Re: [PATCH v2] MODSIGN: Only sign modules if built in-tree

[Rusty Russell <rusty@rustcorp.com.au>]

Josh Boyer <jwboyer@redhat.com> writes:
> On Mon, Nov 05, 2012 at 12:31:39PM +1030, Rusty Russell wrote:
>> Josh Boyer <jwboyer@redhat.com> writes:
>> > diff --git a/Makefile b/Makefile
>> > index 14c93b3..7e27d51 100644
>> > --- a/Makefile
>> > +++ b/Makefile
>> > @@ -722,8 +722,14 @@ export mod_strip_cmd
>> >  ifeq ($(CONFIG_MODULE_SIG),y)
>> >  MODSECKEY = ./signing_key.priv
>> >  MODPUBKEY = ./signing_key.x509
>> > +ifeq ($(KBUILD_EXTMOD),)
>> > +SIGNFAIL = false
>> > +else
>> > +# External builds might not have a signing key, don't break module_install.
>> > +SIGNFAIL = true
>> > +endif # KBUILD_EXTMOD
>> >  export MODPUBKEY
>> > -mod_sign_cmd = perl $(srctree)/scripts/sign-file $(MODSECKEY) $(MODPUBKEY)
>> > +mod_sign_cmd = $(SIGNFAIL) || perl $(srctree)/scripts/sign-file $(MODSECKEY) $(MODPUBKEY)
>> >  else
>> >  mod_sign_cmd = true
>> >  endif
>> 
>> Huh?  'true || perl' never runs perl due to short circuiting.
>
> Ah.  Maybe you were going for "sign all modules if keys are available,
> but don't break external if they aren't" semantics.  I was just skipping
> it entirely for external modules.

Exactly.  This way you get warnings, not failure.  You probably want
that, since you'll (at least) taint your kernel when you load those
modules.

I've applied this in my fixes branch, will push to Linus later this
week.

Thanks,
Rusty.