[0/8] of: automate of_node_put() - new approach to loops.

Message ID 20240211174237.182947-1-jic23@kernel.org
Headers
Series of: automate of_node_put() - new approach to loops. |

Message

Jonathan Cameron Feb. 11, 2024, 5:42 p.m. UTC
  From: Jonathan Cameron <Jonathan.Cameron@huawei.com>

Since RFC:
- Provide a for_each_available_child_of_node_scoped() variant and
  use that whenever we aren't specifically trying to include disabled
  nodes.
- Fix the for_each_child_of_node_scoped() to not use a mix of
  _available_ and other calls.
- Include a few more examples.  The last one is there to show that
  not all uses of the __free(device_node) call are due to the loops.

Thanks to Julia Lawal who also posted coccinelle for both types (loop and
non loop cases)

https://lore.kernel.org/all/alpine.DEB.2.22.394.2401312234250.3245@hadrien/
https://lore.kernel.org/all/alpine.DEB.2.22.394.2401291455430.8649@hadrien/

The cover letter of the RFC includes information on the various approaches
considered.
https://lore.kernel.org/all/20240128160542.178315-1-jic23@kernel.org/

Whilst these macros profduce nice reductions in complexity the loops
still have the unfortunate side effect of hiding the local declaration
of a struct device_node * which is then used inside the loop.

Julia suggested making that a little more visible via
 #define for_each_child_of_node_scoped(parent, struct device_node *, child)
but in discussion we both expressed that this doesn't really make things
all that clear either so I haven't adopted this suggestion.

If the responses to this series are positive I can put the first few
patches on an immutable branch, allowing rapid adoption in other trees
if people want to move quickly. If not we can wait for next cycle and
just take this infrastructure through the IIO tree ready for the 6.9
merge cycle.

I'll be optimistic that we are converging and send out an equivalent
series for fwnode_handle / property.h to replace the previous proposal:
https://lore.kernel.org/all/20240114172009.179893-1-jic23@kernel.org/


Jonathan Cameron (8):
  of: Add cleanup.h based auto release via __free(device_node) markings.
  of: Introduce for_each_*_child_of_node_scoped() to automate
    of_node_put() handling
  of: unittest: Use for_each_child_of_node_scoped()
  iio: adc: fsl-imx25-gcq: Use for_each_available_child_node_scoped()
  iio: adc: rcar-gyroadc: use for_each_available_child_node_scoped()
  iio: adc: ad7124: Use for_each_available_child_of_node_scoped()
  iio: adc: ad7292: Use for_each_available_child_of_node_scoped()
  iio: adc: adi-axi-adc: Use __free(device_node) and guard(mutex)

 drivers/iio/adc/ad7124.c        | 20 ++++++--------------
 drivers/iio/adc/ad7292.c        |  7 ++-----
 drivers/iio/adc/adi-axi-adc.c   | 16 ++++------------
 drivers/iio/adc/fsl-imx25-gcq.c | 13 +++----------
 drivers/iio/adc/rcar-gyroadc.c  | 21 ++++++---------------
 drivers/of/unittest.c           | 11 +++--------
 include/linux/of.h              | 15 +++++++++++++++
 7 files changed, 39 insertions(+), 64 deletions(-)
  

Comments

Andy Shevchenko Feb. 12, 2024, 12:03 p.m. UTC | #1
On Sun, Feb 11, 2024 at 05:42:28PM +0000, Jonathan Cameron wrote:
> From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> 
> Since RFC:
> - Provide a for_each_available_child_of_node_scoped() variant and
>   use that whenever we aren't specifically trying to include disabled
>   nodes.
> - Fix the for_each_child_of_node_scoped() to not use a mix of
>   _available_ and other calls.
> - Include a few more examples.  The last one is there to show that
>   not all uses of the __free(device_node) call are due to the loops.

I'm a bit skeptical about need of this work. What I would prefer to see
is getting rid of OF-centric drivers in IIO. With that, we would need
only fwnode part to be properly implemented.
  
Jonathan Cameron Feb. 16, 2024, 2:47 p.m. UTC | #2
On Mon, 12 Feb 2024 14:03:29 +0200
Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote:

> On Sun, Feb 11, 2024 at 05:42:28PM +0000, Jonathan Cameron wrote:
> > From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> > 
> > Since RFC:
> > - Provide a for_each_available_child_of_node_scoped() variant and
> >   use that whenever we aren't specifically trying to include disabled
> >   nodes.
> > - Fix the for_each_child_of_node_scoped() to not use a mix of
> >   _available_ and other calls.
> > - Include a few more examples.  The last one is there to show that
> >   not all uses of the __free(device_node) call are due to the loops.  
> 
> I'm a bit skeptical about need of this work. What I would prefer to see
> is getting rid of OF-centric drivers in IIO. With that, we would need
> only fwnode part to be properly implemented.
> 

To be honest main reason for doing of first was that they have unit tests :)

The IIO drivers were more of a proving ground than cases I really cared
out cleaning up.  However I'm always of the view that better to make
some improvement now than wait for a perfect improvement later.
 
However one or two are not going to be converted to fwnode handling
any time soon because they make use of phandle based referencing for
driver specific hook ups that isn't going to get generic handling any
time soon.

I'll probably focus on getting the fwnode version of this moving
forwards first though and 'maybe' convert a few of the easier ones
of these over to that framework to reduce how many users of this
we end up with in IIO.

Jonathan
  
Andy Shevchenko Feb. 16, 2024, 3:25 p.m. UTC | #3
On Fri, Feb 16, 2024 at 02:47:56PM +0000, Jonathan Cameron wrote:
> On Mon, 12 Feb 2024 14:03:29 +0200
> Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote:
> > On Sun, Feb 11, 2024 at 05:42:28PM +0000, Jonathan Cameron wrote:

..

> > I'm a bit skeptical about need of this work. What I would prefer to see
> > is getting rid of OF-centric drivers in IIO. With that, we would need
> > only fwnode part to be properly implemented.
> 
> To be honest main reason for doing of first was that they have unit tests :)

fwnode also has KUnit test. Have you considered adding test cases there?

> The IIO drivers were more of a proving ground than cases I really cared
> out cleaning up.  However I'm always of the view that better to make
> some improvement now than wait for a perfect improvement later.

Yes, but in my opinion _in this particular case_ it brings more churn and
some maybe even not good from educational purposes, i.e. one can look at
the current series and think "oh, OF is still in use, let me provide my
driver OF-only (for whatever reasons behind)", while targeting conversion
first will tell people: "hey, there is an agnostic device property framework
that should be used in a new code and that's why we have been converting old
drivers too".

> However one or two are not going to be converted to fwnode handling
> any time soon because they make use of phandle based referencing for
> driver specific hook ups that isn't going to get generic handling any
> time soon.

Sure, exceptions happen.

> I'll probably focus on getting the fwnode version of this moving
> forwards first though and 'maybe' convert a few of the easier ones
> of these over to that framework to reduce how many users of this
> we end up with in IIO.

Thanks!