From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Rafael J. Wysocki" Subject: Re: [PATCH v5 01/48] kernel: Add support for power-off handler call chain Date: Fri, 07 Nov 2014 01:16:54 +0100 Message-ID: <6901931.ThbbTy3BuF@vostro.rjw.lan> References: <1415292213-28652-1-git-send-email-linux@roeck-us.net> <3183593.9HaIZKjCKr@vostro.rjw.lan> <20141106222703.GA6377@roeck-us.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from v094114.home.net.pl ([79.96.170.134]:64830 "HELO v094114.home.net.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1751182AbaKFX4I convert rfc822-to-8bit (ORCPT ); Thu, 6 Nov 2014 18:56:08 -0500 In-Reply-To: <20141106222703.GA6377@roeck-us.net> Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Guenter Roeck Cc: linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org, Alan Cox , Alexander Graf , Andrew Morton , Geert Uytterhoeven , Heiko Stuebner , Lee Jones , Len Brown , Pavel Machek , Philippe =?ISO-8859-1?Q?R=E9tornaz?= , Romain Perier On Thursday, November 06, 2014 02:27:03 PM Guenter Roeck wrote: > On Thu, Nov 06, 2014 at 11:30:59PM +0100, Rafael J. Wysocki wrote: > > On Thursday, November 06, 2014 08:42:45 AM Guenter Roeck wrote: > > > Various drivers implement architecture and/or device specific mea= ns to > > > power off the system. For the most part, those drivers set the g= lobal > > > variable pm_power_off to point to a function within the driver. > > >=20 > > > This mechanism has a number of drawbacks. Typically only one sch= eme > > > to remove power is supported (at least if pm_power_off is used). > > > At least in theory there can be multiple means remove power, some= of > > > which may be less desirable. For example, some mechanisms may onl= y > > > power off the CPU or the CPU card, while another may power off th= e > > > entire system. Others may really just execute a restart sequence > > > or drop into the ROM monitor. Using pm_power_off can also be racy > > > if the function pointer is set from a driver built as module, as = the > > > driver may be in the process of being unloaded when pm_power_off = is > > > called. If there are multiple power-off handlers in the system, r= emoving > > > a module with such a handler may inadvertently reset the pointer = to > > > pm_power_off to NULL, leaving the system with no means to remove = power. > > >=20 > > > Introduce a system power-off handler call chain to solve the desc= ribed > > > problems. This call chain is expected to be executed from the ar= chitecture > > > specific machine_power_off() function. Drivers and architeceture= code > > > providing system power-off functionality are expected to register= with > > > this call chain. When registering a power-off handler, callers c= an > > > provide a priority to control power-off handler execution sequenc= e > > > and thus ensure that the power-off handler with the optimal capab= ilities > > > to remove power for a given system is called first. > > >=20 > > > Cc: Alan Cox > > > Cc: Alexander Graf > > > Cc: Andrew Morton > > > Cc: Geert Uytterhoeven > > > Cc: Heiko Stuebner > > > Cc: Lee Jones > > > Cc: Len Brown > > > Cc: Pavel Machek > > > Cc: Philippe R=C3=A9tornaz > > > Cc: Rafael J. Wysocki > > > Cc: Romain Perier > > > Acked-by: Pavel Machek > > > Acked-by: Heiko Stuebner > > > Signed-off-by: Guenter Roeck > > > --- > > > v5: > > > - Rebase to v3.18-rc3 > > > v4: > > > - Do not use notifiers but internal functions and data structures= to manage > > > the list of power-off handlers. Drop unused parameters from cal= lbacks, and > > > make the power-off function type void. > > > Code to manage and walk the list of callbacks is derived from n= otifier.c. > > > v3: > > > - Rename new file to power_off_handler.c > > > - Replace poweroff in all newly introduced variables and in text > > > with power_off or power-off as appropriate > > > - Replace POWEROFF_PRIORITY_xxx with POWER_OFF_PRIORITY_xxx > > > - Execute power-off handlers without any locks held > > > v2: > > > - poweroff -> power_off > > > - Add defines for default priorities > > > - Use raw notifiers protected by spinlocks instead of atomic noti= fiers > > > - Add register_poweroff_handler_simple > > > - Add devm_register_power_off_handler > > >=20 > > > include/linux/pm.h | 28 ++++ > > > kernel/power/Makefile | 1 + > > > kernel/power/power_off_handler.c | 293 +++++++++++++++++++++++++= ++++++++++++++ > > > 3 files changed, 322 insertions(+) > > > create mode 100644 kernel/power/power_off_handler.c > > >=20 > > > diff --git a/include/linux/pm.h b/include/linux/pm.h > > > index 383fd68..a4d6bf8 100644 > > > --- a/include/linux/pm.h > > > +++ b/include/linux/pm.h > > > @@ -35,6 +35,34 @@ extern void (*pm_power_off)(void); > > > extern void (*pm_power_off_prepare)(void); > > > =20 > > > struct device; /* we have a circular dep with device.h */ > > > + > > > +/* > > > + * Data structures and callbacks to manage power-off handlers > > > + */ > > > + > > > +struct power_off_handler_block { > > > + void (*handler)(struct power_off_handler_block *); > > > + struct power_off_handler_block __rcu *next; > > > + int priority; > > > +}; > > > + > > > +int register_power_off_handler(struct power_off_handler_block *)= ; > > > +int devm_register_power_off_handler(struct device *, > > > + struct power_off_handler_block *); > > > +int register_power_off_handler_simple(void (*function)(void), in= t priority); > > > +int unregister_power_off_handler(struct power_off_handler_block = *); > > > +void do_kernel_power_off(void); > > > +bool have_kernel_power_off(void); > > > + > > > +/* > > > + * Pre-defined power-off handler priorities > > > + */ > > > +#define POWER_OFF_PRIORITY_FALLBACK 0 > > > +#define POWER_OFF_PRIORITY_LOW 64 > > > +#define POWER_OFF_PRIORITY_DEFAULT 128 > > > +#define POWER_OFF_PRIORITY_HIGH 192 > > > +#define POWER_OFF_PRIORITY_HIGHEST 255 > >=20 > > I'm not sure why we need these gaps in the priority space. > >=20 > > I guess it might be possible to use > >=20 > > enum power_off_priority { > > POWER_OFF_PRIORITY_FALLBACK =3D 0, > > POWER_OFF_PRIORITY_LOW, > > POWER_OFF_PRIORITY_DEFAULT, > > POWER_OFF_PRIORITY_HIGH, > > POWER_OFF_PRIORITY_HIGHEST, > > POWER_OFF_PRIORITY_LIMIT, > > }; >=20 > I retained the large number space on purpose, specifically to permit = in-between > priorities. In other words, I want people to be able to say "priority= for this > handler is higher than low but lower than default". After all, the de= fines were > intended as hints, not as a "Thou shall use those and only those prio= rities". Problem with that is how they are supposed to know what priority to use= then. How do I know if my priority is between DEFAULT and HIGH and whether it= is closer to HIGH or closer to DEFAULT? What are the rules? The only rule that seems to be there is "this handler should be tried b= efore that one, so it needs to have a higher priority". But now the question= is how people are going to know which handlers they are competing with and= whether or not they are more "important". > Having said that, the important part is to get the series accepted, s= o I won't > argue if that is what it takes to get an Ack. Let me know. This isn't worth fighting over in my view, so I won't if everyone else = is fine with it. Just feel free to ignore this concern if you don't think it is valid. Rafael