diff mbox

[v2,23/31] arm64: Debugging support

Message ID 1344966752-16102-24-git-send-email-catalin.marinas@arm.com (mailing list archive)
State New, archived
Headers show

Commit Message

Catalin Marinas Aug. 14, 2012, 5:52 p.m. UTC
From: Will Deacon <will.deacon@arm.com>

This patch adds ptrace, debug monitors and hardware breakpoints support.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
---
 arch/arm64/include/asm/debug-monitors.h |   88 +++
 arch/arm64/include/asm/hw_breakpoint.h  |  137 +++++
 arch/arm64/kernel/debug-monitors.c      |  288 ++++++++++
 arch/arm64/kernel/hw_breakpoint.c       |  880 +++++++++++++++++++++++++++++++
 arch/arm64/kernel/ptrace.c              |  834 +++++++++++++++++++++++++++++
 5 files changed, 2227 insertions(+), 0 deletions(-)
 create mode 100644 arch/arm64/include/asm/debug-monitors.h
 create mode 100644 arch/arm64/include/asm/hw_breakpoint.h
 create mode 100644 arch/arm64/kernel/debug-monitors.c
 create mode 100644 arch/arm64/kernel/hw_breakpoint.c
 create mode 100644 arch/arm64/kernel/ptrace.c

Comments

Arnd Bergmann Aug. 15, 2012, 3:07 p.m. UTC | #1
On Tuesday 14 August 2012, Catalin Marinas wrote:

> +const struct user_regset_view *task_user_regset_view(struct task_struct *task)
> +{
> +#ifdef CONFIG_AARCH32_EMULATION
> +	if (test_tsk_thread_flag(task, TIF_32BIT))
> +		return &user_aarch32_view;
> +#endif
> +	return &user_aarch64_view;
> +}

Ah, nice. So you support 64 bit debuggers debugging 32 bit processes, right?

From what I can tell, there is no support for 32 bit processes debugging
64 bit ones. Is that something you plan to add in the future, or do you
consider that out of scope? In either case, a comment would be helpful.

> +long arch_ptrace(struct task_struct *child, long request,
> +		 unsigned long addr, unsigned long data)
> +{
> +	int ret;
> +	unsigned long *datap = (unsigned long __user *)data;
> +
> +	switch (request) {
> +		case PTRACE_GET_THREAD_AREA:
> +			ret = put_user(child->thread.tp_value, datap);
> +			break;
> +
> +#ifdef CONFIG_HAVE_HW_BREAKPOINT
> +		case PTRACE_GETHBPREGS:
> +			ret = ptrace_gethbpregs(child, addr, datap);
> +			break;
> +
> +		case PTRACE_SETHBPREGS:
> +			ret = ptrace_sethbpregs(child, addr, datap);
> +			break;
> +#endif
> +
> +		default:
> +			ret = ptrace_request(child, request, addr, data);
> +			break;
> +	}
> +
> +	return ret;
> +}

Is there a reaons why these are not regsets but have their own ptrace
commands? I believe new architectures should generally not add ptrace
commands any more.

	Arnd
Will Deacon Aug. 16, 2012, 10:47 a.m. UTC | #2
On Wed, Aug 15, 2012 at 04:07:36PM +0100, Arnd Bergmann wrote:
> On Tuesday 14 August 2012, Catalin Marinas wrote:
> 
> > +const struct user_regset_view *task_user_regset_view(struct task_struct *task)
> > +{
> > +#ifdef CONFIG_AARCH32_EMULATION
> > +	if (test_tsk_thread_flag(task, TIF_32BIT))
> > +		return &user_aarch32_view;
> > +#endif
> > +	return &user_aarch64_view;
> > +}
> 
> Ah, nice. So you support 64 bit debuggers debugging 32 bit processes, right?

That should work if the debugger can deal with it, yes.

> From what I can tell, there is no support for 32 bit processes debugging
> 64 bit ones. Is that something you plan to add in the future, or do you
> consider that out of scope? In either case, a comment would be helpful.

That can't really work because the debugger won't be able to manipulate
child pointers properly without us adding a new ptrace interface (and then,
I still wonder about how feasible it really is). I can add a comment.

> > +long arch_ptrace(struct task_struct *child, long request,
> > +		 unsigned long addr, unsigned long data)
> > +{
> > +	int ret;
> > +	unsigned long *datap = (unsigned long __user *)data;
> > +
> > +	switch (request) {
> > +		case PTRACE_GET_THREAD_AREA:
> > +			ret = put_user(child->thread.tp_value, datap);
> > +			break;
> > +
> > +#ifdef CONFIG_HAVE_HW_BREAKPOINT
> > +		case PTRACE_GETHBPREGS:
> > +			ret = ptrace_gethbpregs(child, addr, datap);
> > +			break;
> > +
> > +		case PTRACE_SETHBPREGS:
> > +			ret = ptrace_sethbpregs(child, addr, datap);
> > +			break;
> > +#endif
> > +
> > +		default:
> > +			ret = ptrace_request(child, request, addr, data);
> > +			break;
> > +	}
> > +
> > +	return ret;
> > +}
> 
> Is there a reaons why these are not regsets but have their own ptrace
> commands? I believe new architectures should generally not add ptrace
> commands any more.

I could probably add some regset wrappers about the hbp accessors (which we
have to keep for the compat ptrace interface). I'll have a think as it might
even make sense to have different regsets for breakpoints and watchpoints.

As for the the tls, is it worth having a regset with only one register?

Will
Arnd Bergmann Aug. 16, 2012, 12:49 p.m. UTC | #3
On Thursday 16 August 2012, Will Deacon wrote:
> On Wed, Aug 15, 2012 at 04:07:36PM +0100, Arnd Bergmann wrote:
> > On Tuesday 14 August 2012, Catalin Marinas wrote:

> > From what I can tell, there is no support for 32 bit processes debugging
> > 64 bit ones. Is that something you plan to add in the future, or do you
> > consider that out of scope? In either case, a comment would be helpful.
> 
> That can't really work because the debugger won't be able to manipulate
> child pointers properly without us adding a new ptrace interface (and then,
> I still wonder about how feasible it really is). I can add a comment.

You can already have a 32 bit gdb that is able to do remote debugging of
64 bit processes using a gdb server process. I guess it wouldn't be
too strange to have a ptrace extension to allow the native case as well.
I agree it's not a high priority.

> > > +long arch_ptrace(struct task_struct *child, long request,
> > > +		 unsigned long addr, unsigned long data)
> > > +{
> > > +	int ret;
> > > +	unsigned long *datap = (unsigned long __user *)data;
> > > +
> > > +	switch (request) {
> > > +		case PTRACE_GET_THREAD_AREA:
> > > +			ret = put_user(child->thread.tp_value, datap);
> > > +			break;
> > > +
> > > +#ifdef CONFIG_HAVE_HW_BREAKPOINT
> > > +		case PTRACE_GETHBPREGS:
> > > +			ret = ptrace_gethbpregs(child, addr, datap);
> > > +			break;
> > > +
> > > +		case PTRACE_SETHBPREGS:
> > > +			ret = ptrace_sethbpregs(child, addr, datap);
> > > +			break;
> > > +#endif
> > > +
> > > +		default:
> > > +			ret = ptrace_request(child, request, addr, data);
> > > +			break;
> > > +	}
> > > +
> > > +	return ret;
> > > +}
> > 
> > Is there a reaons why these are not regsets but have their own ptrace
> > commands? I believe new architectures should generally not add ptrace
> > commands any more.
> 
> I could probably add some regset wrappers about the hbp accessors (which we
> have to keep for the compat ptrace interface). I'll have a think as it might
> even make sense to have different regsets for breakpoints and watchpoints.
> 
> As for the the tls, is it worth having a regset with only one register?

Better ask the gdb folks. I'm adding Uli to Cc, maybe he has some insight.

	Arnd
Arnd Bergmann Aug. 17, 2012, 7:06 a.m. UTC | #4
On Thursday 16 August 2012, Arnd Bergmann wrote:
> On Thursday 16 August 2012, Will Deacon wrote:
> > On Wed, Aug 15, 2012 at 04:07:36PM +0100, Arnd Bergmann wrote:
> > > On Tuesday 14 August 2012, Catalin Marinas wrote:
> 
> > > From what I can tell, there is no support for 32 bit processes debugging
> > > 64 bit ones. Is that something you plan to add in the future, or do you
> > > consider that out of scope? In either case, a comment would be helpful.
> > 
> > That can't really work because the debugger won't be able to manipulate
> > child pointers properly without us adding a new ptrace interface (and then,
> > I still wonder about how feasible it really is). I can add a comment.
> 
> You can already have a 32 bit gdb that is able to do remote debugging of
> 64 bit processes using a gdb server process. I guess it wouldn't be
> too strange to have a ptrace extension to allow the native case as well.
> I agree it's not a high priority.
> 
> > > > +long arch_ptrace(struct task_struct *child, long request,
> > > > +		 unsigned long addr, unsigned long data)
> > > > +{
> > > > +	int ret;
> > > > +	unsigned long *datap = (unsigned long __user *)data;
> > > > +
> > > > +	switch (request) {
> > > > +		case PTRACE_GET_THREAD_AREA:
> > > > +			ret = put_user(child->thread.tp_value, datap);
> > > > +			break;
> > > > +
> > > > +#ifdef CONFIG_HAVE_HW_BREAKPOINT
> > > > +		case PTRACE_GETHBPREGS:
> > > > +			ret = ptrace_gethbpregs(child, addr, datap);
> > > > +			break;
> > > > +
> > > > +		case PTRACE_SETHBPREGS:
> > > > +			ret = ptrace_sethbpregs(child, addr, datap);
> > > > +			break;
> > > > +#endif
> > > > +
> > > > +		default:
> > > > +			ret = ptrace_request(child, request, addr, data);
> > > > +			break;
> > > > +	}
> > > > +
> > > > +	return ret;
> > > > +}
> > > 
> > > Is there a reaons why these are not regsets but have their own ptrace
> > > commands? I believe new architectures should generally not add ptrace
> > > commands any more.
> > 
> > I could probably add some regset wrappers about the hbp accessors (which we
> > have to keep for the compat ptrace interface). I'll have a think as it might
> > even make sense to have different regsets for breakpoints and watchpoints.
> > 
> > As for the the tls, is it worth having a regset with only one register?
> 
> Better ask the gdb folks. I'm adding Uli to Cc, maybe he has some insight.

Sorry for the dumb question, but why do you even need PTRACE_GET_THREAD_AREA
for 64 bit tasks? I thought the thread pointer is a GPR, or is this just
for compat tasks?

	Arnd
Will Deacon Aug. 20, 2012, 9:07 a.m. UTC | #5
On Fri, Aug 17, 2012 at 08:06:07AM +0100, Arnd Bergmann wrote:
> Sorry for the dumb question, but why do you even need PTRACE_GET_THREAD_AREA
> for 64 bit tasks? I thought the thread pointer is a GPR, or is this just
> for compat tasks?

