Re: [PATCH 1/2] x86: stop using PER_CPU_VAR(kernel_stack)

[Denys Vlasenko <dvlasenk@redhat.com>]

On 03/23/2015 06:28 PM, Steven Rostedt wrote:
> On Mon, 23 Mar 2015 18:10:01 +0100
> Denys Vlasenko <dvlasenk@redhat.com> wrote:
> 
>> On 03/23/2015 03:18 PM, Steven Rostedt wrote:
>>> On Sat, 21 Mar 2015 21:44:37 +0100
>>> Denys Vlasenko <dvlasenk@redhat.com> wrote:
>>>
>>>> Instead of PER_CPU_VAR(kernel_stack), 64-bit code
>>>> can use PER_CPU_VAR(cpu_tss + TSS_sp0).
>>>
>>> The change log here is lacking an answer to "why". It only states what
>>> it does. What's wrong with using kernel_stack? The change log should
>>> explicitly state that. I have no idea why this patch is needed.
>>
>> Sorry. The reason is:
>>
>> We want to get rid of kernel_stack, since it is redundant:
>> in 64-bits, PER_CPU_VAR(cpu_tss + TSS_sp0) can be used instead,
>> in 32-bits, PER_CPU_VAR(cpu_current_top_of_stack) can be used instead.
> 
> Can we do a:
> 
>  #define cpu_current_top_of_stack (cpu_tss + TSS_sp0)

We already do something similar:

static inline unsigned long current_top_of_stack(void)
{
#ifdef CONFIG_X86_64
        return this_cpu_read_stable(cpu_tss.x86_tss.sp0);
#else
        /* sp0 on x86_32 is special in and around vm86 mode. */
        return this_cpu_read_stable(cpu_current_top_of_stack);
#endif
}