Merge tag 'xtensa-next-20160320' of git://github.com/czankel/xtensa-linux

Pull Xtensa updates from Chris Zankel:
 "Xtensa improvements for 4.6:

   - control whether perf IRQ is treated as NMI from Kconfig
   - implement ioremap for regions outside KIO segment
   - fix ISS serial port behaviour when EOF is reached
   - fix preemption in {clear,copy}_user_highpage
   - fix endianness issues for XTFPGA devices, big-endian cores are now
     fully functional
   - clean up debug infrastructure and add support for hardware
     breakpoints and watchpoints
   - add processor configurations for Three Core HiFi-2 MX and HiFi3
     cpus"

* tag 'xtensa-next-20160320' of git://github.com/czankel/xtensa-linux:
  xtensa: add test_kc705_hifi variant
  xtensa: add Three Core HiFi-2 MX Variant.
  xtensa: support hardware breakpoints/watchpoints
  xtensa: use context structure for debug exceptions
  xtensa: remove remaining non-functional KGDB bits
  xtensa: clear all DBREAKC registers on start
  xtensa: xtfpga: fix earlycon endianness
  xtensa: xtfpga: fix i2c controller register width and endianness
  xtensa: xtfpga: fix ethernet controller endianness
  xtensa: xtfpga: fix serial port register width and endianness
  xtensa: define CONFIG_CPU_{BIG,LITTLE}_ENDIAN
  xtensa: fix preemption in {clear,copy}_user_highpage
  xtensa: ISS: don't hang if stdin EOF is reached
  xtensa: support ioremap for memory outside KIO region
  xtensa: use XTENSA_INT_LEVEL macro in asm/timex.h
  xtensa: make fake NMI configurable

33 files changed, 2593 insertions, 78 deletions

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index 7e9464b0fc00..e832d3e9835e 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -17,6 +17,7 @@ config XTENSA
 	select HAVE_DMA_API_DEBUG
 	select HAVE_FUNCTION_TRACER
 	select HAVE_FUTEX_CMPXCHG if !MMU
+	select HAVE_HW_BREAKPOINT if PERF_EVENTS
 	select HAVE_IRQ_TIME_ACCOUNTING
 	select HAVE_OPROFILE
 	select HAVE_PERF_EVENTS
@@ -138,6 +139,22 @@ config XTENSA_VARIANT_HAVE_PERF_EVENTS
 
 	  If unsure, say N.
 
+config XTENSA_FAKE_NMI
+	bool "Treat PMM IRQ as NMI"
+	depends on XTENSA_VARIANT_HAVE_PERF_EVENTS
+	default n
+	help
+	  If PMM IRQ is the only IRQ at EXCM level it is safe to
+	  treat it as NMI, which improves accuracy of profiling.
+
+	  If there are other interrupts at or above PMM IRQ priority level
+	  but not above the EXCM level, PMM IRQ still may be treated as NMI,
+	  but only if these IRQs are not used. There will be a build warning
+	  saying that this is not safe, and a bugcheck if one of these IRQs
+	  actually fire.
+
+	  If unsure, say N.
+
 config XTENSA_UNALIGNED_USER
 	bool "Unaligned memory access in use space"
 	help
diff --git a/arch/xtensa/Makefile b/arch/xtensa/Makefile
index 709b5748a2d7..e54189427b31 100644
--- a/arch/xtensa/Makefile
+++ b/arch/xtensa/Makefile
@@ -53,9 +53,11 @@ endif
 
 ifeq ($(shell echo __XTENSA_EB__ | $(CC) -E - | grep -v "\#"),1)
 CHECKFLAGS += -D__XTENSA_EB__
+KBUILD_CPPFLAGS += -DCONFIG_CPU_BIG_ENDIAN
 endif
 ifeq ($(shell echo __XTENSA_EL__ | $(CC) -E - | grep -v "\#"),1)
 CHECKFLAGS += -D__XTENSA_EL__
+KBUILD_CPPFLAGS += -DCONFIG_CPU_LITTLE_ENDIAN
 endif
 