The TLS is stored in a co-processor register which is read-only for
userspace. However, it's banked, so accessing the TLS of your child (the guy
you're ptracing) requires a system call.

Will
Will Deacon Aug. 20, 2012, 9:27 a.m. UTC | #6
On Mon, Aug 20, 2012 at 10:07:54AM +0100, Will Deacon wrote:
> On Fri, Aug 17, 2012 at 08:06:07AM +0100, Arnd Bergmann wrote:
> > Sorry for the dumb question, but why do you even need PTRACE_GET_THREAD_AREA
> > for 64 bit tasks? I thought the thread pointer is a GPR, or is this just
> > for compat tasks?
> 
> The TLS is stored in a co-processor register which is read-only for
> userspace.

I should elaborate: the register is RW for AArch64 tasks, RO for aarch32
tasks (although that doesn't affect the need for the ptrace request).

Will
Arnd Bergmann Aug. 20, 2012, 8:10 p.m. UTC | #7
On Monday 20 August 2012, Will Deacon wrote:
> On Mon, Aug 20, 2012 at 10:07:54AM +0100, Will Deacon wrote:
> > On Fri, Aug 17, 2012 at 08:06:07AM +0100, Arnd Bergmann wrote:
> > > Sorry for the dumb question, but why do you even need PTRACE_GET_THREAD_AREA
> > > for 64 bit tasks? I thought the thread pointer is a GPR, or is this just
> > > for compat tasks?
> > 
> > The TLS is stored in a co-processor register which is read-only for
> > userspace.
> 
> I should elaborate: the register is RW for AArch64 tasks, RO for aarch32
> tasks (although that doesn't affect the need for the ptrace request).

So can't you just /add/ that register to the GPR regset?

	Arnd
Will Deacon Aug. 21, 2012, 8:58 a.m. UTC | #8
On Mon, Aug 20, 2012 at 09:10:59PM +0100, Arnd Bergmann wrote:
> On Monday 20 August 2012, Will Deacon wrote:
> > On Mon, Aug 20, 2012 at 10:07:54AM +0100, Will Deacon wrote:
> > > On Fri, Aug 17, 2012 at 08:06:07AM +0100, Arnd Bergmann wrote:
> > > > Sorry for the dumb question, but why do you even need PTRACE_GET_THREAD_AREA
> > > > for 64 bit tasks? I thought the thread pointer is a GPR, or is this just
> > > > for compat tasks?
> > > 
> > > The TLS is stored in a co-processor register which is read-only for
> > > userspace.
> > 
> > I should elaborate: the register is RW for AArch64 tasks, RO for aarch32
> > tasks (although that doesn't affect the need for the ptrace request).
> 
> So can't you just /add/ that register to the GPR regset?

We *could*, but it doesn't feel right to me. The TLS register:

	- Can only be manipulated by msr/mrs instructions
	- Is not part of the PCS
	- Is not accessed explicitly by user applications
	- Can not be written via ptrace

so I think it sticks out like a sore thumb if we bundle it up with the GPRs.

