Quoth Harvey Harrison:
> Ping?
> What: dev->power.power_state
> When: July 2007
> Why: Broken design for runtime control over driver power states, confusing
> driver-internal runtime power management with: mechanisms to support
> system-wide sleep state transitions; event codes that distinguish
> different phases of swsusp "sleep" transitions; and userspace policy
> inputs. This framework was never widely used, and most attempts to
> use it were broken. Drivers should instead be exposing domain-specific
> interfaces either to kernel or to userspace.
> Who: Pavel Machek <[email protected]>
A lot of the infrastructure using that has already been deleted, and
there are some incremental improvements pending for 2.6.25:
- drivers/input/touchscreen/ads7846.c ... patch fixing this
should be in either MM or the input queue
- Documentation/power/devices.txt ... patch fixing this is
in the suspend tree, due to merge RSN
- drivers/spi/spi.c ... patch fixing this is in MM, due to
merge with other SPI patches
- drivers/pcmcia/ds.c ... at least I *think* that patch got sent
But there are still quite a few users left, and a new one was (sigh)
recently added.
- drivers/rtc/rtc-sa1100.c ... new usage, merged last week
- drivers/usb/... has various users, HCDs look easy enough to fix but
the other bits will take more thought
- drivers/ata/... has some too
- drivers/ide/ppc/pmac.c
- drivers/spi/... some controller drivers use this (look easy to fix)
- drivers/scsi/mesh.c
- drivers/input/serio/... has a few users
- ... more ...
I'll probably send in a few more patches for easy stuff in areas
that I touch semi-frequently, but other folk should fix ATA, IDE,
SCSI, SERIO, and so forth. It'd be good if Alan would help fix
the USB stuff too. I'm not sure what Pavel's doing there...
- Dave
On Thu, 31 Jan 2008, David Brownell wrote:
> Quoth Harvey Harrison:
>
> > Ping?
> > What: dev->power.power_state
> > When: July 2007
> > Why: Broken design for runtime control over driver power states, confusing
> > driver-internal runtime power management with: mechanisms to support
> > system-wide sleep state transitions; event codes that distinguish
> > different phases of swsusp "sleep" transitions; and userspace policy
> > inputs. This framework was never widely used, and most attempts to
> > use it were broken. Drivers should instead be exposing domain-specific
> > interfaces either to kernel or to userspace.
> > Who: Pavel Machek <[email protected]>
>
> A lot of the infrastructure using that has already been deleted, and
> there are some incremental improvements pending for 2.6.25:
> But there are still quite a few users left, and a new one was (sigh)
> recently added.
> - drivers/usb/... has various users, HCDs look easy enough to fix but
> the other bits will take more thought
> - drivers/scsi/mesh.c
> I'll probably send in a few more patches for easy stuff in areas
> that I touch semi-frequently, but other folk should fix ATA, IDE,
> SCSI, SERIO, and so forth. It'd be good if Alan would help fix
> the USB stuff too. I'm not sure what Pavel's doing there...
I can look at the USB and SCSI stuff. It shouldn't be all that bad.
The userspace interface has been gone for quite some time now, and most
of the remaining uses of that field are write-only.
Alan Stern
On Thursday 31 January 2008, David Brownell wrote:
> Quoth Harvey Harrison:
>
> > Ping?
> > What: dev->power.power_state
> > When: July 2007
>
> ... there are still quite a few users left, and a new one was (sigh)
> recently added.
>
> - drivers/rtc/rtc-sa1100.c ... new usage, merged last week
Fixed in a patch I just sent.
> - drivers/spi/... some controller drivers use this (look easy to fix)
Ditto.
- Dave
On Friday 01 February 2008, Alan Stern wrote:
>
> I can look at the USB and SCSI stuff. It shouldn't be all that bad.
> The userspace interface has been gone for quite some time now, and most
> of the remaining uses of that field are write-only.
It's that "most" which can make trouble. :)
- Dave