Re: [PATCH] arm64: lib: use C string functions with KASAN enabled.
From: Will Deacon
Date: Fri Sep 14 2018 - 11:27:47 EST
On Tue, Sep 11, 2018 at 04:01:28PM +0300, Andrey Ryabinin wrote:
> On 09/10/2018 04:06 PM, Will Deacon wrote:
> > On Mon, Sep 10, 2018 at 01:53:03PM +0100, Mark Rutland wrote:
> >> On Mon, Sep 10, 2018 at 12:33:22PM +0100, Mark Rutland wrote:
> >>> On Fri, Sep 07, 2018 at 06:48:10PM +0300, Andrey Ryabinin wrote:
> >>>> On 09/07/2018 05:56 PM, Will Deacon wrote:
> >>>>> I don't understand this bit: efistub uses the __pi_ prefixed
> >>>>> versions of the routines, so why do we need to declare them as weak?
> >>>> Weak needed because we can't have two non-weak functions with the same
> >>>> name.
> >>>> Alternative approach would be to never use e.g. "strlen" name for asm
> >>>> implementation of strlen() under CONFIG_KASAN=y. But that would
> >>>> require adding some special ENDPIPROC_KASAN() macro since we want
> >>>> __pi_strlen() to point to the asm_strlen().
> >>> Somehow, what we have today works with CONFIG_FORTIFY_SOURCE, which
> >>> AFAICT would suffer from texactly the same problem with things like
> >>> memcpy.
> FORTIFY_SOURCE seems uses "extern inline" to redefine functions.
> I obviously cannot make the whole lib/string.c 'extern inline'.
> >>> So either we're getting away with that by chance already (and should fix
> >>> that regardless of this patch), or this is not actually a problem.
> >> I now see those functions are marked weak in the assembly
> >> implementation; sorry for the noise.
> >> Regardless, I still think it's preferable to avoid weak wherever
> >> possible.
> > I was thinking along the same lines, but having played around with the code,
> > I agree with Andrey that this appears to be the cleanest solution.
> > Andrey -- could you respin using WEAK instead of .weak, removing any
> > redundant uses of ENTRY in the process? We might also need to throw an
> > ALIGN directive into the WEAK definition.
> Actually I come up with something that looks decent, without using weak symbols, see below.
> "#ifndef CONFIG_KASAN" could be moved to the header. In that ALIAS probably should be renamed to
> something like NOKASAN_ALIAS().
Hmm, to be honest, I'd kinda got used to the version using weak symbols
and I reckon it'd be cleaner still if you respin it using WEAK.