First cut of DTrace for AArch64.

Reviewed by:	andrew, emaste
Sponsored by:	ARM Limited
Differential Revision:	https://reviews.freebsd.org/D2738

8 files changed, 1039 insertions, 3 deletions

diff --git a/sys/cddl/dev/dtrace/aarch64/dtrace_asm.S b/sys/cddl/dev/dtrace/aarch64/dtrace_asm.S
new file mode 100644
index 0000000..9ab4499
--- /dev/null
+++ b/sys/cddl/dev/dtrace/aarch64/dtrace_asm.S
@@ -0,0 +1,173 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License").  You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * $FreeBSD$
+ */
+/*
+ * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#define _ASM
+#define _LOCORE
+
+#include <sys/cpuvar_defs.h>
+#include <sys/dtrace.h>
+
+#include <machine/armreg.h>
+#include <machine/asm.h>
+
+#include "assym.s"
+
+/*
+void dtrace_membar_producer(void)
+*/
+ENTRY(dtrace_membar_producer)
+	RET
+END(dtrace_membar_producer)
+
+/*
+void dtrace_membar_consumer(void)
+*/
+ENTRY(dtrace_membar_consumer)
+	RET
+END(dtrace_membar_consumer)
+
+/*
+dtrace_icookie_t dtrace_interrupt_disable(void)
+*/
+ENTRY(dtrace_interrupt_disable)
+	msr	daifset, #2
+	RET
+END(dtrace_interrupt_disable)
+
+/*
+void dtrace_interrupt_enable(dtrace_icookie_t cookie)
+*/
+ENTRY(dtrace_interrupt_enable)
+	msr	daifclr, #2
+	RET
+END(dtrace_interrupt_enable)
+/*
+uint8_t
+dtrace_fuword8_nocheck(void *addr)
+*/
+ENTRY(dtrace_fuword8_nocheck)
+	ldrb	w0, [x0]
+	RET
+END(dtrace_fuword8_nocheck)
+
+/*
+uint16_t
+dtrace_fuword16_nocheck(void *addr)
+*/
+ENTRY(dtrace_fuword16_nocheck)
+	ldrh	w0, [x0]
+	RET
+END(dtrace_fuword16_nocheck)
+
+/*
+uint32_t
+dtrace_fuword32_nocheck(void *addr)
+*/
+ENTRY(dtrace_fuword32_nocheck)
+	ldr	w0, [x0]
+	RET
+END(dtrace_fuword32_nocheck)
+
+/*
+uint64_t
+dtrace_fuword64_nocheck(void *addr)
+*/
+ENTRY(dtrace_fuword64_nocheck)
+	ldr	x0, [x0]
+	RET
+END(dtrace_fuword64_nocheck)
+
+/*
+void
+dtrace_copy(uintptr_t uaddr, uintptr_t kaddr, size_t size)
+*/
+ENTRY(dtrace_copy)
+	cbz	x2, 2f		/* If len == 0 then skip loop */
+1:
+	ldrb	w4, [x0], #1	/* Load from uaddr */
+	strb	w4, [x1], #1	/* Store in kaddr */
+	sub	x2, x2, #1	/* len-- */
+	cbnz	x2, 1b
+2:
+	RET
+END(dtrace_copy)
+
+/*
+void
+dtrace_copystr(uintptr_t uaddr, uintptr_t kaddr, size_t size,
+    volatile uint16_t *flags)
+XXX: Check for flags?
