[RFC 0/7] Runtime PM updates for autosuspend

[RFC 0/7] Runtime PM updates for autosuspend

[Alan Stern]

Rafael and Greg:

The following series of emails contains my proposed patches for the
runtime PM core, leading up to a general "autosuspend" implementation.  
This is going to be one of those difficult cross-subsystem things,
because the first patch in the series affects sysfs whereas the rest
affect PM.  (In addition, there are two other patches I'm not going to
include here that update USB to take advantage of the new PM
facilities.  Ultimately there will be changes to the SCSI, block, and
IDE layers as well -- they're not written yet.)

We can decide on the best plan for merging these later on.  For now I 
would simply like to post them for review and comments.  These are not 
yet necessarily in final form, but I think they are pretty close.  They 
apply on top of 2.6.36-rc5.

Alan Stern


1/7:	Add sysfs_merge_group() and sysfs_unmerge_group() routines
	for later use by PM and USB.

2/7:	Move the runtime PM accounting routines away from the functions
	that do the actual work.

3/7:	Replace runtime PM boolean arguments with bitflags.

4/7:	Merge the synchronous and asynchronous runtime PM routines.

5/7:	Regroup the public entry points for runtime PM into three
	simple functions.

6/7:	Add the "no_callbacks" flag.

7/7:	Implement autosuspend.

Re: [RFC 0/7] Runtime PM updates for autosuspend

[Rafael J. Wysocki]

On Thursday, September 23, 2010, Alan Stern wrote:
> Rafael and Greg:

Hi,

> The following series of emails contains my proposed patches for the
> runtime PM core, leading up to a general "autosuspend" implementation.  
> This is going to be one of those difficult cross-subsystem things,
> because the first patch in the series affects sysfs whereas the rest
> affect PM.  (In addition, there are two other patches I'm not going to
> include here that update USB to take advantage of the new PM
> facilities.  Ultimately there will be changes to the SCSI, block, and
> IDE layers as well -- they're not written yet.)
> 
> We can decide on the best plan for merging these later on.  For now I 
> would simply like to post them for review and comments.  These are not 
> yet necessarily in final form, but I think they are pretty close.  They 
> apply on top of 2.6.36-rc5.
> 
> Alan Stern
> 
> 
> 1/7:	Add sysfs_merge_group() and sysfs_unmerge_group() routines
> 	for later use by PM and USB.
> 
> 2/7:	Move the runtime PM accounting routines away from the functions
> 	that do the actual work.
> 
> 3/7:	Replace runtime PM boolean arguments with bitflags.
> 
> 4/7:	Merge the synchronous and asynchronous runtime PM routines.
> 
> 5/7:	Regroup the public entry points for runtime PM into three
> 	simple functions.
> 
> 6/7:	Add the "no_callbacks" flag.
> 
> 7/7:	Implement autosuspend.

The series generally looks good to me, except for a few comments, mostly
related to the names of various things I sent in my replies to the specific
patches.

Thanks,
Rafael

Re: [RFC 0/7] Runtime PM updates for autosuspend

[Alan Stern]

On Fri, 24 Sep 2010, Rafael J. Wysocki wrote:

> On Thursday, September 23, 2010, Alan Stern wrote:
> > Rafael and Greg:
> 
> Hi,
> 
> > The following series of emails contains my proposed patches for the
> > runtime PM core, leading up to a general "autosuspend" implementation.  
> > This is going to be one of those difficult cross-subsystem things,
> > because the first patch in the series affects sysfs whereas the rest
> > affect PM.  (In addition, there are two other patches I'm not going to
> > include here that update USB to take advantage of the new PM
> > facilities.  Ultimately there will be changes to the SCSI, block, and
> > IDE layers as well -- they're not written yet.)
> > 
> > We can decide on the best plan for merging these later on.  For now I 
> > would simply like to post them for review and comments.  These are not 
> > yet necessarily in final form, but I think they are pretty close.  They 
> > apply on top of 2.6.36-rc5.
> > 
> > Alan Stern
> > 
> > 
> > 1/7:	Add sysfs_merge_group() and sysfs_unmerge_group() routines
> > 	for later use by PM and USB.
> > 
> > 2/7:	Move the runtime PM accounting routines away from the functions
> > 	that do the actual work.
> > 
> > 3/7:	Replace runtime PM boolean arguments with bitflags.
> > 
> > 4/7:	Merge the synchronous and asynchronous runtime PM routines.
> > 
> > 5/7:	Regroup the public entry points for runtime PM into three
> > 	simple functions.
> > 
> > 6/7:	Add the "no_callbacks" flag.
> > 
> > 7/7:	Implement autosuspend.
> 
> The series generally looks good to me, except for a few comments, mostly
> related to the names of various things I sent in my replies to the specific
> patches.

