netdev - Re: [PATCH net-next 1/2] net: phy: realtek: add PHY driver for RTL8127ATF

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 for Android: free password hash cracker in your pocket

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

Message-ID: <9a21aff5-bbe7-4163-b55b-3e5d4f9496c1@gmail.com>
Date: Fri, 9 Jan 2026 12:10:17 +0100
From: Heiner Kallweit <hkallweit1@...il.com>
To: Daniel Golle <daniel@...rotopia.org>,
 Fabio Baltieri <fabio.baltieri@...il.com>
Cc: Andrew Lunn <andrew@...n.ch>, Andrew Lunn <andrew+netdev@...n.ch>,
 Russell King - ARM Linux <linux@...linux.org.uk>,
 Paolo Abeni <pabeni@...hat.com>, Eric Dumazet <edumazet@...gle.com>,
 David Miller <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
 Vladimir Oltean <vladimir.oltean@....com>,
 Michael Klein <michael@...sekall.de>,
 Realtek linux nic maintainers <nic_swsd@...ltek.com>,
 Aleksander Jan Bajkowski <olek2@...pl>,
 "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next 1/2] net: phy: realtek: add PHY driver for
 RTL8127ATF

On 1/9/2026 2:26 AM, Daniel Golle wrote:
> On Thu, Jan 08, 2026 at 11:20:21PM +0000, Fabio Baltieri wrote:
>> On Thu, Jan 08, 2026 at 10:56:14PM +0000, Daniel Golle wrote:
>>>> +static int rtlgen_sfp_read_status(struct phy_device *phydev)
>>>> +{
>>>> +	int val, err;
>>>> +
>>>> +	err = genphy_update_link(phydev);
>>>> +	if (err)
>>>> +		return err;
>>>> +
>>>> +	if (!phydev->link)
>>>> +		return 0;
>>>> +
>>>> +	val = rtlgen_read_vend2(phydev, RTL_VND2_PHYSR);
>>>
>>> This should be the same as
>>> phy_read(phydev, MII_RESV2); /* on page 0 */
>>> Please try.
>>
>> Tried it on my setup, the two calls do indeed seem to return the same
>> value.
> 
> Thank you for confirming that.
> 
> My understanding at this point is that only register 0x10 to 0x17 are
> actually paged (ie. the 3 bits of freedom in the
> RTL822X_VND2_TO_PAGE_REG apply to all pages), and that seems to apply for
> all 1G, 2.5G and 5G (and 10G?) RealTek PHYs.
> 
Speaking for the internal PHY's (at least for all c45-capabale ones):
There's no actual paging in the PHY. r8169 translates each paged access
to a register in MDIO_MMD_VEND2.

> Hence we do not need to use paged register access for register 0x0...0xf
> and 0x18..0x1e. And the paged operations we do have there right now can
> all be described as registers on MDIO_MMD_VEND2. And maybe that's what
> we should do then, implementing .read_mmd and .write_mmd similar to
> rtl822xb_read_mmd and rtl822xb_write_mmd for all PHYs, with the only
> difference that for older PHYs all MMDs other than MDIO_MMD_VEND2 have
> to be emulated similar to rtlgen_read_mmd and rtl822x_read_mmd.
> 
> The current way we access MDIO_MMD_VEND2 on older PHYs also also fishy
> as it depends on __mdiobus_c45_read as well as the PHY listening to the
> broadcast address 0: Especially for 1GE PHYs not all MDIO controllers
> are capable of Clause-45 access, and listening on address 0 works (at
> best) if there is only one PHY in the bus doing that, and it can be
> disabled via BIT(13) on PHYCR1. For internal PHYs of PCIe NICs this is
> fine, of course, but for standalone PHYs not really.
> 
The access using __mdiobus_c45_read is meant primarily for the internal
PHY's, where the MII bus is provided by r8169, using MAC registers.

> tl;dr: drivers/net/phy/realtek/ has signed up for some serious
> weight-loss program.

Indeed ..