Will
diff mbox

Patch

diff --git a/arch/arm64/include/asm/debug-monitors.h b/arch/arm64/include/asm/debug-monitors.h
new file mode 100644
index 0000000..7eaa0b3
--- /dev/null
+++ b/arch/arm64/include/asm/debug-monitors.h
@@ -0,0 +1,88 @@ 
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __ASM_DEBUG_MONITORS_H
+#define __ASM_DEBUG_MONITORS_H
+
+#ifdef __KERNEL__
+
+#define	DBG_ESR_EVT(x)		(((x) >> 27) & 0x7)
+
+/* AArch64 */
+#define DBG_ESR_EVT_HWBP	0x0
+#define DBG_ESR_EVT_HWSS	0x1
+#define DBG_ESR_EVT_HWWP	0x2
+#define DBG_ESR_EVT_BRK		0x6
+
+enum debug_el {
+	DBG_ACTIVE_EL0 = 0,
+	DBG_ACTIVE_EL1,
+};
+
+/* AArch32 */
+#define DBG_ESR_EVT_BKPT	0x4
+#define DBG_ESR_EVT_VECC	0x5
+
+#define AARCH32_BREAK_ARM	0x07f001f0
+#define AARCH32_BREAK_THUMB	0xde01
+#define AARCH32_BREAK_THUMB2_LO	0xf7f0
+#define AARCH32_BREAK_THUMB2_HI	0xa000
+
+#ifndef __ASSEMBLY__
+struct task_struct;
+
+#define local_dbg_save(flags)							\
+	do {									\
+		typecheck(unsigned long, flags);				\
+		asm volatile(							\
+		"mrs	%0, daif			// local_dbg_save\n"	\
+		"msr	daifset, #8"						\
+		: "=r" (flags) : : "memory");					\
+	} while (0)
+
+#define local_dbg_restore(flags)						\
+	do {									\
+		typecheck(unsigned long, flags);				\
+		asm volatile(							\
+		"msr	daif, %0			// local_dbg_restore\n"	\
+		: : "r" (flags) : "memory");					\
+	} while (0)
+
+#define DBG_ARCH_ID_RESERVED	0	/* In case of ptrace ABI updates. */
+
+u8 debug_monitors_arch(void);
+
+void enable_debug_monitors(enum debug_el el);
+void disable_debug_monitors(enum debug_el el);
+
+void user_rewind_single_step(struct task_struct *task);
+void user_fastforward_single_step(struct task_struct *task);
+
+void kernel_enable_single_step(struct pt_regs *regs);
+void kernel_disable_single_step(void);
+int kernel_active_single_step(void);
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+int reinstall_suspended_bps(struct pt_regs *regs);
+#else
+static inline int reinstall_suspended_bps(struct pt_regs *regs)
+{
+	return -ENODEV;
+}
+#endif
+
+#endif	/* __ASSEMBLY */
+#endif	/* __KERNEL__ */
+#endif	/* __ASM_DEBUG_MONITORS_H */
diff --git a/arch/arm64/include/asm/hw_breakpoint.h b/arch/arm64/include/asm/hw_breakpoint.h
new file mode 100644
index 0000000..d064047
--- /dev/null
+++ b/arch/arm64/include/asm/hw_breakpoint.h
@@ -0,0 +1,137 @@ 
+/*
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+#ifndef __ASM_HW_BREAKPOINT_H
+#define __ASM_HW_BREAKPOINT_H
+
+#ifdef __KERNEL__
+
+struct arch_hw_breakpoint_ctrl {
+	u32 __reserved	: 19,
+	len		: 8,
+	type		: 2,
+	privilege	: 2,
+	enabled		: 1;
+};
+
+struct arch_hw_breakpoint {
+	u64 address;
+	u64 trigger;
+	struct arch_hw_breakpoint_ctrl ctrl;
+};
+
+static inline u32 encode_ctrl_reg(struct arch_hw_breakpoint_ctrl ctrl)
+{
+	return (ctrl.len << 5) | (ctrl.type << 3) | (ctrl.privilege << 1) |
+		ctrl.enabled;
+}
+
+static inline void decode_ctrl_reg(u32 reg,
+				   struct arch_hw_breakpoint_ctrl *ctrl)
+{
+	ctrl->enabled	= reg & 0x1;
+	reg >>= 1;
+	ctrl->privilege	= reg & 0x3;
+	reg >>= 2;
+	ctrl->type	= reg & 0x3;
+	reg >>= 2;
+	ctrl->len	= reg & 0xff;
+}
+
+/* Breakpoint */
+#define ARM_BREAKPOINT_EXECUTE	0
+
+/* Watchpoints */
+#define ARM_BREAKPOINT_LOAD	1
+#define ARM_BREAKPOINT_STORE	2
+#define AARCH64_ESR_ACCESS_MASK	(1 << 6)
+
+/* Privilege Levels */
+#define AARCH64_BREAKPOINT_EL1	1
+#define AARCH64_BREAKPOINT_EL0	2
+
+/* Lengths */
+#define ARM_BREAKPOINT_LEN_1	0x1
+#define ARM_BREAKPOINT_LEN_2	0x3
+#define ARM_BREAKPOINT_LEN_4	0xf
+#define ARM_BREAKPOINT_LEN_8	0xff
+
+/* Kernel stepping */
+#define ARM_KERNEL_STEP_NONE	0
+#define ARM_KERNEL_STEP_ACTIVE	1
+#define ARM_KERNEL_STEP_SUSPEND	2
+
+/*
+ * Limits.
+ * Changing these will require modifications to the register accessors.
+ */
+#define ARM_MAX_BRP		16
+#define ARM_MAX_WRP		16
+#define ARM_MAX_HBP_SLOTS	(ARM_MAX_BRP + ARM_MAX_WRP)
+
+/* Virtual debug register bases. */
+#define AARCH64_DBG_REG_BVR	0
+#define AARCH64_DBG_REG_BCR	(AARCH64_DBG_REG_BVR + ARM_MAX_BRP)
+#define AARCH64_DBG_REG_WVR	(AARCH64_DBG_REG_BCR + ARM_MAX_BRP)
+#define AARCH64_DBG_REG_WCR	(AARCH64_DBG_REG_WVR + ARM_MAX_WRP)
+
+/* Debug register names. */
+#define AARCH64_DBG_REG_NAME_BVR	"bvr"
+#define AARCH64_DBG_REG_NAME_BCR	"bcr"
+#define AARCH64_DBG_REG_NAME_WVR	"wvr"
+#define AARCH64_DBG_REG_NAME_WCR	"wcr"
+
+/* Accessor macros for the debug registers. */
+#define AARCH64_DBG_READ(N, REG, VAL) do {\
+	asm volatile("mrs %0, dbg" REG #N "_el1" : "=r" (VAL));\
+} while (0)
+
+#define AARCH64_DBG_WRITE(N, REG, VAL) do {\
+	asm volatile("msr dbg" REG #N "_el1, %0" :: "r" (VAL));\
+} while (0)
+
+struct task_struct;
+struct notifier_block;
+struct perf_event;
+struct pmu;
+
+extern int arch_bp_generic_fields(struct arch_hw_breakpoint_ctrl ctrl,
+				  int *gen_len, int *gen_type);
+extern int arch_check_bp_in_kernelspace(struct perf_event *bp);
+extern int arch_validate_hwbkpt_settings(struct perf_event *bp);
+extern int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+					   unsigned long val, void *data);
+
+extern int arch_install_hw_breakpoint(struct perf_event *bp);
+extern void arch_uninstall_hw_breakpoint(struct perf_event *bp);
+extern void hw_breakpoint_pmu_read(struct perf_event *bp);
+extern int hw_breakpoint_slots(int type);
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+extern void hw_breakpoint_thread_switch(struct task_struct *next);
+extern void ptrace_hw_copy_thread(struct task_struct *task);
+#else
+static inline void hw_breakpoint_thread_switch(struct task_struct *next)
+{
+}
+static inline void ptrace_hw_copy_thread(struct task_struct *task)
+{
+}
+#endif
+
+extern struct pmu perf_ops_bp;
+
+#endif	/* __KERNEL__ */
+#endif	/* __ASM_BREAKPOINT_H */
diff --git a/arch/arm64/kernel/debug-monitors.c b/arch/arm64/kernel/debug-monitors.c
new file mode 100644
index 0000000..0c3ba9f
--- /dev/null
+++ b/arch/arm64/kernel/debug-monitors.c
@@ -0,0 +1,288 @@ 
+/*
+ * ARMv8 single-step debug support and mdscr context switching.
+ *
+ * Copyright (C) 2012 ARM Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Author: Will Deacon <will.deacon@arm.com>
+ */
+
+#include <linux/cpu.h>
+#include <linux/debugfs.h>
+#include <linux/hardirq.h>
+#include <linux/init.h>
+#include <linux/ptrace.h>
+#include <linux/stat.h>
+
+#include <asm/debug-monitors.h>
+#include <asm/local.h>
+#include <asm/cputype.h>
+#include <asm/system_misc.h>
+
+/* Low-level stepping controls. */
+#define DBG_MDSCR_SS		(1 << 0)
+#define DBG_SPSR_SS		(1 << 21)
+
+/* MDSCR_EL1 enabling bits */
+#define DBG_MDSCR_KDE		(1 << 13)
+#define DBG_MDSCR_MDE		(1 << 15)
+#define DBG_MDSCR_MASK		~(DBG_MDSCR_KDE | DBG_MDSCR_MDE)
+
+/* Determine debug architecture. */
+u8 debug_monitors_arch(void)
+{
+	return read_cpuid(ID_AA64DFR0_EL1) & 0xf;
+}
+
+/*
+ * MDSCR access routines.
+ */
+static void mdscr_write(u32 mdscr)
+{
+	unsigned long flags;
+	local_dbg_save(flags);
+	asm volatile("msr mdscr_el1, %0" :: "r" (mdscr));
+	local_dbg_restore(flags);
+}
+
+static u32 mdscr_read(void)
+{
+	u32 mdscr;
+	asm volatile("mrs %0, mdscr_el1" : "=r" (mdscr));
+	return mdscr;
+}
+
+/*
+ * Allow root to disable self-hosted debug from userspace.
+ * This is useful if you want to connect an external JTAG debugger.
+ */
+static u32 debug_enabled = 1;
+
+static int create_debug_debugfs_entry(void)
+{
+	debugfs_create_bool("debug_enabled", 0644, NULL, &debug_enabled);
+	return 0;
+}
+fs_initcall(create_debug_debugfs_entry);
+
+static int __init early_debug_disable(char *buf)
+{
+	debug_enabled = 0;
+	return 0;
+}
+
+early_param("nodebugmon", early_debug_disable);
+
+/*
+ * Keep track of debug users on each core.
+ * The ref counts are per-cpu so we use a local_t type.
+ */
+static DEFINE_PER_CPU(local_t, mde_ref_count);
+static DEFINE_PER_CPU(local_t, kde_ref_count);
+
+void enable_debug_monitors(enum debug_el el)
+{
+	u32 mdscr, enable = 0;
+
+	WARN_ON(preemptible());
+
+	if (local_inc_return(&__get_cpu_var(mde_ref_count)) == 1)
+		enable = DBG_MDSCR_MDE;
+
+	if (el == DBG_ACTIVE_EL1 &&
+	    local_inc_return(&__get_cpu_var(kde_ref_count)) == 1)
+		enable |= DBG_MDSCR_KDE;
+
+	if (enable && debug_enabled) {
+		mdscr = mdscr_read();
+		mdscr |= enable;
+		mdscr_write(mdscr);
+	}
+}
+
+void disable_debug_monitors(enum debug_el el)
+{
+	u32 mdscr, disable = 0;
+
+	WARN_ON(preemptible());
+
+	if (local_dec_and_test(&__get_cpu_var(mde_ref_count)))
+		disable = ~DBG_MDSCR_MDE;
+
+	if (el == DBG_ACTIVE_EL1 &&
+	    local_dec_and_test(&__get_cpu_var(kde_ref_count)))
+		disable &= ~DBG_MDSCR_KDE;
+
+	if (disable) {
+		mdscr = mdscr_read();
+		mdscr &= disable;
+		mdscr_write(mdscr);
+	}
+}
+
+/*
+ * OS lock clearing.
+ */
+static void clear_os_lock(void *unused)
+{
+	asm volatile("msr mdscr_el1, %0" : : "r" (0));
+	isb();
+	asm volatile("msr oslar_el1, %0" : : "r" (0));
+	isb();
+}
+
+static int __cpuinit os_lock_notify(struct notifier_block *self,
+				    unsigned long action, void *data)
+{
+	int cpu = (unsigned long)data;
+	if (action == CPU_ONLINE)
+		smp_call_function_single(cpu, clear_os_lock, NULL, 1);
+	return NOTIFY_OK;
+}
+
+static struct notifier_block __cpuinitdata os_lock_nb = {
