Re: [PATCH v4 14/17] x86/asm/entry/64: Migrate error and interrupt exit work to C

From: Denys Vlasenko
Date: Fri Jul 03 2015 - 10:37:56 EST


On Thu, Jul 2, 2015 at 6:09 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
> On Thu, Jul 2, 2015 at 5:09 AM, Borislav Petkov <bp@xxxxxxxxx> wrote:
>> On Mon, Jun 29, 2015 at 12:33:46PM -0700, Andy Lutomirski wrote:
>>> Signed-off-by: Andy Lutomirski <luto@xxxxxxxxxx>
>>> ---
>>> arch/x86/entry/entry_64.S | 63 +++++++++++-----------------------------
>>> arch/x86/entry/entry_64_compat.S | 5 ++++
>>> 2 files changed, 22 insertions(+), 46 deletions(-)
>>>
>>> diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
>>> index ce2c9049abef..08a37ec049f0 100644
>>> --- a/arch/x86/entry/entry_64.S
>>> +++ b/arch/x86/entry/entry_64.S
>>> @@ -508,7 +508,16 @@ END(irq_entries_start)
>>>
>>> testb $3, CS(%rsp)
>>> jz 1f
>>> +
>>> + /*
>>> + * IRQ from user mode. Switch to kernel gsbase and inform context
>>> + * tracking that we're in kernel mode.
>>> + */
>>> SWAPGS
>>> +#ifdef CONFIG_CONTEXT_TRACKING
>>> + call enter_from_user_mode
>>> +#endif
>>
>> I think you can make this much cleaner by getting rid of the ifdeffery
>> and pushing it into the enter_from_user_mode() function:
>>
>> __visible void enter_from_user_mode(void)
>> {
>> #ifdef CONFIG_CONTEXT_TRACKING
>>
>> ...
>>
>> #endif
>> }
>>
>> The disadvantage of all that cleanliness is that we get one dumb
>>
>> call enter_from_user_mode
>>
>> there to this abomination:
>>
>> ffffffff810014f0 <enter_from_user_mode>:
>> ffffffff810014f0: e8 db 97 67 00 callq ffffffff8167acd0 <__fentry__>
>> ffffffff810014f5: 55 push %rbp
>> ffffffff810014f6: 48 89 e5 mov %rsp,%rbp
>> ffffffff810014f9: 5d pop %rbp
>> ffffffff810014fa: c3 retq
>> ffffffff810014fb: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1)
>>
>> which sux.
>>
>> We sure could use LTO here.
>>
>
> You mean link-time asm optimizations? Turning off frame pointers in
> leaf functions as long as rbp is still preserved might not be so
> terrible either.
>
> I'm torn on this one. In principle, you're right, or we could have a
> macro CALL_ENTER_FROM_USER_MODE that does nothing if context tracking
> is off. OTOH, that's also kind of messy.

busybox has IF() macros:

IF_CONTEXT_TRACKING(call enter_from_user_mode)

Unlike #if, this needs one line, not three.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/