Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.36 as permitted sender) client-ip=23.128.96.36;
Message-ID: <e1bc93b2-3b68-4675-b1b3-391bc1e5e5d8@redhat.com>
Date:   Tue, 17 Oct 2023 16:45:55 +0200
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] kunit: run test suites only after module initialization
 completes
To:     Jinjie Ruan <ruanjinjie@huawei.com>
Cc:     Brendan Higgins <brendan.higgins@linux.dev>,
        David Gow <davidgow@google.com>, Rae Moar <rmoar@google.com>,
        Shuah Khan <skhan@linuxfoundation.org>,
        linux-kselftest@vger.kernel.org, kunit-dev@googlegroups.com,
        linux-kernel@vger.kernel.org
References: <20231016203548.21993-1-marpagan@redhat.com>
 <8f08ef23-bbe1-bebb-eee5-e202c98dbb44@huawei.com>
Content-Language: en-US
From:   Marco Pagani <marpagan@redhat.com>
In-Reply-To: <8f08ef23-bbe1-bebb-eee5-e202c98dbb44@huawei.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 2023-10-17 03:28, Jinjie Ruan wrote:
> 
> 
> On 2023/10/17 4:35, Marco Pagani wrote:
>> Commit 2810c1e99867 ("kunit: Fix wild-memory-access bug in
>> kunit_free_suite_set()") is causing all test suites to run (when
>> built as modules) while still in MODULE_STATE_COMING. In that state,
>> test modules are not fully initialized and lack sysfs kobjects.
>> This behavior can cause a crash if the test module tries to register
>> fake devices.
>>
>> This patch restores the normal execution flow, waiting for the module
>> initialization to complete before running the test suites.
>> The issue reported in the commit mentioned above is addressed using
>> virt_addr_valid() to detect if the module loading has failed
>> and mod->kunit_suites has not been allocated using kmalloc_array().
>>
>> Fixes: 2810c1e99867 ("kunit: Fix wild-memory-access bug in kunit_free_suite_set()")
>> Signed-off-by: Marco Pagani <marpagan@redhat.com>
>> ---
>>  lib/kunit/test.c | 8 +++++---
>>  1 file changed, 5 insertions(+), 3 deletions(-)
>>
>> diff --git a/lib/kunit/test.c b/lib/kunit/test.c
>> index 421f13981412..1a49569186fc 100644
>> --- a/lib/kunit/test.c
>> +++ b/lib/kunit/test.c
>> @@ -769,12 +769,14 @@ static void kunit_module_exit(struct module *mod)
>>  	};
>>  	const char *action = kunit_action();
>>  
>> +	if (!suite_set.start || !virt_addr_valid(suite_set.start))
>> +		return;
>> +
>>  	if (!action)
>>  		__kunit_test_suites_exit(mod->kunit_suites,
>>  					 mod->num_kunit_suites);
> 
> If the module state is from MODULE_STATE_LIVE to MODULE_STATE_GOING, in
> kunit_module_init() the kunit_exec_run_tests() is executed when action
> is NULL whether kunit_filter_suites() succeeds or not.

If kunit_filter_suites() fails in kunit_module_init(), suite_set is
initialized to {0, 0}. Hence, kunit_exec_run_tests() will not execute
the test suites since num_suites = suite_set->end - suite_set->start
equals 0.

> But in kunit_module_exit() __kunit_test_suites_exit() will not be executed when
> action is NULL if kunit_filter_suites() fails.

If kunit_filter_suites() has previously failed in kunit_module_init(),then
kunit_module_exit() will return before calling __kunit_test_suites_exit()
since suite_set.start has previously been set to 0.

>
>>  
>> -	if (suite_set.start)
>> -		kunit_free_suite_set(suite_set);
>> +	kunit_free_suite_set(suite_set);
>>  }
>>  
>>  static int kunit_module_notify(struct notifier_block *nb, unsigned long val,
>> @@ -784,12 +786,12 @@ static int kunit_module_notify(struct notifier_block *nb, unsigned long val,
>>  
>>  	switch (val) {
>>  	case MODULE_STATE_LIVE:
>> +		kunit_module_init(mod);
>>  		break;
>>  	case MODULE_STATE_GOING:
>>  		kunit_module_exit(mod);
>>  		break;
>>  	case MODULE_STATE_COMING:
>> -		kunit_module_init(mod);
>>  		break;
>>  	case MODULE_STATE_UNFORMED:
>>  		break;
>