On Mon, Jun 29, 2020 at 01:04:31PM -0700, Kees Cook wrote:
> On Mon, Jun 29, 2020 at 12:53:47PM -0700, Nick Desaulniers wrote:
> > On Sun, Jun 28, 2020 at 11:18 PM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> > >
> > > Fix a case of needless quotes in __section(), which Clang doesn't like.
> > >
> > > Acked-by: Will Deacon <will@xxxxxxxxxx>
> > > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> >
> > Yep, I remember bugs from this. Probably should scan the kernel for
> > other instances of this. +Joe for checkpatch.pl validation.
>
> I think the others are safe because they're in macros:
Why does that make it safe -- the commit msg is a bit sparse, but I
assume the problem is that it generates
__attribute__((__section__("\".foo\"")))
from
__section(".foo")
after preprocessing, and clang keeps the quotes in the section name when
generating assembly, while gcc appears to strip them off.
It does that even if nested in another macro, no?
>
> $ git grep -4 '__section("'
> include/linux/compiler.h-# define KENTRY(sym) \
Am I missing something, or is KENTRY unused in the tree?
> include/linux/compiler.h- extern typeof(sym) sym; \
> include/linux/compiler.h- static const unsigned long __kentry_##sym \
> include/linux/compiler.h- __used \
> include/linux/compiler.h: __section("___kentry" "+" #sym ) \
> include/linux/compiler.h- = (unsigned long)&sym;
> --
> include/linux/export.h-#define __ksym_marker(sym) \
> include/linux/export.h: static int __ksym_marker_##sym[0] __section(".discard.ksym") __used
> --
> include/linux/srcutree.h-# define __DEFINE_SRCU(name, is_static) \
> include/linux/srcutree.h- is_static struct srcu_struct name; \
> include/linux/srcutree.h- struct srcu_struct * const __srcu_struct_##name \
> include/linux/srcutree.h: __section("___srcu_struct_ptrs") = &name
>
>
> > Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
>
> Thanks!
>
> --
> Kees Cook
