diff mbox series

[v3] ci: disallow directional formatting

Message ID pull.1071.v3.git.1636031609982.gitgitgadget@gmail.com (mailing list archive)
State Accepted
Commit bc756bc586a67619ec1cf44c753f6744b61f27ca
Headers show
Series [v3] ci: disallow directional formatting | expand

Commit Message

Johannes Schindelin Nov. 4, 2021, 1:13 p.m. UTC
From: Johannes Schindelin <johannes.schindelin@gmx.de>

As described in https://trojansource.codes/trojan-source.pdf, it is
possible to abuse directional formatting (a feature of Unicode) to
deceive human readers into interpreting code differently from compilers.

For example, an "if ()" expression could be enclosed in a comment, but
rendered as if it was outside of that comment. In effect, this could
fool a reviewer into misinterpreting the code flow as benign when it is
not.

It is highly unlikely that Git's source code wants to contain such
directional formatting in the first place, so let's just disallow it.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
    ci: disallow directional formatting
    
    A couple days ago, I stumbled over
    https://siliconangle.com/2021/11/01/trojan-source-technique-can-inject-malware-source-code-without-detection/,
    which details an interesting social-engineering attack: it uses
    directional formatting in source code to pretend to human readers that
    the code does something different than it actually does.
    
    It is highly unlikely that Git's source code wants to contain such
    directional formatting in the first place, so let's disallow it.
    
    Technically, this is not exactly -rc material, but the paper was just
    published, and I want us to be safe.
    
    Changes since v2:
    
     * The pathspec excluding binary files is now used directly instead of
       doing the ls-files | xargs dance.
     * Corrected a code comment: my custom git grep was not PCRE-enabled,
       but Ubuntu's isn't. But git grep -P still does not understand \uNNNN.
     * Even if the *.po files currently pass the check, the script is now
       future-proof by excluding them.
     * Renamed the script to have the .bash extension, to indicate that it
       requires a Bashism (i.e. a printf that understands the \uNNNN
       syntax).
    
    Changes since v1:
    
     * The code was moved into a script in ci/.
     * We use git ls-files now to exclude files marked as binary in the Git
       attributes.

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1071%2Fdscho%2Fcheck-for-utf-8-directional-formatting-v3
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1071/dscho/check-for-utf-8-directional-formatting-v3
Pull-Request: https://github.com/gitgitgadget/git/pull/1071

Range-diff vs v2:

 1:  bbf963695ba ! 1:  80447819de3 ci: disallow directional formatting
     @@ .github/workflows/main.yml: jobs:
           - uses: actions/checkout@v2
           - run: ci/install-dependencies.sh
           - run: ci/run-static-analysis.sh
     -+    - run: ci/check-directional-formatting.sh
     ++    - run: ci/check-directional-formatting.bash
         sparse:
           needs: ci-config
           if: needs.ci-config.outputs.enabled == 'yes'
      
     - ## ci/check-directional-formatting.sh (new) ##
     + ## ci/check-directional-formatting.bash (new) ##
      @@
      +#!/bin/bash
      +
     @@ ci/check-directional-formatting.sh (new)
      +# to deceive reviewers into interpreting code differently from the compiler.
      +# This is intended to run on an Ubuntu agent in a GitHub workflow.
      +#
     -+# `git grep` as well as GNU grep do not handle `\u` as a way to specify UTF-8.
     -+# A PCRE-enabled `git grep` would handle `\u` as desired, but Ubuntu does
     -+# not build its `git` packages with PCRE support.
     ++# To allow translated messages to introduce such directional formatting in the
     ++# future, we exclude the `.po` files from this validation.
     ++#
     ++# Neither GNU grep nor `git grep` (not even with `-P`) handle `\u` as a way to
     ++# specify UTF-8.
      +#
      +# To work around that, we use `printf` to produce the pattern as a byte
      +# sequence, and then feed that to `git grep` as a byte sequence (setting
     @@ ci/check-directional-formatting.sh (new)
      +# U+2066..U+2069: LRI, RLI, FSI and PDI
      +regex='(\u202a|\u202b|\u202c|\u202d|\u202e|\u2066|\u2067|\u2068|\u2069)'
      +
     -+! git ls-files -z ':(attr:!binary)' |
     -+LC_CTYPE=C xargs -0r git grep -Ele "$(LC_CTYPE=C.UTF-8 printf "$regex")" --
     ++! LC_CTYPE=C git grep -El "$(LC_CTYPE=C.UTF-8 printf "$regex")" \
     ++	-- ':(exclude,attr:binary)' ':(exclude)*.po'


