linux-kernel - Re: [PATCH] kunit: cleanup assertion macro internal variables

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] [day] [month] [year] [list]

Message-ID: <CAGS_qxoSn16TgDUvQTUpamu1xzY85Cbqy7Qg94fyKwjE4jw7Fg@mail.gmail.com>
Date:   Fri, 28 Jan 2022 13:36:55 -0800
From:   Daniel Latypov <dlatypov@...gle.com>
To:     Brendan Higgins <brendanhiggins@...gle.com>
Cc:     davidgow@...gle.com, linux-kernel@...r.kernel.org,
        kunit-dev@...glegroups.com, linux-kselftest@...r.kernel.org,
        skhan@...uxfoundation.org
Subject: Re: [PATCH] kunit: cleanup assertion macro internal variables

On Fri, Jan 28, 2022 at 1:21 PM Brendan Higgins
<brendanhiggins@...gle.com> wrote:
>
> On Thu, Jan 27, 2022 at 4:52 PM Daniel Latypov <dlatypov@...gle.com> wrote:
> >
> > All the operands should be tagged `const`.
> > We're only assigning them to variables so that we can compare them (e.g.
> > check if left == right, etc.) and avoid evaluating expressions multiple
> > times.
> >
> > There's no need for them to be mutable.
>
> Agreed.
>
> > Also rename the helper variable `loc` to `__loc` like we do with
> > `__assertion` and `__strs` to avoid potential name collisions with user
> > code.
>
> Probably not necessary since we create a new code block (we are inside
> of an if-statement, do-while-loop, etc), but I don't really care
> either way.

You're totally right that this doesn't matter with our current macros.

given
int loc = 42;
KUNIT_EXPECT_TRUE(test, loc);
KUNIT_EXPECT_EQ(test, loc, 42);

becomes
do {
        if (__builtin_expect(!!(!(!!(loc) == !!true)), 0)) {
                /* we don't use the operands in here, so `loc` is fine */
                static const struct kunit_loc loc = {
                        .file = "lib/kunit/kunit-example-test.c", .line = 25
                };
...
do {
        typeof(loc) __left = (loc);
        typeof(42) __right = (42);
        do {
                /* We never reference the expression again, so `loc` is fine */
                if (__builtin_expect(!!(!(__left == __right)), 0)) {
                        static const struct kunit_loc loc = {
                                .file = "lib/kunit/kunit-example-test.c",
                                .line = 24
                        };

But reminder: this was *not* the case until very recently.
Sau we didn't have my earlier patch to move the `if(!(passed))` check
into the macro.
Then we'd have issues, e.g.
../lib/kunit/kunit-example-test.c: In function ‘example_simple_test’:
../include/kunit/test.h:828:26: error: wrong type argument to unary
exclamation mark
  828 |                         !!(condition) == !!expected_true,
                \
      |
...
../lib/kunit/kunit-example-test.c:25:9: note: in expansion of macro
‘KUNIT_EXPECT_TRUE’
   25 |         KUNIT_EXPECT_TRUE(test, loc);
      |         ^~~~~~~~~~~~~~~~~

So being defensive here lets us change up our implementation more freely.

>
> > Signed-off-by: Daniel Latypov <dlatypov@...gle.com>
>
> Reviewed-by: Brendan Higgins <brendanhiggins@...gle.com>