linux-kernel - Re: Why do we need reset_control_get

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 08 Aug 2016 19:29:24 +0200
From:	Philipp Zabel <p.zabel@...gutronix.de>
To:	Arnd Bergmann <arnd@...db.de>
Cc:	Masahiro Yamada <yamada.masahiro@...ionext.com>,
	Axel Lin <axel.lin@...ics.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Hans de Goede <hdegoede@...hat.com>,
	Maxime Ripard <maxime.ripard@...e-electrons.com>,
	Lee Jones <lee.jones@...aro.org>,
	linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>
Subject: Re: Why do we need reset_control_get_optional() ?

Am Freitag, den 05.08.2016, 17:35 +0200 schrieb Arnd Bergmann:
> On Friday, August 5, 2016 10:55:58 AM CEST Philipp Zabel wrote:
> > Am Samstag, den 30.07.2016, 22:13 +0200 schrieb Arnd Bergmann:
> > > On Friday, July 29, 2016 3:08:15 PM CEST Philipp Zabel wrote:
> > > > Hi Masahiro,
> > > > 
> > > > Am Donnerstag, den 28.07.2016, 19:29 +0900 schrieb Masahiro Yamada:
> > > > [...]
> > > > > However, I think the following makes more sense:
> > > > > 
> > > > > 
> > > > > menuconfig RESET_CONTROLLER
> > > > >         bool "Reset Controller Support"
> > > > >         depends on (ARCH_HAS_RESET_CONTROLLER || COMPILE_TEST)
> > > > >         default y
> > > > >         help
> > > > >           Generic Reset Controller support.
> > > > 
> > > > That looks sensible to me. You'll only have to enable the reset
> > > > controller framework if either some enabled architecture has a reset
> > > > controller (in which case you want the driver for it to be activated by
> > > > default), or if you want to compile test some of the reset drivers.
> > > 
> > > This still doesn't let a platform 'select RESET_FOO', unless they
> > > also select RESET_CONTROLLER and ARCH_HAS_RESET_CONTROLLER.
> > > 
> > > Why do we need to guard all drivers inside of two symbols?
> > 
> > Does the platform have to select RESET_FOO at all? Wouldn't it be enough
> > for RESET_FOO to have "default ARCH_FOO" ?
> 
> It depends on what you want to achieve. With a user-visible option
> and "default ARCH_FOO", you can disable the driver manually, and
> another driver that has "depends on ARCH_FOO" can not rely on this
> one being present as it currently can.
>
> If we do this as
> 
> config RESET_FOO
> 	bool "FOO reset controller" if COMPILE_TEST && !ARCH_FOO
> 	default ARCH_FOO
> 
> then I think we get both: you won't be able to turn it off
> but also get the build testing.

I like it. Automatically enable the mandatory reset controller driver
with its architecture, otherwise don't ask unless COMPILE_TEST is
enabled. For drivers that can be reasonably compiled as a module, it
could be

config RESET_BAR
	tristate "BAR reset controller" if COMPILE_TEST || ARCH_FOO
	default ARCH_FOO

> > Currently ARCH_HAS_RESET_CONTROLLER is used to default y the
> > RESET_CONTROLLER symbol. Maybe we should add another
> > ARCH_REQUIRE_RESET_CONTROLLER and have that select RESET_CONTROLLER,
> > similarly to how it is done for GPIOLIB?
> 
> GPIOLIB just stopped using it, there is now only CONFIG_GPIOLIB
> that can get selected by platforms that need it.

Ok.

[...]
> I never really like the way it was done for gpiolib. I think the
> easiest way would be to have a menu for the reset controllers that
> does not have any dependencies whatsoever, and make the individual
> reset drivers select CONFIG_RESET_CONTROLLER, which then becomes
> a hidden symbol that enables the core code.

I thought making all the driver config symbols visible by default is
problematic, but with "bool ... if COMPILE_TEST" that is addressed
already.

regards
Philipp