ahci: Add some controls on actual LPM capability
cover ahci: Add some controls on actual LPM capability
On some platform, when OS enables LPM by default (eg, min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM; DIPM Slumber request cannot be disallowed if ahci's CAP.PSC is set to '1' and CAP.SSC is cleared to '0', which may cause ahci to be an uncertain state (same for Partial).
In ahci spec, PhyRdy Change cannot coexist with LPM; when CAP.PSC/SSC is cleared to '0', the PxSCTL.IPM field must be programmed to disallow device initiated Partial/Slumber request.
Adds suports to control these cases on actual LPM capability.
Runa Guo-oc (2):
ahci: Add PhyRdy Change control on actual LPM capability
ahci: Add PxSCTL.IPM control on actual LPM capability
drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ drivers/ata/libata-sata.c | 12 +++++++++++- include/linux/libata.h | 4 ++++ 4 files changed, 28 insertions(+), 1 deletion(-)
Signature · 1 line
2.7.4
PATCH 1/2 ahci: Add PhyRdy Change control on actual LPM capability
On some platform, when OS enables LPM by default (eg, min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.
In ahci spec, PhyRdy Change cannot coexist with LPM.
Adds support to control this case on actual LPM capability.
--- drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ include/linux/libata.h | 4 ++++ 3 files changed, 17 insertions(+)
drivers/ata/ahci.c +9 -0
@@ -1870,6 +1870,15 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)else dev_info (&pdev->dev,"SSS flag set, parallel bus scan disabled\n" ); +if (hpriv->cap &HOST_CAP_PART ) + host->flags |=ATA_HOST_PART ; + +if (hpriv->cap &HOST_CAP_SSC ) + host->flags |=ATA_HOST_SSC ; + +if (hpriv->cap2 &HOST_CAP2_SDS ) + host->flags |=ATA_HOST_DEVSLP ; +if (pi.flags &ATA_FLAG_EM )ahci_reset_em (host);
drivers/ata/libata-eh.c +4 -0
@@ -3246,6 +3246,10 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,unsigned int err_mask;int rc; +/* if controller does not support lpm, then sets no LPM flags*/ +if (!(ap->host->flags & (ATA_HOST_PART |ATA_HOST_SSC |ATA_HOST_DEVSLP ))) + link->flags |=ATA_LFLAG_NO_LPM ; +/* if the link or host doesn't do LPM, noop */ if (!IS_ENABLED (CONFIG_SATA_HOST ) || (link->flags &ATA_LFLAG_NO_LPM ) || (ap && !ap->ops->set_lpm))
include/linux/libata.h +4 -0
@@ -216,6 +216,10 @@ enum {ATA_HOST_PARALLEL_SCAN = (1 <<2 ),/* Ports on this host can be scanned in parallel */ ATA_HOST_IGNORE_ATA = (1 <<3 ),/* Ignore ATA devices on this host. */ +ATA_HOST_PART = (1 <<4 ),/* Host support partial.*/ +ATA_HOST_SSC = (1 <<5 ),/* Host support slumber.*/ +ATA_HOST_DEVSLP = (1 <<6 ),/* Host support devslp.*/ +/* bits 24:31 of host->flags are reserved for LLD specific flags */ /* various lengths of time */
Signature · 1 line
2.7.4
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.
Do you mean "...if the ahci adapter does not support LPM." ? If that is what you mean, then min_power should not be set. Mario has patches to fix that.
Quoted 62 lines
In ahci spec, PhyRdy Change cannot coexist with LPM.
Adds support to control this case on actual LPM capability.
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ include/linux/libata.h | 4 ++++ 3 files changed, 17 insertions(+)drivers/ata/ahci.c +9
-0@@ -1870,6 +1870,15 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)else dev_info (&pdev->dev,"SSS flag set, parallel bus scan disabled\n" ); +if (hpriv->cap &HOST_CAP_PART ) + host->flags |=ATA_HOST_PART ; + +if (hpriv->cap &HOST_CAP_SSC ) + host->flags |=ATA_HOST_SSC ; + +if (hpriv->cap2 &HOST_CAP2_SDS ) + host->flags |=ATA_HOST_DEVSLP ; +if (pi.flags &ATA_FLAG_EM )ahci_reset_em (host);drivers/ata/libata-eh.c +4
-0@@ -3246,6 +3246,10 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,unsigned int err_mask;int rc; +/* if controller does not support lpm, then sets no LPM flags*/ +if (!(ap->host->flags & (ATA_HOST_PART |ATA_HOST_SSC |ATA_HOST_DEVSLP ))) + link->flags |=ATA_LFLAG_NO_LPM ; +/* if the link or host doesn't do LPM, noop */ if (!IS_ENABLED (CONFIG_SATA_HOST ) || (link->flags &ATA_LFLAG_NO_LPM ) || (ap && !ap->ops->set_lpm))include/linux/libata.h +4
-0@@ -216,6 +216,10 @@ enum {ATA_HOST_PARALLEL_SCAN = (1 <<2 ),/* Ports on this host can be scanned in parallel */ ATA_HOST_IGNORE_ATA = (1 <<3 ),/* Ignore ATA devices on this host. */ +ATA_HOST_PART = (1 <<4 ),/* Host support partial.*/ +ATA_HOST_SSC = (1 <<5 ),/* Host support slumber.*/ +ATA_HOST_DEVSLP = (1 <<6 ),/* Host support devslp.*/ +/* bits 24:31 of host->flags are reserved for LLD specific flags */ /* various lengths of time */
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.
Yes, I agree with you. But, as we know, link_power_management is a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.
Quoted 1 line
Mario has patches to fix that.
Hmm. How to patch this case ?
Quoted 62 lines
Quoted 61 lines
In ahci spec, PhyRdy Change cannot coexist with LPM.
Adds support to control this case on actual LPM capability.
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ include/linux/libata.h | 4 ++++ 3 files changed, 17 insertions(+)drivers/ata/ahci.c +9
-0@@ -1870,6 +1870,15 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)else dev_info (&pdev->dev,"SSS flag set, parallel bus scan disabled\n" ); +if (hpriv->cap &HOST_CAP_PART ) + host->flags |=ATA_HOST_PART ; + +if (hpriv->cap &HOST_CAP_SSC ) + host->flags |=ATA_HOST_SSC ; + +if (hpriv->cap2 &HOST_CAP2_SDS ) + host->flags |=ATA_HOST_DEVSLP ; +if (pi.flags &ATA_FLAG_EM )ahci_reset_em (host);drivers/ata/libata-eh.c +4
-0@@ -3246,6 +3246,10 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,unsigned int err_mask;int rc; +/* if controller does not support lpm, then sets no LPM flags*/ +if (!(ap->host->flags & (ATA_HOST_PART |ATA_HOST_SSC |ATA_HOST_DEVSLP ))) + link->flags |=ATA_LFLAG_NO_LPM ; +/* if the link or host doesn't do LPM, noop */ if (!IS_ENABLED (CONFIG_SATA_HOST ) || (link->flags &ATA_LFLAG_NO_LPM ) || (ap && !ap->ops->set_lpm))include/linux/libata.h +4
-0@@ -216,6 +216,10 @@ enum {ATA_HOST_PARALLEL_SCAN = (1 <<2 ),/* Ports on this host can be scanned in parallel */ ATA_HOST_IGNORE_ATA = (1 <<3 ),/* Ignore ATA devices on this host. */ +ATA_HOST_PART = (1 <<4 ),/* Host support partial.*/ +ATA_HOST_SSC = (1 <<5 ),/* Host support slumber.*/ +ATA_HOST_DEVSLP = (1 <<6 ),/* Host support devslp.*/ +/* bits 24:31 of host->flags are reserved for LLD specific flags */ /* various lengths of time */
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 6 lines · RunaGuo-oc
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Quoted 7 lines
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.
Users should *never* need to be aware of the HW specs and what can or cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your commit message should clearly mention that. But I still need to spend more time re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect low power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch on rc3 to have the LPM config name revert. We need to fix this LPM mess :)
Quoted 5 lines
Quoted 2 lines
Mario has patches to fix that.
Hmm. How to patch this case ?
Mario's patches modify the beginning of the sata_link_scr_lpm() function to prevent setting an LPM policy that the adapter/drive does not support. This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.
It may also be good to revisit ata_scsi_lpm_store() to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.
Quoted 64 lines
Quoted 63 lines
Quoted 62 lines
In ahci spec, PhyRdy Change cannot coexist with LPM.
Adds support to control this case on actual LPM capability.
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ include/linux/libata.h | 4 ++++ 3 files changed, 17 insertions(+)drivers/ata/ahci.c +9
-0@@ -1870,6 +1870,15 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)else dev_info (&pdev->dev,"SSS flag set, parallel bus scan disabled\n" ); +if (hpriv->cap &HOST_CAP_PART ) + host->flags |=ATA_HOST_PART ; + +if (hpriv->cap &HOST_CAP_SSC ) + host->flags |=ATA_HOST_SSC ; + +if (hpriv->cap2 &HOST_CAP2_SDS ) + host->flags |=ATA_HOST_DEVSLP ; +if (pi.flags &ATA_FLAG_EM )ahci_reset_em (host);drivers/ata/libata-eh.c +4
-0@@ -3246,6 +3246,10 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,unsigned int err_mask;int rc; +/* if controller does not support lpm, then sets no LPM flags*/ +if (!(ap->host->flags & (ATA_HOST_PART |ATA_HOST_SSC |ATA_HOST_DEVSLP ))) + link->flags |=ATA_LFLAG_NO_LPM ; +/* if the link or host doesn't do LPM, noop */ if (!IS_ENABLED (CONFIG_SATA_HOST ) || (link->flags &ATA_LFLAG_NO_LPM ) || (ap && !ap->ops->set_lpm))include/linux/libata.h +4
-0@@ -216,6 +216,10 @@ enum {ATA_HOST_PARALLEL_SCAN = (1 <<2 ),/* Ports on this host can be scanned in parallel */ ATA_HOST_IGNORE_ATA = (1 <<3 ),/* Ignore ATA devices on this host. */ +ATA_HOST_PART = (1 <<4 ),/* Host support partial.*/ +ATA_HOST_SSC = (1 <<5 ),/* Host support slumber.*/ +ATA_HOST_DEVSLP = (1 <<6 ),/* Host support devslp.*/ +/* bits 24:31 of host->flags are reserved for LLD specific flags */ /* various lengths of time */
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 33 lines · Damien Le Moal
Quoted 11 lines · RunaGuo-oc
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.Users should *never* need to be aware of the HW specs and what can or cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your commit message should clearly mention that. But I still need to spend more time re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect low power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch on rc3 to have the LPM config name revert. We need to fix this LPM mess :)
Quoted 3 lines
Quoted 1 line
Mario has patches to fix that.
Hmm. How to patch this case ?
Quoted 8 lines
Mario's patches modify the beginning of the
sata_link_scr_lpm()function to prevent setting an LPM policy that the adapter/drive does not support. This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.It may also be good to revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.
To avoid some confusion in this patch set, I want to explain it here. The patch set involves two LPM related issues, one for the ahci adapter does not support LPM (no partial & slumber & devslp), the other for ahci adapter supports part of LPM(eg, only partial, no slumber & devslp).
The former case is what I metioned in this mail thread. Namely, when LPM is enabled, actions trying to enable this unsupported feature will be failed, but will disable PORT_IRQ_PHYRDY bit at the beginning of the ahci_set_lpm() function, which would make PhyRdy Changed cannot be detected. So I add flags in the ata_eh_set_lpm() function which will not go to the disable operation.
The latter case is what I metioned in "PATCH[2/2]". Namely, if the ahci adapter only supports partial (no slumber & devslp), then when LPM is enabled (eg, min_power), *device* initiated power mode transitions will be enabled with the scontrol register bits setting to indicate no restrictions on LPM transitions. After that, if SSD/HDD sends a DIPM slumber request, it cannot be disallowed by ahci adpter for driver not setting scontrol register bits properly. So I add flags to control it.
Therefore, Mario's patches in the sata_link_scr_lpm() function may fix the issue in "PATCH[2/2]".
Revisit ata_scsi_lpm_store() to prevent the user from setting an unsupported policy may be a way to fix the issue in "PATCH[1/2]", but there may be another case where some operating system manufacturers setting LPM default enable in driver, like the following code in the ahci_init_one() function, also need to control.
if (ap->flags & ATA_FLAG_EM) ap->em_message_type = hpriv->em_msg_type;
+ ap->target_lpm_policy = ATA_LPM_MIN_POWER; ahci_update_initial_lpm_policy(ap, hpriv);
According to my current understanding, the currently submitted patches can solve the above problems, and definitely not the best. I haven't figured out good way to use CAP flags to patch. Hope you can give me some advice, thanks.
Quoted 62 lines
Quoted 61 lines
Quoted 61 lines
Quoted 61 lines
In ahci spec, PhyRdy Change cannot coexist with LPM.
Adds support to control this case on actual LPM capability.
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/ahci.c | 9 +++++++++ drivers/ata/libata-eh.c | 4 ++++ include/linux/libata.h | 4 ++++ 3 files changed, 17 insertions(+)drivers/ata/ahci.c +9
-0@@ -1870,6 +1870,15 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)else dev_info (&pdev->dev,"SSS flag set, parallel bus scan disabled\n" ); +if (hpriv->cap &HOST_CAP_PART ) + host->flags |=ATA_HOST_PART ; + +if (hpriv->cap &HOST_CAP_SSC ) + host->flags |=ATA_HOST_SSC ; + +if (hpriv->cap2 &HOST_CAP2_SDS ) + host->flags |=ATA_HOST_DEVSLP ; +if (pi.flags &ATA_FLAG_EM )ahci_reset_em (host);drivers/ata/libata-eh.c +4
-0@@ -3246,6 +3246,10 @@ static int ata_eh_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,unsigned int err_mask;int rc; +/* if controller does not support lpm, then sets no LPM flags*/ +if (!(ap->host->flags & (ATA_HOST_PART |ATA_HOST_SSC |ATA_HOST_DEVSLP ))) + link->flags |=ATA_LFLAG_NO_LPM ; +/* if the link or host doesn't do LPM, noop */ if (!IS_ENABLED (CONFIG_SATA_HOST ) || (link->flags &ATA_LFLAG_NO_LPM ) || (ap && !ap->ops->set_lpm))include/linux/libata.h +4
-0@@ -216,6 +216,10 @@ enum {ATA_HOST_PARALLEL_SCAN = (1 <<2 ),/* Ports on this host can be scanned in parallel */ ATA_HOST_IGNORE_ATA = (1 <<3 ),/* Ignore ATA devices on this host. */ +ATA_HOST_PART = (1 <<4 ),/* Host support partial.*/ +ATA_HOST_SSC = (1 <<5 ),/* Host support slumber.*/ +ATA_HOST_DEVSLP = (1 <<6 ),/* Host support devslp.*/ +/* bits 24:31 of host->flags are reserved for LLD specific flags */ /* various lengths of time */
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 72 lines · Runa Guo-oc
Quoted 33 lines · Damien Le Moal
Quoted 11 lines · RunaGuo-oc
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.Users should *never* need to be aware of the HW specs and what can or cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your commit message should clearly mention that. But I still need to spend more time re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect low power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch on rc3 to have the LPM config name revert. We need to fix this LPM mess :)
Quoted 3 lines
Quoted 1 line
Mario has patches to fix that.
Hmm. How to patch this case ?
Quoted 8 lines
Mario's patches modify the beginning of the
sata_link_scr_lpm()function to prevent setting an LPM policy that the adapter/drive does not support. This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.It may also be good to revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.To avoid some confusion in this patch set, I want to explain it here. The patch set involves two LPM related issues, one for the ahci adapter does not support LPM (no partial & slumber & devslp), the other for ahci adapter supports part of
LPM(eg, only partial, no slumber & devslp).The former case is what I metioned in this mail thread. Namely, when LPM is enabled, actions trying to enable this unsupported feature will be failed, but will disable
PORT_IRQ_PHYRDYbit at the beginning of theahci_set_lpm()function, which would make PhyRdy Changed cannot be detected. So I add flags in theata_eh_set_lpm()function which will not go to the disable operation.The latter case is what I metioned in "PATCH[2/2]". Namely, if the ahci adapter only supports partial (no slumber & devslp), then when LPM is enabled (eg,
min_power), *device* initiated power mode transitions will be enabled with the scontrol register bits setting to indicate no restrictions on LPM transitions. After that, if SSD/HDD sends a DIPM slumber request, it cannot be disallowed by ahci adpter for driver not setting scontrol register bits properly. So I add flags to control it.Therefore, Mario's patches in the
sata_link_scr_lpm()function may fix the issue in "PATCH[2/2]".Revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy may be a way to fix the issue in "PATCH[1/2]", but there may be another case where some operating system manufacturers setting LPM default enable in driver, like the following code in theahci_init_one()function, also need to control.
