Re: [PATCH 3/3][update] PM / sleep: Introduce command line argument for sleep state enumeration

["Rafael J. Wysocki" <rjw@rjwysocki.net>]

On Monday, May 26, 2014 02:01:14 PM Srivatsa S. Bhat wrote:
> On 05/23/2014 06:33 PM, Rafael J. Wysocki wrote:
> > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > 
> > On some systems the platform doesn't support neither
> > PM_SUSPEND_MEM nor PM_SUSPEND_STANDBY, so PM_SUSPEND_FREEZE is the
> > only available system sleep state.  However, some user space frameworks
> > only use the "mem" and (sometimes) "standby" sleep state labels, so
> > the users of those systems need to modify user space in order to be
> > able to use system suspend at all and that is not always possible.
> >
> 
> So, is this going to be a temporary solution until all the user-space
> frameworks have been fixed? I certainly hope so, because this clearly looks
> like a bug (or a lack of feature) in user-space to me... in the sense that
> those user-space frameworks don't have support for (i.e., don't know how to
> deal with) freeze-only systems yet.
> 
> The more elegant long term solution would have been to teach the kernel to
> export *truly* relative names such as SUSPEND_DEEP, SUSPEND_SHALLOW,
> and SUSPEND_LIGHT or something like that (of course, we can debate on what
> naming would suit best).

I agree and that's a logical next step, so I have a plan to rename the kernel
symbols corresponding to each state so that they reflect the code more closely
(for example, the current PM_SUSPEND_MEM and PM_SUSPEND_STANDBY depend on the
platform to support them, but that is not clear from the symbol naming, and on
many platforms _MEM doesn't mean "suspend-to-RAM" anyway).

The strings in the interface are a somewhat different matter, because user
space already depends on them (which is the source of the problem here), so we
may need to keep them or at least respond to them as expected when written to
/sys/power/state.

I personally think that we should use "sleep", "snooze" and "pause" to reflect
the "sleep depth".  And possibly "hibernate" instead of "disk".

> But this patch continues to keep the names 'SUSPEND_MEM' ("mem") etc., which
> still implies that we are doing Suspend-to-RAM, because the name itself betrays
> that info. So IMHO it doesn't really match what the command-line-switch
> 'relative_sleep_states' says.
> 
> But I understand that this is a quick hack to make existing user-space work
> with systems that support only the freeze state. However, for the reasons
> mentioned above, I hope that this is a temporary solution and we can remove
> or enhance this once all those user-space frameworks have been fixed.

Well, it is supposed to be temporary, but I'm not sure if we can count on all
user space to be fixed in any reasonable time frame (think about Android, for
example).

Rafael