Re: [crash, bisected] Re: [PATCH 3/4] x86_64: Fold pda into per cpu area

[Mike Travis <travis@sgi.com>]

Jeremy Fitzhardinge wrote:
> Mike Travis wrote:
...
>> I'm not sure yet what to put into acpi_save_state_mem:
>>
>>         initial_code = (unsigned long)wakeup_long64;
>> +       /* ZZZ initial_pda = (unsigned long)?; */
>>   
> 
> You'll need to change wakeup_long64 to load the right value into the
> GS_BASE msr anyway.
> 
>    J

I'm afraid I don't quite understand the transitioning of the
ACPI states to figure out the correct thing to do.  My first
inclination would be to:  [sorry, cut and pasted]

--- linux-2.6.tip.orig/arch/x86/kernel/acpi/sleep.c
+++ linux-2.6.tip/arch/x86/kernel/acpi/sleep.c
@@ -89,6 +89,8 @@ int acpi_save_state_mem(void)
 #ifdef CONFIG_SMP
        stack_start.sp = temp_stack + 4096;
 #endif
+       early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
+       initial_pda = (unsigned long)get_cpu_pda(cpu);
        initial_code = (unsigned long)wakeup_long64;
        saved_magic = 0x123456789abcdef0;
 #endif /* CONFIG_64BIT */


But I'd like some confirmation that this is right thing to do...

[This mimics what smpboot.c:do_boot_cpu() does.]

Len - I'll cc you on the full patch submission shortly.

Thanks,
Mike