Use of memmove() in #DF is problematic considered tracing and other
instrumentation.
Remove the memmove() call and simply write out what needs doing; this
even clarifies the code, win-win! The code copies from the espfix64
stack to the normal task stack, there is no possible way for that to
overlap.
Survives selftests/x86, specifically sigreturn_64.
Suggested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Alexandre Chartre <alexandre.chartre@oracle.com>
Acked-by: Andy Lutomirski <luto@kernel.org>
Link: https://lkml.kernel.org/r/20200505134058.863038566@linutronix.de
                regs->ip == (unsigned long)native_irq_return_iret)
        {
                struct pt_regs *gpregs = (struct pt_regs *)this_cpu_read(cpu_tss_rw.x86_tss.sp0) - 1;
+               unsigned long *p = (unsigned long *)regs->sp;
 
                /*
                 * regs->sp points to the failing IRET frame on the
                 * in gpregs->ss through gpregs->ip.
                 *
                 */
-               memmove(&gpregs->ip, (void *)regs->sp, 5*8);
+               gpregs->ip      = p[0];
+               gpregs->cs      = p[1];
+               gpregs->flags   = p[2];
+               gpregs->sp      = p[3];
+               gpregs->ss      = p[4];
                gpregs->orig_ax = 0;  /* Missing (lost) #GP error code */
 
                /*