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  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date:   Sun, 29 Jan 2017 23:43:52 +0100
From:   Sebastian Reichel <sre@...nel.org>
To:     Guenter Roeck <linux@...ck-us.net>
Cc:     Thierry Reding <thierry.reding@...il.com>,
        Laxman Dewangan <ldewangan@...dia.com>,
        Martin Michlmayr <tbm@...ius.com>, linux-pm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] power: reset: Add MAX77620 support

Hi,

On Sun, Jan 29, 2017 at 12:47:57PM -0800, Guenter Roeck wrote:
> On 01/29/2017 12:02 PM, Sebastian Reichel wrote:
> 
> > > 
> > > To keep things simple, I think it would be okay to allow only one of
> > > each type of controller in any running system. It's very unlikely that
> > > board designers would devise two different ways of powering off or
> > > restarting a system, while in a similar way an SoC or CPU would only
> > > ever provide one way to do so. Even if theoretically multiple
> > > possibilities exist, I think the board code should pick which ones are
> > > appropriate.
> > 
> > Using that logic we may also advice, that board-code should only
> > register the board-level reset/poweroff and it's enough to have
> > a callback again... I wonder if that is really feasible.
> > 
> 
> FWIW, it is also not true.

It seems this was misunderstood. I do not expect this to work.

> There is a reason why many of the restart handlers used to have
> code saying "install restart handler, but only if none is
> installed yet". Which of course is racy, and gets more interesting
> if the restart handler installed first is unloaded at a later
> time, leaving the system with no restart handler. Or both are
> unloaded, leaving the system with a pointer to a no longer
> existing handler.
>
> One could then argue that anything implementing a restart handler must
> not unload. Which results in more restrictions. And drivers loaded
> on hardware which don't need it. And more corner cases to deal with.
> And more inconsistencies.
> 
> In reality, many systems or system variants will have more than one means
> to restart it. Yes, board designers do devise multiple ways of powering off
> or restarting a system. There may be and likely are valid reasons for doing
> so; I would not want to claim or suggest that board designers would design
> such hardware without reason. Even "standard" PCs tend to have have more
> than one means to reset it. There _was_ a reason for introducing that
> framework; I didn't just do it for fun.
> 
> However, as I had mentioned before, I am not really interested in this
> topic anymore. Just treat this as my final word of caution, or feel free
> to ignore it. I hope you'll find a much better solution than mine
> to implement "the board code should pick which ones are appropriate".

In case I was unclear: I'm fine with the current state of reboot
code using notifier chain and really thankful for the work. IMHO
it improved the status-quo a lot.

However I'm not fine with the current poweroff stuff and if somebody
offers to implement a solution compatible with Linus (and other people,
which disliked the notifier chain approach): Thanks, please do!

-- Sebastian

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)

Powered by blists - more mailing lists