Re: [PATCH 00/33] Compile-time stack metadata validation

From: Ingo Molnar
Date: Wed Feb 24 2016 - 02:41:10 EST



* Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:

> Hi Ingo,
>
> On Tue, Feb 23, 2016 at 09:14:06AM +0100, Ingo Molnar wrote:
> > So I tried out this latest stacktool series and it looks mostly good for an
> > upstream merge.
> >
> > To help this effort move forward I've applied the preparatory/fix patches that are
> > part of this series to tip:x86/debug - that's 26 out of 31 patches. (I've
> > propagated all the acks that the latest submission got into the changelogs.)
>
> Thanks very much for your review and for applying the fixes!
>
> A few issues relating to the merge:
>
> - The tip:x86/debug branch fails to build because it depends on
> ec5186557abb ("x86/asm: Add C versions of frame pointer macros") which
> is in tip:x86/asm.

Indeed...

> - As Pavel mentioned, the tip-bot seems to be spitting out garbage
> emails from:
> =?UTF-8?B?dGlwLWJvdCBmb3IgSm9zaCBQb2ltYm9ldWYgPHRpcGJvdEB6eXRvci5jb20+?=@xxxxxxxxxx

Yeah, hpa fixed that meanwhile.

Due to the above bad base I rebased the tree (to a x86/asm base), so there will be
a new round of (hopefully readable) tip-bot notifications. I'll push it out after
a bit of testing.

> > 5)
> >
> > Likewise, I think the CONFIG_STACK_VALIDATION=y Kconfig flag does not express that
> > we do exception table checks as well - and it does not express all the other
> > things we may check in object files in the future.
> >
> > Something like CONFIG_CHECK_OBJECT_FILES=y would express it, and the help text
> > would list all the things the tool is able to checks for at the moment.
>
> Hm, I'm not really sure about this. Yes, the tool could potentially do
> other types of checks, but is it necessary to lump them all together
> into a single config option? It does have subcommands after all ;-)

lol ;-)

Ok, I'm fine with CONFIG_STACK_VALIDATION=y as well.

Thanks,

Ingo