Fix PM deadlock and card detection problem
cover Fix PM deadlock and card detection problem
The patchset fixes a deadlock situation of runtime PM methods and minimize card detection "deadtime" during runtime suspending.
Roger Tseng (2):
mfd: rtsx_usb: Fix runtime PM deadlock
mfd: rtsx_usb: Defer autosuspend while card exists
drivers/mfd/rtsx_usb.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-)
Signature · 1 line
2.1.3
PATCH 1/2 mfd: rtsx_usb: Fix runtime PM deadlock
sd_set_power_mode() in derived module drivers/mmc/host/rtsx_usb_sdmmc.c acquires dev_mutex and then calls pm_runtime_get_sync() to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly before rtsx_usb_suspend() acquires the same dev_mutex, both routine deadlock and further hang the driver because pm_runtime_get_sync() waits the pending PM operations.
Fix this by using an empty suspend method. mmc_core always turns the LED off after a request is done and thus it is ok to remove the only rtsx_usb_turn_off_led() here.
--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)
drivers/mfd/rtsx_usb.c +0 -9
@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 1 line
2.1.3
Lee Jones
AppliedRe: [PATCH 1/2] mfd: rtsx_usb: Fix runtime PM deadlock
Quoted 11 lines · Roger Tseng
sd_set_power_mode()in derived moduledrivers/mmc/host/rtsx_usb_sdmmc.cacquiresdev_mutexand then callspm_runtime_get_sync()to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly beforertsx_usb_suspend()acquires the samedev_mutex, both routine deadlock and further hang the driver becausepm_runtime_get_sync()waits the pending PM operations.Fix this by using an empty suspend method.
mmc_corealways turns the LED off after a request is done and thus it is ok to remove the onlyrtsx_usb_turn_off_led()here.
Quoted 6 lines
Cc <[email protected]> # v3.16+Fixes 730876be2566 ("mfd: Add realtek USB card reader driver")Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)
Applied, thanks.
Quoted 20 lines
drivers/mfd/rtsx_usb.c +0
-9@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 1/2] mfd: rtsx_usb: Fix runtime PM deadlock
Quoted 13 lines · Lee Jones
Quoted 11 lines · Roger Tseng
sd_set_power_mode()in derived moduledrivers/mmc/host/rtsx_usb_sdmmc.cacquiresdev_mutexand then callspm_runtime_get_sync()to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly beforertsx_usb_suspend()acquires the samedev_mutex, both routine deadlock and further hang the driver becausepm_runtime_get_sync()waits the pending PM operations.Fix this by using an empty suspend method.
mmc_corealways turns the LED off after a request is done and thus it is ok to remove the onlyrtsx_usb_turn_off_led()here.
Quoted 9 lines
Quoted 6 lines
Cc <[email protected]> # v3.16+Fixes 730876be2566 ("mfd: Add realtek USB card reader driver")Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)Applied, thanks.
I'm sorry but build bot from Intel just reported me that I forgot to delete an unused variable "ucr" between two commits. My bad.
Do I have a chance to send v2?
Quoted 21 lines
Quoted 20 lines
drivers/mfd/rtsx_usb.c +0
-9@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 3 lines
Best regards,
Roger Tseng
���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m����������?�I�Re: [PATCH 1/2] mfd: rtsx_usb: Fix runtime PM deadlock
Quoted 21 lines · Roger Tseng
Quoted 20 lines · Lee Jones
Quoted 17 lines · Roger Tseng
sd_set_power_mode()in derived moduledrivers/mmc/host/rtsx_usb_sdmmc.cacquiresdev_mutexand then callspm_runtime_get_sync()to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly beforertsx_usb_suspend()acquires the samedev_mutex, both routine deadlock and further hang the driver becausepm_runtime_get_sync()waits the pending PM operations.Fix this by using an empty suspend method.
mmc_corealways turns the LED off after a request is done and thus it is ok to remove the onlyrtsx_usb_turn_off_led()here.Cc <[email protected]> # v3.16+Fixes 730876be2566 ("mfd: Add realtek USB card reader driver")Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)
Quoted 6 lines
Quoted 2 lines
Applied, thanks.
I'm sorry but build bot from Intel just reported me that I forgot to delete an unused variable "ucr" between two commits. My bad.
Do I have a chance to send v2?
You're lucky I'm in a good mood. ;)
I fixed it already.
