On Thu, 2009-10-01 at 14:26 -0500, Mike Christie wrote:
> On 10/01/2009 12:08 PM, Vasu Dev wrote:
> > On Wed, 2009-09-30 at 14:59 -0500, micha...@cs.wisc.edu wrote:
> >> - lport = fcoe_if_create(fcoe,&netdev->dev, 0);
> >> + lport = fcoe_if_create(fcoe,&netdev->dev.parent, 0);
> >
> > No ampersand needed here since dev.parent is already a pointer unlink
Ooop typo, I meant unlike not unlink.
> > previously used netdev->dev.
> >
>
> Ughh, how did that even work for me. I will resend with that fixed if
> this is the way we want to go. Thanks.
>
> Do we instead want the parent to be the netdev device so it shows up in
> sysfs like that? I am not sure which is better or right wrt sysfs rules.
I'm also not sure but it seems keeping netdev in sysfs tree would build
complete device tree chain since fcoe i/f is created on top of netdev.
So perhaps would be good from completeness perspective.
> I just know what we want the scsi layer to end up getting the pci
> device's device, but we can do that in different ways if you guys want
> the netdev in the sysfs tree.
Perhaps scsi code should be fixed to skip all netdev instance in
shost->parent chain to find real pci device but only if device
capabilities will be really useful in setting up device queue and I'm
not sure how useful it would be than altering them in slave_alloc or sht
to have fcoe working on any nic with pause capabilities.
Vasu
_______________________________________________
devel mailing list
devel@open-fcoe.org
http://www.open-fcoe.org/mailman/listinfo/devel
