> Subject: Re: [PATCH v3 04/20] i40e: Register a virtbus device to provide RDMA
>
> On Wed, Dec 18, 2019 at 06:57:10PM +0000, Saleem, Shiraz wrote:
> > diff --git a/include/linux/net/intel/i40e_client.h
> > b/include/linux/net/intel/i40e_client.h
> > index 7e147d3..5c81261 100644
> > +++ b/include/linux/net/intel/i40e_client.h
> > @@ -83,11 +83,11 @@ struct i40e_params {
> >
> > /* Structure to hold Lan device info for a client device */ struct
> > i40e_info {
> > struct i40e_client_version version;
>
> I hope this isn't the inter-module versioning stuff we already Nak'd?
This is not used in irdma for checking any versioning.
This and the version in struct i40e_client are not cleaned up because they are used
between i40e <--> i40iw. And the thought was once i40iw is removed, we can send
a follow on patch to remove this versioning + the global
exported functions i40e_register_client() and i40e_unregister_client() used by i40iw.
For the ice driver, we have already removed all versioning.
>
> > u8 lanmac[6];
>
> Is this different from the mac reachable from the netdev?
Hmmm... I think not. This should be possible to get from netdev. We can send a cleanup follow on patch.
> > +struct i40e_virtbus_device {
> > + struct virtbus_device vdev;
> > + struct i40e_info *ldev;
>
> Is the lifetime actually any better? Will ldev be freed and left danling before
> virtbus_device is released?
>
The consumer of the ldev (i.e the rdma driver) will have a
valid ldev when drv->remove() is called as part of the virtbus
device unregister.
It is probably good to set this ldev pointer inside the i40e_virtbus_device
object to NULL in the netdev driver soon after virtbus device unregister.
