[dpdk-dev] [PATCH 00/11] kill global pci device id list

Zhang, Helin helin.zhang at intel.com
Thu Jan 21 02:12:45 CET 2016


Hello David

Yes, long time ago, the same device IDs need to be filled in 3 or 4 places. Then it
would be better to have a centralized one, to avoid missing any.
And yes, it would be good to maintain device IDs per PMD.
As basically we don't expose any base driver header files, and sometimes those
device IDs need to be defined into different meanings, I'd prefer to have each
PMD has its own centralized device ID list header file, like you did for ixgbe.

Thanks a lot!

Regards,
Helin

> -----Original Message-----
> From: David Marchand [mailto:david.marchand at 6wind.com]
> Sent: Saturday, January 16, 2016 11:03 PM
> To: Zhang, Helin
> Cc: dev at dpdk.org; Thomas Monjalon
> Subject: Re: [dpdk-dev] [PATCH 00/11] kill global pci device id list
> 
> Hello Helin,
> 
> On Sun, Jan 10, 2016 at 4:53 PM, Zhang, Helin <helin.zhang at intel.com>
> wrote:
> > Thanks for your huge contribution!
> > May you help to describe more details of why you made these huge
> changes?
> 
> As far as I can see, the only reason why we have a centralised header
> maintained in eal with all pci device ids is the need to identify pci devices that
> require a special treatment before starting a dpdk application (here, bind
> those pci devices to igb_uio / vfio).
> 
> This patchset splits this header into small pieces maintained by the drivers
> themselves, then tries to come up with a way to retrieve those pci device ids
> from the final dpdk application and from the drivers compiled as shared
> libraries.
> 
> With this, supported pci device ids are maintained by the drivers themselves
> rather than eal, pci devices ids requiring uio/vfio binding are still available.
> 
> 
> Regards,
> --
> David Marchand


More information about the dev mailing list