 vardirs := $(patsubst %,arch/xtensa/variants/%/,$(variant-y))
diff --git a/arch/xtensa/boot/dts/kc705.dts b/arch/xtensa/boot/dts/kc705.dts
index c4d17a34ab86..b1f4ee8c9a22 100644
--- a/arch/xtensa/boot/dts/kc705.dts
+++ b/arch/xtensa/boot/dts/kc705.dts
@@ -5,7 +5,7 @@
 / {
 	compatible = "cdns,xtensa-kc705";
 	chosen {
-		bootargs = "earlycon=uart8250,mmio32,0xfd050020,115200n8 console=ttyS0,115200n8 ip=dhcp root=/dev/nfs rw debug memmap=0x38000000";
+		bootargs = "earlycon=uart8250,mmio32native,0xfd050020,115200n8 console=ttyS0,115200n8 ip=dhcp root=/dev/nfs rw debug memmap=0x38000000";
 	};
 	memory@0 {
 		device_type = "memory";
diff --git a/arch/xtensa/boot/dts/xtfpga.dtsi b/arch/xtensa/boot/dts/xtfpga.dtsi
index cd0b9e34adc8..cd45f9c2c448 100644
--- a/arch/xtensa/boot/dts/xtfpga.dtsi
+++ b/arch/xtensa/boot/dts/xtfpga.dtsi
@@ -5,7 +5,7 @@
 	interrupt-parent = <&pic>;
 
 	chosen {
-		bootargs = "earlycon=uart8250,mmio32,0xfd050020,115200n8 console=ttyS0,115200n8 ip=dhcp root=/dev/nfs rw debug";
+		bootargs = "earlycon=uart8250,mmio32native,0xfd050020,115200n8 console=ttyS0,115200n8 ip=dhcp root=/dev/nfs rw debug";
 	};
 
 	memory@0 {
@@ -60,6 +60,8 @@
 			no-loopback-test;
 			reg = <0x0d050020 0x20>;
 			reg-shift = <2>;
+			reg-io-width = <4>;
+			native-endian;
 			interrupts = <0 1>; /* external irq 0 */
 			clocks = <&osc>;
 		};
@@ -67,6 +69,7 @@
 		enet0: ethoc@0d030000 {
 			compatible = "opencores,ethoc";
 			reg = <0x0d030000 0x4000 0x0d800000 0x4000>;
+			native-endian;
 			interrupts = <1 1>; /* external irq 1 */
 			local-mac-address = [00 50 c2 13 6f 00];
 			clocks = <&osc>;
@@ -86,7 +89,8 @@
 			#size-cells = <0>;
 			reg = <0x0d090000 0x20>;
 			reg-shift = <2>;
-			reg-io-width = <1>;
+			reg-io-width = <4>;
+			native-endian;
 			interrupts = <4 1>;
 			clocks = <&osc>;
 
diff --git a/arch/xtensa/include/asm/hw_breakpoint.h b/arch/xtensa/include/asm/hw_breakpoint.h
new file mode 100644
index 000000000000..dbe3053b284a
--- /dev/null
+++ b/arch/xtensa/include/asm/hw_breakpoint.h
@@ -0,0 +1,58 @@
+/*
+ * Xtensa hardware breakpoints/watchpoints handling 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) 2016 Cadence Design Systems Inc.
+ */
+
+#ifndef __ASM_XTENSA_HW_BREAKPOINT_H
+#define __ASM_XTENSA_HW_BREAKPOINT_H
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+
+#include <linux/kdebug.h>
+#include <linux/types.h>
+#include <uapi/linux/hw_breakpoint.h>
+
+/* Breakpoint */
+#define XTENSA_BREAKPOINT_EXECUTE	0
+
+/* Watchpoints */
+#define XTENSA_BREAKPOINT_LOAD		1
+#define XTENSA_BREAKPOINT_STORE		2
+
+struct arch_hw_breakpoint {
+	unsigned long address;
+	u16 len;
+	u16 type;
+};
+
+struct perf_event;
+struct pt_regs;
+struct task_struct;
+
+int hw_breakpoint_slots(int type);
+int arch_check_bp_in_kernelspace(struct perf_event *bp);
+int arch_validate_hwbkpt_settings(struct perf_event *bp);
+int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+				    unsigned long val, void *data);
+
+int arch_install_hw_breakpoint(struct perf_event *bp);
+void arch_uninstall_hw_breakpoint(struct perf_event *bp);
+void hw_breakpoint_pmu_read(struct perf_event *bp);
+int check_hw_breakpoint(struct pt_regs *regs);
+void clear_ptrace_hw_breakpoint(struct task_struct *tsk);
+
+#else
+
+struct task_struct;
+
+static inline void clear_ptrace_hw_breakpoint(struct task_struct *tsk)
+{
+}
+
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* __ASM_XTENSA_HW_BREAKPOINT_H */
diff --git a/arch/xtensa/include/asm/io.h b/arch/xtensa/include/asm/io.h
index 74fed0b4e2c2..c38e5a732d86 100644
--- a/arch/xtensa/include/asm/io.h
+++ b/arch/xtensa/include/asm/io.h
@@ -25,9 +25,12 @@
 