+	.notifier_call = os_lock_notify,
+};
+
+static int __cpuinit debug_monitors_init(void)
+{
+	/* Clear the OS lock. */
+	smp_call_function(clear_os_lock, NULL, 1);
+	clear_os_lock(NULL);
+
+	/* Register hotplug handler. */
+	register_cpu_notifier(&os_lock_nb);
+	return 0;
+}
+postcore_initcall(debug_monitors_init);
+
+/*
+ * Single step API and exception handling.
+ */
+static void set_regs_spsr_ss(struct pt_regs *regs)
+{
+	unsigned long spsr;
+
+	spsr = regs->pstate;
+	spsr &= ~DBG_SPSR_SS;
+	spsr |= DBG_SPSR_SS;
+	regs->pstate = spsr;
+}
+
+static void clear_regs_spsr_ss(struct pt_regs *regs)
+{
+	unsigned long spsr;
+
+	spsr = regs->pstate;
+	spsr &= ~DBG_SPSR_SS;
+	regs->pstate = spsr;
+}
+
+static int single_step_handler(unsigned long addr, unsigned int esr,
+			       struct pt_regs *regs)
+{
+	siginfo_t info;
+
+	/*
+	 * If we are stepping a pending breakpoint, call the hw_breakpoint
+	 * handler first.
+	 */
+	if (!reinstall_suspended_bps(regs))
+		return 0;
+
+	if (user_mode(regs)) {
+		info.si_signo = SIGTRAP;
+		info.si_errno = 0;
+		info.si_code  = TRAP_HWBKPT;
+		info.si_addr  = (void __user *)instruction_pointer(regs);
+		force_sig_info(SIGTRAP, &info, current);
+
+		/*
+		 * ptrace will disable single step unless explicitly
+		 * asked to re-enable it. For other clients, it makes
+		 * sense to leave it enabled (i.e. rewind the controls
+		 * to the active-not-pending state).
+		 */
+		user_rewind_single_step(current);
+	} else {
+		/* TODO: route to KGDB */
+		pr_warning("Unexpected kernel single-step exception at EL1\n");
+		/*
+		 * Re-enable stepping since we know that we will be
+		 * returning to regs.
+		 */
+		set_regs_spsr_ss(regs);
+	}
+
+	return 0;
+}
+
+static int __init single_step_init(void)
+{
+	hook_debug_fault_code(DBG_ESR_EVT_HWSS, single_step_handler, SIGTRAP,
+			      TRAP_HWBKPT, "single-step handler");
+	return 0;
+}
+arch_initcall(single_step_init);
+
+/* Re-enable single step for syscall restarting. */
+void user_rewind_single_step(struct task_struct *task)
+{
+	/*
+	 * If single step is active for this thread, then set SPSR.SS
+	 * to 1 to avoid returning to the active-pending state.
+	 */
+	if (test_ti_thread_flag(task_thread_info(task), TIF_SINGLESTEP))
+		set_regs_spsr_ss(task_pt_regs(task));
+}
+
+void user_fastforward_single_step(struct task_struct *task)
+{
+	if (test_ti_thread_flag(task_thread_info(task), TIF_SINGLESTEP))
+		clear_regs_spsr_ss(task_pt_regs(task));
+}
+
+/* Kernel API */
+void kernel_enable_single_step(struct pt_regs *regs)
+{
+	WARN_ON(!irqs_disabled());
+	set_regs_spsr_ss(regs);
+	mdscr_write(mdscr_read() | DBG_MDSCR_SS);
+	enable_debug_monitors(DBG_ACTIVE_EL1);
+}
+
+void kernel_disable_single_step(void)
+{
+	WARN_ON(!irqs_disabled());
+	mdscr_write(mdscr_read() & ~DBG_MDSCR_SS);
+	disable_debug_monitors(DBG_ACTIVE_EL1);
+}
+
+int kernel_active_single_step(void)
+{
+	WARN_ON(!irqs_disabled());
+	return mdscr_read() & DBG_MDSCR_SS;
+}
+
+/* ptrace API */
+void user_enable_single_step(struct task_struct *task)
+{
+	set_ti_thread_flag(task_thread_info(task), TIF_SINGLESTEP);
+	set_regs_spsr_ss(task_pt_regs(task));
+}
+
+void user_disable_single_step(struct task_struct *task)
+{
+	clear_ti_thread_flag(task_thread_info(task), TIF_SINGLESTEP);
+}
diff --git a/arch/arm64/kernel/hw_breakpoint.c b/arch/arm64/kernel/hw_breakpoint.c
new file mode 100644
index 0000000..5ab825c
--- /dev/null
+++ b/arch/arm64/kernel/hw_breakpoint.c
@@ -0,0 +1,880 @@ 
+/*
+ * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
+ * using the CPU's debug registers.
+ *
+ * Copyright (C) 2012 ARM Limited
+ * Author: Will Deacon <will.deacon@arm.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define pr_fmt(fmt) "hw-breakpoint: " fmt
+
+#include <linux/errno.h>
+#include <linux/hw_breakpoint.h>
+#include <linux/perf_event.h>
+#include <linux/ptrace.h>
+#include <linux/smp.h>
+
+#include <asm/compat.h>
+#include <asm/current.h>
+#include <asm/debug-monitors.h>
+#include <asm/hw_breakpoint.h>
+#include <asm/kdebug.h>
+#include <asm/traps.h>
+#include <asm/cputype.h>
+#include <asm/system_misc.h>
+
+/* Breakpoint currently in use for each BRP. */
+static DEFINE_PER_CPU(struct perf_event *, bp_on_reg[ARM_MAX_BRP]);
+
+/* Watchpoint currently in use for each WRP. */
+static DEFINE_PER_CPU(struct perf_event *, wp_on_reg[ARM_MAX_WRP]);
+
+/* Currently stepping a per-CPU kernel breakpoint. */
+static DEFINE_PER_CPU(int, stepping_kernel_bp);
+
+/* Number of BRP/WRP registers on this CPU. */
+static int core_num_brps;
+static int core_num_wrps;
+
+/* Determine number of BRP registers available. */
+static int get_num_brps(void)
+{
+	return ((read_cpuid(ID_AA64DFR0_EL1) >> 12) & 0xf) + 1;
+}
+
+/* Determine number of WRP registers available. */
+static int get_num_wrps(void)
+{
+	return ((read_cpuid(ID_AA64DFR0_EL1) >> 20) & 0xf) + 1;
+}
+
+int hw_breakpoint_slots(int type)
+{
+	/*
+	 * We can be called early, so don't rely on
+	 * our static variables being initialised.
+	 */
+	switch (type) {
+	case TYPE_INST:
+		return get_num_brps();
+	case TYPE_DATA:
+		return get_num_wrps();
+	default:
+		pr_warning("unknown slot type: %d\n", type);
+		return 0;
+	}
+}
+
+#define READ_WB_REG_CASE(OFF, N, REG, VAL)	\
+	case (OFF + N):				\
+		AARCH64_DBG_READ(N, REG, VAL);	\
+		break
+
+#define WRITE_WB_REG_CASE(OFF, N, REG, VAL)	\
+	case (OFF + N):				\
+		AARCH64_DBG_WRITE(N, REG, VAL);	\
+		break
+
+#define GEN_READ_WB_REG_CASES(OFF, REG, VAL)	\
+	READ_WB_REG_CASE(OFF,  0, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  1, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  2, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  3, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  4, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  5, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  6, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  7, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  8, REG, VAL);	\
+	READ_WB_REG_CASE(OFF,  9, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 10, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 11, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 12, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 13, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 14, REG, VAL);	\
+	READ_WB_REG_CASE(OFF, 15, REG, VAL)
+
+#define GEN_WRITE_WB_REG_CASES(OFF, REG, VAL)	\
+	WRITE_WB_REG_CASE(OFF,  0, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  1, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  2, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  3, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  4, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  5, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  6, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  7, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  8, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF,  9, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 10, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 11, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 12, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 13, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 14, REG, VAL);	\
+	WRITE_WB_REG_CASE(OFF, 15, REG, VAL)
+
+static u64 read_wb_reg(int reg, int n)
+{
+	u64 val = 0;
+
+	switch (reg + n) {
+	GEN_READ_WB_REG_CASES(AARCH64_DBG_REG_BVR, AARCH64_DBG_REG_NAME_BVR, val);
+	GEN_READ_WB_REG_CASES(AARCH64_DBG_REG_BCR, AARCH64_DBG_REG_NAME_BCR, val);
+	GEN_READ_WB_REG_CASES(AARCH64_DBG_REG_WVR, AARCH64_DBG_REG_NAME_WVR, val);
+	GEN_READ_WB_REG_CASES(AARCH64_DBG_REG_WCR, AARCH64_DBG_REG_NAME_WCR, val);
+	default:
+		pr_warning("attempt to read from unknown breakpoint register %d\n", n);
+	}
+
+	return val;
+}
+
+static void write_wb_reg(int reg, int n, u64 val)
+{
+	switch (reg + n) {
+	GEN_WRITE_WB_REG_CASES(AARCH64_DBG_REG_BVR, AARCH64_DBG_REG_NAME_BVR, val);
+	GEN_WRITE_WB_REG_CASES(AARCH64_DBG_REG_BCR, AARCH64_DBG_REG_NAME_BCR, val);
+	GEN_WRITE_WB_REG_CASES(AARCH64_DBG_REG_WVR, AARCH64_DBG_REG_NAME_WVR, val);
+	GEN_WRITE_WB_REG_CASES(AARCH64_DBG_REG_WCR, AARCH64_DBG_REG_NAME_WCR, val);
+	default:
+		pr_warning("attempt to write to unknown breakpoint register %d\n", n);
+	}
+	isb();
+}
+
+/*
+ * Convert a breakpoint privilege level to the corresponding exception
+ * level.
+ */
+static enum debug_el debug_exception_level(int privilege)
+{
+	switch (privilege) {
+	case AARCH64_BREAKPOINT_EL0:
+		return DBG_ACTIVE_EL0;
+	case AARCH64_BREAKPOINT_EL1:
+		return DBG_ACTIVE_EL1;
+	default:
+		pr_warning("invalid breakpoint privilege level %d\n", privilege);
+		return -EINVAL;
+	}
+}
+
+/*
+ * Install a perf counter breakpoint.
+ */
+int arch_install_hw_breakpoint(struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+	struct perf_event **slot, **slots;
+	struct debug_info *debug_info = &current->thread.debug;
+	int i, max_slots, ctrl_reg, val_reg, reg_enable;
+	u32 ctrl;
+
+	if (info->ctrl.type == ARM_BREAKPOINT_EXECUTE) {
+		/* Breakpoint */
+		ctrl_reg = AARCH64_DBG_REG_BCR;
+		val_reg = AARCH64_DBG_REG_BVR;
+		slots = __get_cpu_var(bp_on_reg);
