From mboxrd@z Thu Jan 1 00:00:00 1970 From: Lee Jones Subject: Re: [PATCH v7 3/5] clk: Supply the critical clock {init, enable, disable} framework Date: Thu, 30 Jul 2015 10:50:14 +0100 Message-ID: <20150730095014.GD14642@x1> References: <1437570255-21049-1-git-send-email-lee.jones@linaro.org> <1437570255-21049-4-git-send-email-lee.jones@linaro.org> <20150727072549.GP2564@lukather> <20150727085338.GW3436@x1> <20150728114022.GW2564@lukather> <20150728130055.GV14943@x1> <20150730011932.642.85168@quantum> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Content-Disposition: inline In-Reply-To: <20150730011932.642.85168@quantum> Sender: devicetree-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Michael Turquette Cc: Maxime Ripard , linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, kernel-F5mvAk5X5gdBDgjK7y7TUQ@public.gmane.org, sboyd-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org, devicetree-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, geert-Td1EMuHUCqxL1ZNQvxDV9g@public.gmane.org, s.hauer-bIcnvbaLZ9MEGnE8C9+IrQ@public.gmane.org List-Id: devicetree@vger.kernel.org On Wed, 29 Jul 2015, Michael Turquette wrote: > Quoting Lee Jones (2015-07-28 06:00:55) > > On Tue, 28 Jul 2015, Maxime Ripard wrote: > >=20 > > > On Mon, Jul 27, 2015 at 09:53:38AM +0100, Lee Jones wrote: > > > > On Mon, 27 Jul 2015, Maxime Ripard wrote: > > > >=20 > > > > > On Wed, Jul 22, 2015 at 02:04:13PM +0100, Lee Jones wrote: > > > > > > These new API calls will firstly provide a mechanisms to ta= g a clock as > > > > > > critical and secondly allow any knowledgeable driver to (un= )gate clocks, > > > > > > even if they are marked as critical. > > > > > >=20 > > > > > > Suggested-by: Maxime Ripard > > > > > > Signed-off-by: Lee Jones > > > > > > --- > > > > > > drivers/clk/clk.c | 45 ++++++++++++++++++++++++= ++++++++++++++++++++ > > > > > > include/linux/clk-provider.h | 2 ++ > > > > > > include/linux/clk.h | 30 ++++++++++++++++++++++++= +++++ > > > > > > 3 files changed, 77 insertions(+) > > > > > >=20 > > > > > > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c > > > > > > index 61c3fc5..486b1da 100644 > > > > > > --- a/drivers/clk/clk.c > > > > > > +++ b/drivers/clk/clk.c > > > > > > @@ -46,6 +46,21 @@ static struct clk_core *clk_core_lookup(= const char *name); > > > > > > =20 > > > > > > /*** private data structures ***/ > > > > > > =20 > > > > > > +/** > > > > > > + * struct critical - Provides 'play' over critical clock= s. A clock can be > > > > > > + * marked as critical, meaning that it= should not be > > > > > > + * disabled. However, if a driver whi= ch is aware of the > > > > > > + * critical behaviour wants to control= it, it can do so > > > > > > + * using clk_enable_critical() and clk= _disable_critical(). > > > > > > + * > > > > > > + * @enabled Is clock critical? Once set, doesn't chang= e > > > > > > + * @leave_on Self explanatory. Can be disabled by knowl= edgeable drivers > > > > > > + */ > > > > > > +struct critical { > > > > > > + bool enabled; > > > > > > + bool leave_on; > > > > > > +}; > > > > > > + > > > > > > struct clk_core { > > > > > > const char *name; > > > > > > const struct clk_ops *ops; > > > > > > @@ -75,6 +90,7 @@ struct clk_core { > > > > > > struct dentry *dentry; > > > > > > #endif > > > > > > struct kref ref; > > > > > > + struct critical critical; > > > > > > }; > > > > > > =20 > > > > > > struct clk { > > > > > > @@ -995,6 +1011,10 @@ static void clk_core_disable(struct c= lk_core *clk) > > > > > > if (WARN_ON(clk->enable_count =3D=3D 0)) > > > > > > return; > > > > > > =20 > > > > > > + /* Refuse to turn off a critical clock */ > > > > > > + if (clk->enable_count =3D=3D 1 && clk->critical.lea= ve_on) > > > > > > + return; > > > > > > + > > > > >=20 > > > > > I think it should be handled by a separate counting. Otherwis= e, if you > > > > > have two users that marked the clock as critical, and then on= e of them > > > > > disable it... > > > > >=20 > > > > > > if (--clk->enable_count > 0) > > > > > > return; > > > > > > =20 > > > > > > @@ -1037,6 +1057,13 @@ void clk_disable(struct clk *clk) > > > > > > } > > > > > > EXPORT_SYMBOL_GPL(clk_disable); > > > > > > =20 > > > > > > +void clk_disable_critical(struct clk *clk) > > > > > > +{ > > > > > > + clk->core->critical.leave_on =3D false; > > > > >=20 > > > > > .. you just lost the fact that it was critical in the first p= lace. > > > >=20 > > > > I thought about both of these points, which is why I came up wi= th this > > > > strategy. > > > >=20 > > > > Any device which uses the *_critical() API should a) have knowl= edge of > > > > what happens when a particular critical clock is gated and b) h= ave > > > > thought about the consequences. > > >=20 > > > Indeed. > > >=20 > > > > I don't think we can use reference counting, because we'd need = as > > > > many critical clock owners as there are critical clocks. > > >=20 > > > Which we can have if we replace the call to clk_prepare_enable yo= u add > > > in your fourth patch in __set_critical_clocks. > >=20 > > What should it be replaced with? > >=20 > > > > Cast your mind back to the reasons for this critical clock API.