Quoted 4 lines
if (ap->flags & ATA_FLAG_EM) ap->em_message_type = hpriv->em_msg_type;+
ap->target_lpm_policy=ATA_LPM_MIN_POWER;
This one looks wrong. This is set inside ahci_update_initial_lpm_policy() according to the default kernel config (CONFIG_SATA_MOBILE_LPM_POLICY) and module param + what the drive can do according to ACPI. The problem though is that the adapter capabilities are not checked in that function, so the initial target lpm policy may be wrong.
Since your patch 1/2 adds the hpriv flags indicating the capabilities, you need to use these in ahci_update_initial_lpm_policy() to validate whatever initial policy is asked for by the user.
Quoted 5 lines
ahci_update_initial_lpm_policy(ap, hpriv);According to my current understanding, the currently submitted patches can solve the above problems, and definitely not the best. I haven't figured out good way to use CAP flags to patch. Hope you can give me some advice, thanks.
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 81 lines · Damien Le Moal
Quoted 70 lines · Runa Guo-oc
Quoted 33 lines · Damien Le Moal
Quoted 11 lines · RunaGuo-oc
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.Users should *never* need to be aware of the HW specs and what can or cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your commit message should clearly mention that. But I still need to spend more time re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect low power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch on rc3 to have the LPM config name revert. We need to fix this LPM mess :)
Quoted 3 lines
Quoted 1 line
Mario has patches to fix that.
Hmm. How to patch this case ?
Quoted 8 lines
Mario's patches modify the beginning of the
sata_link_scr_lpm()function to prevent setting an LPM policy that the adapter/drive does not support. This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.It may also be good to revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.To avoid some confusion in this patch set, I want to explain it here. The patch set involves two LPM related issues, one for the ahci adapter does not support LPM (no partial & slumber & devslp), the other for ahci adapter supports part of
LPM(eg, only partial, no slumber & devslp).The former case is what I metioned in this mail thread. Namely, when LPM is enabled, actions trying to enable this unsupported feature will be failed, but will disable
PORT_IRQ_PHYRDYbit at the beginning of theahci_set_lpm()function, which would make PhyRdy Changed cannot be detected. So I add flags in theata_eh_set_lpm()function which will not go to the disable operation.The latter case is what I metioned in "PATCH[2/2]". Namely, if the ahci adapter only supports partial (no slumber & devslp), then when LPM is enabled (eg,
min_power), *device* initiated power mode transitions will be enabled with the scontrol register bits setting to indicate no restrictions on LPM transitions. After that, if SSD/HDD sends a DIPM slumber request, it cannot be disallowed by ahci adpter for driver not setting scontrol register bits properly. So I add flags to control it.Therefore, Mario's patches in the
sata_link_scr_lpm()function may fix the issue in "PATCH[2/2]".Revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy may be a way to fix the issue in "PATCH[1/2]", but there may be another case where some operating system manufacturers setting LPM default enable in driver, like the following code in theahci_init_one()function, also need to control.Quoted 4 lines
if (ap->flags & ATA_FLAG_EM) ap->em_message_type = hpriv->em_msg_type;+
ap->target_lpm_policy=ATA_LPM_MIN_POWER;This one looks wrong. This is set inside
ahci_update_initial_lpm_policy()according to the default kernel config (CONFIG_SATA_MOBILE_LPM_POLICY) and module param + what the drive can do according to ACPI. The problem though is that the adapter capabilities are not checked in that function, so the initial target lpm policy may be wrong.
Quoted 3 lines
Since your patch 1/2 adds the hpriv flags indicating the capabilities, you need to use these in
ahci_update_initial_lpm_policy()to validate whatever initial policy is asked for by the user.
Yes, the above code is not rigorous, existing methods provided by kernel as you said should be used in this case.
In order to use CAP flags to validate user policy, I review the latest kernel LPM policies, here is my understanding:ATA_LPM_UNKNOWN: default policy, no LPMATA_LPM_MAX_POWER: disable LPM (hipm & dipm)ATA_LPM_MED_POWER: enable hipm partialATA_LPM_MED_POWER_WITH_DIPM: enable hipm partial &dipm partial &slumber ATA_LPM_MIN_POWER_WITH_PARTIAL: enable hipm partial &dipm partial&slumber&devslpATA_LPM_MIN_POWER: enable hipm slumber &dipm partial &slumber &devslp
hipm: adpter initiated power mode transitions;
dipm:*device* initiated power mode transitions;
From my comprehension, user policy in [ATA_LPM_MED_POWER, ATA_LPM_MIN_POWER] need to be validated on adapter's capabilities (partial(y/n), slumber(y/n), devslp(y/n)), so, there exits the following cases:
1, (n, n, n), validate it to ATA_LPM_UNKNOWN;
2, (n, n, y), validate it to ATA_LPM_MIN_POWER_WITH_PARTIAL/ATA_LPM_MIN_POWER? 3, (n, y, n), validate it to ATA_LPM_MIN_POWER;
4, (n, y, y), validate it to ATA_LPM_MIN_POWER;
5, (y, n, n), validate it to ATA_LPM_MED_POWER;
6, (y, n, y), validate it to ATA_LPM_MIN_POWER_WITH_PARTIAL;
7, (y, y, n), default user policy;
8, (y, y, y), default user policy;
('y' for support, 'n' for not support)
for case 2, I'm not quiet sure, for which may enable hipm partial/slumber.
The way I provided above is quiet complicated and may be incomplete. It may not be realistic to take all into account, but I think case 1 should be taken seriously for which may cause the above PORT_IRQ_PHYRDY issue.
