[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210319225233.osdqvq27od5nu7wy@treble>
Date: Fri, 19 Mar 2021 17:52:33 -0500
From: Josh Poimboeuf <jpoimboe@...hat.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: x86@...nel.org, jgross@...e.com, mbenes@...e.com,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 14/14] objtool,x86: Rewrite retpoline thunk calls
On Fri, Mar 19, 2021 at 04:56:30PM +0100, Peter Zijlstra wrote:
> On Fri, Mar 19, 2021 at 10:30:26AM -0500, Josh Poimboeuf wrote:
> > On Fri, Mar 19, 2021 at 09:06:44AM +0100, Peter Zijlstra wrote:
> > > > Also doesn't the alternative code already insert nops?
> > >
> > > Problem is that the {call,jmp} *%\reg thing is not fixed length. They're
> > > 2 or 3 bytes depending on which register is picked.
> >
> > Why do they need to be fixed length? Objtool can use sym->len as the
> > alternative replacement length. Then alternatives can add nops as
> > needed.
>
> UNDEF has size 0. That is, unless these symbols exist in the translation
> unit (they do not) we have no clue.
>
> Arguably I could parse the symbol name again and then we know the
> register number and thus if we need REX etc.. but I figured we wanted to
> avoid all that.
Ah, makes sense now.
--
Josh
Powered by blists - more mailing lists