linux-kernel - Re: [PATCH 0/2] DS1374 Watchdog fixes

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]

Message-ID: <20170425161743.GA8443@roeck-us.net>
Date:   Tue, 25 Apr 2017 09:17:43 -0700
From:   Guenter Roeck <linux@...ck-us.net>
To:     Moritz Fischer <moritz.fischer.private@...il.com>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Moritz Fischer <moritz.fischer@...us.com>,
        linux-watchdog@...r.kernel.org, wim@...ana.be,
        a.zummo@...ertech.it, alexandre.belloni@...e-electrons.com,
        rtc-linux@...glegroups.com, alex.williams@...com
Subject: Re: [PATCH 0/2] DS1374 Watchdog fixes

On Tue, Apr 25, 2017 at 07:55:28AM -0700, Moritz Fischer wrote:
> Hi Guenter,
> 
> On Mon, Apr 24, 2017 at 10:03 PM, Guenter Roeck <linux@...ck-us.net> wrote:
> > On 04/24/2017 03:05 PM, Moritz Fischer wrote:
> 
> >> I'm very unhappy with the CONFIG_DRV_RTC_DS1374_WDT way of enabling
> >> the watchdog behavior and currently I'm investigating how to make
> >> that work via DT.
> >>
> >> Watchdog maintainers, do you have an idea on how to do that in a
> >> non breaking fashion?
> >>
> >
> > Depends on what you mean with "non breaking". Just using the normal mfd
> > mechanisms, ie define an mfd cell for each client driver, should work.
> > Do you see any problems with that ? Either case, that doesn't seem
> > to be a watchdog driver problem, or am I missing something ?
> 
> Well so currently watchdog behavior is selected (out of the two options alarm,
> or watchdog) by enabling the configuration option mentioned above.
> If I change this over to use a dt-based approach like dallas,ds1374-mode = <2>;
> to select the behavior in the mfd for example, won't that break people that
> relied on the old behavior? If everyone involved is ok with that, I'm happy
> to just add it to the binding.
> 

Sorry, I must be missing something. Looking into the driver code, my
understanding is that CONFIG_RTC_DRV_DS1374_WDT enables the watchdog in
addition to rtc functionality, not one or the other. Sure you would need
a different configuration option if you were to move the watchdog code into
drivers/watchdog, but other than that I don't really understand the problem.
What is the issue with, for example,

config DS1374_WDT
        bool "Dallas/Maxim DS1374 watchdog timer"
	depends on MFD_DS1374
	help
	  If you say Y here you will get support for the
	  watchdog timer in the Dallas Semiconductor DS1374
	  real-time clock chips.

in drivers/watchdog/Kconfig, and the mfd driver instantiating it like
any other mfd client driver ?

Either case, limiting support to DT based systems seems to be the wrong
approach. There might be Intel platforms using this chip.

> > I don't really see the point of doing that if you plan to move the watchdog
> > part of the driver into the watchdog directory. We for sure won't accept a
> > watchdog driver that does not use the watchdog infrastructure.
> 
> The idea was to fix what's broken currently (this patchset) and then refactor.
> But if you prefer I can do all in one go instead.
> 

It just seemed a waste to me to change/fix a function which is going to
be removed in a subsequent patch (I seem to recall that there was a fix
to the ioctl function).

If/when you move the driver to drivers/watchdog, please make sure that
it doesn't use any instantiation related static variables (ie other than
module parameters).

> >
> > Regarding
> > +       /* WHY? */
> > +       ds1374->wdd.timeout = t;
> >
> > Assuming you mean why the driver has to set the timeout value - not every
> > watchdog hardware supports timeouts in multiples of 1 second. The driver
> > is expected to set the value to the real timeout, not to the timeout asked
> > for by the infrastructure.
> 
> Yeah that branch is work in progress and needs cleanup. Leftover from testing,
> before I had understood why. Branch needs cleanup. It also doesn't really use
> regmap_update_bits etc.
> 

Well, you did enhance the code to use regmap, which by itself is a significant
improvement ...

Thanks,
Guenter