Re: [PATCH 4/5] x86/asm/entry/32: Replace RESTORE_RSI_RDI[_RDX] with open-coded 32-bit reads

[Andy Lutomirski]

On Tue, Jun 9, 2015 at 12:03 PM, Denys Vlasenko <dvlasenk@xxxxxxxxxx> wrote:
> On 06/09/2015 09:01 PM, Andy Lutomirski wrote:
>> On Tue, Jun 9, 2015 at 11:54 AM, Denys Vlasenko <dvlasenk@xxxxxxxxxx> wrote:
>>> This doesn't change much, but this uses shorter 32-bit insns:
>>>
>>>         -48 8b 74 24 68         mov    0x68(%rsp),%rsi
>>>         -48 8b 7c 24 70         mov    0x70(%rsp),%rdi
>>>         -48 8b 54 24 60         mov    0x60(%rsp),%rdx
>>>         +8b 74 24 68            mov    0x68(%rsp),%esi
>>>         +8b 7c 24 70            mov    0x70(%rsp),%edi
>>>         +8b 54 24 60            mov    0x60(%rsp),%edx
>>>
>>> Since these are the only uses of RESTORE_RSI_RDI[_RDX], drop these macros.
>>>
>>
>> It probably doesn't matter for these fast paths, but, for the full
>> slow path return, we really do need to restore the full pt_regs.
>> After all, the syscall we're returning from might be sigreturn.
>
> This is compat 32-bit syscall handling code.
> IIUC we do not restore high half of any registers for 32-bit tasks.
>
> Am I missing something?

Yes -- 64-bit tasks can call 32-bit compat syscalls.  In fact, we
should really excise the entire concept of "64-bit tasks" and "32-bit
tasks" from the kernel.  The things that have bitness are the current
syscall (TS_COMPAT), CS, the mm, and the signal context.  The task
should be agnostic.

--Andy
--
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/