| Message ID | 1428696929-2988-1-git-send-email-mjg59@coreos.com |
|---|---|
| State | Not Applicable |
| Delegated to: | David Miller |
| Headers | show |
On Fri, Apr 10, 2015 at 01:15:29PM -0700, Matthew Garrett wrote: > Intel publish a document on designing energy efficient SATA devices at > http://www.intel.com/content/dam/doc/reference-guide/sata-devices-implementation-recommendations.pdf > which recommends that ALPE be set, ASPE be cleared and that DIPM be enabled > on the device. Right now we have no policy that matches that - medium_power > does not enable DIPM and min_power sets ASPE. Add a new low_power policy to > implement these recommendations. I've left devslp disabled because I'm not > yet sure what Intel do in that case. With luck this will provide reasonable > power savings without causing the device breakages we occasionally see with > the min_power policy. I suspect the aggressive methods actually don't buy us much, except for the obsessive mode switchings, in terms of power saving compared to something more reasonable provided via DIPM and I'm not sure it's a good idea to introduce yet another mode. I'd be all in for converting min_power mode to DIPM for cases where we know this actually works rather than introducing yet another mode. Or if there's a clear power consumption disadvantage to DIPM, let's conver medium_power to that. Thanks.
On Mon, Apr 13, 2015 at 7:32 AM, Tejun Heo <tj@kernel.org> wrote: > On Fri, Apr 10, 2015 at 01:15:29PM -0700, Matthew Garrett wrote: >> Intel publish a document on designing energy efficient SATA devices at >> http://www.intel.com/content/dam/doc/reference-guide/sata-devices-implementation-recommendations.pdf >> which recommends that ALPE be set, ASPE be cleared and that DIPM be enabled >> on the device. Right now we have no policy that matches that - medium_power >> does not enable DIPM and min_power sets ASPE. Add a new low_power policy to >> implement these recommendations. I've left devslp disabled because I'm not >> yet sure what Intel do in that case. With luck this will provide reasonable >> power savings without causing the device breakages we occasionally see with >> the min_power policy. > > I suspect the aggressive methods actually don't buy us much, except > for the obsessive mode switchings, in terms of power saving compared > to something more reasonable provided via DIPM and I'm not sure it's a > good idea to introduce yet another mode. I'd be all in for converting > min_power mode to DIPM for cases where we know this actually works > rather than introducing yet another mode. Or if there's a clear power > consumption disadvantage to DIPM, let's conver medium_power to that. Ok - my only real concern there was that we might end up changing the semantics of medium_power enough that it would break somebody. I'll post a patch to do that instead and I guess we can see what happens - worst case we revert it and add a new one. -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello, On Mon, Apr 13, 2015 at 10:52:59AM -0700, Matthew Garrett wrote: > Ok - my only real concern there was that we might end up changing the > semantics of medium_power enough that it would break somebody. I'll > post a patch to do that instead and I guess we can see what happens - > worst case we revert it and add a new one. Yeah, that sounds like a good plan to me. Thanks.
diff --git a/Documentation/scsi/link_power_management_policy.txt b/Documentation/scsi/link_power_management_policy.txt index d18993d..e095e03 100644 --- a/Documentation/scsi/link_power_management_policy.txt +++ b/Documentation/scsi/link_power_management_policy.txt @@ -1,13 +1,8 @@ This parameter allows the user to set the link (interface) power management. -There are 3 possible options: +There are 4 possible options: Value Effect ---------------------------------------------------------------------------- -min_power Tell the controller to try to make the link use the - least possible power when possible. This may - sacrifice some performance due to increased latency - when coming out of lower power states. - max_performance Generally, this means no power management. Tell the controller to have performance be a priority over power management. @@ -16,4 +11,11 @@ medium_power Tell the controller to enter a lower power state when possible, but do not enter the lowest power state, thus improving latency over min_power setting. +low_power Tell the controller to enter a lower power state + when possible, also enabling power management on + the target device. +min_power Tell the controller to try to make the link use the + least possible power when possible. This may + sacrifice some performance due to increased latency + when coming out of lower power states. diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c index 61a9c07..5814bef 100644 --- a/drivers/ata/libahci.c +++ b/drivers/ata/libahci.c @@ -701,9 +701,9 @@ static int ahci_set_lpm(struct ata_link *link, enum ata_lpm_policy policy, if (hpriv->cap & HOST_CAP_ALPM) { u32 cmd = readl(port_mmio + PORT_CMD); + cmd &= ~(PORT_CMD_ASP | PORT_CMD_ALPE); if (policy == ATA_LPM_MAX_POWER || !