[dpdk-dev] [PATCH 1/2] eal: internalize devargs parsing

[Gaëtan Rivet]

On Sun, Jul 09, 2017 at 06:56:25AM -0400, Jan Blunck wrote:
> On Sun, Jul 9, 2017 at 6:46 AM, Jan Blunck <jblunck at infradead.org> wrote:
> > On Sun, Jul 9, 2017 at 6:28 AM, Thomas Monjalon <thomas at monjalon.net> wrote:
> >> 09/07/2017 11:28, Jan Blunck:
> >>> This is going to get cleaned up in future releases so lets not expose
> >>> this API unnecessarily.
> >>>
> >>> Signed-off-by: Jan Blunck <jblunck at infradead.org>
> >>> ---
> >>> --- a/lib/librte_eal/linuxapp/eal/rte_eal_version.map
> >>> +++ b/lib/librte_eal/linuxapp/eal/rte_eal_version.map
> >>> @@ -211,7 +211,6 @@ DPDK_17.08 {
> >>>  EXPERIMENTAL {
> >>>       global:
> >>>
> >>> -     rte_eal_devargs_parse;
> >>>       rte_eal_hotplug_add;
> >>>       rte_eal_hotplug_remove;
> >>
> >> We need to expose it temporarily as experimental in order to let
> >> the failsafe driver calling it.
> >> When reworked in 17.11, this function or an equivalent should allow
> >> failsafe (and maybe bonding) to parse slave declaration.
> >
> > There is no need for the failsafe PMD to parse it to a rte_devargs
> > because it can just pass the string to the hotplug API.
> 
> Just in case it isn't clear enough: the failsafe PMD is using the
> rte_devargs structure to get the bus name. The bus name is explicitly
> given in the hotplug API. There is really no need to deal with
> rte_devargs here.

The fail-safe PMD uses the rte_devargs parsing helper to find the
bus name from a device argument handled to it.

As you point out, the bus name is explicitly given in the hotplug API.
In order to be able to explicitly give the bus name, the caller has to
know it. This is done by parsing devargs and using the result.

-- 
Gaëtan Rivet
6WIND