Hi Matan,
Thanks for the patches. A remark however:
On Sun, Jan 07, 2018 at 09:45:50AM +0000, Matan Azrad wrote:
> Fail-safe PMD sub devices management is based on ethdev port mechanism.
> So, the sub-devices management structures are exposed to other DPDK
> entities which may use them in parallel to fail-safe PMD.
>
> Use the new port ownership mechanism to avoid multiple managments of
> fail-safe PMD sub-devices.
>
I think your implementation does not work with several fail-safe
instances, have you tested this configuration?
It should be possible for a user to create any number of fail-safe
instances. The minimum would be to allow for multiple fail-safe
side-by-side, but ideally it should also support a recursive
configuration:
+-----------+
|fail-safe |
| |
| |
+-+ +--+
| | | |
| +-----------+ |
| |
+-------v----+ +-----v-----+
|fail-safe | | |
| | | |
| | | |
| | | |
+-+ +-+ | |
| +------------+ | +-----------+
| |
+-----v-----+ +-----v-----+
| | | |
| | | |
| | | |
| | | |
| | | |
+-----------+ +-----------+
If I am not mistaken on this, then you need to generate different
owner-ids for each fail-safe instances.
I suggest using the full fail-safe instance name, as they are already
assured to be different from each other by the EAL, and you thus won't
need to generate IDs on the fly, as well as declare a global owner-id
prefix.
--
Gaëtan Rivet
6WIND
