Re: [PATCH 16/21] x86/entry/64: Use a per-CPU trampoline stack for IDT entries

[Andy Lutomirski]

> On Dec 1, 2017, at 9:06 AM, Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx> wrote:
>
>> On 11/27/2017 02:45 AM, Ingo Molnar wrote:
>> --- a/arch/x86/kernel/cpu/common.c
>> +++ b/arch/x86/kernel/cpu/common.c
>> @@ -1645,11 +1645,13 @@ void cpu_init(void)
>>    setup_cpu_entry_area(cpu);
>>
>>    /*
>> -     * Initialize the TSS.  Don't bother initializing sp0, as the initial
>> -     * task never enters user mode.
>> +     * Initialize the TSS.  sp0 points to the entry trampoline stack
>> +     * regardless of what task is running.
>>     */
>>    set_tss_desc(cpu, &get_cpu_entry_area(cpu)->tss.x86_tss);
>>    load_TR_desc();
>> +    load_sp0((unsigned long)&get_cpu_entry_area(cpu)->tss +
>> +         offsetofend(struct tss_struct, SYSENTER_stack));
>>
>>    load_mm_ldt(&init_mm);
>
> Does this also mean that our stack dumps will say "<SYSENTER>" in oopses?

Only if we oops while we're on the trampoline stack.  If the oops is
due to a bug that isn't in the entry code, then we won't see it.

>
>> [   30.811750] CR2: fffffffffdeb2f98 CR3: 0000000423fae001 CR4: 00000000001607e0
>> [   30.819712] Call Trace:
>> [   30.822442]  <SYSENTER>
>> [   30.825170]  trace_hardirqs_on_thunk+0x1c/0x1c
> ...
>> [   31.000571] R13: 0000000000000050 R14: 0000000000000076 R15: 00007f59f76f2d60
>> [   31.008533]  </SYSENTER>
>
> Should we change that string to something more descriptive?

I suppose we could rename it to "ENTRY_TRAMPOLINE" or something like that.