From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew Lunn <andrew@lunn.ch>
Subject: Re: [PATCH v2] net: phy: mdio-gpio: fix access that may sleep
Date: Wed, 14 Nov 2018 21:46:54 +0100
Message-ID: <20181114204654.GB32305@lunn.ch>
References: <20181114061703.11026-1-ms@dev.tdt.de>
 <20181114063703.13379-1-ms@dev.tdt.de>
 <20181114070528.GD17052@lunn.ch>
 <248deac95f412b925b0de44ead2ffec6@dev.tdt.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        f.fainelli@gmail.com, davem@davemloft.net
To: Martin Schiller <ms@dev.tdt.de>
Return-path: <linux-kernel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <248deac95f412b925b0de44ead2ffec6@dev.tdt.de>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: netdev.vger.kernel.org

On Wed, Nov 14, 2018 at 08:43:49AM +0100, Martin Schiller wrote:
> On 2018-11-14 08:05, Andrew Lunn wrote:
> >On Wed, Nov 14, 2018 at 07:37:03AM +0100, Martin Schiller wrote:
> >>This commit re-enables support for slow GPIO pins. It was initially
> >>introduced by commit
> >>	2d6c9091ab7630dfcf34417c6683ce4764d7d40a
> >>and got lost by commit
> >>	7e5fbd1e0700f1bdb94508f84ec2aeb01eed7b12
> >
> >Hi Martin
> >
> >Was it really lost? It looks like _cansleep() just adds an extra check
> >might_sleep_if(extra_checks), but it does not change any
> >functionality.
> 
> Well, you are right, the functionality itself is not broken, but using
> the NON _cansleep() functions on GPIOs that have the cansleep flag set,
> this leads to a lot of kernel warnings/backtraces which makes the system
> in fact useless.
> 
> Thats the WARN_ON() here:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpio/gpiolib.c?h=v4.20-rc2#n2992
> 
> and here:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpio/gpiolib.c?h=v4.20-rc2#n3304

Hi Martin

Right, this is really what you are fixing. So the commit message
should talk about these WARN_ON being hit.

> >
> >So the change itself is O.K, i'm just not too sure about the commit
> >message.
> >
> >	Andrew
> 
> Hmm, ok. What would you suggest for a better commit message?

Up until commit XXX, the _cansleep variants of the gpio_ API was
used. After that commit and the change to gpiod_ API, the _cansleep()
was dropped. This then results in WARN_ON() when used with GPIO
devices which do sleep. Add back the _cansleep() to avoid this.

This is something which is useful in stable. So please base this fix
on DaveM net tree, not net-next, and include a Fixes: tag for the
patch which broke it.

      Thanks
	Andrew