scsi: ufs: Reduce the power mode change timeout
Subsystems (2)
Bart Van Assche
reviewerPATCH 1/1 scsi: ufs: Reduce the power mode change timeout
The current power mode change timeout (180 s) is so large that it can cause a watchdog timer to fire. Reduce the power mode change timeout to 10 seconds.
--- drivers/ufs/core/ufshcd.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/ufs/core/ufshcd.c +8 -1
@@ -8776,6 +8776,8 @@ static int ufshcd_set_dev_pwr_mode(struct ufs_hba *hba,struct scsi_device *sdp;unsigned long flags;int ret, retries; +unsigned long deadline; +int32_t remaining;spin_lock_irqsave (hba->host->host_lock, flags); sdp = hba->ufs_device_wlun;
@@ -8808,9 +8810,14 @@ static int ufshcd_set_dev_pwr_mode(struct ufs_hba *hba,* callbacks hence set the RQF_PM flag so that it doesn't resume the * already suspended childs. */ + deadline = jiffies +10 *HZ ;for (retries =3 ; retries >0 ; --retries) { + ret = -ETIMEDOUT ; + remaining = deadline - jiffies; +if (remaining <=0 ) +break ; ret =scsi_execute (sdp, cmd,DMA_NONE ,NULL ,0 ,NULL , &sshdr, -START_STOP_TIMEOUT ,0 ,0 ,RQF_PM ,NULL ); + remaining /HZ ,0 ,0 ,RQF_PM ,NULL );if (!scsi_status_is_check_condition (ret) || !scsi_sense_valid (&sshdr) || sshdr.sense_key !=UNIT_ATTENTION )
Quoted 10 lines · Bart Van Assche
The current power mode change timeout (180 s) is so large that it can cause a watchdog timer to fire. Reduce the power mode change timeout to 10 seconds.
Signed-off-by Bart Van Assche <[email protected]>--- drivers/ufs/core/ufshcd.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
Avri Altman
reviewerQuoted 3 lines
The current power mode change timeout (180 s) is so large that it can cause a watchdog timer to fire. Reduce the power mode change timeout to 10 seconds.
Maybe also worth noting that it was invented 20 years ago for scsi ioctl, And is less applicable for nowadays ufs devices
Quoted 11 lines
Signed-off-by Bart Van Assche <[email protected]>--- drivers/ufs/core/ufshcd.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)drivers/ufs/core/ufshcd.c +0
-0
drivers/ufs/core/ufshcd.c static int ufshcd_set_dev_pwr_mode(struct · line 8781
Quoted 6 lines
drivers/ufs/core/ufshcd.c +2
-0@@ -8776,6 +8776,8 @@ static int ufshcd_set_dev_pwr_mode(structufs_hba *hba, struct scsi_device *sdp;unsigned long flags;int ret, retries; +unsigned long deadline; +int32_t remaining;
Maybe use ktime api, like its done throughout the driver?
Thanks,
Avri
Quoted 20 lines
drivers/ufs/core/ufshcd.c +6
-1spin_lock_irqsave (hba->host->host_lock, flags); sdp = hba->ufs_device_wlun;@@ -8808,9 +8810,14 @@ static int ufshcd_set_dev_pwr_mode(structufs_hba *hba, * callbacks hence set the RQF_PM flag so that it doesn't resume the * already suspended childs. */ + deadline = jiffies +10 *HZ ;for (retries =3 ; retries >0 ; --retries) { + ret = -ETIMEDOUT ; + remaining = deadline - jiffies; +if (remaining <=0 ) +break ; ret =scsi_execute (sdp, cmd,DMA_NONE ,NULL ,0 ,NULL , &sshdr, -START_STOP_TIMEOUT ,0 ,0 ,RQF_PM ,NULL ); + remaining /HZ ,0 ,0 ,RQF_PM ,NULL ); if (!scsi_status_is_check_condition(ret) || !scsi_sense_valid(&sshdr) ||sshdr.sense_key != UNIT_ATTENTION)
Bart Van Assche
reviewerQuoted 12 lines · Avri Altman
Quoted 3 lines
The current power mode change timeout (180 s) is so large that it can cause a watchdog timer to fire. Reduce the power mode change timeout to 10 seconds.
Maybe also worth noting that it was invented 20 years ago for scsi ioctl, And is less applicable for nowadays ufs devices
Quoted 6 lines
Signed-off-by Bart Van Assche <[email protected]>--- drivers/ufs/core/ufshcd.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
Quoted 13 lines
drivers/ufs/core/ufshcd.c static int ufshcd_set_dev_pwr_mode(struct · line 8781
Quoted 11 lines
drivers/ufs/core/ufshcd.c +2
-0@@ -8776,6 +8776,8 @@ static int ufshcd_set_dev_pwr_mode(structufs_hba *hba, struct scsi_device *sdp;unsigned long flags;int ret, retries; +unsigned long deadline; +int32_t remaining;Maybe use ktime api, like its done throughout the driver?
Hi Avri,
Calling ktime_get() is much more expensive than reading the jiffies counter. That's why I prefer to use the jiffies counter if the timeout is significantly larger than 1 / HZ and if the accuracy of 1 / HZ is sufficient.
Thanks,
Bart.