+*/
+ENTRY(dtrace_copystr)
+	cbz     x2, 2f          /* If len == 0 then skip loop */
+ 
+1:	ldrb    w4, [x0], #1    /* Load from uaddr */
+	strb    w4, [x1], #1    /* Store in kaddr */
+	cbz     w4, 2f          /* If == 0 then break */
+	sub     x2, x2, #1      /* len-- */
+	cbnz    x2, 1b
+2:
+	RET
+END(dtrace_copystr)
+
+/*
+uintptr_t
+dtrace_caller(int aframes)
+*/
+ENTRY(dtrace_caller)
+	mov	x0, #-1
+	RET
+END(dtrace_caller)
+
+/*
+uint32_t
+dtrace_cas32(uint32_t *target, uint32_t cmp, uint32_t new)
+*/
+ENTRY(dtrace_cas32)
+1:	ldxr	w3, [x0]	/* Load target */
+	cmp	w3, w1		/* Check if *target == cmp */
+	bne	2f		/* No, return */
+	stxr	w12, w2, [x0]	/* Store new to target */
+	cbnz	w12, 1b		/* Try again if store not succeed */
+2:	mov	w0, w3		/* Return the value loaded from target */
+	RET
+END(dtrace_cas32)
+
+/*
+void *
+dtrace_casptr(volatile void *target, volatile void *cmp, volatile void *new)
+*/
+ENTRY(dtrace_casptr)
+1:	ldxr	x3, [x0]	/* Load target */
+	cmp	x3, x1		/* Check if *target == cmp */
+	bne	2f		/* No, return */
+	stxr	w12, x2, [x0]	/* Store new to target */
+	cbnz	w12, 1b		/* Try again if store not succeed */
+2:	mov	x0, x3		/* Return the value loaded from target */
+	RET
+END(dtrace_casptr)
diff --git a/sys/cddl/dev/dtrace/aarch64/dtrace_isa.c b/sys/cddl/dev/dtrace/aarch64/dtrace_isa.c
new file mode 100644
index 0000000..dcda440
--- /dev/null
+++ b/sys/cddl/dev/dtrace/aarch64/dtrace_isa.c
@@ -0,0 +1,287 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License").  You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * $FreeBSD$
+ */
+/*
+ * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+#include <sys/cdefs.h>
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/kernel.h>
+#include <sys/stack.h>
+#include <sys/pcpu.h>
+
+#include <machine/frame.h>
+#include <machine/md_var.h>
+#include <machine/reg.h>
+
+#include <vm/vm.h>
+#include <vm/vm_param.h>
+#include <vm/pmap.h>
+
+#include <machine/atomic.h>
+#include <machine/db_machdep.h>
+#include <machine/md_var.h>
+#include <machine/vmparam.h>
+#include <machine/stack.h>
+#include <ddb/db_sym.h>
+#include <ddb/ddb.h>
+#include <sys/kdb.h>
+
+#include "regset.h"
+
+/*
+ * Wee need some reasonable default to prevent backtrace code
+ * from wandering too far
+ */
+#define	MAX_FUNCTION_SIZE 0x10000
+#define	MAX_PROLOGUE_SIZE 0x100
+
+uint8_t dtrace_fuword8_nocheck(void *);
+uint16_t dtrace_fuword16_nocheck(void *);
+uint32_t dtrace_fuword32_nocheck(void *);
+uint64_t dtrace_fuword64_nocheck(void *);
+
+void
+dtrace_getpcstack(pc_t *pcstack, int pcstack_limit, int aframes,
+    uint32_t *intrpc)
+{
+	struct unwind_state state;
+	int scp_offset;
+	register_t sp;
+	int depth;
+
+	depth = 0;
+
+	if (intrpc != 0) {
+		pcstack[depth++] = (pc_t) intrpc;
+	}
+
+	aframes++;
+
+	__asm __volatile("mov %0, sp" : "=&r" (sp));
+
+	state.fp = (uint64_t)__builtin_frame_address(0);
+	state.sp = sp;
+	state.pc = (uint64_t)dtrace_getpcstack;
+
+	while (depth < pcstack_limit) {
+		if (unwind_frame(&state))
+			break;
+
+		if (!INKERNEL(state.pc) || !INKERNEL(state.fp))
+			break;
+
+		/*
+		 * NB: Unlike some other architectures, we don't need to
+		 * explicitly insert cpu_dtrace_caller as it appears in the
+		 * normal kernel stack trace rather than a special trap frame.
