[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230522134823.GA18381@nucnuc.mle>
Date: Mon, 22 May 2023 15:48:23 +0200
From: David Epping <david.epping@...singlinkelectronics.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>, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, UNGLinuxDriver@...rochip.com
Subject: Re: [PATCH net 3/3] net: phy: mscc: enable VSC8501/2 RGMII RX clock
On Sun, May 21, 2023 at 06:59:53PM +0100, Russell King (Oracle) wrote:
> On Sat, May 20, 2023 at 06:06:03PM +0200, David Epping wrote:
> > +static int vsc85xx_rgmii_enable_rx_clk(struct phy_device *phydev,
> > + u32 rgmii_cntl)
> > +{
> > + int rc, phy_id;
> > +
> > + phy_id = phydev->drv->phy_id & phydev->drv->phy_id_mask;
> > + if (PHY_ID_VSC8501 != phy_id && PHY_ID_VSC8502 != phy_id)
> > + return 0;
>
> As you are accessing the phy_id in the phy_driver struct, isn't it
> already true that this will be initialised to constants such as
> PHY_ID_VSC8501 or PHY_ID_VSC8502? In which case, why would you need
> to mask it with drv->phy_id_mask ?
Yes you are right. I copied the code from the vsc85xx_config_init()
function in the same driver, but the extra masking is not necessary.
It seems to be the phy_id in the struct phy_device which is read from
the MDIO bus and thus needs masking. phy_id in struct phy_driver seems
compile time defined and already masked.
I'll adjust my patch.
> > +
> > + mutex_lock(&phydev->lock);
> > +
> > + rc = phy_modify_paged(phydev, MSCC_PHY_PAGE_EXTENDED_2, rgmii_cntl,
> > + VSC8502_RGMII_RX_CLK_DISABLE, 0);
> > +
> > + mutex_unlock(&phydev->lock);
>
> What is the purpose of taking this lock? phy_modify_paged() will do its
> read-modify-write access and page accesses under the MDIO bus lock,
> which should be all that's required to guarantee an atomic update.
Yes, I copied this from the vsc85xx_rgmii_set_skews() function in the
same driver accessing the same register in the same context.
But maybe it is used there because of repeated phydev->interface
accesses, which may otherwise change during function execution?
I'll remove the locks from my patch.
Thanks for your feedback.
Powered by blists - more mailing lists