+		max_slots = core_num_brps;
+		reg_enable = !debug_info->bps_disabled;
+	} else {
+		/* Watchpoint */
+		ctrl_reg = AARCH64_DBG_REG_WCR;
+		val_reg = AARCH64_DBG_REG_WVR;
+		slots = __get_cpu_var(wp_on_reg);
+		max_slots = core_num_wrps;
+		reg_enable = !debug_info->wps_disabled;
+	}
+
+	for (i = 0; i < max_slots; ++i) {
+		slot = &slots[i];
+
+		if (!*slot) {
+			*slot = bp;
+			break;
+		}
+	}
+
+	if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot"))
+		return -ENOSPC;
+
+	/* Ensure debug monitors are enabled at the correct exception level.  */
+	enable_debug_monitors(debug_exception_level(info->ctrl.privilege));
+
+	/* Setup the address register. */
+	write_wb_reg(val_reg, i, info->address);
+
+	/* Setup the control register. */
+	ctrl = encode_ctrl_reg(info->ctrl);
+	write_wb_reg(ctrl_reg, i, reg_enable ? ctrl | 0x1 : ctrl & ~0x1);
+
+	return 0;
+}
+
+void arch_uninstall_hw_breakpoint(struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+	struct perf_event **slot, **slots;
+	int i, max_slots, base;
+
+	if (info->ctrl.type == ARM_BREAKPOINT_EXECUTE) {
+		/* Breakpoint */
+		base = AARCH64_DBG_REG_BCR;
+		slots = __get_cpu_var(bp_on_reg);
+		max_slots = core_num_brps;
+	} else {
+		/* Watchpoint */
+		base = AARCH64_DBG_REG_WCR;
+		slots = __get_cpu_var(wp_on_reg);
+		max_slots = core_num_wrps;
+	}
+
+	/* Remove the breakpoint. */
+	for (i = 0; i < max_slots; ++i) {
+		slot = &slots[i];
+
+		if (*slot == bp) {
+			*slot = NULL;
+			break;
+		}
+	}
+
+	if (WARN_ONCE(i == max_slots, "Can't find any breakpoint slot"))
+		return;
+
+	/* Reset the control register. */
+	write_wb_reg(base, i, 0);
+
+	/* Release the debug monitors for the correct exception level.  */
+	disable_debug_monitors(debug_exception_level(info->ctrl.privilege));
+}
+
+static int get_hbp_len(u8 hbp_len)
+{
+	unsigned int len_in_bytes = 0;
+
+	switch (hbp_len) {
+	case ARM_BREAKPOINT_LEN_1:
+		len_in_bytes = 1;
+		break;
+	case ARM_BREAKPOINT_LEN_2:
+		len_in_bytes = 2;
+		break;
+	case ARM_BREAKPOINT_LEN_4:
+		len_in_bytes = 4;
+		break;
+	case ARM_BREAKPOINT_LEN_8:
+		len_in_bytes = 8;
+		break;
+	}
+
+	return len_in_bytes;
+}
+
+/*
+ * Check whether bp virtual address is in kernel space.
+ */
+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 = get_hbp_len(info->ctrl.len);
+
+	return (va >= TASK_SIZE) && ((va + len - 1) >= TASK_SIZE);
+}
+
+/*
+ * Extract generic type and length encodings from an arch_hw_breakpoint_ctrl.
+ * Hopefully this will disappear when ptrace can bypass the conversion
+ * to generic breakpoint descriptions.
+ */
+int arch_bp_generic_fields(struct arch_hw_breakpoint_ctrl ctrl,
+			   int *gen_len, int *gen_type)
+{
+	/* Type */
+	switch (ctrl.type) {
+	case ARM_BREAKPOINT_EXECUTE:
+		*gen_type = HW_BREAKPOINT_X;
+		break;
+	case ARM_BREAKPOINT_LOAD:
+		*gen_type = HW_BREAKPOINT_R;
+		break;
+	case ARM_BREAKPOINT_STORE:
+		*gen_type = HW_BREAKPOINT_W;
+		break;
+	case ARM_BREAKPOINT_LOAD | ARM_BREAKPOINT_STORE:
+		*gen_type = HW_BREAKPOINT_RW;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Len */
+	switch (ctrl.len) {
+	case ARM_BREAKPOINT_LEN_1:
+		*gen_len = HW_BREAKPOINT_LEN_1;
+		break;
+	case ARM_BREAKPOINT_LEN_2:
+		*gen_len = HW_BREAKPOINT_LEN_2;
+		break;
+	case ARM_BREAKPOINT_LEN_4:
+		*gen_len = HW_BREAKPOINT_LEN_4;
+		break;
+	case ARM_BREAKPOINT_LEN_8:
+		*gen_len = HW_BREAKPOINT_LEN_8;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/*
+ * 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->ctrl.type = ARM_BREAKPOINT_EXECUTE;
+		break;
+	case HW_BREAKPOINT_R:
+		info->ctrl.type = ARM_BREAKPOINT_LOAD;
+		break;
+	case HW_BREAKPOINT_W:
+		info->ctrl.type = ARM_BREAKPOINT_STORE;
+		break;
+	case HW_BREAKPOINT_RW:
+		info->ctrl.type = ARM_BREAKPOINT_LOAD | ARM_BREAKPOINT_STORE;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* Len */
+	switch (bp->attr.bp_len) {
+	case HW_BREAKPOINT_LEN_1:
+		info->ctrl.len = ARM_BREAKPOINT_LEN_1;
+		break;
+	case HW_BREAKPOINT_LEN_2:
+		info->ctrl.len = ARM_BREAKPOINT_LEN_2;
+		break;
+	case HW_BREAKPOINT_LEN_4:
+		info->ctrl.len = ARM_BREAKPOINT_LEN_4;
+		break;
+	case HW_BREAKPOINT_LEN_8:
+		info->ctrl.len = ARM_BREAKPOINT_LEN_8;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/*
+	 * On AArch64, we only permit breakpoints of length 4, whereas
+	 * AArch32 also requires breakpoints of length 2 for Thumb.
+	 * Watchpoints can be of length 1, 2, 4 or 8 bytes.
+	 */
+	if (info->ctrl.type == ARM_BREAKPOINT_EXECUTE) {
+		if (is_compat_task()) {
+			if (info->ctrl.len != ARM_BREAKPOINT_LEN_2 &&
+			    info->ctrl.len != ARM_BREAKPOINT_LEN_4)
+				return -EINVAL;
+		} else if (info->ctrl.len != ARM_BREAKPOINT_LEN_4) {
+			/*
+			 * FIXME: Some tools (I'm looking at you perf) assume
+			 *	  that breakpoints should be sizeof(long). This
+			 *	  is nonsense. For now, we fix up the parameter
+			 *	  but we should probably return -EINVAL instead.
+			 */
+			info->ctrl.len = ARM_BREAKPOINT_LEN_4;
+		}
+	}
+
+	/* Address */
+	info->address = bp->attr.bp_addr;
+
+	/*
+	 * Privilege
+	 * Note that we disallow combined EL0/EL1 breakpoints because
+	 * that would complicate the stepping code.
+	 */
+	if (arch_check_bp_in_kernelspace(bp))
+		info->ctrl.privilege = AARCH64_BREAKPOINT_EL1;
+	else
+		info->ctrl.privilege = AARCH64_BREAKPOINT_EL0;
+
+	/* Enabled? */
+	info->ctrl.enabled = !bp->attr.disabled;
+
+	return 0;
+}
+
+/*
+ * Validate the arch-specific HW Breakpoint register settings.
+ */
+int arch_validate_hwbkpt_settings(struct perf_event *bp)
+{
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+	int ret;
+	u64 alignment_mask, offset;
+
+	/* Build the arch_hw_breakpoint. */
+	ret = arch_build_bp_info(bp);
+	if (ret)
+		return ret;
+
+	/*
+	 * Check address alignment.
+	 * We don't do any clever alignment correction for watchpoints
+	 * because using 64-bit unaligned addresses is deprecated for
+	 * AArch64.
+	 *
+	 * AArch32 tasks expect some simple alignment fixups, so emulate
+	 * that here.
+	 */
+	if (is_compat_task()) {
+		if (info->ctrl.len == ARM_BREAKPOINT_LEN_8)
+			alignment_mask = 0x7;
+		else
+			alignment_mask = 0x3;
+		offset = info->address & alignment_mask;
+		switch (offset) {
+		case 0:
+			/* Aligned */
+			break;
+		case 1:
+			/* Allow single byte watchpoint. */
+			if (info->ctrl.len == ARM_BREAKPOINT_LEN_1)
+				break;
+		case 2:
+			/* Allow halfword watchpoints and breakpoints. */
+			if (info->ctrl.len == ARM_BREAKPOINT_LEN_2)
+				break;
+		default:
+			return -EINVAL;
+		}
+
+		info->address &= ~alignment_mask;
+		info->ctrl.len <<= offset;
+	} else {
+		if (info->ctrl.type == ARM_BREAKPOINT_EXECUTE)
+			alignment_mask = 0x3;
+		else
+			alignment_mask = 0x7;
+		if (info->address & alignment_mask)
+			return -EINVAL;
+	}
+
+	/*
+	 * Disallow per-task kernel breakpoints since these would
+	 * complicate the stepping code.
+	 */
+	if (info->ctrl.privilege == AARCH64_BREAKPOINT_EL1 && bp->hw.bp_target)
+		return -EINVAL;
+
+	return 0;
+}
+
+/*
+ * Enable/disable all of the breakpoints active at the specified
+ * exception level at the register level.
+ * This is used when single-stepping after a breakpoint exception.
+ */
+static void toggle_bp_registers(int reg, enum debug_el el, int enable)
+{
+	int i, max_slots, privilege;
+	u32 ctrl;
+	struct perf_event **slots;
+
+	switch (reg) {
+	case AARCH64_DBG_REG_BCR:
+		slots = __get_cpu_var(bp_on_reg);
+		max_slots = core_num_brps;
+		break;
+	case AARCH64_DBG_REG_WCR:
+		slots = __get_cpu_var(wp_on_reg);
+		max_slots = core_num_wrps;
+		break;
+	default:
+		return;
+	}
+
+	for (i = 0; i < max_slots; ++i) {
+		if (!slots[i])
+			continue;
+
+		privilege = counter_arch_bp(slots[i])->ctrl.privilege;
+		if (debug_exception_level(privilege) != el)
+			continue;
+
+		ctrl = read_wb_reg(reg, i);
+		if (enable)
+			ctrl |= 0x1;
+		else
+			ctrl &= ~0x1;
+		write_wb_reg(reg, i, ctrl);
+	}
+}
+
+/*
+ * Debug exception handlers.
+ */
+static int breakpoint_handler(unsigned long unused, unsigned int esr,
+			      struct pt_regs *regs)
+{
+	int i, step = 0, *kernel_step;
+	u32 ctrl_reg;
+	u64 addr, val;
+	struct perf_event *bp, **slots;
+	struct debug_info *debug_info;
+	struct arch_hw_breakpoint_ctrl ctrl;
+
+	slots = (struct perf_event **)__get_cpu_var(bp_on_reg);
+	addr = instruction_pointer(regs);
+	debug_info = &current->thread.debug;
+
+	for (i = 0; i < core_num_brps; ++i) {
+		rcu_read_lock();
+
+		bp = slots[i];
+
+		if (bp == NULL)
+			goto unlock;
+
+		/* Check if the breakpoint value matches. */
+		val = read_wb_reg(AARCH64_DBG_REG_BVR, i);
+		if (val != (addr & ~0x3))
+			goto unlock;
+
+		/* Possible match, check the byte address select to confirm. */
+		ctrl_reg = read_wb_reg(AARCH64_DBG_REG_BCR, i);
+		decode_ctrl_reg(ctrl_reg, &ctrl);
+		if (!((1 << (addr & 0x3)) & ctrl.len))
+			goto unlock;
+
+		counter_arch_bp(bp)->trigger = addr;
+		perf_bp_event(bp, regs);
+
+		/* Do we need to handle the stepping? */
+		if (!bp->overflow_handler)
+			step = 1;
+unlock:
+		rcu_read_unlock();
+	}
+
+	if (!step)
+		return 0;
+
+	if (user_mode(regs)) {
+		debug_info->bps_disabled = 1;
+		toggle_bp_registers(AARCH64_DBG_REG_BCR, DBG_ACTIVE_EL0, 0);