Perhaps, I could refer to Mario's patches later (I have not found yet on kernel/git ^_^).
Quoted 7 lines
Quoted 5 lines
ahci_update_initial_lpm_policy(ap, hpriv);According to my current understanding, the currently submitted patches can solve the above problems, and definitely not the best. I haven't figured out good way to use CAP flags to patch. Hope you can give me some advice, thanks.
Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
Quoted 100 lines · Runa Guo-oc
Quoted 81 lines · Damien Le Moal
Quoted 70 lines · Runa Guo-oc
Quoted 33 lines · Damien Le Moal
Quoted 11 lines · RunaGuo-oc
Quoted 4 lines · Damien Le Moal
Quoted 2 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, PhyRdy Change cannot be detected if ahci supports no LPM.Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.Users should *never* need to be aware of the HW specs and what can or cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your commit message should clearly mention that. But I still need to spend more time re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect low power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch on rc3 to have the LPM config name revert. We need to fix this LPM mess :)
Quoted 3 lines
Quoted 1 line
Mario has patches to fix that.
Hmm. How to patch this case ?
Quoted 8 lines
Mario's patches modify the beginning of the
sata_link_scr_lpm()function to prevent setting an LPM policy that the adapter/drive does not support. This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.It may also be good to revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.To avoid some confusion in this patch set, I want to explain it here. The patch set involves two LPM related issues, one for the ahci adapter does not support LPM (no partial & slumber & devslp), the other for ahci adapter supports part of
LPM(eg, only partial, no slumber & devslp).The former case is what I metioned in this mail thread. Namely, when LPM is enabled, actions trying to enable this unsupported feature will be failed, but will disable
PORT_IRQ_PHYRDYbit at the beginning of theahci_set_lpm()function, which would make PhyRdy Changed cannot be detected. So I add flags in theata_eh_set_lpm()function which will not go to the disable operation.The latter case is what I metioned in "PATCH[2/2]". Namely, if the ahci adapter only supports partial (no slumber & devslp), then when LPM is enabled (eg,
min_power), *device* initiated power mode transitions will be enabled with the scontrol register bits setting to indicate no restrictions on LPM transitions. After that, if SSD/HDD sends a DIPM slumber request, it cannot be disallowed by ahci adpter for driver not setting scontrol register bits properly. So I add flags to control it.Therefore, Mario's patches in the
sata_link_scr_lpm()function may fix the issue in "PATCH[2/2]".Revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy may be a way to fix the issue in "PATCH[1/2]", but there may be another case where some operating system manufacturers setting LPM default enable in driver, like the following code in theahci_init_one()function, also need to control.Quoted 4 lines
if (ap->flags & ATA_FLAG_EM) ap->em_message_type = hpriv->em_msg_type;+
ap->target_lpm_policy=ATA_LPM_MIN_POWER;This one looks wrong. This is set inside
ahci_update_initial_lpm_policy()according to the default kernel config (CONFIG_SATA_MOBILE_LPM_POLICY) and module param + what the drive can do according to ACPI. The problem though is that the adapter capabilities are not checked in that function, so the initial target lpm policy may be wrong.Quoted 3 lines
Since your patch 1/2 adds the hpriv flags indicating the capabilities, you need to use these in
ahci_update_initial_lpm_policy()to validate whatever initial policy is asked for by the user.Yes, the above code is not rigorous, existing methods provided by kernel as you said should be used in this case.
In order to use CAP flags to validate user policy, I review the latest kernel LPM policies, here is my understanding:
ATA_LPM_UNKNOWN: default policy, no LPMATA_LPM_MAX_POWER: disable LPM (hipm & dipm)ATA_LPM_MED_POWER: enable hipm partialATA_LPM_MED_POWER_WITH_DIPM: enable hipm partial &dipm partial &slumberATA_LPM_MIN_POWER_WITH_PARTIAL: enable hipm partial &dipm partial&slumber&devslpATA_LPM_MIN_POWER: enable hipm slumber &dipm partial &slumber &devslp
hipm: adpter initiated power mode transitions;
dipm:*device* initiated power mode transitions;
Quoted 3 lines
From my comprehension, user policy in [
ATA_LPM_MED_POWER,ATA_LPM_MIN_POWER]
need to be validated on adapter's capabilities (partial(y/n), slumber(y/n), devslp(y/n)), so, there exits the following cases:
Note that devslp is a device side feature too. See ata_dev_config_devslp() in libata-core.c. So even if the adapter supports devslp, if the drive does not, devslp should not be enabled on the port.
Quoted 11 lines
1, (n, n, n), validate it to
ATA_LPM_UNKNOWN;
2, (n, n, y), validate it to ATA_LPM_MIN_POWER_WITH_PARTIAL/ATA_LPM_MIN_POWER? 3, (n, y, n), validate it toATA_LPM_MIN_POWER;
4, (n, y, y), validate it toATA_LPM_MIN_POWER;
5, (y, n, n), validate it toATA_LPM_MED_POWER;
6, (y, n, y), validate it toATA_LPM_MIN_POWER_WITH_PARTIAL;
7, (y, y, n), default user policy;
8, (y, y, y), default user policy;
('y' for support, 'n' for not support)for case 2, I'm not quiet sure, for which may enable hipm partial/slumber.
For all above cases, the default should be to use the default configured policy defined by SATA_MOBILE_LPM_POLICY or the ahci module parameter but corrected to match what the adapter & device support, including the eventual NO LPM horkage flag. Mario's patch started addressing that, but that patch can be improved using yours.
Quoted 3 lines
The way I provided above is quiet complicated and may be incomplete. It may not be realistic to take all into account, but I think case 1 should be taken seriously for which may cause the above
PORT_IRQ_PHYRDYissue.
yes.
