Re: [dpdk-stable] [PATCH] net/ring: fix unchecked return value

[Kevin Laatz <kevin.laatz@intel.com>]

On 25/09/2020 13:43, Ferruh Yigit wrote:
> On 9/22/2020 6:20 PM, Kevin Laatz wrote:
>> Add a check for the return value of the sscanf call in
>> parse_internal_args(), returning an error if we don't get the expected
>> result.
>>
>> Coverity issue: 362049
>> Fixes: 96cb19521147 ("net/ring: use EAL APIs in PMD specific API")
>> Cc: stable@dpdk.org
>>
>> Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
>> ---
>>   drivers/net/ring/rte_eth_ring.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ring/rte_eth_ring.c 
>> b/drivers/net/ring/rte_eth_ring.c
>> index 40fe1ca4ba..62060e46ce 100644
>> --- a/drivers/net/ring/rte_eth_ring.c
>> +++ b/drivers/net/ring/rte_eth_ring.c
>> @@ -539,7 +539,8 @@ parse_internal_args(const char *key __rte_unused, 
>> const char *value,
>>       struct ring_internal_args **internal_args = data;
>>       void *args;
>>   -    sscanf(value, "%p", &args);
>> +    if (sscanf(value, "%p", &args) != 1)
>> +        return -1;
>
> I am aware that this is just to fix the coverity error to check the 
> return value, BUT :)
>
> The internal args is mainly to pass the information get by API 
> ('rte_eth_from_ring()') to ring probe. And the main information to 
> pass is the ring.
> It may be possible to pass the ring_name only and eliminate internal 
> args completely, the driver already has a way to pass ring name:
> "nodeaction=name:node:ATTACH" devargs.
>
> If you have time, can you please check if it can be possible to fill 
> and pass the nodeaction devargs in 'rte_eth_from_rings()' and 
> eliminate the 'internal' devargs completely?
> If it works, as a bonus it will resolve above coverity issue by 
> removing it :)
>
Hi Ferruh,

It seems to be used for more than just that - after the parsing, the 
internal args are passed to do_eth_dev_ring_create() as multiple parameters.

- Kevin