 #ifdef CONFIG_MMU
 
+void __iomem *xtensa_ioremap_nocache(unsigned long addr, unsigned long size);
+void __iomem *xtensa_ioremap_cache(unsigned long addr, unsigned long size);
+void xtensa_iounmap(volatile void __iomem *addr);
+
 /*
  * Return the virtual address for the specified bus memory.
- * Note that we currently don't support any address outside the KIO segment.
  */
 static inline void __iomem *ioremap_nocache(unsigned long offset,
 		unsigned long size)
@@ -36,7 +39,7 @@ static inline void __iomem *ioremap_nocache(unsigned long offset,
 	    && offset - XCHAL_KIO_PADDR < XCHAL_KIO_SIZE)
 		return (void*)(offset-XCHAL_KIO_PADDR+XCHAL_KIO_BYPASS_VADDR);
 	else
-		BUG();
+		return xtensa_ioremap_nocache(offset, size);
 }
 
 static inline void __iomem *ioremap_cache(unsigned long offset,
@@ -46,7 +49,7 @@ static inline void __iomem *ioremap_cache(unsigned long offset,
 	    && offset - XCHAL_KIO_PADDR < XCHAL_KIO_SIZE)
 		return (void*)(offset-XCHAL_KIO_PADDR+XCHAL_KIO_CACHED_VADDR);
 	else
-		BUG();
+		return xtensa_ioremap_cache(offset, size);
 }
 #define ioremap_cache ioremap_cache
 
@@ -60,6 +63,13 @@ static inline void __iomem *ioremap(unsigned long offset, unsigned long size)
 
 static inline void iounmap(volatile void __iomem *addr)
 {
+	unsigned long va = (unsigned long) addr;
+
+	if (!(va >= XCHAL_KIO_CACHED_VADDR &&
+	      va - XCHAL_KIO_CACHED_VADDR < XCHAL_KIO_SIZE) &&
+	    !(va >= XCHAL_KIO_BYPASS_VADDR &&
+	      va - XCHAL_KIO_BYPASS_VADDR < XCHAL_KIO_SIZE))
+		xtensa_iounmap(addr);
 }
 
 #define virt_to_bus     virt_to_phys
diff --git a/arch/xtensa/include/asm/irqflags.h b/arch/xtensa/include/asm/irqflags.h
index 8e090c709046..407606e576f8 100644
--- a/arch/xtensa/include/asm/irqflags.h
+++ b/arch/xtensa/include/asm/irqflags.h
@@ -13,6 +13,7 @@
 #define _XTENSA_IRQFLAGS_H
 
 #include <linux/types.h>
+#include <asm/processor.h>
 
 static inline unsigned long arch_local_save_flags(void)
 {
diff --git a/arch/xtensa/include/asm/processor.h b/arch/xtensa/include/asm/processor.h
index 83e2e4bc01ba..d2e40d39c615 100644
--- a/arch/xtensa/include/asm/processor.h
+++ b/arch/xtensa/include/asm/processor.h
@@ -78,22 +78,20 @@
 #define XTENSA_INTLEVEL_MASK(level) _XTENSA_INTLEVEL_MASK(level)
 #define _XTENSA_INTLEVEL_MASK(level) (XCHAL_INTLEVEL##level##_MASK)
 
-#define IS_POW2(v) (((v) & ((v) - 1)) == 0)
+#define XTENSA_INTLEVEL_ANDBELOW_MASK(l) _XTENSA_INTLEVEL_ANDBELOW_MASK(l)
+#define _XTENSA_INTLEVEL_ANDBELOW_MASK(l) (XCHAL_INTLEVEL##l##_ANDBELOW_MASK)
 
 #define PROFILING_INTLEVEL XTENSA_INT_LEVEL(XCHAL_PROFILING_INTERRUPT)
 