Quoted 2 lines
Perhaps, I could refer to Mario's patches later (I have not found yet on kernel/git ^_^).
Mario's patch is here:
https://lore.kernel.org/all/[email protected]/
Can you add that patch into your series with eventual modifications to better check the adapter's CAP bits ?
Mario ? Are you OK with that ?
Signature · 2 lines
Damien Le Moal Western Digital Research
RE: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capability
[Public]
Quoted 191 lines
-----Original Message-----
From: Damien Le Moal <[email protected]>
Sent: Wednesday, May 11, 2022 03:15
To: Runa Guo-oc <[email protected]>; [email protected]; Limonciello, Mario <[email protected]>; Cobe Chen <[email protected]>; Tim Guo <[email protected]>; TonyW Wang <[email protected]>; Leo Liu <[email protected]>
Subject: Re: [PATCH 1/2] ahci: Add PhyRdy Change control on actual LPM capabilityQuoted 7 lines · Runa Guo-oc
Quoted 6 lines · Damien Le Moal
Quoted 5 lines · Runa Guo-oc
Quoted 4 lines · Damien Le Moal
Quoted 3 lines · RunaGuo-oc
Quoted 2 lines · Damien Le Moal
Quoted 1 line · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power),Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 2 lines
Quoted 1 line
then, PhyRdy Change cannot be detected if ahci supports no LPM.
Do you mean "...if the ahci adapter does not support LPM." ?
Yes.
Quoted 1 line
If that is what you mean, then
min_powershould not be set.Yes, I agree with you. But, as we know,
link_power_managementis a user policy which can be modified, if some users are not familiar with ahci spec, then the above case may happen.Users should *never* need to be aware of the HW specs and what can
or
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
cannot be done with a particular adapter/drive. User actions trying to enable an unsupported feature should be failed, always.
In your case though, quickly checking the AHCI specs, the scontrol register bits you change seem to be for preventing *device* initiated power mode transitions, not user/host initiated operations. Your
commit
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
message should clearly mention that. But I still need to spend more
time
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
re-reading the specs to confirm. Will do that next week.
Since you added the CAP flags, these flags should be used to detect
low
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
power policies that can be allowed for user actions.
Mario,
Please rebase and repost your patches ! I rebased the for-5.19 branch
on
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
rc3 to have the LPM config name revert. We need to fix this LPM mess
:)
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
Quoted 2 lines
Mario has patches to fix that.
Hmm. How to patch this case ?
Mario's patches modify the beginning of the
sata_link_scr_lpm()function
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
to prevent setting an LPM policy that the adapter/drive does not
support.
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
This together with the correct bits set/reset in the scr register will only allow LPM transitions that are supported.
It may also be good to revisit
ata_scsi_lpm_store()to prevent the user from setting an unsupported policy. Currently, that uselessly triggers an EH sequence.Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
To avoid some confusion in this patch set, I want to explain it here. The patch set involves two LPM related issues, one for the ahci adapter does not support LPM (no partial & slumber & devslp), the other for ahci adapter supports part of
LPM(eg, only partial, no slumber & devslp).The former case is what I metioned in this mail thread. Namely, when
LPM is
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
enabled, actions trying to enable this unsupported feature will be failed, but will disable
PORT_IRQ_PHYRDYbit at the beginning of the
ahci_set_lpm()Quoted 1 line
Quoted 1 line
Quoted 1 line
function, which would make PhyRdy Changed cannot be detected. So I
add flags
Quoted 1 line
Quoted 1 line
Quoted 1 line
in the
ata_eh_set_lpm()function which will not go to the disableoperation.
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
The latter case is what I metioned in "PATCH[2/2]". Namely, if the ahci adapter only supports partial (no slumber & devslp), then when LPM is
enabled
Quoted 1 line
Quoted 1 line
Quoted 1 line
(eg,
min_power), *device* initiated power mode transitions will beenabled
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
with the scontrol register bits setting to indicate no restrictions on LPM transitions. After that, if SSD/HDD sends a DIPM slumber request, it
cannot
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
be disallowed by ahci adpter for driver not setting scontrol register bits properly. So I add flags to control it.
Therefore, Mario's patches in the
sata_link_scr_lpm()function may fixthe
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
issue in "PATCH[2/2]".
Revisit
ata_scsi_lpm_store()to prevent the user from setting anunsupported
Quoted 1 line
Quoted 1 line
Quoted 1 line
policy may be a way to fix the issue in "PATCH[1/2]", but there may be
another
Quoted 1 line
Quoted 1 line
Quoted 1 line
case where some operating system manufacturers setting LPM default
enable in
Quoted 1 line
Quoted 1 line
Quoted 1 line
driver, like the following code in the
ahci_init_one()function, also needto
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines
control.
if (ap->flags & ATA_FLAG_EM) ap->em_message_type = hpriv->em_msg_type;+
ap->target_lpm_policy=ATA_LPM_MIN_POWER;This one looks wrong. This is set inside
ahci_update_initial_lpm_policy()according to the default kernel config(
CONFIG_SATA_MOBILE_LPM_POLICY) andQuoted 1 line
Quoted 1 line
module param + what the drive can do according to ACPI. The problem
though is
Quoted 4 lines
Quoted 4 lines
that the adapter capabilities are not checked in that function, so the initial target lpm policy may be wrong.
Since your patch 1/2 adds the hpriv flags indicating the capabilities, you
need
Quoted 1 line
Quoted 1 line
to use these in
ahci_update_initial_lpm_policy()to validate whateverinitial
Quoted 3 lines
Quoted 1 line
policy is asked for by the user.
Yes, the above code is not rigorous, existing methods provided by kernel as
you
Quoted 8 lines
said should be used in this case.