Your points were all well taken.  I'll revise the patches and submit
them.  Do you think it's too late for them to appear in the next merge
window?

Alan Stern

Re: [RFC 0/7] Runtime PM updates for autosuspend

[Rafael J. Wysocki]

On Friday, September 24, 2010, Alan Stern wrote:
> On Fri, 24 Sep 2010, Rafael J. Wysocki wrote:
> 
> > On Thursday, September 23, 2010, Alan Stern wrote:
> > > Rafael and Greg:
> > 
> > Hi,
> > 
> > > The following series of emails contains my proposed patches for the
> > > runtime PM core, leading up to a general "autosuspend" implementation.  
> > > This is going to be one of those difficult cross-subsystem things,
> > > because the first patch in the series affects sysfs whereas the rest
> > > affect PM.  (In addition, there are two other patches I'm not going to
> > > include here that update USB to take advantage of the new PM
> > > facilities.  Ultimately there will be changes to the SCSI, block, and
> > > IDE layers as well -- they're not written yet.)
> > > 
> > > We can decide on the best plan for merging these later on.  For now I 
> > > would simply like to post them for review and comments.  These are not 
> > > yet necessarily in final form, but I think they are pretty close.  They 
> > > apply on top of 2.6.36-rc5.
> > > 
> > > Alan Stern
> > > 
> > > 
> > > 1/7:	Add sysfs_merge_group() and sysfs_unmerge_group() routines
> > > 	for later use by PM and USB.
> > > 
> > > 2/7:	Move the runtime PM accounting routines away from the functions
> > > 	that do the actual work.
> > > 
> > > 3/7:	Replace runtime PM boolean arguments with bitflags.
> > > 
> > > 4/7:	Merge the synchronous and asynchronous runtime PM routines.
> > > 
> > > 5/7:	Regroup the public entry points for runtime PM into three
> > > 	simple functions.
> > > 
> > > 6/7:	Add the "no_callbacks" flag.
> > > 
> > > 7/7:	Implement autosuspend.
> > 
> > The series generally looks good to me, except for a few comments, mostly
> > related to the names of various things I sent in my replies to the specific
> > patches.
> 
> Your points were all well taken.  I'll revise the patches and submit
> them.  Do you think it's too late for them to appear in the next merge
> window?

No, it's not.  I actually would like to push them for .37.

Thanks,
Rafael

Re: [RFC 0/7] Runtime PM updates for autosuspend

[Alan Stern]

On Fri, 24 Sep 2010, Rafael J. Wysocki wrote:

> On Friday, September 24, 2010, Alan Stern wrote:
> > On Fri, 24 Sep 2010, Rafael J. Wysocki wrote:
> > 
> > > On Thursday, September 23, 2010, Alan Stern wrote:
> > > > Rafael and Greg:
> > > 
> > > Hi,
> > > 
> > > > The following series of emails contains my proposed patches for the
> > > > runtime PM core, leading up to a general "autosuspend" implementation.  
> > > > This is going to be one of those difficult cross-subsystem things,
> > > > because the first patch in the series affects sysfs whereas the rest
> > > > affect PM.  (In addition, there are two other patches I'm not going to
> > > > include here that update USB to take advantage of the new PM
> > > > facilities.  Ultimately there will be changes to the SCSI, block, and
> > > > IDE layers as well -- they're not written yet.)
> > > > 
> > > > We can decide on the best plan for merging these later on.  For now I 
> > > > would simply like to post them for review and comments.  These are not 
> > > > yet necessarily in final form, but I think they are pretty close.  They 
> > > > apply on top of 2.6.36-rc5.
> > > > 
> > > > Alan Stern
> > > > 
> > > > 
> > > > 1/7:	Add sysfs_merge_group() and sysfs_unmerge_group() routines
> > > > 	for later use by PM and USB.
> > > > 
> > > > 2/7:	Move the runtime PM accounting routines away from the functions
> > > > 	that do the actual work.
> > > > 
> > > > 3/7:	Replace runtime PM boolean arguments with bitflags.
> > > > 
> > > > 4/7:	Merge the synchronous and asynchronous runtime PM routines.
> > > > 
> > > > 5/7:	Regroup the public entry points for runtime PM into three
> > > > 	simple functions.
> > > > 
> > > > 6/7:	Add the "no_callbacks" flag.
> > > > 
> > > > 7/7:	Implement autosuspend.
> > > 
> > > The series generally looks good to me, except for a few comments, mostly
> > > related to the names of various things I sent in my replies to the specific
> > > patches.
> > 
> > Your points were all well taken.  I'll revise the patches and submit
> > them.  Do you think it's too late for them to appear in the next merge
> > window?
> 
> No, it's not.  I actually would like to push them for .37.

Okay, here they come.  I'll skip the 0/7 message; this thread can serve 
in its place.

Alan Stern