diff --git a/sys/arm64/arm64/elf32_machdep.c b/sys/arm64/arm64/elf32_machdep.c
--- a/sys/arm64/arm64/elf32_machdep.c
+++ b/sys/arm64/arm64/elf32_machdep.c
@@ -51,6 +51,9 @@
 #include <sys/vnode.h>
 
 #include <machine/elf.h>
+#ifdef VFP
+#include <machine/vfp.h>
+#endif
 
 #include <compat/freebsd32/freebsd32_util.h>
 
@@ -251,6 +254,10 @@
 	tf->tf_x[14] = imgp->entry_addr;
 	tf->tf_elr = imgp->entry_addr;
 	tf->tf_spsr = PSR_M_32;
+
+#ifdef VFP
+	vfp_reset_state(td, td->td_pcb);
+#endif
 }
 
 void
diff --git a/sys/arm64/arm64/machdep.c b/sys/arm64/arm64/machdep.c
--- a/sys/arm64/arm64/machdep.c
+++ b/sys/arm64/arm64/machdep.c
@@ -552,6 +552,7 @@
 exec_setregs(struct thread *td, struct image_params *imgp, uintptr_t stack)
 {
 	struct trapframe *tf = td->td_frame;
+	struct pcb *pcb = td->td_pcb;
 
 	memset(tf, 0, sizeof(struct trapframe));
 
@@ -559,6 +560,12 @@
 	tf->tf_sp = STACKALIGN(stack);
 	tf->tf_lr = imgp->entry_addr;
 	tf->tf_elr = imgp->entry_addr;
+
+#ifdef VFP
+	vfp_reset_state(td, pcb);
+#endif
+
+	/* TODO: Shouldn't we also reset pcb_dbg_regs? */
 }
 
 /* Sanity check these are the same size, they will be memcpy'd to and fro */
diff --git a/sys/arm64/arm64/vfp.c b/sys/arm64/arm64/vfp.c
--- a/sys/arm64/arm64/vfp.c
+++ b/sys/arm64/arm64/vfp.c
@@ -33,6 +33,7 @@
 #ifdef VFP
 #include <sys/param.h>
 #include <sys/systm.h>
+#include <sys/limits.h>
 #include <sys/kernel.h>
 #include <sys/malloc.h>
 #include <sys/pcpu.h>
@@ -199,6 +200,26 @@
 	critical_exit();
 }
 
+/*
+ * Reset the FP state to avoid leaking state from the parent process across
+ * execve() (and to ensure that we get a consistent floating point environment
+ * in every new process).
+ */
+void
+vfp_reset_state(struct thread *td, struct pcb *pcb)
+{
+	critical_enter();
+	bzero(&pcb->pcb_fpustate.vfp_regs, sizeof(pcb->pcb_fpustate.vfp_regs));
+	KASSERT(pcb->pcb_fpusaved == &pcb->pcb_fpustate,
+	    ("pcb_fpusaved should point to pcb_fpustate."));
+	pcb->pcb_fpustate.vfp_fpcr = initial_fpcr;
+	pcb->pcb_fpustate.vfp_fpsr = 0;
+	pcb->pcb_vfpcpu = UINT_MAX;
+	pcb->pcb_fpflags = 0;
+	vfp_discard(td);
+	critical_exit();
+}
+
 void
 vfp_restore_state(void)
 {
diff --git a/sys/arm64/arm64/vm_machdep.c b/sys/arm64/arm64/vm_machdep.c
--- a/sys/arm64/arm64/vm_machdep.c
+++ b/sys/arm64/arm64/vm_machdep.c
@@ -108,7 +108,6 @@
 	td2->td_pcb->pcb_sp = (uintptr_t)td2->td_frame;
 	td2->td_pcb->pcb_fpusaved = &td2->td_pcb->pcb_fpustate;
 	td2->td_pcb->pcb_vfpcpu = UINT_MAX;
-	td2->td_pcb->pcb_fpusaved->vfp_fpcr = initial_fpcr;
 
 	/* Setup to release spin count in fork_exit(). */
 	td2->td_md.md_spinlock_count = 1;
diff --git a/sys/arm64/include/vfp.h b/sys/arm64/include/vfp.h
--- a/sys/arm64/include/vfp.h
+++ b/sys/arm64/include/vfp.h
@@ -68,6 +68,7 @@
 
 void	vfp_init(void);
 void	vfp_discard(struct thread *);
+void	vfp_reset_state(struct thread *, struct pcb *);
 void	vfp_restore_state(void);
 void	vfp_save_state(struct thread *, struct pcb *);
 
diff --git a/sys/arm64/linux/linux_sysvec.c b/sys/arm64/linux/linux_sysvec.c
--- a/sys/arm64/linux/linux_sysvec.c
+++ b/sys/arm64/linux/linux_sysvec.c
@@ -57,6 +57,10 @@
 #include <compat/linux/linux_util.h>
 #include <compat/linux/linux_vdso.h>
 
+#ifdef VFP
+#include <machine/vfp.h>
+#endif
+
 MODULE_VERSION(linux64elf, 1);
 
 const char *linux_kplatform;
@@ -360,6 +364,10 @@
         regs->tf_lr = 0xffffffffffffffff;
 #endif
         regs->tf_elr = imgp->entry_addr;
+
+#ifdef VFP
+	vfp_reset_state(td, td->td_pcb);
+#endif
 }
 
 int