diff --git a/sys/arm64/arm64/genassym.c b/sys/arm64/arm64/genassym.c
index 22696796e69..b54f446e83b 100644
--- a/sys/arm64/arm64/genassym.c
+++ b/sys/arm64/arm64/genassym.c
@@ -53,8 +53,6 @@ ASSYM(PC_CURPCB, offsetof(struct pcpu, pc_curpcb));
 ASSYM(PC_CURTHREAD, offsetof(struct pcpu, pc_curthread));
 ASSYM(PC_SSBD, offsetof(struct pcpu, pc_ssbd));
 
-/* Size of pcb, rounded to keep stack alignment */
-ASSYM(PCB_SIZE, roundup2(sizeof(struct pcb), STACKALIGNBYTES + 1));
 ASSYM(PCB_SINGLE_STEP_SHIFT, PCB_SINGLE_STEP_SHIFT);
 ASSYM(PCB_REGS, offsetof(struct pcb, pcb_x));
 ASSYM(PCB_X19, PCB_X19);
diff --git a/sys/arm64/arm64/locore.S b/sys/arm64/arm64/locore.S
index b200aa93c28..bd61b485edf 100644
--- a/sys/arm64/arm64/locore.S
+++ b/sys/arm64/arm64/locore.S
@@ -128,8 +128,7 @@ virtdone:
 
 	/* Set up the stack */
 	adrp	x25, initstack_end
-	add	x25, x25, :lo12:initstack_end
-	sub	sp, x25, #PCB_SIZE
+	add	sp, x25, :lo12:initstack_end
 
 	/* Zero the BSS */
 	ldr	x15, .Lbss
diff --git a/sys/arm64/arm64/machdep.c b/sys/arm64/arm64/machdep.c
index 6b6dd2c510a..d33208a0819 100644
--- a/sys/arm64/arm64/machdep.c
+++ b/sys/arm64/arm64/machdep.c
@@ -131,6 +131,7 @@ uintptr_t boot_canary = 0x49a2d892bc05a0b1ul;
 #endif
 
 static struct trapframe proc0_tf;
+static struct pcb pcb0;
 
 int early_boot = 1;
 int cold = 1;
@@ -443,7 +444,7 @@ init_proc0(void *kstack)
 #if defined(PERTHREAD_SSP)
 	thread0.td_md.md_canary = boot_canary;
 #endif
-	thread0.td_pcb = (struct pcb *)td_kstack_top(&thread0) - 1;
+	thread0.td_pcb = &pcb0;
 	thread0.td_pcb->pcb_flags = 0;
 	thread0.td_pcb->pcb_fpflags = 0;
 	thread0.td_pcb->pcb_fpusaved = &thread0.td_pcb->pcb_fpustate;
diff --git a/sys/arm64/arm64/vm_machdep.c b/sys/arm64/arm64/vm_machdep.c
index 29a1c8667e7..3789dd318c4 100644
--- a/sys/arm64/arm64/vm_machdep.c
+++ b/sys/arm64/arm64/vm_machdep.c
@@ -27,8 +27,8 @@
 
 #include "opt_platform.h"
 
-#include <sys/param.h>
 #include <sys/systm.h>
+#include <sys/kernel.h>
 #include <sys/limits.h>
 #include <sys/proc.h>
 #include <sys/sf_buf.h>
@@ -61,6 +61,8 @@
  */
 cpu_reset_hook_t cpu_reset_hook = psci_reset;
 
+static uma_zone_t pcb_zone;
+
 /*
  * Finish a fork operation, with process p2 nearly set up.
  * Copy and update the pcb, set up the stack so that the child
@@ -260,20 +262,20 @@ cpu_thread_exit(struct thread *td)
 void
 cpu_thread_alloc(struct thread *td)
 {
+	td->td_pcb = uma_zalloc(pcb_zone, M_WAITOK);
 	ptrauth_thread_alloc(td);
 }
 
 void
 cpu_thread_new_kstack(struct thread *td)
 {
-	td->td_pcb = (struct pcb *)td_kstack_top(td) - 1;
-	td->td_frame = (struct trapframe *)STACKALIGN(
-	    (struct trapframe *)td->td_pcb - 1);
+	td->td_frame = (struct trapframe *)td_kstack_top(td) - 1;
 }
 
 void
 cpu_thread_free(struct thread *td)
 {
+	uma_zfree(pcb_zone, td->td_pcb);
 }
 
 void
@@ -333,3 +335,11 @@ cpu_sync_core(void)
 	 * return from ELx is a context synchronization event.
 	 */
 }
+
+static void
+pcbinit(void *dummy __unused)
+{
+	pcb_zone = uma_zcreate("pcb", sizeof(struct pcb), NULL, NULL, NULL,
+	    NULL, UMA_ALIGNOF(struct pcb), 0);
+}
+SYSINIT(pcbinit, SI_SUB_INTRINSIC, SI_ORDER_ANY, pcbinit, NULL);
diff --git a/sys/arm64/include/stack.h b/sys/arm64/include/stack.h
index 0aa483e15e6..23e7a5af27d 100644
--- a/sys/arm64/include/stack.h
+++ b/sys/arm64/include/stack.h
@@ -39,11 +39,9 @@ bool unwind_frame(struct thread *, struct unwind_state *);
 
 #ifdef _SYS_PROC_H_
 
-#include <machine/pcb.h>
-
 #define	GET_STACK_USAGE(total, used) do {				\
 	struct thread *td = curthread;					\
-	(total) = ptoa(td->td_kstack_pages) - sizeof(struct pcb);	\
+	(total) = ptoa(td->td_kstack_pages);				\
 	(used) = td->td_kstack + (total) - (char *)&td;			\
 } while (0)
 
@@ -51,7 +49,7 @@ static __inline bool
 kstack_contains(struct thread *td, vm_offset_t va, size_t len)
 {
 	return (va >= (vm_offset_t)td->td_kstack && va + len >= va &&
-	    va + len <= (vm_offset_t)td_kstack_top(td) - sizeof(struct pcb));
+	    va + len <= (vm_offset_t)td_kstack_top(td));
 }
 #endif	/* _SYS_PROC_H_ */