linux-hardening - Re: [PATCH] kunit: memcpy: Split slow memcpy tests into MEMCPY_SLOW_KUNIT

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: <CABVgOSkeCHF3Nrcd6eWQ=kFaaTo19+hStYdZg1ngQm7dGyBxHA@mail.gmail.com>
Date:   Sat, 14 Jan 2023 12:31:51 +0800
From:   David Gow <davidgow@...gle.com>
To:     Kees Cook <keescook@...omium.org>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Guenter Roeck <linux@...ck-us.net>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Nathan Chancellor <nathan@...nel.org>,
        linux-hardening@...r.kernel.org, Vlastimil Babka <vbabka@...e.cz>,
        Daniel Latypov <dlatypov@...gle.com>,
        Josh Poimboeuf <jpoimboe@...nel.org>,
        Geert Uytterhoeven <geert+renesas@...der.be>,
        Miguel Ojeda <ojeda@...nel.org>,
        Isabella Basso <isabbasso@...eup.net>,
        Dan Williams <dan.j.williams@...el.com>,
        Rasmus Villemoes <linux@...musvillemoes.dk>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] kunit: memcpy: Split slow memcpy tests into MEMCPY_SLOW_KUNIT_TEST

On Sat, 14 Jan 2023 at 08:54, Kees Cook <keescook@...omium.org> wrote:
>
> Since the long memcpy tests may stall a system for tens of seconds
> in virtualized architecture environments, split those tests off under
> CONFIG_MEMCPY_SLOW_KUNIT_TEST so they can be separately disabled.
>
> Reported-by: Guenter Roeck <linux@...ck-us.net>
> Link: https://lore.kernel.org/lkml/20221226195206.GA2626419@roeck-us.net
> Reviewed-by: Nick Desaulniers <ndesaulniers@...gle.com>
> Reviewed-and-tested-by: Guenter Roeck <linux@...ck-us.net>
> Reviewed-by: David Gow <davidgow@...gle.com>
> Cc: Andrew Morton <akpm@...ux-foundation.org>
> Cc: Nathan Chancellor <nathan@...nel.org>
> Cc: linux-hardening@...r.kernel.org
> Signed-off-by: Kees Cook <keescook@...omium.org>
> ---

Thanks: this is okay as-is, but I left a few suggestions below, having
tried it out a bit more.

Daniel's comment about kunit_skip() aborting the test is the only real
fix, but depending on how easy you'd want to make turning these tests
off, there are a couple of possible tweaks to the Kconfig option.

Cheers,
-- David

> v2: fix tristate to bool
> v1: https://lore.kernel.org/lkml/20230107040203.never.112-kees@kernel.org
> ---
>  lib/Kconfig.debug  |  9 +++++++++
>  lib/memcpy_kunit.c | 15 ++++++++++++---
>  2 files changed, 21 insertions(+), 3 deletions(-)
>
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index c2c78d0e761c..f90637171453 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -2621,6 +2621,15 @@ config MEMCPY_KUNIT_TEST
>
>           If unsure, say N.
>
> +config MEMCPY_SLOW_KUNIT_TEST
> +       bool "Include exhaustive memcpy tests" if !KUNIT_ALL_TESTS

I think it'd be better to not include the "if !KUNIT_ALL_TESTS" here,
because it's very convenient to be able to use:
./tools/testing/kunit/kunit.py run memcpy --kconfig_add
CONFIG_MEMCPY_SLOW_KUNIT_TEST=n
to override it.

That does undermine the way KUNIT_ALL_TESTS works a bit, though it
depends if you want to consider this a new set of tests, or just an
option for the existing ones.

> +       depends on MEMCPY_KUNIT_TEST
> +       default KUNIT_ALL_TESTS

Does this default work for everyone? Personally, I think these tests
aren't slow enough that we'd want them disabled from an "all tests"
build by default. So I'd keep it as-is.

> +       help
> +         Some memcpy tests are quite exhaustive in checking for overlaps
> +         and bit ranges. These can be very slow, so they are split out
> +         as a separate config.
> +
>  config IS_SIGNED_TYPE_KUNIT_TEST
>         tristate "Test is_signed_type() macro" if !KUNIT_ALL_TESTS
>         depends on KUNIT
> diff --git a/lib/memcpy_kunit.c b/lib/memcpy_kunit.c
> index 89128551448d..5a545e1b5dbb 100644
> --- a/lib/memcpy_kunit.c
> +++ b/lib/memcpy_kunit.c
> @@ -307,8 +307,12 @@ static void set_random_nonzero(struct kunit *test, u8 *byte)
>         }
>  }
>
> -static void init_large(struct kunit *test)
> +static int init_large(struct kunit *test)
>  {
> +       if (!IS_ENABLED(CONFIG_MEMCPY_SLOW_KUNIT_TEST)) {
> +               kunit_skip(test, "Slow test skipped. Enable with CONFIG_MEMCPY_SLOW_KUNIT_TEST=y");
> +               return -EBUSY;

As Daniel notes, it shouldn't be necessary to return here, as
kunit_skip() will abort the test. (It's a macro, so isn't marked
__noreturn itself, but kunit_try_catch_throw(), which it uses, is.)

> +       }
>
>         /* Get many bit patterns. */
>         get_random_bytes(large_src, ARRAY_SIZE(large_src));
> @@ -319,6 +323,8 @@ static void init_large(struct kunit *test)
>
>         /* Explicitly zero the entire destination. */
>         memset(large_dst, 0, ARRAY_SIZE(large_dst));
> +
> +       return 0;
>  }
>
>  /*
> @@ -327,7 +333,9 @@ static void init_large(struct kunit *test)
>   */
>  static void copy_large_test(struct kunit *test, bool use_memmove)
>  {
> -       init_large(test);
> +
> +       if (init_large(test))
> +               return;
>
>         /* Copy a growing number of non-overlapping bytes ... */
>         for (int bytes = 1; bytes <= ARRAY_SIZE(large_src); bytes++) {
> @@ -472,7 +480,8 @@ static void memmove_overlap_test(struct kunit *test)
>         static const int bytes_start = 1;
>         static const int bytes_end = ARRAY_SIZE(large_src) + 1;
>
> -       init_large(test);
> +       if (init_large(test))
> +               return;
>
>         /* Copy a growing number of overlapping bytes ... */
>         for (int bytes = bytes_start; bytes < bytes_end;
> --
> 2.34.1
>

Download attachment "smime.p7s" of type "application/pkcs7-signature" (4003 bytes)