Quoted 22 lines
Quoted 21 lines
Quoted 20 lines
drivers/mfd/rtsx_usb.c +0
-9@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
Re: [PATCH 1/2] mfd: rtsx_usb: Fix runtime PM deadlock
Quoted 25 lines · Lee Jones
Quoted 23 lines · Roger Tseng
Quoted 22 lines · Lee Jones
Quoted 11 lines · Roger Tseng
sd_set_power_mode()in derived moduledrivers/mmc/host/rtsx_usb_sdmmc.cacquiresdev_mutexand then callspm_runtime_get_sync()to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly beforertsx_usb_suspend()acquires the samedev_mutex, both routine deadlock and further hang the driver becausepm_runtime_get_sync()waits the pending PM operations.Fix this by using an empty suspend method.
mmc_corealways turns the LED off after a request is done and thus it is ok to remove the onlyrtsx_usb_turn_off_led()here.Quoted 6 lines
Cc <[email protected]> # v3.16+Fixes 730876be2566 ("mfd: Add realtek USB card reader driver")Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)Applied, thanks.
Quoted 9 lines
Quoted 4 lines
I'm sorry but build bot from Intel just reported me that I forgot to delete an unused variable "ucr" between two commits. My bad.
Do I have a chance to send v2?
You're lucky I'm in a good mood. ;)
I fixed it already.
That's great, thanks! And thus patch 2/2 also needs to be changed accordingly.
By the way, the build bot reported again about an undefined variable build error in 2/2(I'm sorry for this again). I put the overall updated content of 2/2 here for you if you're going to fix it, or I can also re-send it individually.
drivers/mfd/rtsx_usb.c +0 -0
@@ -681,9 +681,27 @@ static void rtsx_usb_disconnect(struct
usb_interface *intf)
#ifdef CONFIG_PM
static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t
message)
{
+ struct rtsx_ucr *ucr =
+ (struct rtsx_ucr *)usb_get_intfdata(intf);
+ u16 val = 0;
+
dev_dbg(&intf->dev, "%s called with pm message 0x%04x\n",
__func__, message.event);
+ if (PMSG_IS_AUTO(message)) {
+ if (mutex_trylock(&ucr->dev_mutex)) {
+ rtsx_usb_get_card_status(ucr, &val);
+ mutex_unlock(&ucr->dev_mutex);
+
+ /* Defer the autosuspend if card exists */
+ if (val & (SD_CD | MS_CD))
+ return -EAGAIN;
+ } else {
+ /* There is an ongoing operation*/
+ return -EAGAIN;
+ }
+ }
+
return 0;
}
Quoted 23 lines
Quoted 22 lines
Quoted 21 lines
Quoted 20 lines
drivers/mfd/rtsx_usb.c +0
-9@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 3 lines
Best regards,
Roger Tseng
���{.n�+�������+%�����ݶ��w��{.n�+����{��G�����{ay�ʇڙ�,j��f���h��������z_�(�階�ݢj"���m������G������?����&���~��iO��z��v�^�m����������?�I�Re: [PATCH 1/2] mfd: rtsx_usb: Fix runtime PM deadlock
Quoted 35 lines · Roger Tseng
Quoted 25 lines · Lee Jones
Quoted 23 lines · Roger Tseng
Quoted 22 lines · Lee Jones
Quoted 11 lines · Roger Tseng
sd_set_power_mode()in derived moduledrivers/mmc/host/rtsx_usb_sdmmc.cacquiresdev_mutexand then callspm_runtime_get_sync()to make sure the device is awake while initializing a newly inserted card. Once it is called during suspending state and explicitly beforertsx_usb_suspend()acquires the samedev_mutex, both routine deadlock and further hang the driver becausepm_runtime_get_sync()waits the pending PM operations.Fix this by using an empty suspend method.
mmc_corealways turns the LED off after a request is done and thus it is ok to remove the onlyrtsx_usb_turn_off_led()here.Quoted 6 lines
Cc <[email protected]> # v3.16+Fixes 730876be2566 ("mfd: Add realtek USB card reader driver")Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 9 --------- 1 file changed, 9 deletions(-)Applied, thanks.
Quoted 9 lines
Quoted 4 lines
I'm sorry but build bot from Intel just reported me that I forgot to delete an unused variable "ucr" between two commits. My bad.
Do I have a chance to send v2?
You're lucky I'm in a good mood. ;)
I fixed it already.
Quoted 7 lines
That's great, thanks! And thus patch 2/2 also needs to be changed accordingly.
