[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <01cf4367-1c2d-4920-a387-3ba38b83dd66@lunn.ch>
Date: Wed, 16 Aug 2023 22:13:10 +0200
From: Andrew Lunn <andrew@...n.ch>
To: Serge Semin <fancer.lancer@...il.com>
Cc: Heiner Kallweit <hkallweit1@...il.com>,
Russell King <linux@...linux.org.uk>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
Francesco Dolcini <francesco.dolcini@...adex.com>,
netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC net] Revert "net: phy: Fix race condition on link status
change"
> > So i suggest you change phy_process_error() to remove the lock.
>
> This doable.
>
> > Maybe
> > add a test to ensure the lock is actually held, and do a phydev_err()
> > if not.
>
> This can't be done since phy_state_machine() calls phy_error_precise()
> which calls phy_process_error() with no phy_device.lock held. Printing the
> error in that case would mean an error in the Networking PHY subsystem
> itself.
>
> Do you suggest to take the lock before calling phy_error_precise() then?
Thanks for digging into the details.
phy_error_precise() is used in exactly one place. So i would actually
put the lock inside it. And maybe move the comment about not using the
function with the lock already held here :-)
Andrew
Powered by blists - more mailing lists