= One > > > > of the most important intentions of this API is the requirement > > > > mitigation for each of the critical clocks to have an owner > > > > (driver). > > > >=20 > > > > With regards to your second point, that's what 'critical.enable= d' > > > > is for. Take a look at clk_enable_critical(). > > >=20 > > > I don't think this addresses the issue, if you just throw more > > > customers at it, the issue remain with your implementation. > > >=20 > > > If you have three customers that used the critical API, and if on= of > > > these calls clk_disable_critical, you're losing leave_on. > >=20 > > That's the idea. See my point above, the one you replied "Indeed" > > to. So when a driver uses clk_disable_critical() it's saying, "I k= now > > why this clock is a critical clock, and I know that nothing terribl= e > > will happen if I disable it, as I have that covered". So then if i= t's > > not the last user to call clk_disable(), the last one out the door > > will be allowed to finally gate the clock, regardless whether it's > > critical aware or not. > >=20 > > Then, when we come to enable the clock again, the critical aware us= er > > then re-marks the clock as leave_on, so not critical un-aware user = can > > take the final reference and disable the clock. > >=20 > > > Which means that if there's one of the two users left that calls > > > clk_disable on it, the clock will actually be disabled, which is > > > clearly not what we want to do, as we have still a user that want= the > > > clock to be enabled. > >=20 > > That's not what happens (at least it shouldn't if I've coded it up > > right). The API _still_ requires all of the users to give-up their > > reference. > >=20 > > > It would be much more robust to have another count for the critic= al > > > stuff, initialised to one by the __set_critical_clocks function. > >=20 > > If I understand you correctly, we already have a count. We use the > > original reference count. No need for one of our own. > >=20 > > Using your RAM Clock (Clock 4) as an example > > -------------------------------------------- > >=20 > > Early start-up: > > Clock 4 is marked as critical and a reference is taken (ref =3D=3D= 1) > >=20 > > Driver probe: > > SPI enables Clock 4 (ref =3D=3D 2) > > I2C enables Clock 4 (ref =3D=3D 3) > >=20 > > Suspend (without RAM driver's permission): > > SPI disables Clock 4 (ref =3D=3D 2) > > I2C disables Clock 4 (ref =3D=3D 1) > > /* > > * Clock won't be gated because: > > * .leave_on is True - can't dec final reference >=20 > I am clearly missing the point. The clock won't be gated because the > enable_count is still 1! What does .leave_on do here? The point of _this_ (the extended) part of the API is so that the clock _can_ be turned off. Without the possibility to disable =2Eleave_on and the logic with accompanies it (i.e. clk_disable_critical()) the clock will _never_ be gated. > > */ > >=20 > > Suspend (with RAM driver's permission): > > /* Order is unimportant */ > > SPI disables Clock 4 (ref =3D=3D 2) > > RAM disables Clock 4 (ref =3D=3D 1) /* Won't turn off here (ref >= 0) > > I2C disables Clock 4 (ref =3D=3D 0) /* (.leave_on =3D=3D False) l= ast ref can be taken */ > > /* > > * Clock will be gated because: > > * .leave_on is False, so (ref =3D=3D 0) >=20 > Again, .leave_on does nothing new here. We gate the clock because the > reference count is 0. It's the fact that .leave_on has been disabled in clk_disable_critical() that allows the final reference to be taken. > > */ > >=20 > > Resume: > > /* Order is unimportant */ > > SPI enables Clock 4 (ref =3D=3D 1) > > RAM enables Clock 4 and re-enables .leave_on (ref =3D=3D 2) > > I2C enables Clock 4 (ref =3D=3D 3) >=20 > Same again. As soon as RAM calls clk_enable_critical the ref count go= es > up. .leave_on does nothing as far as I can tell. The all works becaus= e > of the reference counting, which already exists before this patch > series. So fundamentally you're right in what you say. All you really need to disable a critical clock is write a knowledgeable driver, which is intentionally unbalanced i.e. just calls clk_disable(). All this extended API really does is makes the process more official and ensures that an unintentionally unbalanced driver doesn't bugger up the running platform. We could also add a new WARN() to say that said driver is unbalanced, as it just tried to turn off a critical clock. What do you think is best? --=20 Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org =E2=94=82 Open source software for ARM SoCs =46ollow Linaro: Facebook | Twitter | Blog -- To unsubscribe from this list: send the line "unsubscribe devicetree" i= n the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html