+		 */
+		if (aframes > 0) {
+			aframes--;
+		} else {
+			pcstack[depth++] = state.pc;
+		}
+
+	}
+
+	for (; depth < pcstack_limit; depth++) {
+		pcstack[depth] = 0;
+	}
+}
+
+void
+dtrace_getupcstack(uint64_t *pcstack, int pcstack_limit)
+{
+
+	printf("IMPLEMENT ME: %s\n", __func__);
+}
+
+int
+dtrace_getustackdepth(void)
+{
+
+	printf("IMPLEMENT ME: %s\n", __func__);
+
+	return (0);
+}
+
+void
+dtrace_getufpstack(uint64_t *pcstack, uint64_t *fpstack, int pcstack_limit)
+{
+
+	printf("IMPLEMENT ME: %s\n", __func__);
+}
+
+/*ARGSUSED*/
+uint64_t
+dtrace_getarg(int arg, int aframes)
+{
+
+	printf("IMPLEMENT ME: %s\n", __func__);
+
+	return (0);
+}
+
+int
+dtrace_getstackdepth(int aframes)
+{
+	struct unwind_state state;
+	int scp_offset;
+	register_t sp;
+	int depth;
+	int done;
+
+	depth = 1;
+	done = 0;
+
+	__asm __volatile("mov %0, sp" : "=&r" (sp));
+
+	state.fp = (uint64_t)__builtin_frame_address(0);
+	state.sp = sp;
+	state.pc = (uint64_t)dtrace_getstackdepth;
+
+	do {
+		done = unwind_frame(&state);
+		if (!INKERNEL(state.pc) || !INKERNEL(state.fp))
+			break;
+		depth++;
+	} while (!done);
+
+	if (depth < aframes)
+		return (0);
+	else
+		return (depth - aframes);
+}
+
+ulong_t
+dtrace_getreg(struct trapframe *rp, uint_t reg)
+{
+
+	printf("IMPLEMENT ME: %s\n", __func__);
+
+	return (0);
+}
+
+static int
+dtrace_copycheck(uintptr_t uaddr, uintptr_t kaddr, size_t size)
+{
+
+	if (uaddr + size > VM_MAXUSER_ADDRESS || uaddr + size < uaddr) {
+		DTRACE_CPUFLAG_SET(CPU_DTRACE_BADADDR);
+		cpu_core[curcpu].cpuc_dtrace_illval = uaddr;
+		return (0);
+	}
+
+	return (1);
+}
+
+void
+dtrace_copyin(uintptr_t uaddr, uintptr_t kaddr, size_t size,
+    volatile uint16_t *flags)
+{
+
+	if (dtrace_copycheck(uaddr, kaddr, size))
+		dtrace_copy(uaddr, kaddr, size);
+}
+
+void
+dtrace_copyout(uintptr_t kaddr, uintptr_t uaddr, size_t size,
+    volatile uint16_t *flags)
+{
+
+	if (dtrace_copycheck(uaddr, kaddr, size))
+		dtrace_copy(kaddr, uaddr, size);
+}
+
+void
+dtrace_copyinstr(uintptr_t uaddr, uintptr_t kaddr, size_t size,
+    volatile uint16_t *flags)
+{
+
+	if (dtrace_copycheck(uaddr, kaddr, size))
+		dtrace_copystr(uaddr, kaddr, size, flags);
+}
+
+void
+dtrace_copyoutstr(uintptr_t kaddr, uintptr_t uaddr, size_t size,
+    volatile uint16_t *flags)
+{
+
+	if (dtrace_copycheck(uaddr, kaddr, size))
+		dtrace_copystr(kaddr, uaddr, size, flags);
+}
+
+uint8_t
+dtrace_fuword8(void *uaddr)
+{
+
+	if ((uintptr_t)uaddr > VM_MAXUSER_ADDRESS) {
+		DTRACE_CPUFLAG_SET(CPU_DTRACE_BADADDR);
+		cpu_core[curcpu].cpuc_dtrace_illval = (uintptr_t)uaddr;
+		return (0);
+	}
+
+	return (dtrace_fuword8_nocheck(uaddr));
+}
+
+uint16_t
+dtrace_fuword16(void *uaddr)
+{
+
+	if ((uintptr_t)uaddr > VM_MAXUSER_ADDRESS) {
+		DTRACE_CPUFLAG_SET(CPU_DTRACE_BADADDR);
+		cpu_core[curcpu].cpuc_dtrace_illval = (uintptr_t)uaddr;
+		return (0);
+	}
+
+	return (dtrace_fuword16_nocheck(uaddr));
+}
+
+uint32_t
+dtrace_fuword32(void *uaddr)
+{
+
+	if ((uintptr_t)uaddr > VM_MAXUSER_ADDRESS) {
+		DTRACE_CPUFLAG_SET(CPU_DTRACE_BADADDR);
+		cpu_core[curcpu].cpuc_dtrace_illval = (uintptr_t)uaddr;
+		return (0);
+	}
+
+	return (dtrace_fuword32_nocheck(uaddr));
+}
+
+uint64_t
+dtrace_fuword64(void *uaddr)
+{
+
+	if ((uintptr_t)uaddr > VM_MAXUSER_ADDRESS) {
+		DTRACE_CPUFLAG_SET(CPU_DTRACE_BADADDR);
+		cpu_core[curcpu].cpuc_dtrace_illval = (uintptr_t)uaddr;
+		return (0);
+	}
+
+	return (dtrace_fuword64_nocheck(uaddr));
+}
diff --git a/sys/cddl/dev/dtrace/aarch64/dtrace_subr.c b/sys/cddl/dev/dtrace/aarch64/dtrace_subr.c
new file mode 100644
index 0000000..9cf5bc4
--- /dev/null
+++ b/sys/cddl/dev/dtrace/aarch64/dtrace_subr.c
@@ -0,0 +1,311 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License").  You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * $FreeBSD$
+ *
+ */
+/*
+ * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/types.h>
+#include <sys/kernel.h>
+#include <sys/malloc.h>
+#include <sys/kmem.h>
+#include <sys/smp.h>
+#include <sys/dtrace_impl.h>
+#include <sys/dtrace_bsd.h>
+#include <machine/armreg.h>
+#include <machine/clock.h>
+#include <machine/frame.h>
+#include <machine/trap.h>
+#include <vm/pmap.h>
+
+extern uintptr_t 	dtrace_in_probe_addr;
+extern int		dtrace_in_probe;
+extern dtrace_id_t	dtrace_probeid_error;
+extern int (*dtrace_invop_jump_addr)(struct trapframe *);
+extern void dtrace_getnanotime(struct timespec *tsp);
+
+int dtrace_invop(uintptr_t, uintptr_t *, uintptr_t);
+void dtrace_invop_init(void);
+void dtrace_invop_uninit(void);
+
+typedef struct dtrace_invop_hdlr {
+	int (*dtih_func)(uintptr_t, uintptr_t *, uintptr_t);
+	struct dtrace_invop_hdlr *dtih_next;
+} dtrace_invop_hdlr_t;
+
+dtrace_invop_hdlr_t *dtrace_invop_hdlr;
+
+int
+dtrace_invop(uintptr_t addr, uintptr_t *stack, uintptr_t eax)
+{
+	dtrace_invop_hdlr_t *hdlr;
+	int rval;
+
+	for (hdlr = dtrace_invop_hdlr; hdlr != NULL; hdlr = hdlr->dtih_next)
+		if ((rval = hdlr->dtih_func(addr, stack, eax)) != 0)
+			return (rval);
+
+	return (0);
+}
+
+
+void
+dtrace_invop_add(int (*func)(uintptr_t, uintptr_t *, uintptr_t))
+{
+	dtrace_invop_hdlr_t *hdlr;
+
+	hdlr = kmem_alloc(sizeof (dtrace_invop_hdlr_t), KM_SLEEP);
+	hdlr->dtih_func = func;
+	hdlr->dtih_next = dtrace_invop_hdlr;
+	dtrace_invop_hdlr = hdlr;
+}
+
+void
+dtrace_invop_remove(int (*func)(uintptr_t, uintptr_t *, uintptr_t))
