Re: [dpdk-dev] [PATCH 5/5] sfc: don't use RTE_LOGTYPE_PMD

[Ferruh Yigit <ferruh.yigit@intel.com>]

On 2/27/2019 11:24 AM, Andrew Rybchenko wrote:
> On 2/27/19 2:21 PM, Ferruh Yigit wrote:
>> On 2/26/2019 9:34 PM, Stephen Hemminger wrote:
>>> The sfc driver was still using RTE_LOGTYPE_PMD which was superseded
>>> by local logging.
>>>
>>> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
>>> ---
>>>   drivers/net/sfc/sfc.c | 4 ++--
>>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/net/sfc/sfc.c b/drivers/net/sfc/sfc.c
>>> index 898603884fa0..2cd7126015fd 100644
>>> --- a/drivers/net/sfc/sfc.c
>>> +++ b/drivers/net/sfc/sfc.c
>>> @@ -1115,12 +1115,12 @@ sfc_register_logtype(const struct rte_pci_addr *pci_addr,
>>>   		++lt_prefix_str_size; /* Reserve space for prefix separator */
>>>   		lt_str_size_max = lt_prefix_str_size + PCI_PRI_STR_SIZE + 1;
>>>   	} else {
>>> -		return RTE_LOGTYPE_PMD;
>>> +		return sfc_logtype_driver;
>>>   	}
>>>   
>>>   	lt_str = rte_zmalloc("logtype_str", lt_str_size_max, 0);
>>>   	if (lt_str == NULL)
>>> -		return RTE_LOGTYPE_PMD;
>>> +		return sfc_logtype_driver;
>>>   
>>>   	strncpy(lt_str, lt_prefix_str, lt_prefix_str_size);
>>>   	lt_str[lt_prefix_str_size - 1] = '.';
>>>
>> Overall I think it is good idea to remove RTE_LOGTYPE_PMD, but sfc has a few
>> more usage of it around same manner, as a fallback value if allocating dynamic
>> one fails.
>>
>>
>> Andrew,
>>
>> Can be possible to update this sfc patch to completely eliminate RTE_LOGTYPE_PMD
>> usage? What do you think?
>>
>> Thanks,
>> ferruh
> 
> I'm OK to use  sfc_logtype_driverif dynamic log type register fails, but
> what should I do if sfc_logtype_driverregister fails?

Right. Same concern is valid for all dynamic log registration but we are
ignoring it.

What do you think instead of each pmd/lib cover the error case,
'rte_log_register_type_and_pick_level()' & 'rte_log_register()' always return
valid value. Like if they fail internally, return 'RTE_LOGTYPE_GENERIC' kind of
value.