[dpdk-dev] [PATCH 2/4] bus/vdev: bus scan by multi-process channel

Tan, Jianfeng jianfeng.tan at intel.com
Tue Mar 6 01:50:57 CET 2018


Hi Anatoly,

> -----Original Message-----
> From: Burakov, Anatoly
> Sent: Monday, March 5, 2018 5:37 PM
> To: Tan, Jianfeng; dev at dpdk.org
> Cc: Richardson, Bruce; Ananyev, Konstantin; thomas at monjalon.net;
> maxime.coquelin at redhat.com; Yigit, Ferruh
> Subject: Re: [PATCH 2/4] bus/vdev: bus scan by multi-process channel
> 
> On 04-Mar-18 3:30 PM, Jianfeng Tan wrote:
> > To scan the vdevs in primary, we send request to primary process
> > to obtain the names for vdevs.
> >
> > Only the name is shared from the primary. In probe(), the device
> > driver is supposed to locate (or request more) the detail
> > information from the primary.
> >
> > Signed-off-by: Jianfeng Tan <jianfeng.tan at intel.com>
> > ---
> 
> Is there much point in having private vdevs? Granted, i'm not exactly a
> heavy user of vdev's, but to me this would seem like a way to introduce
> more confusion. How do i tell which devices are shared between
> processes, and which are private to one process? Can i control which one
> do i get? To me it would seem like it would be better to just switch all
> vdevs to being shared.

Yes, that’s the final target: to make every vdev shared between primary and secondary process.

However, now most kinds of the vdevs do not support multi-process. For those devices,

- If they are firstly probed in primary, then we will share the rte_eth_dev_data to the secondary, so that the secondary can get stats or pdump the port.
- If they are firstly probed in secondary, considering it's mostly used by the secondary process, so we will allocate the "port id" exclusively, and keep it in that secondary process privately.

Thanks,
Jianfeng


More information about the dev mailing list