Re: [PATCH 3/4] x86: open-code register save/restore in trace_hardirqs thunks

[Andy Lutomirski]

On Sat, Jan 10, 2015 at 2:00 PM, Borislav Petkov <bp@xxxxxxxxx> wrote:
> On Sat, Jan 10, 2015 at 09:14:03PM +0100, Denys Vlasenko wrote:
>> From 2f636e0a92db898f2bdb592027aa302fcb32a326 Mon Sep 17 00:00:00 2001
>> From: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
>> To: linux-kernel@xxxxxxxxxxxxxxx
>> Subject: [PATCH 3/4] x86: open-code register save/restore in trace_hardirqs thunks
>>
>> This is a preparatory patch for change in "struct pt_regs"
>> handling in entry_64.S.
>>
>> trace_hardirqs thunks were (ab)using a part of pt_regs
>> handling code, namely SAVE_ARGS/RESTORE_ARGS macros,
>> to save/restore registers across C function calls.
>>
>> Since SAVE_ARGS is going to be changed, open-code
>> register saving/restoring here. Take a page from thunk_32.S
>> and use push/pop insns instead of movq, they are far shorter:
>> 1 or 2 bytes versus 5, and no need for insns to adjust %rsp:
>>
>>    text          data     bss     dec     hex filename
>>     333            40       0     373     175 thunk_64_movq.o
>>     104            40       0     144      90 thunk_64_push_pop.o
>>
>> Incidentally, this removes a bit of dead code:
>> one SAVE_ARGS was used just to emit a CFI annotation,
>> but it also generated unreachable assembly insns.
>>
>> Signed-off-by: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
>> CC: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
>> CC: Oleg Nesterov <oleg@xxxxxxxxxx>
>> CC: "H. Peter Anvin" <hpa@xxxxxxxxx>
>> CC: Borislav Petkov <bp@xxxxxxxxx>
>> CC: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
>> CC: Frederic Weisbecker <fweisbec@xxxxxxxxx>
>> CC: X86 ML <x86@xxxxxxxxxx>
>> CC: Alexei Starovoitov <ast@xxxxxxxxxxxx>
>> CC: Will Drewry <wad@xxxxxxxxxxxx>
>> CC: Kees Cook <keescook@xxxxxxxxxxxx>
>> CC: linux-kernel@xxxxxxxxxxxxxxx
>> ---
>>  arch/x86/lib/thunk_64.S | 46 ++++++++++++++++++++++++++++++++++++++++++----
>>  1 file changed, 42 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/lib/thunk_64.S b/arch/x86/lib/thunk_64.S
>> index b30b5eb..8ec443a 100644
>> --- a/arch/x86/lib/thunk_64.S
>> +++ b/arch/x86/lib/thunk_64.S
>> @@ -17,9 +17,27 @@
>>       CFI_STARTPROC
>>
>>       /* this one pushes 9 elems, the next one would be %rIP */
>> -     SAVE_ARGS
>> +     pushq_cfi %rdi
>> +     CFI_REL_OFFSET rdi, 0
>
> Btw, why the second CFI annotation?
>
> pushq_cfi does already CFI_ADJUST_CFA_OFFSET 8. Can't we use one and
> hide it in the macro?

By my imperfect understanding of CFI:

CFI_ADJUST_CFA_OFFSET means that the offset between rsp and the
"canonical frame address" is increased by 8 (because we just
subtracted 8 from rsp) and CFI_REL_OFFSET reg, 0 means that the
unwinder can find reg at offset 0 + (cfa offset here) from the CFA.

IOW, one is to keep the stack frame tracking consistent and the other
is to tell the unwinder about the register we just saved.

--Andy

>
> Btw, patch boots fine in the guest.
>
> --
> Regards/Gruss,
>     Boris.
>
> Sent from a fat crate under my desk. Formatting is fine.
> --



-- 
Andy Lutomirski
AMA Capital Management, LLC
--
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/