Re: [PATCH 4/4] ethdev: add traffic manager query function

[Bruce Richardson <bruce.richardson@intel.com>]

On Sun, Sep 22, 2024 at 05:26:30PM +0100, Ferruh Yigit wrote:
> On 8/6/2024 4:24 PM, Bruce Richardson wrote:
> > +/**
> > + * Return information about a traffic management node
> > + *
> > + * Return information about a hierarchy node, using the same format of parameters
> > + * as was passed to the rte_rm_node_add() function.
> > + * Each of the "out" parameters pointers (except error) may be passed as NULL if the
> > + * information is not needed by the caller. For example, to one may check if a node id
> > + * is in use by:
> > + *
> > + *  struct rte_tm_error error;
> > + *  int ret = rte_tm_node_query(port, node_id, NULL, NULL, NULL, NULL, NULL, &error);
> > + *  if (ret == ENOENT) ...
> > + *
> > + * @param[in] port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param[in] node_id
> > + *   Node ID. Should be a valid node id.
> > + * @param[out] parent_node_id
> > + *   Parent node ID.
> > + * @param[out] priority
> > + *   Node priority. The highest node priority is zero. Used by the SP algorithm
> > + *   running on the parent of the current node for scheduling this child node.
> > + * @param[out] weight
> > + *   Node weight. The node weight is relative to the weight sum of all siblings
> > + *   that have the same priority. The lowest weight is one. Used by the WFQ
> > + *   algorithm running on the parent of the current node for scheduling this
> > + *   child node.
> > + * @param[out] level_id
> > + *   The node level in the scheduler hierarchy.
> > + * @param[out] params
> > + *   Node parameters, as would be used when creating the node.
> > + * @param[out] error
> > + *   Error details. Filled in only on error, when not NULL.
> > + * @return
> > + *   0 on success, non-zero error code otherwise.
> > + *   -EINVAL - port or node id value is invalid
> > + *   -ENOENT - no node exists with the provided id
> > + */
> > +int
> > +rte_tm_node_query(uint16_t port_id,
> > +	uint32_t node_id,
> > +	uint32_t *parent_node_id,
> > +	uint32_t *priority,
> > +	uint32_t *weight,
> > +	uint32_t *level_id,
> > +	struct rte_tm_node_params *params,
> > +	struct rte_tm_error *error);
> > +
> >
> 
> No objection to get an TM node query API overall, but it would be good
> to get more comment on the what correct API should be, we are missing it.
> Both because it is not discussed much, and it is first release, better
> to add this API as experimental.
> 

Yes, good point, it should be added to version.map file and marked as
experimental. Happy to take any feedback on what the API should be.

> Also we should have an implementation in driver and a sample application
> usage (testpmd?) with new API. Are these planned separately for this
> release, or can it be available part of next version of this patch?
> 
I have an implemented for ice driver in [1]. On testpmd side, I never
thought to do so, because it was more for me when writing test code around
the APIs than for actual end-users. [As I explain, this API should save app
developers the work of storing a copy of the TM hierarchy in the app code
too]. However, I think it may be no harm to do a testpmd call for it, it
may be useful for debugging.

[1] https://patches.dpdk.org/project/dpdk/patch/20240812152815.1132697-2-bruce.richardson@intel.com/

> Finally, does it worth documenting this in release notes, as just a
> query API I am not sure if this a notable feature, but just a reminder.
> 

Don't think it requires an RN item.

Will perhaps do a new revision of this patchset without this final patch,
and then submit this patch as a separate one for tracking. I don't think
the other patches to mark things const should be blocked by discussion on
this.

Regards,

/Bruce