Skip to contentSkip to conversation

scsi: ufs: Fix power drain and hci quirk for WriteBooster

PATCH Applied Stanley Chu 3 participants 6 messages Dec 22, 2020 – Jan 8, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover scsi: ufs: Fix power drain and hci quirk for WriteBooster

Hi,
This series fixes two WriteBooster issues,
1. Fix a corner case that device flush capability is not disabled during system suspend
2. Fix the checking of UFSHCI quirk UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

Stanley Chu (2):
scsi: ufs: Fix possible power drain during system suspend
scsi: ufs: Relax the condition of
UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

 drivers/scsi/ufs/ufshcd.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
Signature · 5 lines
2.18.0
_______________________________________________
Linux-mediatek mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-mediatek
#1 lore raw

PATCH v2 1/2 scsi: ufs: Fix possible power drain during system suspend

Currently if device needs to do flush or BKOP operations, the device VCC power is kept during runtime-suspend period.

However, if system suspend is happening while device is runtime-suspended, such power may not be disabled successfully.

The reasons may be,

1. If current PM level is the same as SPM level, device will keep runtime-suspended by ufshcd_system_suspend().

2. Flush recheck work may not be scheduled successfully during system suspend period. If it can wake up the system, this is also not the intention of the recheck work.

To fix this issue, simply runtime-resume the device if the flush is allowed during runtime suspend period. Flush capability will be disabled while leaving runtime suspend, and also not be allowed in system suspend period.

Fixes 51dd905bd2f6 ("scsi: ufs: Fix WriteBooster flush during runtime suspend")
---
 drivers/scsi/ufs/ufshcd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +2 -1
index e221add25a7e..9d61dc3eb842 100644
@@ -8903,7 +8903,8 @@ int ufshcd_system_suspend(struct ufs_hba *hba)
 	if ((ufs_get_pm_lvl_to_dev_pwr_mode(hba->spm_lvl) ==
 	     hba->curr_dev_pwr_mode) &&
 	    (ufs_get_pm_lvl_to_link_pwr_state(hba->spm_lvl) ==
-	     hba->uic_link_state))
+	     hba->uic_link_state) &&
+	     !hba->dev_info.b_rpm_dev_flush_capable)
 		goto out;
 
 	if (pm_runtime_suspended(hba->dev)) {
Signature · 5 lines
2.18.0
_______________________________________________
Linux-mediatek mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-mediatek
Reviewed-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] scsi: ufs: Fix possible power drain during system suspend

Quoted 32 lines · Stanley Chu

Currently if device needs to do flush or BKOP operations, the device VCC power is kept during runtime-suspend period.

However, if system suspend is happening while device is runtime-suspended, such power may not be disabled successfully.

The reasons may be,

1. If current PM level is the same as SPM level, device will keep runtime-suspended by ufshcd_system_suspend().

2. Flush recheck work may not be scheduled successfully during system suspend period. If it can wake up the system, this is also not the intention of the recheck work.

To fix this issue, simply runtime-resume the device if the flush is allowed during runtime suspend period. Flush capability will be disabled while leaving runtime suspend, and also not be allowed in system suspend period.

Fixes 51dd905bd2f6 ("scsi: ufs: Fix WriteBooster flush during runtime suspend")
---
 drivers/scsi/ufs/ufshcd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +0 -0
index e221add25a7e..9d61dc3eb842 100644

drivers/scsi/ufs/ufshcd.c int ufshcd_system_suspend(struct ufs_hba *hba) · line 8910

Quoted 9 lines
drivers/scsi/ufs/ufshcd.c +2 -1
@@ -8903,7 +8903,8 @@ int ufshcd_system_suspend(struct ufs_hba *hba)
 	if ((ufs_get_pm_lvl_to_dev_pwr_mode(hba->spm_lvl) ==
 	     hba->curr_dev_pwr_mode) &&
 	    (ufs_get_pm_lvl_to_link_pwr_state(hba->spm_lvl) ==
-	     hba->uic_link_state))
+	     hba->uic_link_state) &&
+	     !hba->dev_info.b_rpm_dev_flush_capable)
 		goto out;
 
 	if (pm_runtime_suspended(hba->dev)) {

_______________________________________________
Linux-mediatek mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-mediatek

#3 lore raw

PATCH v2 2/2 scsi: ufs: Relax the condition of UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

In reply to Stanley Chu

UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL is aimed to skip enabling fWriteBoosterBufferFlushEn while WriteBooster is initializing. Therefore it is better to apply the checking during WriteBooster initialization only.

