Re: [PATCH 3/7] crypto: testmgr: fix warning

[Jarod Wilson <jarod@redhat.com>]

On 10/19/09 9:58 AM, Jarod Wilson wrote:
> On 10/19/09 9:52 AM, Jiri Kosina wrote:
>> On Mon, 19 Oct 2009, Felipe Contreras wrote:
>>
>>> crypto/testmgr.c: In function ?test_cprng?:
>>> crypto/testmgr.c:1204: warning: ?err? may be used uninitialized in
>>> this function
>>>
>>> Signed-off-by: Felipe Contreras<felipe.contreras@gmail.com>
>>> ---
>>> crypto/testmgr.c | 2 +-
>>> 1 files changed, 1 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/crypto/testmgr.c b/crypto/testmgr.c
>>> index 6d5b746..1f2357b 100644
>>> --- a/crypto/testmgr.c
>>> +++ b/crypto/testmgr.c
>>> @@ -1201,7 +1201,7 @@ static int test_cprng(struct crypto_rng *tfm,
>>> struct cprng_testvec *template,
>>> unsigned int tcount)
>>> {
>>> const char *algo = crypto_tfm_alg_driver_name(crypto_rng_tfm(tfm));
>>> - int err, i, j, seedsize;
>>> + int err = 0, i, j, seedsize;
>>> u8 *seed;
>>> char result[32];
>>
>> As it is not obvious to me immediately why/whether tcount couldn't be
>> zero
>> (which would cause uninitialized use of 'err'), I am not merging this
>> through trivial tree. Herbert?
>
> I believe I'm the guilty party who wrote the code in question.
> Initializing err to 0 isn't correct. tcount should always be at least 1,
> if its 0, test_cprng has been called with invalid parameters. I believe
> err would best be initialized to -EINVAL, lest the caller think they
> were successful.

...and I need to re-read said code more carefully. tcount is 
desc->suite.cprng.count, which is ANSI_CPRNG_AES_TEST_VECTORS, which is 
#define'd to 6, and is the count of how many cprng test vectors there 
are. If someone changes that to 0, then I guess a retval of 0 would 
actually be correct, since no tests failed...

So yeah, I rescind my claim that initializing err to 0 is incorrect, I 
think that's just fine.

-- 
Jarod Wilson
jarod@redhat.com