[Libva] [PATCH V1][libva] fix check NULL issues reported by Klockwork

[Chen, Peng C]

-----Original Message-----
From: Zhao, Yakui 
Sent: Monday, May 23, 2016 4:06 PM
To: Chen, Peng C <peng.c.chen at intel.com>
Cc: libva at lists.freedesktop.org
Subject: Re: [Libva] [PATCH V1][libva] fix check NULL issues reported by Klockwork

On 05/23/2016 03:02 PM, Chen, Peng C wrote:
>
>
> -----Original Message-----
> From: Zhao, Yakui
> Sent: Monday, May 23, 2016 2:51 PM
> To: Chen, Peng C<peng.c.chen at intel.com>
> Cc: libva at lists.freedesktop.org
> Subject: Re: [Libva] [PATCH V1][libva] fix check NULL issues reported 
> by Klockwork
>
> On 05/23/2016 01:46 PM, peng.chen wrote:
>> Signed-off-by: peng.chen<peng.c.chen at intel.com>
>> ---
>>    va/va_trace.c | 16 ++++++++++++++--
>>    1 file changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/va/va_trace.c b/va/va_trace.c index 994422e..398ca1a
>> 100644
>> --- a/va/va_trace.c
>> +++ b/va/va_trace.c
>> @@ -501,7 +501,10 @@ static void FILE_NAME_SUFFIX(
>>        int tmp = strnlen(env_value, max_size);
>>        int left = max_size - tmp;
>>        struct timeval tv;
>> -    int size = strlen(suffix_str);
>> +    int size = 0;
>> +
>> +    if(suffix_str)
>> +        strlen(suffix_str);
>>
>>        if(left<   (size + 8 + 10))
>>            return;
>> @@ -735,8 +738,14 @@ void va_TraceInit(VADisplay dpy)
>>        struct va_trace *pva_trace = calloc(sizeof(struct va_trace), 1);
>>        struct trace_context *trace_ctx = calloc(sizeof(struct 
>> trace_context), 1);
>>
>> -    if (pva_trace == NULL || trace_ctx == NULL)
>> +    if (pva_trace == NULL || trace_ctx == NULL) {
>> +        if(pva_trace)
>> +            free(pva_trace);
>> +
>> +        if(trace_ctx)
>> +            free(trace_ctx);
>>            return;
>> +    }
>>
>>        if (va_parseConfig("LIBVA_TRACE",&env_value[0]) == 0) {
>>            pva_trace->fn_log_env = strdup(env_value); @@ -1474,6
>> +1483,9 @@ void va_TraceCreateBuffer (
>>        VABufferID *buf_id		/* out */
>>    )
>>    {
>> +    if(!buf_id)
>> +        return;
>
> Maybe the zero buf_id is the valid buffer id.
> The check should be changed as the following:
>          if (buf_id == VA_INVALID_ID)
>              return
> [Peng] here it means the pointer, not the value itself.

[Yakui]: Although it is the pointer, we still can check whether it is valid before trying to trace the corresponding info.  In such case it can avoid the tracing of invalid buffer.
[peng]:yes, it is important. I will double check whether invalid buffer id is rightly handled. Thanks!

>
>> +
>>        DPY2TRACECTX(dpy, context, VA_INVALID_ID);
>>
>>        add_trace_buf_info(pva_trace, context, *buf_id);
>