+{
+	dtrace_invop_hdlr_t *hdlr, *prev;
+
+	hdlr = dtrace_invop_hdlr;
+	prev = NULL;
+
+	for (;;) {
+		if (hdlr == NULL)
+			panic("attempt to remove non-existent invop handler");
+
+		if (hdlr->dtih_func == func)
+			break;
+
+		prev = hdlr;
+		hdlr = hdlr->dtih_next;
+	}
+
+	if (prev == NULL) {
+		ASSERT(dtrace_invop_hdlr == hdlr);
+		dtrace_invop_hdlr = hdlr->dtih_next;
+	} else {
+		ASSERT(dtrace_invop_hdlr != hdlr);
+		prev->dtih_next = hdlr->dtih_next;
+	}
+
+	kmem_free(hdlr, 0);
+}
+
+/*ARGSUSED*/
+void
+dtrace_toxic_ranges(void (*func)(uintptr_t base, uintptr_t limit))
+{
+
+	printf("IMPLEMENT ME: dtrace_toxic_ranges\n");
+}
+
+void
+dtrace_xcall(processorid_t cpu, dtrace_xcall_t func, void *arg)
+{
+	cpuset_t cpus;
+
+	if (cpu == DTRACE_CPUALL)
+		cpus = all_cpus;
+	else
+		CPU_SETOF(cpu, &cpus);
+
+	smp_rendezvous_cpus(cpus, smp_no_rendevous_barrier, func,
+	    smp_no_rendevous_barrier, arg);
+}
+
+static void
+dtrace_sync_func(void)
+{
+
+}
+
+void
+dtrace_sync(void)
+{
+
+	dtrace_xcall(DTRACE_CPUALL, (dtrace_xcall_t)dtrace_sync_func, NULL);
+}
+
+/*
+ * DTrace needs a high resolution time function which can
+ * be called from a probe context and guaranteed not to have
+ * instrumented with probes itself.
+ *
+ * Returns nanoseconds since boot.
+ */
+uint64_t
+dtrace_gethrtime()
+{
+	struct timespec curtime;
+
+	nanouptime(&curtime);
+
+	return (curtime.tv_sec * 1000000000UL + curtime.tv_nsec);
+
+}
+
+uint64_t
+dtrace_gethrestime(void)
+{
+	struct timespec current_time;
+
+	dtrace_getnanotime(&current_time);
+
+	return (current_time.tv_sec * 1000000000UL + current_time.tv_nsec);
+}
+
+/* Function to handle DTrace traps during probes. See arm64/arm64/trap.c */
+int
+dtrace_trap(struct trapframe *frame, u_int type)
+{
+	/*
+	 * A trap can occur while DTrace executes a probe. Before
+	 * executing the probe, DTrace blocks re-scheduling and sets
+	 * a flag in it's per-cpu flags to indicate that it doesn't
+	 * want to fault. On returning from the probe, the no-fault
+	 * flag is cleared and finally re-scheduling is enabled.
+	 *
+	 * Check if DTrace has enabled 'no-fault' mode:
+	 *
+	 */
+
+	if ((cpu_core[curcpu].cpuc_dtrace_flags & CPU_DTRACE_NOFAULT) != 0) {
+		/*
+		 * There are only a couple of trap types that are expected.
+		 * All the rest will be handled in the usual way.
+		 */
+		switch (type) {
+		case EXCP_DATA_ABORT:
+			/* Flag a bad address. */
+			cpu_core[curcpu].cpuc_dtrace_flags |= CPU_DTRACE_BADADDR;
+			cpu_core[curcpu].cpuc_dtrace_illval = 0;
+
+			/*
+			 * Offset the instruction pointer to the instruction
+			 * following the one causing the fault.
