[dpdk-dev] [PATCH] cryptodev: fix cryptodev start return value

[Pavan Nikhilesh Bhagavatula]

On Wed, Jun 07, 2017 at 03:54:23PM +0000, Trahe, Fiona wrote:
> Hi Pavan,
>
> > -----Original Message-----
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Pavan Nikhilesh
> > Sent: Wednesday, June 7, 2017 11:37 AM
> > To: dev at dpdk.org
> > Cc: Doherty, Declan <declan.doherty at intel.com>; Pavan Nikhilesh Bhagavatula
> > <pbhagavatula at caviumnetworks.com>
> > Subject: [dpdk-dev] [PATCH] cryptodev: fix cryptodev start return value
> >
> > From: Pavan Nikhilesh Bhagavatula <pbhagavatula at caviumnetworks.com>
> >
> > If cryptodev has already started it should return -EBUSY instead of 0
> > when rte_cryptodev_start is called.
> >
> > Fixes: d11b0f30df88 ("cryptodev: introduce API and framework for crypto devices")
> >
> > Signed-off-by: Pavan Nikhilesh <pbhagavatula at caviumnetworks.com>
> > ---
> >  lib/librte_cryptodev/rte_cryptodev.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/lib/librte_cryptodev/rte_cryptodev.c b/lib/librte_cryptodev/rte_cryptodev.c
> > index b65cd9c..c815038 100644
> > --- a/lib/librte_cryptodev/rte_cryptodev.c
> > +++ b/lib/librte_cryptodev/rte_cryptodev.c
> > @@ -1000,7 +1000,7 @@ rte_cryptodev_start(uint8_t dev_id)
> >  	if (dev->data->dev_started != 0) {
> >  		CDEV_LOG_ERR("Device with dev_id=%" PRIu8 " already started",
> >  			dev_id);
> > -		return 0;
> > +		return -EBUSY;
> It makes sense to me to return 0/success in this case, as the end result is the
> same, the device is successfully started.
> But I don't feel strongly about it if there's a good argument for making the change?

I do agree with this but from an application perspective when the API
is called again after the device has already started (without calling
the stop API) it would mean that there is an underlying issue with
the application's business logic and it would go undetected, so I feel
that we should strictly enforce this scenario as an error.

> However, as it is an API change doesn't it need to be flagged in a release before the change is made?

I don't think that this would be an API change as it doesn't deprecate
the existing ABI.
Any thoughts about this from the community are welcome as the same
issue affects multiple core libraries (crytodev, ethdev, eventdev).
>
>
> }
> >
> >  	diag = (*dev->dev_ops->dev_start)(dev);
> > --
> > 2.7.4
>