+
+		/* If we're already stepping a watchpoint, just return. */
+		if (debug_info->wps_disabled)
+			return 0;
+
+		if (test_thread_flag(TIF_SINGLESTEP))
+			debug_info->suspended_step = 1;
+		else
+			user_enable_single_step(current);
+	} else {
+		toggle_bp_registers(AARCH64_DBG_REG_BCR, DBG_ACTIVE_EL1, 0);
+		kernel_step = &__get_cpu_var(stepping_kernel_bp);
+
+		if (*kernel_step != ARM_KERNEL_STEP_NONE)
+			return 0;
+
+		if (kernel_active_single_step()) {
+			*kernel_step = ARM_KERNEL_STEP_SUSPEND;
+		} else {
+			*kernel_step = ARM_KERNEL_STEP_ACTIVE;
+			kernel_enable_single_step(regs);
+		}
+	}
+
+	return 0;
+}
+
+static int watchpoint_handler(unsigned long addr, unsigned int esr,
+			      struct pt_regs *regs)
+{
+	int i, step = 0, *kernel_step, access;
+	u32 ctrl_reg;
+	u64 val, alignment_mask;
+	struct perf_event *wp, **slots;
+	struct debug_info *debug_info;
+	struct arch_hw_breakpoint *info;
+	struct arch_hw_breakpoint_ctrl ctrl;
+
+	slots = (struct perf_event **)__get_cpu_var(wp_on_reg);
+	debug_info = &current->thread.debug;
+
+	for (i = 0; i < core_num_wrps; ++i) {
+		rcu_read_lock();
+
+		wp = slots[i];
+
+		if (wp == NULL)
+			goto unlock;
+
+		info = counter_arch_bp(wp);
+		/* AArch32 watchpoints are either 4 or 8 bytes aligned. */
+		if (is_compat_task()) {
+			if (info->ctrl.len == ARM_BREAKPOINT_LEN_8)
+				alignment_mask = 0x7;
+			else
+				alignment_mask = 0x3;
+		} else {
+			alignment_mask = 0x7;
+		}
+
+		/* Check if the watchpoint value matches. */
+		val = read_wb_reg(AARCH64_DBG_REG_WVR, i);
+		if (val != (addr & ~alignment_mask))
+			goto unlock;
+
+		/* Possible match, check the byte address select to confirm. */
+		ctrl_reg = read_wb_reg(AARCH64_DBG_REG_WCR, i);
+		decode_ctrl_reg(ctrl_reg, &ctrl);
+		if (!((1 << (addr & alignment_mask)) & ctrl.len))
+			goto unlock;
+
+		/*
+		 * Check that the access type matches.
+		 * 0 => load, otherwise => store
+		 */
+		access = (esr & AARCH64_ESR_ACCESS_MASK) ? HW_BREAKPOINT_W :
+			 HW_BREAKPOINT_R;
+		if (!(access & hw_breakpoint_type(wp)))
+			goto unlock;
+
+		info->trigger = addr;
+		perf_bp_event(wp, regs);
+
+		/* Do we need to handle the stepping? */
+		if (!wp->overflow_handler)
+			step = 1;
+
+unlock:
+		rcu_read_unlock();
+	}
+
+	if (!step)
+		return 0;
+
+	/*
+	 * We always disable EL0 watchpoints because the kernel can
+	 * cause these to fire via an unprivileged access.
+	 */
+	toggle_bp_registers(AARCH64_DBG_REG_WCR, DBG_ACTIVE_EL0, 0);
+
+	if (user_mode(regs)) {
+		debug_info->wps_disabled = 1;
+
+		/* If we're already stepping a breakpoint, just return. */
+		if (debug_info->bps_disabled)
+			return 0;
+
+		if (test_thread_flag(TIF_SINGLESTEP))
+			debug_info->suspended_step = 1;
+		else
+			user_enable_single_step(current);
+	} else {
+		toggle_bp_registers(AARCH64_DBG_REG_WCR, DBG_ACTIVE_EL1, 0);
+		kernel_step = &__get_cpu_var(stepping_kernel_bp);
+
+		if (*kernel_step != ARM_KERNEL_STEP_NONE)
+			return 0;
+
+		if (kernel_active_single_step()) {
+			*kernel_step = ARM_KERNEL_STEP_SUSPEND;
+		} else {
+			*kernel_step = ARM_KERNEL_STEP_ACTIVE;
+			kernel_enable_single_step(regs);
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * Handle single-step exception.
+ */
+int reinstall_suspended_bps(struct pt_regs *regs)
+{
+	struct debug_info *debug_info = &current->thread.debug;
+	int handled_exception = 0, *kernel_step;
+
+	kernel_step = &__get_cpu_var(stepping_kernel_bp);
+
+	/*
+	 * Called from single-step exception handler.
+	 * Return 0 if execution can resume, 1 if a SIGTRAP should be
+	 * reported.
+	 */
+	if (user_mode(regs)) {
+		if (debug_info->bps_disabled) {
+			debug_info->bps_disabled = 0;
+			toggle_bp_registers(AARCH64_DBG_REG_BCR, DBG_ACTIVE_EL0, 1);
+			handled_exception = 1;
+		}
+
+		if (debug_info->wps_disabled) {
+			debug_info->wps_disabled = 0;
+			toggle_bp_registers(AARCH64_DBG_REG_WCR, DBG_ACTIVE_EL0, 1);
+			handled_exception = 1;
+		}
+
+		if (handled_exception) {
+			if (debug_info->suspended_step) {
+				debug_info->suspended_step = 0;
+				/* Allow exception handling to fall-through. */
+				handled_exception = 0;
+			} else {
+				user_disable_single_step(current);
+			}
+		}
+	} else if (*kernel_step != ARM_KERNEL_STEP_NONE) {
+		toggle_bp_registers(AARCH64_DBG_REG_BCR, DBG_ACTIVE_EL1, 1);
+		toggle_bp_registers(AARCH64_DBG_REG_WCR, DBG_ACTIVE_EL1, 1);
+
+		if (!debug_info->wps_disabled)
+			toggle_bp_registers(AARCH64_DBG_REG_WCR, DBG_ACTIVE_EL0, 1);
+
+		if (*kernel_step != ARM_KERNEL_STEP_SUSPEND) {
+			kernel_disable_single_step();
+			handled_exception = 1;
+		} else {
+			handled_exception = 0;
+		}
+
+		*kernel_step = ARM_KERNEL_STEP_NONE;
+	}
+
+	return !handled_exception;
+}
+
+/*
+ * Context-switcher for restoring suspended breakpoints.
+ */
+void hw_breakpoint_thread_switch(struct task_struct *next)
+{
+	/*
+	 *           current        next
+	 * disabled: 0              0     => The usual case, NOTIFY_DONE
+	 *           0              1     => Disable the registers
+	 *           1              0     => Enable the registers
+	 *           1              1     => NOTIFY_DONE. per-task bps will
+	 *                                   get taken care of by perf.
+	 */
+
+	struct debug_info *current_debug_info, *next_debug_info;
+
+	current_debug_info = &current->thread.debug;
+	next_debug_info = &next->thread.debug;
+
+	/* Update breakpoints. */
+	if (current_debug_info->bps_disabled != next_debug_info->bps_disabled)
+		toggle_bp_registers(AARCH64_DBG_REG_BCR,
+				    DBG_ACTIVE_EL0,
+				    !next_debug_info->bps_disabled);
+
+	/* Update watchpoints. */
+	if (current_debug_info->wps_disabled != next_debug_info->wps_disabled)
+		toggle_bp_registers(AARCH64_DBG_REG_WCR,
+				    DBG_ACTIVE_EL0,
+				    !next_debug_info->wps_disabled);
+}
+
+/*
+ * CPU initialisation.
+ */
+static void reset_ctrl_regs(void *unused)
+{
+	int i;
+
+	for (i = 0; i < core_num_brps; ++i) {
+		write_wb_reg(AARCH64_DBG_REG_BCR, i, 0UL);
+		write_wb_reg(AARCH64_DBG_REG_BVR, i, 0UL);
+	}
+
+	for (i = 0; i < core_num_wrps; ++i) {
+		write_wb_reg(AARCH64_DBG_REG_WCR, i, 0UL);
+		write_wb_reg(AARCH64_DBG_REG_WVR, i, 0UL);
+	}
+}
+
+static int __cpuinit hw_breakpoint_reset_notify(struct notifier_block *self,
+						unsigned long action,
+						void *hcpu)
+{
+	int cpu = (long)hcpu;
+	if (action == CPU_ONLINE)
+		smp_call_function_single(cpu, reset_ctrl_regs, NULL, 1);
+	return NOTIFY_OK;
+}
+
+static struct notifier_block __cpuinitdata hw_breakpoint_reset_nb = {
+	.notifier_call = hw_breakpoint_reset_notify,
+};
+
+/*
+ * One-time initialisation.
+ */
+static int __init arch_hw_breakpoint_init(void)
+{
+	core_num_brps = get_num_brps();
+	core_num_wrps = get_num_wrps();
+
+	pr_info("found %d breakpoint and %d watchpoint registers.\n",
+		core_num_brps, core_num_wrps);
+
+	/*
+	 * Reset the breakpoint resources. We assume that a halting
+	 * debugger will leave the world in a nice state for us.
+	 */
+	smp_call_function(reset_ctrl_regs, NULL, 1);
+	reset_ctrl_regs(NULL);
+
+	/* Register debug fault handlers. */
+	hook_debug_fault_code(DBG_ESR_EVT_HWBP, breakpoint_handler, SIGTRAP,
+			      TRAP_HWBKPT, "hw-breakpoint handler");
+	hook_debug_fault_code(DBG_ESR_EVT_HWWP, watchpoint_handler, SIGTRAP,
+			      TRAP_HWBKPT, "hw-watchpoint handler");
+
+	/* Register hotplug notifier. */
+	register_cpu_notifier(&hw_breakpoint_reset_nb);
+
+	return 0;
+}
+arch_initcall(arch_hw_breakpoint_init);
+
+void hw_breakpoint_pmu_read(struct perf_event *bp)
+{
+}
+
+/*
+ * Dummy function to register with die_notifier.
+ */
+int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+				    unsigned long val, void *data)
+{
+	return NOTIFY_DONE;
+}
diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c
new file mode 100644
index 0000000..816b1b5
--- /dev/null
+++ b/arch/arm64/kernel/ptrace.c
@@ -0,0 +1,834 @@ 
+/*
+ * Based on arch/arm/kernel/ptrace.c
+ *
+ * By Ross Biro 1/23/92
+ * edited by Linus Torvalds
+ * ARM modifications Copyright (C) 2000 Russell King
+ * Copyright (C) 2012 ARM Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/ptrace.h>
+#include <linux/user.h>
+#include <linux/security.h>
+#include <linux/init.h>
+#include <linux/signal.h>
+#include <linux/uaccess.h>
+#include <linux/perf_event.h>
+#include <linux/hw_breakpoint.h>
+#include <linux/regset.h>
+#include <linux/tracehook.h>
+#include <linux/elf.h>
+
+#include <asm/compat.h>
+#include <asm/debug-monitors.h>
+#include <asm/pgtable.h>
+#include <asm/traps.h>
+#include <asm/system_misc.h>
+
+/*
+ * TODO: does not yet catch signals sent when the child dies.
+ * in exit.c or in signal.c.
+ */
+
+/*
+ * Called by kernel/ptrace.c when detaching..
+ */
+void ptrace_disable(struct task_struct *child)
+{
+}
+
+/*
+ * Handle hitting a breakpoint.
+ */
+static int ptrace_break(struct pt_regs *regs)
+{
+	siginfo_t info;
+
+	info.si_signo = SIGTRAP;
+	info.si_errno = 0;
+	info.si_code  = TRAP_BRKPT;
+	info.si_addr  = (void __user *)instruction_pointer(regs);
+
+	force_sig_info(SIGTRAP, &info, current);
+	return 0;
+}
+
+static int arm64_break_trap(unsigned long addr, unsigned int esr,