---
 drivers/scsi/ufs/ufshcd.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/scsi/ufs/ufshcd.c +2 -4
index 9d61dc3eb842..e50b19925236 100644
@@ -289,7 +289,8 @@ static inline void ufshcd_wb_config(struct ufs_hba *hba)
 	if (ret)
 		dev_err(hba->dev, "%s: En WB flush during H8: failed: %d\n",
 			__func__, ret);
-	ufshcd_wb_toggle_flush(hba, true);
+	if (!(hba->quirks & UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL))
+		ufshcd_wb_toggle_flush(hba, true);
 }
 
 static void ufshcd_scsi_unblock_requests(struct ufs_hba *hba)
@@ -5401,9 +5402,6 @@ static int ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set)
 
 static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, bool enable)
 {
-	if (hba->quirks & UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL)
-		return;
-
 	if (enable)
 		ufshcd_wb_buf_flush_enable(hba);
 	else
Signature · 1 line
2.18.0
Reviewed-by on v2 2/2
#4 lore raw

Re: [PATCH v2 2/2] scsi: ufs: Relax the condition of UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

Quoted 35 lines · Stanley Chu

UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL is aimed to skip enabling fWriteBoosterBufferFlushEn while WriteBooster is initializing. Therefore it is better to apply the checking during WriteBooster initialization only.

---
 drivers/scsi/ufs/ufshcd.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/scsi/ufs/ufshcd.c +2 -1
index 9d61dc3eb842..e50b19925236 100644
@@ -289,7 +289,8 @@ static inline void ufshcd_wb_config(struct ufs_hba 
 *hba)
 	if (ret)
 		dev_err(hba->dev, "%s: En WB flush during H8: failed: %d\n",
 			__func__, ret);
-	ufshcd_wb_toggle_flush(hba, true);
+	if (!(hba->quirks & UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL))
+		ufshcd_wb_toggle_flush(hba, true);
 }
 

static void ufshcd_scsi_unblock_requests(struct ufs_hba *hba)

drivers/scsi/ufs/ufshcd.c +0 -3
@@ -5401,9 +5402,6 @@ static int
 ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set)
 
 static inline void ufshcd_wb_toggle_flush(struct ufs_hba *hba, bool 
 enable)
 {
-	if (hba->quirks & UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL)
-		return;
-
 	if (enable)

drivers/scsi/ufs/ufshcd.c static int ufshcd_wb_toggle_flush_during_h8(struct ufs_hba *hba, bool set) · line 5407

Quoted 2 lines
drivers/scsi/ufs/ufshcd.c +0 -0
 		ufshcd_wb_buf_flush_enable(hba);
 	else

Thanks for the change, I was thinking about the same.

_______________________________________________
Linux-mediatek mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-mediatek

#5 lore raw
In reply to Stanley Chu
Quoted 10 lines · Stanley Chu

This series fixes two WriteBooster issues,
1. Fix a corner case that device flush capability is not disabled during system suspend
2. Fix the checking of UFSHCI quirk UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

Stanley Chu (2):
scsi: ufs: Fix possible power drain during system suspend
scsi: ufs: Relax the condition of
UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL

[...]

Applied to 5.11/scsi-fixes, thanks!

[1/2] scsi: ufs: Fix possible power drain during system suspend

      https://git.kernel.org/mkp/scsi/c/1d53864c3617
[2/2] scsi: ufs: Relax the condition of UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL
      https://git.kernel.org/mkp/scsi/c/21acf4601cc6
Signature · 6 lines
Martin K. Petersen	Oracle Linux Engineering

_______________________________________________
Linux-mediatek mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-mediatek