 /* LOCKLEVEL defines the interrupt level that masks all
  * general-purpose interrupts.
  */
-#if defined(CONFIG_XTENSA_VARIANT_HAVE_PERF_EVENTS) && \
-	defined(XCHAL_PROFILING_INTERRUPT) && \
-	PROFILING_INTLEVEL == XCHAL_EXCM_LEVEL && \
-	XCHAL_EXCM_LEVEL > 1 && \
-	IS_POW2(XTENSA_INTLEVEL_MASK(PROFILING_INTLEVEL))
-#define LOCKLEVEL (XCHAL_EXCM_LEVEL - 1)
+#if defined(CONFIG_XTENSA_FAKE_NMI) && defined(XCHAL_PROFILING_INTERRUPT)
+#define LOCKLEVEL (PROFILING_INTLEVEL - 1)
 #else
 #define LOCKLEVEL XCHAL_EXCM_LEVEL
 #endif
+
 #define TOPLEVEL XCHAL_EXCM_LEVEL
 #define XTENSA_FAKE_NMI (LOCKLEVEL < TOPLEVEL)
 
@@ -132,11 +130,10 @@ struct thread_struct {
 	unsigned long bad_vaddr; /* last user fault */
 	unsigned long bad_uaddr; /* last kernel fault accessing user space */
 	unsigned long error_code;
-
-	unsigned long ibreak[XCHAL_NUM_IBREAK];
-	unsigned long dbreaka[XCHAL_NUM_DBREAK];
-	unsigned long dbreakc[XCHAL_NUM_DBREAK];
-
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	struct perf_event *ptrace_bp[XCHAL_NUM_IBREAK];
+	struct perf_event *ptrace_wp[XCHAL_NUM_DBREAK];
+#endif
 	/* Make structure 16 bytes aligned. */
 	int align[0] __attribute__ ((aligned(16)));
 };
diff --git a/arch/xtensa/include/asm/regs.h b/arch/xtensa/include/asm/regs.h
index 4ba9f516b0e2..881a1134a4b4 100644
--- a/arch/xtensa/include/asm/regs.h
+++ b/arch/xtensa/include/asm/regs.h
@@ -28,6 +28,7 @@
 /*  Special registers.  */
 
 #define SREG_MR			32
+#define SREG_IBREAKENABLE	96
 #define SREG_IBREAKA		128
 #define SREG_DBREAKA		144
 #define SREG_DBREAKC		160
@@ -103,6 +104,8 @@
 
 /*  DEBUGCAUSE register fields.  */
 
+#define DEBUGCAUSE_DBNUM_MASK		0xf00
+#define DEBUGCAUSE_DBNUM_SHIFT		8	/* First bit of DBNUM field */
 #define DEBUGCAUSE_DEBUGINT_BIT		5	/* External debug interrupt */
 #define DEBUGCAUSE_BREAKN_BIT		4	/* BREAK.N instruction */
 #define DEBUGCAUSE_BREAK_BIT		3	/* BREAK instruction */
diff --git a/arch/xtensa/include/asm/thread_info.h b/arch/xtensa/include/asm/thread_info.h
index 9ad12c617184..7be2400f745a 100644
--- a/arch/xtensa/include/asm/thread_info.h
+++ b/arch/xtensa/include/asm/thread_info.h
@@ -111,6 +111,7 @@ static inline struct thread_info *current_thread_info(void)
 #define TIF_MEMDIE		5	/* is terminating due to OOM killer */
 #define TIF_RESTORE_SIGMASK	6	/* restore signal mask in do_signal() */
 #define TIF_NOTIFY_RESUME	7	/* callback before returning to user */
+#define TIF_DB_DISABLED		8	/* debug trap disabled for syscall */
 
 #define _TIF_SYSCALL_TRACE	(1<<TIF_SYSCALL_TRACE)
 #define _TIF_SIGPENDING		(1<<TIF_SIGPENDING)
diff --git a/arch/xtensa/include/asm/timex.h b/arch/xtensa/include/asm/timex.h
index ca929e6a38b5..f9b389d4e973 100644
--- a/arch/xtensa/include/asm/timex.h
+++ b/arch/xtensa/include/asm/timex.h
@@ -12,19 +12,16 @@
 #include <asm/processor.h>
 #include <linux/stringify.h>
 
