Re: [PATCH v4 2/5] x86/alternative: add indirect call patching

[Juergen Gross]

On 14.11.23 15:18, Peter Zijlstra wrote:
> On Tue, Nov 14, 2023 at 02:47:15PM +0100, Borislav Petkov wrote:
> > On Tue, Nov 14, 2023 at 01:50:28PM +0100, Peter Zijlstra wrote:
> > > This loads the function target from the pv_ops table. We can't otherwise
> > > do this.
> >
> > On Tue, Nov 14, 2023 at 01:56:37PM +0100, Juergen Gross wrote:
> > > It is replacing an _indirect_ call with a _direct_ one, taking the
> > > call target from the pointer used by the indirect call.
> >
> > Then this is not just a ALT_FLAG_CALL. This is something special. The
> > flag definition needs a better name along with an explanation what it
> > does, perhaps best with an example from the final vmlinux - not from the
> > object file:
> >
> > call   *0x0(%rip)
> >
> > ==>
> >
> > call   *0x0
> >
> > where the offsets haven't been linked in yet.
>
> Well, a random absolute address isn't going to be any better or worse
> than 0. But perhaps adding the relocation as a comment helps?
>
>
>     ff 15 00 00 00 00       call   *0x0(%rip)  #  R_X86_64_PC32    pv_ops+0x4
> into:
>     e8 00 00 00 00 90	   call   +0          #  R_X86_64_PC32 *(pv_ops+0x04)
>
>
> > If this is going into the generic infrastructure, then it better be
> > explained properly so that other stuff can potentially use it too.
>
> ALT_FLAG_DEREFERENCE_INDIRECT_CALL ?

ALT_FLAG_MAKE_CALL_DIRECT ?

> I'm going to already raise my hand and say that's too long ;-)

I think the length of the name doesn't matter that much, as it is used only
in the patching code and via the ALT_CALL() macro (at least as long as Boris
doesn't ask me to change the macro name, too).


Juergen

Attachment: OpenPGP_0xB0DE9DD628BF132F.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature