)) {
+		set_er(1u << (xtensa_get_ext_irq_no(d->hwirq) -
+					HW_IRQ_MX_BASE), MIENG);
+	} else {
+		mask = __this_cpu_read(cached_irq_mask) & ~mask;
+		__this_cpu_write(cached_irq_mask, mask);
+		set_sr(mask, intenable);
+	}
+}
+
+static void xtensa_mx_irq_unmask(struct irq_data *d)
+{
+	unsigned int mask = 1u << d->hwirq;
+
+	if (mask & (XCHAL_INTTYPE_MASK_EXTERN_EDGE |
+				XCHAL_INTTYPE_MASK_EXTERN_LEVEL)) {
+		set_er(1u << (xtensa_get_ext_irq_no(d->hwirq) -
+					HW_IRQ_MX_BASE), MIENGSET);
+	} else {
+		mask |= __this_cpu_read(cached_irq_mask);
+		__this_cpu_write(cached_irq_mask, mask);
+		set_sr(mask, intenable);
+	}
+}
+
+static void xtensa_mx_irq_enable(struct irq_data *d)
+{
+	variant_irq_enable(d->hwirq);
+	xtensa_mx_irq_unmask(d);
+}
+
+static void xtensa_mx_irq_disable(struct irq_data *d)
+{
+	xtensa_mx_irq_mask(d);
+	variant_irq_disable(d->hwirq);
+}
+
+static void xtensa_mx_irq_ack(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intclear);
+}
+
+static int xtensa_mx_irq_retrigger(struct irq_data *d)
+{
+	set_sr(1 << d->hwirq, intset);
+	return 1;
+}
+
+static int xtensa_mx_irq_set_affinity(struct irq_data *d,
+		const struct cpumask *dest, bool force)
+{
+	unsigned mask = 1u << cpumask_any(dest);
+
+	set_er(mask, MIROUT(d->hwirq - HW_IRQ_MX_BASE));
+	return 0;
+
+}
+
+static struct irq_chip xtensa_mx_irq_chip = {
+	.name		= "xtensa-mx",
+	.irq_enable	= xtensa_mx_irq_enable,
+	.irq_disable	= xtensa_mx_irq_disable,
+	.irq_mask	= xtensa_mx_irq_mask,
+	.irq_unmask	= xtensa_mx_irq_unmask,
+	.irq_ack	= xtensa_mx_irq_ack,
+	.irq_retrigger	= xtensa_mx_irq_retrigger,
+	.irq_set_affinity = xtensa_mx_irq_set_affinity,
+};
+
+int __init xtensa_mx_init_legacy(struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_legacy(NULL, NR_IRQS, 0, 0,
+				&xtensa_mx_irq_domain_ops,
+				&xtensa_mx_irq_chip);
+	irq_set_default_host(root_domain);
+	secondary_init_irq();
+	return 0;
+}
+
+static int __init xtensa_mx_init(struct device_node *np,
+		struct device_node *interrupt_parent)
+{
+	struct irq_domain *root_domain =
+		irq_domain_add_linear(np, NR_IRQS, &xtensa_mx_irq_domain_ops,
+				&xtensa_mx_irq_chip);
+	irq_set_default_host(root_domain);
+	secondary_init_irq();
+	return 0;
+}
+IRQCHIP_DECLARE(xtensa_mx_irq_chip, "xtensa,mx", xtensa_mx_init);
diff --git a/include/linux/irqchip/xtensa-mx.h b/include/linux/irqchip/xtensa-mx.h
new file mode 100644
index 0000000..9c3b6ec
--- /dev/null
+++ b/include/linux/irqchip/xtensa-mx.h
@@ -0,0 +1,17 @@
+/*
+ * Xtensa MX interrupt distributor
+ *
+ * Copyright (C) 2002 - 2013 Tensilica, Inc.
+ *
+ * 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.
+ */
+
+#ifndef __LINUX_IRQCHIP_XTENSA_MX_H
+#define __LINUX_IRQCHIP_XTENSA_MX_H
+
+struct device_node;
+int xtensa_mx_init_legacy(struct device_node *interrupt_parent);
+
+#endif /* __LINUX_IRQCHIP_XTENSA_MX_H */
-- 
1.8.1.4

Re: [Linux-Xtensa] [PATCHv2 14/17] xtensa: add MX irqchip

[Baruch Siach]

Hi Max,

On Sun, Dec 01, 2013 at 09:38:47AM +0400, Max Filippov wrote:
> MX is an interrupt distributor used in some SMP-capable xtensa
> configurations.
> 
> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
> ---

[...]

> +IRQCHIP_DECLARE(xtensa_mx_irq_chip, "xtensa,mx", xtensa_mx_init);

This should probably better named "cdns,xtensa-mx", according to the ePAPR 1.1 
recommendation.

baruch

-- 
     http://baruch.siach.name/blog/                  ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch@tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -

Re: [Linux-Xtensa] [PATCHv2 14/17] xtensa: add MX irqchip

[Max Filippov]

On Sun, Dec 1, 2013 at 11:20 AM, Baruch Siach <baruch@tkos.co.il> wrote:
> Hi Max,
>
> On Sun, Dec 01, 2013 at 09:38:47AM +0400, Max Filippov wrote:
>> MX is an interrupt distributor used in some SMP-capable xtensa
>> configurations.
>>
>> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
>> ---
>
> [...]
>
>> +IRQCHIP_DECLARE(xtensa_mx_irq_chip, "xtensa,mx", xtensa_mx_init);
>
> This should probably better named "cdns,xtensa-mx", according to the ePAPR 1.1
> recommendation.

Right, I've sent an update for both MX and PIC xtensa irqchips.

-- 
Thanks.
-- Max