Re: [PATCH] nvmem: fix unregistering device in nvmem_register() error path

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, Dec 22, 2021 at 10:00:03AM +0100, Rafał Miłecki wrote:
> On 22.12.2021 09:38, Johan Hovold wrote:
> > On Wed, Dec 22, 2021 at 08:44:44AM +0100, Greg Kroah-Hartman wrote:
> > > On Tue, Dec 21, 2021 at 06:46:01PM +0100, Rafał Miłecki wrote:
> > > > On 21.12.2021 17:06, Greg Kroah-Hartman wrote:
> > > > > On Tue, Dec 21, 2021 at 04:45:50PM +0100, Rafał Miłecki wrote:
> > > > > > From: Rafał Miłecki <rafal@milecki.pl>
> > > > > > 
> > > > > > 1. Drop incorrect put_device() calls
> > > > > > 
> > > > > > If device_register() fails then underlaying device_add() takes care of
> > > > > > calling put_device() if needed. There is no need to do that in a driver.
> > > > > 
> > > > > Did you read the documentation for device_register() that says:
> > > > > 
> > > > >    * NOTE: _Never_ directly free @dev after calling this function, even
> > > > >    * if it returned an error! Always use put_device() to give up the
> > > > >    * reference initialized in this function instead.
> > > > 
> > > > I clearly tried to be too smart and ignored documentation.
> > > > 
> > > > I'd say device_add() behaviour is rather uncommon and a bit unintuitive.
> > > > Most kernel functions are safe to assume to do nothing that requires
> > > > cleanup if they fail.
> > > > 
> > > > E.g. if I call platform_device_register() and it fails I don't need to
> > > > call anything like platform_device_put(). I just free previously
> > > > allocated memory.
> > > 
> > > And that is wrong.
> > 
> > It seems Rafał is mistaken here too; you certainly need to call
> > platform_device_put() if platform_device_register() fail, even if many
> > current users do appear to get this wrong.
> 
> Yes I was! Gosh I made up that "platform_device_put()" name and only
> now I realized it actually exists!
> 
> I stand by saying this design is really misleading. Even though
> platform_device_put() was obviously a bad example.
> 
> Please remember I'm just a minor kernel developer however in my humble
> opinion behaviour of device_register() and platform_device_register()
> should be changed.
> 
> If any function fails I expect:
> 1. That function to clean up its mess if any
> 2. Me to be responsible to clean up my mess if any
> 
> This is how "most" code (whatever it means) works.
> 1. If POSIX snprintf() fails I'm not expected to call *printf_put() sth
> 2. If POSIX bind() fails I'm not expected to call bind_put() sth
> 3. (...)
> 
> I'm not sure if those are the best examples but you should get my point.

I do understand, and for platform_device_register() I agree with you.

But for device_register() we can not do this as the driver core is not
the "owner" of the structure being passed into it.  If you call
device_register() you are bus and you have to know how to handle an
error here as there is usually much more that needs to be done that a
device_put() can not do by the core.

Yes, it's well down on the "Rusty's API usability scale", but it is
documented well and in a number of places for device_register().

platform_device_register() is not documented, and that's not good, so we
should fix it up.  Although there's the larger issue of everyone using
static 'struct device' for this which is yet-another-reason I hate the
platform device code.

thanks,

greg k-h