[PATCH v2 3/3] net/ice: do not include private headers in testpmd
Burakov, Anatoly
anatoly.burakov at intel.com
Thu Oct 15 17:29:08 CEST 2026
On 10/15/2026 5:21 PM, Bruce Richardson wrote:
> On Thu, Oct 15, 2026 at 05:00:38PM +0200, Burakov, Anatoly wrote:
>> On 10/15/2026 4:52 PM, Medvedkin, Vladimir wrote:
>>> Hi Anatoly,
>>>
>>> One nit: the buffer must be allocated with malloc, could you please
>>> reflect this information in doxygen comments?
>>
>> @Bruce
>>
>> Perhaps this can be fixed on apply?
>>
>> diff --git a/drivers/net/intel/ice/rte_pmd_ice.h
>> b/drivers/net/intel/ice/rte_pmd_ice.h
>> index dccfdcd5f4..881dc4cc56 100644
>> --- a/drivers/net/intel/ice/rte_pmd_ice.h
>> +++ b/drivers/net/intel/ice/rte_pmd_ice.h
>> @@ -27,8 +27,9 @@ extern "C" {
>> * @param port
>> * The port identifier of the Ethernet device.
>> * @param buff
>> - * Pointer to an allocated buffer. On success, it may be replaced by a
>> - * newly allocated buffer; the caller must free the resulting buffer.
>> + * Pointer to a `malloc()`-allocated buffer. On success, it will be freed
>> by
>> + * `free()` and replaced by a newly `malloc()`-allocated buffer; the
>> caller
>> + * must free the resulting buffer.
>
> That is very strange behaviour! That looks like something a future patch
> should fix, but ok to change the documentation on apply here.
>
> /Bruce
Agreed. My best guess is that the intention was to return success even
if the buffer was too small, by passing pointer-to-pointer so that we
can reallocate buffer of an appropriate size (as opposed to introducing
a "size" API...). We will address this later.
--
Thanks,
Anatoly
More information about the dev
mailing list