linux-kernel - Re: [net-next RFC PATCH 1/6] net: phy: add support for defining multiple PHY IDs in PHY driver

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <65d26c13.df0a0220.63f42.d8e6@mx.google.com>
Date: Sun, 18 Feb 2024 21:44:03 +0100
From: Christian Marangi <ansuelsmth@...il.com>
To: "Russell King (Oracle)" <linux@...linux.org.uk>
Cc: Andrew Lunn <andrew@...n.ch>, Heiner Kallweit <hkallweit1@...il.com>,
	"David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
	Florian Fainelli <florian.fainelli@...adcom.com>,
	Broadcom internal kernel review list <bcm-kernel-feedback-list@...adcom.com>,
	Robert Marko <robimarko@...il.com>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@....com>,
	Nipun Gupta <nipun.gupta@....com>,
	Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
	Puneet Gupta <puneet.gupta@....com>,
	Abhijit Gangurde <abhijit.gangurde@....com>,
	Umang Jain <umang.jain@...asonboard.com>, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [net-next RFC PATCH 1/6] net: phy: add support for defining
 multiple PHY IDs in PHY driver

On Sun, Feb 18, 2024 at 08:34:16PM +0000, Russell King (Oracle) wrote:
> On Sun, Feb 18, 2024 at 09:27:22PM +0100, Christian Marangi wrote:
> > On Sun, Feb 18, 2024 at 09:10:30PM +0100, Andrew Lunn wrote:
> > > > > > +	phy_dev_id = (struct mdio_device_id *)&phydev->dev_id;
> > > > > 
> > > > > Why this cast? Try to write code that doesn't need casts.
> > > > > 
> > > > 
> > > > This cast is needed to keep the dev_id const in the phy_device struct so
> > > > that other are warned to not modify it and should only be handled by
> > > > phy_probe since it's the one that fills it.
> > > > 
> > > > Alternative is to drop const and drop the warning.
> > > 
> > > Can you propagate the const. Make phy_dev_id point to a const?
> > >
> > 
> > Mhh not following, I tried changing to const struct mdio_device_id *phy_dev_id
> > but that results in memcpy complain (dest is void * not const) and
> > writing in read-only for the single PHY part (the else part)
> > 
> > An alternative might be to make dev_id a pointer in struct phy_device
> > and dynamically allocate a mdio_device_id for the case of single PHY
> > (else case). That effectively remove the need of this cast but I would
> > love to skip checking for -ENOMEM (this is why i made that local)
> > 
> > If it's OK to dynamically allocate for the else case then I will make
> > this change. I just tested this implementation and works correctly with
> > not warning.
> 
> Why do we need memcpy() etc - as I demonstrated in my proposal, it's
> not necessary if we introduce a mdio_device_id within struct phy_driver
> and we can just store a const pointer to the mdio_device_id that
> matched. That was very much an intentional decision in my proposal to
> make the code simple.
>

With the allocated mdio_devic_id it would result in this snipped

	const struct mdio_device_id *driver_dev_id;
	struct mdio_device_id *dev_id;
	int err = 0;

	phydev->drv = phydrv;
	/* Fill the mdio_device_id for the PHY istance.
	 * If PHY driver provide an array of PHYs, search the right one,
	 * in the other case fill it with the phy_driver data.
	 */
	if (phy_driver_match(phydrv, phydev, &driver_dev_id) && driver_dev_id) {
		/* If defined, overwrite the PHY driver dev name with a
		 * more specific one from the matching dev_id.
		 */
		phydev->dev_id = driver_dev_id;
		if (driver_dev_id->name)
			drv->name = driver_dev_id->name;
	} else {
		dev_id = kzalloc(sizeof(*phydev->dev_id), GFP_KERNEL);
		if (!dev_id) {
			err = -ENOMEM;
			goto out;
		}
		dev_id->phy_id = phydrv->phy_id;
		dev_id->phy_id_mask = phydrv->phy_id_mask;
		dev_id->name = phydrv->name;
		phydev->dev_id = dev_id;
	}

Is it ok? (in phy.h the thing is const struct mdio_device_id *ids)
I don't really like modifying phy_driver too much.

-- 
	Ansuel