In order to use CAP flags to validate user policy, I review the latest kernel LPM policies, here is my understanding:
ATA_LPM_UNKNOWN: default policy, no LPMATA_LPM_MAX_POWER: disable LPM (hipm & dipm)ATA_LPM_MED_POWER: enable hipm partialATA_LPM_MED_POWER_WITH_DIPM: enable hipm partial &dipm partial&slumber
Quoted 1 line
ATA_LPM_MIN_POWER_WITH_PARTIAL: enable hipm partial &dipmpartial&slumber&devslp
Quoted 1 line
ATA_LPM_MIN_POWER: enable hipm slumber &dipm partial &slumber&devslp
Quoted 4 lines
hipm: adpter initiated power mode transitions;
dipm:*device* initiated power mode transitions;From my comprehension, user policy in [
ATA_LPM_MED_POWER,
ATA_LPM_MIN_POWER]Quoted 2 lines
need to be validated on adapter's capabilities (partial(y/n), slumber(y/n), devslp(y/n)), so, there exits the following cases:
Note that devslp is a device side feature too. See
ata_dev_config_devslp()inlibata-core.c. So even if the adapter supports devslp, if the drive does not, devslp should not be enabled on the port.Quoted 2 lines
1, (n, n, n), validate it to
ATA_LPM_UNKNOWN;
2, (n, n, y), validate itto ATA_LPM_MIN_POWER_WITH_PARTIAL/ATA_LPM_MIN_POWER?
Quoted 9 lines
3, (n, y, n), validate it to
ATA_LPM_MIN_POWER;
4, (n, y, y), validate it toATA_LPM_MIN_POWER;
5, (y, n, n), validate it toATA_LPM_MED_POWER;
6, (y, n, y), validate it toATA_LPM_MIN_POWER_WITH_PARTIAL; 7, (y, y, n), default user policy;
8, (y, y, y), default user policy;
('y' for support, 'n' for not support)for case 2, I'm not quiet sure, for which may enable hipm partial/slumber.
For all above cases, the default should be to use the default configured policy defined by
SATA_MOBILE_LPM_POLICYor the ahci module parameter but corrected to
match what the adapter & device support, including the eventual NO LPM horkage
flag. Mario's patch started addressing that, but that patch can be improved using yours.Quoted 3 lines
The way I provided above is quiet complicated and may be incomplete. It may not be realistic to take all into account, but I think case 1 should be taken seriously for which may cause the above
PORT_IRQ_PHYRDYissue.
yes.
Quoted 2 lines
Perhaps, I could refer to Mario's patches later (I have not found yet on kernel/git ^_^).
Mario's patch is here:
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.
kernel.org%2Fall%2F20220404194510.9206-2- mario.limonciello%40amd.com%2F&data=05%7C01%7Cmario.limonciell o%40amd.com%7C2ff34a9405444559699208da33264e52%7C3dd8961fe4884e 608e11a82d994e183d%7C0%7C0%7C637878536864920264%7CUnknown%7CT WFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLC JXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=izsUMCw%2BOzV7myxDH JBqmWnzInVJ2CJz3wfsdgVmDdY%3D&reserved=0
Quoted 4 lines
Can you add that patch into your series with eventual modifications to better check the adapter's CAP bits ?
Mario ? Are you OK with that ?
I think that's fine. I've obviously gotten side tracked and didn't get a chance to rebase and send them up. but I do want to bring to your attention that I just found out there was a regression from the original patch that set the policy to mobile.
I was hoping to figure out whether it's just a stable artifact or if it's a real problem with mainline kernel before giving to much thought to setting ALL boards to this new policy. It might be a single outlier, or it might be a crystal ball - TBD.
This regression happened because it got brought back to 5.4-stable, and as it turns out the exact same FCH controller ID from the client silicon is used in another product. It's an ASUS server system with AMD Epyc processor.
The regression is specifically along hotplug, that hotplug no longer works with the more aggressive policies.
The regression (including the bisect) is reported here:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1971576
There is active open tests to figure out whether the issue happens in mainline kernel as well or if it's just from 5.4 backport.
My suggestion is that you pull my patches into your series and lets keep iterating Them, but let's hold off pulling into 5.19 until that's figured out, especially if we have to revert 380cd49e207ba4 to solve that problem then we shouldn't push this policy across all boards (unfortunately).
Quoted 5 lines
--
Damien Le Moal
Western Digital Research
PATCH 2/2 ahci: Add PxSCTL.IPM control on actual LPM capability
In reply to Runa Guo-ocOn some platform, when OS enables LPM by default (eg, min_power), then, DIPM slumber request cannot be disallowed if ahci's CAP.PSC is set to '1' and CAP.SSC is cleared to '0', which may cause ahci to be an uncertain state (same for Partial).
In ahci spec, when CAP.PSC/SSC is cleared to '0', the PxSCTL.IPM field must be programmed to disallow device initiated Partial/ Slumber request.
Adds support to control this case on actual LPM capability.
--- drivers/ata/libata-sata.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/ata/libata-sata.c +11 -1
@@ -394,9 +394,19 @@ int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy,case ATA_LPM_MED_POWER_WITH_DIPM :case ATA_LPM_MIN_POWER_WITH_PARTIAL :case ATA_LPM_MIN_POWER : -if (ata_link_nr_enabled (link) >0 ) +if (ata_link_nr_enabled (link) >0 ) {/* no restrictions on LPM transitions */ scontrol &= ~(0x7 <<8 ); + +/* if controller does not support partial, then disallows it, +* the same for slumber +*/ +if (!(link->ap->host->flags &ATA_HOST_PART )) + scontrol |= (0x1 <<8 ); + +if (!(link->ap->host->flags &ATA_HOST_SSC )) + scontrol |= (0x2 <<8 ); + }else {/* empty port, power off */ scontrol &= ~0xf ;
Signature · 1 line
2.7.4
Re: [PATCH 2/2] ahci: Add PxSCTL.IPM control on actual LPM capability
Quoted 10 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, DIPM slumber request cannot be disallowed if ahci's CAP.PSC is set to '1' and CAP.SSC is cleared to '0', which may cause ahci to be an uncertain state (same for Partial).In ahci spec, when CAP.PSC/SSC is cleared to '0', the PxSCTL.IPM field must be programmed to disallow device initiated Partial/ Slumber request.