+			 */
+			frame->tf_elr += 4;
+			return (1);
+		default:
+			/* Handle all other traps in the usual way. */
+			break;
+		}
+	}
+
+	/* Handle the trap in the usual way. */
+	return (0);
+}
+
+void
+dtrace_probe_error(dtrace_state_t *state, dtrace_epid_t epid, int which,
+    int fault, int fltoffs, uintptr_t illval)
+{
+
+	dtrace_probe(dtrace_probeid_error, (uint64_t)(uintptr_t)state,
+	    (uintptr_t)epid,
+	    (uintptr_t)which, (uintptr_t)fault, (uintptr_t)fltoffs);
+}
+
+static int
+dtrace_invop_start(struct trapframe *frame)
+{
+	int data, invop, reg, update_sp;
+	register_t arg1, arg2;
+	register_t *sp;
+	int offs;
+	int tmp;
+	int i;
+
+	invop = dtrace_invop(frame->tf_elr, (uintptr_t *)frame, frame->tf_elr);
+
+	tmp = (invop & LDP_STP_MASK);
+	if (tmp == STP_64 || tmp == LDP_64) {
+		sp = (register_t *)frame->tf_sp;
+		data = invop;
+		arg1 = (data >> ARG1_SHIFT) & ARG1_MASK;
+		arg2 = (data >> ARG2_SHIFT) & ARG2_MASK;
+
+		offs = (data >> OFFSET_SHIFT) & OFFSET_MASK;
+
+		switch (tmp) {
+		case STP_64:
+			if (offs >> (OFFSET_SIZE - 1))
+				sp -= (~offs & OFFSET_MASK) + 1;
+			else
+				sp += (offs);
+			*(sp + 0) = frame->tf_x[arg1];
+			*(sp + 1) = frame->tf_x[arg2];
+			break;
+		case LDP_64:
+			frame->tf_x[arg1] = *(sp + 0);
+			frame->tf_x[arg2] = *(sp + 1);
+			if (offs >> (OFFSET_SIZE - 1))
+				sp -= (~offs & OFFSET_MASK) + 1;
+			else
+				sp += (offs);
+			break;
+		default:
+			break;
+		}
+
+		/* Update the stack pointer and program counter to continue */
+		frame->tf_sp = (register_t)sp;
+		frame->tf_elr += INSN_SIZE;
+		return (0);
+	}
+
+	if ((invop & B_MASK) == B_INSTR) {
+		data = (invop & B_DATA_MASK);
+		/* The data is the number of 4-byte words to change the pc */
+		data *= 4;
+		frame->tf_elr += data;
+		return (0);
+	}
+
+	if (invop == RET_INSTR) {
+		frame->tf_elr = frame->tf_lr;
+		return (0);
+	}
+
+	return (-1);
+}
+
+void
+dtrace_invop_init(void)
+{
+
+	dtrace_invop_jump_addr = dtrace_invop_start;
+}
+
+void
+dtrace_invop_uninit(void)
+{
+
+	dtrace_invop_jump_addr = 0;
+}
diff --git a/sys/cddl/dev/dtrace/aarch64/regset.h b/sys/cddl/dev/dtrace/aarch64/regset.h
new file mode 100644
index 0000000..f99b48f
--- /dev/null
+++ b/sys/cddl/dev/dtrace/aarch64/regset.h
@@ -0,0 +1,51 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License, Version 1.0 only
+ * (the "License").  You may not use this file except in compliance
+ * with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * $FreeBSD$ 
+ */
+/*
+ * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*	Copyright (c) 1990, 1991 UNIX System Laboratories, Inc. */
+
+/*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T		*/
+/*	All Rights Reserved	*/
+
+#ifndef	_REGSET_H
+#define	_REGSET_H
+
+/*
+ * #pragma ident	"@(#)regset.h	1.11	05/06/08 SMI"
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Place here */
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif	/* _REGSET_H */
diff --git a/sys/cddl/dev/fbt/aarch64/fbt_isa.c b/sys/cddl/dev/fbt/aarch64/fbt_isa.c
new file mode 100644
index 0000000..94de1e8
--- /dev/null
+++ b/sys/cddl/dev/fbt/aarch64/fbt_isa.c