-#define _INTLEVEL(x)	XCHAL_INT ## x ## _LEVEL
-#define INTLEVEL(x)	_INTLEVEL(x)
-
 #if XCHAL_NUM_TIMERS > 0 && \
-	INTLEVEL(XCHAL_TIMER0_INTERRUPT) <= XCHAL_EXCM_LEVEL
+	XTENSA_INT_LEVEL(XCHAL_TIMER0_INTERRUPT) <= XCHAL_EXCM_LEVEL
 # define LINUX_TIMER     0
 # define LINUX_TIMER_INT XCHAL_TIMER0_INTERRUPT
 #elif XCHAL_NUM_TIMERS > 1 && \
-	INTLEVEL(XCHAL_TIMER1_INTERRUPT) <= XCHAL_EXCM_LEVEL
+	XTENSA_INT_LEVEL(XCHAL_TIMER1_INTERRUPT) <= XCHAL_EXCM_LEVEL
 # define LINUX_TIMER     1
 # define LINUX_TIMER_INT XCHAL_TIMER1_INTERRUPT
 #elif XCHAL_NUM_TIMERS > 2 && \
-	INTLEVEL(XCHAL_TIMER2_INTERRUPT) <= XCHAL_EXCM_LEVEL
+	XTENSA_INT_LEVEL(XCHAL_TIMER2_INTERRUPT) <= XCHAL_EXCM_LEVEL
 # define LINUX_TIMER     2
 # define LINUX_TIMER_INT XCHAL_TIMER2_INTERRUPT
 #else
diff --git a/arch/xtensa/include/asm/traps.h b/arch/xtensa/include/asm/traps.h
index 28f33a8b7f5f..2e69aa4b843f 100644
--- a/arch/xtensa/include/asm/traps.h
+++ b/arch/xtensa/include/asm/traps.h
@@ -65,4 +65,21 @@ static inline void spill_registers(void)
 #endif
 }
 
+struct debug_table {
+	/* Pointer to debug exception handler */
+	void (*debug_exception)(void);
+	/* Temporary register save area */
+	unsigned long debug_save[1];
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	/* Save area for DBREAKC registers */
+	unsigned long dbreakc_save[XCHAL_NUM_DBREAK];
+	/* Saved ICOUNT register */
+	unsigned long icount_save;
+	/* Saved ICOUNTLEVEL register */
+	unsigned long icount_level_save;
+#endif
+};
+
+void debug_exception(void);
+
 #endif /* _XTENSA_TRAPS_H */
diff --git a/arch/xtensa/include/uapi/asm/ptrace.h b/arch/xtensa/include/uapi/asm/ptrace.h
index ee17aa842fdf..6ccbd9e38e35 100644
--- a/arch/xtensa/include/uapi/asm/ptrace.h
+++ b/arch/xtensa/include/uapi/asm/ptrace.h
@@ -72,6 +72,8 @@
 #define PTRACE_SETREGS		13
 #define PTRACE_GETXTREGS	18
 #define PTRACE_SETXTREGS	19
+#define PTRACE_GETHBPREGS	20
+#define PTRACE_SETHBPREGS	21
 
 
 #endif /* _UAPI_XTENSA_PTRACE_H */
diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile
index 4db730290d2d..c31f5d5afc7d 100644
--- a/arch/xtensa/kernel/Makefile
+++ b/arch/xtensa/kernel/Makefile
@@ -8,12 +8,12 @@ obj-y := align.o coprocessor.o entry.o irq.o pci-dma.o platform.o process.o \
 	 ptrace.o setup.o signal.o stacktrace.o syscall.o time.o traps.o \
 	 vectors.o
 
-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-$(CONFIG_SMP) += smp.o mxhead.o
 obj-$(CONFIG_XTENSA_VARIANT_HAVE_PERF_EVENTS) += perf_event.o
+obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
 
 AFLAGS_head.o += -mtext-section-literals
 AFLAGS_mxhead.o += -mtext-section-literals
diff --git a/arch/xtensa/kernel/asm-offsets.c b/arch/xtensa/kernel/asm-offsets.c
index b123ace3b67c..8e10e357ee32 100644
--- a/arch/xtensa/kernel/asm-offsets.c
+++ b/arch/xtensa/kernel/asm-offsets.c
@@ -23,6 +23,7 @@
 #include <linux/kbuild.h>
 
 #include <asm/ptrace.h>
+#include <asm/traps.h>
 #include <asm/uaccess.h>
 
 int main(void)
@@ -117,5 +118,16 @@ int main(void)
 	DEFINE(_CLONE_UNTRACED, CLONE_UNTRACED);
 	DEFINE(PG_ARCH_1, PG_arch_1);
 