Adds support to control this case on actual LPM capability.
s/Adds/Add
Overall, I need to reread the specs to confirm all this.
Quoted 11 lines
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/libata-sata.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)drivers/ata/libata-sata.c +0
-0
drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 398
Quoted 6 lines
drivers/ata/libata-sata.c +1
-1@@ -394,9 +394,19 @@ int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy,case ATA_LPM_MED_POWER_WITH_DIPM :case ATA_LPM_MIN_POWER_WITH_PARTIAL :case ATA_LPM_MIN_POWER : -if (ata_link_nr_enabled (link) >0 ) +if (ata_link_nr_enabled (link) >0 ) { /* no restrictions on LPM transitions */> scontrol &= ~(0x7 << 8);
Given that the added code below adds restrictions, this comment is strange. Better change it to something like:
/* Assume no restrictions on LPM transitions */
drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 403
Quoted 4 lines
drivers/ata/libata-sata.c +4
-0+ +/* if controller does not support partial, then disallows it, +* the same for slumber +*/
Please correctly format the comment and check the grammar. Some like below is easier to read.
/* * If the controller does not support partial or * slumber then disallow these transitions. */
drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 410
Quoted 7 lines
drivers/ata/libata-sata.c +6
-0+if (!(link->ap->host->flags &ATA_HOST_PART )) + scontrol |= (0x1 <<8 ); + +if (!(link->ap->host->flags &ATA_HOST_SSC )) + scontrol |= (0x2 <<8 ); + }else {
Please do not leave this else here. Put it on the same line as the closing bracket and enclose the below statements in brackets too.
drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 412
Quoted 2 lines
drivers/ata/libata-sata.c +0
-0/* empty port, power off */ scontrol &= ~0xf ;
} else {
/* empty port, power off */
scontrol &= ~0xf;
}Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH 2/2] ahci: Add PxSCTL.IPM control on actual LPM capability
Quoted 6 lines · Damien Le Moal
Quoted 5 lines · Runa Guo-oc
On some platform, when OS enables LPM by default (eg,
min_power), then, DIPM slumber request cannot be disallowed if ahci's CAP.PSC is set to '1' and CAP.SSC is cleared to '0', which may cause ahci to be an uncertain state (same for Partial).
Quoted 6 lines
Quoted 5 lines
In ahci spec, when CAP.PSC/SSC is cleared to '0', the PxSCTL.IPM field must be programmed to disallow device initiated Partial/ Slumber request.
Adds support to control this case on actual LPM capability.
s/Adds/Add
Sorry, here should use 'Add' instead of 'Adds'.
Quoted 1 line
Overall, I need to reread the specs to confirm all this.
Ok.
Quoted 43 lines
Quoted 10 lines
Signed-off-by Runa Guo-oc <[email protected]>--- drivers/ata/libata-sata.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)drivers/ata/libata-sata.c +0
-0drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 398
Quoted 6 lines
drivers/ata/libata-sata.c +1
-1@@ -394,9 +394,19 @@ int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy,case ATA_LPM_MED_POWER_WITH_DIPM :case ATA_LPM_MIN_POWER_WITH_PARTIAL :case ATA_LPM_MIN_POWER : -if (ata_link_nr_enabled (link) >0 ) +if (ata_link_nr_enabled (link) >0 ) { /* no restrictions on LPM transitions */> scontrol &= ~(0x7 << 8);Given that the added code below adds restrictions, this comment is strange. Better change it to something like:
/* Assume no restrictions on LPM transitions */drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 403
Quoted 4 lines
drivers/ata/libata-sata.c +4
-0+ +/* if controller does not support partial, then disallows it, +* the same for slumber +*/ Please correctly format the comment and check the grammar. Some like below is easier to read.
/* * If the controller does not support partial or * slumber then disallow these transitions. */drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 410
Quoted 7 lines
drivers/ata/libata-sata.c +6
-0+if (!(link->ap->host->flags &ATA_HOST_PART )) + scontrol |= (0x1 <<8 ); + +if (!(link->ap->host->flags &ATA_HOST_SSC )) + scontrol |= (0x2 <<8 ); + }else {Please do not leave this else here. Put it on the same line as the closing bracket and enclose the below statements in brackets too.
Quoted 6 lines
drivers/ata/libata-sata.c int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy, · line 412
Quoted 2 lines
drivers/ata/libata-sata.c +0
-0/* empty port, power off */ scontrol &= ~0xf ;} else { /* empty port, power off */ scontrol &= ~0xf; }
I'll change it like below,
+ if (ata_link_nr_enabled(link) > 0) {
- /* no restrictions on LPM transitions */
+ /* Assume no restrictions on LPM transitions */
scontrol &= ~(0x7 << 8);
+ /*
+ * If the controller does not support partial or
+ * slumber then disallow these transitions.
+ */
+ if (!(link->ap->host->flags & ATA_HOST_PART))
+ scontrol |= (0x1 << 8);
+
+ if (!(link->ap->host->flags & ATA_HOST_SSC))
+ scontrol |= (0x2 << 8);
} else {
/* empty port, power off */
scontrol &= ~0xf;
}