@@ -0,0 +1,180 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * Portions Copyright 2006-2008 John Birrell jb@freebsd.org
+ * Portions Copyright 2013 Justin Hibbits jhibbits@freebsd.org
+ * Portions Copyright 2013 Howard Su howardsu@freebsd.org
+ * Portions Copyright 2015 Ruslan Bukin <br@bsdpad.com>
+ *
+ * $FreeBSD$
+ */
+
+/*
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+#include <sys/cdefs.h>
+#include <sys/param.h>
+
+#include <sys/dtrace.h>
+
+#include "fbt.h"
+
+#define	AARCH64_BRK		0xd4200000
+#define	AARCH64_BRK_IMM16_SHIFT	5
+#define	AARCH64_BRK_IMM16_VAL	(0x40d << AARCH64_BRK_IMM16_SHIFT)
+#define	FBT_PATCHVAL		(AARCH64_BRK | AARCH64_BRK_IMM16_VAL)
+#define	FBT_ENTRY	"entry"
+#define	FBT_RETURN	"return"
+
+int
+fbt_invop(uintptr_t addr, uintptr_t *stack, uintptr_t rval)
+{
+	struct trapframe *frame;
+	solaris_cpu_t *cpu;
+	fbt_probe_t *fbt;
+
+	frame = (struct trapframe *)stack;
+	cpu = &solaris_cpu[curcpu];
+	fbt = fbt_probetab[FBT_ADDR2NDX(addr)];
+
+	for (; fbt != NULL; fbt = fbt->fbtp_hashnext) {
+		if ((uintptr_t)fbt->fbtp_patchpoint == addr) {
+			fbt->fbtp_invop_cnt++;
+			cpu->cpu_dtrace_caller = addr;
+
+			dtrace_probe(fbt->fbtp_id, frame->tf_x[0],
+			    frame->tf_x[1], frame->tf_x[2],
+			    frame->tf_x[3], frame->tf_x[4]);
+
+			cpu->cpu_dtrace_caller = 0;
+			return (fbt->fbtp_savedval);
+		}
+	}
+
+	return (0);
+}
+
+void
+fbt_patch_tracepoint(fbt_probe_t *fbt, fbt_patchval_t val)
+{
+
+	*fbt->fbtp_patchpoint = val;
+	cpu_icache_sync_range((vm_offset_t)fbt->fbtp_patchpoint, 4);
+}
+
+int
+fbt_provide_module_function(linker_file_t lf, int symindx,
+    linker_symval_t *symval, void *opaque)
+{
+	fbt_probe_t *fbt, *retfbt;
+	uint32_t *target, *start;
+	uint32_t *instr, *limit;
+	const char *name;
+	char *modname;
+	int offs;
+
+	modname = opaque;
+	name = symval->name;
+
+	/* Check if function is excluded from instrumentation */
+	if (fbt_excluded(name))
+		return (0);
+
+	instr = (uint32_t *)(symval->value);
+	limit = (uint32_t *)(symval->value + symval->size);
+
+	/* Look for stp (pre-indexed) operation */
+	for (; instr < limit; instr++) {
+		if ((*instr & LDP_STP_MASK) == STP_64)
+			break;
+	}
+
+	if (instr >= limit)
+		return (0);
+
+	fbt = malloc(sizeof (fbt_probe_t), M_FBT, M_WAITOK | M_ZERO);
+	fbt->fbtp_name = name;
+	fbt->fbtp_id = dtrace_probe_create(fbt_id, modname,
+	    name, FBT_ENTRY, 3, fbt);
+	fbt->fbtp_patchpoint = instr;
+	fbt->fbtp_ctl = lf;
+	fbt->fbtp_loadcnt = lf->loadcnt;
+	fbt->fbtp_savedval = *instr;
+	fbt->fbtp_patchval = FBT_PATCHVAL;
+	fbt->fbtp_rval = DTRACE_INVOP_PUSHM;
+	fbt->fbtp_symindx = symindx;
+
+	fbt->fbtp_hashnext = fbt_probetab[FBT_ADDR2NDX(instr)];
+	fbt_probetab[FBT_ADDR2NDX(instr)] = fbt;
+
+	lf->fbt_nentries++;
+
+	retfbt = NULL;
+again:
+	for (; instr < limit; instr++) {
+		if (*instr == RET_INSTR)
+			break;
+		else if ((*instr & B_MASK) == B_INSTR) {
+			offs = (*instr & B_DATA_MASK);
+			offs *= 4;
+			target = (instr + offs);
+			start = (uint32_t *)symval->value;
+			if (target >= limit || target < start)
+				break;
+		}
+	}
+
+	if (instr >= limit)
+		return (0);
+
+	/*
+	 * We have a winner!