+			    struct pt_regs *regs)
+{
+	return ptrace_break(regs);
+}
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+/*
+ * Convert a virtual register number into an index for a thread_info
+ * breakpoint array. Breakpoints are identified using positive numbers
+ * whilst watchpoints are negative. The registers are laid out as pairs
+ * of (address, control), each pair mapping to a unique hw_breakpoint struct.
+ * Register 0 is reserved for describing resource information.
+ */
+static int ptrace_hbp_num_to_idx(long num)
+{
+	if (num < 0)
+		num = (ARM_MAX_BRP << 1) - num;
+	return (num - 1) >> 1;
+}
+
+/*
+ * Returns the virtual register number for the address of the
+ * breakpoint at index idx.
+ */
+static long ptrace_hbp_idx_to_num(int idx)
+{
+	long mid = ARM_MAX_BRP << 1;
+	long num = (idx << 1) + 1;
+	return num > mid ? mid - num : num;
+}
+
+/*
+ * Handle hitting a HW-breakpoint.
+ */
+static void ptrace_hbptriggered(struct perf_event *bp,
+				struct perf_sample_data *data,
+				struct pt_regs *regs)
+{
+	struct arch_hw_breakpoint *bkpt = counter_arch_bp(bp);
+	long num;
+	int i;
+	siginfo_t info;
+
+	for (i = 0; i < ARM_MAX_HBP_SLOTS; ++i)
+		if (current->thread.debug.hbp[i] == bp)
+			break;
+
+	num = (i == ARM_MAX_HBP_SLOTS) ? 0 : ptrace_hbp_idx_to_num(i);
+
+	info.si_signo	= SIGTRAP;
+	info.si_errno	= (int)num;
+	info.si_code	= TRAP_HWBKPT;
+	info.si_addr	= (void __user *)(bkpt->trigger);
+
+	force_sig_info(SIGTRAP, &info, current);
+}
+
+/*
+ * Unregister breakpoints from this task and reset the pointers in
+ * the thread_struct.
+ */
+void flush_ptrace_hw_breakpoint(struct task_struct *tsk)
+{
+	int i;
+	struct thread_struct *t = &tsk->thread;
+
+	for (i = 0; i < ARM_MAX_HBP_SLOTS; i++) {
+		if (t->debug.hbp[i]) {
+			unregister_hw_breakpoint(t->debug.hbp[i]);
+			t->debug.hbp[i] = NULL;
+		}
+	}
+}
+
+void ptrace_hw_copy_thread(struct task_struct *task)
+{
+	memset(&task->thread.debug, 0, sizeof(struct debug_info));
+}
+
+static u32 ptrace_get_hbp_resource_info(void)
+{
+	u8 num_brps, num_wrps, debug_arch, wp_len;
+	u32 reg = 0;
+
+	num_brps	= hw_breakpoint_slots(TYPE_INST);
+	num_wrps	= hw_breakpoint_slots(TYPE_DATA);
+
+	debug_arch	= debug_monitors_arch();
+	wp_len		= 8;		/* Reserved on AArch64 */
+	reg		|= debug_arch;
+	reg		<<= 8;
+	reg		|= wp_len;
+	reg		<<= 8;
+	reg		|= num_wrps;
+	reg		<<= 8;
+	reg		|= num_brps;
+
+	return reg;
+}
+
+static struct perf_event *ptrace_hbp_create(struct task_struct *tsk, int type)
+{
+	struct perf_event_attr attr;
+
+	ptrace_breakpoint_init(&attr);
+
+	/*
+	 * Initialise fields to sane defaults
+	 * (i.e. values that will pass validation).
+	 */
+	attr.bp_addr	= 0;
+	attr.bp_len	= HW_BREAKPOINT_LEN_4;
+	attr.bp_type	= type;
+	attr.disabled	= 1;
+
+	return register_user_hw_breakpoint(&attr, ptrace_hbptriggered, NULL,
+					   tsk);
+}
+
+static int ptrace_gethbpregs(struct task_struct *tsk, long num,
+			     unsigned long  __user *data)
+{
+	u64 addr_reg;
+	u32 ctrl_reg;
+	int idx, ret = 0;
+	struct perf_event *bp;
+	struct arch_hw_breakpoint_ctrl arch_ctrl;
+
+	if (num == 0) {
+		ctrl_reg = ptrace_get_hbp_resource_info();
+		if (put_user(ctrl_reg, (u32 __user *)data))
+			ret = -EFAULT;
+	} else {
+		idx = ptrace_hbp_num_to_idx(num);
+		if (idx < 0 || idx >= ARM_MAX_HBP_SLOTS)
+			return -EINVAL;
+
+		bp = tsk->thread.debug.hbp[idx];
+		arch_ctrl = counter_arch_bp(bp)->ctrl;
+
+		if (is_compat_task()) {
+			/*
+			 * Fix up the len because we may have adjusted
+			 * it to compensate for an unaligned address.
+			 */
+			while (!(arch_ctrl.len & 0x1))
+				arch_ctrl.len >>= 1;
+		}
+
+		if (num & 0x1) {
+			addr_reg = bp ? bp->attr.bp_addr : 0;
+			if (put_user(addr_reg, data))
+				ret = -EFAULT;
+		} else {
+			ctrl_reg = bp ? encode_ctrl_reg(arch_ctrl) : 0;
+			if (put_user(ctrl_reg, (u32 __user *)data))
+				ret = -EFAULT;
+		}
+	}
+
+	return ret;
+}
+
+static int ptrace_sethbpregs(struct task_struct *tsk, long num,
+			     unsigned long __user *data)
+{
+	int idx, gen_len, gen_type, implied_type, ret;
+	u64 user_addr;
+	u32 user_ctrl;
+	struct perf_event *bp;
+	struct arch_hw_breakpoint_ctrl ctrl;
+	struct perf_event_attr attr;
+
+	if (num == 0)
+		return 0;
+	else if (num < 0)
+		implied_type = HW_BREAKPOINT_RW;
+	else
+		implied_type = HW_BREAKPOINT_X;
+
+	idx = ptrace_hbp_num_to_idx(num);
+	if (idx < 0 || idx >= ARM_MAX_HBP_SLOTS)
+		return -EFAULT;
+
+	bp = tsk->thread.debug.hbp[idx];
+	if (!bp) {
+		bp = ptrace_hbp_create(tsk, implied_type);
+		if (IS_ERR(bp))
+			return PTR_ERR(bp);
+		tsk->thread.debug.hbp[idx] = bp;
+	}
+
+	attr = bp->attr;
+
+	if (num & 0x1) {
+		/* Address */
+		if (get_user(user_addr, data))
+			return -EFAULT;
+		attr.bp_addr = user_addr;
+	} else {
+		/* Control */
+		if (get_user(user_ctrl, (u32 __user *)data))
+			return -EFAULT;
+		decode_ctrl_reg(user_ctrl, &ctrl);
+		ret = arch_bp_generic_fields(ctrl, &gen_len, &gen_type);
+		if (ret)
+			return ret;
+
+		if ((gen_type & implied_type) != gen_type)
+			return -EINVAL;
+
+		attr.bp_len	= gen_len;
+		attr.bp_type	= gen_type;
+		attr.disabled	= !ctrl.enabled;
+	}
+
+	return modify_user_hw_breakpoint(bp, &attr);
+}
+#endif	/* CONFIG_HAVE_HW_BREAKPOINT */
+
+static int gpr_get(struct task_struct *target,
+		   const struct user_regset *regset,
+		   unsigned int pos, unsigned int count,
+		   void *kbuf, void __user *ubuf)
+{
+	struct user_pt_regs *uregs = &task_pt_regs(target)->user_regs;
+	return user_regset_copyout(&pos, &count, &kbuf, &ubuf, uregs, 0, -1);
+}
+
+static int gpr_set(struct task_struct *target, const struct user_regset *regset,
+		   unsigned int pos, unsigned int count,
+		   const void *kbuf, const void __user *ubuf)
+{
+	int ret;
+	struct user_pt_regs newregs;
+
+	ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, &newregs, 0, -1);
+	if (ret)
+		return ret;
+
+	if (!valid_user_regs(&newregs))
+		return -EINVAL;
+
+	task_pt_regs(target)->user_regs = newregs;
+	return 0;
+}
+
+/*
+ * TODO: update fp accessors for lazy context switching (sync/flush hwstate)
+ */
+static int fpr_get(struct task_struct *target, const struct user_regset *regset,
+		   unsigned int pos, unsigned int count,
+		   void *kbuf, void __user *ubuf)
+{
+	struct user_fpsimd_state *uregs;
+	uregs = &target->thread.fpsimd_state.user_fpsimd;
+	return user_regset_copyout(&pos, &count, &kbuf, &ubuf, uregs, 0, -1);
+}
+
+static int fpr_set(struct task_struct *target, const struct user_regset *regset,
+		   unsigned int pos, unsigned int count,
+		   const void *kbuf, const void __user *ubuf)
+{
+	int ret;
+	struct user_fpsimd_state newstate;
+
+	ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, &newstate, 0, -1);
+	if (ret)
+		return ret;
+
+	target->thread.fpsimd_state.user_fpsimd = newstate;
+	return ret;
+}
+
+enum aarch64_regset {
+	REGSET_GPR,
+	REGSET_FPR,
+};
+
+static const struct user_regset aarch64_regsets[] = {
+	[REGSET_GPR] = {
+		.core_note_type = NT_PRSTATUS,
+		.n = sizeof(struct user_pt_regs) / sizeof(u64),
+		.size = sizeof(u64),
+		.align = sizeof(u64),
+		.get = gpr_get,
+		.set = gpr_set
+	},
+	[REGSET_FPR] = {
+		.core_note_type = NT_PRFPREG,
+		.n = sizeof(struct user_fpsimd_state) / sizeof(u32),
+		/*
+		 * We pretend we have 32-bit registers because the fpsr and
+		 * fpcr are 32-bits wide.
+		 */
+		.size = sizeof(u32),
+		.align = sizeof(u32),
+		.get = fpr_get,
+		.set = fpr_set
+	},
+};
+
+static const struct user_regset_view user_aarch64_view = {
+	.name = "aarch64", .e_machine = EM_AARCH64,
+	.regsets = aarch64_regsets, .n = ARRAY_SIZE(aarch64_regsets)
+};
+
+#ifdef CONFIG_AARCH32_EMULATION
+enum compat_regset {
+	REGSET_COMPAT_GPR,
+	REGSET_COMPAT_VFP,
+};
+
+static int compat_gpr_get(struct task_struct *target,
+			  const struct user_regset *regset,
+			  unsigned int pos, unsigned int count,
+			  void *kbuf, void __user *ubuf)
+{
+	int ret = 0;
+	unsigned int i, start, num_regs;
+
+	/* Calculate the number of AArch32 registers contained in count */
+	num_regs = count / regset->size;
+
+	/* Convert pos into an register number */
+	start = pos / regset->size;
+
+	if (start + num_regs > regset->n)
+		return -EIO;
+
+	for (i = 0; i < num_regs; ++i) {
+		unsigned int idx = start + i;
+		void *reg;
+
+		switch (idx) {
+		case 15:
+			reg = (void *)&task_pt_regs(target)->pc;
+			break;
+		case 16:
+			reg = (void *)&task_pt_regs(target)->pstate;
+			break;
+		case 17:
+			reg = (void *)&task_pt_regs(target)->orig_x0;
+			break;
+		default:
+			reg = (void *)&task_pt_regs(target)->regs[idx];
+		}
+
+		ret = copy_to_user(ubuf, reg, sizeof(compat_ulong_t));
+
+		if (ret)
+			break;
+		else
+			ubuf += sizeof(compat_ulong_t);
+	}
+
+	return ret;
+}
+
+static int compat_gpr_set(struct task_struct *target,
+			  const struct user_regset *regset,
+			  unsigned int pos, unsigned int count,
+			  const void *kbuf, const void __user *ubuf)
+{
+	struct pt_regs newregs;
+	int ret = 0;
+	unsigned int i, start, num_regs;
+
+	/* Calculate the number of AArch32 registers contained in count */
+	num_regs = count / regset->size;
+
+	/* Convert pos into an register number */
+	start = pos / regset->size;
+
