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
@@ -38,6 +38,8 @@
 #include <sys/pcpu.h>
 #include <sys/proc.h>
 
+#include <vm/uma.h>
+
 #include <machine/armreg.h>
 #include <machine/md_var.h>
 #include <machine/pcb.h>
@@ -57,6 +59,9 @@
 	struct vfpstate	 state;
 };
 
+static	uma_zone_t fpu_save_area_zone;
+static	struct vfpstate *fpu_initialstate;
+
 void
 vfp_enable(void)
 {
@@ -291,12 +296,37 @@
 
 	/* Disable to be enabled when it's used */
 	vfp_disable();
+}
+
+static void
+vfp_sysinit(const void *dummy __unused)
+{
+	uint64_t pfr;
+
+	/* Check if there is a vfp unit present */
+	pfr = READ_SPECIALREG(id_aa64pfr0_el1);
+	if ((pfr & ID_AA64PFR0_FP_MASK) == ID_AA64PFR0_FP_NONE)
+		return;
+
+	fpu_save_area_zone = uma_zcreate("VFP_save_area",
+	    sizeof(struct vfpstate), NULL, NULL, NULL, NULL,
+	    __alignof(struct vfpstate) - 1, 0);
+	fpu_initialstate = uma_zalloc(fpu_save_area_zone, M_WAITOK | M_ZERO);
+
+	/* Ensure the VFP is enabled before accessing it in vfp_store */
+	vfp_enable();
+	vfp_store(fpu_initialstate);
+
+	/* Disable to be enabled when it's used */
+	vfp_disable();
 
-	if (PCPU_GET(cpuid) == 0)
-		thread0.td_pcb->pcb_fpusaved->vfp_fpcr = VFPCR_INIT;
+	/* Zero the VFP registers but keep fpcr and fpsr */
+	bzero(fpu_initialstate->vfp_regs, sizeof(fpu_initialstate->vfp_regs));
+
+	thread0.td_pcb->pcb_fpusaved->vfp_fpcr = VFPCR_INIT;
 }
 
-SYSINIT(vfp, SI_SUB_CPU, SI_ORDER_ANY, vfp_init, NULL);
+SYSINIT(vfp, SI_SUB_CPU, SI_ORDER_ANY, vfp_sysinit, NULL);
 
 struct fpu_kern_ctx *
 fpu_kern_alloc_ctx(u_int flags)
@@ -433,4 +463,27 @@
 	curpcb = curthread->td_pcb;
 	return ((curpcb->pcb_fpflags & PCB_FP_KERN) != 0);
 }
+
+/*
+ * FPU save area alloc/free/init utility routines
+ */
+struct vfpstate *
+fpu_save_area_alloc(void)
+{
+
+	return (uma_zalloc(fpu_save_area_zone, M_WAITOK));
+}
+
+void
+fpu_save_area_free(struct vfpstate *fsa)
+{
+
+	uma_zfree(fpu_save_area_zone, fsa);
+}
+
+void
+fpu_save_area_reset(struct vfpstate *fsa)
+{
+	memcpy(fsa, fpu_initialstate, sizeof(*fsa));
+}
 #endif
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
@@ -94,6 +94,10 @@
 int fpu_kern_thread(u_int);
 int is_fpu_kern_thread(u_int);
 
+struct vfpstate	*fpu_save_area_alloc(void);
+void	fpu_save_area_free(struct vfpstate *fsa);
+void	fpu_save_area_reset(struct vfpstate *fsa);
+
 /* Convert to and from Aarch32 FPSCR to Aarch64 FPCR/FPSR */
 #define VFP_FPSCR_FROM_SRCR(vpsr, vpcr) ((vpsr) | ((vpcr) & 0x7c00000))
 #define VFP_FPSR_FROM_FPSCR(vpscr) ((vpscr) &~ 0x7c00000)