+	 */
+	fbt = malloc(sizeof (fbt_probe_t), M_FBT, M_WAITOK | M_ZERO);
+	fbt->fbtp_name = name;
+	if (retfbt == NULL) {
+		fbt->fbtp_id = dtrace_probe_create(fbt_id, modname,
+		    name, FBT_RETURN, 3, fbt);
+	} else {
+		retfbt->fbtp_next = fbt;
+		fbt->fbtp_id = retfbt->fbtp_id;
+	}
+	retfbt = fbt;
+
+	fbt->fbtp_patchpoint = instr;
+	fbt->fbtp_ctl = lf;
+	fbt->fbtp_loadcnt = lf->loadcnt;
+	fbt->fbtp_symindx = symindx;
+	if ((*instr & B_MASK) == B_INSTR)
+		fbt->fbtp_rval = DTRACE_INVOP_B;
+	else
+		fbt->fbtp_rval = DTRACE_INVOP_RET;
+	fbt->fbtp_savedval = *instr;
+	fbt->fbtp_patchval = FBT_PATCHVAL;
+	fbt->fbtp_hashnext = fbt_probetab[FBT_ADDR2NDX(instr)];
+	fbt_probetab[FBT_ADDR2NDX(instr)] = fbt;
+
+	lf->fbt_nentries++;
+
+	instr++;
+	goto again;
+}
diff --git a/sys/cddl/dev/fbt/aarch64/fbt_isa.h b/sys/cddl/dev/fbt/aarch64/fbt_isa.h
new file mode 100644
index 0000000..5552f31
--- /dev/null
+++ b/sys/cddl/dev/fbt/aarch64/fbt_isa.h
@@ -0,0 +1,30 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ *
+ * $FreeBSD$
+ *
+ */
+
+#ifndef _FBT_ISA_H_
+#define _FBT_ISA_H_
+
+typedef uint32_t fbt_patchval_t;
+
+#endif
diff --git a/sys/cddl/dev/lockstat/lockstat.c b/sys/cddl/dev/lockstat/lockstat.c
index a4e4efd..71a8632 100644
--- a/sys/cddl/dev/lockstat/lockstat.c
+++ b/sys/cddl/dev/lockstat/lockstat.c
@@ -43,9 +43,8 @@
 #include <sys/dtrace.h>
 #include <sys/lockstat.h>
 
-#if defined(__i386__) || defined(__amd64__) || \
-	defined(__mips__) || defined(__powerpc__) || \
-	defined(__arm__)
+#if defined(__aarch64__) || defined(__amd64__) || defined(__arm__) || \
+    defined(__i386__) || defined(__mips__) || defined(__powerpc__)
 #define LOCKSTAT_AFRAMES 1
 #else
 #error "architecture not supported"
diff --git a/sys/cddl/dev/profile/profile.c b/sys/cddl/dev/profile/profile.c
index dd13397..80a0eca 100644
--- a/sys/cddl/dev/profile/profile.c
+++ b/sys/cddl/dev/profile/profile.c
@@ -140,6 +140,11 @@ struct profile_probe_percpu;
 #define	PROF_ARTIFICIAL_FRAMES	10
 #endif
 
+#ifdef __aarch64__
+/* TODO: verify */
+#define	PROF_ARTIFICIAL_FRAMES	10
+#endif
+
 typedef struct profile_probe {
 	char		prof_name[PROF_NAMELEN];
 	dtrace_id_t	prof_id;