[dpdk-dev] [PATCH] eventdev: define the default value for dequeue timeout

[Jerin Jacob]

-----Original Message-----
> Date: Thu, 18 May 2017 19:45:44 +0530
> From: Hemant Agrawal <hemant.agrawal at nxp.com>
> To: Jerin Jacob <jerin.jacob at caviumnetworks.com>, dev at dpdk.org
> CC: bruce.richardson at intel.com, harry.van.haaren at intel.com,
>  gage.eads at intel.com, nipun.gupta at nxp.com, narender.vangati at intel.com
> Subject: Re: [PATCH] eventdev: define the default value for dequeue timeout
> User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:45.0) Gecko/20100101
>  Thunderbird/45.8.0
> 
> On 5/18/2017 2:18 PM, Jerin Jacob wrote:
> > Defining the value 0 as default value for dequeue timeout
> > will help the application reduce the configuration setup
> > if the application is interested only in default
> > timeout value.
> > 
> > Signed-off-by: Jerin Jacob <jerin.jacob at caviumnetworks.com>
> > ---
> > This patch will fix following error found in the event_pipeline RFC application
> > http://dpdk.org/dev/patchwork/patch/23799/ with event_octeontx HW driver.
> > 
> > EVENTDEV: rte_event_dev_configure() line 379: dev0 invalid
> > dequeue_timeout_ns=0 min_dequeue_timeout_ns=853 max_dequeue_timeout_ns=873813
> > ---
> >  drivers/event/octeontx/ssovf_evdev.c | 2 ++
> >  lib/librte_eventdev/rte_eventdev.c   | 5 +++--
> >  lib/librte_eventdev/rte_eventdev.h   | 1 +
> >  3 files changed, 6 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/event/octeontx/ssovf_evdev.c b/drivers/event/octeontx/ssovf_evdev.c
> > index c80a44379..5499b1bf7 100644
> > --- a/drivers/event/octeontx/ssovf_evdev.c
> > +++ b/drivers/event/octeontx/ssovf_evdev.c
> > @@ -194,6 +194,8 @@ ssovf_configure(const struct rte_eventdev *dev)
> > 
> >  	ssovf_func_trace();
> >  	deq_tmo_ns = conf->dequeue_timeout_ns;
> > +	if (deq_tmo_ns == 0)
> > +		deq_tmo_ns = edev->min_deq_timeout_ns;
> 
> '0' should mean don't wait?

Yes. I think, we can leave that to the driver for the minimum supported
dequeue timeout for the given platform(PMD). OCTEONTX PMD needs
different treatment for "no-wait" case, hence setting the minimum value
that PMD supports.

Any other comments on API change?


> 
> > 
> >  	if (conf->event_dev_cfg & RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT) {
> >  		edev->is_timeout_deq = 1;
> > diff --git a/lib/librte_eventdev/rte_eventdev.c b/lib/librte_eventdev/rte_eventdev.c
> > index 20afc3f0e..8cafffe03 100644
> > --- a/lib/librte_eventdev/rte_eventdev.c
> > +++ b/lib/librte_eventdev/rte_eventdev.c
> > @@ -369,9 +369,10 @@ rte_event_dev_configure(uint8_t dev_id,
> > 
> >  	/* Check dequeue_timeout_ns value is in limit */
> >  	if (!(dev_conf->event_dev_cfg & RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT)) {
> > -		if (dev_conf->dequeue_timeout_ns < info.min_dequeue_timeout_ns
> > +		if (dev_conf->dequeue_timeout_ns &&
> > +		    (dev_conf->dequeue_timeout_ns < info.min_dequeue_timeout_ns
> >  			|| dev_conf->dequeue_timeout_ns >
> > -				 info.max_dequeue_timeout_ns) {
> > +				 info.max_dequeue_timeout_ns)) {
> >  			RTE_EDEV_LOG_ERR("dev%d invalid dequeue_timeout_ns=%d"
> >  			" min_dequeue_timeout_ns=%d max_dequeue_timeout_ns=%d",
> >  			dev_id, dev_conf->dequeue_timeout_ns,
> > diff --git a/lib/librte_eventdev/rte_eventdev.h b/lib/librte_eventdev/rte_eventdev.h
> > index 94284337d..f39fbc6b9 100644
> > --- a/lib/librte_eventdev/rte_eventdev.h
> > +++ b/lib/librte_eventdev/rte_eventdev.h
> > @@ -409,6 +409,7 @@ struct rte_event_dev_config {
> >  	 * This value should be in the range of *min_dequeue_timeout_ns* and
> >  	 * *max_dequeue_timeout_ns* which previously provided in
> >  	 * rte_event_dev_info_get()
> > +	 * The value 0 is allowed, in which case, default dequeue timeout used.
> >  	 * @see RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT
> >  	 */
> >  	int32_t nb_events_limit;
> > 
> 
>