Re: Re: [PATCH 0/9] added helper macros to remove duplicate code from probe functions of the platform drivers

From: Vinod Koul
Date: Tue Sep 24 2019 - 15:28:36 EST


On 21-09-19, 20:27, Satendra Singh Thakur wrote:
> On Wed, Sep 18, 2019 at 3:57 PM, Vinod Koul wrote:
> > On 15-09-19, 12:30, Satendra Singh Thakur wrote:
> > > 1. For most of the platform drivers's probe include following steps
> > >
> > > -memory allocation for driver's private structure
> > > -getting io resources
> > > -io remapping resources
> > > -getting irq number
> > > -registering irq
> > > -setting driver's private data
> > > -getting clock
> > > -preparing and enabling clock
> >
> > And we have perfect set of APIs for these tasks!
> Hi Vinod,
> Thanks for the comments.
> You are right, we already have set of APIs for these tasks.
> The proposed macros combine the very same APIs to remove
> duplicate/redundant code.
> A new driver author can use these macros to easily write probe

Nope they can write each APIs, know the exact flow and do it!

> function without having to worry about signatures of internal APIs.
> In the past, people have combined some of them e.g.
> a) clk_prepare_enable combines clk_prepare and clk_enable

As it is clock, it is called in sequence, whereas different drivers may
have different sequencing.

Btw I am not for adding maanged irq functions, they are really not the
correct way to manage!

> b) devm_platform_ioremap_resource combines
> platform_get_resource (for type IORESOURCE_MEM)
> and devm_ioremap_resource
> c) module_platform_driver macro encompasses module_init/exit
> and driver_register/unregister functions.

All examples you are quoting do a set of functions (clk, resources etc
and not do N things!

> The basic idea is to simplyfy coding.
> > > 2. We have defined a set of macros to combine some or all of
> > > the above mentioned steps. This will remove redundant/duplicate
> > > code in drivers' probe functions of platform drivers.
> >
> > Why, how does it help and you do realize it also introduces bugs
> This will make probe function shorter by removing repeated code.
> This will also reduce bugs caused due to improper handling
> of failure cases because of these reasons:
> a) If the developer calls each API individualy one might miss
> proper handling of the failure for some API; Whereas the macro
> properly handles failure of each API.
> b) Macros are devres compatible which leaves less room for
> memory leaks.

No we review the code and if we miss we fix later!

> Yes, the macros which enable clock and request irqs
> might cause bugs if they are not used carefully.
> For instance, enabling the clock or requesting the irq
> earlier than actually required. So, the macros with _clk
> and _irq, _all suffix should be used carefully.

Precisely!


> Please let me know if I miss any specific type of bug
> here.
> >
> > > devm_platform_probe_helper(pdev, priv, clk_name);
> > > devm_platform_probe_helper_clk(pdev, priv, clk_name);
> > > devm_platform_probe_helper_irq(pdev, priv, clk_name,
> > > irq_hndlr, irq_flags, irq_name, irq_devid);
> > > devm_platform_probe_helper_all(pdev, priv, clk_name,
> > > irq_hndlr, irq_flags, irq_name, irq_devid);
> > > devm_platform_probe_helper_all_data(pdev, priv, clk_name,
> > > irq_hndlr, irq_flags, irq_name, irq_devid);
> > >
> > > 3. Code is made devres compatible (wherever required)
> > > The functions: clk_get, request_irq, kzalloc, platform_get_resource
> > > are replaced with their devm_* counterparts.
> >
> > We already have devres APIs for people to use!
> Yes, we have devres APIs and many drivers do use them.
> But still there are many which don't use them.
> The proposed macros provides just another way to use devres APIs.
> > >
> > > 4. Few bugs are also fixed.
> >
> > Which ones..?
> The bug is that the failure of request_irq
> is not handled properly in mtk-hsdma.c. This is fixed in patch [5/9].
> https://lkml.org/lkml/2019/9/15/35

Please send the fix individually and I would be glad to review.

--
~Vinod