[dpdk-dev] [PATCH] eventdev: add dev id checks to config functions

[Van Haaren, Harry]

> From: Jerin Jacob [mailto:jerin.jacob at caviumnetworks.com]
> Sent: Monday, September 4, 2017 6:21 AM
> To: Van Haaren, Harry <harry.van.haaren at intel.com>
> Cc: dev at dpdk.org
> Subject: Re: [PATCH] eventdev: add dev id checks to config functions
> 
> -----Original Message-----
> > Date: Mon, 24 Jul 2017 13:48:20 +0100
> > From: Harry van Haaren <harry.van.haaren at intel.com>
> > To: dev at dpdk.org
> > CC: jerin.jacob at caviumnetworks.com, Harry van Haaren
> >  <harry.van.haaren at intel.com>
> > Subject: [PATCH] eventdev: add dev id checks to config functions
> > X-Mailer: git-send-email 2.7.4
> >
> > This commit adds checks to verify the device ID is valid
> > to the following functions. Given that they are non-datapath,
> > these checks are always performed.
> 
> Makes sense.

Great - lets discuss implementation ;)


> > This commit also updates the event/octeontx test-cases to
> > have the correct signed-ness, as the API has changes this
> > change is required in order to compile.
> >
> > Suggested-by: Jesse Bruni <jesse.bruni at intel.com>
> > Signed-off-by: Harry van Haaren <harry.van.haaren at intel.com>
> >
> > ---
> > @@ -1288,9 +1293,10 @@ worker_ordered_flow_producer(void *arg)
> >  static inline int
> >  test_producer_consumer_ingress_order_test(int (*fn)(void *))
> >  {
> > -	uint8_t nr_ports;
> > +	int16_t nr_ports;
> >
> > -	nr_ports = RTE_MIN(rte_event_port_count(evdev), rte_lcore_count() - 1);
> > +	nr_ports = RTE_MIN(rte_event_port_count(evdev),
> > +			(int)rte_lcore_count() - 1);
> 
> While I agree on the problem statement, I am trying to see
> 1/ an API symmetrical to ethdev APIs. Similar problem solved in a differently in
> ethdev. see rte_eth_dev_adjust_nb_rx_tx_desc().
> Just want to make sure, all the APIs across ethdev, eventdev looks same
> 
> 2/ How to get rid of above typecasting
> 
> Considering above two points and following the
> rte_eth_dev_adjust_nb_rx_tx_desc() pattern. How about,
> 
> Removing,
> rte_event_port_dequeue_depth()
> rte_event_port_enqueue_depth()
> rte_event_port_count()
> 
> rte_event_queue_count()
> rte_event_queue_priority()
> 
> and change to,
> 
> int rte_event_port_attr_get(uint8_t dev_id, uint8_t port_id,
> uint8_t *enqueue_depth /*out */, uint8_t *dequeue_depth /* out*/, uin8_t *count /*out*/);
> 
> int rte_event_queue_attr_get(uint8_t dev_id, uint8_t port_id,
> uin8_t *prio /* out */, uint8_t *count /*out */);
> 
> or something similar.

Hmm, I don't like that we'd have to break ABI every time we want to add an item to attr_get().. so adding a parameter "attr_id" would allow adding events in future. This solution feels a bit like a re-implementation of the xstats API..

Thoughts? -H


enum {
  PORT_COUNT,
  PORT_DEQUEUE_DEPTH,
  PORT_ENQUEUE_DEPTH,
}

/* retrieve value of port 
int rte_event_port_attr_get(uint8_t dev_id, uint8_t port_id, uint32_t attr_id, uint32_t *attr_value /* out */);