Device Initiated Power Management, which is defined

in SATA 2.5 can be enabled for disks which support it.

This patch enables DIPM when the user sets the link

power management policy to "min_power".
Additionally, libata drivers can define a function

(enable_pm) that will perform hardware specific actions to

enable whatever power management policy the user set up

for Host Initiated Power management (HIPM).

This power management policy will be activated after all

disks have been enumerated and intialized.  Drivers should

also define disable_pm, which will turn off link power

management, but not change link power management policy.
Signed-off-by:  Kristen Carlson Accardi <kristen.c.accardi@intel.com>

---

 Documentation/scsi/link_power_management_policy.txt |   19 +

 drivers/ata/libata-core.c                           |  194 +++++++++++++++++++-

 drivers/ata/libata-eh.c                             |    5

 drivers/ata/libata-scsi.c                           |   83 ++++++++

 include/linux/ata.h                                 |    7

 include/linux/libata.h                              |   25 ++

 6 files changed, 322 insertions(+), 11 deletions(-)
Index: libata-dev/drivers/ata/libata-scsi.c

===================================================================

--- libata-dev.orig/drivers/ata/libata-scsi.c	2007-09-24 13:43:10.000000000 -0700

+++ libata-dev/drivers/ata/libata-scsi.c	2007-09-24 13:46:22.000000000 -0700

@@ -110,6 +110,78 @@ static struct scsi_transport_template at

 };
+static const struct {

+	enum link_pm	value;

+	char		*name;

+} link_pm_policy[] = {

+	{ NOT_AVAILABLE, "max_performance" },

+	{ MIN_POWER, "min_power" },

+	{ MAX_PERFORMANCE, "max_performance" },

+	{ MEDIUM_POWER, "medium_power" },

+};

+

+const char *ata_scsi_link_pm_policy(enum link_pm policy)

+{

+	int i;

+	char *name = NULL;

+

+	for (i = 0; i < ARRAY_SIZE(link_pm_policy); i++) {

+		if (link_pm_policy[i].value == policy) {

+			name = link_pm_policy[i].name;

+			break;

+		...
[ message continues ]

applied as the attached two patches to jgarzik/libata-dev.git#alpm
open issues:
1) need to check ata_dev_set_feature() return value in

ata_dev_set_dipm() and do something useful with it
2) as the name implies, this probably better belongs in ata_link.
3) however, the feature is tightly coupled to the host controller.  in

theory PMP -might- do this, but I think its unlikely.  as such I was OK

with the present arrangement.
4) there has been some discussion of software-initiated device/link

power management, but I think this should go in, in parallel with those

discussions.  ALPM

	* is quite self-contained

	* gives a noticable power savings
I'm definitely interested in seeing somebody pursue software-initiated

link PM as well...
	Jeff

^

		|

are you sure this

should be 76?
we can also change the first statement a bit:
and:
-

On Tue, 25 Sep 2007 01:12:32 +0200
I feel this is equivalent functionality and not as readable.

-

Poke around for Alan Cox's cleanup of this area of linux/ata.h.
It converts several macros to inline functions (encouraged), and also

illustrates a nice, clean way of testing an ID word's validity.

[obviously the final implementation varies, depending on that ID word's

history]
Alan or Andrew, got a copy somewhere?  My feeble search skills don't

seem to turn it up at the moment, even though I had a copy in my hands

quite recently.
	Jeff
-

Its in -mm and I thought you put a copy in your tree after I said it

hadn't upset anything in -mm ?
Alan

-

it didn't apply straightaway to my tree for whatever reason, IIRC
-

int foo(int i, int j) {
        return !(i & 8) || !j;

}
int moo(int i, int j) {
        return !((i & 8) && j);

}
gcc -O2 -S:
.globl foo

        .type   foo, @function

foo:

        shrl    $3, %edi

        xorl    $1, %edi

        testl   %esi, %esi

        sete    %al

        orl     %eax, %edi

        andl    $1, %edi

        movl    %edi, %eax

        ret

.globl moo

        .type   moo, @function

moo:

        shrl    $3, %edi

        xorl    $1, %edi

        testl   %esi, %esi

        sete    %al

        orl     %eax, %edi

        andl    $1, %edi

        movl    %edi, %eax

        ret
- Davide
-

Indeed, no difference, except for the eye. 
do you not consider it an improvement or do you not want to change it?

or

don't you consider it an improvement and want to change it?
-