Re: [dpdk-dev] [PATCH v4] ethdev: add sanity checks in control APIs

[Kevin Traynor <ktraynor@redhat.com>]

On 16/04/2021 08:00, Min Hu (Connor) wrote:
> Thanks Kevin,
> 	all is fixed in v6, please review it, thanks.
> 	Some comments are below.
> 
> 在 2021/4/15 20:04, Kevin Traynor 写道:
>> On 15/04/2021 01:52, Min Hu (Connor) wrote:
>>> This patch adds more sanity checks in control path APIs.
>>>
>>
>> Hi Connor,
>>
>> A few general comments,
>>
>> --
>> Some of the functions have unit tests, you could consider adding unit
>> tests for the new checks. Considering the checks are not subtle and
>> unlikely to be messed up in future, not adding unit tests is not a
>> blocker imho.
>>
>> --
>> It took me a while to get what you meant with "by NULL". It's usage
>> seems like in "Death by taxes". Perhaps "because NULL ptr" would be a
>> better way to phrase this generically, but I think it is more useful to
>> say what is NULL.
>>
>> e.g. "Failed to convert NULL to string\n" is very generic and would be
>> better as "Failed to convert NULL link to string\n" . ok, still a bit
>> generic but more of a clue.
>>
>> I won't comment on each log message individually but I've added a few
>> suggestions here and there.

Thanks, I think it looks a lot nicer to read in v6 my completely
subjective biased opinion :-)

>> --
>>
>> Did you check the usage of these functions in DPDK, and if the return
>> value is handled ok? e.g. RTE_ETH_FOREACH_MATCHING_DEV will keep calling
>> iterator functions. I'm not sure that having a return check is needed in
>> that case, but there could be other cases where you want to take some
>> different action now.
>>
> As iterator functions are all APIs, they may be used by APP directly.
> I think param check is necessary.

The point is that it would continue to call the functions even after it
caught this error, so would continue to print error messages. Yes, that
is much better than a seg fault and maybe in this case that is ok. I
will leave it to maintainers to decided.

I was just wondering if there was additional things similar to this in
DPDK where handling these new errors could now be improved too. I don't
think it has to be a prerequisite for this patch, as this patch is still
an improvement.

>> some other comments inlined,
>>
>>> Fixes: 214ed1acd125 ("ethdev: add iterator to match devargs input")
>>> Fixes: 3d98f921fbe9 ("ethdev: unify prefix for static functions and variables")
>>> Fixes: 0366137722a0 ("ethdev: check for invalid device name")
>>> Fixes: d948f596fee2 ("ethdev: fix port data mismatched in multiple process model")
>>> Fixes: 5b7ba31148a8 ("ethdev: add port ownership")
>>> Fixes: f8244c6399d9 ("ethdev: increase port id range")