+	/* struct debug_table */
+	DEFINE(DT_DEBUG_EXCEPTION,
+	       offsetof(struct debug_table, debug_exception));
+	DEFINE(DT_DEBUG_SAVE, offsetof(struct debug_table, debug_save));
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	DEFINE(DT_DBREAKC_SAVE, offsetof(struct debug_table, dbreakc_save));
+	DEFINE(DT_ICOUNT_SAVE, offsetof(struct debug_table, icount_save));
+	DEFINE(DT_ICOUNT_LEVEL_SAVE,
+	       offsetof(struct debug_table, icount_level_save));
+#endif
+
 	return 0;
 }
diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S
index db5c1765b413..fe8f7e7efb9d 100644
--- a/arch/xtensa/kernel/entry.S
+++ b/arch/xtensa/kernel/entry.S
@@ -543,6 +543,12 @@ common_exception_return:
 #endif
 
 5:
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	_bbci.l	a4, TIF_DB_DISABLED, 7f
+	movi	a4, restore_dbreak
+	callx4	a4
+7:
+#endif
 #ifdef CONFIG_DEBUG_TLB_SANITY
 	l32i	a4, a1, PT_DEPC
 	bgeui	a4, VALID_DOUBLE_EXCEPTION_ADDRESS, 4f
@@ -789,39 +795,99 @@ ENTRY(debug_exception)
 
 	movi	a2, 1 << PS_EXCM_BIT
 	or	a2, a0, a2
-	movi	a0, debug_exception	# restore a3, debug jump vector
 	wsr	a2, ps
-	xsr	a0, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
 
 	/* Switch to kernel/user stack, restore jump vector, and save a0 */
 
 	bbsi.l	a2, PS_UM_BIT, 2f	# jump if user mode
 
 	addi	a2, a1, -16-PT_SIZE	# assume kernel stack
+3:
+	l32i	a0, a3, DT_DEBUG_SAVE
+	s32i	a1, a2, PT_AREG1
 	s32i	a0, a2, PT_AREG0
 	movi	a0, 0
-	s32i	a1, a2, PT_AREG1
 	s32i	a0, a2, PT_DEPC		# mark it as a regular exception
+	xsr	a3, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
 	xsr	a0, depc
 	s32i	a3, a2, PT_AREG3
 	s32i	a0, a2, PT_AREG2
 	mov	a1, a2
+
+	/* Debug exception is handled as an exception, so interrupts will
+	 * likely be enabled in the common exception handler. Disable
+	 * preemption if we have HW breakpoints to preserve DEBUGCAUSE.DBNUM
+	 * meaning.
+	 */
+#if defined(CONFIG_PREEMPT_COUNT) && defined(CONFIG_HAVE_HW_BREAKPOINT)
+	GET_THREAD_INFO(a2, a1)
+	l32i	a3, a2, TI_PRE_COUNT
+	addi	a3, a3, 1
+	s32i	a3, a2, TI_PRE_COUNT
+#endif
+
+	rsr	a2, ps
+	bbsi.l	a2, PS_UM_BIT, _user_exception
 	j	_kernel_exception
 
 2:	rsr	a2, excsave1
 	l32i	a2, a2, EXC_TABLE_KSTK	# load kernel stack pointer
-	s32i	a0, a2, PT_AREG0
+	j	3b
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	/* Debug exception while in exception mode. This may happen when
+	 * window overflow/underflow handler or fast exception handler hits
+	 * data breakpoint, in which case save and disable all data
+	 * breakpoints, single-step faulting instruction and restore data
+	 * breakpoints.
+	 */
+1:
+	bbci.l	a0, PS_UM_BIT, 1b	# jump if kernel mode
+
+	rsr	a0, debugcause
+	bbsi.l	a0, DEBUGCAUSE_DBREAK_BIT, .Ldebug_save_dbreak
+
+	.set	_index, 0
+	.rept	XCHAL_NUM_DBREAK
+	l32i	a0, a3, DT_DBREAKC_SAVE + _index * 4
+	wsr	a0, SREG_DBREAKC + _index
+	.set	_index, _index + 1
+	.endr
+
+	l32i	a0, a3, DT_ICOUNT_LEVEL_SAVE
+	wsr	a0, icountlevel
+
+	l32i	a0, a3, DT_ICOUNT_SAVE
+	xsr	a0, icount
+
+	l32i	a0, a3, DT_DEBUG_SAVE
+	xsr	a3, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
+	rfi	XCHAL_DEBUGLEVEL
+
+.Ldebug_save_dbreak:
+	.set	_index, 0
+	.rept	XCHAL_NUM_DBREAK
 	movi	a0, 0
-	s32i	a1, a2, PT_AREG1
-	s32i	a0, a2, PT_DEPC
-	xsr	a0, depc
-	s32i	a3, a2, PT_AREG3
-	s32i	a0, a2, PT_AREG2
-	mov	a1, a2
-	j	_user_exception
+	xsr	a0, SREG_DBREAKC + _index
+	s32i	a0, a3, DT_DBREAKC_SAVE + _index * 4
+	.set	_index, _index + 1
+	.endr
 
