diff --git a/arch/i386/kernel/entry.S b/arch/i386/kernel/entry.S
index 1e45ff292bc..3c73dc865ea 100644
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -245,6 +245,9 @@ syscall_exit:
 
 restore_all:
 	movl EFLAGS(%esp), %eax		# mix EFLAGS, SS and CS
+	# Warning: OLDSS(%esp) contains the wrong/random values if we
+	# are returning to the kernel.
+	# See comments in process.c:copy_thread() for details.
 	movb OLDSS(%esp), %ah
 	movb CS(%esp), %al
 	andl $(VM_MASK | (4 << 8) | 3), %eax
diff --git a/arch/i386/kernel/process.c b/arch/i386/kernel/process.c
index c36fedf40e9..36145efc61b 100644
--- a/arch/i386/kernel/process.c
+++ b/arch/i386/kernel/process.c
@@ -405,7 +405,17 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 	childregs->esp = esp;
 
 	p->thread.esp = (unsigned long) childregs;
-	p->thread.esp0 = (unsigned long) (childregs+1);
+	/*
+	 * The below -8 is to reserve 8 bytes on top of the ring0 stack.
+	 * This is necessary to guarantee that the entire "struct pt_regs"
+	 * is accessable even if the CPU haven't stored the SS/ESP registers
+	 * on the stack (interrupt gate does not save these registers
+	 * when switching to the same priv ring).
+	 * Therefore beware: accessing the xss/esp fields of the
+	 * "struct pt_regs" is possible, but they may contain the
+	 * completely wrong values.
+	 */
+	p->thread.esp0 = (unsigned long) (childregs+1) - 8;
 
 	p->thread.eip = (unsigned long) ret_from_fork;