By the way, the build bot reported again about an undefined variable build error in 2/2(I'm sorry for this again). I put the overall updated content of 2/2 here for you if you're going to fix it, or I can also re-send it individually.
Please re-send this patch and prefix it with [OVERWRITE].
In future make sure you test your submission thoroughly prior to sending, as my time is very limited and with all due respect I could make better use of it than this.
Quoted 60 lines
drivers/mfd/rtsx_usb.c +18
-0@@ -681,9 +681,27 @@ static void rtsx_usb_disconnect(struct usb_interface *intf)#ifdef CONFIG_PM static int rtsx_usb_suspend (struct usb_interface *intf,pm_message_t message) { +struct rtsx_ucr *ucr = + (struct rtsx_ucr *)usb_get_intfdata (intf); +u16 val =0 ; +dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); +if (PMSG_IS_AUTO (message)) { +if (mutex_trylock (&ucr->dev_mutex)) { +rtsx_usb_get_card_status (ucr, &val); +mutex_unlock (&ucr->dev_mutex); + +/* Defer the autosuspend if card exists */ +if (val & (SD_CD |MS_CD )) +return -EAGAIN ; + }else { +/* There is an ongoing operation*/ +return -EAGAIN ; + } + } +return 0 ;}
Quoted 23 lines
Quoted 22 lines
Quoted 21 lines
Quoted 20 lines
drivers/mfd/rtsx_usb.c +0
-9@@ -687,15 +687,6 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); -/* -* Call to make sure LED is off during suspend to save more power. -* It is NOT a permanent state and could be turned on anytime later. -* Thus no need to call turn_on when resunming. -*/ -mutex_lock (&ucr->dev_mutex); -rtsx_usb_turn_off_led (ucr); -mutex_unlock (&ucr->dev_mutex); -return 0 ; }
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
PATCH 2/2 mfd: rtsx_usb: Defer autosuspend while card exists
In reply to Roger TsengA card insertion happens after the lastest polling before reader is suspended may never have a chance to be detected. Under current 1-HZ polling interval setting in mmc_core, the worst case of such undetectablility is about 1 second.
To further reduce the undetectability, detect card slot again in suspend method and defer the autosuspend if the slot is loaded. The default 2 second autosuspend delay of USB subsystem should let the next polling detects the card.
--- drivers/mfd/rtsx_usb.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
drivers/mfd/rtsx_usb.c +14 -0
@@ -687,6 +687,20 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); +if (PMSG_IS_AUTO (message)) { +if (mutex_trylock (&ucr->dev_mutex)) { +rtsx_usb_get_card_status (ucr, &val); +mutex_unlock (&ucr->dev_mutex); + +/* Defer the autosuspend if card exists */ +if (val & (SD_CD |MS_CD )) +return -EAGAIN ; + }else { +/* There is an ongoing operation*/ +return -EAGAIN ; + } + } +return 0 ; }
Signature · 1 line
2.1.3
Lee Jones
AppliedRe: [PATCH 2/2] mfd: rtsx_usb: Defer autosuspend while card exists
Quoted 10 lines · Roger Tseng
A card insertion happens after the lastest polling before reader is suspended may never have a chance to be detected. Under current 1-HZ polling interval setting in
mmc_core, the worst case of such undetectablility is about 1 second.To further reduce the undetectability, detect card slot again in suspend method and defer the autosuspend if the slot is loaded. The default 2 second autosuspend delay of USB subsystem should let the next polling detects the card.
Quoted 4 lines
Signed-off-by Roger Tseng <[email protected]>--- drivers/mfd/rtsx_usb.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
Applied, thanks.
Quoted 25 lines
drivers/mfd/rtsx_usb.c +14
-0@@ -687,6 +687,20 @@ static int rtsx_usb_suspend(struct usb_interface *intf, pm_message_t message)dev_dbg (&intf->dev,"%s called with pm message 0x%04x\n" , __func__, message.event); +if (PMSG_IS_AUTO (message)) { +if (mutex_trylock (&ucr->dev_mutex)) { +rtsx_usb_get_card_status (ucr, &val); +mutex_unlock (&ucr->dev_mutex); + +/* Defer the autosuspend if card exists */ +if (val & (SD_CD |MS_CD )) +return -EAGAIN ; + }else { +/* There is an ongoing operation*/ +return -EAGAIN ; + } + } +return 0 ; }
Signature · 4 lines
Lee Jones Linaro STMicroelectronics Landing Team Lead Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog