Received-SPF: pass (google.com: domain of linux-kernel+bounces-95212-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-ID: <fd210d04-cc98-4e29-a051-bdb7e1367494@collabora.com>
Date: Thu, 7 Mar 2024 14:22:27 +0500
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Cc: Muhammad Usama Anjum <usama.anjum@collabora.com>,
 Eric Biederman <ebiederm@xmission.com>, Shuah Khan <shuah@kernel.org>,
 Mark Brown <broonie@kernel.org>, linux-mm@kvack.org,
 linux-kernel@vger.kernel.org, linux-kselftest@vger.kernel.org,
 "kernel@collabora.com" <kernel@collabora.com>
Subject: Re: [Test Failure Report] exec: Test failures in execveat
To: Kees Cook <keescook@chromium.org>
References: <02c8bf8e-1934-44ab-a886-e065b37366a7@collabora.com>
 <202403051256.7A50FE28E7@keescook>
Content-Language: en-US
From: Muhammad Usama Anjum <usama.anjum@collabora.com>
In-Reply-To: <202403051256.7A50FE28E7@keescook>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

On 3/6/24 2:00 AM, Kees Cook wrote:
> On Tue, Mar 05, 2024 at 07:20:27PM +0500, Muhammad Usama Anjum wrote:
>> Hello,
>>
>> I've been running execveat (execveat.c) locally on v6.1 and next-20240228.
>> It has flaky test case. There are some test cases which fail consistently.
>> The comment (not very clear) on top of failing cases is as following:
>>
>> /*
>>  * Execute as a long pathname relative to "/".  If this is a script,
>>  * the interpreter will launch but fail to open the script because its
>>  * name ("/dev/fd/5/xxx....") is bigger than PATH_MAX.
>>  *
>>  * The failure code is usually 127 (POSIX: "If a command is not found,
>>  * the exit status shall be 127."), but some systems give 126 (POSIX:
>>  * "If the command name is found, but it is not an executable utility,
>>  * the exit status shall be 126."), so allow either.
>>  */
>> The file name is just less than PATH_MAX (4096) and we are expecting the
>> execveat() to fail with particular 99 or 127/128 error code. But kernel is
>> returning 1 error code. Snippet from full output:
>>
>> # child 3493092 exited with 1 not 99 nor 99
>> # child 3493094 exited with 1 not 127 nor 126
>>
>> I'm not sure if test is wrong or the kernel has changed the return error codes.
> 
> The error code is actually coming from the script interpreter (in this
> case, "/bin/sh"). On my system, /bin/sh is /bin/dash, and I see the
> failure. If I manually change "script" to use "#!/bin/bash", the test
> passes for me.
> 
> Since lots of other selftests appears to depend on /bin/bash, I think
> the right fix is simply:
> 
> 
> diff --git a/tools/testing/selftests/exec/execveat.c b/tools/testing/selftests/exec/execveat.c
> index bf79d664c8e6..0546ca24f2b2 100644
> --- a/tools/testing/selftests/exec/execveat.c
> +++ b/tools/testing/selftests/exec/execveat.c
> @@ -393,7 +393,7 @@ static int run_tests(void)
>  static void prerequisites(void)
>  {
>  	int fd;
> -	const char *script = "#!/bin/sh\nexit $*\n";
> +	const char *script = "#!/bin/bash\nexit $*\n";
>  
>  	/* Create ephemeral copies of files */
>  	exe_cp("execveat", "execveat.ephemeral");
> 
> 
> Can you test this and let me know if this fixes it for you?
I've tested this patch. Still getting same failures.

> 
> Thanks for the report!
> 
> -Kees
> 

-- 
BR,
Muhammad Usama Anjum