-	/* Debug exception while in exception mode. */
+	movi	a0, XCHAL_EXCM_LEVEL + 1
+	xsr	a0, icountlevel
+	s32i	a0, a3, DT_ICOUNT_LEVEL_SAVE
+
+	movi	a0, 0xfffffffe
+	xsr	a0, icount
+	s32i	a0, a3, DT_ICOUNT_SAVE
+
+	l32i	a0, a3, DT_DEBUG_SAVE
+	xsr	a3, SREG_EXCSAVE + XCHAL_DEBUGLEVEL
+	rfi	XCHAL_DEBUGLEVEL
+#else
+	/* Debug exception while in exception mode. Should not happen. */
 1:	j	1b	// FIXME!!
+#endif
 
 ENDPROC(debug_exception)
 
diff --git a/arch/xtensa/kernel/head.S b/arch/xtensa/kernel/head.S
index 9ed55649ac8e..bc4f4bf05099 100644
--- a/arch/xtensa/kernel/head.S
+++ b/arch/xtensa/kernel/head.S
@@ -128,7 +128,7 @@ ENTRY(_startup)
 	wsr	a0, icountlevel
 
 	.set	_index, 0
-	.rept	XCHAL_NUM_DBREAK - 1
+	.rept	XCHAL_NUM_DBREAK
 	wsr	a0, SREG_DBREAKC + _index
 	.set	_index, _index + 1
 	.endr
@@ -197,11 +197,6 @@ ENTRY(_startup)
 	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
diff --git a/arch/xtensa/kernel/hw_breakpoint.c b/arch/xtensa/kernel/hw_breakpoint.c
new file mode 100644
index 000000000000..b35656ab7dbd
--- /dev/null
+++ b/arch/xtensa/kernel/hw_breakpoint.c
@@ -0,0 +1,317 @@
+/*
+ * Xtensa hardware breakpoints/watchpoints handling 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) 2016 Cadence Design Systems Inc.
+ */
+
+#include <linux/hw_breakpoint.h>
+#include <linux/log2.h>
+#include <linux/percpu.h>
+#include <linux/perf_event.h>
+#include <variant/core.h>
+
+/* Breakpoint currently in use for each IBREAKA. */
+static DEFINE_PER_CPU(struct perf_event *, bp_on_reg[XCHAL_NUM_IBREAK]);
+
+/* Watchpoint currently in use for each DBREAKA. */
+static DEFINE_PER_CPU(struct perf_event *, wp_on_reg[XCHAL_NUM_DBREAK]);
+
+int hw_breakpoint_slots(int type)
+{
+	switch (type) {
+	case TYPE_INST:
+		return XCHAL_NUM_IBREAK;
+	case TYPE_DATA:
+		return XCHAL_NUM_DBREAK;
+	default:
+		pr_warn("unknown slot type: %d\n", type);
+		return 0;
+	}
+}
+
+int arch_check_bp_in_kernelspace(struct perf_event *bp)
+{
+	unsigned int len;
+	unsigned long va;
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+
+	va = info->address;
+	len = bp->attr.bp_len;
+
+	return (va >= TASK_SIZE) && ((va + len - 1) >= TASK_SIZE);
+}
+
+/*
+ * Construct an arch_hw_breakpoint from a perf_event.
+ */
+static int arch_build_bp_info(struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+
+	/* Type */
+	switch (bp->attr.bp_type) {
+	case HW_BREAKPOINT_X:
+		info->type = XTENSA_BREAKPOINT_EXECUTE;
+		break;
+	case HW_BREAKPOINT_R:
+		info->type = XTENSA_BREAKPOINT_LOAD;
+		break;
+	case HW_BREAKPOINT_W:
+		info->type = XTENSA_BREAKPOINT_STORE;
+		break;
+	case HW_BREAKPOINT_RW:
+		info->type = XTENSA_BREAKPOINT_LOAD | XTENSA_BREAKPOINT_STORE;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Len */
+	info->len = bp->attr.bp_len;
