On 4/3/2019 5:27 PM, Thomas Monjalon wrote:
> 03/04/2019 18:07, Mohammad Abdul Awal:
>> Null value for parameter name will cause segfault for the
>> strnlen and strcmp functions.
>
> I'm not sure we want such obvious checks for all APIs.
> Here I would say yes.
These are internal functions, not APIs.
I am for verifying input for (all) APIs but not for internal functions, drivers
should call them and they are in our control, if they are passing NULL we can
fix them :)
>
>> Fixes: 0b33b68d12 ("ethdev: export allocate function")
>> Fixes: 942661004c ("ethdev: export secondary attach function")
>> Cc: sta...@dpdk.org
>>
>> Signed-off-by: Mohammad Abdul Awal <mohammad.abdul.a...@intel.com>
>> ---
>> --- a/lib/librte_ethdev/rte_ethdev.c
>> +++ b/lib/librte_ethdev/rte_ethdev.c
>> @@ -440,6 +440,11 @@ rte_eth_dev_allocate(const char *name)
>> struct rte_eth_dev *eth_dev = NULL;
>> size_t name_len;
>>
>> + if (name == NULL) {
>> + RTE_ETHDEV_LOG(ERR, "Null pointer is specified\n");
>
> This is a very generic error message.
> It might be "Fail to allocate port with empty name"
>
>> @@ -492,6 +497,11 @@ rte_eth_dev_attach_secondary(const char *name)
>> uint16_t i;
>> struct rte_eth_dev *eth_dev = NULL;
>>
>> + if (name == NULL) {
>> + RTE_ETHDEV_LOG(ERR, "Null pointer is specified\n");
>
> "Fail to attach port with empty name"
>
>
>
