linux-kernel - Re: [PATCH v4 14/18] static_call: Add static_cond

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 5 May 2020 11:48:29 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Nick Desaulniers <ndesaulniers@...gle.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Rasmus Villemoes <linux@...musvillemoes.dk>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Jason Baron <jbaron@...mai.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...nel.org>, Nadav Amit <namit@...are.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Andy Lutomirski <luto@...nel.org>,
        Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
        "H.J. Lu" <hjl.tools@...il.com>,
        clang-built-linux <clang-built-linux@...glegroups.com>
Subject: Re: [PATCH v4 14/18] static_call: Add static_cond_call()

On Tue, May 5, 2020 at 11:28 AM Nick Desaulniers
<ndesaulniers@...gle.com> wrote:
>
> Changing
> void *func = READ_ONCE(name.func); \
> to
> void *func = &READ_ONCE(name.func); \

What? That makes no sense.

Yes,

    void *func = foo;

and

   void *func = &foo;

are the same thing, _if_ "foo" is an actual function, because then
"foo" degrades from a function to a pointer to a function as part of C
type semantics.

But that's not the case here. READ_ONCE(name.func) isn't a function -
it's a pointer to a function. So it doesn't degrade to anything at
all, and adding a '&' in front ot it completely changes the meaning of
the expression. So now the '&' changes it from "pointer to a function"
to "pointer to a pointer to a function", and the end result is not the
same thing any more.

Without the "&" it will call the function "bar" (which is the function
pointer that was assigned).

With the "&", it will not not call a function at all, it will do a
call to an address that is actually data of type "struct
static_call_key".

That's also why the NULL pointer check goes away: now the pointer is a
pointer to static data, which can never be NULL.

That said, I found it interesting that the volatile read also goes
away. That struck me as strange. But then I thought about it somem
more, and realized that the '&' basically just peels off the '*', so
now there isn't any actual volatile access any more, which is why the
read went away too.

Anyway, adding that '&' completely changes the meaning of the test.

Your initial reaction that "you can't compile away the read and the
test of NULL" was correct, I think.

                   Linus