+	if (info->len < 1 || info->len > 64 || !is_power_of_2(info->len))
+		return -EINVAL;
+
+	/* Address */
+	info->address = bp->attr.bp_addr;
+	if (info->address & (info->len - 1))
+		return -EINVAL;
+
+	return 0;
+}
+
+int arch_validate_hwbkpt_settings(struct perf_event *bp)
+{
+	int ret;
+
+	/* Build the arch_hw_breakpoint. */
+	ret = arch_build_bp_info(bp);
+	return ret;
+}
+
+int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+				    unsigned long val, void *data)
+{
+	return NOTIFY_DONE;
+}
+
+static void xtensa_wsr(unsigned long v, u8 sr)
+{
+	/* We don't have indexed wsr and creating instruction dynamically
+	 * doesn't seem worth it given how small XCHAL_NUM_IBREAK and
+	 * XCHAL_NUM_DBREAK are. Thus the switch. In case build breaks here
+	 * the switch below needs to be extended.
+	 */
+	BUILD_BUG_ON(XCHAL_NUM_IBREAK > 2);
+	BUILD_BUG_ON(XCHAL_NUM_DBREAK > 2);
+
+	switch (sr) {
+#if XCHAL_NUM_IBREAK > 0
+	case SREG_IBREAKA + 0:
+		WSR(v, SREG_IBREAKA + 0);
+		break;
+#endif
+#if XCHAL_NUM_IBREAK > 1
+	case SREG_IBREAKA + 1:
+		WSR(v, SREG_IBREAKA + 1);
+		break;
+#endif
+
+#if XCHAL_NUM_DBREAK > 0
+	case SREG_DBREAKA + 0:
+		WSR(v, SREG_DBREAKA + 0);
+		break;
+	case SREG_DBREAKC + 0:
+		WSR(v, SREG_DBREAKC + 0);
+		break;
+#endif
+#if XCHAL_NUM_DBREAK > 1
+	case SREG_DBREAKA + 1:
+		WSR(v, SREG_DBREAKA + 1);
+		break;
+
+	case SREG_DBREAKC + 1:
+		WSR(v, SREG_DBREAKC + 1);
+		break;
+#endif
+	}
+}
+
+static int alloc_slot(struct perf_event **slot, size_t n,
+		      struct perf_event *bp)
+{
+	size_t i;
+
+	for (i = 0; i < n; ++i) {
+		if (!slot[i]) {
+			slot[i] = bp;
+			return i;
+		}
+	}
+	return -EBUSY;
+}
+
+static void set_ibreak_regs(int reg, struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+	unsigned long ibreakenable;
+
+	xtensa_wsr(info->address, SREG_IBREAKA + reg);
+	RSR(ibreakenable, SREG_IBREAKENABLE);
+	WSR(ibreakenable | (1 << reg), SREG_IBREAKENABLE);
+}
+
+static void set_dbreak_regs(int reg, struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+	unsigned long dbreakc = DBREAKC_MASK_MASK & -info->len;
+
+	if (info->type & XTENSA_BREAKPOINT_LOAD)
+		dbreakc |= DBREAKC_LOAD_MASK;
+	if (info->type & XTENSA_BREAKPOINT_STORE)
+		dbreakc |= DBREAKC_STOR_MASK;
+
+	xtensa_wsr(info->address, SREG_DBREAKA + reg);
+	xtensa_wsr(dbreakc, SREG_DBREAKC + reg);
+}
+
+int arch_install_hw_breakpoint(struct perf_event *bp)
+{
+	int i;
+
+	if (counter_arch_bp(bp)->type == XTENSA_BREAKPOINT_EXECUTE) {
+		/* Breakpoint */
+		i = alloc_slot(this_cpu_ptr(bp_on_reg), XCHAL_NUM_IBREAK, bp);
+		if (i < 0)
+			return i;
+		set_ibreak_regs(i, bp);
+
+	} else {
+		/* Watchpoint */
+		i = alloc_slot(this_cpu_ptr(wp_on_reg), XCHAL_NUM_DBREAK, bp);
+		if (i < 0)
+			return i;
+		set_dbreak_regs(i, bp);
+	}
+	return 0;
+}