diff --git a/kernel/arch/x86_64/Signal.S b/kernel/arch/x86_64/Signal.S
index 2b82cb511..0e24f639b 100644
--- a/kernel/arch/x86_64/Signal.S
+++ b/kernel/arch/x86_64/Signal.S
@@ -2,6 +2,11 @@
 
 #include <sys/syscall.h>
 
+// stack contains
+//    return address
+//    signal number
+//    signal handler
+
 .global signal_trampoline
 signal_trampoline:
 	pushq %rax
@@ -20,6 +25,8 @@ signal_trampoline:
 	pushq %r14
 	pushq %r15
 
+	// This is 16 byte aligned
+
 	movq 128(%rsp), %rdi
 	movq 120(%rsp), %rax
 	call *%rax
@@ -46,4 +53,5 @@ signal_trampoline:
 
 	addq $16, %rsp
 
-	ret
+	// return over red-zone
+	ret $128
diff --git a/kernel/kernel/Thread.cpp b/kernel/kernel/Thread.cpp
index c6a30f478..800c11349 100644
--- a/kernel/kernel/Thread.cpp
+++ b/kernel/kernel/Thread.cpp
@@ -83,6 +83,7 @@ namespace Kernel
 		thread->m_rip = (uintptr_t)entry;
 
 		// Initialize stack for returning
+		write_to_stack(thread->m_rsp, nullptr); // alignment
 		write_to_stack(thread->m_rsp, thread);
 		write_to_stack(thread->m_rsp, &Thread::on_exit);
 		write_to_stack(thread->m_rsp, data);
@@ -179,6 +180,7 @@ namespace Kernel
 		{
 			// FIXME: don't use PageTableScope
 			PageTableScope _(process().page_table());
+			write_to_stack(m_rsp, nullptr); // alignment
 			write_to_stack(m_rsp, this);
 			write_to_stack(m_rsp, &Thread::on_exit);
 			write_to_stack(m_rsp, nullptr);
@@ -206,6 +208,7 @@ namespace Kernel
 		{
 			// FIXME: don't use PageTableScope
 			PageTableScope _(process().page_table());
+			write_to_stack(m_rsp, nullptr); // alignment
 			write_to_stack(m_rsp, this);
 			write_to_stack(m_rsp, &Thread::on_exit);
 			write_to_stack(m_rsp, m_process);
@@ -275,6 +278,7 @@ namespace Kernel
 			// FIXME: signal trampoline should take a hash etc
 			//        to only allow marking signals done from it
 			m_handling_signal = signal;
+			return_rsp += 128; // skip possible red-zone
 			write_to_stack(return_rsp, return_rip);
 			write_to_stack(return_rsp, signal);
 			write_to_stack(return_rsp, signal_handler);