[PATCH v8 1/2] scsi: ufs: Enable power management for wlun

Adrian Hunter adrian.hunter at intel.com
Mon Mar 1 19:12:55 GMT 2021


On 1/03/21 8:10 pm, Asutosh Das wrote:
> On Mon, Mar 01 2021 at 05:23 -0800, Adrian Hunter wrote:
>> On 26/02/21 1:37 am, Asutosh Das wrote:
>>> @@ -8901,43 +9125,14 @@ static int ufshcd_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
>>>              goto vendor_suspend;
>>>      }
>>
>> The ufshcd_reset_and_restore() in ufshcd_resume() will also change the power
>> mode of the UFS device to active.  Until the UFS device is also resumed and
>> then suspended, it will not return to a low power mode.
>>
>>
> Umm, sorry, I didn't understand this comment.
> Say, the UFS device was reset in ufshcd_reset_and_restore() it'd be a hardware
> reset and the UFS device would move to Powered On mode and then to Active power
> mode, when it is ready to begin initialization. And from this state it should
> move to all other legal states.
> Before entering system suspend ufshcd_system_suspend(), the ufs device is
> runtime resumed in ufshcd_suspend_prepare().
> 
> Please can you explain a bit more on this issue that you see?

Say you runtime resume the host controller, and
ufshcd_reset_and_restore() makes the UFS device active,
but the UFS device is still runtime suspended.


Example:

Add a debugfs file to show the current power mode:

diff --git a/drivers/scsi/ufs/ufs-debugfs.c b/drivers/scsi/ufs/ufs-debugfs.c
index dee98dc72d29..700b88df0866 100644
--- a/drivers/scsi/ufs/ufs-debugfs.c
+++ b/drivers/scsi/ufs/ufs-debugfs.c
@@ -48,6 +48,7 @@ void ufs_debugfs_hba_init(struct ufs_hba *hba)
 {
        hba->debugfs_root = debugfs_create_dir(dev_name(hba->dev), ufs_debugfs_root);
        debugfs_create_file("stats", 0400, hba->debugfs_root, hba, &ufs_debugfs_stats_fops);
+       debugfs_create_u32("curr_dev_pwr_mode", 0400, hba->debugfs_root, (u32 *)&hba->curr_dev_pwr_mode);
 }

 void ufs_debugfs_hba_exit(struct ufs_hba *hba)


# grep -H . /sys/bus/pci/drivers/ufshcd/0000\:00\:12.5/rpm*
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/rpm_lvl:6
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/rpm_target_dev_state:DEEPSLEEP
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/rpm_target_link_state:OFF
# cat /sys/kernel/debug/ufshcd/0000\:00\:12.5/curr_dev_pwr_mode
4
# echo on > /sys/devices/pci0000:00/0000:00:12.5/power/control
# cat /sys/kernel/debug/ufshcd/0000\:00\:12.5/curr_dev_pwr_mode
1
# grep -H . /sys/bus/pci/drivers/ufshcd/0000\:00\:12.5/host*/target*/*/power/runtime_status
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/host1/target1:0:0/1:0:0:0/power/runtime_status:suspended
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/host1/target1:0:0/1:0:0:49456/power/runtime_status:suspended
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/host1/target1:0:0/1:0:0:49476/power/runtime_status:suspended
/sys/bus/pci/drivers/ufshcd/0000:00:12.5/host1/target1:0:0/1:0:0:49488/power/runtime_status:suspended

So UFS devices is runtime suspended and should in DeepSleep, but it is active.



More information about the linux-arm-kernel mailing list