Date:	Fri, 17 May 2013 00:56:41 +0200
From:	Tomasz Figa <tomasz.figa@...il.com>
To:	Heiko Stübner <heiko@...ech.de>
Cc:	Doug Anderson <dianders@...omium.org>,
	Kukjin Kim <kgene.kim@...sung.com>,
	Olof Johansson <olof@...om.net>,
	Stephen Warren <swarren@...dotorg.org>,
	Thomas Abraham <thomas.abraham@...aro.org>,
	Linus Walleij <linus.walleij@...aro.org>,
	Prathyush K <prathyush.k@...sung.com>,
	linux-samsung-soc <linux-samsung-soc@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/2] pinctrl: samsung: fix suspend/resume functionality

On Friday 17 of May 2013 00:30:38 Heiko Stübner wrote:
> Am Freitag, 17. Mai 2013, 00:08:34 schrieb Tomasz Figa:
> > On Thursday 16 of May 2013 14:51:53 Doug Anderson wrote:
> > > Tomasz,
> > > 
> > > On Thu, May 16, 2013 at 2:27 PM, Tomasz Figa <tomasz.figa@...il.com>
> > 
> > wrote:
> > > > OK. I will be fine to go with your patches, after addressing the
> > > > comments. In the end it's good that you posted them, as reviewing
> > > > them allowed me to find even better ways of doing some things than
> > > > I
> > > > had in mine ;) .
> > > 
> > > Yes.  I often find that the best way to review code is to think
> > > about
> > > how I would implement it myself.  Certainly I think we've ended up
> > > with something better / less buggy this way.  ;)
> > > 
> > > > How all of this works is basically a good question. I couldn't
> > > > find
> > > > any
> > > > mention about pins switching from power down to normal mode in the
> > > > documentation, but maybe there is a small side note somewhere,
> > > > which I
> > > > could miss.
> > > > 
> > > > On S3C6410, for example, there are two modes. State is switched to
> > > > power down mode automatically, but can be switched out either
> > > > automatically on wake-up (exact timing is unknown to me) or by
> > > > clearing a special bit, depending on value of other special bit.
> > > > 
> > > > IMHO this is rather important, so we should find out how it work
> > > > on
> > > > other SoCs and make the code account for it.
> > > 
> > > Agreed that it's important.  ...but it's also good not to have tons
> > > of
> > > complexity when it's not needed.  It sounds like S3C6410 could be
> > > handled OK by just using the special bits and waiting to take things
> > > out of power down mode.
> > > 
> > > ...thinking about it, all SoCs that have power down modes (which you
> > > _must_ have if your pinctrl state is lost across a low power) would
> > > be
> > > slightly broken if they didn't have a bit to switch out of power
> > > down
> > > mode.  Otherwise you're asking for at least some type of glitch
> > > because you'll end up in the default state of pins for a little
> > > while
> > > during resume.
> > > 
> > > That's not to say that there aren't broken SoCs out there and it's
> > > entirely possible that people even designed systems around them
> > > (knowing that the default state of each pin after wakeup is not
> > > harmful to whatever is connected to that pin).  If there are any
> > > cases
> > > like this then they would need the special code like my V1 patch
> > > had.
> > > Do you know of any SoCs like this that we need to support on kernel
> > > 3.10 and higher?
> > 
> > Hmm, I just checked documentation of S3C2440 and S3C2416 they seem to
> > retain GPIO settings completely in sleep mode. This would mean that
> > they don't require any suspend/resume support in pinctrl driver.
> > Heiko, can you confirm this?
> 
> Hmm, my system does not have a working suspend right now, but looking at
> the legacy code (mach-s3c24xx/pm.c, etc) tells me that the gpio banks
> were never saved during suspend.
> 
> And as there were (and still are) systems with working suspend around,
> I'd assume that you're correct that the pins retain their state.
> 
> Is the same true for the s3c64xx, as I didn't find any gpio suspend
> handling for it either.

Seems like I need some sleep, as I'm already starting to overlook large 
blobs of code. 

Originally, GPIO suspend/resume handlers have been configured in 
drivers/gpio/gpio-samsung.c, by setting pm field of samsung_gpio_chip 
struct to point to appropriate samsung_gpio_pm struct, which contains 
pointers to save and resume callbacks.

In result, samsung_gpio_pm_2bit_* or samsung_gpio_pm_4bit_* have been 
used, depending on bank type, on all SoCs.

Now since the documentation states that wake-up reset doesn't reset GPIO 
registers (at least on S3C2440 and S3C2416), I wonder what is the correct 
way of handling them.

Best regards,
Tomasz

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists