[dpdk-dev] [PATCH] app/testpmd: app/testpmd: add device removal command

[Gaëtan Rivet]

On Wed, Aug 23, 2017 at 06:18:34PM +0200, Thomas Monjalon wrote:
> 23/08/2017 17:09, Gaëtan Rivet:
> > Hello Raslan,
> > 
> > On Wed, Aug 23, 2017 at 05:37:04PM +0300, Raslan Darawsheh wrote:
> > > Added hotplug in testpmd, to be able to test hotplug function
> > > in the PMD's.
> > > 
> > > Signed-off-by: Raslan Darawsheh <rasland at mellanox.com>
> [...]
> > > --- a/app/test-pmd/cmdline.c
> > > +++ b/app/test-pmd/cmdline.c
> > > @@ -716,6 +716,9 @@ static void cmd_help_long_parsed(void *parsed_result,
> > >  			"port config (port_id|all) l2-tunnel E-tag"
> > >  			" (enable|disable)\n"
> > >  			"    Enable/disable the E-tag support.\n\n"
> > > +
> > > +			" device remove (device)\n"
> > > +			"    Remove a device"
> > 
> > I think it should still be a part of the "port" command set (port
> > attach|detach|stop|close, etc).
> 
> I tend to disagree.
> As far as I know, we use port for ethdev or cryptodev.
> Here we want to deal with EAL rte_device.
> 

I see, that makes sense.
I will redo the review with that in mind.

> > This would probably be easier to understand for users.
> 
> [...]
> > Continuing on using the port ...
> > format, then the port_id should allow to remove it instead of the device
> > identifier.
> > Using the device identifier will complexify your implementation.
> [...]
> > 	eth_dev = &rte_eth_devices[port_id];
> > 	bus = rte_bus_find_by_device(eth_dev->device);
> 
> Note that we are going to remove eth_dev->device which implies eth_dev
> but maybe also more device interfaces for the same HW.
> That's why I think we need to distinguish port and device somehow.

-- 
Gaëtan Rivet
6WIND