[dpdk-dev] [PATCH v4] ethdev: replace bus specific struct with generic dev

De Lara Guarch, Pablo pablo.de.lara.guarch at intel.com
Wed Apr 4 19:57:12 CEST 2018



> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ferruh Yigit
> Sent: Tuesday, April 3, 2018 10:50 AM
> To: David Marchand <david.marchand at 6wind.com>; santosh
> <santosh.shukla at caviumnetworks.com>
> Cc: dev at dpdk.org; Shreyansh Jain <shreyansh.jain at nxp.com>; Legacy, Allain
> (Wind River) <allain.legacy at windriver.com>; Tomasz Duszynski
> <tdu at semihalf.com>; Thomas Monjalon <thomas at monjalon.net>
> Subject: Re: [dpdk-dev] [PATCH v4] ethdev: replace bus specific struct with
> generic dev
> 
> On 4/3/2018 10:06 AM, David Marchand wrote:
> > On Mon, Apr 2, 2018 at 6:13 PM, santosh
> > <santosh.shukla at caviumnetworks.com> wrote:
> >> On Friday 30 March 2018 08:59 PM, David Marchand wrote:
> >>> I can see we enforce the driver name by putting it after the call to
> >>> .dev_infos_get.
> >>> http://dpdk.org/browse/dpdk/tree/lib/librte_ether/rte_ethdev.c#n2399
> >>>
> >>> octeontx pmd seems to try to do something about it:
> >>> http://dpdk.org/browse/dpdk/tree/drivers/net/octeontx/octeontx_ethde
> >>> v.c#n622
> >>>
> >>> Not sure it does something, might be a thing to cleanup.
> >>>
> >>>
> >> In case, if your referring to driver_name update then indeed its a
> >> cleanup [1].
> >>
> >> Otherwise, I don't see any issue with v4 Or may be /I /misunderstood
> >> your comment.
> >
> > I agree there is no fundamental issue.
> >
> >     dev_info->device = dev->device;
> >
> >     RTE_FUNC_PTR_OR_RET(*dev->dev_ops->dev_infos_get);
> >     (*dev->dev_ops->dev_infos_get)(dev, dev_info);
> >     dev_info->driver_name = dev->device->driver->name;
> >
> > If somebody (I mean some pmd out there) has a usecase with
> > dev_info->device != dev->device, why not.
> 
> Intentional let drivers update this variable although I don't also see any use case
> of it.
> 
> This variable was set by PMDs before this patch, so I don't see any reason to be
> so strict here.
> 
> If driver does anything ethdev will set dev_info->device for it, if it want to
> overwrite, for any reason, it will have the capability.

Looks good to me. Will do the same for cryptodev and bbdev.
The only thing that I am missing here is an update in documentation,
adding the ABI Change in release notes.

Apart from it:

Acked-by: Pablo de Lara <pablo.de.lara.guarch at intel.com>

> 
> >
> > Thomas ?
> >
> >



More information about the dev mailing list