[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190104132438.GB11435@lunn.ch>
Date: Fri, 4 Jan 2019 14:24:38 +0100
From: Andrew Lunn <andrew@...n.ch>
To: Marek Vasut <marex@...x.de>
Cc: netdev@...r.kernel.org, "David S . Miller" <davem@...emloft.net>,
Nisar Sayed <Nisar.Sayed@...rochip.com>,
Woojung Huh <Woojung.Huh@...rochip.com>,
Florian Fainelli <f.fainelli@...il.com>,
linux-usb@...r.kernel.org
Subject: Re: [PATCH 10/19] usbnet: smsc95xx: Replace smsc95xx_mdio_read()
with phy_read()
> > All this crossover code should be moved into the PHY driver.
>
> Fine, this can be done in a subsequent patch though, right ? I'd like to
> keep the changes small, so if something breaks, it could be bisected easily.
Hi Marek
The problem is, do you have a regression because of the changes you
made here, or because you have two things controlling the PHY, the MAC
driver and the PHY driver, which are not coordinating?
I don't know if you can do a piecemeal conversation like this. I don't
remember anybody else doing anything like this before.
What you can do piecemeal is add new functionality to the existing PHY
driver. If you break the PHY driver we can then bisect it to find out
which change broke it. Once the PHY driver has everything you need,
then swap the MAC driver to use phylib.
Thanks
Andrew
Powered by blists - more mailing lists