[PATCH v2 2/3] arm64: Enable BTI for main executable as well as the interpreter

[Dave Martin]

On Thu, Jun 10, 2021 at 02:19:05PM +0100, Mark Brown wrote:
> On Wed, Jun 09, 2021 at 04:17:13PM +0100, Dave Martin wrote:
> > On Fri, Jun 04, 2021 at 12:24:49PM +0100, Mark Brown wrote:
> 
> > > -		if (system_supports_bti() && has_interp == is_interp &&
> > > -		    (*p & GNU_PROPERTY_AARCH64_FEATURE_1_BTI))
> > > -			arch->flags |= ARM64_ELF_BTI;
> > > +		if (system_supports_bti() &&
> > > +		    (*p & GNU_PROPERTY_AARCH64_FEATURE_1_BTI)) {
> > > +			if (is_interp) {
> > > +				arch->flags |= ARM64_ELF_INTERP_BTI;
> > > +			} else {
> > > +				arch->flags |= ARM64_ELF_EXEC_BTI;
> > > +			}
> 
> > Nit: surplus curlies? (coding-style.rst does actually say to drop them
> > when all branches of an if are single-statement one-liners -- I had
> > presumed I was just being pedantic...)
> 
> I really think this hurts readability with the nested if inside
> another if with a multi-line condition.

So long as there is a reason rather than it being purely an accident of
editing, that's fine.

(Though if the nested if can be flattened so that this becomes a non-
issue, that's good too :)

> > > -	if (prot & PROT_EXEC)
> > > -		prot |= PROT_BTI;
> > > +		if (state->flags & ARM64_ELF_EXEC_BTI && !is_interp)
> > > +			prot |= PROT_BTI;
> > > +	}
> 
> > Is it worth adding () around the bitwise-& expressions?  I'm always a
> > little uneasy about the operator precedence of binary &, although
> > without looking it up I think you're correct.
> 
> Sure.  I'm fairly sure the compiler would've complained about
> this case if it were ambiguous, I'm vaguely surprised it didn't
> already.

I was vaguely surprised too -- though I didn't try to compile this
myself yet.  Anyway, not a huge deal.  Adding a helper to generate the
appropriate mask would make this issue go away in any case, but so long
as you're confident this is being evaluated as intended I can take your
word for it.

> > Feel free to adopt if this appeals to you, otherwise I'm also fine with
> > your version.)
> 
> I'll see what I think when I get back to looking at this
> properly.

Ack -- again, this was just a suggestion.  I can also live with your
original code if you ultimately decide to stick with that.

Cheers
---Dave