issue patch in next net/eth: fix link handling

Jean-Christophe PLAGNIOL-VILLARD plagnioj at jcrosoft.com
Fri Sep 28 05:14:14 EDT 2012


On 10:55 Fri 28 Sep     , Sascha Hauer wrote:
> On Fri, Sep 28, 2012 at 10:27:16AM +0200, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > On 09:50 Fri 28 Sep     , Sascha Hauer wrote:
> > > On Fri, Sep 28, 2012 at 04:28:21AM +0200, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > > > HI,
> > > > 
> > > > 	The patch is next
> > > > 	net/eth: fix link handling
> > > > 
> > > > 	was NEVER send to the ML
> > > > 
> > > > 	IIRC I was the author of the first version and this disapear
> > > > 
> > > > 	Uwe and I just get this discussion on the kernel ML about patch update
> > > > 
> > > 
> > > I was basically pissed off because I got the strong feeling that I spent
> > > more time reviewing and testing the patch than you initially spent
> > > writing it in the first place. The second version still stored apples
> > > in edev->phydev->link and bananas in edev->carrier, but still did a
> > > edev->carrier = dev->link.
> > I did this on purpose as I do want to store the link and later export it via
> > env as I get a patch here for 2 wifi driver where I'll not use the phylib
> > 
> > so store the carrier is the correct way
> 
> Whatever it is, adding a variable to an ethernet device and then
> manipulating it in both the phylib and the ethernet code is desastrous.
> It must be clear everytime who owns a variable. Doing a
> 
> eth_current->carrier = CARRIER_UNKNOW;
> 
> in the ethernet code, and then:
> 
> edev->carrier = dev->link;
> 
> in the phy code is a recipe for spaghetti code.
> 
so we need to do call net_carrier in phylib as done in the kernel
the net framework mamange the carrier by itself

as example in wifi phy up does not mean carrier on

Best Regards,
J.



More information about the barebox mailing list