 .github/workflows/main.yml           |  1 +
 ci/check-directional-formatting.bash | 27 +++++++++++++++++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100755 ci/check-directional-formatting.bash


base-commit: 0cddd84c9f3e9c3d793ec93034ef679335f35e49

Comments

Ævar Arnfjörð Bjarmason Nov. 4, 2021, 1:48 p.m. UTC | #1
On Thu, Nov 04 2021, Johannes Schindelin via GitGitGadget wrote:

> From: Johannes Schindelin <johannes.schindelin@gmx.de>
> [...]
>      * Corrected a code comment: my custom git grep was not PCRE-enabled,
>        but Ubuntu's isn't. But git grep -P still does not understand \uNNNN.
>      * Even if the *.po files currently pass the check, the script is now
>        future-proof by excluding them.
> [...]

> +# This is intended to run on an Ubuntu agent in a GitHub workflow.
> +#
> +# To allow translated messages to introduce such directional formatting in the
> +# future, we exclude the `.po` files from this validation.
> +#
> +# Neither GNU grep nor `git grep` (not even with `-P`) handle `\u` as a way to
> +# specify UTF-8.
> +#
> +# To work around that, we use `printf` to produce the pattern as a byte
> +# sequence, and then feed that to `git grep` as a byte sequence (setting
> +# `LC_CTYPE` to make sure that the arguments are interpreted as intended).
> +#
> +# Note: we need to use Bash here because its `printf` interprets `\uNNNN` as
> +# UTF-8 code points, as desired. Running this script through Ubuntu's `dash`,
> +# for example, would use a `printf` that does not understand that syntax.
> +
> +# U+202a..U+2a2e: LRE, RLE, PDF, LRO and RLO
> +# U+2066..U+2069: LRI, RLI, FSI and PDI
> +regex='(\u202a|\u202b|\u202c|\u202d|\u202e|\u2066|\u2067|\u2068|\u2069)'
> +
> +! LC_CTYPE=C git grep -El "$(LC_CTYPE=C.UTF-8 printf "$regex")" \
> +	-- ':(exclude,attr:binary)' ':(exclude)*.po'
>
> base-commit: 0cddd84c9f3e9c3d793ec93034ef679335f35e49

So this is still using not-PCRE instead of the much simpler PCRE-enabled
one I suggested in[1] because your locally-built git doesn't link to
libpcre?

At the very least that comment is still quite confusing. I.e. it starts
out by saying that it's meant to run in CI where we've got PCRE, but
then goes on to describe an elaborate workaround that's only needed with
a not-PCRE grep.

Would be less confusing to understand why it's so complex as:

   # This is intended to run on an Ubuntu agent in a GitHub
   # workflow. We've got PCRE there, so all of the below could also be:
   #
   #    ! git -P grep -nP '[\N{U+202a}-\N{U+202e}\N{U+2066}-\N{U+2069}]' ':!(attr:binary)' ':!**po/**'
   #
   #
   # But the author wanted to run this locally on a system that didn't
   # have PCRE, So [go on to describe elaborate bash / "git grep -E" /
   # LC_* tweaking workaround....]
   [...]

B.t.w. I think that **po/** exclusion is closer to what you want,
i.e. "a 'po' dir anywhere in our tree". It'll also exclude this if we
e.g. end up using these in language-specific README files there or
whatever.

1. https://lore.kernel.org/git/211103.86zgqlhzvz.gmgdl@evledraar.gmail.com/ or:

   ! git -P grep -nP '[\N{U+202a}-\N{U+202e}\N{U+2066}-\N{U+2069}]' ':!(attr:binary)'
Junio C Hamano Nov. 4, 2021, 5:19 p.m. UTC | #2
Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:

>> +# U+202a..U+2a2e: LRE, RLE, PDF, LRO and RLO
>> +# U+2066..U+2069: LRI, RLI, FSI and PDI
>> +regex='(\u202a|\u202b|\u202c|\u202d|\u202e|\u2066|\u2067|\u2068|\u2069)'
>> +
>> +! LC_CTYPE=C git grep -El "$(LC_CTYPE=C.UTF-8 printf "$regex")" \
>> +	-- ':(exclude,attr:binary)' ':(exclude)*.po'
> ...
>    ! git -P grep -nP '[\N{U+202a}-\N{U+202e}\N{U+2066}-\N{U+2069}]' ':!(attr:binary)'

So you are comparing

 * requiring bash and C.UTF-8 locale to be available

vs

 * requiring git built with PCRE

assuming that "Dscho says doesn't work with PCRE and you say it
works with PCRE" is resolved?  They seem roughly the same
difficulty to me.
Ævar Arnfjörð Bjarmason Nov. 8, 2021, 6:49 p.m. UTC | #3
On Thu, Nov 04 2021, Junio C Hamano wrote:

> Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:
>
>>> +# U+202a..U+2a2e: LRE, RLE, PDF, LRO and RLO
>>> +# U+2066..U+2069: LRI, RLI, FSI and PDI
>>> +regex='(\u202a|\u202b|\u202c|\u202d|\u202e|\u2066|\u2067|\u2068|\u2069)'
>>> +
>>> +! LC_CTYPE=C git grep -El "$(LC_CTYPE=C.UTF-8 printf "$regex")" \
>>> +	-- ':(exclude,attr:binary)' ':(exclude)*.po'
>> ...
>>    ! git -P grep -nP '[\N{U+202a}-\N{U+202e}\N{U+2066}-\N{U+2069}]' ':!(attr:binary)'
>
> So you are comparing
>
>  * requiring bash and C.UTF-8 locale to be available
>
> vs
>
>  * requiring git built with PCRE
>
> assuming that "Dscho says doesn't work with PCRE and you say it
> works with PCRE" is resolved?  They seem roughly the same
> difficulty to me.

We can hard depend on a git built with PCRE, since the point of this
thing is to run in GitHub CI, Ubuntu builds git with PCRE, and that's
unlikely to ever change.

The caveats around PCRE that still somewhat persist around PCRE are due
to a misunderstanding, i.e. I think Johannes tried the \uXXXX syntax,
which we don't opt git-grep into, but as shown above you can just use
the other universally supported PCRE syntax for referring to Unicode
codepoints.
Junio C Hamano Nov. 8, 2021, 8:08 p.m. UTC | #4
Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:

>> So you are comparing
>>
>>  * requiring bash and C.UTF-8 locale to be available
>>
>> vs
>>
>>  * requiring git built with PCRE
>>
>> assuming that "Dscho says doesn't work with PCRE and you say it
>> works with PCRE" is resolved?  They seem roughly the same
>> difficulty to me.
>
> We can hard depend on a git built with PCRE, since the point of this
> thing is to run in GitHub CI, Ubuntu builds git with PCRE, and that's
> unlikely to ever change.

Yes, so is the availability of bash and C.UTF-8 for the same reason:
we are talking about controlled environment.  That is what I meant
by "roughly the same difficulty to me".

FWIW, I am OK with either approach, as I find the patch in question
is just as readable as any rewrite that would use "grep -P", so...
Ævar Arnfjörð Bjarmason Nov. 9, 2021, 1:34 p.m. UTC | #5
On Mon, Nov 08 2021, Junio C Hamano wrote:

> Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes:
>
>>> So you are comparing
>>>
>>>  * requiring bash and C.UTF-8 locale to be available
>>>
>>> vs
>>>
>>>  * requiring git built with PCRE
>>>
>>> assuming that "Dscho says doesn't work with PCRE and you say it
>>> works with PCRE" is resolved?  They seem roughly the same
>>> difficulty to me.
>>
>> We can hard depend on a git built with PCRE, since the point of this
>> thing is to run in GitHub CI, Ubuntu builds git with PCRE, and that's
>> unlikely to ever change.
>
> Yes, so is the availability of bash and C.UTF-8 for the same reason:
> we are talking about controlled environment.  That is what I meant
> by "roughly the same difficulty to me".
>
> FWIW, I am OK with either approach, as I find the patch in question
> is just as readable as any rewrite that would use "grep -P", so...

To each his own I guess :) I do find the simple regex of:

    '[\N{U+202a}-\N{U+202e}\N{U+2066}-\N{U+2069}]'

Much easier to understand than something using printf, shell
interpolation, and needing to switch around LC_CTYPE to two different
values twice on one line.

But anyway, that's a matter of taste.

What isn't is the issue I noted at the bottom of [1], i.e. if we're
relying on '(attr:binary)' we should probably start with an assertion
that our idea of "binary" matches reality, or perhaps go back to the -I
heuristic.

Because now we've got at least one binary non-attribute-marked file, and
if files like that ever get updated they might start matching this
pattern. Maybe not a big deal, but someone updating those might
confusingly trip over this otherwise...

1. https://lore.kernel.org/git/211103.86zgqlhzvz.gmgdl@evledraar.gmail.com/
diff mbox series

Patch

diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
index 6ed6a9e8076..deda12db3a9 100644
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -289,6 +289,7 @@  jobs:
     - uses: actions/checkout@v2
     - run: ci/install-dependencies.sh
     - run: ci/run-static-analysis.sh
+    - run: ci/check-directional-formatting.bash
   sparse:
     needs: ci-config
     if: needs.ci-config.outputs.enabled == 'yes'
diff --git a/ci/check-directional-formatting.bash b/ci/check-directional-formatting.bash
new file mode 100755
index 00000000000..e6211b141aa
--- /dev/null
+++ b/ci/check-directional-formatting.bash
@@ -0,0 +1,27 @@ 
+#!/bin/bash
+
+# This script verifies that the non-binary files tracked in the Git index do
+# not contain any Unicode directional formatting: such formatting could be used
+# to deceive reviewers into interpreting code differently from the compiler.
+# This is intended to run on an Ubuntu agent in a GitHub workflow.
+#
+# To allow translated messages to introduce such directional formatting in the
+# future, we exclude the `.po` files from this validation.
+#
+# Neither GNU grep nor `git grep` (not even with `-P`) handle `\u` as a way to
+# specify UTF-8.
+#
+# To work around that, we use `printf` to produce the pattern as a byte
+# sequence, and then feed that to `git grep` as a byte sequence (setting
+# `LC_CTYPE` to make sure that the arguments are interpreted as intended).
+#
+# Note: we need to use Bash here because its `printf` interprets `\uNNNN` as
+# UTF-8 code points, as desired. Running this script through Ubuntu's `dash`,
+# for example, would use a `printf` that does not understand that syntax.
+
+# U+202a..U+2a2e: LRE, RLE, PDF, LRO and RLO
+# U+2066..U+2069: LRI, RLI, FSI and PDI
+regex='(\u202a|\u202b|\u202c|\u202d|\u202e|\u2066|\u2067|\u2068|\u2069)'
+
+! LC_CTYPE=C git grep -El "$(LC_CTYPE=C.UTF-8 printf "$regex")" \
+	-- ':(exclude,attr:binary)' ':(exclude)*.po'