Re: [RFC 2/2] x86/pti/64: Remove the SYSCALL64 entry trampoline

[Dave Hansen]

On 07/22/2018 10:45 AM, Andy Lutomirski wrote:
>  static void __init pti_clone_user_shared(void)
>  {
> +	unsigned cpu;
> +
>  	pti_clone_p4d(CPU_ENTRY_AREA_BASE);
> +
> +	for_each_possible_cpu(cpu) {
> +		/*
> +		 * The SYSCALL64 entry code needs to be able to find the
> +		 * thread stack and needs one word of scratch space in which
> +		 * to spill a register.  All of this lives in the TSS, in
> +		 * the sp1 and sp2 slots.
> +		 */

I had to remind myself about the r/o cpu_entry_area alias here.  Should
we maybe call it out explicitly?

/*
 * The TSS is also mapped read-only into the cpu_entry_area.
 * The cpu_entry_area copy is used r/o by the hardware for the
 * hardware stack switching, like interrupt entry.
 *
 * The copies being mapped here are the normal r/w per-cpu
 * areas.  We need r/w because we spill a register here.
 */

BTW, since we have this alias, do we still *need* the r/o cpu_entry_area
alias?  Or do we still get some value from keeping the thing referenced
by hardware r/o?