Re: [dpdk-dev] [PATCH v5 10/10] qede: Enable PMD build

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

On Wed, Apr 20, 2016 at 04:43:44PM +0000, Harish Patil wrote:
> 
> >On Wed, Apr 20, 2016 at 10:51:06AM +0200, Thomas Monjalon wrote:
> >> 2016-04-20 00:14, Harish Patil:
> >> > >2016-03-31 19:15, Rasesh Mody:
> >> > >> --- a/config/common_base
> >> > >> +++ b/config/common_base
> >> > >> +CONFIG_RTE_LIBRTE_QEDE_RX_COAL_US=24
> >> > >> +CONFIG_RTE_LIBRTE_QEDE_TX_COAL_US=48
> >> > >
> >> > >It looks to be some tuning which could be done at runtime. Isn't it?
> >> >  
> >> > That’s right. Can you please suggest if there is any better option to
> >>make
> >> > it a runtime? There is no PMD API for that.
> >> 
> >> There are some devargs for that.
> >> For PCI dev, it can be passed in the whitelist option.
> >> We should remove this limitation by having a devargs API (and command
> >>line
> >> options) independent of whitelisting.
> >
> >But back to the original setting. Are these likely to be values that are
> >tunable
> >or need to be tunable by the user?
> 
> > 
> This is a tunable which is equivalent of ethtool -c (in linux) which
> controls the rate at which status block is updated.
> 
> >If not, I see little reason to make them
> >run-time configurable - they could be defines inside the driver itself.
> 
> There are defines already which set them to the defaults. The reason to
> expose it as config option is that the user don’t have to change the
> driver and rather just control it via config file. For now, I can remove
> it till we find a better alternative to make those run time. Please
> confirm and I can have it removed in next patch submission.
> 
Removing config options is always a good thing. As you say we can always find a
way to make them tunable in a later patchset.

/Bruce