From: Peter Zijlstra
> Sent: 18 March 2022 18:02
>
> On Fri, Mar 18, 2022 at 06:47:32PM +0100, Peter Zijlstra wrote:
> > On Fri, Mar 18, 2022 at 06:28:37PM +0100, Peter Zijlstra wrote:
> > > > Related to this, I don't see anything in arch/x86/kernel/static_call.c that
> > > > limits this code to x86-64:
> > > >
> > > > if (func == &__static_call_return0) {
> > > > emulate = code;
> > > > code = &xor5rax;
> > > > }
> > > >
> > > >
> > > > On 32-bit, it will be patched as "dec ax; xor eax, eax" or something like
> > > > that. Fortunately it doesn't corrupt any callee-save register but it is not
> > > > just a bit funky, it's also not a single instruction.
> > >
> > > Urggghh.. that's fairly yuck. So there's two options I suppose:
> > >
> > > 0x66, 0x66, 0x66, 0x31, 0xc0
> >
> > Argh, that turns into: xorw %ax, %ax.
> >
> > Let me see if there's another option.
>
> Amazingly:
>
> 0x2e, 0x2e, 0x2e, 0x31, 0xc0
>
> seems to actually work.. I've build and ran and decoded the below on
> 32bit and 64bit (arguably on the same 64bit host).
Not really amazing...
In 64bit mode all accesses to 32bit registers zero the
high bits.
So 'xor %eax,%eax' zeros all of %rax in 64bit mode.
So three segment override prefixes will extend it to 5 bytes.
Think I'd pick the FS or GS override (0x64 or 0x65).
Just in case someone decides that CS/DS/ES/SS prefix will
mean something else in 64bit mode.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
