> On 8/26/2016 5:54 PM, Kamil Rytarowski wrote:
>> From: Kamil Rytarowski <kamil.rytarowski at caviumnetworks.com>
>>
>> In case of the multiprocess mode a shared nicvf struct between processes
>> cannot point with the eth_dev pointer to master device, therefore remove it
>> allong with references to it refactoring the code where needed.
>
> Patch subject says fix multiprocess support in stat, but it seems doing
> more than just for stats fix.
>
> Overall eliminates private_data->eht_dev link.
>
> I guess this is because eth_dev->data is shared for primary and
> secondary processes, and this makes impossible to use separate
> private_data for primary and secondaries.
>
> So this patch looks like keeping separate copy of private_data (nic) and
> using eth_dev and nic structs for functions instead of using
> eth_dev->data->private_data.
>
> If above correct, can you please updated patch subject?
>
> Also another approach can be allocating "data" independently and
> overwrite eth_dev->data with this per each process, this also makes
> eth_dev->data->private_data usable for each process.
>
>>
>> Fixes: 7413feee662d ("net/thunderx: add device start/stop and close")
>>
>> Signed-off-by: Maciej Czekaj <maciej.czekaj at caviumnetworks.com>
>> Signed-off-by: Kamil Rytarowski <kamil.rytarowski at caviumnetworks.com>
>> Signed-off-by: Zyta Szpak <zyta.szpak at semihalf.com>
>> Signed-off-by: Slawomir Rosek <slawomir.rosek at semihalf.com>
>> Signed-off-by: Radoslaw Biernacki <rad at semihalf.com>
>> Signed-off-by: Jerin Jacob <jerin.jacob at caviumnetworks.com>
>> ---
>
>
>Yes, this patch enables multi-process support for primary and secondary functions by eliminating the pointer from private struct to eth_dev. Will extend the comment in v2.