+	if (start + num_regs > regset->n)
+		return -EIO;
+
+	newregs = *task_pt_regs(target);
+
+	for (i = 0; i < num_regs; ++i) {
+		unsigned int idx = start + i;
+		void *reg;
+
+		switch (idx) {
+		case 15:
+			reg = (void *)&newregs.pc;
+			break;
+		case 16:
+			reg = (void *)&newregs.pstate;
+			break;
+		case 17:
+			reg = (void *)&newregs.orig_x0;
+			break;
+		default:
+			reg = (void *)&newregs.regs[idx];
+		}
+
+		ret = copy_from_user(reg, ubuf, sizeof(compat_ulong_t));
+
+		if (ret)
+			goto out;
+		else
+			ubuf += sizeof(compat_ulong_t);
+	}
+
+	if (valid_user_regs(&newregs.user_regs))
+		*task_pt_regs(target) = newregs;
+	else
+		ret = -EINVAL;
+
+out:
+	return ret;
+}
+
+static int compat_vfp_get(struct task_struct *target,
+			  const struct user_regset *regset,
+			  unsigned int pos, unsigned int count,
+			  void *kbuf, void __user *ubuf)
+{
+	struct user_fpsimd_state *uregs;
+	compat_ulong_t fpscr;
+	int ret;
+
+	uregs = &target->thread.fpsimd_state.user_fpsimd;
+
+	/*
+	 * The VFP registers are packed into the fpsimd_state, so they all sit
+	 * nicely together for us. We just need to create the fpscr separately.
+	 */
+	ret = user_regset_copyout(&pos, &count, &kbuf, &ubuf, uregs, 0,
+				  VFP_STATE_SIZE - sizeof(compat_ulong_t));
+
+	if (count && !ret) {
+		fpscr = (uregs->fpsr & VFP_FPSCR_STAT_MASK) |
+			(uregs->fpcr & VFP_FPSCR_CTRL_MASK);
+		ret = put_user(fpscr, (compat_ulong_t *)ubuf);
+	}
+
+	return ret;
+}
+
+static int compat_vfp_set(struct task_struct *target,
+			  const struct user_regset *regset,
+			  unsigned int pos, unsigned int count,
+			  const void *kbuf, const void __user *ubuf)
+{
+	struct user_fpsimd_state *uregs;
+	compat_ulong_t fpscr;
+	int ret;
+
+	if (pos + count > VFP_STATE_SIZE)
+		return -EIO;
+
+	uregs = &target->thread.fpsimd_state.user_fpsimd;
+
+	ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, uregs, 0,
+				 VFP_STATE_SIZE - sizeof(compat_ulong_t));
+
+	if (count && !ret) {
+		ret = get_user(fpscr, (compat_ulong_t *)ubuf);
+		uregs->fpsr = fpscr & VFP_FPSCR_STAT_MASK;
+		uregs->fpcr = fpscr & VFP_FPSCR_CTRL_MASK;
+	}
+
+	return ret;
+}
+
+static const struct user_regset aarch32_regsets[] = {
+	[REGSET_COMPAT_GPR] = {
+		.core_note_type = NT_PRSTATUS,
+		.n = COMPAT_ELF_NGREG,
+		.size = sizeof(compat_elf_greg_t),
+		.align = sizeof(compat_elf_greg_t),
+		.get = compat_gpr_get,
+		.set = compat_gpr_set
+	},
+	[REGSET_COMPAT_VFP] = {
+		.core_note_type = NT_ARM_VFP,
+		.n = VFP_STATE_SIZE / sizeof(compat_ulong_t),
+		.size = sizeof(compat_ulong_t),
+		.align = sizeof(compat_ulong_t),
+		.get = compat_vfp_get,
+		.set = compat_vfp_set
+	},
+};
+
+static const struct user_regset_view user_aarch32_view = {
+	.name = "aarch32", .e_machine = EM_ARM,
+	.regsets = aarch32_regsets, .n = ARRAY_SIZE(aarch32_regsets)
+};
+#endif /* CONFIG_AARCH32_EMULATION */
+
+const struct user_regset_view *task_user_regset_view(struct task_struct *task)
+{
+#ifdef CONFIG_AARCH32_EMULATION
+	if (test_tsk_thread_flag(task, TIF_32BIT))
+		return &user_aarch32_view;
+#endif
+	return &user_aarch64_view;
+}
+
+long arch_ptrace(struct task_struct *child, long request,
+		 unsigned long addr, unsigned long data)
+{
+	int ret;
+	unsigned long *datap = (unsigned long __user *)data;
+
+	switch (request) {
+		case PTRACE_GET_THREAD_AREA:
+			ret = put_user(child->thread.tp_value, datap);
+			break;
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+		case PTRACE_GETHBPREGS:
+			ret = ptrace_gethbpregs(child, addr, datap);
+			break;
+
+		case PTRACE_SETHBPREGS:
+			ret = ptrace_sethbpregs(child, addr, datap);
+			break;
+#endif
+
+		default:
+			ret = ptrace_request(child, request, addr, data);
+			break;
+	}
+
+	return ret;
+}
+
+#ifdef CONFIG_AARCH32_EMULATION
+
+#include <linux/compat.h>
+
+int aarch32_break_trap(struct pt_regs *regs)
+{
+	unsigned int instr;
+	bool bp = false;
+	void __user *pc = (void __user *)instruction_pointer(regs);
+
+	if (compat_thumb_mode(regs)) {
+		/* get 16-bit Thumb instruction */
+		get_user(instr, (u16 __user *)pc);
+		if (instr == AARCH32_BREAK_THUMB2_LO) {
+			/* get second half of 32-bit Thumb-2 instruction */
+			get_user(instr, (u16 __user *)(pc + 2));
+			bp = instr == AARCH32_BREAK_THUMB2_HI;
+		} else {
+			bp = instr == AARCH32_BREAK_THUMB;
+		}
+	} else {
+		/* 32-bit ARM instruction */
+		get_user(instr, (u32 __user *)pc);
+		bp = (instr & ~0xf0000000) == AARCH32_BREAK_ARM;
+	}
+
+	if (bp)
+		return ptrace_break(regs);
+	return 1;
+}
+
+static int compat_ptrace_read_user(struct task_struct *tsk, compat_ulong_t off,
+				   compat_ulong_t __user *ret)
+{
+	compat_ulong_t tmp;
+
+	if (off & 3)
+		return -EIO;
+
+	if (off == PT_TEXT_ADDR)
+		tmp = tsk->mm->start_code;
+	else if (off == PT_DATA_ADDR)
+		tmp = tsk->mm->start_data;
+	else if (off == PT_TEXT_END_ADDR)
+		tmp = tsk->mm->end_code;
+	else if (off < sizeof(compat_elf_gregset_t))
+		return copy_regset_to_user(tsk, &user_aarch32_view,
+					   REGSET_COMPAT_GPR, off,
+					   sizeof(compat_ulong_t), ret);
+	else if (off >= COMPAT_USER_SZ)
+		return -EIO;
+	else
+		tmp = 0;
+
+	return put_user(tmp, ret);
+}
+
+static int compat_ptrace_write_user(struct task_struct *tsk, compat_ulong_t off,
+				    compat_ulong_t val)
+{
+	int ret;
+
+	if (off & 3 || off >= COMPAT_USER_SZ)
+		return -EIO;
+
+	if (off >= sizeof(compat_elf_gregset_t))
+		return 0;
+
+	ret = copy_regset_from_user(tsk, &user_aarch32_view,
+				    REGSET_COMPAT_GPR, off,
+				    sizeof(compat_ulong_t),
+				    &val);
+	return ret;
+}
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+static int compat_ptrace_gethbpregs(struct task_struct *tsk, compat_long_t num,
+				    compat_ulong_t __user *data)
+{
+	int ret;
+	unsigned long kdata;
+
+	mm_segment_t old_fs = get_fs();
+	set_fs(KERNEL_DS);
+	ret = ptrace_gethbpregs(tsk, (long)num, &kdata);
+	set_fs(old_fs);
+
+	if (!ret)
+		ret = put_user(kdata, data);
+
+	return ret;
+}
+
+static int compat_ptrace_sethbpregs(struct task_struct *tsk, compat_long_t num,
+				    compat_ulong_t __user *data)
+{
+	int ret;
+	unsigned long kdata = 0;
+	mm_segment_t old_fs = get_fs();
+
+	ret = get_user(kdata, data);
+
+	if (!ret) {
+		set_fs(KERNEL_DS);
+		ret = ptrace_sethbpregs(tsk, (long)num, &kdata);
+		set_fs(old_fs);
+	}
+
+	return ret;
+}
+#endif	/* CONFIG_HAVE_HW_BREAKPOINT */
+
+long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
+			compat_ulong_t caddr, compat_ulong_t cdata)
+{
+	unsigned long addr = caddr;
+	unsigned long data = cdata;
+	void __user *datap = compat_ptr(data);
+	int ret;
+
+	switch (request) {
+		case PTRACE_PEEKUSR:
+			ret = compat_ptrace_read_user(child, addr, datap);
+			break;
+
+		case PTRACE_POKEUSR:
+			ret = compat_ptrace_write_user(child, addr, data);
+			break;
+
+		case PTRACE_GETREGS:
+			ret = copy_regset_to_user(child,
+						  &user_aarch32_view,
+						  REGSET_COMPAT_GPR,
+						  0, sizeof(compat_elf_gregset_t),
+						  datap);
+			break;
+
+		case PTRACE_SETREGS:
+			ret = copy_regset_from_user(child,
+						    &user_aarch32_view,
+						    REGSET_COMPAT_GPR,
+						    0, sizeof(compat_elf_gregset_t),
+						    datap);
+			break;
+
+		case PTRACE_GET_THREAD_AREA:
+			ret = put_user((compat_ulong_t)child->thread.tp_value,
+				       (compat_ulong_t __user *)datap);
+			break;
+
+		case PTRACE_SET_SYSCALL:
+			task_pt_regs(child)->syscallno = data;
+			ret = 0;
+			break;
+
+		case COMPAT_PTRACE_GETVFPREGS:
+			ret = copy_regset_to_user(child,
+						  &user_aarch32_view,
+						  REGSET_COMPAT_VFP,
+						  0, VFP_STATE_SIZE,
+						  datap);
+			break;
+
+		case COMPAT_PTRACE_SETVFPREGS:
+			ret = copy_regset_from_user(child,
+						    &user_aarch32_view,
+						    REGSET_COMPAT_VFP,
+						    0, VFP_STATE_SIZE,
+						    datap);
+			break;
+
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+		case PTRACE_GETHBPREGS:
+			ret = compat_ptrace_gethbpregs(child, addr, datap);
+			break;
+
+		case PTRACE_SETHBPREGS:
+			ret = compat_ptrace_sethbpregs(child, addr, datap);
+			break;
+#endif
+
+		default:
+			ret = compat_ptrace_request(child, request, addr,
+						    data);
+			break;
+	}
+
+	return ret;
+}
+#endif /* CONFIG_AARCH32_EMULATION */
+
+static int __init ptrace_break_init(void)
+{
+	hook_debug_fault_code(DBG_ESR_EVT_BRK, arm64_break_trap, SIGTRAP,
+			      TRAP_BRKPT, "ptrace BRK handler");
+	return 0;
+}
+core_initcall(ptrace_break_init);
+
+
+asmlinkage int syscall_trace(int dir, struct pt_regs *regs)
+{
+	unsigned long saved_reg;
+
+	if (!test_thread_flag(TIF_SYSCALL_TRACE))
+		return regs->syscallno;
+
+	if (test_thread_flag(TIF_32BIT)) {
+		/* AArch32 uses ip (r12) for scratch */
+		saved_reg = regs->regs[12];
+		regs->regs[12] = dir;
+	} else {
+		/*
+		 * Save X7. X7 is used to denote syscall entry/exit:
+		 *   X7 = 0 -> entry, = 1 -> exit
+		 */
+		saved_reg = regs->regs[7];
+		regs->regs[7] = dir;
+	}
+
+	if (dir)
+		tracehook_report_syscall_exit(regs, 0);
+	else if (tracehook_report_syscall_entry(regs))
+		regs->syscallno = ~0UL;
+
+	if (test_thread_flag(TIF_32BIT))
+		regs->regs[12] = saved_reg;
+	else
+		regs->regs[7] = saved_reg;
+
+	return regs->syscallno;
+}