On 09/01/2020 10:52, Jan Beulich wrote:
>>>> --- a/xen/arch/x86/smpboot.c
>>>> +++ b/xen/arch/x86/smpboot.c
>>>> @@ -554,7 +554,7 @@ static int do_boot_cpu(int apicid, int cpu)
>>>> printk("Booting processor %d/%d eip %lx\n",
>>>> cpu, apicid, start_eip);
>>>>
>>>> - stack_start = stack_base[cpu];
>>>> + stack_start = stack_base[cpu] + STACK_SIZE - sizeof(struct cpu_info);
>>>>
>>>> /* This grunge runs the startup process for the targeted processor. */
>>> Further down smp_prepare_cpus() has
>>>
>>> stack_base[0] = stack_start;
>>>
>>> which I think you need to also adjust (or replace, e.g. by giving
>>> stack_base[] an initializer).
>> In practice, this variable is never used because we never offline the BSP.
> traps.c uses stack_base[], for example, and hence it needs to be
> correct also for the BSP. Even more important is perhaps
> get_cpu_current()'s use.
>
>> However, the code as-is is correct. The value in stack_start has
>> changed in this patch, but is still the correct value for the BSP.
> But it's no longer the stack base (which is what stack_base[] holds
> for all other CPUs), or else you wouldn't have had a need to change
> do_boot_cpu().
Oh - of course. I'm being dense.
To avoid a move into data, lets just go with & ~(STACK_SIZE - 1) here.
~Andrew
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