(hints & ATA_LPM_HIPM)) { - cmd &= ~(PORT_CMD_ASP | PORT_CMD_ALPE); cmd |= PORT_CMD_ICC_ACTIVE; writel(cmd, port_mmio + PORT_CMD); @@ -720,7 +720,6 @@ static int ahci_set_lpm(struct ata_link *link, enum ata_lpm_policy policy, writel(cmd, port_mmio + PORT_CMD); } } - /* set aggressive device sleep */ if ((hpriv->cap2 & HOST_CAP2_SDS) && (hpriv->cap2 & HOST_CAP2_SADM) && diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c index 23dac3b..39e03d5 100644 --- a/drivers/ata/libata-core.c +++ b/drivers/ata/libata-core.c @@ -3662,6 +3662,7 @@ int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, scontrol &= ~(0x1 << 8); scontrol |= (0x6 << 8); break; + case ATA_LPM_LOW_POWER: case ATA_LPM_MIN_POWER: if (ata_link_nr_enabled(link) > 0) /* no restrictions on LPM transitions */ diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c index d2029a4..d4476fa 100644 --- a/drivers/ata/libata-eh.c +++ b/drivers/ata/libata-eh.c @@ -3411,7 +3411,7 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy, return 0; /* - * DIPM is enabled only for MIN_POWER as some devices + * DIPM is enabled only for LOW_POWER and MIN_POWER as some devices * misbehave when the host NACKs transition to SLUMBER. Order * device and link configurations such that the host always * allows DIPM requests. @@ -3432,7 +3432,7 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy, hints &= ~ATA_LPM_HIPM; /* disable DIPM before changing link config */ - if (policy != ATA_LPM_MIN_POWER && dipm) { + if (policy < ATA_LPM_LOW_POWER && dipm) { err_mask = ata_dev_set_feature(dev, SETFEATURES_SATA_DISABLE, SATA_DIPM); if (err_mask && err_mask != AC_ERR_DEV) { @@ -3473,9 +3473,12 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy, if (ap && ap->slave_link) ap->slave_link->lpm_policy = policy; - /* host config updated, enable DIPM if transitioning to MIN_POWER */ + /* + * host config updated, enable DIPM if transitioning to LOW_POWER or + * MIN_POWER + */ ata_for_each_dev(dev, link, ENABLED) { - if (policy == ATA_LPM_MIN_POWER && !no_dipm && + if (policy >= ATA_LPM_LOW_POWER && !no_dipm && ata_id_has_dipm(dev->id)) { err_mask = ata_dev_set_feature(dev, SETFEATURES_SATA_ENABLE, SATA_DIPM); diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c index b061ba2..66c6946 100644 --- a/drivers/ata/libata-scsi.c +++ b/drivers/ata/libata-scsi.c @@ -108,6 +108,7 @@ static const char *ata_lpm_policy_names[] = { [ATA_LPM_UNKNOWN] = "max_performance", [ATA_LPM_MAX_POWER] = "max_performance", [ATA_LPM_MED_POWER] = "medium_power", + [ATA_LPM_LOW_POWER] = "low_power", [ATA_LPM_MIN_POWER] = "min_power", }; diff --git a/include/linux/libata.h b/include/linux/libata.h index 6b08cc1..6cf823f 100644 --- a/include/linux/libata.h +++ b/include/linux/libata.h @@ -509,6 +509,7 @@ enum ata_lpm_policy { ATA_LPM_UNKNOWN, ATA_LPM_MAX_POWER, ATA_LPM_MED_POWER, + ATA_LPM_LOW_POWER, ATA_LPM_MIN_POWER, };
Intel publish a document on designing energy efficient SATA devices at http://www.intel.com/content/dam/doc/reference-guide/sata-devices-implementation-recommendations.pdf which recommends that ALPE be set, ASPE be cleared and that DIPM be enabled on the device. Right now we have no policy that matches that - medium_power does not enable DIPM and min_power sets ASPE. Add a new low_power policy to implement these recommendations. I've left devslp disabled because I'm not yet sure what Intel do in that case. With luck this will provide reasonable power savings without causing the device breakages we occasionally see with the min_power policy. Signed-off-by: Matthew Garrett <mjg59@coreos.com> --- Documentation/scsi/link_power_management_policy.txt | 14 ++++++++------ drivers/ata/libahci.c | 3 +-- drivers/ata/libata-core.c | 1 + drivers/ata/libata-eh.c | 11 +++++++---- drivers/ata/libata-scsi.c | 1 + include/linux/libata.h | 1 + 6 files changed, 19 insertions(+), 12 deletions(-)